j business and behavioral sciences - semantic · pdf filep.o. box 10367, san bernardino, ca...

JOURNAL OF BUSINESS AND BEHAVIORAL SCIENCES

VOLUME 8, NUMBER 1 FALL 2001

Management Education in a Globalized World ......................................................................................................................................Mahendra Raj

Ranking Differences Among the Sharpe Ratio and Related Total Portfolio Risk Adjusted Performance Measures .............................................................................................................................Bruce D. Bagamery

The Highest Short-Term IPO Return & its Trend: The “A” Share Experience on the Shanghai Stock Exchange ....................................................................................................................... Anthony Yanxiang Gu

A Credit Scoring Model for Subprime Debt .........................................................................................................................Angeline M. Lavin

Accountancy: A Profession Evolving From an Art ................................................................... Kel-Ann S. Eyler, Teresa T. King and Elliott L. Slocum

The Use of the Delphi Method to Estimate Audit & Detection Probabilities ................................................................................................................................. Carol M. Fischer

The Effect of Journal Quality on Accounting Faculty Remuneration: Does University Research Emphasis Matter? ............................................................. Todd L. Sayre, Carol M. Graham and James R. Hasselback

Revisiting the Organizational Commitment Model in a Public Sector Organization ...............................................................................................................................Anthony F. Chelte

NCHICA: A Strategic Approach to Implementing a State-wide Healthcare Information System ..................................................................................................Mary A. Curran and Kent E. Curran

Uncertainty and the Design of Department of Defense Contracts ...............................................................................................Gerald M. Groshek and David J. Rose

Explaining Customer Response to Service Recovery: An Assessment of the Psychometric Properties & Predictive Validity of Measurement Scales for Exchange-Theoretical Constructs ............................................................................... Manfred M. Maute and William R. Forrester, Jr.

Customer Relationship Management & Consumer Correspondence: A Theoretical Approach ............................................................................................Michael McCall and Donald W. Eckrich

An Investigation of Voluntary Disclosures, Sizes and Foreign Sales of Irish Firms Approach ................................................................................................ Sinéad O�Connor & Robyn Lawrence

3

JOURNAL OF BUSINESS AND BEHAVIORAL SCIENCES P.O. Box 10367, San Bernardino, CA 92423-0367: Tel 909-748-6287 Fax 909-335-9279 Email: [email protected] http://www.asbbs.org ____________________ISSN 1099-5374_______________________

Editor-in-Chief Wali I. Mondal

University of Redlands

Managing Editor Mahendra Raj

The Robert Gordon University, Scotland

Editorial Assistant Diane E. Wilson

The Robert Gordon University, Scotland

Editorial Board Paul Bursik

St. Norbert College

Robyn Lawrence University of Scranton

Gerald Calvasina University of Southern Utah

Michele Matherly University of North Carolina, Wilmington

Richard Davis California State University, Chico

Mark Puclik University of Illinois, Springfield

Harold Fletcher Loyola College, Maryland

Kevin Quinn St. Norbert College

Gerald Groshek University of Redlands

Darshan Sachdeva Cal St University, Long Beach

Roxanne Johnson

University of Scranton Paula Weber

St. Cloud State University

William Kehoe University of Virginia

Linda Whitten Skyline College

The Journal of Business and Behavioral Sciences is an official publication of the American Society of Business and Behavioral Sciences (ASBBS). Papers published in the Journal went through a blind refereed review process prior to acceptance for publication. Thanks are due to anonymous referees. The Society�s National Annual Meeting is held in February of each year in Las Vegas and the International Meeting is held in August of each year in Europe. Visit www.asbbs.org for information regarding the Society�s meetings and Publications.

4


Volume 8, Number 1 Fall 2001

TABLE OF CONTENTS

MANAGEMENT EDUCATION IN A GLOBALIZED WORLD

Mahendra Raj 6

RANKING DIFFERENCES AMONG THE SHARPE RATIO AND RELATED TOTAL PORTFOLIO RISK ADJUSTED PERFORMANCE MEASURES

Bruce D. Bagamery 10

THE HIGHEST SHORT-TERM IPO RETURN & ITS TREND: THE �A� SHARE EXPERIENCE ON THE SHANGHAI STOCK EXCHANGE

Anthony Yanxiang Gu 22

A CREDIT SCORING MODEL FOR SUBPRIME DEBT Angeline M. Lavin 32

ACCO UNTANCY: A PROFESSION EVOLVING FROM AN ART Kel-Ann S. Eyler, Teresa T. King, and Elliott L. Slocum 50

THE USE OF THE DELPHI METHOD TO ESTIMATE AUDIT & DETECTION PROBABILITIES

Carol M. Fischer 67

THE EFFECT OF JOURNAL QUALITY ON ACCOUNTING FACULTY REMUNERATION: DOES UNIVERSITY RESEARCH EMPHASIS MATTER?

Todd L. Sayre, Carol M. Graham, and James R. Hasselback 77

REVISITING THE ORGANIZATIONAL COMMITMENT MODEL IN A PUBLIC SECTOR ORGANIZATION

Anthony F. Chelte 93

NCHICA: A STRATEGIC APPROACH TO IMPLEMENTING A STATE-WIDE HEALTHCARE INFORMATION SYSTEM

Mary A. Curran and Kent E. Curran 106

UNCERTAINTY AND THE DESIGN OF DEPARTMENT OF DEFENSE CONTRACTS

Gerald M. Groshek and David J. Rose 114

5


Volume 8, Number 1 Fall 2001

TABLE OF CONTENTS (CONTINUED)

EXPLAINING CUSTOMER RESPONSE TO SERVICE RECOVERY: AN ASSESSMENT OF THE PSYCHOMETRIC PROPERTIES & PREDICTIVE VALIDITY OF MEASUREMENT SCALES FOR EXCHANGE- THEORETICAL CONSTRUCTS

Manfred M. Maute and William R. Forrester, Jr. 127

CUSTOMER RELATIONSHIP MANAGEMENT & CONSUMER CORRESPONDENCE: A THEORETICAL APPROACH

Michael McCall and Donald W. Eckrich 138

AN INVESTIGATION OF VOLUNTARY DISCLOSURES, SIZES AND FOREIGN SALES OF IRISH FIRMS APPROACH

Sinéad O�Connor & Robyn Lawrence 146

6

MANAGEMENT EDUCATION IN A GLOBALIZED WORLD

Mahendra Raj1

The Robert Gordon University, United Kingdom. The past thirty years or so has been by and large a very smooth and easy going period for business schools. There was a strong demand for their services, i.e., imparting functional business skills to young graduates enabling them to start a career in the business world. The placement opportunities for the students were also good since most of the west and US in particular had more or less steady economic growth. The staff of the business schools had it particularly good as they enjoyed high salaries relative to other subjects. All good things must come to an end and this state of affairs is no exception. However, this does not imply a prediction of doom for business schools. On the contrary, the future is going to offer the schools more challenges as well as opportunities for those that take them. These developments are in most cases extraneous and so the schools have no choice but to respond to them. We have already started to experience some of the influences that are likely to drastically alter the state of business education. Business students in the past have typically been bright young boys/girls fresh from school/college with very little or no work experience. The schools catered to them by providing knowledge and skills in the functional areas of management that enabled these young students to embark on a business career. This is no longer the case now as more and more mature and employed students decide to pursue management studies to further their career. COMPOSITION OF STUDENTS: During the last few decades the business schools have seen vast numbers of smart young boys/girls go through their corridors in the pursuit of business education. There is no indication that these numbers are likely to fall drastically in the immediate future. However, there has been a parallel growth in the demand for business education from a different segment of the population. These are the working managers with several years� experience behind them. There is an increasing realization amongst them that there is a gap in their knowledge which only a business degree can fill. Many of them view a business degree, as something missing in their portfolio of skills that would enable them to advance in their chosen business. These students cannot afford the luxury of full-time study and so choose to go for part-time or open learning courses. The demand from this group has been increasing at a rapid rate and business schools everywhere are quickly responding to this opportunity. The needs of this group of learners are different from those of the traditional students and in general they tend to be much more demanding. The part-time students are mature and experienced managers who want to update their existing skills and

1 Keynote address, 8th Annual Conference, ASBBS.

Journal of Business and Behavioral Sciences

7

learn new softer skills such as HR, communication, negotiation, board room, etc. These managers have to juggle their time between their job, studies and family. Lectures and other modes of delivery used in the past are not appropriate for this group and so the schools need to develop new modes of delivery to suit them. Internationalization of Business Education: The last couple of decades have seen the Internationalization of business. Firms have expanded internationally in search of new markets. Many firms have moved production abroad to avail of cheaper labor, raw materials or to be nearer markets. All this has created a need for employees who are equipped to work abroad � business graduates who understand the culture and can speak the language. The rest of the western economies realized this need much earlier. The large domestic job market of US had insulated it from this trend until recently. The training provided by the western business schools was inadequate and unable to prepare the students for working in emerging economies. So, businesses were increasingly recruiting foreign students who were educated in the west. The business schools have now responded by increasingly taking on the responsibility of filling this need. Many schools now include a number of international business related modules into their core curriculum. Others offer specialized degrees in international business. Schools have started emphasizing international experience for both the students and faculty. Several schools in both the US and Europe have exchange programs for students and staff. Foreign language requirements are being introduced into many courses. Cultural education is being imparted indirectly by increasing the mix of international students and faculty in the schools. Globalization of Management Education: The nineties have brought about the globalization of business. The largest five hundred firms in the world are now so big and their operations so widely spread across the globe that it is becoming irrelevant to attribute a nationality to them. Many of them originated in one country but are now owned or controlled by groups in another. Companies have been gobbling up each other through mergers and acquisitions and of late this trend has spread to the emerging countries as well. These trends coupled with the rapid growth of the emerging economies have given them an insatiable appetite for well-trained business graduates. Most of these developing nations do not have adequate number of educational institutions to meet this requirement. This demand for business graduates and the shortage of institutions offering business education has resulted in large number students from these countries pursuing business education in US, UK and other western countries. It would be futile to talk about changes in business education without mentioning the impact of IT revolution. Internet has made it easy to access information about courses from anywhere in the world. Several new ways of communicating to the students are now available besides the traditional lectures that make location irrelevant. Faculty and students can communicate to each other through email, chat sessions, discussion group and video conferencing. Lecture

8

notes, video presentations, power point presentations, cases, assignments and examinations can be made available through the Internet, CDs and videos. Even assessments can be done through the Internet making it possible for students from most parts of the world to do a business degree from the west. The cozy period of the last three decades when funding was not a problem is coming to an end for the business schools. Government is cutting down on the support for Universities in the western world. Schools seeking other sources of funding are turning towards foreign students who are an abundant source of funding, especially of the kind without strings attached. Schools from the US, UK, Canada and Australia compete with each other to recruit the best students from developing countries like China, India, Brazil and Malaysia. This has reached a stage where there are regular exhibitions manned by these Universities seeking to attract foreign students. It is no longer a surprising thing to go to China or Malaysia and meet some one with a degree from a Harvard who has actually never been to US. This is a trend that is catching on with some big names involved. U of Chicago for instance, offers part-time MBA from three centers around the world. This is done through several weeks of intense workshops supplemented by Internet materials etc. Some schools have gone a step further by franchising their degrees to local institutions. The basic teaching is done by faculty of the local institution according to the course outline provided by the franchiser. The franchiser�s faculties visit the site to conduct intense workshops and also to do the assessment. Supporting material is provided through the Internet. Open learning is probably the one mode of delivery that has benefited the greatest from the IT advancements. Universities all over the World have taken to this mode of delivery in a big way. This is also the one mode of delivery that is likely to have the most impact on globalization of business education. This enables students from any part of the world to obtain a prestigious degree from the west without having to leave the country. Experienced and mature students, both domestic and international, are likely to prefer open learning as this offers a more flexibility and convenience. FUTURE TRENDS: All these changes obviously mean that the good old times of the past are gone. An uncertain future lies ahead. Like all forecasting, trying to predict the future of business education may be a futile exercise but the process of doing so makes us more prepared for the unpredictable. Given the trends that we are currently observing it may be a reasonable extrapolation to assume that part-time and open learning students are likely to increase. Location and buildings will become less and less relevant as the modes of delivery incorporate the benefits of the information revolution. At the undergraduate level the content of the course may change but at a steady pace. There will more stress on global aspects of business; greater mix of students and faculty and exchange programs will become a standard feature. Different modes of delivery will be incorporated but the bulk of the student population will continue to be full time students and less experienced. At the postgraduate level there is likely to be a bifurcation of degrees


9

with the less experienced students going for the full time MBA, MSc or MA as is common in Europe and Australia. The experienced learners seeking softer managerial skills would opt for the Executive MBA. The top schools will use the new modes of delivery to become more global and exploit the benefits of brand name. This is already happening with Chicago and many such top schools already having centers around the world. More and more schools will start offering business degrees all over the world. So, then how would schools differentiate from each other? The key I believe lies in research. Business is obviously not going to spend money and resources on research that does not immediately result in a tangible product. So, the onus of the leading the research in the field rests on business schools. Schools would have to emphasize research by faculty to position itself as a center of knowledge. In a world where conventional teaching and lectures are becoming obsolete, stressing on teaching quality alone cannot help a school survive. Research will provide them with the brand name and recognition that they can market worldwide. The faculty also would find that it is in their personal interest to build up their own profile and make themselves marketable.

10

RANKING DIFFERENCES AMONG THE SHARPE RATIO AND RELATED TOTAL PORTFOLIO RISK

ADJUSTED PERFORMANCE MEASURES

Bruce D. Bagamery Central Washington University – Lynnwood Center

ABSTRACT: This paper examines ranking differences of six risk adjusted portfolio performance measures: the Sharpe Ratio (Sharpe, 1966), and five related measures, including the Excess Return Sharpe Ratio (Sharpe, 1994), the Information Ratio (Grinold, 1989), the M-Square, or Risk Adjusted Return (Modigliani & Modigliani, 1997), the Excess Standard-Deviation-Adjusted Return (Statman, 2000), and Graham-Harvey Measure 2 (Graham & Harvey, 1997). These performance measures focus on total risk, where risk is measured as standard deviation. Sensitivity analyses are used to assess differences between these performance measures over relevant parameter values. Five of the measures (all except the Information Ratio) produce nearly identical values of performance in most cases. Exceptions occur for low volatility portfolios; in those cases, the Graham-Harvey Measure 2 is appropriate. INTRODUCTION: Ever since its introduction, the Sharpe Ratio (Sharpe, 1966) has been the predominant measure of portfolio risk adjusted return when consideration focuses upon the investor�s total portfolio. In investments textbooks (e.g., Bodie, Kane, & Marcus, 1999), and in empirical studies it is the most likely performance measure to be discussed, when risk is measured as standard deviation of the entire portfolio. However, in the past few years, several other related total risk adjusted performance measures have been developed. These include the Excess Return Sharpe Ratio (Sharpe, 1994), the Information Ratio (Grinold, 1989), the M-Square, or Risk Adjusted Return (Modigliani & Modigliani, 1997), the Excess Standard-Deviation-Adjusted Return (Statman, 2000), and Graham-Harvey Measure 2 (Graham & Harvey, 1997). The purpose of this study is to discuss these other measures and then to examine how closely they are related and under what conditions they provide different information regarding portfolio performance. The paper is arranged as follows: First, each of the six risk adjusted performance measures is defined and discussed. Then, for three pairs of measures, the difference between the two measures is determined as a function of differences between returns, standard deviations and correlations of the portfolio and benchmark, and sensitivity analyses are performed to assess the magnitude of the differences. Magnitudes of these differences are examined with reference to estimated annual values over the 1926-1996 periods, and annualized estimated monthly values from the Morningstar Principia Pro Mutual Funds database. Conclusions appear in the final section.


11

ORIGINAL SHARPE RATIO, OSR: The Original Sharpe Ratio OSRP for portfolio P, introduced as the reward-to-variability ratio in Sharpe (1966), is defined as

OSRP = Mean(RP � RF) / Stdev(RP), where RP and RF are respectively the total percentage returns on the portfolio P and a risk free asset F, and Mean (.) denotes the arithmetic mean. Stdev (RP) is the standard deviation of returns on portfolio P. The risk free asset F is often specified as three month Treasury Bills, even though in his 1966 study, Sharpe used ten-year Treasury bonds as his risk free rate proxy. This formulation is consistent with the interpretation that a line can be drawn in ex post return � standard deviation space between the risk free asset F and each portfolio P; Bodie, Kane, & Marcus (1999, p. 184) define this line as a �capital allocation line.� This concept is just a generalization of the idea of an ex post capital market line as discussed in Sharpe (1970), and more recently in Sharpe, Alexander, & Bailey (1999), where the relevant risky portfolio is now specified to be an overall market portfolio M. Essentially, the value of the Original Sharpe Ratio for any portfolio P equals the slope of the capital allocation line for that portfolio. More steeply sloped lines indicate greater risk-adjusted performance. This original formulation of the Sharpe Ratio appears in most investments texts. Along with several other performance measures, the Original Sharpe Ratio OSR is illustrated in Figure 1. Portfolios P, M, and F denote respectively the analyzed portfolio, the market portfolio, and the risk free asset. Assuming a zero standard deviation for the risk free return defines a reference risk free portfolio at point F' on the graph. The Original Sharpe Ratios OSRM and OSRP are shown as the slopes of the straight lines F'M and F'P. These slopes are the ratios of excess return to standard deviation for the market portfolio and for portfolio P. Portfolio P outperforms the market portfolio M because it lies on an OSR line with steeper slope.

EXCESS RETURN SHARPE RATIO, XRSR: A second formulation of the Sharpe ratio was called the reward to variability ratio in Sharpe (1975), the Sharpe Ratio in Sharpe (1994), and the Excess Return Sharpe Ratio in Sharpe (1996). Using Sharpe�s (1996) notation, the Excess Return Sharpe Ratio XRSRP for portfolio P is defined as

XRSRP = Mean (RP � RF) / Stdev (RP � RF), the ratio of the arithmetic mean of excess returns of portfolio P over the risk free asset F to the standard deviation of those excess returns. Sharpe (1994) emphasized that the excess (or differential) return can represent the result of a zero-investment strategy, a strategy that can be established using no money by taking a long position in the portfolio and a short position in the riskless asset. This Excess Return Sharpe Ratio formulation has not appeared in many investments texts, but did appear in some earlier editions of Sharpe�s investments text, for example Sharpe (1985).

Bagamery, Bruce

INFORMATION RATIO, IR: Sharpe (1994) also pointed out that the definition of the Excess Return Sharpe Ratio above in terms of excess return over a risk free rate can be generalized as an excess return RB over a benchmark appropriate for the particular portfolio in question. In this form, where RB replaces RF, we obtain the information ratio introduced in Grinold (1989), and exposited thoroughly in Grinold & Kahn (2000):

IRP = Mean(RP � RB) / Stdev(RP � RB).

In the information ratio framework, the focus is on active portfolio

management, and the excess return over the benchmark are called the active return. Goodwin (1998) shows that, for a linear market model,

(RPt � RFt) = αP + βP (RBt � RFt) + εt , when the benchmark is the market index, and when the portfolio in question is required to have a βP = 1.0, each excess return over the benchmark can be expressed as

(RPt � RBt) = αP + εt . The arithmetic mean of these excess returns, Mean (RPt � RBt), equals αP,

the portfolio�s market model alpha. The standard deviation of these returns over the benchmark, Stdev (εt), is also called ωP, the portfolio�s idiosyncratic risk, or

Return

Standard Deviation

RAPP

RP

RZ

RF

RM = RAPM

σF σM σP

F

M

GH2P

OSRP

OSRM

P

eSDARP

F'

Z

L

Z'

Figure 1. Graphical Illustration of Original Sharpe Ratio (OSR P), Risk Adjusted Performance(RAPP), Excess Standard-Deviation-Adjusted Return (eSDARP), and Graham-Harvey Measure 2(GH2P).


13

market model residual risk. Therefore we can also express the portfolio�s information ratio as its alpha-omega ratio, IRP = αP / ωP . Thus, we can interpret the information ratio as the portfolio manager�s skill in security selection measured by the portfolio�s alpha, divided by the additional residual risk the portfolio has taken on as a result of the manager�s security selections. Goodwin (1998) points out that the information ratio has also been called the signal to noise ratio for an active manager, the return to variability ratio, and the appraisal ratio. M-SQUARE, OR RISK ADJUSTED RETURN, RAP: One shortcoming of the Sharpe Ratio is that it may be difficult for some people to interpret, since it is expressed as a ratio of excess return to variability. This difficulty motivated Modigliani & Modigliani (1997) to introduce their measure of Risk Adjusted Performance (RAP), which is also known as the M-Square (or M2) measure, after its authors. This measure ranks exactly the same as the Original Sharpe Ratio, but has a more intuitively obvious interpretation in terms of percentage return instead of Sharpe�s excess return per unit risk.

Algebraically, the M-Square risk adjusted performance RAPP is RAPP = Mean(RF) + Mean(RP � RF) (Stdev(RM) / Stdev(RP) ), and it is denominated in units of percentage return. Substituting in the above expression gives RAPP in terms of the portfolio�s Original Sharpe Ratio OSRP: RAPP = Mean(RF) + Stdev(RM) OSRP, which is denominated in units of percentage return. This expression emphasizes the strict linear dependence between the RAP measure and the Original Sharpe Ratio.

In Figure 1, the RAP measure for portfolio P, RAPP, places the portfolio on its capital allocation line, which has a slope equal to the portfolio�s Original Sharpe Ratio OSRP. Then we form portfolio L, which levers or unlevers the portfolio using the risk free asset F' until it has the same standard deviation as the market return. The risk adjusted performance RAPP for portfolio P is the total return on that levered portfolio L having an original Sharpe Ratio ORSP and a standard deviation equal to that of the market Stdev(RM). For the market portfolio, its risk-adjusted performance RAPM simply equals the realized market return RM. RAP measures total return, not excess return.

EXCESS STANDARD-DEVIATION-ADJUSTED RETURN, eSDAR: Frequently, it is more illuminating to define the risk adjusted performance measure as an excess return over the market return. This provides an answer to the question, �How much better did my portfolio perform than the benchmark portfolio?� Statman (2000) defines a measure called eSDAR, the �excess standard-deviation-adjusted return� which he initially developed in ex-ante form in Statman (1987). For portfolio P, eSDARP can be expressed as eSDARP = RAPP � RAPM = (OSRP � OSRM) Stdev(RM).

Bagamery, Bruce

14

In Figure 1, eSDARp is identified as the distance ML, the vertical distance between the Original Sharpe Ratio lines for portfolio P and for the market portfolio. If portfolio P underperformed the market, the distance LM would indicate a negative amount. The Excess Standard-Deviation-Adjusted Return eSDAR produces rankings identical to those based on the Risk Adjusted Performance measure RAP, and to those based upon the Original Sharpe Ratio, OSR. The identical rankings occur because the three measures are exact linear transformations of each other. Therefore, which measure is selected depends solely on whether the user prefers to denominate performance as 1)excess return (over the risk free rate) per unit of standard deviation (OSR), 2)total return when scaled to have the same standard deviation as the market (RAP), or 3)excess return (over the market) when scaled to have the same standard deviation as the market (eSDAR). GRAHAM-HARVEY MEASURE 2, GH2: The Original Sharpe Ratio, M-Square, and eSDAR, all assume that the risk free asset has zero standard deviation, and thus it is uncorrelated with all other portfolios. For many equity portfolios that assumption may be appropriate, but it is less appropriate for low volatility portfolios such as money market funds. The Graham-Harvey Measure 2, or GH2 (Graham & Harvey, 1997), relaxes that assumption, allowing the "risk free" asset to have positive standard deviation and non-zero correlations with risky portfolios. Their analysis starts by drawing an ex post efficient frontier line in return � standard deviation space through portfolio P and the (now not entirely) risk free asset F. Bodie, Kane, & Marcus (1999) refer to this curved line as a �portfolio opportunity set.� The exact degree of curvature of the line depends upon the correlation between portfolios P and F; the lower the correlation, the more curved the line. For perfect positive correlation, the portfolio opportunity set is a straight line connecting P and F.

Graham and Harvey then define a portfolio which we can call Z, a

combination of P and F, which has the same standard deviation (and variance) as

the market:

VAR(RZ) = VAR(RM).

Then the two-asset portfolio variance equation, VAR(RZ) = w2VAR(RP) + 2w(1-w)COV(RP,RF) + (1-w)2VAR(RF) = VAR(RM), is solved (using the quadratic formula) for w, the weight of portfolio P in the combination portfolio Z. Once the weight w is obtained, then the return RZ on the combination portfolio is obtained as, RZ = w RP + (1-w) RF. The Graham-Harvey GH2P measure for portfolio P is defined as an excess return: the return RZ on this combination portfolio minus the return on the market portfolio RM: GH2P = RZ � RM.


15

It is an excess return relative to the market, just like the eSDARP measure described above. In Figure 1, the curved line FZP traces out some of the portfolio opportunity set. Portfolio Z on the line has the same standard deviation as the market. The length of the vertical line ZM shows GH2P, the Graham-Harvey measure 2 for portfolio P. In this case GH2P exceeds eSDARP, which was identified above as the length of the line LM. However, for high values of Corr(RP,RF), it is possible that GH2P will be less than eSDARP. This case is illustrated by the dotted line portfolio opportunity set FZ'P, where the distance Z'M is less than LM. The magnitudes of the differences between the Graham Harvey Measure 2 and the Excess Standard-Deviation-Adjusted Return are examined below.

Therefore, to compute Graham-Harvey Measure 2, we find the portfolio Z on the efficient frontier line between portfolio P and the risk free portfolio F that has the same standard deviation as the market portfolio M; the return difference between that portfolio Z and the market M is GH2P. In contrast, to compute Graham-Harvey Measure 1 (Graham & Harvey, 1997), we find a portfolio, say Q, on the efficient frontier line between the market and the risk free asset that has the same standard deviation as portfolio P; the return difference between portfolio P and portfolio Q is GH1P. GH1 values are not comparable between portfolios if the portfolios have different standard deviations; therefore Graham-Harvey Measure 1 is not examined further in this study.

ORIGINAL SHARPE RATIO VS EXCESS RETURN SHARPE RATIO: If the risk free asset is indeed riskless, with Stdev (RF) = 0, then the Original and Excess Return Sharpe Ratios are equivalent. The two measures differ only in their denominators. The relation between the two standard deviations Stdev(RP�RF) and Stdev(RP) is defined by Stdev(RP�RF)= Sqrt[Stdev2(RP)�2 Stdev(RP) Stdev(RF) Corr(RP,RF)+Stdev2(RF)], where Corr(RP,RF) is the correlation between the returns on P and F. Empirically, for equities, the two standard deviation terms are nearly the same; primarily because of the very low variability of the risk free return during the past several years. To assess the extent of the difference between the two measures, we can examine the ratio of the two standard deviations, which also is the ratio of the two Sharpe Ratios: Stdev(RP � RF) / Stdev(RP) = OSRP / XRSRP. For any given value of risk free asset standard deviation Stdev(RF), the extent to which the ratio deviates from 1.0 reflects the effect of 1)the correlation Corr(RP,RF) between the portfolio and the risk free asset and 2)the portfolio standard deviation Stdev(RP). Values of the ratio between 0.9 and 1.1 falls with a range where there is less than a 10% difference in the magnitudes of the two different measures. Figure 2 displays a sensitivity analysis of the OSR / XRSR ratio for Stdev(RF) = 3.35%, which is the historical value over the 1926-1996 period

Bagamery, Bruce

16

(Bodie, Kane, & Marcus (1999), p. 135). The figure displays Corr(RP,RF) from �0.50 to 0.50; for a sample of 1499 Morningstar domestic equity mutual funds over the 60-month period from January 1995 through December 1999, the correlations between their returns and the three-month T-Bill return ranged from a minimum of �0.264 to a maximum of 0.360 with an average value of 0.013. Each of the eight lines on the figure corresponds to a Stdev(RP) ranging from 5% to 40%; for the same sample of 1499 funds over the 1995-1999 period, annualized standard deviations ranged from 3.9% to 38.2%, with an average value of 15.4%.

Figure 2. Sensitivity of the OSR/XRSR Ratio to Corr(Rp,Rf) and Stdev(Rp) for Stdev(Rf) = 3.35%, the 1926-1996 Historical Value.

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

-0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50

Corr(Rp,Rf), Correlation between Portfolio and Risk Free Asset Return

OSR

p/XR

SRp,

Rat

io o

f Orig

inal

Sha

rpe

Rat

io

to E

xces

s R

etur

n Sh

arpe

Rat

io

Stdev(Rp)= 5% Stdev(Rp)= 10% Stdev(Rp)= 15% Stdev(Rp)= 20%Stdev(Rp)= 25% Stdev(Rp)= 30% Stdev(Rp)= 35% Stdev(Rp)= 40%

It is clear from Figure 2 that, for the historical risk free rate of 3.35%, the Original and Excess Return Sharpe Ratios have values within 10% of each other for most combinations of correlation and portfolio standard deviation. The only exceptions occur for combinations of low portfolio standard deviations and low correlations. The OSR/XRSR ratio exceeds 1.1 for Stdev(RP) = 5% and Corr(RP,RF) below about .18, and for Stdev(RP) = 10% and Corr(RP,RF) below about -.15.

In recent years, the risk free asset standard deviation Stdev(RF) has been much lower than the historical value of 3.35%. During the 1995-1999 period, the annualized T-Bill standard deviation is much closer to zero at a value of 0.111%. Figure 2A shows that Stdev(RF) = 0.111% results in much lower deviations of the OSR/XRSR ratio, within a range of 0.99 to 1.01, implying that OSR and XRSR never differ by more than 1%. This result is consistent with the findings of Modigliani & Modigliani (1997, p.53) who find that Stdev(RP) is �very close� to


17

Stdev(RP � RF). Thus, for practical purposes, there is little difference between using the Original Sharpe Ratio and the Excess Return Sharpe Ratio.

GRAHAM-HARVEY MEASURE 2 VS EXCESS STANDARD-DEVIATION-ADJUSTED RETURN: The Graham-Harvey Measure 2 has been shown to be conceptually superior to those measures directly linearly related to the Original Sharpe Ratio, including the MM RAP measure and eSDAR. Essentially, those measures reflect a special case of GH2, in which Stdev(RF) = Corr (RP,RF) = 0. Both GH2 and eSDAR define an excess return measure over the market or benchmark portfolio. However, the conceptual superiority of GH2 comes at the cost of obtaining additional input data and performing more tedious computations. Is there a sufficient difference between the values of the two measures to merit the additional computational burden of GH2?

Figure 2A. Sensitivity of the OSR/XRSR Ratio to Corr(Rp,Rf) and Stdev(Rp) for Stdev(Rf) = 0.110657%, the 1995-1999 value.

0.985

0.990

0.995

1.000

1.005

1.010

1.015

-0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50


OSR

p/XR

SRp,

Rat

io o

f Orig

inal

Sha

rpe

Rat

io

to E

xces

s R

etur

n Sh

arpe

Rat

io


Figure 3 shows how the difference between GH2 and eSDAR varies for different values of Corr (RP,RF) and Stdev(RP), given a portfolio return of 15%, and using 1926-1996 historical values for the returns and standard deviations of the risk free asset and market portfolio from Bodie, Kane & Marcus (1999). Differences increase as the portfolio-risk free correlation moves away from zero in either direction; however, for values of Stdev(Rp) of 20% and higher, the differences are not much more than 20 basis points (0.2%) over the correlation range of �0.5 to 0.5. Since the S&P 500 standard deviation was 20.39% over that period, the majority of portfolios had standard deviations greater than 20%, and therefore the difference between GH2 and eSDAR would be 20 basis points or less.

Bagamery, Bruce

18

However, as the portfolio standard deviation decreases, the differences between GH2 and eSDAR become greater and greater, with very large differences for Stdev(RP) values of 10% and 5% over wide ranges of correlation values. This result is consistent with Graham & Harvey's (1997, p. 66) concern that "misleading inferences about the performance of low-volatility funds" could occur if GH2 is not used. Based on these historical values, a reasonable rule of thumb would be that the Graham-Harvey Measure 2 should be used instead of one of the Original Sharpe Ratio related measures if the annual portfolio standard deviation is 15% or less. Figure 3 assumed a portfolio total return RP of 15%. As RP decreases below 15%, the GH2-eSDAR differences decline, and as RP increases above 15%, the differences increase, but only slightly.

Figure 3. Sensitivity of the Difference between GH2 and eSDAR to Corr(Rp,Rf) and Stdev(Rp) , Given Rp = 15.00% , Using 1926-1996 Annual Values Rm=

12.50% , Rf= 3.76% , Stdev(Rm)= 20.39% , and Stdev(Rf)= 3.35% .

-1.0%

-0.8%

-0.6%

-0.4%

-0.2%

0.0%

0.2%

0.4%

0.6%

0.8%

1.0%

-0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50


GH

2 - e

SDA

R (%

ann

ualiz

ed re

turn

)


Figure 3A replicates the sensitivity analysis using the recent 1995-1999 annualized values of RM = 29.77%, RF = 5.25%, Stdev(RM) = 13.94% and Stdev(RF) = 0.11% from the Morningstar Principia Pro database. For an assumed RP = 20%, the differences between GH2 and eSDAR are much smaller. Even for RP = 5%, the difference is within 20 basis points for relevant correlation values. And, except for Stdev(RP) = 5%, the differences are two basis points (0.02%) or less. This reduction in differences is due primarily to the much smaller recent value of Stdev(RF) of 0.11% compared to the historical value of 3.35%. As in the previous analysis, the differences decline (increase) as the assumed RP declines (increases) from 20%. In addition, the previously discussed sample of 1499 Morningstar equity mutual funds yielded nearly identical values of GH2 and eSDAR, with the maximum and minimum differences equal to 69.1 and �50.2 basis points respectively. Furthermore, the Spearman rank correlation across the 1499 funds


19

between GH2 and eSDAR was 0.9999978, indicating almost perfect ranking correspondence. Therefore, as long as the portfolio under examination is not a very low standard deviation portfolio, such as a money market mutual fund, the difference between the calculated value of the Graham-Harvey Measure 2 and the excess Standard-Deviation-Adjusted Return will likely be at most just a few basis points.

Figure 3A. Sensitivity of the Difference between GH2 and eSDAR to Corr(Rp,Rf) and Stdev(Rp) , Given Rp = 20.00% , Using 1995-1999 Annualized Values Rm=

29.77% , Rf= 5.25% , Stdev(Rm)= 13.94% , and Stdev(Rf)= 0.11% .

-0.4%

-0.3%

-0.2%

-0.1%

0.0%

0.1%

0.2%

0.3%

0.4%

-0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50


GH

2 - e

SDA

R (%

ann

ualiz

ed re

turn

)


EXCESS RETURN SHARPE RATIO VS INFORMATION RATIO: A third ranking comparison of performance measures merits only a brief discussion. This comparison is between the Excess Return Sharpe Ratio XRSR and the information ratio IR. As described above, both of these measures are ratios of a mean excess return to the standard deviation of the excess return. XRSR computes the excess returns relative to the return on a risk-free asset, while IR computes excess returns relative to the return on an appropriate benchmark, which can be the market index if we are assessing performance of funds with a target beta of 1.0. The conditions under which ranking similarities exist are much more difficult to assess because the assessments are made against two different benchmarks that are not particularly highly correlated. For example, over the 60 months ending December 1999, the correlation between returns on the S&P 500 and three month Treasury Bills was practically zero at 0.001506.

In fact, as Modigliani & Modigliani (1997, p. 52) indicate, the two measures really assess different kinds of performance. Using the risk-free asset as a benchmark is more appropriate from the investor's point of view, where additional return on average can only be obtained by a willingness to accept

Bagamery, Bruce

20

greater levels of risk. On the other hand, using a market index as a benchmark is more appropriate for a fund manager responsible for constructing a portfolio designed to mirror (or track) the returns on the benchmark. Therefore, a sensitivity analysis of differences between the Information Ratio and the Excess Return Sharpe Ratio does not provide any useful information, since the two measures measure different types of performance.

IMPLICATIONS FOR PORTFOLIO MANAGEMENT: Investors and portfolio managers can more accurately assess portfolio performance when performance measures explicitly account for risk. Implementing these risk adjusted performance measures is facilitated if they are more easily understood by users. Since both the Risk Adjusted Performance measure and the Excess Standard-Deviation-Adjusted Return are denominated in percentage return, they are superior to the Sharpe measures, which are denominated in return per unit of standard deviation. Implementing the more precise Graham-Harvey Measure 2 (GH2) requires additional data and computations as described above. However, using recent monthly data, the only time this increased precision is necessary is for the minority of low variability portfolios that have annual standard deviations below ten percent; the rest of the time there is no difference between the GH2 rankings and those of the measures that assume constant returns on the risk free asset. CONCLUSIONS: This study has reviewed six different portfolio risk-adjusted performance measures where risk is defined as total risk. Three of these measures, the Original Sharpe Ratio, the Risk Adjusted Performance measure, and Excess Standard-Deviation-Adjusted Return, all rank identically because they are linear transformations of each other. The study provided a graphical comparison of these three measures with each other and with the Graham-Harvey Measure 2.

A sensitivity analysis of the differences between the Original Sharpe Ratio and the Excess Return Sharpe Ratio showed that the differences are small, but will be largest for low values of portfolio standard deviation combined with highly negative values of portfolio correlation with the risk free asset. A sensitivity analysis of the differences between Graham-Harvey Measure 2 and the Excess Standard-Deviation-Adjusted Return showed very small differences, with differences increasing for portfolios with low standard deviations. Sensitivity analyses of the differences between the Excess Return Sharpe Ratio and the Information Ratio are not appropriate since information ratio focuses on a different concept of performance from the other measures.

The primary conclusion from the sensitivity analyses is that, unless the portfolios under examination have very low standard deviations, it makes little difference whether one uses Original Sharpe Ratio (or one of its two transformations Risk Adjusted Performance or Excess Standard-Deviation-Adjusted Return), Excess Return Sharpe Ratio, or Graham-Harvey Measure 2. All five measures provide nearly identical performance rankings.


21

This study could be extended in several ways by examining additional data sets. First, the funds examined were domestic equity funds; examining a sample of money market funds would show more directly the extent to which these lower volatility funds would actually benefit from the additional computations necessary for the more accurate GH2 computation instead of the simpler eSDAR value. Second, because international mutual funds should display lower variability due to the benefits of increased diversification, a sample of international funds could also be examined to see if the GH2 measure differs from the eSDAR measure. Finally, the study focused on annual and monthly returns; would the results be different for weekly or daily returns? REFERENCES Bodie, Z., Kane, A. & Marcus, A.J. (1999). Investments, 4th edition. New York,

McGraw-Hill Companies, Inc. Goodwin, T.H. (1998). The Information Ratio. Financial Analysts Journal, 54

(4), 34-43. Graham, J. R., & Harvey, C.R. (1997). Grading the Performance of Market-

Timing Newsletters. Financial Analysts Journal, 53 (6), 54-66. Grinold, R. C. (1989). The Fundamental Law of Active Management. Journal of

Portfolio Management, 15 (3), 30-37. Grinold, R. C. & Kahn, R.N. (2000). Active Portfolio Management, 2nd edition.

New York, McGraw-Hill. Modigliani, F, & Modigliani, L. (1997). Risk-Adjusted Performance: How to

Measure It and Why. Journal of Portfolio Management, 23 (2), 45-54. Sharpe, W.F. (1996). The Excess Return Sharpe Ratio.

<http://www.sharpe.stanford.edu/mia_rr5.htm>, January 2, (Accessed January 18, 2000).

Sharpe, W.F. (1994). The Sharpe Ratio. Journal of Portfolio Management, 21 (1), 49-58.

Sharpe, W.F. (1985). Investments, 3rd edition. Englewood Cliffs, NJ, Prentice Hall.

Sharpe, W.F. (1975). Adjusting for Risk in Portfolio Performance Measurement. Journal of Portfolio Management, 1 (2), 7-19.

Sharpe, W.F. (1970). Portfolio Theory and Capital Markets, New York, McGraw-Hill.

Sharpe, W.F. (1966). Mutual Fund Performance. Journal of Business: A Supplement, 39 (1, part 2), 119-138.

Sharpe, W.F., Alexander, G.J. & Bailey, J.V. (1999). Investments, 6th edition. Upper Saddle River, NJ, Prentice Hall, Inc.

Statman, M. (2000). Socially Responsible Mutual Funds. Financial Analysts Journal, 56 (3), 30-39.

Statman, M. (1987). How Many Stocks Make a Diversified Portfolio? Journal of Financial and Quantitative Analysis, 22 (3), 353-363.

22

THE HIGHEST SHORT-TERM IPO RETURN AND ITS TREND: THE “A” SHARE EXPERIENCE ON THE

SHANGHAI STOCK EXCHANGE

Anthony Yanxiang Gu State University of New York

ABSTRACT: The �A� Share IPOs in China exhibited the highest short-term returns compared to IPOs around the world and the returns have revealed a downward trend toward a typical level. Possible reasons for the highest returns include excess demand for new shares, underwriter�s strong risk aversion, and extremely long institutional lags. Rapid growth in IPOs accommodating to the excess demand for new shares, decreasing and now normal institutional lags, lower rates of inflation, lower risk in the stock market, and investors� experienced investment behavior may partially explain the trend.

INTRODUCTION: The �A� share IPOs in mainland China have exhibited the highest short-term returns that have ever recorded in the IPO history, i.e., 647 percent on average in 1993 (Chau, Ciccotello and Grant, 1994) and 217 percent in 1994 (Gu and Jin, 1998). Initial returns in other emerging markets include, 80.3 percent in Malaysia (Isa, 1993), 78.5 percent in Brazil (Aggarwal, Leal and Hernadez, 1993), 78 percent and 58 percent, respectively in Korea (Dhatt, Kim, and Lim, 1993, and Kim, Krinsky, and Lee, 1993), 45 percent in Taiwan (Chen, 1992), and 58 percent in Thailand (Wethyavivorn and Koo-Smith, 1991). In developed markets, the average initial returns range from 9.5 percent in France to 17.4 percent in the U.S.A1. (Loughran, Ritter and Rydqvist, 1994, and Ritter, 1997).

This study examines whether the world�s highest short-term IPO returns are temporary and approaching some typical level. The study also identifies the trend of the returns and analyzes the factors that may contribute to the highest returns and the trend. There are two forms of common stocks in China: �A� shares and �B� shares. The �A� shares are available only to domestic investors and are priced/traded in the Chinese currency, the Yuan. �B� shares are issued only to foreign investors (available to domestic investors from February 20, 2001) and are priced/traded in U.S. dollars on the Shanghai Stock Exchange (SHSE) and in Hong Kong dollars on the Shenzhen Stock Exchange (SZSE). Both �A� and �B� shares carry the same voting rights and dividends, with �A� share dividends paid in Chinese yuan and �B� share dividends paid in either U.S. or Hong Kong dollars, adjusted for exchange rates. The SHSE officially opened on December

1 Data updated in September 2000 on Ritter�s website.


23

19, 1990 with four listed �A� shares. The trading of �B� shares started in February 1992 on the SHSE. By the end of 1998, the trading floor in SHSE was handling 438 listed companies (528 stocks including both �A� and �B� shares) with market capitalization of 1,062.6 billion Chinese yuan, equivalent to 13.3 percent of GDP (Shanghai Securities Yearbook). As of April 19, 2001, there were 597 �A� shares and 55 �B� shares listed on the SHSE, with a total market capitalization of 3,031 billion yuan (about 365 billion US dollars), of which �A� shares account for 2,961 billion yuan and �B� shares 70 billion US dollars (Shanghai Securities Daily). The plan of the paper is as follows. Section II presents the estimation of the short-term returns and analyzes the factors that contribute to the super high returns. Section III reports the trend of the returns and provides some explanations for the trend. Section IV concludes the paper.

THE HIGHEST SHORT-TERM RETURNS: This research includes 478 companies that went IPO from 1984 (the first one in China) until September 26, 2000 and are traded on the SHSE. Daily open and close prices are provided by the Shanghai Stock Exchange. Issue prices, dates of issue and first trade are from Shanghai Securities Yearbook, and from Shanghai Pudong Development Bank. The short-term returns are defined and computed as follows:

Initial return = (first trade price - issue price)/issue price

First day return = (first trading day close price - issue price)/issue price The average initial return is 398 percent and the average first day return is

406 percent. Particularly, stocks issued in 1988, 1989, and 1991 have an average initial return over 1,000 percent, and only two years, 1994 and 1996, out of the fifteen years experienced average initial and first day returns below 100 percent.

There are several possible explanations for the extraordinarily high short-term returns or underpricing in the Chinese IPO markets, particularly before mid 1990s. First, excess demand for the new shares is obvious in China�s stock markets. Researches on the demand side of stock markets include Stulz and Wasserfallen (1995), Domowitz, Glen and Madhaven (1997), and Chen, Lee and Rui (2001). Before early 1990s, bank deposits and Treasury bonds were the major investment instruments available to Chinese people. The estimated national savings rate is about 45 percent of GDP. For Chinese savers, the real returns on bank deposits and Treasury bonds were actually negative due to the relatively high rates of inflation (Miurin and Sommariva, 1993) before 1993 and slightly above zero from 1993 to 1996 (Chen, Lee and Rui, 2001). Under the circumstances, stocks were the only instruments that could be expected to provide real returns, but the equity investment opportunities were scarce. As of the end of 1993, the total market capitalization of both SHSE and SZSE was 353.1 billion yuan, only 10.2 percent of the GDP (China Financial Outlook, The People�s Bank of China, 1999). This was very low compared to 349.8 percent in Hong Kong,

Gu, Anthony

89.2 percent in Taiwan and 107 percent in Thailand (Emerging Stock Markets Factbook, 1995. International Finance Corporation). When investment opportunities are scarce, IPO shares probably provide the best opportunity to earn higher returns. However, due to the serious imbalance of supply and demand, the �A� shares are distributed through a lottery system. Winners are selected via a random number generating scheme and are entitled to purchase (usually five hundred) shares at the issue price, but only a small percentage of the subscribers would win the lottery. In some cases, the �hit rate� could be as low as 0.13 percent (Shanghai Securities Yearbook, 1999). The lack of investment opportunities leads to pent-up demand for stocks (Chau, Ciccotello and Grant, 1994, and Gu and Jin, 1998). Once a new issue reaches the secondary market, the large number of unsuccessful bidders in the primary market would be willing to pay high prices for the shares and therefore drive up the prices. Chen, Lee, and Rui, (2001) also point out that the highly speculative behavior of Chinese investors may push up �A� share prices.. As a further evidence of the relation between the high short-term return and the excess demand for new shares, the short-term returns on the �B� shares are much lower than that on the �A� shares. The valuation methods for pricing the IPOs are the same for both the �A� shares and the �B� shares, but the average initial return was 9.36 percent and the average first day return was 15.43 percent respectively on the �B� shares from 1992 to 1996 (Gu and Jin, 1999). For the same time period, the average initial and first day returns on the �A� share IPOs were 148 percent and 145 percent, respectively. A major reason for the sharp difference is that the demand for the �B� shares in the secondary market is much lower than that for the �A� shares. �B� share investors have ample investment opportunities, they are more risk averse and require higher rates of return. Similarly, in their research of the price behavior of the Chinese �A� and �B� share markets, Chen, Lee, and Rui, (2001) find that �A� shares exhibit premiums over �B� shares.

Second, underwriter�s strong risk aversion also contributes to the biggest underpricing. In the U.S., the underwriting procedure includes both best effort and firm commitment. In China, the underwriting procedure is only firm commitment, under which the underwriter guarantees the issuing company the issue price. When pricing a new issue, an underwriter would estimate earnings and growth rates of the company, compare the new issue with previous issues in the same industry if available, and then assign a somewhat arbitrary multiple to the estimated earnings. The underwriter also looks at the market during the issue, it would determine a bigger multiple in a bull market, a smaller multiple in a bear market. If the IPO firm wants to have a higher issue price (a greater multiple) than the underwriter has suggested, the underwriter would require a significant additional charge. Also, the low quality of accounting reports and the exaggerating window dressing of some firms, particularly private firms for IPO, creates large information asymmetry between the underwriters and the issuers, which would cause the underwriter to underprice the issue. Underpricing caused by information asymmetry is discussed earlier by Baron (1982). Furthermore,


25

underwriters may try to protect themselves against possible lawsuits by disgruntled investors with underpricing (Tinic, 1988). The underwriters in China have a stronger incentive to underprice the new issue because they face greater risks.

Third, institutional lag (number of days between the date of issue and the date of first trade) may result in underpricing if the stock market is rising between the fixing of the offering price and the beginning of trading (Kunz and Aggarwal, 1994). Among the IPOs issued before 1992, most have very long institutional lags that average over 1,000 days. The longest institutional lag is almost 10 years from the date of issue to the date of first trade (November 1988 � May 1998).

To test the above explanations, I regress the initial and first-day returns against the relevant variables. The regression model is specified as follows

Ri = α + β1NIT+ β2YSO + β3MV + β4INF + β5IL

Where, Ri = initial and first day rates of returns; NIT = natural logarithm of number of new issues started trading on the

Shanghai Stock Exchange in the year, which proxies for the supply of new shares to investors;

YSO = natural logarithm of number of years since the opening of the Shanghai Stock Exchange, which indicates investment bankers� and investors� experience;

MV = market volatility measured by the natural logarithm of the standard deviation of the daily value of the �A� share index, which approximates risk; INF = annual rate of inflation; and IL = natural logarithm of institutional lag in number of days. Results of the regression analysis are reported in Table 1. As shown in the

table, The short-term returns are significantly negatively connected to number of new shares on the SHSE. Greater number of new shares on the market represents greater supplies, which accommodates the excess demand for new shares hence results in relatively lower initial and firs-day returns, and vice versa.

The negative coefficient of the variable YSO (years since opening), though insignificant, indicates a negative connection between short-term returns and the number of years since the opening of the SHSE. This may imply that short-term returns decline as investment bankers and investors gain experience of IPOs during the early years of stock market in China. Over time, more investors learned to wait once they realize the extraordinarily sharp price jump in the first day and the poor long-term performance of IPOs

Gu, Anthony

26

Table 1. Estimated Impacts of Selected Independent Variables Dependent Intercept Number Years Market Rate of Institut'l Variables of Issues Opening Volatility Inflation Lag Initial Return 1.1457 -0.8711 0.3285 0.1687 4.8075 0.6222 R2: 0.11 -0.639 (-2.374)** (-0.735) (-0.564) (2.161)** (5.229)*** 1st Day Return 1.3247 -1.0009 0.3565 0.2074 5.1148 0.6185 R2: 0.13 -0.775 (-2.860)*** (-0.836) (-0.727) (2.410)** (5.450)***(T-values in parentheses) ** Significant at the 5% level *** Significant at the 1% level

The regression analysis shows some positive relation between short-term

returns and market volatility but the coefficient for the variable MV (market volatility) is not significant. This may imply that the level of market risk is not a major factor of underpricing for the data set.

The rate of inflation has a significantly positive impact on the short-term returns. The result shows that the demand for stocks is stronger during years of higher inflation, which indirectly supports the explanation that excess demand for new shares contributes to the high short-term returns.

The significant coefficients for the variable institutional lag provide evidence that institutional lag has a significantly positive impact on the short-term IPO returns. As indicated earlier, some of the IPOs took several years from date of issue to date of first trade. During the time period or institutional lag, the company would have been growing, and more information about the company would be available to investors. Thus the value of the stock should increase. As shown in Figure I, short-term returns displayed by issue year are generally much higher from 1984 through 1991. During the same time period, the institutional lags were much longer than that of the later issues.

A TREND TOWARD A TYPICAL LEVEL: As shown in Figures I and II, there is a declining trend in the short-term IPO returns until 1994 shown by issue year (Figure I) and until 1996 displayed by first trading year (Figure II). Sine then, the returns have been fluctuating around some typical levels, that is, 125 for average initial returns and 114 for average first day returns, respectively. Note that the returns are still well above 100 percent on average, and are still the highest among all markets.

Here I try to offer several explanations for the downward trend. First, the increasing supply of new shares accommodates the excess demand for new shares. There have been a growing number of companies issuing stocks along with economic reform or privatization in China. As of the end of 1995, five years after the opening of the SHSE, there were 188 companies listed on the SHSE. As of the end of September 2000, there were 541 companies listed on the SHSE. The


27

Figure II. Average Short-term Return by First Trading Year

50%

150%

250%

350%

450%

550%

650%

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

First Trading Year

Ret

urn

avreage initial return

average firs day retyrn

Figure I. Average Short-term Return by Issue Year (Raw)

0%

200%

400%

600%

800%

1000%

1200%

1400%

1600%

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

Year of Issue

Ret

urn

Average Initial Return

Average First Day Return

Gu, Anthony

28

second five-year period has 3 times as many IPOs as that of the first five-year period on the SHSE.

Also, as of the end of 1998, the middle year of the second five years, the total market capitalization of both SHSE and SZSE was 1,950.5 billion yuan, 24.52 percent of the GDP (China Financial Outlook, The People�s Bank of China, 1999, Shanghai Securities Daily). The market capitalization is over 5.52 times that of 1993, the middle year of the first five years, and the percentage of GDP is 2.4 times that of 1993. More IPOs each year represent greater supplies of new shares to investors, which should partially accommodate the excess demand for new shares, and thus should moderate the extraordinarily high short-term returns. As one may note, there is a big jump in the returns in 1995 exhibited by first trading year as shown in Figure II. A major reason may be that Chinese investors were even more eager to invest in inflation-protect stocks after they had experienced the highest explicit inflation since 1950, 22 percent in 1994 (China Financial Outlook, The People�s Bank of China, 1999), but there were only eight new issues in 1995. Correspondingly, there is also a hump for 1995 in Figure 1. After 1995, there have been over 50 IPOs each year, and the inflation rate has been declining (negative from 1998 to 2001).

Second, decrease in the length of institutional lags reduces the extent of price appreciation during the time period. In the previous section we report that a super long institutional lag significantly contributes to the super high short-term returns. The average length of institutional lags declined sharply after the opening of the SHSE, from 1,335 days in 1990 to 21 days in 1996, then increased to 53 days in 1999 and dropped to 8 days in 2000. The short to normal institutional lags would contain much less upward pressure on the initial and first day returns.

Third, investors� experienced investment behavior for new issues offers lower bid prices in the first trading day. Realizing the phenomenon that most of the IPOs exhibit extremely high initial and first day prices, which would decline afterwards, more experienced investors would fill their demand for new shares sometime after the first trading day. This experienced waiting would reduce the excess demand for new shares and thus result in lower initial and first day returns.

Fourth, lower risk to underwriters may help to reduce the extent of underpricing. During the second five years as compared to the first five years of the last decade, China�s political environment looks more stable, its international relationships are improving, and the Chinese economy experienced one of the highest growth in the world (over 7 percent per year even during the Asian crises). In the stock market, the volatility of stock returns reveals an apparent downward trend during the decade. The average standard deviation of monthly and daily returns on the SHSE �A� share index for the second five years are about half the average for the first five years. Meanwhile, the quality of accounting reports has been improved over time under the pressure of entering the World Trading Organization. Over time, underwriters have gained more experience and confidence, which would help them reduce excess underpricing.

Certainly, there are more funds available for investing in stocks as people�s incomes increase over time, which would further strengthen the already


29

strong demand for stocks. This may partially explain the still extraordinarily high and lasting short-term IPO returns in China.

CONCLUSION: The short-term returns on �A� share IPOs on SHSE exhibited a downward trend during the first half of the last decade and then have been fluctuating moderately around a typical level of about 120. Still, the �A� share IPOs in China show the highest short-term returns among all markets around the world. Pent-up demand for new shares, long institutional lag, and underwriter�s strong risk aversion may have contributed to the super high returns. Rapid growth in IPOs as an accommodation to the excess demand for new shares, decreased to normal institutional lags, lower risk in the stock market, and investors maturing investment behavior may partially explain the trend. We need to research on IPOs in other emerging markets that experienced extraordinarily high short-term IPO returns, such as Brazil, Korea, Malaysia, Taiwan and Thailand, and see if there is any trend with their maturing markets since early 1990s, because most of the reports used data before 1992. We also need to research for the reasons why short-term IPO returns are so much different among markets. For example, in emerging markets, the average initial return in China (398 percent) is so much higher than that of Malaysia (80.3 percent). And in developed markets, the average initial return in the U.S.A. (17.4 percent) is so much higher than that of France (9.5 percent). It is obviously incorrect to say that the Chinese investment bankers are less accurate than Malaysian and Brazilian ones; or the U.S. investment bankers are less accurate than the French ones in pricing IPOs. Finally, we need to find whether short-term IPO returns fluctuate around a typical level though the typical levels are so much different among markets. ACKNOWLEDGMENT The author wants to thank Qin Jing and Shi, Feng with Pudong Development Bank for providing most of the data. REFERENCES Aggarwal, R., R. Leal and L. Hernadez, (1993). The Aftermarket Performance of

Initial Public Offerings in Latin America, Financial Management, 22, 42-53.

Bailey, W., (1994). Risk and Return on China's New Stock Markets: Some Preliminary Evidence.Pacific-Basin Finance Journal, 2, 243-260.

Baron, D., (1982) A Model of the Demand of Investment Banking Advising and Distribution Services for New Issues. Journal of Finance 37, 955-976.

Gu, Anthony

30

Chau, C.T., C.S. Ciccotello and C.T. Grant, (1996) IPOs in Previously Centrally-Planned Economies: Chinese A-Share Evidence. Working Paper.

Checchi D., (1993). Creation of Financial Markets in (previously) Centrally � Planned Economies, Journal of Banking and Finance, 17,819-847. Chen, G. M., Bong-Soo Lee and Oliver Rui, (2001). Foreign ownership restrictions and market segmentation in China�s stock markets, Journal of Financial Research, 24, 133-155 Chen, Hsuan-chi and Jay R. Ritter, (2000). The Seven Percent Solution, Journal of Finance 55, 1105 - 1131. Chen, H.L., (1992). The Price Behavior of IPOs in Taiwan. Working Paper. Chen, Yea-Mow, (1996). The Performance of the Chinese IPO Market. Working

Paper, San Francisco State University, Series 95-02. Dhatt, M.S. Y.H. Kim, and U. Lim, (1993). The Short-run and Long-run

Performance of Korean IPOs: 1980-1990, Working Paper, University of Cincinnati and Yonsei University.

Gu, Yanxiang and Zhenhu Jin, (1998). The Determinants of IPO Performance: Evidence from Chinese �A� Share Market.� Proceedings of the American Society of Business and Behavioral Sciences 1998 Annual Meeting, 146 - 151. Gu, Yanxiang and Zhenhu Jin, (1999). The Short - Term Return on Chinese B Share IPO and Their Determinants. Midwest Review of Finance and Insurance, 13(1): 111 - 120. Ibbotson, R.G., J.L. Sindela and J.R. Ritter, (1988). Initial Public Offerings.

Journal of Applied Corporate Finance 7, 6-14. Isa, M. Md., (1993). List of Malaysian IPOs, Working Paper, University of Malaysia. Kahn, Charles, George Pennacchi, and Ben Sopranzetti, (1999). Bank deposit rate clustering: Theory and empirical evidence, Journal of Finance 54, 2185-2214. Kim, J., I. Krinsky, and J. Lee, (1993). Motives for Going Public and

Underpricing: New Findings from Korea. Journal of Business Finance and Accounting, 20, 195-211.

Kunz, R.M. and R. Aggarwal, (1994). Why Initial Public Offerings are Underpriced: Evidence from Switzerland. Journal of banking and Finance,18.

Loughran, T., J.R. Ritter and K. Rydqvist, (1994). Initial Public Offerings: International Insights. Pacific-Basin Finance Journal 2, 165-199 (data updated in September 2000 on Ritter�s website).

Miurin, P. and A. Sommariva, (1993). The Financial Reforms in Central and Eastern European Countries and in China. Journal of Banking and Finance, 17, 883-911.

Ritter, J.R., (1984). Signaling and the Value of Unseasoned New Issues: A Comment. Journal of Finance 39, 1231-1237.


31

Ritter, J.R., (1991). The Long Run Performance of Initial Public Offerings. Journal of Finance 46, 3-27.

Ri t te r , J .R . , (1997) , In i t ia l Publ ic Offer ings . Working Paper . Shanghai Securities Yearbook, various years, Shanghai Institute of Social Sciences and the Shanghai Stock Exchange. Tinic, M., (1988). Anatomy of Initial Public Offerings of Common Stock. Journal

of Finance, 43, Sep., 789-822. Wethyavivorn, K. and Y. Koo-Smith, (1991). Initial Public Offerings in

Thailand, 1988-1989: Price and Return Patterns, in S. G. Rhee and R.P. Chang, eds., Pacific-Basin Capital Markets Research, Vol. II (North- Holland, Amsterdam).

32

A CREDIT SCORING MODEL FOR SUBPRIME DEBT

Angeline M. Lavin University of South Dakota

ABSTRACT: This paper develops six credit scoring models that can be used by subprime lenders to make decisions about targeting potential borrowers and granting credit. To date, the topic of credit scoring for subprime debt is largely unexplored. The data used in this study were obtained from a subprime lender that purchases consumer installment loans, credit card debt, and auto loans that have been written-off by prime lenders. Three of the models use both account and credit bureau data, and three of the models use account data only. Characteristics that are consistently found to be significant predictors of borrower behavior in the subprime market include: participation in an automatic payment plan, credit score, length of time at residence, number of times to collection, number of previous charge-offs, and the existence of bank or retail tradelines on the borrower�s credit bureau report. During the development stage, the models are able to classify subprime accounts as �good� or �bad� risks with approximately 78% accuracy. When the six models are tested on the holdout sample, the classification accuracy decreases to approximately 45%. The average Type I error is 6.7% for the development sample but increases to 31.5% for the holdout sample. The holdout sample results indicate that the models which utilize account data only perform better than the models that use both account and credit bureau data Thus, it appears that incurring the extra costs associated with obtaining credit bureau data may not be warranted. INTRODUCTION AND PURPOSE: Credit scoring is a statistical method used to predict the probability that a borrower will default or become delinquent. According to Rusnak (1994) and Mester (1997), credit scoring models have three purposes: 1) application scoring, deciding who should be given credit; 2) behavior scoring, determining who is likely to default once credit has been granted; and 3) solicitation targeting, determining from whom to solicit applications. Credit scoring is predicated on the assumption that a borrower�s future behavior is expected to be similar to his/her past behavior. Since there are no absolute criteria that determine good or bad credit performance, credit scoring uses historical data to determine which characteristics are associated with good and bad credit performance. Future applicants are screened based on those characteristics. Credit scoring was first developed by Fair, Isaac, and Company for the consumer lending market nearly 50 years ago (Sullivan, 1994). Over the past 25 years credit scoring has become widely used in issuing credit cards, auto loans, and home equity loans (Mester, 1997). In 1994, Fair, Isaac, and Company began making inroads in the mortgage industry (Steinbach, 1998). While the traditional Fair Isaac credit scores are available to subprime lenders, these scores do not focus specifically on characteristics that might be important to determining the risk of a loan to a


33

borrower already categorized as �subprime.� The purpose of this paper is to focus on the development of a credit scoring model for subprime debt.

Despite the widespread use of statistical techniques in the consumer credit industry, the published literature on credit scoring is sparse. The lack of published work in this area can likely be attributed to the nature of the lending industry. First, superior credit scoring techniques provide lenders with a competitive advantage. Therefore, lenders are unlikely to divulge their methods. Second, confidentiality of applicant information must be maintained (Hand and Henley, 1997). Finally, the majority of all consumer credit scorecards are proprietary products developed by Fair Isaac or Experian and sold to lenders worldwide.

THE SUBPRIME LENDING MARKET: According to a March 1998 article in the Wall Street Journal, an increasingly broad spectrum of American households qualifies as subprime either due to their bad credit or over-indebtedness (Hube, 1998). While the ratio of debt payments to family income increased by 6% between 1992 and 1995, the percentage of debtors with payments more than 60 days past due increased by 17%. In the United States between 1993 and 1997, the number of bankruptcies increased by 43%. (U.S. Census Bureau, 1998). As a result of these trends, there is increasing demand for subprime lenders that purchase written-off debt at pennies on the dollar from finance companies, banks, and major credit card companies and offer the delinquent borrowers the opportunity to repair their tarnished credit histories. For example, they may offer the delinquent borrower an unsecured credit card with a line of credit equal to the amount of debt written-off by the original lender. As soon as the delinquent debtor makes his or her first payment on the new card, the subprime lender reports to the three major credit reporting agencies that the outstanding account has been satisfied. As the debtor pays off the balance on the new card, he or she gains the ability to charge to the card, up to the amount of the original line of credit. The subprime lenders that offer these services need a method for objectively evaluating subprime credit applications.

When a subprime lender purchases a portfolio of �written-off� debt from the original lender, the subprime lender receives information about the delinquent debtors including yearly income, phone number, length of employment, type of credit, length of time at residence, and date of last payment on the original account. The subprime lender then contacts the debtor to offer him/her the opportunity to rehabilitate his/her credit history. The purpose of this paper is to study the characteristics of good and bad accounts and to develop a credit scoring model that will assist subprime lenders in differentiating between those potential borrowers who are �good� credit risks and those who are �bad� credit risks. Specifically, the models developed in this paper will assist subprime lenders in three areas: 1) improving the rate at which customers who are �good� credit risks are converted to the rehabilitation program (open/active accounts); 2) decreasing the number of debtors who agree to the rehabilitation program, participate for awhile, and then switch back to a traditional credit card (unwind); and 3)

Lavin, Angeline

decreasing the number of debtors who agree to the program but never make a single payment (charge-off).1 These goals will be achieved by using a probit analysis to develop models that predict the probability that an account will fall into one of three categories: open/active, unwound, or charged-off.

DATA: Credit scoring is a statistical technique for estimating probabilities of default based on historical loan performance data and borrower characteristics. Credit scoring models are developed using a sample of applicants to whom credit has already been granted. The data should include the values of the characteristics of each account (credit bureau and/or credit application information) as well as the true class of each account (�good� or �bad� risk class). The data used in this analysis was obtained from the database of a subprime lender in the United States2 during the fall of 1997 and spring of 1998. This particular subprime lender purchases charged-off credit card, auto loan, and other installment loan accounts from prime lenders. According to Hand and Henley (1997), one problem encountered in the development of credit scoring models is that generally only those who were accepted for credit will be followed up to find out whether they turned out to be good or bad risks. This problem may cause the sample studied to be a biased sample from the overall population of applicants. The data used in this study avoid that problem to a certain extent; at the time these data were obtained, this subprime lender was not doing any credit scoring. Therefore, anyone who was in the portfolio of written-off credit debt purchased by the subprime lender was offered the opportunity to receive credit. The only people who were not in the database of converted accounts were those people who could not be contacted or those who refused the opportunity to rehabilitate their credit histories.

The primary development sample consisted of 3,000 randomly selected accounts from the universe of all of the subprime lender�s accounts in the fall of 1997. The original sample included 1,000 open/active accounts, 1,000 unwound accounts, and 1,000 charged-off accounts. In order to study the characteristics of �good� versus �bad� accounts, information was obtained from two sources: 1) account data provided by the lender who originally charged-off the delinquent account and 2) credit bureau reports. Account Data • Accrued interest charged-off (INTEREST) • Annual gross income (INC)

1 Although this model is developed using data obtained from a subprime lender that focuses on debt rehabilitation, subprime lenders that provide credit cards, auto loans, and other installment loans to borrowers with imperfect credit histories can also use it. In those cases, the model can assist with increasing the proportion of loans that are granted to borrowers who are �good� credit risks and minimizing the proportion of loans granted to borrowers who are more likely to become delinquent or default. 2 The subprime lender has requested anonymity.


35

• Automatic payment program participation (APP)3 • Credit score (CSCORE) • Date of last payment of original account (DLASTPAY) • Home: own/rent/other (RESIDENCE)4 • Length of time at residence (LRES) • Monthly payment to subprime lender (MONPAY) • Monthly payment to subprime lender divided by monthly income (MONRAT) • Number of times to collection (COLL) • Principal charged-off (PRIN) • Settlement amount (SETT) • State of account [geography] (STATE) • Total due charged-off (DUE) • Type of credit [credit card, consumer installment, or auto loan]5(TYPE) Credit Bureau Data • Number of times in bankruptcy (BANKRUPT) • Months since bankruptcy discharge (MDISCHG) • Number of charge-offs (CHGOFFS) • Number of tradelines (TRADES) • Types of tradelines [bank, retail, utility, etc.] (BANK, RETAIL, UTILITY)

After removing the accounts with incomplete account data, the development sample consisted of 1,914 accounts: 672 open/active accounts, 605 unwound accounts, and 637 charged-off accounts. The accounts in this sample came from 45 of the 50 United States and the District of Columbia.6 The annual incomes listed for the final sample of accounts ranged from a low of $0 to a high of $64,000. Credit bureau data was sought for the 1,914 accounts. After removing the accounts with incomplete credit bureau data, there were 740 accounts left in the sample: 273 open/active, 237 unwound accounts, and 230 charged-off accounts. The models were developed using the data set that includes only the account data (1,914 accounts) as well as the data set that includes both account and credit bureau data (740 accounts).

3 The APP variable was coded as a dummy variable where �1� denotes that the borrower signed up for the automatic payment plan and �2� denotes that he/she did not. 4 The RESIDENCE variable was coded as follows: 1=own, 2=rent, and 3=other. 5 The TYPE variable was coded as follows: 1=credit card, 2=consumer installment loan, and 3=auto loan. 6 The states of Alaska, Hawaii, North Dakota, Utah and Vermont were not represented in the sample.

Lavin, Angeline

36

METHOD: The purpose of this paper is to develop a set of models that will aid subprime lenders in differentiating between potential borrowers who are likely to be good credit risks (maintain open/active accounts) and those who are likely to be poor credit risks (open accounts that later unwind or are charged-off). A qualitative dependent variable model is an appropriate tool to use in a situation such as this where the intent is to consider the occurrence or non-occurrence of an event. In a qualitative dependent variable or binary choice model, y is a dichotomous random variable that takes on the value of 1 if the event occurs (the account unwinds or is charged-off) and 0 if it does not. The probability of the event occurring depends on a vector of independent variables (x) that measure the attributes of the alternatives and a vector of unknown parameters (β). The data set includes ni repetitions of the same choice situation (�good� versus �bad� accounts). According to Judge, Hill, Griffiths, Lutkepohl, and Lee (1988), if p� is

the proportion of times that alternative 1 was chosen in ni trials, then p� is the estimator of the true probability Pi

(1) ( ) iiiii exFePp +=+= β'� where the random error ei has mean zero and variance Pi(1-Pi)/ni. Equation (1) follows from the fact that p� is an unbiased and consistent estimator of Pi.

In order to constrain β'ix to lie between 0 and 1 as a probability should, a probit or logit model can be used as the functional form for F(.). The probit model uses a cumulative normal distribution that is symmetric around 0 and has variance equal to 1 while the logit model uses a logistic distribution that is also symmetric around 0 but has variance equal to π2/3. A probit model was chosen for this analysis.7

In the probit model, the cumulative distribution function is used to translate β'ix , which has the range ),( +∞−∞ , into a 0-1 range so that Prob

)()1( ' βiii xFPY === . More formally, the probability, Pi, of yi , the observed random variable, taking the value of 1 is

(2) dt2texp

21P

2x

i

'i

−

π= ∫

β

∞−

.

Because the probit model is nonlinear, it must be estimated using maximum likelihood estimation. Given the assumption of independent observations, the log-likelihood function for a sample of n observations is the following:

7 The cumulative normal and logistic distributions are very similar, and it is difficult to distinguish between the two unless the data set contains an extremely large number of observations. According to Amemiya (1981), in a binary choice model the choice between the logit and the probit models is irrelevant except when the data are heavily concentrated in the tails.


37

(3) [ ]∑∑==

β−−+β=βn

1i

'ii

n

1i

'ii )x(F1ln)y1()x(Flny)(L

where F(.) is the standard normal cumulative density function (Judge et al, 1988). The value of yi is assigned to be 0 if the account is open/active or 1 if the account unwound or was charged-off. The maximum likelihood estimate of the coefficients in the β vector is obtained by maximizing L with respect to β. This method also produces estimates of the asymptotic t-ratios that can be used to test the significance of each individual coefficient included in the probit model. After estimation of the model is complete, three summary measures can be used to assess the model's performance.8 The first hypothesis of interest is the null hypothesis that all of the coefficients jointly equal zero. This hypothesis is the analogue of the traditional F-test for the existence of a relationship within a linear regression model and is tested within a nonlinear model using the likelihood ratio test

(4) [ ] 2max* ~2 jLL χλ −−= . In Equation (4) L* is the maximum of the log likelihood [Equation (3)] when the restrictions of the null hypothesis have been imposed upon the β vector, and Lmax is the maximum of the unrestricted log likelihood. The likelihood ratio statistic has an asymptotic 2

1−kχ distribution. If the null hypothesis can not be rejected, then the group of independent variables has no influence on the probability of the event occurring. A pseudo- or quasi-R2 measure known as the McFadden R2 can be used to measure the percentage of the variation in the dependent variable explained by the characteristics included in the probit model. The McFadden R2 is computed as

(5) *

max2

LLR =

where L* and Lmax are defined as above. If Lmax=0, then R2=1, and the probit model [Equation (1)] fits the data perfectly in the sense that 1� =ip for the observations where 1=iy , and 0� =ip for the observations where 0=iy . Prediction success of the model is the basis for the third performance measure. A prediction for observation i is considered to be correct if the following holds:

(6) 5.0� ≥ip when 1=iy (7) 5.0� <ip when 0=iy .

Thus, a percentage of "correct predictions" out of the sample of n observations can be computed.

8 See Judge et al, 1998 for a more detailed explanation of these performance measures.

Lavin, Angeline

38

In this analysis, we will develop six binary choice models. Models 1 and 4 will predict whether an account is more likely to be a �good� risk, an open/active account, or a �bad� risk, an unwound or charged-off account. The Model 1 predictions will be based on both account and credit bureau data while the Model 4 predictions will be based on account data alone. Models 1 and 4 can be used to identify the accounts that are the most desirable credit risks. In addition to a general prediction of whether an account is more likely to be a �good� or a �bad� risk, it may also be informative to know whether a �bad� risk is more prone to unwinding or charging-off. Models 2 and 5 will predict whether an account is more likely to charge-off or remain open/active, and Models 3 and 6 will predict whether an account is more likely to unwind or remain open/active. Models 2 and 3 will use both account and credit bureau data while Models 5 and 6 will use only account data. The models that are developed in this study can be used to make decisions about whether credit should be granted to a potential borrower with a certain set of characteristics.9 DEVELOPMENT SAMPLE RESULTS: The models developed in this section of the paper focus on differentiating between accounts that are predicted to be "good" risks (open/active accounts), and accounts that are predicted to be "bad" risks (accounts that either unwind or are charged-off.) The models may serve a variety of purposes for subprime lenders, depending on the situation. First, the models can be used in the solicitation stage by subprime lenders who purchase portfolios of written-off debt from prime lenders or offer credit cards to individuals with imperfect credit histories. Subprime lenders can use the model to determine which accounts or applications are most likely to result in open/active accounts, and they can target those individuals first, offering them the option to rehabilitate their debt or obtain a credit card. Second, for subprime lenders that make auto or installment loans to borrowers with impaired credit histories, the model can be used for application scoring, or deciding when credit should be granted. Third, if credit has already been granted, the models can be used for behavior scoring, or predicting who is likely to default once credit has been granted. Accounts that are more likely to unwind or charge-off can then be watched or targeted in an attempt to prevent the "bad" outcome from occurring. As explained in the previous section, this paper uses a probit model to determine which combination of factors best predicts delinquency or default and how much weight should be given to each factor. Binary choice models, such as the probit model, are limited to differentiating between two outcomes. Therefore, the first model developed in this paper focuses on determining the characteristics that help to distinguish open/active accounts from accounts that either unwind or charge-off.

9 While there are methods available (conditional logit and probit analysis) to address the occurrence of multiple alternative events (Hausman and Wise, 1978), this paper will focus on the development of binary choice models because they are easier to estimate and implement in real world applications.


39

Model 1: Differentiating between open/active and unwound/charged-off accounts using account and bureau data Initially, all characteristics listed in the Data section under Account Data and Credit Bureau Data were included in the regression model. Open/active accounts were designated with a �0� while unwound or charged-off accounts were designated with a �1�. The development sample for Model 1 includes 740 accounts: 273 (36.89%) open/active accounts and 467 (63.11%) accounts that either unwound or were charged-off. The asymptotic t-ratios computed during the initial maximum likelihood estimation of the probit model were used to eliminate some characteristics that were not significant by themselves.10 After eliminating those variables, the regression model was re-estimated and produced the following sets of coefficients and t-ratios. The set of coefficients presented for each model is the set of coefficients that maximized the prediction success rate for that model using the development sample.

When interpreting the coefficients for a probit model, the sign of the coefficient describes the direction of the impact on the probability of the outcome�s occurrence. In this research, a positive coefficient means that the characteristic increases the probability of the bad outcome. However, the magnitude of the coefficient cannot be interpreted directly as it is in a traditional regression model because the distribution used for the probit model is a cumulative normal density function.

Consistent with expectations, the greater the number of times that a borrower has previously been referred to collections (COLL) or had an account charged-off (CHGOFFS), the higher the probability of the account unwinding or charging-off. The more bank (BANK) or retail (RETAIL) tradelines that a consumer has, the lower the probability that his or her account will unwind or charge-off, possibly because consumers must have relatively good credit histories to obtain bank and retail tradelines. Also consistent with expectations, borrowers who do not sign-up for the automatic payment plan (APP=2) have a higher probability of unwinding or charging-off. The positive sign associated with the credit score (CSCORE) and length of time in residence (LRES) characteristics makes sense given that this is a subprime debt sample. Individuals with higher credit scores or who have lived in their residence for a longer period of time are more likely to obtain credit from �prime� lenders at better interest rates. Thus, accounts with those characteristics are more likely to unwind or charge-off.

The significance of the likelihood ratio test indicates that the null hypothesis (all of the coefficients jointly equal zero) can be rejected. The R2 measure indicates that 37% of the variability in the probability of an account unwinding or charging-off can be explained by the set of included independent variables. The model predicted 22.84% of the accounts would be open/active (0) while the remaining 77.16% would unwind or charge-off (1).

10 Some insignificant characteristics were retained in the model because they increased the model�s prediction success rate.

Lavin, Angeline

40

Table 1: Model 1 Results COEFFICIENTS AND T-RATIOS

Variable Coefficient T-Ratio COLL 2.5836 2.4751* INC 2.0675 x 10-6 3.325 x 10-1

LRES 1.1340 x 10-2 1.9405* APP 2.7706 7.4277*

MONRAT 3.5877 1.1738 CSCORE 7.6204 x 10-2 4.3195*

DLASTPAY 2.7558 x 10-2 9.353 x 10-1 CHGOFFS 1.0426 x 10-1 3.6358*

BANK -8.98 x 10-2 -3.8484* RETAIL -1.4891 x 10-1 -4.6682*

BANKRUPT 3.3538 x 10-1 8.843 x 10-1 MDISCHG -1.7329 x 10-2 -1.5103

CONSTANT -60.082 -1.0245 Summary Measures of Performance

Likelihood Ratio Test 361.64** McFadden R2 0.37

Prediction Success Rate 79.73% Prediction Success

Actual 0 1 Total

0 0.1973 0.0311 0.2284 1 0.1716 0.6000 0.7716

Predicted Total 0.3689 0.6311 1.0000

** Denotes significance at 1% level. * Denotes significance at the 5% level. The model correctly classified 19.73% of the open/active accounts as "good"

accounts and incorrectly classified 17.16% of the open/active accounts as �bad� accounts. It appears that the model was better at accurately identifying �bad� accounts, as it correctly identified 60% of the accounts as unwound or charged-off accounts and incorrectly classified only 3.11% of the �bad� accounts as open/active accounts. This model is quite conservative, erring on the side of predicting that an account is �bad� rather than �good.� According to the prediction success table, the potential for a Type I error (pursuing an account that is likely to unwind/charge-off or granting credit to a borrower who is likely to become delinquent) is small (3.11%). The results of Model 1 are useful for the initial stages of evaluation when decisions are made with respect to targeting accounts and granting credit. After credit has been granted, it may be useful to classify accounts separately according to the probability that they will charge-off or unwind. In particular, accounts that are likely to unwind may be targeted with lower interest rates in order to encourage the borrower to maintain the account. The data used to estimate Model 2 consists of account and credit bureau data on 503 accounts: 273 open/active accounts and 230 charged-off accounts; the data for Model 3 includes 510 accounts: 273 open/active and 237 unwound accounts.


41

Model 2: Differentiating between open/active and charged-off accounts using account and bureau data

The Model 2 results are consistent with those from Model 1 except that number of times in collection (COLL) and participation in the automatic payment plan (APP) are not significant predictors of account charge-off. Again, the null that the coefficients jointly equal zero can be rejected, and the R2 measure indicates that the included independent variables explain 46% of the variation in the probability that an account will charge-off. The prediction success rate for Model 2 is slightly better than for Model 1. Like Model 1, Model 2 is conservative; incorrectly predicting that only 10.93% of the accounts would be open/active when they were actually charged-off accounts.


Variable Coefficient T-Ratio COLL 5.7085 4.7924 x 10-2 LRES 1.3725 x 10-2 1.7596* APP 6.2210 2.4491 x 10-2

MONRAT 3.3217 1.1085 CSCORE 8.4927 x 10-2 4.1748*

CHGOFFS 1.1444 x 10-1 3.0384* BANK -8.1296 x 10-2 -2.6883*

RETAIL -2.05 x 10-1 -4.2780* BANKRUPT 5.2354 x 10-1 1.2166 MDISCHG -1.1744 x 10-2 -9.6636 x 10-1

CONSTANT -12.694 -2.4988 x 10-2 Summary Measures of Performance

Likelihood Ratio Test 320.17 ** McFadden R2 0.46


Actual 0 1 Total

0 0.4553 0.1093 0.5646 1 0.0875 0.3479 0.4354

Predicted Total 0.5427 0.4573 1.0000

** Denotes significance at 1% level. * Denotes significance at the 5% level. Model 3: Differentiating between open/active and unwound accounts using account and bureau data

The Model 3 results are also consistent with those from Model 1 with respect to significant variables except that number of times in collection (COLL) and length of time at residence (LRES) are not significant predictors of account unwind. Again, the null hypothesis that the coefficients jointly equal zero can be rejected. The R2 measure for Model 3 is much lower than the R2 for Model 1 or Model 2. The prediction success rate for Model 3 is only slightly lower than the success rate for Models 1 and 2. Like Models 1 and 2, Model 3 is conservative, incorrectly predicting that only 7.25% of the accounts would be open/active when they were actually unwound accounts.

Lavin, Angeline

42

In summary, the results of Models 1, 2 and 3 indicate that when both account and credit bureau data are used, the credit store (CSCORE), the number of charge-offs (CHGOFFS), the number of bank tradelines (BANK), and the number of retail tradelines (RETAIL) are all significant predictors of �good� versus �bad� accounts. Number of times to collection (COLL) is significant in Model 1 only, while length of time at residence (LRES) is significant for both Models 1 and 2, and participating in an automatic payment plan (APP) is significant for both Models 1 and 3.


Variable Coefficient T-Ratio COLL 2.6492 1.2888 LRES 1.0119 x 10-2 1.5879 INC -3.9679 x 10-7 -5.3354 x 10-2 APP 2.4754 6.6412*

MONRAT 3.5431 8.8084 x 10-1 CSCORE 6.9111 x 10-2 3.3231*

CHGOFFS 1.1502 x 10-1 3.5952* BANK -9.8744 x 10-2 -3.6381*

RETAIL -1.3334 x 10-1 -3.8680* DLASTPAY 1.6633 x 10-2 4.9069 x 10-1 MDISCHG -3.1715 x 10-2 -1.8546

CONSTANT -37.969 -5.6282 x 10-1 Summary Measures of Performance

Likelihood Ratio Test 206.50** McFadden R2 0.29


Actual 0 1 Total

0 0.3726 0.0725 0.4451 1 0.1627 0.3922 0.5549

Predicted Total 0.5353 0.4647 1.0000

** Denotes significance at 1% level. * Denotes significance at the 5% level. Models 4-6 differ from the first three models in that the latter three use only

account data and, therefore, do not necessitate the purchase of credit bureau data. The performance of Models 4-6 will be compared to Models 1- 3 to determine whether excluding the credit bureau data decreases the prediction success rate of the models. The development sample for Model 4 includes 1,914 accounts: 672 open/active accounts, 605 unwound accounts, and 637 charged-off accounts.

Model 4: Differentiating between open/active and unwound/charged-off accounts using account data only

Comparing Model 4 to Model 1 indicates that the length of time in residence (LRES), participation in an automatic payment plan (APP), and the credit score


43

(CSCORE) are significant predictors of a account status no matter which data set is used. In Model 4, the ratio of the monthly payment for the charged-off debt to the monthly income (MONRAT) is also positive and significant. In other words, the higher the ratio of debt payment to income, the more likely the account is to charge-off or unwind. The null hypothesis that the coefficients jointly equal zero can be rejected for Model 4. The prediction success rate for Model 4 is exactly the same as the prediction success rate for Model 1, but the R2 measure for Model 4 is slightly lower. The Type I error rate for Model 4 is especially impressive at only 0.21%. Model 4 incorrectly predicted that only 4 accounts out of a sample of 1,914 would be open/active when they actually charged-off or unwound. According to these results, excluding credit bureau data does not impair model performance and actually improves the Type I error rate.


Variable Coefficient T-Ratio COLL 4.2579 0.4149 LRES 9.5228 x 10-2 2.6932* APP 2.7722 13.646*

MONRAT 2.5152 2.19* CSCORE 5.2574 x 10-1 5.6101*

DLASTPAY -5.0296 x 10-2 -0.35207 CONSTANT 4.6838 0.16468

Summary Measures of Performance Likelihood Ratio Test 901.83**

McFadden R2 0.36 Prediction Success Rate 79.73%

Prediction Success Actual

0 1 Total 0 0.1505 0.0021 0.1526 1 0.2006 0.6468 0.8474

Predicted Total 0.3511 0.6489 1.0000

** Denotes significance at 1% level. * Denotes significance at the 5% level. The results of Model 4, like the results for Model 1, are useful for the

initial stages of account evaluation, when decisions are being made with respect to which accounts should be targeted or which borrowers should be granted credit. However, after credit has been granted, it may be useful to classify accounts separately according to the probability that they will charge-off or unwind. Knowing that an account is likely to charge-off or unwind offers the lender the chance to intervene before the potential �bad� outcome occurs. The data used to estimate Model 5 consists of account data on 1,309 accounts: 672 open/active accounts and 637 charged-off accounts; the data for Model 6 includes 1,277 accounts: 672 open/active and 605 unwound accounts.

Lavin, Angeline

44

Model 5: Differentiating between open/active and charged-off accounts using account data only

The credit score (CSCORE) and length of time at residence (LRES) are significant predictors of account charge-off no matter which data set is used. However, when the credit bureau data is eliminated (Model 5), participating in the automatic payment plan (APP), number of times in collection (COLL), and the ratio of monthly debt payment to monthly income (MONRAT) all become significant predictors of charge-off. The null hypothesis that the coefficients jointly equal zero can be rejected for Model 5. Both the R2 measure and the prediction success rate for Model 5 are lower than the corresponding numbers for Model 2. The probability of a Type I error increases slightly from 10.98% in Model 2 to 14.13% in Model 5. These results suggest that excluding the credit bureau data causes a slight decline in model performance.


Variable Coefficient T-Ratio COLL 3.4356 3.9979* LRES 1.1266 x 10-2 2.5526* APP 2.5125 9.3857*

MONRAT 2.7407 2.0656* CSCORE 6.5627 x 10-2 5.9773*

TYPE 3.1764 x 10-2 2.8065 x 10-1 DLASTPAY 2.6094 x 10-2 1.4874 CONSTANT -57.432 -1.6419




0 1 Total 0 0.4370 0.1413 0.5783 1 0.0764 0.3453 0.4217

Predicted Total 0.5134 .4866 1.0000

** Denotes significance at 1% level. * Denotes significance at the 5% level. Model 6: Differentiating between open/active and unwound accounts using account data only

The credit score (CSCORE) and participating in the automatic payment plan (APP) are significant predictors of account unwind no matter which data set is used. However, when the credit bureau data is eliminated (Model 6), number of times in collection (COLL), length of time at residence (LRES), annual income (INC), and the type of account [credit card, consumer installment or auto loan] (TYPE) become significant predictors of an account�s propensity to unwind. The positive relationship between annual income and the probability of account


45

unwind makes sense for the subprime market. Those borrowers with higher levels of income may have more opportunities to borrow from prime lenders. Thus, borrowers with higher incomes do not follow through with the debt rehabilitation program once they have agreed to it. The positive relationship between the �type� variable and the probability of an account unwinding suggests that auto loans (TYPE=3) are more likely to unwind than credit card debt (TYPE=1). The null hypothesis that the coefficients jointly equal zero can be rejected for Model 6. The R2 measure is the same for both Models 3 and 6, but the prediction success rate is lower for Model 6. The probability of a Type I error is actually lower (4.62% for Model 6 vs. 7.25% for Model 3) when the credit bureau data is excluded.


Variable Coefficient T-Ratio COLL 2.8929 3.1075* LRES 9.3803 x 10-3 2.3361* APP 2.6364 10.293*

MONRAT 1.1977 7.6841 x 10-1 CSCORE 4.3286 x 10-2 4.1777*

TYPE 2.4035 x 10-1 2.1030* INC -8.0026 x 10-6 -1.8318*

DLASTPAY -1.5075 x 10-2 -8.6384 x 10-1 CONSTANT 24.436 7.0226 x 10-1




0 1 Total 0 0.2968 0.0462 0.3430 1 0.2294 0.4276 0.6570

Predicted Total 0.5262 0.4738 1.0000

** Denotes significance at 1% level. * Denotes significance at the 5% level.

In summary, the results of Models 4, 5, and 6 suggest that when account data are used alone, the credit store (CSCORE), length of time in residence (LRES), and participating in an automatic payment plan (APP) are all significant predictors of �good� versus �bad� accounts. The ratio of monthly payment to monthly income (MONRAT) is significant in Models 4 and 5, while the number of times a borrower has previously been referred to collections (COLL) is significant in Models 5 and 6. The type of [credit card, consumer installment or auto loan] (TYPE) and annual income of the borrower (INC) are significant in Model 6 only.

Lavin, Angeline

46

HOLDOUT SAMPLE RESULTS: The summary measures of performance discussed in the previous section provide one measure of model performance. The second aspect of performance is how the model and its decision criterion work with accounts that were not included in the original development sample. According to Altman and Haldeman (1995), the errors reported for the holdout sample are more representative of a model�s expected performance than the error rates of the development sample. The holdout sample used in this study was obtained in the spring of 1998 and includes both account and credit bureau data. The sample consists of 728 accounts: 292 open/active accounts, 200 unwound accounts, and 236 charged-off accounts.

Table 7 compares each model�s prediction success rate for the development sample to the prediction success rate for the holdout sample.

Table 7: Prediction Success Rates: Development Sample vs. Holdout Sample Development Sample Holdout Sample

Model 1 79.73% 39.84% Model 2 80.32% 46.21% Model 3 76.47% 42.68% Model 4 79.73% 52.06% Model 5 78.23% 45.83% Model 6 72.44% 45.53% The average prediction success rates for the development and holdout

samples are 77.82% and 45.36%, respectively. The lower rate for the holdout sample may be explained by the limited development sample that was available. Many credit scoring development databases contain five or more years of data on 100,000 or more accounts.

It is interesting to note that on average, the models that were developed using only account data [Models 4-6] perform better on the holdout sample (47.81% average prediction success) than the models that included both account and credit bureau data [Models 1-3] (42.91% average prediction success). For the holdout sample, the prediction success rate for Model 4 was over 12% better than the rate for Model 1. The Type I error rate is an important gauge of model performance because it explains the probability of the model predicting that an account is open/active when it actually charges-off or unwinds. Table 8 compares the Type I error rates for the development and holdout samples. The average Type I error rate for the development sample is 6.71% while the average Type I error rate for the holdout sample is 31.5%. In the holdout sample, the average Type I error rate was much lower for the models that used only account data [Models 4-6] (25.76%) when compared to the models that used both account and credit bureau data [Models 1-3] (37.24%). This information, combined with the prediction success rates in Table 7, suggests that incurring the extra costs associated with obtaining credit bureau data may not be warranted.


47

Table 8: Type I Error Rates: Development Sample vs. Holdout Sample Development Sample Holdout Sample

Model 1 3.11% 42.58% Model 2 10.93% 35.42% Model 3 7.25% 33.74% Model 4 0.21% 23.76% Model 5 14.13% 29.74% Model 6 4.62% 23.78% Table 9 reveals that Models 1-3 predict substantially more open/active

accounts for the holdout sample than there are in the actual sample. For example, Model 1 predicted that 65.11% of the accounts were open/active when only 40.11% of the accounts were actually open/active. Models 4-6 appear to do a better job distinguishing between �good� and �bad� accounts. �Training� the model on a larger development sample may improve model performance. The lack of readily available data may be one reason that there has been very little published on credit scoring for subprime debt.

The potential for improvement in model performance is also supported by the relatively low R2 measures reported for Models 1-6 using the development sample. The R2 measures varied from a low of 0.29 for Models 3 and 6 to a high of 0.46 for Model 2. These low R2 statistics indicate that there is a significant amount of variability in the dependent variable that is not accounted for by the independent variables included in each model. Furthermore, the existence of large constant terms in all of the models except Model 4 suggests that there may be important predictive variables that are not included in the current data set and models.

Table 9: Prediction Success Table for the Holdout Sample for Models 1– 3 MODEL 1

Actual 0 1 TOTAL

0 0.2253 0.4258 0.6511 1 0.1758 0.1731 0.3489

Predicted

Total 0.4011 0.5989 1.0000 MODEL 2

ACTUAL 0 1 Total

0 0.3693 0.3542 0.7235 1 0.1837 0.0928 0.2765

Predicted

Total 0.5530 0.4470 1.0000 MODEL 3

Actual 0 1 Total

0 0.3577 0.3374 0.6951 1 0.2358 0.0691 0.3049

Predicted Total 0.5935 0.4065 1.0000

Lavin, Angeline

48

Table 9: Prediction Success Table for the Holdout Sample for Models 4-6 MODEL 4

ACTUAL 0 1 Total

0 0.1593 0.2376 0.3969 1 0.2418 0.3613 0.6031

Predicted Total 0.4011 0.5989 1.0000

MODEL 5 ACTUAL

0 1 Total 0 0.3087 0.2974 0.6061 1 0.2443 0.1496 0.3939

Predicted Total 0.5530 0.4470 1.0000

MODEL 6 Actual

0 1 Total 0 0.2866 0.2378 0.5244 1 0.3069 0.1687 0.4756

Predicted Total 0.5935 0.4065 1.0000

CONCLUSION AND APPLICATIONS: Six models are developed in this paper to investigate the effects of various account and credit bureau characteristics on the propensity of a subprime account to unwind and/or charge-off. Models 1, 2 and 3 use both account and credit bureau data while Models 4, 5 and 6 focus on account data only, eliminating the need to purchase credit bureau data. Models 1 and 4 are general models for predicting which potential borrowers are most likely to evolve into open/active accounts. Models 2, 3, 5 and 6 classify the outcomes into separate groups. Models 2 and 5 measure the likelihood that an account is more likely to charge-off rather than remain open/active while Models 3 and 6 measure whether an account is more likely to unwind or remain open/active.

During the development stage, the models are able to classify subprime accounts as �good� or �bad� risks with 78% accuracy. Participation in an automatic payment plan (APP), credit score (CSCORE), length of time at residence (LRES), and number of times to collection (COLL) are consistently found to be significant predictors of an account unwinding or charging-off include. Participation in an automatic payment plan (APP) is particularly important for predicting whether an account will unwind or remain open/active. The credit bureau data characteristics that are consistently found to be significant predictors of whether an account will unwind or charge-off include the number of previous charge-offs (CHGOFFS) and the existence of bank (BANK) or retail (RETAIL) tradelines on the borrower�s credit bureau report.

When the six models are tested on the holdout sample, the classification accuracy decreases to 45%. Furthermore, the average Type I error rate increases from 6.7% (development sample) to 31.5% (holdout sample). In the holdout sample, the models developed using only account data perform better (47.81% prediction success) than the models that included both account data and credit


49

bureau data (42.91% prediction success). The average Type I error rate was much lower for the models that used only the account data (25.76%) in the holdout sample when compared to the models that used both account and credit bureau data (37.24%). These results suggest that incurring the additional costs of obtaining credit bureau data may not be warranted.

The models developed in this paper can be used for a variety of purposes. Model 4 can be used in the solicitation stage by subprime lenders who purchase portfolios of written-off debt from prime lenders or offer credit to individuals with imperfect credit histories. Subprime lenders can use the model to determine which potential borrowers are most likely to establish open/active accounts so they can target those individuals first, offering them debt rehabilitation or credit card services. For subprime lenders who make auto or installment loans to borrowers with impaired credit, Model 4 can be used to decide when credit should be granted. Third, if credit has already been granted, Models 5 and 6 can be used for predicting which accounts are likely to unwind or charge-off. Accounts that are more likely to unwind or charge-off can be watched or targeted in an attempt to prevent the "bad" outcome from occurring.

REFERENCES: Altman, E.I. and R.B. Haldeman (1995). Corporate Credit-Scoring Models:

Approaches and Tests for Successful Implementation. The Journal of Commercial Lending, 77,10-22.

Amemiya T. (1981). Qualitative Response Models: A Survey. Journal of Economic Literature, 19, 1483-1536. Hand, D.J. and W.E. Henley (1997). Statistical Classification Methods in

Consumer Credit Scoring: A Review. Journal of the Royal Statistical Society, 160, 522-541. Hausman, J.A. and D.A. Wise (1978). A Conditional Probit Model for Qualitative Choice: Discrete Decisions Recognizing Interdependence and Heterogeneous Preferences. Econometrica, 46, 403-426. Hube, K. (1998, March 18). In the Wild West of Subprime Lending, Borrowers

Have to Dodge Many Bullets. The Wall Street Journal. Judge, G.G., R.C. Hill, W.E. Griffiths, H. Lutkepohl, and T. Lee (1988).

Introduction to the Theory and Practice of Econometrics. (2nd ed.). New York: Wiley & Sons.

Mester, L.J. (1997). What�s the Point of Credit Scoring? Federal Reserve Bank of Philadelphia Business Review, 3-16.

Steinbach, G.H. (1998). Making Risk-Based Pricing Work. Mortgage Banking, 58, 10-22.

Rusnak, R.L. (1994). Consumer Credit-Scoring: Are There Lessons for Commercial Lenders? The Journal of Commercial Lending, 76, 37-42.

U.S. Census Bureau. (1998). The Official Statistics Statistical Abstract of the United States: 1998. Washington, DC: U.S. Government Printing Office.

50

ACCOUNTANCY: A PROFESSION EVOLVING FROM AN ART

Kel-Ann S. Eyler

Brenau University Teresa T. King

Wesleyan College Elliott L. Slocum

Georgia State University ABSTRACT: Since the establishment and evolution of accountancy in the United States during the last century, the profession has been criticized for its apparent unwillingness to establish enforceable generally accepted accounting standards as well as to provide general universal principles upon which to base the standards. Accountancy's role in society was developing and being defined as a profession evolving from an art and achieving some qualities of a science. This paper focuses on some of the social, economic, political, and professional criticisms occurring in this period, and the profession's responses to those criticisms. INTRODUCTION: The significant changes in business and economic conditions, social expectations, and political agendas experienced in the United States, especially following World War II, led to severe criticism of accounting practices use in financial reporting and threatened accountancy�s role as a profession. The turmoil was such that some recommended (and many feared) that government would take over the role of establishing accounting standards for financial reporting, which would have reduced accountancy to little more than a skilled occupation. As a result, changes were initiated in the service image of accountancy and the source of authority for accounting standards.

This paper is primarily concerned with the change in the source of authority for accounting principles which support the body of knowledge that is critical to accountancy's professional status. Factors that support accountancy as a profession are briefly presented. A review is provided of key decisions and actions taken by accountancy during the 1920s and 1930s regarding establishing a body of knowledge and related accounting principles. Accountancy�s relationship to stockholders and other users of financial reports is also reviewed related to the profession's inability to deal with the increasingly complex business environment and the social, economic, and political demands that followed World War II. Events from 1950 through 1970 leading to the profession�s evolving from the status of an art are briefly chronicled and a number of issues are discussed. ACCOUNTANCY AS A PROFESSION: Accountancy is used to denote the profession, and traditionally public accounting is the area of accounting that


51

possesses the combination of characteristics associated with professions. A profession (Greenwood, 1978) is an organized group which constantly interacts with society, performs social functions through a network of formal and informal relationships, and creates a subculture to which adherence is required for career success. Professions possess a systematic theory, authority, community sanction, ethical codes, and a culture. It is the degree to which an organized group possesses these attributes that places the group along a continuum from a skilled occupation to a well-recognized profession. Accountancy is perhaps best viewed as a relatively young profession in the United States. Certainly accountancy as reflected by public accounting is more than a skilled occupation.

A profession must have a set of values which promote some idea of service to society (Buckley, 1978) and must be concerned about the service-ideal image. Reduction of this image will lead to less autonomy and therefore to less authority. Public accounting has (Montagna, 1974) variously been described as serving society, serving the public interest, and serving or protecting investors. The character of the service philosophy influences how accountancy serves diverse interests. Van Riper (1994) observes that a fundamental philosophical debate currently exists regarding what information the public needs and how it should be reported. Should accountancy be primarily concerned with the reliability and objectivity of reported information? If so, it would exercise neutrality and self-discipline in measuring and reporting economic activity objectively such that information can be used with confidence for economic decisions. Alternatively, should accountancy focus on the possible economic and social consequences of information? In this case, accountancy would report information based on the concern for the economic health of enterprises, industries, or even society at large.

Greenwood (1978) states that the skills that distinguish a profession are derived from and supported by a body of knowledge that has been �organized into an internally consistent system, called a body of theory� (pp. 51-52). The body of theory represents abstract propositions that describe the general nature of the phenomena in which the professional is interested and serves as a basis on which to rationalize operations into concrete situations. Professional skills must be supported by the underlying theories of the skills. Apprenticeship becomes inadequate as a means to achieve the body of theory and must yield to formalized education. Greenwood (1978) states that "application of the scientific method to the service-related problems of the profession" (p. 52) is required to generate a valid theory that provides a solid foundation for professional techniques. Continued use of the scientific method is needed to reinforce rationality, which in turn nurtures the scientific method. Rationality encourages critical attitude toward the theoretical system as opposed to a reverential attitude. Thus, rationality leads to a willingness to change the body of theory whenever a more valid approach is identified regardless of tradition or past experience. THE PROFESSION PRIOR TO 1950: Early in the Twentieth Century, at the end the progressive era (Previts and Merino, 1998) many in society and

Eyler, King, and Slocum

governments were critical of business practices and believed that better communication between business and society through uniform financial reporting would help to reduce inefficient and harmful practices. No authoritative guidelines existed for financial reporting to stockholders and creditors. Financial reports and accounting practices largely reflected the personal experiences and training, mostly through proprietary schools and apprenticeship, of individual accountants. Leaders of accountancy recognized a need for sound accounting practices and improved financial reporting. Since accountancy did not possess legal authority to establish or to enforce sound accounting practice, it needed to cooperate with government agencies or groups who had such authority. The American Institute of Accountants (Institute, AIA) first cooperated with the Federal Trade Commission leading to the publication of Uniform Accounting in 1917 (Previts and Merino, 1998; Federal Reserve Board, 1917). The publication provided some helpful guidelines for form and content of financial statements.

By 1930, however, little had been accomplished that provided sound accounting practices in financial reporting and support of a body of knowledge. At this time, the Institute worked with the New York Stock Exchange on accounting matters to achieve greater uniformity of accounting principles and scope of audits (Flegm, 1984). The cooperative effort successfully led to the publication of Audits of Corporate Accounts (1934) and to the establishment of five broad accounting principles.

The Securities and Exchange Commission (SEC), as a regulatory agency, depended on the development and general acceptance of a body of principles (Sanders, 1936). The success of the work with the New York Stock Exchange probably influenced the SEC (Carey, 1970) to allow the accounting profession to establish accounting principles and practices fundamental to financial reporting. However, by 1938, little had been accomplished to formulate a cohesive body of principles, and the profession failed to accept its responsibility to investors and reduce the extensive variety of accounting methods then in use (Flegm, 1984).

The academically oriented American Accounting Association (AAA) criticized practicing accountants for their failure to establish an authoritative statement of essential principles and assume full responsibility for preparation of sound and informative financial reports ("A Statement of Objectives . . .," 1936). The AAA's publication of A Tentative Statement of Accounting Principles Affecting Corporate Reports (1936) was viewed by practitioners as an attempt to establish a uniform code of accounting principles (Carey, 1970). The Institute responded by publishing A Statement of Accounting Principles, (Sanders, Hatfield and Moore, 1938) which codified existing accounting practices, and reaffirmed that management was responsible for deciding what information should be reported in financial statements and how it should be shown (Carey, 1970).

In 1938, the Institute�s committee on accounting procedure (CAP) (AIA, 1938 Yearbook), due to widespread demand for greater uniformity in accounting, was enlarged and a research division established to prepare studies of particular questions for the purpose of issuing pronouncements on specific procedures and practices. The CAP�s (AICPA, 1961) principal objectives were to reduce


53

differences and inconsistencies in accounting practices, to further develop and recognize generally accepted accounting principles, and to issue Accounting Research Bulletins as criteria for accountants choosing accounting practices used in financial statements. Authority for opinions issued by the CAP rested on general acceptability. Thus, the notion of generally accepted accounting principles (Van Riper, 1994) was established which focused, with the SEC's acquiescence, on general guidelines rather than enforceable rules.

As a volunteer organization, CAP�s progress was erratic, and many of its conclusions were not supported by research. The SEC and the Institute cooperated in following a "common-law" approach and ruled on accounting questions as they arose. The CAP was criticized for lack of industry representation and experimentation with new ideas (Jennings, 1958). The same problems were continually reconsidered, and other problems were unsolved. In a process expressly dependent on consensus, setting of principles was hampered by the lack of consensus needed to approve controversial pronouncements. Corporate management frequently objected to proposed changes because they did not want to give up their right to choose the accounting methodology used to prepare financial statements. The CAP could not ignore corporate influence with the SEC, stock exchanges, and the commercial press. Additionally, auditors were at liberty to deviate, except in cases of compliance with the SEC and NYSE requirements, if they accepted the burden of justification for departure from CAP pronouncements (Carey, 1970).

Accounting Research Bulletins (ARBs) issued from 1938 to 1959 helped to eliminate many abuses and to narrow alternative methodologies. Pronouncements involving form, disclosure, and levels of detail were successful, but those regarding asset or liability valuation and related revenue and expense recognition failed. The process was more political than theoretical (Previts and Merino, 1998). No comprehensive statement of accounting principles was prepared on which accountants and users of financial statements could rely. The piecemeal approach (Storey, June, 1964) made a significant contribution in the past; however, further contribution was limited because of the inability to "lay a general conceptual foundation for accounting practice" (p. 52). CRISIS AND TURMOIL: Shortly after the end of World War II, accountancy was faced with an increasingly more complex social and business environment. Aggressive governmental agencies and public concern about the quality of financial reporting brought heavy pressure on the profession from regulators, academics, financial analysts, financial press, and from the profession itself to achieve greater uniformity or comparability in financial reporting. The SEC, Federal Trade Commission, and other agencies pursued policies to protect investors based on the theme of uniformity of accounting practices. Problems of fraud and continued availability of alternative methodologies plagued the profession. The myth of precision and uniformity in accounting and financial reporting was reinforced and exaggerated through wide use and misunderstanding of the meaning of "generally accepted accounting principles" (Flegm, 1984).


54

Various ways (Byrne, 1937) of preparing financial statements, classification of data, and methods of disclosure became the subject of many rules and regulations that were "frequently dignified improperly with the title of 'accounting principles'" (p. 375). Accountancy was more alert to the need to improve financial reporting than corporate management who had provided little support in efforts to change accounting terminology and methods. Stans (1949), in writing about weaknesses in accounting practices and financial reporting, stated that �. . . it is a sad but true commentary that only the enforcement powers of the Securities and Exchange Commission and the New York Stock Exchange over listed companies have brought about as much compliance as does exist. Unfortunately, there are a great many rugged individuals among management who do not concede the right of the independent accounting profession to develop principles or narrow the methods of presentation in financial statements� (pp. 469-470). A number of leaders in accountancy expressed particular concern about the public�s perception of the profession (for examples, see Carey, 1948; �What the Public Thinks about Financial Statements,� 1947; Stans, 1949; and, �The Need for Agreement on Basic Accounting Procedure,� 1947). A survey conducted for the Controllership Foundation (cited in Stans, 1948, p.101) reported that the public did not use or understand accounting information, and many did not believe the accounting data that was provided. Thus, financial statements were questioned as effective means of providing information to the public. Printed media began to suggest that annual reports of corporate income were not useful to investors. Newspapers wrote about �accounting double-talk." Bankers increasingly criticized the incomprehensibility of financial statements. Labor leaders had become outspokenly critical of the �mysteries� of corporate accounting. Although the profession�s leaders believed accountancy to be of great strategic importance in the economic environment and indispensable to business, the public, and government agencies, they realized that the profession would be in jeopardy if the public did not comprehend or appreciate the quality of services provided. The flood of criticism concerning accounting procedures and financial reporting came at a time when various business organizations and economic writers were trying to explain the function of profits. Advertising agencies were also attempting to explain the merits of the free enterprise system to a public that was �rapidly becoming conscious that the results of corporate operations are vitally important to its own welfare, and that it needs clear and reliable information about business profits� (�The Need for Agreement on Basic Accounting Procedures,� 1947, p. 455). Paton (January, 1948) and Stans (1949) were among those who wrote about the �economic illiteracy� prevalent in the years following World War II. Paton concluded accountants had a vital stake in free enterprise but had not distinguished themselves as defenders of free enterprise. He recommended that accountancy take advantage of its unique position and knowledge to explain the operations of private businesses. Paton (April, 1948) emphasized that accountancy should provide realistic recording,


55

disclose economic values, and abandon convention when the integrity of financial reporting was constrained by their use. He stated that �the primary function of accounting is to furnish significant, useful information to managers and owners of business enterprises regarding assets, liabilities, revenues, costs of production, income and so on� (p. 278). Thus, he believed that accounting should stress careful, competent measurement and not blindly follow conventions such as conservatism. Frustrations rose due to a number of controversial accounting issues. Representatives on the CAP from the larger accounting firms philosophically differed on the controversial issue of uniformity versus flexibility in the choice of accounting methods (Zeff, 1999). Stans (1948) stated that the public (meaning investor, employee, customer, and the general citizen), attempting to find simple answers from the income statement, considered the statement to be incomplete, inconclusive, ambiguous, and generally incomprehensible with obvious differences in terminology, form, construction, principle, and designation of results. The accountant on the other hand would recognize the income accounting practices to be proper and justifiable due to the past standards "under which consistency is paramount and 'general acceptance' the prime criterion of procedures" (p. 102). Stans questioned the continuing validity of "general acceptance," one of the profession's "sacred cows":

The answer seems to be that the public, now highly interested in financial accounting, wants a more direct and forthright standard, with much more emphasis on uniformity, comparability and understandability, than the vagueness of "general acceptance" by experts. After all, the idea of "generally accepted practice" is a passive concept, and implies more the existence of areas of permitted professional deviations than an objective standard of public service, especially since there is no mechanism by which individual public objection as to any accounting practice could be expressed or tallied. Is it, then, time for the ranges of permissible practice to be further narrowed in the public interest? And are the technical colloquialisms and conventions of accounting wholly justifiable? Do the existing procedures of income determination and reporting meet the standard of public service that is required by accounting's social responsibility (p. 102)?

McLaren (1948) noted that the demand for greater uniformity in

accounting had come about with the focus on accounting issues from the views of current buyers and sellers of corporate securities rather than from the view of a continuing owner. Comparisons of financial statements of companies within the same type of business were frequently misleading due to the dissimilarities of reporting practices. He stated that �an unfortunate manifestation of the search for improved reporting techniques stems from the absence of a framework of authoritative standards within which the corporate picture can be painted� (p. 382). In an editorial in the Journal of Accountancy ("Uniformity and


56

Comparability in Financial Statements", 1950), the staff noted that the debate regarding uniformity and comparability had existed for a long time. The fundamental question remained about how much standardization, comparability, and clarification of accounts can be achieved before the real differences between business environments are distorted. Since great difficulty was encountered in reaching agreement, perhaps instead of lamenting the lack of uniformity and comparability, "it might be worthwhile to try to eliminate as much of the variations as could reasonably be expected to disappear under persuasion or agreement" (p. 2). Thus, the consensus or persuasion approach continued to be supported as a means of selecting accounting procedures for financial reporting. By the mid-1950s, the accounting profession was in turmoil, and establishment of accounting standards in the private sector was in peril. Differences among accounting professionals and businessmen due to differing philosophies were evidenced in the over-simplified terms of uniformity, flexibility, or comparability. Storey (1964) said that a significant, though not very perceptible, change in attitude occurred in the mid-1950s. The change was based on an increased concern for outside users of financial statements and was manifested in the insistence that such statements be fair to all users. Accountants spoke of comparability of financial statements rather than uniformity of accounting principles. Many accountants continued to focus their criticism on such issues as the cost principle and conservatism in the belief that accountancy had been deficient in informing the general public of the nature of free enterprise and had used accepted accounting practices to overstate profits of business. Others feared that accounting was too investor-oriented, or worse, too client-oriented, and sought greater social responsibility. Most importantly, the profession had failed to establish a cohesive set of accounting principles.

Perhaps many agreed with McLaren (1948) that many of accountancy's problems occurred because no framework of authoritative standards existed. The failure of the CAP to develop a cohesive set of accounting principles continued to be a basis of criticism of accountancy. Spacek (1958), one of the vocal critics within the profession, concluded that, although the requirements for improvement of financial reporting had increased greatly, the progress of defining sound accounting principles significantly failed to keep pace. Those who said that accounting principles could not be defined were only avoiding the issue. "If it were true that we could not define generally accepted principles so the public can understand us, we might as well write our certificate in a foreign language . . ." (p. 43). "So long as the accounting profession cannot recognize accounting principles where they are highly important to the public welfare . . . , the professional accountant will continue to stumble over his own feet" (p. 46).

Independence (Montagna, 1974) is the general rule of the Code of Professional Ethics considered to be most important to the profession. Maintenance of independence is manifested in the resulting audit opinion. By attesting that the financial statements are in accordance with generally accepted accounting principles, the certified public accountant becomes responsible to stockholders, grantors of credit, government agencies, and other parties who are


57

directly interested. This responsibility to third parties is a unique function which distinguishes public accounting from other professions. Thus, the criticisms of Leonard Spacek (1958) and others regarding the lack of definition of what constituted generally accepted accounting principles and appropriate financial reporting represented issues fundamental to accountancy.

Many believe (Montagna, 1974) that the body of knowledge is the unique component on which the authority of the profession rests and that the practitioner�s autonomy is derived from superior knowledge in the related areas. The practitioner's technical skill is derived from the body of knowledge and requires reflection and highly abstract reasoning. Thus, abstract principles are applied to situations to obtain a judgment of reality. If the practitioner cannot specify the basic underlying principles which justify his superior knowledge, his authority is likely to be questioned. Yet, the body of knowledge of public accounting traditionally has been viewed as mostly experiential because auditing and its accounting base are constructed almost totally from inductive processes.

Much importance (Montagna, 1974) was attached to the body of knowledge during the early 1960s as reflected in the publishing of Horizons for a Profession (Roy and MacNeill, 1967). This study said that the profession must broaden its base of knowledge and learn to adapt to social and technological change with respect to the development of knowledge. The method of developing knowledge should be changed to include deduction. This recommendation continues to be an issue as it had been in the efforts of the Research Division during the 1960s. Practitioners (Montagna, 1974) generally favor the experimental method or the experience based inductive approach, whereas academics favor the deductive or scientific method. A CHANGE IN ATTITUDE? Alvin R. Jennings, (1958), President of the Institute, established a special committee in December, 1957 (�Special Committee on Accounting Research Program,� 1958) with a charge to consider a new approach which would achieve three things: (1) to undertake accounting research, (2) to promulgate accounting principles, and (3) to secure adherence to the principles. Based on the special committee�s recommendations, the Accounting Principles Board (APB, Board) was established in 1959 along with a new and more elaborate Research Division (Carey, 1970; Flegm, 1984). A change in attitude was signaled when the special committee proposed that the Institute�s general purpose should be to advance the development of accounting principles that were more than a survey of existing practices, determine appropriate practice to narrow differences and inconsistencies, and take the lead with regard to unsettled and controversial issues (�Report to Council of the Special Committee on Research Program,� 1958). A proactive approach in the development of accounting principles, and a conceptual framework on which to base principles would be undertaken. Research would support the deliberations of the APB, and research would include academic research and deductive reasoning (�Report to Council of the Special Committee on Research Program,� 1958). However, the authority for pronouncements continued to rely on persuasion, or the general


58

acceptance model, which conflicted with the expectation of the academic/scientific research approach and the rationality and critical attitude associated with such research.

Dr. Maurice Moonitz, a respected accounting academic, was appointed as director of the Research Division. A conceptual framework for financial accounting consisting of four levels was planned: postulates, principles, rules or other guides, and research. However, Accounting Research Study (ARS), No. 1, The Basic Postulates of Accounting published in 1961 (Moonitz, 1961) and ARS, No. 3, A Tentative Set of Broad Accounting Principles for Business Enterprises in 1962 (Sprouse and Moonitz, 1962), were rejected by accounting practitioners. As ARS No. 3 was in process, Weldon Powell (1961), chairman of the Board, stated that research in accounting is by nature applied, and accounting principles must be practical if they are to be generally accepted. He recommended, as several other practitioners had, that the Research Division should first study prevailing practice and establish what principles were actually being applied and what procedures were actually being followed. Thus, a shift in the Institute�s attitude from academic research and deductive reasoning to the previous positions of practice based on research and inductive reasoning occurred. The Board said that ARS, Nos. 1 and 3 were conscientious attempts to resolve major accounting issues but contained �inferences and recommendations in part of a speculative and tentative nature. . . . while these studies are a valuable contribution to accounting thinking, they are too radically different from present generally accepted accounting principles for acceptance at this time� (Accounting Principles Board, 1962). Widespread criticisms of ARS, No. 3 occurred because of its focus on the asset and liability measurement model which conflicted with the traditional cost and revenue allocation model (Previts and Merino, 1998) and likely �signaled the demise of any attempts to adopt a normative research orientation for the APB� (p. 313).

The Institute appointed Paul Grady, a distinguished accounting practitioner who opposed the academic approach and ARS No. 3 (Grady, 1962; Previts and Merino, 1998), as director of the Research Division and initiated the project which led to ARS, No. 7, Inventory of Generally Accepted Accounting Principles for Business Enterprises (Inventory) (Grady, 1965). The Inventory largely identified principles or practices, on which general agreement existed, and developed a useful classification of these principles. Its detailed discussion helped practitioners to understand why these practices were considered acceptable and provided concise sources for citing authorities in support of the practice. Thus, the study provided the only comprehensive discussion of accounting principles for which authoritative support existed. As was the case of the Sanders, Hatfield, Moore monograph in 1938, the Inventory was a codification of existing practice (Previts and Merino, 1998) that proved valuable to practitioners (Carey, 1970), as well as non-threatening. However, the Inventory did not provide the conceptual framework needed by the APB to deal successfully with the changing and increasingly complex social and business environment.


59

During the 1960s, widespread speculation in the securities market occurred, and a new breed of promoters and managers took advantage of inadequate accounting principles to inflate earnings and promote various types of ventures (Olson, 1982). Thus, managers were not supportive of efforts by the Board to improve asset and liability measurement and to improve the quality of earnings reporting. Since the SEC continued to be relatively neutral in its support of the Board and the Board had no authority to enforce its pronouncements, practitioners often sought to serve their clients rather than to protect third party users.

Emotions ran high between those seeking comparability and those opposed to uniformity. Attempts by the APB to gain industry cooperation in achieving comparability in earnings more often resulted in opposition to proposals rather than suggestions for solution. Although some observers believed that total comparability could not be achieved because judgment was required in the application of accounting principles, the Institute�s Council had in 1964 and 1965 declared that unnecessary obstacles to comparability should be eliminated. The SEC continued to press for greater action to reduce alternative methodology and to increase comparability (Carey, 1970).

The APB (Flegm, 1984) struggled to satisfy the SEC, investors, and business simultaneously, which was increasingly difficult to do. The public�s expectations concerning problem-solving rose faster than the accounting profession�s ability, in a business environment growing more sophisticated and complex, to respond. Businessmen were dissatisfied because pressures led the APB to issue opinions in areas that they did not believe were needed and which impinged on the accounting practices of companies. Olson (1982) noted that managers were unhappy about having to comply with expanding reporting and disclosure requirements and believed that they should have more say in the standard-setting process. Other critics believed the Board was too slow in acting to solve problems, and the pronouncements did not go far enough. In summary, the profession and the Board did not have a clear vision of their service philosophy nor the authority to enforce its pronouncements.

The APB quickly encountered problems in 1962 with APB No. 2, �Accounting for the �Investment Credit.�� Public accountants, business representatives, and government extensively criticized the APB�s decision to use the �deferral� approach (Carey, 1970). Indeed one might argue that the �deferral� approach is preferred and supported by the traditional cost allocation model. The selection of one best approach to account for investment credit conformed to the objective of narrowing the variety of methodology and providing comparability of reported results. However, the decision proved to be a serious error in judgment because the Board should have known that the majority of businesses would want to use the �flow through� approach rather than the �deferral� approach (Grady, 1979). A significant number of public accounting firms refused to follow the APB�s pronouncement, and since authority depended on consensus and general acceptability, there was little the APB could do about the situation. An attempt two years later to amend the situation with the issuance


60

of APB No. 4, which allowed either method, also drew extensive criticism in the financial press. A CHANGE OF AUTHORITY: The Board recognized that it lacked authority to enforce its pronouncements, and the SEC appeared to be taking a wait and see attitude (Carey, 1970). The original proposals for establishment of the CAP and APB recommended that pronouncements be binding on membership. However, consensus or persuasion was ultimately chosen over compulsion as the means of achieving general acceptance and authoritative support. Thus, the APB found itself in situations comparable to those that led to the demise of the CAP; achieving a consensus in deliberations and for implementation of pronouncements proved to be very difficult.

The Board recommended that Institute members disclose deviations from principles approved by the Board and proposed amendments of auditing standards and the code of professional ethics to assure compliance. The executive committee agreed with the Board's objectives and issued a special report entitled "Status of Pronouncements of Accounting Principles Board" (Carey, 1970; "Proposal to Council Would Define Authority of APB Pronouncements," 1964) which would establish "force and effect of pronouncements" of the APB. Although the Board (Jennings, 1964) had only intended that material departures from Board pronouncements be reported in some satisfactory manner, the executive committee's motion (Carey, 1970) was: "A pronouncement of the Board constitutes the only generally accepted accounting principle for purposes of expressing an opinion on financial statements, unless and until rescinded by Council" (p. 113). The motion would have, in effect, given substantial authoritative support to each pronouncement of the Board and provided an authority or power to enforce compliance; this was a radical change from the traditional provision that authority rested with general acceptability derived from persuasion and consensus through practice.

After a long and vigorous debate, a compromise was reached that material departures from APB opinions should be disclosed and that a special committee be appointed to formulate recommendations regarding implementation ("Council Adopts Resolution on APB Opinions", 1964). The special committee's recommendations established that generally accepted accounting principles had substantial authoritative support, that APB Opinions constituted substantial authoritative support, that other methodologies may also have substantial authoritative support, and that no distinction should be made between ARBs and APBs. Failure to disclose a material departure from an APB Opinion would be considered substandard reporting ("American Institute Council Acts on Recommendations for Disclosure of Departures from APB Opinions," 1964). An amended proposal was approved on October 2, 1964, and the special committee's final report was submitted in May, 1965. Most of its recommendations were accepted by the executive committee, Board, and Council (Carey, 1970); however, the Code of Professional Ethics was not amended at that time.


61

The concept of general acceptability as substantial authoritative support for alternative accounting practices in similar circumstances had been challenged. Although the Code of Professional Ethics was not amended, Article 2.02 of the Code required disclosure of any material departure from generally accepted accounting principles, and the Council had declared APB Opinions and ARBs to be generally accepted accounting principles (Carey, 1970). Thus, in effect, the basis of authority for pronouncements of the APB had begun to change. Perhaps most accounting practitioners did not fully realize such a change was occurring because they continued to think in terms of generally accepted accounting principles as having derived substantial authoritative support from experience and consensus achieved through use in financial reporting.

The accounting profession and the APB continued to encounter great criticism as financial news media began to focus on the setting of accounting standards. Major business failures and lawsuits against public accounting firms led to criticism of auditor performance and accounting principles. Questions were once more raised regarding the profession's ability to establish accounting standards and to impose them on the entire business community (Olson, 1982). The profession responded with the establishment of the Financial Accounting Foundation, the Financial Accounting Standards Advisory Council, and the Financial Accounting Standards Board in 1972 to succeed the Accounting Principles Board. Establishment of the Financial Accounting Standards Board (FASB) did not end the turbulent times for accountancy. If anything, the problems facing the profession increased and intensified. The SEC took actions that undermined the FASB's role of setting standards. A number of highly publicized frauds eroded confidence in the accounting profession and the standard setting process. Senate Subcommittee hearings led to the staff report called The Accounting Establishment which was highly critical of the accounting profession (Olson, 1982). However, Olson (1982) states that accountancy reached a new maturity during the 1970s because the responsibilities imposed by the securities acts of the 1930s were more fully realized than before. Prior to 1970, the profession had focused its attention on the needs of management. With the third party litigations begun in the 1960s and intensification of third party liability in the 1970s, accountancy began to focus more on the profession's responsibilities on shareholders, credit grantors, and other third parties. SOME CONCLUDING OBSERVATIONS: In the context of evolving professions, the events and debates within accountancy during the years following World War II are not particularly unusual. Members of professions (Montagna, 1974; Byrne, 1937) have generally taken opposing views regarding the body of knowledge and the fundamentals which support it. The conflicts between practical and theoretical are common to professions. The accounting practitioner, being more conservative, is less willing to accept rapid change and more likely to emphasize the inductive logic of experience. Accounting practitioners are likely to argue that financial accounting and auditing depend on judgment, objectivity,


62

and honesty, none of which can be regulated. The theorist or academic seeks to further the intellectual discipline through the deductive method of science and apply abrupt change to accounting practice. Academics have argued for the need for authoritative principles. Of course these and other characterizations are generalizations because few individuals fall in the extremes of these ideologies. However, practitioners necessarily tend to be realists like their clients and see accounting as an art and themselves as artists shaping the financial picture.

Opposing views also stem from the issue of art versus science approaches to accounting as an intellectual discipline (Montagna, 1974; Byrne, 1937). The debate of art versus science influences the issue of change in the profession's body of knowledge and source of authority. These issues and an analysis of their implications for accountancy are summarized in Table 1. Proof that a body of organized knowledge has reached the rank of an established science is the authoritative character of the generalizations derived from the body of knowledge and which become known as scientific laws or principles. Conflict centers on the level of certainty and the consequent loss of judgment as uncertainty is reduced. Those who support accounting as an art view establishment of the body of knowledge as derived from the experience of accounting based on inductive analysis and the existence of uncertainty. Support of accounting as a science focuses on the use of deductive analysis to establish the body of knowledge and the reduction of uncertainty. Formulation of accounting principles based on deductive research removes uncertainty and allows the profession or institution control over the external rules used to provide services. Whereas, the maintenance of uncertainty allows the expert or individual to exercise judgment in the choice of rules that govern the services provided. Debate regarding the APB centered on the source of authority and the power derived from the authority. Neither the practitioner, nor the manager, wanted to relinquish the authority derived from individual judgment in the choice of accounting methodology used in financial reporting.

Formal external rules are the most important rules governing firms and the profession. Social control of professional bodies by means of external rules is known as professional bureaucracy in whom change is dependent on variations in the extent and specificity of external rules. However, generally accepted accounting principles were traditionally viewed by Institute membership as informal external rules (Montagna, 1974), and as such, did not represent a set of rules that were binding on Institute membership. Montagna thought that with the compilation of principles in ARS, No. 7, generally accepted accounting principles were in the process of being formalized. However, such was not the case because ARS, No. 7 and other ARSs did not represent official pronouncements of the Board. These research studies were designed to be discussion instruments and to document accounting problems. Although informative, ARSs were not authoritative and binding on Institute members.


63

TABLE 1: IMPACT OF ART VERSUS SCIENCE ON ACCOUNTANCY ART SCIENCE Source of Body of Knowledge

Derived from Experience Inductive Reasoning/ Applied Research Education/ Apprenticeship Piecemeal Approach

Academic Research Deductive Reasoning Formal Education Conceptual Framework

Authority Piecemeal consensus or use Persuasion Individual/Expert

Cohesive derived from supporting research Compulsion Institutional control

Characteristics of Knowledge

Uncertainty Judgmental Informal rules or Guidance

Certain Rational Formalization of Rules

Characteristics of Professional

Practical Conservative Traditional Reverential attitude Status quo

Theoretical Innovative Adaptive Critical attitude Accepts rapid change

Service Ideal/Philosophy

Client focus Situational Comparability

Public/Society focus Universal Application Uniformity

The issue of authority to enforce standards was directly addressed by the Institute in 1973. The revised Code of Professional Ethics contained Rule 203. This rule mandated that Institute members comply with the accounting standards established by the FASB. This action was significant because it explicitly brought the standards directly under the profession's disciplinary machinery (Olson, 1982). The SEC's adoption of Accounting Series Release (ASR), No. 150 in 1973 entitled "Statement of Policy on the Establishment and Improvement of Accounting Principles and Standards," strengthened its support and reliance upon self-regulation of the accounting profession. ASR, No. 150 stated that "principles, standards, and practices promulgated by the FASB in its Statements and Interpretations will be considered by the Commission as having substantial authoritative support, and those contrary to such FASB promulgations will be considered to have no such support" (pp. 280-281, United States Securities and Exchange Commission, 1976). ASR, No. 150 also identified the issue of service responsibility as one of assuring that investors are provided with information necessary to make informed investment decisions.


64

REFERENCES: Accounting Principles Board (April 13, 1962). Statement by the Accounting Principles Board. New York: American Institute of Certified Public Accountants. American Institute of Accountants (1939). 1938 Yearbook. New York: American

Institute of Accountants. American Institute of Certified Public Accountants (1961). Accounting Research

and Terminology Bulletins, Final Edition. New York: American Institute of Certified Public Accountants.

American Institute Council acts on recommendations for disclosure of departures from APB Opinions (1964). Journal of Accountancy, November, 11-12.

A tentative statement of accounting principles affecting corporate reports (1936). The Accounting Review, June, 187-191.

A statement of objectives of the American Accounting Association (1936). The Accounting Review, March, 1-4.

Audits of Corporate Accounts (1934). New York: American Institute of Certified Public Accountants. Bryne, Gilbert R. (1937). To what extent can the practice of accounting be

reduced to rules and standards? The Journal of Accountancy, November, 364-379.

Buckley, John W. (1978). An exploration of professional identity. Ethics in the Accounting Profession, (Stephen E. Loeb, Editor) Santa Barbara: John Wiley & Sons reprinted from Chapter 3 In Search of Identity: An Inquiry Into Identity Issues in Accounting (1972) The California Certified Public Accountants Foundation for Education and Research.

Carey, John L. (1948). Public opinion of the accounting profession. The Journal of Accountancy, January, 59-64.

Carey, John L. (1970) The Rise of the Accounting Profession: To Responsibility and Authority, 1937-1969. New York: American Institute of Certified Public Accountants.

Council adopts resolution on APB Opinions (1964). The Journal of Accountancy, June, 9.

Federal Reserve Board (1917). Uniform Accounting. Washington: Government Printing Office.

Flegm, Eugene H. (1984). Accounting: How to Meet the Challenges of Relevance and Regulation. New York: John Wiley & Sons, Inc.

Grady, Paul (1962). The quest for accounting principles. The Journal of Accountancy, May, 45-50.

Grady, Paul (1965). Inventory of Generally Accepted Accounting Principles for Business Enterprises, Accounting Research Study No. 7. New York: American Institute of Certified Public Accountants.

Grady, Paul (1979). The accounting profession since World War II. Collected Papers of the Thirty-First Annual Meeting, (Irving L. Fantl, Editor) Southeast Regional Meeting, American Accounting Association. Miami, Florida International University, April 27-28, 18-24.


65

Greenwood, Ernest (1978). Attributes of a profession. Ethics in the Accounting Profession. (Stephen E. Loeb, Editor) Santa Barbara: John Wiley & Sons. Reprinted from Social Work (July, 1957). National Association of Social Workers.

Jennings, Alvin R. (1958). Present-day challenges in financial reporting. The Journal of Accountancy, January, 28-34.

Jennings, Alvin R. (1964). Opinions of the Accounting Principles Board. The Journal of Accountancy, August, 27-33.

McLaren, N. Loyall (1948). The public accountant and annual reports of corporations. The Journal of Accountancy, May, 381-385.

Moonitz, Maurice (1961). The Basic Postulates of Accounting, Accounting Research Study No. 1. New York: American Institute of Certified Public Accountants, Inc.

Montagna, Paul D. (1974). Certified Public Accounting: A Sociological View of a Profession in Change. Houston: Scholars Book Co.

The need for agreement on basic accounting procedures (1947). The Journal of Accountancy, June, 455. Olson, Wallace E. (1982). The Accounting Profession Years of Trial:

1969-1980. New York: American Institute of Certified Public Accountants. Paton, W. A. (1948). Accounting procedures and private enterprise. The Journal

of Accountancy, April, 278-291. Paton, W. A. (1948). The accountant and private enterprise. The Journal of

Accountancy, January, 44-58. Powell, Weldon (1961). Report on the accounting research activities of the

American Institute of Certified Public Accountants. The Accounting Review, January, 26-31.

Previts, Gary John and Barbara Dubis Merino (1998). A History of Accountancy in the United States: The Cultural Significance of Accounting. Columbus: Ohio State University Press.

Proposal to Council would define authority of APB pronouncements. (1964). The Journal of Accountancy, April, 9-12.

Report to Council of the Special Committee on Research Program. (1958). The Journal of Accountancy, December, 62-68.

Roy, Robert H. and James H. MacNeill (1967). Horizons for a Profession New York: American Institute of Certified Public Accountants. Sanders, T. H. (1936). Influences of the Securities and Exchange Commission

upon accounting principles. Accounting Review, March, 66-74. Sanders, Thomas Henry, Henry Rand Hatfield, and Underhill Moore (1938). A

Statement of Accounting Principles. New York: American Institute of Accountants, Haskins & Sells Foundation, Inc.

Spacek, Leonard (1958), Can we define generally accepted accounting principles? The Journal of Accountancy, December, 40-47.

Special Committee on Accounting Research Program. (1958), The Journal of Accountancy, August, 77.


66

Sprouse, Robert T. and Moonitz, Maurice (1962). A Tentative Set of Broad Accounting Principles for Business Enterprises. Accounting Research Study No. 3. New York: American Institute of Certified Public Accountants.

Stans, Maurice H. (1948). How new standards of financial reporting grow from social responsibility of Accountants. The Journal of Accountancy, August, 98-106.

Stans, Maurice H. (1949). Weaknesses in present accounting which inhibit understanding of free enterprise. The Journal of Accountancy, December, 466-471.

Storey, Reed K (1964). The Search for Accounting Principles: Today's Problems In Perspective. New York: American Institute of Certified Public Accountants, Inc.

Storey, Reed K. (1964). Accounting principles: AAA and AICPA. The Journal of Accountancy, June, 47-55.

Van Riper, Robert (1994). Setting Standards for Financial Reporting. Westport, Connecticut: Quorum Books.

Uniformity and comparability in financial statements (1950). The Journal of Accountancy, July, 1-2.

United States Securities and Exchange Commission (1973). Accounting Series Release No. 150. December 20, 1973.

United States Securities and Exchange Commission (1976). Accounting Series Releases: Compilation of Releases to 195.

What the public thinks about financial statements (1947). The Journal of Accountancy, June, 487-489.

Zeff, Stephen A. (1999). The evolution of the conceptual framework for business enterprises in the United States. Accounting Historians Journal, December, 89-131.


67

THE USE OF THE DELPHI METHOD TO ESTIMATE AUDIT AND DETECTION PROBABILITIES

Carol M. Fischer

St. Bonaventure University

ABSTRACT: This paper discusses the importance of developing measures that can be used to assess the accuracy of taxpayers� detection probability perceptions. It describes the application of the Delphi technique in this context, presents the results, and concludes with a discussion of how this method might be used to advance our understanding of taxpayers� audit and detection probability perceptions. The Delphi method utilizes a group of experts who respond to a questionnaire regarding uncertain or subjective measurements. An iterative approach allows panel members to change their responses based on anonymous feedback from other expert panel members. Final estimates are obtained when the panel members� responses stabilize. This study employs a Delphi panel to obtain estimates of audit and detection probabilities from experienced tax practitioners. To further assess the validity of the estimates provided by the Delphi Panel, the panel was arbitrarily divided into two �sub-panels� prior to soliciting the first set of estimates. Each of the sub-panels was run independently and developed a separate set of audit and detection probabilities. These probability estimates were then compared and found to be generally consistent. Detection rates for failure to report income subject to third party reporting were very high, while the detection rates for failure to report other cash income and for overstating business expenses were much lower. Detection rates generally increased with income, and were higher for larger amounts of noncompliance. However, the findings also indicated that in some instances the detection probability construct is not well-defined, suggesting the need for future research. INTRODUCTION: Recent estimates suggest that the annual loss of tax revenues from taxpayer noncompliance in the United States approaches $195 billion (Hamilton, 1998). Because noncompliance imposes very real costs on society, academic researchers have been joined by policymakers and practitioners in trying to gain a better understanding of taxpayer behavior. One variable that has consistently been associated with compliance behavior is detection probability. Yet, �... almost nothing is known about how taxpayers assess detection risks using information they have,� (Roth, Scholz, and Witte, 1989, p. 6). This study explores the use of the Delphi Method (Linstone and Turoff, 1975) to estimate audit and detection probabilities, which can serve as a basis for assessing the accuracy of taxpayers� perceptions and the extent to which taxpayers distinguish between audit and detection probabilities.

Fischer, Carol

68

The remainder of this paper is organized as follows. The next section provides a brief review of the taxpayer compliance literature, focusing on the effect of audit and detection probabilities on compliance behavior. This is followed by a section identifying the research question and describing the research design. The next section presents the results, and the final section discusses the implications and limitations of this research and identifies opportunities for further research. THEORY AND LITERATURE REVIEW: A large body of research has focused on the determinants of taxpayer compliance behavior. There have been several reviews of the taxpayer compliance literature (see, for example, Jackson and Milliron, 1986; Long and Swingen, 1991; Cuccia, 1994; and Andreoni et al., 1998). Theoretical models of taxpayer compliance are typically based on a financial self-interest model, which focuses on the costs and benefits associated with the compliance decision. Because it is generally acknowledged that taxpayer compliance behavior is also affected by taxpayers� attitudes and social norms, other models incorporate social psychological variables (see, for example, Lewis, 1982 and Weigel et al., 1987). Importantly, despite differences in these models, each predicts that increased detection probability increases taxpayer compliance. Although the importance of the perceived detection probability variable has long been acknowledged by researchers, measurement of this variable has been particularly troublesome (Fischer et al., 1992). This is due, in part, to differences in the way the variable has been operationalized in different research studies. Detection probability is defined as the likelihood that the IRS will detect a taxpayer�s noncompliance through any of its enforcement programs (including in-person audits, correspondence audits generated by computer matching of third party reports, tips received through the informant program, and other programs). Perceived detection probability is the taxpayer’s belief about the likelihood that the IRS will detect the taxpayer�s noncompliance through any of its enforcement programs. There are two important elements to this definition: first, detection probability is distinct from audit probability; second, perceived detection probability is distinct from objective (or actual) detection probability. Much of the prior research examining the relationship between detection probability and taxpayer compliance has not made these important distinctions, thus limiting the external validity of such research (Fischer et al., 1992). The accuracy of taxpayers� detection probability perceptions is important because only perceived detection probability can be expected to influence taxpayers� behavior. If taxpayers underestimate detection probabilities, policy initiatives designed to enhance taxpayer awareness of IRS enforcement programs might improve compliance. Conversely, if taxpayers overestimate detection probabilities, the IRS might be disinclined to provide more detailed information concerning their enforcement activities to the public. Finally, if taxpayers� detection probability perceptions are accurate and are found to be positively associated with compliance behavior, then increased funding of IRS enforcement programs would be expected to increase compliance.


69

An important barrier to assessing the accuracy of taxpayers� detection probability perceptions is the difficulty of measuring actual detection probabilities. Roth et al. (1989, p. 98-99) discuss several reasons why it is not possible to assess the accuracy of taxpayers� perceptions of noncompliance detection probability by comparing perceived detection probabilities to published overall audit rates. First, the audit rate varies considerably across audit classes. Second, published audit rates are for in-person audits and therefore do not incorporate the rate at which noncompliance is detected through such techniques as computer matching. Third, audit and contact rates distort detection probabilities because audits and computer matching are imperfect. Fourth, the overall audit or contact rate for all returns may be much lower than that for noncompliant returns because the DIF formula (used by the IRS to identify individual tax returns with the highest potential for noncompliance) enables the IRS to target returns that are more likely to yield audit adjustments. Finally, the relevant audit rate depends on the amount and type of noncompliance, and disaggregated audit and contact rate data are not available. This study provides a first step in developing disaggregated measures of actual detection probability. RESEARCH DESIGN: This study employs the policy Delphi method (Dalkey and Helmer, 1963; Fusfeld and Foster, 1971; Linstone and Turoff, 1975) to develop disaggregated measures of �actual� audit and detection probabilities. The Delphi method utilizes a group of experts who respond to a questionnaire regarding uncertain or subjective measurements. The responses of the participants are aggregated and a summary is provided to each panel member as feedback. Individual responses are not attributed to particular participants; this maintains the anonymity of individual panel members, helping to ensure that the group is not heavily influenced by any single member. Panel members revise their responses to the original questions on the basis of the feedback, and the responses are once more aggregated by the researcher. This iterative process continues until the responses (estimates) stabilize. Although aggregate historical tax audit rates (for in-person audits) are published, providing a measure of overall audit probability, disaggregated audit rates are not published by the IRS. In addition, detection probability is even more difficult to assess than audit probability because detailed data about the percentage of returns targeted for IRS action by means other than audits is unavailable. In the absence of concrete data about actual disaggregated audit and detection probabilities, a Delphi panel of tax experts was used to obtain estimates of these variables from experienced tax practitioners whose dealings with the IRS make them knowledgeable about IRS enforcement procedures and provide them with many data points upon which to base their estimates. This study employed dual Delphi panels to increase confidence in the Delphi Panel estimates. The Panel was arbitrarily divided into two distinct sub-panels, Panel A and Panel B, each of which generated a separate set of audit and detection probability estimates. There was no crossover of feedback between the panels. This facilitated the examination of the following research question:

Fischer, Carol

70

RQ: Did the two sub-panels’ audit and detection probability estimates converge? If the two sub-panels� estimates are similar, the credibility of the disaggregated audit and detection probabilities generated by the Delphi panel would be enhanced. The Delphi instrument employed in this study provided a brief description of three hypothetical noncompliance scenarios in which a taxpayer (1) failed to report income subject to third-party reporting; (2) failed to report income not subject to third-party reporting; and (3) overstated business expenses (see Appendix 1). The instrument was revised several times on the basis of review by accounting professionals and pilot testing. For each of the noncompliance scenarios, panel members were required to estimate audit and detection probabilities for the hypothetical taxpayer assuming several different adjusted gross income levels and two different levels of noncompliance: �high� ($5,000) and �low�($500). Both sub-panels were administered in an identical manner. A summary of panelists� comments in response to open-ended questions and summary statistics (minimum, maximum, median) describing their audit and detection probability estimates were provided to all panel members in subsequent iterations. (Panel members were provided feedback from their own panel only.) The panel members were then asked to re-assess the probabilities on the basis of this information. The entire process occurred over a 60-day period. Potential participants for the Delphi panel were identified through personal contacts with senior tax managers and tax partners at a number of public accounting firms and a review of a Pennsylvania Institute of Certified Public Accountants listing of tax specialists. A total of 29 tax experts agreed to participate in this Delphi study after personal contacts from the researcher; 23 (79.3 percent) of these individuals actually participated and 22 (75.9 percent) provided usable quantitative responses. Importantly, all 22 of the practitioners who completed the first round instrument responded to the second round Delphi instrument. The Delphi panel members provided background information, which was used to develop the profile in Table 1. Participants represented four of the then-Big 6 firms, one other national firm, and twelve regional or local firms.

TABLE 1. A PROFILE OF THE DELPHI PANEL Combined Panel Panel A Panel B # of individuals 22 11 11 Mean # of years experience as a tax practitioner 14 14 15 # with IRS experience 7 5 2 # with National Firm 10 6 4 # with Regional or Local Firm 12 5 7 # of Partners 13 7 6 Table 1 demonstrates that the panel members were highly experienced tax experts and included several individuals with significant IRS work experience in addition to their public accounting experience. Table 1 also shows that the demographic


71

profiles of the two sub-panels were similar. The experience level of participants compares quite favorably with prior studies using participants from the public accounting profession. RESULTS: Participants in the Delphi study estimated a total of 95 audit and detection probabilities for a base case and three noncompliance cases. In the base case, panel members estimated the overall audit rate for the hypothetical taxpayer at five different income levels. As discussed below, the purpose of including the base case was to provide some means of comparing the experts� responses to published audit rates, in an effort to roughly assess the presumed expertise of the panel members. The noncompliance cases were used to develop the disaggregated detection rates that were of primary interest in this study. In each of the noncompliance cases, participants estimated the (a) audit rate, (b) detection rate if audited, and (c) detection rate if not audited for five different income levels and for two different levels of noncompliance. This facilitated the calculation of 10 overall detection probability rates for each case (five income levels each with a �low� and �high� amount of noncompliance). Consistent with the method of analysis used in most prior Delphi studies (Linstone and Turoff, 1975), the median of the estimates provided by the group members was considered to be the group estimate. Group estimates were calculated separately for Panel A and Panel B, and an overall group estimate was calculated by pooling the responses of the participants in both panels (the �Combined Panel�). The majority of panel members (17 individuals, 77% of all participants) made some changes in their estimates after considering feedback from the first round, thus suggesting that they took the Delphi process seriously. These changes generally served to bring the individual estimates closer to the group estimate (median). From Round 1 to Round 2, the range between the minimum and maximum estimates decreased for 65 (68%) of the Panel A and 50 (53%) of the Panel B estimates. Changes in individual panel members� estimates did not significantly affect the overall group estimates, however. Comparison of the group estimates (medians) between rounds one and two (using the Brown-Mood Median Test) determined that for each of the 95 audit and detection probability estimates there was no statistical difference between the round one estimate and the round two estimate. This was true for both of the individual sub-panels considered separately and for the Combined Panel. Thus, a third round of estimates from the Delphi panel was not collected. Based on Brown-Mood Median tests, there were no significant differences between the Panel A and Panel B estimates of overall audit rates. The panel data was therefore pooled to develop audit rate estimates for the Combined Panel. As shown in Table 2, the Combined Panel�s overall audit rate estimates increase with adjusted gross income from 0.50% for taxpayers in the lowest income category to 3.26% for those in the highest income category.

Fischer, Carol

72

Table 2. Estimated Audit Rates (Combined Panel)

vs. Published IRS Audit Rates Estimated Published IRS Audit Rates Reported AGI Audit Rate Non-1040A Sch. C $0 to 24,999 0.50% 0.61% 1.45% $25,000 to $49,999 0.95% 0.64% 1.85%2 $50,000 to $99,999 1.20% 1.11% $100,000 to $249,999 2.25% 5.26%1 3.63%1 $250,000 and over 3.26% 1 $100,000 and over 2 $25,000 to $99,999 Similarly, published IRS audit rates for the same year for Non-1040A taxpayers increase with total positive income from 0.61% to 5.26%. Audit rates for Schedule C taxpayers during the same timeframe increased with total gross receipts from 1.45% to 3.63%. These audit rates had not yet been published at the time the Delphi Panel members participated in the study. Thus, the panel�s estimates are reasonably close to subsequently published examination rate data, suggesting that the panel members� estimates are well-informed. The primary issue of interest in this study was the comparison of the two sub-panels� disaggregated detection probability estimates. Overall detection probabilities were calculated as follows: (audit rate x detection probability if audited) + ((1-audit rate) x detection probability if not audited). These are the detection probabilities that are discussed throughout the following sections. Although Panels A and B provided similar estimates in most instances, there were some areas of disagreement between the two panels. A comparison of the two sub-panels� estimates in round one yielded statistically significant differences for only four of the 30 overall detection rate estimates (all at a level of p < 0.05). All four of these differences were in case 2, failure to report income not subject to third party reporting. In round two, however, the two panels� estimates moved in opposite directions for some cases, resulting in a greater number of differences between the detection rate estimates of Panels A and B (see Table 3). In this round, 13 of the 30 estimates differed at a statistically significant level (p < 0.05). As shown by Table 3, these differences were concentrated in the estimates for detection rates when there is a small amount of noncompliance in cases 1 and 2, both concerned with failure to report income. Despite the unexpected divergence of the panels� detection rate estimates in some instances, both panels� estimates exhibited similar patterns. Both panels provide estimates that are consistent in several respects: (1) detection rates are much higher for income subject to third party reporting than for other cash income and for business expenses; (2) detection rates for both income not subject to third party reporting and business expenses are very low; (3) detection rates are higher


73

for larger amounts of noncompliance; and (4) detection rates generally increase with adjusted gross income. All of these patterns are clearly exhibited by the data in Table 3.

Table 3. Estimated Overall Detection Rates (Medians): Round 2 Case 1: Failure to Report Cash Income Subject to Third Party Reporting Unreported Income $500 $5000 Reported AGI Panel A Panel B Panel A Panel B $0 to $24,999 33.300% 85.040% * 80.195% 89.950% $25,000 to $49,999 35.045% 89.400% * 80.254% 89.700% $50,000 to $99,999 40.050% 80.400% * 80.390% 89.700% $100,000 to $249,999 45.050% 80.800% * 89.800% 89.913% $250,000 and over 48.000% 81.576% * 88.760% 89.900% Case 2: Failure to Report Cash Income Not Subject to Third Party Reporting Unreported Income $500 $5000 Reported AGI Panel A Panel B Panel A Panel B $0 to $24,999 0.010% 0.518% * 1.029% 1.193% $25,000 to $49,999 0.016% 0.690% * 0.748% 1.980% $50,000 to $99,999 0.016% 1.080% * 0.745% 2.011% * $100,000 to $249,999 0.020% 0.960% * 0.957% 2.213% * $250,000 and over 0.025% 0.975% * 1.440% 1.720% Case 3: Overstatement of Business Expenses Overstated Expenses $500 $5000 Reported AGI Panel A Panel B Panel A Panel B $0 to $24,999 0.150% 0.318% 0.260% 0.995% $25,000 to $49,999 0.160% 0.548% 0.480% 0.995% $50,000 to $99,999 0.323% 0.644% * 0.799% 1.698% $100,000 to $249,999 0.544% 0.643% 1.600% 1.848% $250,000 and over 0.500% 0.690% 2.475% 2.197% * Panel A estimate differs significantly from Panel B estimate (p < 0.05) Overall detection rate = ((audit rate x detection rate if audited) + ((1-audit rate) x detection rate if not audited) The panels� detection rate estimates clearly underscore the importance of specifying the type of noncompliance when studying detection probability and noncompliance. The detection rates shown for Case 1, failure to report income subject to third party reporting, are very high, typically in excess of 80 percent. However, detection rates for Case 2, failure to report income not subject to third party reporting, are very low, ranging from 0.1 to 2.2 percent. Detection rates for Case 3, overstatement of business expenses, are also very low, ranging from 0.2 to 2.5 percent.

Fischer, Carol

74

In summary, the findings suggest that the two panels� detection probability estimates were similar in many respects, enhancing the credibility of the results. However, several of the disaggregated detection probability estimates did not converge, suggesting that the detection probability construct is not well defined. Detection rates are difficult to estimate, even for highly experienced tax professionals. The pattern of results indicates greater uncertainty among tax experts when estimating detection rates for small amounts of noncompliance. This is important because it is reasonable to believe that many taxpayers could face noncompliance opportunities involving small amounts. CONCLUSIONS, IMPLICATIONS, AND LIMITATIONS: The attempt to measure disaggregated audit and detection probabilities through application of the Delphi technique represents a first effort in this direction. The validity of the estimates calculated in this manner cannot be assessed, as actual IRS data on disaggregated detection probabilities is not available. Although the panel members were selected carefully to ensure that they had considerable and diverse experience with the IRS�s enforcement programs, their estimates of audit and detection probabilities were necessarily subjective. These estimates may not be the best proxies for actual audit and detection probabilities. However, overall audit rate estimates were reasonable relative to published rates, and two distinct panels generated detection rate estimates that exhibited similar patterns and were more likely than not to be equivalent. Thus, the method seems to provide some promise for future research. A Delphi panel comprised of IRS personnel might be used to develop detection rate estimates. These estimates could be compared to those of tax practitioners and taxpayers to provide further evidence concerning the accuracy of their perceptions. Determining actual detection rates is important not only for research purposes; it could also facilitate the enhancement of IRS enforcement programs by identifying the types of noncompliance that are currently most and least likely to be detected. The results of this study suggest that resources to fund the IRS�s enforcement programs must be increased. If the Delphi panel�s detection rate estimates are reasonably accurate, then actual detection probabilities for two common types of noncompliance � failure to report cash income not subject to third-party reporting and overstatement of business expenses � are extremely low (ranging from 0.1% to 2.5%). Even if the Delphi panel estimates are low, the perception among tax experts that the IRS�s detection rates are so low should be of great concern to the IRS and Congress. Unless the IRS is able to increase their enforcement efforts considerably, it is doubtful that these perceptions will change. This hurts the IRS�s image, it undermines public confidence in the tax system, and it could potentially exacerbate the noncompliance problem. The finding that detection rates vary significantly based on the type and amount of noncompliance underscores the importance of recognizing that there is not a single detection rate that influences compliance behavior. Rather, the detection rate varies for different types of compliance and for taxpayers in


75

different economic situations. This points to the need for researchers to carefully specify the detection rate construct when studying the effects of detection probability on taxpayer compliance. Further research on the measurement of detection probabilities will facilitate the assessment of the accuracy of taxpayer perceptions. This is an important element of taxpayer compliance research, since policymakers can exploit the connection between perceived detection probability and compliance only if they clearly understand the accuracy and determinants of taxpayer perceptions. The Delphi Method offers one approach to making further progress in this direction. REFERENCES Andreoni, J., Erard, B., & Feinstein, J. (1998). Tax Compliance. Journal of

Economic Literature, 36, 818-60. Cuccia, A.D. (1994). The Economics of Tax Compliance: What Do We Know

and Where Do We Go? Journal of Accounting Literature, 13, 81-116. Dalkey, N., & Helmer, O. (1963). An Experimental Application of the Delphi

Method to the Use of Experts. Management Science, 9, 458-67. Fischer, C.M., Wartick, M., & Mark, M.M. (1992). Detection Probability and

Taxpayer Compliance: A Review of the Literature. Journal of Accounting Literature, 11, 1-46.

Fusfeld, A., & Foster, R. (1971). The Delphi Technique: Survey and Comments. Business Horizons, 14, 63-74.

Hamilton, A. (May 25, 1998). The Tax Gap Game and Inklings of a Focus on Noncompliance. Tax Notes, 79, 933.

Jackson, B.R., & Milliron, V.C. (1986). Tax Compliance Research: Findings, Problems, and Prospects. Journal of Accounting Literature, 5, 125-65.

Lewis, A. (1982). The Psychology of Taxation. New York: St. Martin�s Press. Linstone, H. A., & Turoff, M. (eds.). (1975). The Delphi Method: Techniques and

Applications. Reading, Mass.: Addison-Wesley. Long, S.B., & Swingen, J.A. (1991). Taxpayer Compliance: Setting New Agendas

for Research. Law and Society Review, 25, 637-83. Roth, J.A., Scholz, J.T., & Witte, A.D. (eds.). (1989). Taxpayer Compliance,

Volume 1: An Agenda for Research Philadelphia: University of Pennsylvania Press.

Weigel, R., Hessing, D., & Elffers, H. (1987). Tax Evasion Research: A Critical Appraisal and Theoretical Model. Journal of Economic Psychology, 8, 215-35.

Fischer, Carol

76

APPENDIX 1: BASE CASE & NONCOMPLIANCE CASES Base Case: Overall Audit Rates Assume the following: (a) A form 1040 has been filed. There is wage, interest, and dividend income,

and the taxpayer has itemized deductions (Schedules A and B have been filed).

(b) The taxpayer also has a material amount of personal service income (from, for example, consulting, child care, home improvement, or other services), and has filed a Schedule C, which reflects a profit.

In addition, assume that all income subject to third party reporting has been declared, and there are no computational errors on the return. For each of the following levels of reported adjusted gross income (AGI): out of 1,000 tax returns, estimate the number of returns that would be selected for office or field examination of any item on the tax return. (Followed by the AGI levels shown on Table 2) Case 1. Failure to report cash income subject to third party reporting In addition to assumptions (a) and (b) listed in the base case, assume that the personal service income is subject to third party reporting (on Form 1099 Misc.). The taxpayer received Form 1099s reporting all of this income, but some of the income was not reported on the Schedule C. Assuming that the amount of unreported income is $500 and $5000, respectively, estimate the audit and detection rates for each level of reported adjusted gross income. Case 2. Failure to report cash income not subject to third party reporting In addition to assumptions (a) and (b) listed in the base case, assume that none of the personal service income is subject to third party reporting. Some of the income was not reported on the Schedule C. Assuming that the amount of unreported income is $500 and $5000, respectively, estimate the audit and detection rates for each level of reported adjusted gross income. Case 3. Overstatement of business expenses In addition to assumptions (a) and (b) listed in the base case, assume that all personal service income is properly reported on Schedule C, but business expenses (e.g., office expenses, supplies expense, utilities) are overstated, resulting in an understatement of business income. Assuming that the amount of overstated business expenses is $500 and $5000, respectively, estimate the audit and detection rates for each level of reported adjusted gross income.


77

THE EFFECT OF JOURNAL QUALITY ON ACCOUNTING FACULTY REMUNERATION: DOES

UNIVERSITY RESEARCH EMPHASIS MATTER?

Todd L. Sayre Carol M. Graham

University of San Francisco James R. Hasselback

Florida State University INTRODUCTION: Journal article publications play an important role in the reward structure of faculty members in academic institutions. Many studies document a positive relationship between number of publications and faculty salaries (e.g., Zivney and Bertin 1992; DeLorme et al. 1979). A smaller number of studies investigate the impact of journal quality on faculty salaries (e.g., Tuckman and Leahey 1975; Swidler and Goldreyer 1998). Despite its importance, only Gomez-Mejia and Balkin (1992) address the notion that the research emphasis of institutions affects the valuation of journal article publications. Our study investigates an issue similar to that of Gomez-Mejia and Balkin, focusing on accounting academics. In the context of agency theory, Gomez-Mejia and Balkin (1992) argue that organizational objectives should influence the emphasis that principals place on particular performance dimensions in determining pay. They hypothesize that institutions that grant doctoral degrees are more likely to reward faculty members for research productivity than are non-doctoral granting institutions. For a sample of 353 professors of management, they use surveys to collect salary information and other data. The results indicate that while both doctoral and non-doctoral granting institutions reward professors for publishing in high-tier journals equally, only the non-doctoral granting institutions reward professors for publishing in low-tier journals. We investigate whether faculty pay varies as a joint function of university research emphasis and the publication journal quality. We find that only those institutions with high, rather than low, research emphasis, show significant differences in pay between high and low journal quality publications. In addition, we find that both types of institutions consider the quantity of publication in determining faculty pay. These findings empirically support the commonly heard phrases in high-tier institutions such as, �Three or more publications in top-tier journals are needed for tenure,� as well as in low-tier institutions such as, �The quantity of publications is more important than their quality.� Our study improves on similar studies that investigate the relationship between publications and salary in several respects. First, earlier studies report results from very small samples, which were collected from only a handful of institutions. In contrast, our sample consists of 910 accounting professors from

Sayre, Graham, and Hasselback

78

126 institutions. Second, Gomez-Mejia and Balkin (1992), the study most similar to this paper, used surveys to collect salary information and other data. Although the surveys provided for a large sample, concerns arise regarding the accuracy of the data. For example, a sampling bias would result if the professors who were proud of their publication records or salaries were more inclined to complete the survey. To avoid such problems, we collected the salary information for our sample from the annual budgets of universities. RESEARCH METHOD: We requested the 1995 academic year budgets for the accounting departments of every public U.S. four-year academic institution listed in Hasselback�s Accounting Faculty Directory (1996). A total of 126 schools provided us with usable budgets. Candidates for our study were restricted to individuals included in the 1995 academic year budget who 1) held PhD�s or DBA�s as of 1995 according to Hasselback�s Accounting Faculty Directory (1995); and, 2) held the rank of professor, associate professor, or assistant professor in 1995 according to the budget. Department heads and deans were excluded. The final sample totaled 910 accounting professors. For each faculty member we collected (1) the nine-month pay for the 1995 academic year, (2) the rank achieved for the 1995 academic year (e.g., assistant professor), (3) the number of years employed in an academic accounting department after receiving the terminal degree, and (4) information pertaining to the quantity and quality of journal articles published through 1995. The budgets typically listed the pay for each faculty member and the number of months for which the pay applied (e.g., nine months). We clarified any uncertainties regarding the budget information by calling the appropriate institutional budget department. To insure the integrity of the budget information, one author made all the necessary calls. For several reasons, we decided that the dependent variable reported in our results should exclude summer support and stipends and thus reflect only nine-month pay. First, many budgets include neither summer support nor stipends. Second, summer support and stipends typically fluctuate from year to year based on variables exogenous to our model (e.g., signing bonuses for new faculty). Finally, unless the independent variables vary systematically with summer support and/or stipends, considering only nine-month salaries should not bias our results. We limited our sample to the professorial ranks of full professor, associate professor, and assistant professor. Department heads and deans were eliminated since we had no control variables related to the additional pay for such administrative jobs. The categories denoted in departmental budgets were used when given; otherwise, they were obtained from Hasselback�s Accounting Faculty Directory (1995).


79

We determined, again from Hasselback�s Accounting Faculty Directory (1995), the number of academic years each individual had been employed at an academic institution since receiving the terminal degree.1 Both rank and years of employment data were collected since each might impact pay. Three computerized databases were used to build the database of articles examined in this study. Pacific Research Company publishes two databases: Database of Accounting Research, which contains the listings of 47 accounting journals, and Database of Finance Research, which contains the listings of 40 finance journals. The Economic Literature Database (1996), which contains information pertaining to 300 economics, finance, accounting, real estate, and insurance journals, was utilized to collect data on articles published in 47 additional journals. A total of 134 academic journals were examined for authored papers. None of the databases give credit for notes, letters to the editor, departmental articles, or other instances where the author�s name does not appear in the listed table of contents.2 Full credit was given, however, for co-authored works. We separated journals into those of high and low quality. To determine the quality of the published articles, we assigned weights to the journals in our database primarily based on seven recent articles that ranked subsets of the journals in our list (i.e., Schroeder et al. 1988; Hull and Wright 1990; Hall and Ross 1991; Smith 1994; Brown and Huefner 1994; Alexander and Mabry 1994; Jolly et al. 1995). Hull and Wright (1990) ranked 39 journals by having respondents assign a rating to each journal with respect to The Journal of Accountancy, which previously had been assigned a rating of 1. Information contained in the six remaining studies was then used to assign ratings to 27 journals not included in the Hull and Wright study. We estimated the ratings for each of these journals based on their proximity to journals previously rated in the Hull and Wright study. This procedure resulted in ratings for 66 of the journals considered in the current study. Similar to the procedure used in Morris et al. (1990), these journals were then separated into eleven clusters, with all journals in a given cluster receiving the same rating. The resulting ratings ranged from a high of 2.25 to a low of .70. Most of the remaining journals not included in any previously cited study were assigned a weight of .70.

1 The Hassleback directory provides data on when an individual received his/her degree. Sometimes, this is after they were initially employed in academics. The difference is typically one year, but can be many more. Per discussions with James Hassleback, the percentage where large differences occur is small and does not appear to vary systematically with gender. 2 We personally checked minor problems such as author misspellings, name changes, use of initials rather than full first names, and instances where more than one author shared a given name. We checked the actual articles in our university libraries and were able to resolve all differences.


80

Based on the quality ratings, we classified journals as either high-tier, those with ratings of 1.60 and higher, and low-tier. Table 1 contains the journals considered in our study, along with their ratings and classification.3

TABLE 1

RATED ACADEMIC JOURNALS CLASSIFIED AS EITHER HIGH-TIER OR LOW-TIER Journal Rating Tier Journal Rating Tier

Accounting Review1 2.25 High Review of Financial Studies3 1.00 Low Journal of Accounting Research1 2.25 High Tax Adviser1 1.00 Low Journal of Accounting & Economics1 2.00 High Advances in Accounting1 0.95 Low Journal of Finance1 2.00 High International Journal of Accounting

Education & Research1 0.95 Low

Accounting, Organizations & Society1 1.60 High Journal of Accounting Education1 0.95 Low Contemporary Accounting Research2 1.60 High Advances in International

Accounting2 0.90 Low

Journal of Accounting, Auditing & Finance1 1.60 High Advances in Taxation2 0.90 Low Journal of Business1 1.60 High Critical Perspectives on Accounting2 0.90 Low Journal of Financial & Quantitative Analysis1

1.60 High Journal of Banking and Finance3 0.90 Low

Journal of Financial Economics3 1.60 High Journal of Financial Research3 0.90 Low Journal of the American Taxation Association1

1.60 High Journal of Information Systems1 0.90 Low

Management Science1 1.60 High Journal of Portfolio Management3 0.90 Low Auditing: A Journal of Practice & Theory1 1.35 Low Research in Accounting Regulation3 0.90 Low Decision Sciences1 1.35 Low Research in Government &

Nonprofit Accounting2 0.90 Low

Harvard Business Review1 1.35 Low Taxation for Accountants1 0.90 Low Journal of Accounting & Public Policy1 1.35 Low Taxes�The Tax Magazine1 0.90 Low Journal of Business, Finance & Accounting1 1.35 Low Accounting & Finance5 0.85 Low Journal of Management Accounting Research2

1.35 Low Accounting Educators Journal2 0.85 Low

National Tax Journal1 1.35 Low Accounting Historians Journal1 0.85 Low Journal of Taxation1 1.25 Low Advances in Accounting

Information Systems5 0.85 Low

Abacus1 1.15 Low Advances in Public Interest Accounting2

0.85 Low

Accounting & Business Research1 1.15 Low British Accounting Review5 0.85 Low Accounting Horizons1 1.15 Low Financial Management1 0.85 Low Behavioral Research in Accounting2 1.15 Low International Tax Journal1 0.85 Low Journal of Accounting Literature1 1.15 Low Management Accounting1 0.85 Low Accounting, Auditing & Accountability4 1.00 Low The CPA Journal1 0.85 Low Financial Analysts Journal1 1.00 Low Corporate Accounting/Financial

Manager4 0.80 Low

Issues in Accounting Education1 1.00 Low Georgia Journal of Accounting1 0.80 Low Journal of Accountancy1 1.00 Low Government Accountants Journal4 0.80 Low Journal of Corporate Taxation1 1.00 Low Journal of Cost Analysis4 0.80 Low

3 An article co-authored by two subjects included in our sample would be denoted in Table 5 as two publications.


81

Table 1 Continued Related Academic Journals Classified as either High-Tier or Low-Tier Accounting Education: Journal of Practice, Theory, & Research

0.70 Low Journal of Corporate Accounting & Finance

0.70 Low

Accounting Inquiries 0.70 Low Journal of Corporate Finance 0.70 Low Accounting Perspectives 0.70 Low Journal of Education for Business 0.70 Low Advances in Financial Planning & Forecasting

0.70 Low Journal of Empirical Finance 0.70 Low

Advances in Futures & Options Research 0.70 Low Journal of Financial Education 0.70 Low Advances in Investment Analysis & Portfolio Management

0.70 Low Journal of Financial Engineering 0.70 Low

Advances in Management Accounting 0.70 Low Journal of Financial Intermediation 0.70 Low Advances in Management Accounting 0.70 Low Journal of Financial Planning 0.70 Low Advances in Math Programming & Financial Plan

0.70 Low Journal of Financial Services Research

0.70 Low

Advances in Pacific Basin Business, Economics & Finance

0.70 Low Journal of Fixed Income 0.70 Low

Advances in Quantitative Analysis of Finance & Accounting

0.70 Low Journal of Futures Market 0.70 Low

Advances in Quantitative Analysis of Finance & Accounting

0.70 Low Journal of Housing Research 0.70 Low

Advances in Working Capital Management 0.70 Low Journal of International Accounting, Auditing & Taxation

0.70 Low

American Economic Review 0.70 Low Journal of International Financial Management3

0.70 Low

Applied Financial Economics 0.70 Low Journal of International Money & Finance

0.70 Low

AREUER 0.70 Low Journal of Investing 0.70 Low Atlantic Economic Review 0.70 Low Journal of Money, Credit &

Banking3 0.70 Low

Corporate Controller 0.70 Low Journal of Multinational Financial Management

0.70 Low

Financial Markets, Institutions & Instruments

0.70 Low Journal of Real Estate Finance & Economics

0.70 Low

Financial Practice in Education 0.70 Low Journal of Real Estate Literature 0.70 Low Financial Review3 0.70 Low Journal of Real Estate Research 0.70 Low Financial Services Review 0.70 Low Journal of Real Estate Taxation 0.70 Low Geneva Papers on Risk & Insurance Theory 0.70 Low Journal of Cost Management 0.70 Low Global Finance Journal 0.70 Low Journal of Economics and Finance 0.70 Low Info Systems in Accounting, Finance & Management

0.70 Low Journal of Risk & Insurance 0.70 Low

International Journal of Finance 0.70 Low Journal of Risk & Uncertainty 0.70 Low International Review of Economics & Finance

0.70 Low Journal of Small Business Finance 0.70 Low

International Review of Financial Analysis 0.70 Low Journal of Taxation of Investment1 0.70 Low Journal of Cost Management2 0.70 Low Mathematical Finance 0.70 Low Journal of International Financial Management & Accounting

0.70 Low New York CPA 0.70 Low

Journal of Applied Corporate Finance 0.70 Low Oil & Gas Tax Quarterly4 0.70 Low Journal of Bank Research 0.70 Low Pacific Basin Finance Journal 0.70 Low Journal of Commercial Bank Lending 0.70 Low Perspectives on Local Public

Finance & Public Policy 0.70 Low


82

Quarterly Review of Economics & Finance

0.70 Low Research on Accounting Ethics 0.70 Low

Real Estate Finance 0.70 Low Review of Accounting Studies 0.70 Low Recent Developments in Banking & Finance

0.70 Low Review of Financial Economics 0.70 Low

Research in 3rd World Accounting 0.70 Low Review of Futures Market 0.70 Low Research in Finance 0.70 Low Review of Quantitative Finance &

Accounting 0.70 Low

Research in Financial Services 0.70 Low Review of Research in Banking & Finance

0.70 Low

Research in International Business & Finance

0.70 Low

1 Journal included in Hull and Wright (1990) study 2 Journal included in Brown and Huefner (1994) study 3 Journal included in Alexander and Mabry (1994) study 4 Journal included in Jolly et al. (1995) study 5 Journal included in Smith (1994) study We also rated each doctoral program from which our sample of faculty members received their terminal degrees. We use the six categories reported in A Classification of Institutions of Higher Education (Carnegie Foundation for the Advancement of Teaching, 1994) to classify the sample into high-tier and low-tier institutions. We classified the first three categories as high-tier institutions and the last three categories as low-tier institutions. Any institutions not included in the Carnegie classification were also considered low-tier institutions. The Carnegie classification scheme categorizes institutions of higher learning based on the level of degree offered, the comprehensiveness of the espoused mission, and the level of federal support. The categories included in our sample follow:

(1) Research I Institutions offer a full range of baccalaureate programs, are committed to graduate education through the doctorate degree, and give high priority to research. These institutions receive at least $40 million in federal support and award 50 or more doctoral degrees annually.

(2) Research II Institutions offer a full range of baccalaureate programs, are committed to graduate education through the doctorate degree, and give high priority to research. These institutions receive at least $15.5 million to $40 million annually in federal support and award 50 or more doctoral degrees.

(3) Doctorate Granting I Institutions offer a full range of baccalaureate programs and offer the doctorate degree. Annually, these institutions award 40 or more doctoral degrees in at least five academic disciplines.


83

(4) Doctorate Granting II Institutions offer a full range of baccalaureate programs and offer the doctorate degree. Annually, these institutions award 20 or more doctoral degrees in at least one discipline or at least 10 doctoral degrees in three or more disciplines.

(5) Master's Comprehensive I Universities offer baccalaureate programs and are committed to graduate education through the master's degree. Annually, they award 40 or more masters� degrees in at least three disciplines.

(6) Master's Comprehensive Universities II offer a full range of baccalaureate programs and are committed to graduate education through the master's degree. Annually, they award 20 or more masters� degrees in at least one discipline.

Table 2 shows the institutions included in our sample and their Carnegie classification.

TABLE 2 Carnegie Classification of Institutions in our Sample

Classification Name of Institution Classification Name of Institution

1 Arizona State University 1 University of Wisconsin

1 Colorado State University 1 Utah State University

1 Indiana University 1 Virginia Commonwealth University

1 Iowa State University 1 Wayne State College

1 Louisiana State University 2 Kansas State University

1 Michigan State University 2 Mississippi State University

1 Ohio State University 2 Oklahoma State University

1 Purdue University 2 Southern Illinois University

1 University of Arizona 2 University of California-Riverside

1 University of California-Berkeley 2 University of Oklahoma

1 University of California-Davis 2 University of Toledo

1 University of California-Irvine 2 University of Wyoming

1 University of California-Los Angeles 3 Ball State University

1 University of Colorado-Boulder 3 Georgia State University

1 University of Florida 3 North Carolina-Greensboro

1 University of Georgia 3 Northern Arizona University

1 University of Illinois 3 Northern Illinois University

1 University of Iowa 3 University of Missouri-Kansas City

1 University of Maryland 3 University of North Texas

1 University of Massachusetts 3 University of Southern Mississippi

1 University of Minnesota 3 University of Texas-Arlington

1 University of North Carolina 4 Cleveland State University

1 University of Texas-Austin 4 Florida Atlantic University


84

Table 2 Continued: Carnegie Classification of Institutions in our sample

Classification Name of Institution Classification Name of Institution

4 Idaho State University 5 Southern Oregon State College 4 Indiana State University 5 Southwest Missouri State University 4 Louisiana Technical University 5 Southwest Texas State University 4 North Dakota State University 5 Stephen F Austin State University 4 University of Alaska-Fairbanks 5 Tarleton State University 4 University of Central Florida 5 Texas A&M-Corpus Christi 4 University of Colorado-Denver 5 University of Central Arkansas 4 University of Nevada-Reno 5 University of Colorado-

Colorado Springs

4 University of New Hampshire 5 University of Houston-Clear Lake 4 University of South Dakota 5 University of Houston-Downtown 4 Wichita State University 5 University of Minnesota-Duluth 5 Appalachian State University 5 University of Nevada-Las Vegas 5 Arkansas State University 5 University of North Florida 5 Auburn University-Montgomery 5 University of Northern Iowa 5 Boise State University 5 University of Texas of Permian Basin 5 California State University-

Long Beach 5 University of Texas-El Paso

5 Central Missouri State University 5 University of Texas-Pan American 5 Columbus College 5 University of Texas-San Antonio 5 East Carolina University 5 University of Texas-Tyler 5 East Tennessee State University 5 University of Wisconsin-Eau Claire 5 East Texas State University 5 University of Wisconsin-La Cross 5 Eastern Illinois University 5 University of Wisconsin-Oshkosh 5 Eastern Kentucky University 5 University of Wisconsin-Platteville 5 Eastern New Mexico University 5 University of Wisconsin-Stevens Point 5 Georgia College 5 University of Wisconsin-Superior 5 Georgia Southern University 5 West Georgia College 5 Henderson State University 5 West Texas A&M University 5 Lamar University 5 Western Carolina University 5 Louisiana State in Shreveport 5 Western Washington University 5 McNeese State University 5 Youngstown State University 5 Midwestern State University 6 Fort Lewis College 5 Morehead State University 6 Missouri Southern State College 5 Nicholls State University 6 Missouri Western State College 5 North Carolina-Charlotte 6 Truman State 5 Northeast Louisiana University 6 University of Illinois-Springfield 5 Northwest Missouri State 6 University of Southern Colorado 5 Pittsburg State University 6 University of Southern Indiana 5 Radford University 6 University of Texas-Dallas 5 Saginaw Valley State University 6 University of Wisconsin-Parkside 5 Southeastern Louisiana University 6 Western State College of Colorado


85

We used the Hasselback and Reinstein (1995) rankings of accounting doctoral programs to rate the institutions from which each faculty member had received his or her terminal degree. Hasselback and Reinstein ranked 73 doctoral programs by the number of articles per accounting graduate, weighted both for co-authorship and journal quality. Rankings were based on the articles written by the 1978-1992 graduates, which had been published in 41 major accounting journals. Journal quality was determined in a manner similar to that used in the present study. The ratings of doctoral granting institutions range from a high of .53 to a low of .01. All institutions not included in the Hasselback and Reinstein study, such as a foreign school, were assigned a rating of .01. Table 3 contains the ratings for the graduate programs related to our sample, along with the number of graduates included in our sample from each program. Note that 163 individuals (17.9% of the sample) earned their terminal degrees from one of the top 12 institutions (average rating = 38.5). In contrast, 33 individuals (3.6% of the sample) earned their doctorates from the 25 institutions that were assigned a rating of .01.

Table 3. Quality of Graduate Program Ratings1

Academic Institution Rating Number of Graduates


University of Chicago 0.53 12 University of Alabama 0.21 23 Stanford University 0.51 8 Columbia University 0.20 2 Carnegie Mellon University 0.47 6 Florida State University 0.20 8 University of California- Berkeley

0.39 8 University of Colorado- Boulder

0.20 12

University of Rochester 0.38 5 University of Maryland 0.20 2 Case Western Reserve University

0.37 2 University of Massachusetts 0.20 6

University of Michigan 0.36 20 University of Oklahoma 0.20 29 University of Kansas 0.33 10 University of California-L.A. 0.19 5 University of Oregon 0.33 10 University of North Carolina 0.19 15 University of Florida 0.32 20 University of Pennsylvania 0.19 3 University of Illinois 0.32 35 University of Wisconsin 0.19 31 Ohio State University 0.31 27 Memphis State University 0.18 7 University of Iowa 0.28 16 SUNY-Buffalo 0.18 1 University of Pittsburgh 0.28 3 University of Arizona 0.18 14 Cornell University 0.27 4 University of Southern

California 0.18 10

Harvard University 0.27 2 University of Texas-Austin 0.18 53 University of Washington 0.25 13 Michigan State University 0.16 23 University of Tennessee 0.24 14 Virginia Technological 0.16 13 University of Minnesota 0.23 14 University of Utah 0.14 4 Indiana University 0.22 22 Georgia State University 0.12 15 Arizona State University 0.21 15 Texas Tech University 0.12 18 Northwestern University 0.21 5 University of Mississippi 0.12 17 Penn State University 0.21 12 George Washington

University 0.11 4


86

Table 3. Quality of Graduate Program Ratings1 (continued)



Texas A&M University 0.11 22 University of Texas-Arlington

0.02 3

Boston University 0.10 2 Cleveland State University 0.01 1 Drexel University 0.10 1 Georgia Institute Technology 0.01 1 Syracuse University 0.10 1 Lehigh University 0.01 1 University of Houston 0.10 14 North Carolina State

University 0.01 1

University of Cincinnati 0.09 4 Odense Universitet 0.01 3 University of Kentucky 0.09 21 Rensselaer Polytechnical

Institute 0.01 1

University of Missouri-Columbia

0.09 35 Southern Illinois University 0.01 1

CUNY-Baruch College 0.07 1 SUNY-Albany 0.01 1 Purdue University 0.07 5 The American University 0.01 1 University of Arkansas 0.07 35 University College of Wales 0.01 1 University of North Texas 0.07 22 University of British

Columbia 0.01 4

University of South Carolina 0.07 13 University of California-Irvine

0.01 1

Louisiana State University 0.06 29 University of Connecticut 0.01 1 University of Georgia 0.06 15 University of Denver 0.01 1 Virginia Commonwealth University

0.06 2 University of Hawaii-Manoa 0.01 1

Kent State University 0.05 5 University of Northern Colorado

0.01 2

Mississippi State University 0.05 10 University of South Florida 0.01 1 Saint Louis University 0.05 4 University of Texas-Dallas 0.01 1 University of Nebraska 0.04 19 Washington State University 0.01 3 Louisiana Technical University

0.02 16 Miscellaneous foreign universities (6)

0.01 6

Total 910 1 Source of ratings: Hasselback and Reinstein (1995).

RESULTS: Table 4, Panel A provides descriptive statistics of high-tier and low-tier institutions. The total sample consists of 910 faculty members from 126 institutions. The faculty averaged a 9-month salary of $68,355, high-tier publications of 1.63, and low-tier publications of 3.23. In addition, the average number of years since receiving a terminal degree was 13.21 and the average quality of graduate program was .18. Of the 910 accounting faculty comprising the sample, 475 worked at high-tier institutions while 435 worked at low-tier institutions. Compared to those at low-tier institutions, the accounting faculty at high-tier institutions have, on average, higher salaries ($75,160 vs. $60,924), more publications in both high-tier (2.75 vs. .40) and low-tier journals (4.19 vs. 2.19), and terminal degrees from higher quality graduate programs (QGP = .21 vs. 14). Table 4 Panel B shows the results of t-tests on these differences between the high-


87

tier and low-tier institutions. The tests indicate that all of these differences are significant at the .01 level. The remaining variable, average years since being granted a doctorate, was higher for high-tier institutions (13.75 vs. 12.63), but the difference was only weakly significant (t-value = -2.7, p = .039). Table 4 also shows averages across professorial ranks. Not unexpectedly, average pay, number of years since receiving a doctorate, and low-tier and high-tier publications all increase when moving from the assistant the full professor rank. A chi-square test of independence indicates that the proportion of faculty at the professorial ranks of assistant, associate, and full professor were not significantly different between high-tier and low-tier institutions (Chi-square = 2.68, p = .262). Moreover, Panel B indicates that, as the total sample, for each of the professorial ranks, t-tests for the differences in pay, quality of graduate program, and publications between high-tier and low-tier institutions are significant at the .01 level. Not surprisingly, the difference in years since receiving a doctorate is not significant for assistant and associate ranks, but is significant at the .05 level for full professors.

TABLE 4: Characteristics the Sample and Test Statistics

Panel A: Characteristics of sample7

Assistant Professor Associate Professor Institution Type1

n Pay2 QGP3 Yrs4 LoPub5 HiPub6 n Pay2 QGP3 Yrs4 LoPub5 HiPub6

Research 1 81 68847 0.23 4.69 1.12 1.27 100 72851 0.23 13.46 3.85 2.82 Research 2 19 60333 0.16 4.79 2.00 0.21 31 64377 0.17 11.03 0.38 0.65 Doctoral 1 30 62341 0.16 5.07 1.60 0.27 31 62422 0.15 13.87 3.07 0.61 High-tier 130 66101 0.21 4.79 1.36 0.89 162 69234 0.20 13.07 3.69 1.98 Doctoral 2 33 61405 0.17 4.45 0.70 0.45 27 64534 0.15 15.04 3.70 0.63 Master's 1 89 55021 0.13 5.01 0.83 0.15 120 58657 0.14 12.97 2.32 0.17 Master's 2 9 53672 0.16 5.11 0.11 0.00 12 53541 0.15 8.00 0.50 0.17 Low-tier 131 56536 0.14 4.88 0.75 0.21 159 59269 0.14 12.94 2.42 0.25 Total 261 61301 0.17 4.84 1.05 0.55 321 64298 0.17 13.01 3.06 1.12

Full Professor Total Institution

Type1 n Pay2 QGP3 Yrs4 LoPub5 HiPub6 n Pay2 QGP3 Yrs4 LoPub5 HiPub6

Research 1 121 92066 0.25 20.95 6.71 6.21 302 79476 0.24 14.11 4.27 3.76 Research 2 18 77446 0.17 18.56 6.61 2.00 68 66706 0.17 11.28 4.03 0.88 Doctoral 1 44 76319 0.19 20.91 6.48 1.91 105 68222 0.17 14.30 4.08 1.06 High-tier 183 86841 0.23 20.70 6.65 4.76 475 75160 0.21 13.75 4.19 2.75 Doctoral 2 35 73958 0.18 20.71 5.80 2.09 95 66919 0.17 13.45 3.43 1.11 Master's 1 106 64425 0.13 18.81 2.42 0.33 315 59571 0.13 12.69 1.93 0.22 Master's 2 4 63590 0.11 19.00 2.75 0.00 25 55196 0.15 8.72 0.72 0.08 Low-tier 145 66703 0.14 19.28 3.24 0.75 435 60924 0.14 12.63 2.19 0.40 Total 328 77939 0.19 20.07 5.14 2.99 910 68355 0.18 13.21 3.23 1.63


88

Panel B: Statistical tests for differences between high-tier and low-tier institutions

Assistant Professors Associate Professors Full Professors All Professors Variable t-value Sig. t-value Sig. t-value Sig. t-value Sig. Pay2 -10.58 .000 -9.96 .000 -11.69 .000 -15.52 .000 QGP3 -5.15 .000 -5.68 .000 -7.35 .000 -10.67 .000 Yrs4 .194 .846 -.191 .849 -2.12 .035 -2.7 .039 LoPub5 -3.21 .000 -3.81 .000 -5.57 .000 -7.19 .000 HiPub6 -5.26 .000 -8.52 .000 -7.32 .000 -10.37 .000 1 Source: Carnegie Foundation for the Advancement of Teaching (1994) 2 Average 9-month salary. 3 Average quality of graduate program. 4 Average number of years since being granted a doctorate. 5 Average number of publications in the low-tier journals listed on Table 1. 6 Average number of publications in the low-tier journals listed on Table 1. 7 Professorial Rank chi-square statistic = 2.68, p = .262 Table 5 shows the results of the regressions for high-tier and low-tier institutions. The regressions for both high-tier and low-tier institutions indicate the coefficients for QGP, LOWJRLS, HIGHJNLS, and PROF are significant and that GEN, YEARS, and ASOC are not significant. The variance inflation factors indicate no serious multicollinearity among variables.4 Other diagnostic tests for multicollinearity, including eigenvalues, condition indices, and variance-decomposition proportions, also indicated no serious problems. The explanation for an insignificant GEN coefficient is discussed in Sayre et al. (2000). Briefly, they find that gender affects accounting faculty pay through differences in productivity and experience (e.g., publications, rank, and years worked), but not independently. Therefore, since on average males produced more and worked longer, their salaries were higher. The insignificant YEARS variable is consistent with Moore et al(1998) who document that when they controlled for quantity and quality of faculty productivity, seniority failed to affect pay. Comparing the coefficients, the two types of institutions appear to pay about the same amount for a publication in a high-tier journal, but differ in what they pay for a publication in a low-tier journal. The coefficients for the HIGHJRNLS variable indicate that a publication in a high-tier journal results in $1,574 for a low-tier institution and $1,968 for a high-tier institution. Ninety-five percent confidence intervals for the coefficients imply that the coefficients do not significantly differ. The coefficients for the LOWJRNLS variable indicate that a

4 Neter et al. (1985, p. 392) state that, �the largest variance inflation factor among all X variables is often used as an indicator of severity of multicollinearity. A maximum variance inflation factor in excess of 10 is often taken as an indication that multicollinearity may be unduly influencing the least squares estimates. The variable with the highest variance inflation factor is 4.42, which is well below this cutoff.


89

publication in a low-tier journal results in $892 for a low-tier institution and $292 for a high-tier institution. The 95% confidence intervals imply that the coefficients significantly differ from each other as well as from the coefficients related to HIGHJRNLS for high-tier and low-tier institutions. These results taken together suggest that both high-tier and low-tier institutions pay more for publications in high-tier journals and that the pay is comparable; however, compared to low-tier institutions, high-tier institutions pay less for publications in low-tier journals. The coefficients indicate that as compared to a publication in a low-tier journal, high-tier institutions pay about 7 times more for a publication in a high-tier journal while low-tier institutions pay only twice as much. This difference in the amount by which high and low-tier institutions reward publications in high-tier journals combined with the average difference in publications in high-tier journals explains a large part of the difference in average pay.

TABLE 5 Regression Results

Low-tier institutions (See Table 2): Unstandardized

Coefficients 95% Confidence

Interval for B

Model

B Standard

Error

t

Sig. Lower Bound

Upper Bound

VIF Constant 54,368 988 55.00 0.00 52,425 56,311 GEN -823 931 -0.88 0.38 -2,653 1,007 1.11 QGP 12,321 4,175 2.95 0.00 4,114 20,527 1.09 YEARS -59 66 -0.90 0.37 -188 70 2.14 LOWJRLS 892 110 8.11 0.00 676 1,108 1.12 HIGHJNLS 1,574 277 5.68 0.00 1,030 2,118 1.12 ASOC 1,596 1,058 1.51 0.13 -483 3,674 1.95 FULL 7,775 1,343 5.79 0.00 5,135 10,415 3.01

R2 = .391 High-tier institutions (see Table 2):

Unstandardized Coefficients

95% Confidence Interval for B

Model

B Standard

Error

t

Sig. Lower Bound

Upper Bound

VIF Constant 61,274 1,367 44.83 0.00 58,588 63,960 GEN 883 1,317 0.67 0.50 -1,706 3,472 1.11 QGP 13,283 4,195 3.17 0.00 5,040 21,527 1.12 YEARS -67 91 -0.73 0.46 -246 112 2.64 LOWJRLS 292 110 2.65 0.01 75 509 1.28 HIGHJNLS 1,968 113 17.35 0.00 1,745 2,191 1.25 ASOC 1,002 1,434 0.70 0.48 -1,815 3,819 2.18 FULL 12,569 1,988 6.32 0.00 8,664 16,475 4.42

R2 = .628


90

Variable Definitions: Dependent variable (PAY) = Average 9-month salary for year. GEN = Categorical variable (1 = female; 0 = male) QGP = Rating of doctoral granting institution YEARS = Number of years since obtaining doctorate LOWJNLS = number of publications in low-tier journals listed in Table 1. HIGHJNLS = number of publications in high--tier journals listed in Table 1. RANK = Series of two variables denoting an aspect of attained rank (1 = yes; 0 = no) ASOC = Associate FULL = Full VIF = Variance inflation factor The average number of publications in high-tier journals is 2.75 for high-tier institutions and .40 for low-tier institutions. Multiplying these averages by their coefficients yields $5,412 for high-tier institutions and $630 for low-tier institutions. The difference of $4,782 explains over 1/3 of the total $14,236 difference in average pay between high and low-tier institutions. In contrast, publications in low-tier journals add more to average pay in low-tier institutions than they add to pay in high-tier institutions. The average number of publications in low-tier journals is 4.19 for high-tier institutions and 2.19 for low-tier institutions. Multiplying these averages by their coefficients equals $1,223 for high-tier institutions and $1,953 for low-tier institutions. Finally, the independent variables explain more of the variation in salaries for high-tier institutions than they do for low-tier institutions (R2 = .628 vs. .391). This implies that high-tier institutions rely on performance measures included in our model to a greater degree than do low-tier institutions. This brings us to some of the shortcomings of our study. LIMITATIONS: Many limitations are inherent in this study. The sample collected for this study only includes public institutions. Thus, no direct inference to pay for publications in private institutions can be drawn from our results. Also, we made no attempt to measure individual performance at any level other than research productivity. Boyer (1992) proposes that scholarship include four interlocking components: discovery, integration, application, and teaching. Thus, while our model explained 62.8% of all variance in pay for high-tier institutions it explained only 39.1% of all variance in pay for low-tier institutions. Including teaching evaluations in the model would probably result in greatly increasing R squared for low-tier institutions. Moreover, categorization of published research in terms of topics investigated and methodology utilized would enhance the explanatory power of the model for both types of institutions. CONCLUSION: Although considerable research investigates the determinants of faculty pay, few consider journal quality and only one (i.e., Gomez-Mejia and Balkin 1992) considers the research emphasis of institutions. Moreover, previous


91

studies base their results on very small samples collected from only a handful of institutions and/or survey data (i.e., Gomez-Mejia and Balkin 1992). We investigate the effects of high-tier and low-tier publications on the salaries of accounting faculty in institutions with high and low levels of research emphasis. Our sample consists of 910 accounting professors from 126 institutions. We collected the salary information for our sample from the annual budgets of each university. To our knowledge no earlier study uses such a large sample of objective salary data in investigating the determinants of faculty pay. This study suggests that while both high-tier and low-tier institutions consider number of publications in determining faculty pay, only high-tier institutions account for the journal quality. In addition, while both types of institutions pay a comparable amount for a publication in a high-tier journal, high-tier institutions pay less than low-tier institutions for publication in low-tier journals. Combining the difference in pay and publications between the faculty at high-tier and low-tier institutions explains over 1/3 of the total difference between the average salaries. REFERENCES: Alexander, J.C., Jr., and Mabry, R.H. (1994). Relative significance of journals,

authors, and articles cited in financial research. Journal of Finance 49: 697-712.

Becker, G.S. (1964). Human Capital. Chicago, Illinois: The University of Chicago Press.

Bertin, W., and Zivney, T. (1992). The determinants of finance faculty salaries: The 1991-1992 FMA salary survey, Financial Practice and Education 2: 19-30.

Boyer, E. L. (1992). Scholarship reconsidered: priorities of the professoriate. Issues in Accounting Education 7 (1), 87-91.

Brown, L. D., and Huefner, R.J. (1994). The familiarity with and perceived quality of accounting journals: Views of senior accounting faculty in leading U.S. MBA programs. Contemporary Accounting Research 11: 23-50.

Cargile, B.R., and Bublitz, B. (1986). Factors contributing to published research by accounting faculties. The Accounting Review, 61 (1): 158-178.

Carnegie Foundation for the Advancement of Teaching. (1994). A Classification of Institutions of Higher Education. Princeton, NJ.

Cox, C.T., Boze, K.M. and Schwendig, L. (1987). Academic accountants: A study of faculty characteristics and career activities. Journal of Accounting Education, 5 (Spring): 59-76.

DeLorme, C., Hill, R.C. and Wood, N. (1979). Analysis of a quantitative method of determining faculty salaries. Journal of Economic Education 11, 20-5.

Gomez-Mejia L.R., and Balkin, D.B. (1992). Determinants of faculty pay: An agency theory perspective. Academy of Management Journal. 35: 921-955.

Hall, T.W., and Ros, W.R. (1991). Contextual effect in measuring accounting faculty perceptions of Accounting journals: An empirical test and updated journal rankings. Advances in Accounting: 161-182.


92

Hasselback, J.R. (1995). Accounting Faculty Directory. Englewood Cliffs, New Jersey: Prentice Hall.

________________. (1996). Accounting Faculty Directory. Englewood Cliffs, New Jersey: Prentice Hall.

________________, and Reinstein, A. (1995). Assessing accounting doctoral programs by their graduates� research productivity. Advances in Accounting, 13: 61-86.

Holland, R.G., and Arrington, C.E. (1987). Issues influencing the decisions of accounting faculty to relocate. Issues in Accounting Education, 2 (1): 57-71.

Hull, R.P., and Wright, G.B. (1990). Faculty perceptions of journal quality: An update. Accounting Horizons, March: 77-98.

Jolly, S.A., Schroeder, R.G. and Spear, R.K. (1995). An empirical investigation of the relationship between journal quality ratings and promotion and tenure decisions. Accounting Educators� Journal, 7 (2): 47-68.

Kida, T.E., and Mannino,R.C. (1980). Job selection criteria of accounting Ph.D. students and faculty members. The Accounting Review, 55 (3): 491-500.

Moore, W.J., Newman, R.J., and Turnbull, G.K. (1998). Do academic salaries decline with seniority? Journal of Labor Economics 16 (2), 352 - 66.

Morris, J.L., Cudd, R.M., and Crain, J.L. (1990). The potential bias in accounting journal ratings: evidence concerning journal-specific bias. The Accounting Educators� Journal 3 (1): 46-55.

Neter, J., Wasserman, W., and Kutner, M.H. (1985). Applied Linear Statistical Models. Homewood, Illinois: Irwin.

Sayre, T.L., Holmes, S.A., Hasselback, J.R., Strawser, R.H, Rowe, B.J., (2000). The association of gender with the salaries of accounting academics. Journal of Accounting Education 18, 2000.

Schroeder, R.G., Payne, D.D., and Harris, D.G., (1988). Perceptions of accounting publications outlets. The Accounting Educators� Journal, Fall: 1-17.

Schultz, J.J., Meade, J.A., and Khurana, I. (1989). The changing roles of teaching, research, and service in the promotion and tenure decisions for accounting faculty. Issues in Accounting Education, 4 (1): 109-119.

Smith, L.M.. 1994. Relative contributions of professional journals to the field of accounting. Accounting Educators� Journal, Spring: 1-31.

Street, D.L., Baril, C.P., and Benke, Jr., R.L. (1993). Research, teaching, and service in promotion and tenure decisions of accounting faculty. Journal of Accounting Education, 11: 43-60.

Swidler, S., and Goldreyer, E. (1998). The value of a finance journal publication. Journal of Finance, 53: 351-364.

Tuckman, H., and Leahey, J. (1975). What is an article worth? Journal of Political Economy 83: 951-968.

Zivney, T.L., and Bertin, W.J. (1992). Publish or perish: What the competition is really doing. Journal of Finance 47: 295-329.


93

REVISITING THE ORGANIZATIONAL COMMITMENT MODEL IN A PUBLIC SECTOR ORGANIZATION

Anthony F. Chelte

Western New England College

ABSTRACT: The paper examines the levels of organizational commitment for a large sample of employees within a public organization. An analysis of the traditional model of organizational commitment is presented including examinations of the relationship between commitment and role related characteristics. The study then moves to an exploration of the outcomes of the organizational commitment model with an emphasis on absenteeism, turnover, job stay, and motivation. The study examines eight specific job classifications within the organization. Tests for specific relationships that may validate the commitment model are examined. A multivariate analysis of the antecedents-outcomes aspects of the commitment model is tested. The results have implications for organizational development strategies.

INTRODUCTION: In a widely accepted paradigm in organizational theory, members and their organizations are seen in an exchange relationship. There is a reciprocal �give and take� in which participants exchange physical or mental effort in return for rewards. March and Simon (1958) in their seminal work in this area, suggested that this relationship is characterized by two general forms � production and participation (Angle and Perry, 1981). Empirical assessments of the linkages between employee attitudes and organizationally relevant behaviors have met with mixed results, however. Research along these lines has tended to focus on two key elements that are believed to affect productivity and the quality of work life. These two issues, job satisfaction and organizational commitment, have been the subject of numerous studies. For example, over 500 studies have been published using organizational commitment as a focal variable since the mid 1970�s (Eby and Freeman, 1999).

This research stems from an Organizational Employee Opinion Survey that was designed to provide baseline data on issues critical to workplace effectiveness and the concomitant implications for general human resource policy in the organization. The organization is a large state university with several distinct job levels including administrative, faculty, blue-collar, professional staff, and clerical. The focus of this paper is to assess the utility of the organizational commitment model in explaining organizationally important work outcomes. To this end, we examine the antecedents and consequences of organizational commitment. This is important. The majority of previous research has examined either the antecedents or the consequences of commitment but not both in the same study (Kacmar and Carlson, 1999). The present study extends past research

Chelte, Anthony

94

by considering both simultaneously. It goes further by examining an entire organizational setting. We look at the employee group as a whole and then focus on each of eight occupational levels within the organization. Organizational commitment has been defined in numerous ways. Some conceptualize it as the �strength of involvement one has with the organization (Brown, 1969; Hall & Schneider, 1972; Mowday, Steers & Porter, 1979). An emphasis on congruence between personal and organizational goals and values leading to commitment has been considered (Buchanan, 1974), and approaches have viewed commitment through an exchange theory framework (Becker, 1960; Meyer and Allen, 1984). The driving force, however, for virtually all studies using organizational commitment is an interest in the psychological attachment and individual has to an organization (Kacmar and Carlson, 1999). Organizational commitment is operationalized here as an affective component. That is, the employee�s feelings of obligation to stay with organization: feelings resulting from the internalization of normative pressures exerted on an individual prior to entry or following entry (Allen and Meyer, 1990). The construct is measured through the organizational commitment questionnaire (OCQ) developed by Porter et al. (1974). This scale was substantially revised through the work of Meyer and Allen (1991) who found that the original instrument assumed that commitment took only one form, affective commitment. Meyer and Allen found that organizational commitment may take three distinct forms: affective, continuance, or normative commitment. The present study examines only one of these dimensions, affective commitment, and explores various relationships between the level of commitment and key workplace indicators. The focus on organizational commitment attempts to establish reliable linkages between employee attitudes and organizationally relevant behaviors (Angle and Perry, 1981; Balfour and Wechsler, 1991; 1996; Blau and Boal, 1989; Brown, 1996; Kacmar and Carlson, 1999; Mathieu and Zajac, 1990; Meyer and Allen, 1991). Considerable attention has been directed to this construct as it is assumed to be a relatively stable employee attribute (Angle and Perry, 1981; Porter et al, 1974). A resurgence of Becker�s side-bet thesis appears to confirm that stability of commitment is manifest in the development process that an individual undergoes within an organization. Among practitioners organizational commitment is assumed to develop over time as individuals come to attach themselves and certain behaviors to the organization in exchange for various rewards (Chelte, 1983). Differing theoretical frameworks have characterized the approach to the study of commitment (cf. Kacmar and Carlson, 1999; Lok and Crawford, 1999; Sagie, 1998). However, the focus of the commitment literature appears to emphasize the notion that organizationally committed individuals tend to be more motivated toward pursuing organizational goals. A meta analysis of the commitment literature demonstrates that there are relationships between commitment and tenure in the organization, job performance, absenteeism, and turnover. However, more recent studies suggest that the original Organizational


95

Commitment Questionnaire is one-dimensional and that commitment may have three dimensions. The OCQ does tap the affective dimension of commitment as discussed by Meyer and Allen (1991). The affective dimension refers to identification with, involvement in, and emotional attachment to the organization, in the sense that employees with strong affective commitment remain with the organization because they want to do so. Two other dimensions of commitment, continuance commitment and normative commitment, yield a broader perspective. Continuance commitment refers to commitment based on employees� recognition of the costs associated with leaving the organization. This dimension is relatively analogous to Becker�s side-bet hypothesis. Normative commitment refers to commitment based on a sense of obligation to the organization. Therefore, those with strong normative commitment remain with the organization because they feel they ought to do so. While the original commitment scale developed by Mowday et al (1979) has been criticized for lack of homogeneity (cf. Benkhoff, 1997), the OCQ was used in this study and is shown to have only one factor loading suggesting homogeneity of the scale�s items. THE ORGANIZATION: The university is located in the Northeastern United States and employs approximately 5000 persons ranging from administrative personnel to unskilled blue-collar employees. The organization is the largest employer in the area. The intent of the original study was to provide broad base-line data through an Employee Opinion Survey that captured employee attitudes and perceptions on key human resource and OD variables. The findings would be integrated with the human resources strategic planning process in an attempt to address concerns and to develop strategic initiatives to improve the working climate in the organization. Of the 1,440 persons in the sample, 779 returned completed questionnaires. Employee categories were then coded to form eight (8) occupational categories for further analysis. Table 1 provides the details of the breakdown of employment categories. TABLE 1. OCCUPATIONAL CATEGORY MARGINALS

Category N Percent Administrators 45 5.9Faculty 158 20.7White Collar Upper 46 6.0White Collar Middle 175 22.9Upper Clerical 148 19.4Other Clerical 68 8.9Skilled and Foremen 38 4.9Semi and Unskilled 85 11.1Totals 763 99.8

Chelte, Anthony

96

A demographic overview of the sample shows that there are 367 men and 408 women (4 respondents did not provide their gender); Mean years of education is 14.8, more than half of the sample are married (490), 168 are single and the remaining are divorced, widowed, or separated. Better than 90 percent of the sample are full-time employees of the organization. The median number of years employed at the university is 7.04 and the median age of the group is 39.1 years. Forty one percent of the respondents report at least one child currently residing in their home.

MEASUREMENT: Organizational commitment was conceptualized following the OCQ. This instrument is constructed of fourteen statements to which respondents indicate agreement or disagreement. Commitment scores range from 1 to 4, with 4 indicating the highest level of commitment. A factor analysis examined whether these fourteen items tap the same dimension. This analysis confirmed previous findings in the literature and is consistent with research on affective commitment (cf. Eby and Freeman, 1999). All fourteen items loaded on factor 1 with an Eignevalue of 6.866. A further analysis was performed to assess the internal consistency or reliability of the items. This analysis produced a coefficient alpha of .924 confirming the internal reliability of the items. Data was collected on the antecedent variables discussed widely in the literature and on three specific outcome variables. We attempt here a partial test of the organizational commitment model proposed by Mowday et al. (1982). The elements of the model include: (1) antecedents that lead to initial commitment; (2) commitment itself; and (3) outcomes, which are particularly useful for organizational maintenance (turnover, attendance and motivation). Several hypotheses are derived from the model and are based on a sizeable literature (cf. Kacmar and Carlson, 1999): Antecedents based on personal characteristics:

1. Organizational commitment is positively related to age. 2. Education is inversely related to commitment. 3. Women are more committed than men. 4. Married persons are more committed than are singles. 5. Persons with children in the home are more committed than those without

children. Antecedents based on work-role characteristics:

1. Role conflict is inversely related to commitment. Outcomes (or consequences) derived from organizational commitment

1. Absenteeism is inversely related to commitment 2. Tenure in the organization is positively related to commitment 3. Turnover intention is inversely related to commitment 4. Motivation to perform is positively related to commitment


97

FINDINGS: The mean level of organizational commitment for this organization is reported in Table 2. Taken as a whole, the employees report a moderately positive level of commitment (3.2 on a 5 point scale). A correlation analysis was performed for the commitment scale and several independent variables. Several of the findings confirm the conclusions in the general literature (cf. Eby and Freeman, 1999).

TABLE 2 SAMPLE-WIDE COMMITMENT SCORES

Scale Score Percent 1.0 � 1.5 8.0 1.6 � 2.0 10.0 2.1 � 2.5 20.0 2.6 � 3.0 22.0 3.1 � 3.5 31.0 3.6 � 4.0 9.0 N 779 Mean 3.2 S.D. .712

Organizational commitment is positively related to age. Older employees are more strongly committed than younger employees. (r = .249, p< .001). The coefficient is of moderate strength, is in the predicted direction and is significant. It may be that older employees have fewer options for alternative employment and thus psychologically attach themselves to the host organization. This is also consistent with what Becker had labeled side-bets that accrue over time limiting the ability to disengage from an employer. The relationship between education and commitment is not consistent with the literature (r = -.048). It has been found that persons with greater educational attainment are less apt to commit themselves to an organization. This may be a result of the increased expectations associated with higher levels of education that are not met by employing organizations. In the current case, the relationship is not significant. Gender has been found to be related to organizational commitment. In the present study, however, there is no significant difference between men and women (t = 1.20; n.s.). The correlation analysis indicates that, if anything, women are less committed that are men in this organization. The coefficient is extremely weak (r = -.006), however and is not significant. What is interesting is that the sign of the coefficient is not in the anticipated direction. This raises some questions about the usual findings between gender and commitment. It has been suggested that married persons are more organizationally committed than are single people (cf. Kacmar and Carlson, 1999). This is also consistent with a �side-bet� approach offered by Becker (1960). The relationship is confirmed in this study (r = -.149, p < .001). Married persons are generally

Chelte, Anthony

98

more committed in this organization. Perhaps married persons have more at stake and thus have �invested� significantly more in the employment relationship. This may be due to issues of longer-term security. Paralleling the previous finding, it has been argued that individuals with children at home are more likely commit themselves to an organization. We are unable to confirm this relationship here, however (r = .037, n.s.). Two items were used to assess the relationship between commitment and role-related characteristics. The first measures role conflict. This aspect of job-role in relation to commitment does gain some support from the present study. The coefficient is positive and significant (r = .142, p< .001), suggesting that employees who perceive their roles as relatively conflict-free are more committed to the organization. The second role-related variable is perceptions of how interesting the work is. A moderately positive relationship is found here (r = .279, p < .001) suggesting that the more interesting the work is, the greater the psychological commitment to the organization. Three outcome level variables were examined relative to organizational commitment. The literature has clearly suggested that absenteeism, turnover, and motivation to perform are directly associated with commitment (cf. Eby and Freeman, 1999; Kacmar and Carlson, 1999). Taken together, these outcomes represent potential costs to the organization and are therefore relevant for OD consideration. Absenteeism is not closely associated with commitment in this study (r = .022). Much has been written concerning the link between commitment and absenteeism and has yielded varying results. We will return to this linkage when we analyze a predictive model later in this paper. Length of tenure with the organization has been presumed to be positively related to commitment. The more committed an individual, the longer one is likely to be associated with the organization. We do find a relationship between commitment and tenure (r = .140, p < .001). Turnover and commitment are the most studied phenomenon in the organizational commitment literature (cf. Eby and Freeman, 1999; Kacmar and Carlson, 1999). The findings seem to hold regardless of organization type. We found a moderate relationship between organizational commitment and turnover intention (r = .291, p < .001). Work performance (and motivation to perform) has been attributed to commitment. That is, �high performers� are presumably more committed to the organization. We attempted to assess motivational level as an indicator of performance intention. We utilized three measures of this behavioral intention operationalized as �involvement and effort� in one�s job. The amount of �extra effort� individual invested in their jobs was not related to organizational commitment here (r = -.039, n.s.). The effort dimension was also reflected in an item that asked respondents to compare how hard they work relative to others in the organization. Perceptions that they work harder could be construed as perceptions of effort expended on the job. We found no relationship between this construct and commitment. Finally, we employed a third measure of effort and


99

involvement on the job. Here we were interested in how individuals perceived the passing of time on the job. The idea is that if time was perceived as moving slowly, there would be less interest and effort expended (Mowday et al, 1982). Here we find a moderate relationship (*r = .259, p < .001). To summarize thus far, we find modest support for the bivariate relationships among commitment and age, marital status, role conflict, interesting work, intent to turn over, tenure and time passage. These findings are based on zero-order correlations and parallel the existing literature. We are unable to replicate the relationships among commitment and education, gender, children, absenteeism and two measures of effort and involvement in ones job. The failure to replicate and the generally weak correlations may be due more to the simple correlation technique than to real effects. To test this proposition, we will examine a partial organizational commitment model in a multivariate framework. THE PARTIAL COMMITMENT MODEL: The Organizational Commitment model consists of role-related characteristics, structural characteristics and work experiences that lead to the development of commitment. These elements are the antecedents of commitment. The model provides for outcomes stemming from increased levels of commitment that include the desire to remain with the organization, remaining with the organization, attendance, and job effort (motivation to perform). The model positions commitment as the mediating variable between the set of antecedents and outcomes that are important to organizations. Organizational commitment is viewed as a linear composite of the antecedent variables (Mowday et al., 1982). We tested this through a regression model to determine whether the model holds for this organization. We included ten antecedent variables in the equation (Age, education, role conflict, marital status, tenure, perceived work effort relative to others, level of interesting work, and compensation. Despite the emphasis in the literature on the model�s predictive power, just slightly more than 14 percent of the variance in commitment is accounted for by the ten antecedent variables in the equation. Of these ten, only three are statistically significant (degree of role conflict, age, and level of interesting work � Table 3). TABLE 3 SAMPLE-WIDE REGRESSION OF ANTECEDENTS AND ORGANIZATIONAL COMMITMENT

Statistic Marital Status

Children at Home Education

Role Conflict

Job Tenure

Interesting Work Age Income

b -.050 .004 -.006 .068 .009 .177 .009 -.011 SE .044 .012 .004 .018 .003 .024 .002 .014 B -.042 .012 -.054 .124 .012 .259 .210 -.038 R2 .144

AR2 .134 N 779 F 14.48

Constant 2.183

Chelte, Anthony

100

The second part of the model, a combination of antecedents and organizational commitment leading to the organizational outcomes, was tested. The literature suggests that the model should provide utility in understanding organizationally important outcomes such as turnover intention, absenteeism, and effort (motivation to perform). The model performs relatively well in this regard. With all variables in the equation, approximately 25 percent of the variance in turnover intention is explained (Table 4). TABLE 4 SAMPLE-WIDE REGRESSION OF ORGANIZATIONAL COMMITMENT & TURNOVER INTENTION

Stat Marital Status

Org Comm-itment Children Education

Role Conflict

Job Tenure

Inter-esting Work Age

In-come

b -.057 -.331 -.180 .224 .147 -.326 .211 .204 .272 SE .068 .052 .012 .078 .022 .058 .0333 .024 .027 B -.028 -.195 -.030 .102 .016 -.248 .018 .268 .054 R2 .279.

AR2 .271 N 779 F 33.12

Constant 3.88

A surprising finding emerges when we examined the outcome of absenteeism. The model explains slightly more than 3 percent of the variance in absenteeism.(Table 5). This is a very surprising finding given the attention that has been given to the relationship between organizational commitment and absenteeism. The final outcome examined was motivation to perform. The model explains approximately 9 percent of the variance (Table 6). TABLE 5 SAMPLE-WIDE REGRESSION OF ORGANIZATIONAL COMMITMENT & ABSENTEEISM

Stat Marital Status


Role Conflict

Job Tenure


In-come

b -.325 -.772 -.255 -.107 -.137 .116 -.714 .158 -.433 SE .052 .041 .014 .052 .

029 .034 .026 .029 .015

B -.025 -.001 -.063 -.075 -.022 .135 -.094 .032 -.130 R2 .054

AR2 .043 N 779 F 4.9

Constant 10.25


101

TABLE 6 SAMPLE-WIDE REGRESSION OF ORGANIZATIONAL COMMITMENT & MOTIVATION

Stat Marital Status


Role Conflict

Job Tenure


In-come

b .111 .176 .244 -.455 .568 .640 -.176 .238 -.730 SE .051 .049 .014 .051 .028 .033 .025 .032 .014 B .081 .015 .058 -.031 .089 .072 -.225 .046 -.210 R2 .116

AR2 .106 N 779 F 11.285

Constant 2.12

It appears that on a sample-wide basis for this organization, the Organizational Commitment Model is not particularly powerful in demonstrating what Mowday et al. (1982) and others (cf. Eby and Freeman, 1999; Kacmar and Carlson, 1999) refer to as employee-organization linkages. Of the hypothesized pool of predictor variables only a limited number emerge as significant factors. We next tested whether or not the model would be more or less appropriate controlling for occupational category. We ran separate regression analyses for each of the eight occupational groupings in this organization. For the antecedent � organizational commitment part of the model for Administrators none of the variables are significant. The outcomes prediction of the model for this category of employees yields mixed results. Over 40 percent of the variance in turnover the model explains intention. For absenteeism, little more than 11 percent of the variance is explained. For the final outcome measure, motivation to perform, the model accounts for 13 percent of the variance in the outcome variable. The second occupational group, faculty, is examined next. The antecedent � organizational regression accounts for 11 percent of the variance. The outcomes regression model for turnover intention provides support for the Mowday model. Over 27 percent of the variance in turnover intention is explained by the combination of antecedents and organizational commitment. The model does not fare well in explaining absenteeism for this group. The model accounts for slightly more than 12 percent of the variance in motivation to perform. As with the administrator category, the commitment model for the faculty group yielded similar results. Only for the turnover intention outcome does the model receive more than modest support. Upper Level White Collar: Strong support for the antecedent-organizational commitment model appears for this group. Over 48 percent of the variance in organizational commitment is explained by the antecedent variables. For the outcome side of the model, 29 percent of the variance in intention to

Chelte, Anthony

102

turnover is explained providing additional support for the model. The model does not explain absenteeism. The model does not explain the last outcome variable, motivation to perform. For this group of employees, it appears that the model has significant explanatory power for the antecedent-commitment link and for the impact on the outcome of turnover intention. Middle White Collar: No empirical support for the antecedent-organizational commitment partial model for this group of employees can be found. There is some support for the turnover intention outcome (R2 = .22). The other two outcomes (absenteeism and motivation to perform) are not supported by the model here. Upper Clerical: Little more than 12 percent of the variance is accounted for by the antecedent � organizational commitment model for this group of employees. The outcome, turnover intention, is supported by the model (R2 = .36); the model accounts for 10 percent of the variance in absenteeism and 16 percent of the variance in motivation to perform is explained. TABLE 7 R2 FOR OCCUPATIONAL CATEGORIES

Occupational Category Antecedents

Turnover Intention Absenteeism

Motivation to Perform

Administrators .239 .538 .293 .307 Faculty .155 .314 .084 .241 White Collar Upper .576 .434 .124 .234 White Collar Middle .112 .259 .081 .372 Upper Clerical .172 .400 .159 .213 Other Clerical .272 .359 .283 .266 Skilled and Foremen .403 .507 .268 .456 Semi and Unskilled .331 .391 .090 .083

Other Clerical: The first part of the model (antecedent � organizational commitment) explains 17 percent of the variance in organizational commitment. When we examine the outcomes, we find that slightly more than 26 percent of the variance in turnover intention is explained. For the absenteeism measure, the model accounts of 17 percent of the variance. For the outcome of motivation to perform, the model provides for 15 percent of the variance in motivation. Skilled Blue Collar & Supervisors: For this category of employees, the combination of antecedents explains 24 percent of the variance in organizational commitment. The turnover intention outcome is supported by the model (R2 = .349). The absenteeism outcome does not receive support from this model for this group. The motivation to perform outcome receives modest support (R2 = .282). Semi-Skilled and Un-Skilled: The antecedent portion of the model receives modest support (R2 = .26). Turnover intention also is supported as an outcome of the model (R2 = .318). No support is found for the outcomes of absenteeism or motivation to perform.


103

DISCUSSION: Organizational commitment is fostered and strengthened by the favorable treatment provided by an organization to its employees. As such, commitment can be an important barometer of the quality of work life that is experienced by organizational members. Theoretically, it also can help organizations understand key performance variables such as absenteeism, turnover, and motivation. Overall, for this large organization, the commitment model receives mixed support. While the commitment level for the employees of the organization is on the positive side of neutrality, the analyses provide only partial empirical support for the predictive or explanatory power of the model. Of a more general theoretical interest, are the occupational category analyses. Here, we examined eight occupational categories within the organization based on the standard model of organizational commitment. We found that absenteeism is much too complex to be explained by the commitment model. The weak predictive power of the model may be due in part to the way in which absenteeism was measured. We did not differentiate between voluntary and involuntary absences. This may be an important distinction and may have had an impact on the analyses here. Each of the tests of the model�s utility was evaluated on the basis of empirical fit between the model and this organization�s employees� responses. When we looked at the regression coefficients for each of the elements tested in the model particularly for each sub group of employees, there is a great deal of variability in the model�s utility. The model appears to be appropriate in some respects but not in others. For virtually all of the employee subgroups, the support for the antecedent-commitment linkage is mixed. The model fares well for the antecedent-commitment-outcome linkage for turnover intention. The model is somewhat stable for motivation to perform but provides virtually no support for the outcome of absenteeism. In summary, with the exception of the intention to turnover, the model does not appear to be a stable indicator across employee categories for this organization. CONCLUSIONS: It appears from this study that there are substantive differences within the organization when examining different employee categories. This does provide guidance to policy makers within the organization that not all employees are in the same pot. That is, there appears to be significant differences based on occupational category in terms of levels of commitment, what contributes to commitment formation, and the consequences of this commitment. The only consistent finding across all occupational groups is the outcome of turnover intention. Here the findings are robust and significant. It appears that there is a strong support for this part of the model. However, the antecedents and other outcomes of the model are not widely supported in this study. It may well be that the type of organization has an influence on whether commitment is a salient factor in the employment relationship. Only through carefully constructed comparative studies can we be certain as to why these linkages do not appear here. Further, there appear to be substantive differences between the occupational groups relative to certain antecedent and outcome

Chelte, Anthony

104

variables. This provides further support for the tentative conclusion that within the same organization, different structural and other variables have an impact on the employment relationship. While the current study does not confirm a considerable amount of previous research in this area, it does provide a basis for further investigation as to what underlies the observed differences. Finally, it may be that the other dimensions of organizational commitment may have a stronger relationship to the antecedent and outcome variables examined here. We have restricted our analysis to the affective dimension of organizational commitment. It may be that normative and/or continuous dimensions of commitment are in play here. Further research is urged in examining levels of occupational groups within the same organization. REFERENCES Allen, N. and Meyer, J. (1990). The measurement of antecedents of affective,

continuance and normative commitment to the organization. Journal of Occupational Psychology. 63, 1-18.

Angle, H.L., & Perry, J.L. (1981). An empirical assessment of organizational commitment and organizational effectiveness. Administrative Science Quarterly, 26, 1-14.

Angle, H.L., & Perry, J.L. (1983). Organizational commitment: Individual and organizational influences. Work and Occupations, 10, 123-146. Balfour, D.L., & Wechsler, B. (1990). Organizational commitment: a

reconceptualization and empirical test of public-private difference. Review of Public Personnel Administration, 10, 23-40.

Balfour, D.L., & Wechsler, B. (1996). Organizational commitment: antecedents and outcomes in public organizations. Public Productivity and management Review, 29, 256-277.

Becker, H.S. (1960). Notes on the concept of commitment. American Journal of Sociology, 66, 32-40. Blau, G. (1985). The measurement and prediction of career commitment. Journal

of Occupational Psychology. 58, 277-288. Brown, M.E. (1969). Identification and some conditions of organizational

involvement. Administrative Science Quarterly. 14, 346-355. Brown, R.B. (1996). Organizational commitment: Clarifying the concept and

simplifying the existing construct typology. Journal of Vocational Behavior. 49, 230-251.

Buchanan, B. (1974). Building organizational commitment: The socialization of managers in work organizations. Administrative Science Quarterly, 19, 533-546.

Chelte, A.F. (1983). Organizational commitment, job satisfaction and the quality of work life. Dissertation. University of Massachusetts, Amherst, MA.

Eby, L. & Freeman, D. (1999). Motivational bases of affective organizational commitment: a partial test of an integrative theoretical model. Journal of Occupational and Organizational Psychology. 72, 463-484.


105

Hall, D. & Schneider, B.(1972). Correlates of organizational identification as a function of career patter and organizational type. Administrative Science Quarterly. 15, 176-190. Kacmar, K.M. & Carlson, D.S. (1999). Antecedents and consequences of

organizational commitment: a comparison of two scales. Educational and Psychological Measurement. 59, 976-985.

March, J. & Simon, H.(1958). Organizations. New York: Wiley. Mathieu, J.E., & Zajac, D.M. (1990). A review and meta-analysis of the antecedents, correlates and consequences of organizational commitment. Psychological Bulletin 108, 171-194. Meyer, J. and Allen, C. (1984). Testing the side-bet theory of organizational commitment: some methodological considerations. Journal of Applied Psychology. 69, 372-398. Mowday, R., Steers, R. & Porter, L. (1979). The measurement of organizational

commitment. Journal of Vocational Behavior. 14, 224-247. Mowday, R., Steers, R., & Porter, L. (1982). Employee-Organization Linkages:

The Psychology of Commitment, Absenteeism, and Turnover. New York: Academic Press.

Porter, M., Steers, R., Mowday, R. & Boulian (1974). Organizational commitment, job satisfaction and turnover among psychiatric technicians. Journal of Applied Psychology. 59, 603-609.

Sagie, A. (1998). Employee absenteeism, organizational commitment, and job satisfaction: another look. Journal of Vocational Behavior, 52, 156-171.

106

NCHICA: A STRATEGIC APPROACH TO IMPLEMENTING A STATEWIDE HEALTHCARE

INFORMATION SYSTEM

Mary A. Curran Kent E. Curran

University of North Carolina at Charlotte

ABSTRACT: Recent governmental passage of the Health Insurance Portability and Accountability Act (HIPPA) has focused both public and private organizations on issues related to the development of a comprehensive healthcare information network. In light of this action, the question that comes to the forefront most often is �How can government and the private sector come together to solve the problems of universal health information data interchange?� This paper reports on the pioneering approach taken by the State of North Carolina in trying to deal with the critical issues related to the development of integrated and interactive healthcare information systems. A historical and evaluative approach explains the origins, evolution, successes and failures of the North Carolina Healthcare Information and Communications Alliance (NCHICA).

OVERVIEW: Governmental legislation shows a steadily growing focus on healthcare information. The provision of high quality healthcare at a reasonable cost to all citizens is a common goal at both the state and federal level in the United States. Both of these areas need detailed, current information on the quality, quantity, content, and cost of rendered healthcare services. This paper analyzes one state�s strategic plan to deal with critical issues revolving around healthcare information. Discussion includes the mission, formation, successes and limitations of this approach for dealing with these needs in the electronic age.

The North Carolina Healthcare Information and Communication Alliance (NCHICA) was established in 1994 by gubernatorial executive order. The Governor directed the formation of an organization to facilitate the development and implementation of a statewide healthcare information system to improve the delivery, quality, accessibility and efficiency of healthcare service in North Carolina. NCHICA was incorporated in 1994 as a not-for-profit organization. This charter created a unique model for public/private collaboration in healthcare: a privately funded, not-for-profit corporation open to a range of health provider organizations and healthcare IT corporations. With both private and public membership, each constituent is viewed as bringing various assets and commitments needed to achieve NCHICA goals. Members represent information system venders, healthcare providers, relevant state departments and divisions, and other contributors. These groups and �former competitors� joined to tackle some of the social and technological challenges faced in information sharing.


107

Because NCHICA was to be self-sustaining and not dependent on government financing, constituents (initially 18 private and public organizations) paid annual membership dues. These pro-rated dues included monetary and in-kind contributions. Since its inception, NCHICA has functioned under the same organizational hierarchy. The main governing body consists of a Board of Directors that serves and answers to the membership. This group sets the policies and overall direction of the organization. The Board meets on a regular basis (initially every month) to establish, plan, direct, and evaluate NCHICA�s goals and activities. Board members can designate individuals or organizations to serve as Board advisors, in Working Groups or on various Committees. Non-administrative strategic and policy actions of the Board of Directors are monitored by a Standing Advisory Committee (SAC). Because of its oversight function, committee members can not serve on the Board of Directors. These participants represent various healthcare professional societies and associations. Members of the SAC give NCHICA insight into the healthcare provider perspective. The SAC reviews NCHICA policies and strategies with current healthcare practices and information technology to �serve public rather than private interests� (NCHICA, July 1995). To manage NCHICA in the time between Board meetings, there is an Executive Committee. Members are elected from the Board of Directors and this committee may exercise the Board�s authority in the management of the Corporation. It can set operational policy and has responsibility for the �business� of the Board, while subject to Board oversight. Other work within the corporation is handled by various groups and committees that are established by the Board of Directors.

NCHICA Organization and Activities

Standing Advisory Committee

Executive Committee

Board of Directors

NCHICA Members

�Committees� are formed by the NCHICA Board and/or Bylaws. They monitor and steer certain activities, provide the Board and Executive Committee with advice, and make suggestions to influence strategic direction and organizational composition. Committees are generally permanently constituted, because they serve an ongoing role and their membership is by appointment and/or election, depending on committee and position. Focus Groups are also established by the Board. These groups are composed of member-affiliated individuals with an interest in particular NCHICA-related information or activities. They serve to alert members about

Curran and Curran

new developments, areas of concern, and issues needing attention in the Focus Group�s area of expertise. Membership is by mutual consent of the Focus Group. Each group sets their own agendas, resources, etc Ad Hoc Groups are composed of member-affiliated individuals who, with approval of the Board, come together to address particular questions, issues, or problems. Ad Hoc�s are formed when members see advantages to working together for a defined purpose and over a limited time: to answer a question, to provide recommendations on an issue, and/or to solve a problem. Membership in Ad Hoc Committees is also by mutual consent. From its inception, NCHICA has focused on fulfilling its charter. Early corporation projects and concerns explored technologies to integrate highly fragmented healthcare delivery and information systems. Efforts began to achieve consensus for computerized patient records (CPRs) throughout the healthcare system. Groups worked to implement standards and develop user interfaces for a digital patient record for hospital and office. One of these standardization efforts was directed toward a statewide master patient index (MPI). Based on open system architecture, the MPI would allow access to limited/relevant patient information held by other provider/agencies. While working on the MPI, several concerns emerged along with a growing realization that many historic, inherent problems needed a different approach. Several overriding reasons emerged for the difficulties encountered in providing a comprehensive healthcare information and telecommunications network: 1. Such network development requires collaboration between both public and

private sectors to create information, telecommunication, and telemedicine technologies applicable to healthcare in all settings;

2. There has been no overriding incentive for public and private sector organizations to overcome the difficulties mentioned above;

3. No one group has the power, resources, and expertise to accomplish the task: it requires collaboration across public and private sectors; and,

4. There has been no standardization of system parameters, databases, patient records, or terminologies.

Another major concern identified with the MPI was that of patient privacy and confidentiality. The missing answer to the question of �how to guarantee patient privacy� continues to stifle progress with electronic medical records. Because of existing limitations in privacy and confidentiality legislation, NICHICA drafted and proposed laws to protect the patient�s information and rights. This legislation would ground the development of an effective healthcare system by strengthening privacy and confidentiality, while allowing maximum access for information for clinical research and health policy management. For the health system to function effectively, there is a need for an interactive and real time exchange of medical information. Before NCHICA, no mechanism was available to overcome the social, business, and technological challenges inherent in enabling such an information exchange.


109

Systems for state and local governments to exchange electronic data may not exist, and if they do exist, they may not interact. Taxonomies among healthcare providers and payers vary widely both among and within the groups. For example, nursing identifies patient care data in a different manner and with a different perspective than medicine. Therefore, NCHICA began to focus on projects that supported both vertical and horizontal integration. These ventures were scaleable, but accommodate a wide and growing range of domain specific applications. In addition to the need for information technology (IT) solutions to many healthcare problems, NCHICA�s Board identified that there is a lack of knowledge about information technology within the healthcare community. In response to this knowledge deficit, the corporation has (since 1995) consistently provided various educational programs and annual conferences to the healthcare and IT community. Over the past six years of NCHICA�s existence, various interest groups and committees have formed and evolved within the corporation. The current organizational chart reflects this expanded and diverse IT focus.

NCHICA Organization and Activities

Standing Advisory Committee

Education

Annual Conference/Exhibition

Conferences/Workshops

Management Committee

HIPPA AdminSimp

Privacy

Transactions Codes & Identifiers

Security

Awareness Ed Training

Nominating Committee

Clinical Applications Technology Infrastructure

STEER/NCEED Project Telehealth Initiative

PAIRS RWJF HealthKey Project

Regional PAIRS NBC Terrorism

Organ/Tissue Donation Emergency Response Network

DEEDS & Claims

Deliverables Committee

Executive Committee

Board of Directors

NCHICA Members

Work continues toward a computerized patient record (CPR). The professional societies and associations that are represented on the Standing Advisory Committee of NCHICA committed to have each of their groups adopt a goal of �Paperless, person-centered healthcare by 2010� (NCHICA, 1999). The promise of building a common vision with articulated barriers and methods of overcoming these barriers is very exciting and may prove to be crucial in bringing healthcare professionals into the process of reaching this goal (NCHICA, 2001).

Curran and Curran

110

Telecommunications and telemedicine have remained a strong area for NCHICA. Current efforts include the development of engineering standards, clinical standards, and protocols for implementing telemedicine for emergency and chronic care consults (NCHICA, 1998). Passage of the Privacy Rules for the Health Insurance Portability and Accountability Act (HIPAA) earlier this year, has drastically increased the incentive to solve the problem of a comprehensive healthcare information network. The HIPAA regulations finally mandate the interchange of data between different organizations. There is no longer a question as to whether network standards will be set. The query has now become �what is the best approach to satisfy HIPAA regulations and even go beyond them?� NCHICA has been on the forefront of this legislation and its implementation.

POSITIVES/NEGATIVES: The strength of NCHCA lies in its ability to bring people together to accomplish something they could not do alone. It is an organization of organizations (NCHICA, 2001). Since its formation, NCHICA has consistently followed its mission. Adherence has occurred through projects, educational programs, consultation, and affiliations that combine the talents donated by the various member organizations. This merging of abilities and member efforts has given NCHICA success. NCHICA�s statewide strides in information technology are expected to continue to increase because of its successful accomplishments with projects like Provider Access to Immunization Registry Securely (PAiRS) and Standardization and Electronic Transmission of Emergency Records (STEER). Telehealth projects are also seeing statewide adoption. Deployment and use of communications and video technology is expanding and rural healthcare is being improved through remote clinical assistance. At the national level, the corporation has provided resources for, and participated in, work on the Health Insurance Portability & Accountability Act (HIPAA) through consultation, a NCHICA HIPAA Implementation Planning Tack Force and the Robert Wood Johnson Foundation HealthKey Multi-state Project. HealthKey looks at a technology roadmap for the security of networks and transactions by utilizing a Public Key-encryption Infrastructure (PKI) and other technologies. NCHICA functions if its members and their parent organizations are willing to �donate� time and money to various tasks, groups, projects, etc. Members are willing to participate if they perceive some derived benefit from their efforts. In an effort to market to new members and inform current members of the benefits of staying members, NCHICA has had to work on a marketing strategy and to improve communications (Table 1). For example, over time, NCHICA has reduced the frequency of Board of Director meetings from monthly to quarterly because members stated that they did not believe more frequent meetings were beneficial. Membership dues have also decreased over time. Part of the dues decrease is attributable to the growing number of �paying� members,


111

but another part was in response to questions raised by members about the cost benefit ratio of remaining in NCHICA.

Table 1 NEEDS/ BENEFITS SUMMARY FOR NCHICA’S MEMBERS The following summary outlines what experience and research indicate as key business needs, issues and priorities of NCHICA�s members. Since 1994, NCHICA has made contributions in each area that will benefit members.

Members� Needs NCHICA�s Contributions Collaborative Environment among Healthcare Industry Providers and IT Partners

Sponsors non-competitive, solution-sharing task forces and education. (e.g., Y2K and HIPAA member planning) Membership with technology focus ~ leads to product and efficiency improvements. Provide input in IT regulatory policy developments, on behalf of members. (e.g., health information security and privacy)

Healthcare Industry IT Standards Implementation Guidance

E-commerce transactions ~ focused on claims, remittance and eligibility. Led inter-operability for computer-based records ~ cost-effective approach. Led concept development for statewide master patient index system. Internet use compliance guidelines ~ key role with HCFA development. Key participant for EMR, digital signature legislation and privacy.

IT Enabled Solutions to Improve the Quality of Healthcare

Development of database and data exchange for Emergency Medical System outcomes management ~ pilot project grant funded. (STEER) Sponsorship of public-private, Internet-based immunization registry (PAiRS). Initiate goal of paperless health record by 2010 and adoption by professional organizations.

Accountable and Cost-Effective IT Leadership

Privately-funded, non-profit status. Accountable to member-selected Board. Over 70% of support is from member fees.

In spite of its many accomplishments, NCHICA has areas that need improvement. The number of paid staff is limited, but organizational activities are growing. As NCHICA�s focus expands from a state to regional and national level, more staff are needed. The consumers of healthcare have very limited representation in NCHICA. This missing group forms an important part of information technology: recipients and data providers. However, they currently have no real �voice� to reach the group. Model legislation for the privacy and security of electronic health information established by NCHICA serves as a template for other state and federal efforts, but legislative resolution in North Carolina has not been achieved. NCHICA support for further development of privacy and confidentiality laws must continue. However, this perseverance necessitates continued expenditure of resources and efforts.

Curran and Curran

112

CONSIDERATIONS/CONCERNS: Funding and membership have always been a concern for NCHICA and these concerns are interrelated. Membership dues create the financial basis for organizations activities; i.e., the more members the more dues. However, the largest dues come from for-profit member companies. Even though NCHICA is a not-for-profit corporation, it must provide some benefit or �profit� to obtain and maintain members. While there have been some slight variations, enrollment efforts show a steady increase since the corporation�s beginnings and current membership is cited as 185 organizations.

Membership dues alone cannot finance NCHICA�s activities; therefore, the corporation has had to pursue other was to obtain money. One of these sources has been grants: private and federal. Although NCHICA has had some success, the future of this type of income source is restrictive, inconsistent and unusable for long term planning. Other revenue sources have emerged from NCHICA projects and activities. The creation of HIPAA EarlyViewTM security self-assessment, gap-analysis tool has generated income and set a model for the development of similar tools for privacy and security. Educational programs and seminars provide further opportunities for income, but attendance is inconsistent and planning is labor-intensive Making a �profit� must be carefully handled in a not-for-profit corporation to avoid endangering the incorporation. Because, many groups can be involved in a NCHICA project, questions arise about who owns the products or intellectual property rights and who should receive any income.


113

CONCLUSIONS: Improvement in healthcare will occur through the utilization of advanced information, telecommunications, and telemedicine technologies. NCHICA showing itself to be an effective strategy for effecting the ongoing development and implementation of open architecture, interoperable, integrated and interactive information systems in healthcare settings. These system efforts are beginning to create fully articulated linkages in a statewide healthcare information network and to collaborate with systems and standards for national and international healthcare information. One of the interesting and forward looking characteristics of this North Carolina solution is that it was begun well before many of the current federal regulations and has served as a role model for other initiatives. NCHICA has actively promoted the advancement and integration of information technology into the healthcare industry. While problems and concerns will probably remain an organizational unpleasant fact, NCHICA may stay successful. With the right strategy, this �organization of organizations� could continue to improve healthcare through information technology and secure communications in ways that none of its members would achieve alone. REFERENCES NCHICA (July 26, 1995). Intellectual property policy of the North Carolina Healthcare

Information and Communication Alliance, Inc. NCHICA (1998). NCHICA Annual Report 1998. NCHICA (1999). NCHICA Annual Report 1999. NCHICA (2001). NCHICA Annual Report 2001.

114

UNCERTAINTY AND THE DESIGN OF DEPARTMENT OF DEFENSE CONTRACTS

Gerald M. Groshek

University of Redlands David J. Rose

Naval Postgraduate School ABSTRACT: The defense contracting problem has been perceived as one of constructing optimal risk-sharing arrangements while providing incentives to control costs. This paper develops a model of the effects of extrinsic or environmental uncertainty on the design of DoD contracts. The conventional interest with risk-sharing mechanisms is extended to demonstrate how sources of extrinsic uncertainty in both production and consumption combine to determine contract type. Within specific contract types, the relative level of extrinsic uncertainty for each of the contract participants influences its ability to define contract terms. INTRODUCTION: Variety in contracting mechanisms is a characteristic of Department of Defense (DoD) procurement. The types of contracts and the conditions under which each applies are subject to numerous DoD regulations, guidance, and procedural specifications (see the Defense Acquisition Deskbook and DFARS). The DoD�s objectives and the nature of the product determine whether the contract will employ cooperative or non-cooperative interaction between the DoD and its suppliers. For standard, well-defined products, the DoD�s objectives are uni-dimensional and center on the issue of cost. Here, fixed-price contracts are stipulated, if not required. Alternatively, when a product is undefined, the DoD�s objectives are multi-dimensional and include quality, delivery, capability, as well as price. To control this uncertainty, cooperative contracting mechanisms are employed. Initial conditions of uncertainty, then, have an a priori effect on the degree of cooperation contained in DoD contracts. Many studies of defense contracting focus on how the DoD and its suppliers interact to allocate responsibility for production uncertainties while controlling information asymmetries that favor the supplier (Sandler and Hartley 1995 provide an overview of the literature). Where possible, the use of competitive mechanisms such as sealed-bids and fixed-price contracts is advocated (Anton and Yao 1990, Gansler 1989a 1989b, Laffont 1986). However, the recognition that defense markets are often far from the competitive ideal and the frequency of sole-source procurement has motivated the design of efficient incentive and cost-sharing contracts (McAfee and McMillan 1986, De Mayo 1983, Cummins 1977). In markets for �first of a kind� products, the DoD�s responsibility for production uncertainties, as reflected in cost-plus-fixed-fee arrangements, often becomes complete. Bos and Lufesmann (1996) note that the


115

government often abandons ex ante fixed prices to renegotiate prices upward in order to achieve optimal results. Others point to increased regulatory control as offering an alternative, but not costless, approach to controlling supplier behavior (Austin and Larkey 1992, Kovacic 1992, Hartley 1991, Turpin 1989). The defense contracting problem is largely perceived as one of constructing optimal risk-sharing arrangements while providing incentives to control costs. This paper takes a different track by developing a model that illustrates the effects of various sources of uncertainty on the choice of DoD contracts. Along the way, we note that information asymmetries constitute only part of the overall uncertainty in contracting. We expand the conventional interest with incentive and risk-sharing mechanisms to demonstrate how distinct sources of environmental or extrinsic uncertainty --in production and in consumption-- combine to determine contract type. From a static perspective, the model is instructive in demonstrating when initial contract types follow a uni-dimensional, non-cooperative pattern versus a multi-dimensional, cooperative form. Dynamically, the model illustrates the effect of temporal shifts in uncertainty on the evolution in contract terms and type. The next section clarifies the intrinsic and extrinsic sources of uncertainty in which one can classify the extant literature on agent characteristics, market structure, and product life-cycle. This is followed with a model that links extrinsic uncertainty in production and in consumption with the subsequent choice of initial contract type. Before concluding with policy suggestions, we introduce a dynamic perspective and connect changes in extrinsic uncertainty with the effects on the DoD and its suppliers as well as to evolution in contract terms. UNCERTAINTY: To better interpret the effect of uncertainty on contract design, we borrow the distinction between intrinsic and extrinsic uncertainty applied in the monetary policy credibility literature (see Blackburn and Christensen 1989 for a review). An application of these two sources of uncertainty links the sources of uncertainty with the perspective of each contract participant and the resulting contract type. Intrinsic uncertainty involves one participant�s ignorance of the other�s attributes and inclinations. In contrast, extrinsic uncertainty arises from the existence of random marketplace interaction, undetectable events (i.e., technological changes, political shifts), and exogenous shocks (i.e., factor shortages) that may alter specific production and demand functions.

The standard principal-agent information asymmetries between the DoD and its suppliers provide the main source of intrinsic uncertainty. The DoD, as principal, may be unable to discern or verify the true production capabilities and preferences of its suppliers or agents (Cyert and March 1963, Cummins 1977). Conversely, intrinsic uncertainty over consumption specifications are assumed to be negligible because the DoD is assumed to gain nothing by concealing its preferences and utility. Although there might exist timing and quantity factors that are subject to shifting political influences, these are addressed subsequently under extrinsic uncertainty. Information asymmetries create intrinsic uncertainty

Groshek and Rose

for the principal (DoD) both before and after a contract is awarded. The manner in which the DoD manages intrinsic uncertainty depends on the degree of extrinsic uncertainty.

Unlike its intrinsic counterpart, extrinsic uncertainty does not result from information asymmetries between the DoD and its suppliers. It arises instead from the state of technology and the influence of exogenous, stochastic events that are unknown to and uncontrolled by either party (McAfee and McMillian 1986). Anton and Yao (1990) survey the research on defense contracts and implicitly connect the type of contract to the level of extrinsic uncertainty in production. Under the assumption that the DoD is the primary consumer, they tie the structure and evolution of the market to the level of extrinsic uncertainty in production. However, since conditions and events affect the production function and the demand function differently, a precise picture of the relation between extrinsic uncertainty, market structure, and contract type requires the separate consideration of extrinsic production uncertainty and extrinsic consumption uncertainty.

Items with negligible levels of extrinsic uncertainty are typified by routine production and consumption functions and a large number of actors in competitive markets. In this environment, non-cooperative interaction between the DoD and its suppliers is used to manage the residual intrinsic uncertainty. The design of an optimal contract selection mechanism centers on the consequences to both parties of the price that results from various auction types (Myerson 1981). Issues relating to non-price characteristics such as quality and timeliness are assumed to be resolved by the non-cooperative, competitive interaction between agents in the market. The use of uni-dimensional, price-centered (fixed-price) contract mechanisms results.

Conversely, in initial stages of development, basic questions concerning the manner in which output is produced and consumed remain unanswered. The effort required to address these issues reduces the number of agents willing or able to engage in production. Likewise, items with undefined characteristics and effectiveness attract few if any consumers willing or able to chance an untested product. In an environment of high extrinsic production uncertainty, it is difficult for the DoD to distinguish that portion of its information deficit attributable to intrinsic versus extrinsic sources. A high degree of extrinsic uncertainty also means that the DoD�s concerns are multi-dimensional and might encompass design, quality, delivery, as well as price. Several studies point to the necessity of cooperative interaction �reflected in the application of incentive or cost-reimbursement contracts� in the presence of multi-dimensional objectives (see Goldberg 1977, McAfee and McMillan 1988, and Thiel 1988 among others).

Because the DoD�s mission translates into demand for a range of items with innumerable combinations of uncertain exogenous production and consumption parameters, a well-defined set of contract types has developed to manage the persistence of intrinsic uncertainty (see Department of Defense 1996). Because extrinsic uncertainty is neither constant nor persistent, contracting mechanisms adapt to changes in the intensity of extrinsic uncertainty. However, the DoD�s commitment to maintain a qualitative lead over potential threat sources


117

and the desire of contractors to fulfill unmet or unexpressed needs means that cutting-edge items with a high degree of extrinsic uncertainty are continuously generated. Since, the source and behavior of extrinsic uncertainty guide the choice of contracting mechanism, understanding its form is essential. A MODEL OF EXTRINSIC UNCERTAINTY AND CONTRACT TYPE: We proceed with the distinction between intrinsic and extrinsic uncertainty and the recognition that the latter affects producers and consumers differently. The initial objective of the model is to illustrate how the presence of distinct extrinsic production and extrinsic consumption uncertainty combine to influence the selection of contract type. To focus on the effects of extrinsic uncertainties, we assume intrinsic uncertainty to be a random, but persistent, component that varies with the agent. As such, the following is not concerned with the specific design of optimal contract mechanisms, but rather with how extrinsic uncertainty directs the choice between contract types. This assumption permits us to emphasize the differences in extrinsic uncertainty and its impact on the principal and agent.

Let the extrinsic uncertainty in production for a supplier be εp and the extrinsic uncertainty in consumption for the DoD be εc such that both εp and εc are bound between zero and unity. As the degree of uncertainty decreases (increases), εp and εc tend to zero (unity).

To measure the effect of extrinsic uncertainty in production (εp) on the ex ante determination of price, we note that price is composed of estimated costs and a target profit. The producer seeks to maximize profit (π), which Cummins (1977) describes as:

π = ay + x(y-B) [1] where: a = the fixed percentage of the negotiated cost estimate required by the

producer, y = initial total production cost estimate, x = agent�s share of any cost overrun 0< x <1, B = actual cost.

Cross (1968) finds that the percentage profit fee differs among contract types, which implies that a�the after-tax, real marginal rate of return on private investment�differs with the level of extrinsic uncertainty. It follows that the ex ante determination of the profit rate is composed of the normal real rate-of-return possible in an alternative private sector activity (θ) plus a risk premium (δp) associated with the expected return on similar investments (a=θ +δp). Formally, δp = g(εp), where g is a strictly monotonically increasing function of εp.

When extrinsic production uncertainty is negligible: εp tends to 0; δp = 0; y = B; and x = 1. As production uncertainty approaches zero, the return required by the producer or agent is equal to the normal rate of return π=ay=θy. In this case, the DoD engages the non-cooperative interaction between agents to motivate

Groshek and Rose

118

a true revelation of the agents� abilities and preferences. Competitive bidding mechanisms, sealed bids, or alternative Prisoners' Dilemmas might be employed to set the interests of agents against each other to arrive at an outcome favorable to the DoD.

Alternatively, as extrinsic production uncertainty approaches unity (εp !1), objectives are multi-dimensional (McAfee and McMillan 1987a), thereby precluding non-cooperative interaction between principal and agent. The initial rate of profit, determined ex ante during contract negotiations, includes a risk premium (δp>0). Greater levels of extrinsic production uncertainty are accompanied by higher risk-adjusted return requirements and a subsequent profit rate in excess of the normal rate of return.

This relationship is illustrated in Figure 1 with extrinsic uncertainty in production (εp) along the vertical axis below the origin and the corresponding rate of return in excess of the normal rate (δp) measured along the horizontal axis right of the origin. Because the origin represents the normal rate of return (θ), points to the right of the origin represent the risk premiums on production activities when εp>0. The downward sloping curve from the origin reflects the assumption of a risk-averse agent. At the origin, the level of extrinsic uncertainty is zero thereby requiring no premium above the normal rate of return to induce production (g(0)=0). As εp ! 1, the risk premium increases (δp ! ∞). Figure 1: Extrinsic Uncertainty, Premiums, and Feasible Contracts

+

+ +

+

δc

δp

εp

εc

F

F'

cf

pf g

h c'f = h(cf)

(cf , pf)

p'f = g(pf)

(p'f ,c'f)=(g(pf),h(cf))

0

A


119

CONSUMER'S PERSPECTIVE: A corresponding objective function can be established for the DoD. Justification for government programs is found by assigning monetary values to benefits and costs to arrive at a measure of net social welfare (W) such that:

W = v - c [2] v =λc+δc c = ay + (1-x)(y-B) + y

where: v = projected social benefit from project, c = projected cost of project completion, λ c = normal rate-of-social-return for government procurement, δc = consumption premium.

As shown in Equation [2], the net social welfare of a project depends on the effect that extrinsic production uncertainty (the influence of εp on a via δp) has on the cost of project completion (c). Additionally, we posit that a project's net social welfare depends on extrinsic consumption uncertainty (εc) via the project's social benefit (v). A consumption premium (δc) is included to account for the value of externalities resulting from project completion. These effects, such as positive interactive effects (i.e. the continued economic viability of strategic suppliers) or international consequences, increase the value of the project above that reflected in the normal rate-of-social-return on government procurement (λ). The DoD might be willing to pay a consumption premium that reflects this higher value.

The northwest quadrant of Figure 1 shows the degree of extrinsic uncertainty in consumption (εc) relative to the consumption premium that the DoD is willing to pay (δc). Formally, δc = h(εc), where h is a strictly monotonically decreasing function of εc, with h(0) = V0 and h(1) = 0. As extrinsic consumption uncertainty increases, the willingness to pay a consumption premium above normal rate-of-social-return on government procurements decreases.

COMBINED UNCERTAINTIES: A STATIC LOOK AT INITIAL CONTRACTS: The effect of extrinsic uncertainty in production and consumption is demonstrated by combining the relation between εp and δp in the southeast quadrant of Figure 1 with that between εc and δc in the northwest quadrant. The southwest quadrant contains all possible combinations of extrinsic producer and consumer uncertainties leading to the set of possible negotiating environments in which the producer and consumer might find themselves. Moving down (left) from the origin represents increasing extrinsic uncertainty for the producer (DoD). The northeast quadrant shows the image of points in the southwest quadrant under the transformations by {g(εp), h(εc)}. Points to the right of (above) the origin represent increasing risk (consumption) premiums required by the producer (acceptable to the DoD).

Groshek and Rose

120

An upward sloping 45-degree line from the origin (0A) represents combinations of equal producer and consumer premiums with points above (below) corresponding to greater (smaller) premiums acceptable to the DoD relative to that required by the producer. The model combines values for γp and γc and the subsequent relation between δp and δc to illustrate the determination of specific contract types to deal with intrinsic uncertainty. The type of contract that results from various combinations of εp and εc depends on whether the translated point in lies above or below 0A. The initial selection of contracts is bound by extreme levels of extrinsic uncertainty. There exists some level of extrinsic uncertainty in both production pf and consumption cf beyond which each party is unwilling to negotiate. Beyond these points, the producer requires a risk premium (δp!∞) while the DoD recognizes no positive externality to warrant a consumption premium (δc!0). The remaining points in the southwest quadrant are called the feasible region (F) where:

There is a corresponding feasible region F’ in the northeast quadrant which is the image of F under {g(εp), h(εc)} formally defined as:

Each environmental point in F corresponds to a set of risk and consumption premiums, which determines a contract type between the producer and consumer in F’. When mapped into the northeast quadrant, high levels of extrinsic uncertainty result in an inability to conclude a contract. The DoD and its suppliers engage in alternative consumption and production opportunities. When the degree of extrinsic uncertainty for both parties is nonexistent, the point in F’ lies along the vertical axis above 0A. Depending on the value of positive consumption externalities, the DoD may be willing to pay δc =V0, however, it is not necessary because δp = 0. Here, the DoD relies on market competition to alleviate any residual intrinsic uncertainty such as moral hazard. The non-cooperative market interaction between suppliers also ensures that prices reflect true production costs and that non-price objectives are met without the need for monitoring programs to ensure quality control or incentive contracts to motivate production in an economic or efficient manner. As εp increases, there exists a point (p0) up to which the corresponding level δp is insignificantly different from zero with the producer willing to accept complete risk. Therefore, at insignificant levels of εp where δp<δc , a corresponding line k is projected into F’ to designate a region of uni-dimensional, fixed-price contracts that focus on cost minimization and are labeled FP in Figure 2.

( ){ } ,~0,~0:~ , ~p fpfcc pcF ≤≤≤≤= εεεε

( ) ( ) ( )( ){ }=∧=∋∈∃= ppccpcpc ghFF εδεδεεδδ ,, :'


121

Figure 2: Effects of Extrinsic Uncertainty on Contract Type

Conversely, there exists a level (c0) at which consumption uncertainty is very low. Because of the DoD�s well-defined demand and potential for positive externalities, it assumes the bulk, if not all, of the extrinsic production uncertainty in the form of a cost-plus-fixed-fee contract (δc!V0). As with a fixed-price contract, there exists a range delineated by the line l wherein cost-plus-fixed-fee arrangements predominate. This region is designated as C+ in Figure 2. Clearly this type of contract is costly from the perspective of intrinsic uncertainty. Kovacic (1991) finds that the potential for cost maximization is high under cost-plus mechanisms because the government accepts all the risks. As a means to distinguish that portion of cost overruns (B > y in Equation 1) due to extrinsic sources, the DoD might employ monitoring or impose information sharing and ex post cost recuperation requirements on the agent. The remaining region in F’ above 0A reflects the environment discussed most in contracting literature. In this region, the DoD�s objectives are multi-dimensional and interaction between the DoD and its suppliers is cooperative. When the level of extrinsic uncertainty is greater than zero, but less in consumption than in production (c0 < ε c < εp < pf), the resulting contract is of the incentive type. In this area (labeled I), the producer�s risk relative to the DoD�s demand uncertainty is greater. The DoD foresees positive consumption externalities sufficient to offset the risk premiums required by the producers. To motivate production, the DoD must design a contract which transfers extrinsic production risk back to itself. This could be in the form of a greater cost-overrun sharing percentage for the government or higher negotiated profit fee. McAfee and McMillan (1986) find that the optimal contract type in this environment is

+

+

+

+

δc

δp

εp

εc

F

F'

c0

FP

C+ lk

I

RO

h

g

V0

q'

q

p0

Groshek and Rose

122

usually an incentive contract, may be fixed price, and is never cost-plus. Figure 2 bears this out since there are few if any products that match the underlying combinations of extrinsic production and consumption uncertainty to warrant a cost-plus contract. When the level of extrinsic uncertainty is positive, but more in consumption than in production ( p0 < εp < εc < cf), the resulting contract maps below 0A in F’. A cursory analysis might conclude that no agreement and thus no production occurs because the risk premium that the DoD is willing to pay is below the amount the producer requires for production. However, the rapidly growing literature on the use of option models to examine the value of information and timing in non-financial investments suggests otherwise. Specifically, we identify this region as one where real-option mechanisms may allow the postponement of a decision until more information is revealed (Dixit and Pindyck 1994, 1995 and Trigeorgis 1996). In the area labeled RO the DoD seeks concessions from the producer or the producer offers to bear the costs of production uncertainty to maintain the potential for future DoD procurement. Potential producers may offer to provide pilot programs, develop prototypes, or promote dual uses that might sufficiently lower the level of εc. Alternatively, producers may accept losses in initial production contracts to maintain flexibility in anticipation of returns from follow-on contracts. The DoD might also wish to preserve its options by funding follow-on research and development or by maintaining �warm� production lines (operating at less than full capacity). The DoD�s advanced concept technology demonstration (ACTD) is one such real-option program that seeks to shorten the period needed to evaluate unfamiliar products (see Sherman 1998). The real-options type of contract allows for a cooperative reduction in extrinsic uncertainty on both sides with the supplier accepting more risk. In either case the DoD values the flexibility or option to initiate (or expand) consumption at some point in the future. EXTRINSIC UNCERTAINTY AND TEMPORAL CHANGES IN CONTRACT TYPE: The map of contract types that emerges from Figure 2 provides a static picture of the relation between extrinsic uncertainties and contract types. This section introduces a dynamic perspective by considering the effects of temporal shifts in extrinsic uncertainty on changes in contract terms for a specific product. The model provides insight on the evolution of the relationship between a principal (DoD) and an agent (producer) as extrinsic uncertainties change. It also indicates the conditions necessary to move from one contract type to another. The exact point at which shifts occur, however, is a subject for further empirical study. From any point q in Figure 2 such that an incentive contract results at q’ in F’, we can demonstrate the effect on contract type that results from isolated changes in εp and εc. With a given level of εc, the probability of evolving to a fixed price contract (FP) increases as the level of εp decreases (d(q’,k!0


123

=>P(FPf I)!1). Alternatively, should εp increase with a static level of εc, the contract converts to a real-option (d(q’,0A)!0 =>P(ROf I)!1). At extreme levels of extrinsic production uncertainty (εp > pf), failure to conclude a contract results. However, it is evident that a cost-plus arrangement cannot result from point q’ solely due to reductions in εp. By holding εp constant while decreasing εc, the DoD is constrained to provide incentive terms more favorable to the supplier. At some point, this might require the use of a cost-plus-award-fee structure in place of cost-plus-incentive-fee or fixed-price-incentive terms. Alternatively, the contractor might require the DoD to assume a greater share of cost overruns ( x from Equation 1). Ultimately, decreases in εc will result in a cost-plus arrangement (d(q’,l)!0 =>P(C+f I) !1). A sufficient increase in εc (with static εp) acts like an increase in εp (with static εc) with the contract evolving to a real-options arrangement (d(q’,0A) !0=>P(ROf I)!1). Before reaching this point the terms of the incentive contract become less favorable to the supplier. Similar changes to contract type can be demonstrated from any point in F’ by adjusting εp, εc, or both. Figure 3: Post-Cold War Effects on Contract Type

The end of the cold-war had the effect of shifting the DoD�s perspective towards consumption risk. As demonstrated in Figure 3 by the movement from h to h*, the potential for positive externalities and, therefore, the DoD�s willingness to pay a consumption premium has declined at all levels of extrinsic consumption uncertainty. The cold-war combination of extrinsic uncertainties at point r resulted in an incentive contract at r’. However, the post-cold war realities, given constant εc and εp, results in movement to a real-options arrangement at point r”.

+

+

+

+

δc

δp

εp

εc

h

F'FP

C+ lk A

0

r

r'

r'' RO

I

h*

F

g

+

Groshek and Rose

124

Contracts previously under real-options arrangements might be canceled in the post-cold war environment while cost-plus fixed fee arrangements would tend to disappear. It should be noted, however, that it is not possible to impose a fixed-price contract on the supplier merely due to the shift in h. As demonstrated above, fixed-price contracts require a sufficiently small degree of εp. Absent a decline in here, the DoD might find little success in emphasizing the greater use of fixed-price contracts with its suppliers. CONCLUSION: Contracting in the Department of Defense continues to be the subject of considerable interest. The amount of money involved and the peculiar relationship between the DoD and its suppliers have motivated the establishment of a variety of contract mechanisms. The existing literature on this relationship does not address the important role played by environmental uncertainties. This paper demonstrated the initial determination of contract type and terms as the result of the degree of extrinsic uncertainty faced by both contract participants. The model relies on a clear distinction between the consumption and production sources of uncertainty. The use of non-cooperative, competitive contracting mechanisms is feasible in the absence of extrinsic uncertainty where the DoD concentrates on uni-dimensional, cost minimization objectives. When extrinsic uncertainty is present, objectives are multi-dimensional and the use of cooperative contracting mechanisms is necessary to manage the persistence of intrinsic uncertainty. As the level of extrinsic consumption and production uncertainties shifts between the DoD and its suppliers, contract terms and type change. As extrinsic uncertainty in production decreases relative to that in consumption, the result will be an incentive contract more favorable to the DoD. The opposite occurs when uncertainty in consumption declines by more than the extrinsic uncertainty in production. The impact is notable when considering recent changes in defense demand. The post-cold war environment has reduced the DoD�s requirement for a number of defense items. In turn, the DoD recognizes fewer positive consumption externalities resulting from the acquisition of new systems and is less willing to pay a premium toward their production. Cost-plus-incentive contracts are no longer warranted, and many programs once governed by incentive contracts are now of the real-option type. However, this shift towards the horizontal axis in our graphical model of uncertainty does not mean that the remaining incentive contracts can be recast within a fixed-price type. Caution should prevail in seeking cost reductions through the application of uni-dimensional contracts and non-cooperative procedures in an environment of continued significant extrinsic production uncertainty.


125

REFERENCES Anton, J., and Yao, D. (1990). Measuring the effectiveness of competition in

defense procurement: a survey of the empirical literature. Journal of Policy Analysis and Management, 9(1), 60-79.

Austin, R., and Larkey, P. (1992). The unintended consequences of micromanagement: the case study of procuring mission critical computer resources. Policy Sciences, 25(1), 3-28.

Blackburn, K., and Christensen, M. (1989). Monetary policy and policy credibility: theories and evidence. Journal of Economic Literature, March, 1-45.

Bos, D., and Lulfesmann, C. (1996) The Hold-up Problem in Government Contracting. Scandinavian Journal of Economics, 98(1), 53-74.

Cross, J. (1968). A reappraisal of cost incentives in defense contracts, Western Economic Journal, June, 205-225.

Cummins, J.M. (1977). Incentive contracting for national defense: the problem of optimal risk sharing, Bell Journal of Economics, 168-184

Cyert, R., and March, J. (1963). A Behavioral Theory of the Firm, Englewood Cliffs, Prentice Hall.

De Mayo, P. (1983). Bidding on new ship construction, in Ehglebrecht-Wiggans, R., et.al., Auction, Bidding, and Contracting, New York, New York University Press.

Department of Defense. (1996). DFARS-Part 216, Types of Contracts, DoD. Dixit, A.K. and Pindyck ,R.S. (1994). Investment under Uncertainty, Princeton

University Press, Princeton, N.J. Dixit, A.K. and Pindyck ,R.S. (1995). The options approach to capital

investment. Harvard Business Review, May-June 1995, pp.105-115. Gansler, J. (1989a). Affording Defense, Cambridge, MA, MIT Press. Gansler, J. (1989b). Affording defense: the changes that are needed. National

Contract Management Journal, 23(1), 1-22. Goldberg, V. (1977). Competitive bidding and the production of precontract

Information. Bell Journal of Economics, 250-61. Hartley, K. (1991). The Economics of Defense Policy, London, Brasseys. Kovacic, W. (1991). Commitment in regulation: defense contracting and

extensions to price caps. Journal of Regulatory Economics, 3(3), 219-40. Kovacic, W. (1992), Regulatory controls as barriers to entry in government

procurement. Policy Sciences, 25, 29-42. Laffont, J. (1986). Towards a normative theory of incentive contracts between

government and private firms. Economic Journal, Supmnt. 97, 17-31. McAfee, R.P., and McMillan, J. (1986). Bidding for contracts: a principal-agent

analysis. Rand Journal of Economics, 17(3), 326-338. McAfee, R.P., and McMillan, J. (1987a). Competition for agency contacts.

Rand Journal of Economics, 18(2), 269-307. McAfee, R.P., and McMillan, J. (1987b). Auctions and bidding. Journal of

Economic Literature, June, 699-738.

Groshek and Rose

126

McAfee, R.P., and McMillan, J. (1988). Multi-dimensional incentive compatibility and mechanism design. Journal of Economic Theory, 46, 335-54.

Myerson, R. (1981). Optimal auction design. Mathematics of Operations Research, 6, 58-73.

Sandler, T., and Hartley, K. (1995). Procurement: theories, evidence, and policies, Chapter in The Economics of Defense, Cambridge University Press, 113-155.

Sherman, J. (1998). Tally up. Armed Forces Journal, July, 20-1. Thiel, S. (1988). Multi-dimensional auctions. Economics Letters, 28, 37-40. Trigeorgis, L. (1996). Real Options: Managerial Flexibility and Strategy in

Resource Allocation. Cambridge, MA: The MIT Press. Turpin, C. (1989). Government Procurement and Contracts. London, Longman.


127

EXPLAINING CUSTOMER RESPONSES TO SERVICE RECOVERY: AN ASSESSMENT OF THE

PSYCHOMETRIC PROPERTIES AND PREDICTIVE VALIDITY OF MEASUREMENT SCALES FOR EXCHANGE-THEORETICAL CONSTRUCTS

Manfred F. Maute

University of Northern British Columbia William R. Forrester, Jr.

Kennesaw State University ABSTRACT: Even though service failure and recovery are widely regarded as leading causes of customer defection, relatively little is known about how customers decide to maintain or terminate a service relationship once a service failure has occurred. This study uses the investment model of ongoing relationships to advance a theoretical explanation of how service recovery affects customer responses to service failure. Measures are developed for exchange-theoretical constructs and the predictive validity of the model is tested. Results indicate that measurement scales exhibited acceptable psychometric properties. A test of the model using the new measures affirmed relationships between exchange-theoretical constructs such as current relationship rewards, costs, and investments and service recovery satisfaction and customer defection. INTRODUCTION: Research increasingly suggests that customers evaluate service failures and recovery efforts very critically (Bitner et al, 1990; Tax et al, 1998). Keveaney�s critical incident study (1995) identified service failure and recovery as two leading causes of customer decisions to switch service providers. Empirical research has begun to focus attention on how customers respond to service failures (see Johnson, 1995; Levesque & McDougall; 2000). Nevertheless, Smith et al (1999) have argued that new theoretical frameworks are required if the study of service recovery is to move beyond the identification of effective recovery strategies to examine the underlying �psychological mechanism� that leads customers to maintain or terminate service relationships once a service failure has occurred. In this paper, we report the results of a study that utilizes the investment model of ongoing relationships to develop a theoretical explanation of how service recovery acts on exchange-theoretical constructs such as relationship rewards, costs, and investments to influence customer responses to service failure. Measures for exchange-theoretical constructs are developed, the psychometric properties of these measures are assessed, and the predictive validity of the investment model is evaluated by empirically testing for the effects of relationship rewards, costs, investments, and alternative relationship value on service recovery satisfaction and customer defection.

Maute and Forrester

128

BACKGROUND: Service Failure Recovery: Service failures are challenging events for service providers, exacerbating the potential for customer defection on one hand, while creating opportunities to restore satisfaction and loyalty on the other. In a study of service complaint handling, Tax at al. (1998) found that most complaining customers were dissatisfied with their complaint experience. Keveaney (1995) examined 838 switching episodes and identified service failure as the leading cause of customer defections. While service failures can become flash points that give rise to switching, customer satisfaction and retention can be just as adversely affected by a shoddy recovery effort. Ineffective service recovery was cited by slightly more than one third of Keveaney�s respondents as the motivation for switching service providers. Fortunately for service firms, well-executed recoveries not only restore, but also enhance customer satisfaction and retention. Tax et al (1998) concluded that the manner in which customer complaints were resolved was positively related to service provider trust and commitment. Spreng et al (1995) studied an interstate moving service and concluded that service recovery satisfaction had a larger positive effect on customer satisfaction and retention than the negative effect of the service failure. The Investment Model: The investment model of ongoing relationships is introduced as an alternative theoretical framework to explain service recovery effects on customer responses to service failure. According to the model, individuals participate in relationships to maximize personal outcomes and exchange-theoretical constructs such as rewards, costs, and investments are key determinants of behavior in relationship dyads (see Kelley &Thibault, 1978). The value of a relationship is determined by relationship rewards and costs relative to a comparison level, with the comparison level representing what individuals have come to expect from relationships generally (Rusbult, 1980). The value of a relationship increases as rewards increase, costs decrease, or the comparison level decreases (Rusbult et al, 1982).

Satisfaction refers to the positive affect associated with a service relationship and is theorized to be a function of current relationship value (Rusbult, 1980). Relationship commitment comprises the intention to maintain the relationship and is affected not only by current relationship value, but also by relationship investments (e.g., resources �put into� the relationship) and the value of the best alternative relationship (e.g., rewards less costs for the best alternative partner) (Rusbult et al, 1982). Higher relationship value, larger investments, and lower alternative relationship value increase commitment (Rusbult et al, 1986; Rusbult et al., 1988).

Customer Responses to Service Failure: Our explanation of the

psychological mechanism that translates service recovery effects into customer decisions to stay or leave service relationships is based on interpersonal and organizational relationship research (see Rusbult, 1980; Allen, 1984; Rusbult et al 1982; Rosse, 1988).


129

Studies have found that exchange-theoretical constructs such as relationship rewards, costs, and investments predict not only relationship commitment, but also behavioral responses to decline such as exit, voice, and loyalty (Rusbult et al, 1982; Rusbult et al, 1988). Although service relationships are generally shorter in duration, more limited in scope, and easier to terminate than interpersonal and organizational relationships, the shift from a transactional to a relational view of marketing (see Morgan and Hunt, 1994; Gwinner et al, 1998) implies that the investment model may offer new insights into how service recovery affects customer retention in the aftermath of a service failure.

We consider the effects of exchange-theoretical constructs on two outcomes that bear directly on the ability of service marketers to retain customers. Service recovery satisfaction is an affective judgment about the degree to which a recovery meets or exceeds customer expectations (Smith et al., 1999). Customer defection, also referred to as exit by Hirschman (1970, p. 29), occurs when individuals �disassociate themselves from the object of their dissatisfaction� and is manifested in service relationships when customers switch suppliers. METHODS: Service recovery research is problematic because service failures are difficult to predict and the factors that affect customer responses are numerous and complex. A service recovery experiment, conducted by survey, was chosen to generate a representative set of recovery events, eliminate the time, managerial, and ethical, problems of observing or enacting service failures in field settings, and reduce the potential for retrospective, self-report bias (see Smith et al, 1999). Subjects and Procedure: The study was presented to subjects as a survey on customer satisfaction with airlines. Eighty-two business students who rated themselves as experienced airline travelers and well-informed consumers of airline services participated for course credit. After reading a description of a service transaction in which the focal airline canceled a flight due to a technical problem, participants were exposed to four different recovery scenarios (e.g., high and low customer loyalty strategies crossed with high and low service recovery strategies). Participants were then instructed to assume the role of the protagonist, a passenger much like themselves who was denied the airline service, and asked to respond to a series of questions about the event (see Appendix 1 for manipulations). Service recovery satisfaction and likelihood of customer defection were measured with a series of multi-item scales adapted to an airline service context from Rusbult (1980). Subject indicated their degree of satisfaction with the airline�s handling of the problem with seven-point semantic differential scales anchored by four pairs of descriptors (displeased � pleased, satisfied � dissatisfied, unhappy � happy, and does a good job, does a poor job). Subjects were asked to indicate the likelihood that they would respond to the service failure with specific behaviors on a seven-point scale (�Very Unlikely��Very Likely�). Defection items included �I would fly with another airline next time� (reverse coded); �I would be reluctant to fly with (the focal airline) again�; and �I would not fly with (the focal airline) again�. Subjects indicated their perceptions of

Maute and Forrester

130

relationship rewards and costs, relationship investments, and rewards from alternative relationships using 7 point Likert type agree-disagree scales. Cumulative results from all four experimental conditions are reported here as a reflection of the variety of recovery strategies that customers could encounter in actual service failure situations. RESULTS: Psychometric Properties of Measurement Scales: Measures for relationship rewards, costs, investments, and alternative relationship value were developed according to Churchill�s measurement paradigm (1979). Representative items were identified from the interpersonal relationship literature and in discussions with experienced airline travelers (refer to Appendix 2 for scale items). Measures were purified by subjecting the scale to exploratory factor analysis and assessing the degree of association among items comprising the rewards, costs, investments, and alternative relationship value sub-scales. Table 1 presents the results of a principal components analysis of the sixteen-item scale. Four components with eigenvalues greater than one accounted for 74 percent of the variation in the scale. Items loaded strongly and predictably on current relationship rewards (component 1) and alternative relationship value (component 2). However, neither the current relationship costs, nor the current relationship investments sub-scales were found to be one-dimensional. On the basis of further analysis, two items that loaded on several components were removed from the cost and investment sub-scales. TABLE 1: PRINCIPAL COMPONENTS ANALYSIS OF EXCHANGE-THEORETICAL CONSTRUCTS ___________________________________________________________________________ Exchange-Theoretical Principal Components1

Constructs 1 2 3 4 ____________________________________________________________________________ Current Relationship Rewards

Q1 .910 .116 .174 .473 Q2 .895 .266 .175 .359 Q3 .846 .240 .186 .344 Q4 .901 .104 .182 .427

Current Relationship Costs Q5 .377 .182 .106 .760

Q6 .348 .255 .854 Current Relationship Investments

Q9 .353 .108 .668 Q10 .153 .103 .880

Alternative Relationship Rewards Q13 .122 .933 .118 Q14 .147 .925 .113 Q15 .306 .796 .285 Q16 .173 .922 .210

_______________________________________________________________ 1 Factor loadings less than .10 are not reported.


131

The second step in measure purification involved an evaluation of the degree of association between items comprising the four sub-scales. Table 2 summarizes the inter-item and item to total correlations for the relationship rewards, costs, investments, and alternative relationship value sub-scales. Strong and statistically significant intercorrelations among items and between individual items and summated values were observed for each sub-scale. The internal stability of the sub-scales as represented by coefficient α was found to satisfy acceptable levels for exploratory research (see Nunnally, 1977). TABLE 2: MEANS, INTER-ITEM, AND ITEM TO TOTAL CORRELATIONS FOR EXCHANGE-THEORETICAL CONSTRUCTS _________________________________________________________________ Inter-Item Item to Total Coefficient

Correlations Means Correlations Alpha ______________________________________________________________________________ Current Relationship Rewards 3.32 (1.41)1 .926

Q1 3.50 (1.51) .906 Q2 .827 3.36 (1.60) .936 Q3 .665 .752 3.09 (1.53) .862 Q4 .789 .801 .701 3.37 (1.61) .912

Current Relationship Costs 4.26 (1.22) .698

Q5 4.34 (1.57) .706 Q6 .513 4.57 (1.43) .770

Current Relationship Investments 3.88 (1.37) .602

Q9 4.09 (1.62) .842 Q10 .424 3.67 (1.64) .846

Alternative Relationship Value 4.59 (1.23) .923

Q13 4.58 (1.40) .925 Q14 .889 4.70 (1.34) .920 Q15 .636 .632 4.40 (1.39) .835 Q16 .810 .802 .742 4.68 (1.34) .929

________________________________________________________________________ 1 Standard deviation. Maximally different measures of exchange-theoretical constructs were not obtained, so it was not possible to assess convergent validity. Discriminant validity was examined by computing correlations between similar measures of different constructs. Low to moderate correlations ranging from .006 (p > .957) (investments and alternative relationship value) up to -.450 (p < .01) (current relationship rewards and costs) were revealed by this analysis. The absence of consistently strong correlations indicated that the four exchange-theoretical constructs were distinct. Predictive Validity of Exchange-Theoretical Constructs: Service recovery satisfaction and customer defection were regressed on current relationship

Maute and Forrester

132

rewards, costs, investments, and alternative relationship value in order to assess the predictive validity of the investment model. Results are presented in Table 3. TABLE 3: REGRESSION OF SERVICE RECOVERY SATISFACTION AND CUSTOMER DEFECTION ON EXCHANGE-THEORETICAL CONSTRUCTS ______________________________________________________________ Exchange-Theoretical Standardized t Value F Value Significance R Square Predictors Coefficient ________________________________________________________________ Service Recovery Satisfaction 27.12 .000 .608 CR1 Rewards .791 8.80 .000 CR Costs -.045 -0.52 .604 CR Investments -.079 - 0.99 .324 AR2 Value .059 0.74 .459 Customer Defection 13.85 .000 .453 CR Rewards -.452 -4.21 .000 CR Costs .263 2.52 .014 CR Investments -.195 -2.04 .046 AR Value -.034 -0.36 .720 _________________________________________________________________________ 1 Relationship with current service provider. 2 Relationship with alternative service provider.

Exchange-theoretical variables were significant predictors of customer responses to service failure, accounting for 60 percent and 45 percent of the variation in service recovery satisfaction (F=27.12, p < .000) and customer defection (F=13.85, p < .000), respectively. Current relationship rewards (t = 8.80, p < .000) increased service recovery satisfaction, but none of the other relationship variables was a statistically significant predictor of this service recovery response. Customer defection (t= -4.21, p < 000) was negatively related to current relationship rewards (t= -4.21, p < 000) and current relationship investments (t=-2.04, p < .046), and positively associated with relationship costs (t=2.52, p < .014). Alternative relationship value was not significantly related to customer defection. DISCUSSION: Psychometric Properties of Measures: Results suggest that although exchange-theoretical measures exhibited generally acceptable psychometric properties, the effort to translate relationship cost and investment items from an interpersonal to a service relationship context was not completely satisfactory (see Rusbult et al., 1982 for examples of original items). Items that probed for costs associated with relationship uncertainty and asked for a comparative assessment of the negative and positive aspects of the exchange relationship without referring to specific service failure costs loaded on several components. Similar results were achieved with investment items that referred to �sacrifices� made to ensure the continuation of the exchange relationship and asked participants to reflect on how their reluctance to


133

terminate the relationship. Concerns about the dimensionality of the cost and investment sub-scales, perhaps exacerbated by the relatively small sample size, led to the removal of these items from each sub-scale. In elaborating these scales further, it appears that new items tapping specific rather than general relationship costs and investments will be required to improve the measures. Predictive Validity of the Investment Model: Empirical relationships between exchange-theoretical constructs and service recovery responses were affirmed. These results provide support for the presence of a psychological mechanism that translates service recovery into background evaluations of relationship rewards, costs, and investments that subsequently influence customer decisions to maintain or terminate a service relationship. Among the exchange-theoretical variables, current relationship rewards was disproportionately important as a predictor of service recovery satisfaction and customer defection, increasing the former and decreasing the latter.

Only relationship rewards was significantly related to increased service recovery satisfaction. Since relationship rewards and costs determine current relationship value, the fact that relationship costs was not associated with recovery satisfaction was somewhat surprising. A possible explanation for this result is that a hypothetical service failure of the type and magnitude depicted in the recovery scenarios is more effective at imposing non-financial (e.g., emotional frustration, anxiety) than financial costs on study participants. Customers may be prepared to accept relationship rewards as a financial offset for service failure costs that are not easily mitigated or reimbursed by the service provider. Consistent with prior interpersonal and organizational relationship research, relationship rewards and investments decreased defection, while relationship costs increased the likelihood that customers would end the relationship with the current service provider. Much as was the case for recovery satisfaction, relationship rewards was the dominant predictor as reflected in a standardized regression coefficient approximately twice the magnitude of other relationship variables. In this instance, however, the positive effect of relationship rewards on defection was counterbalanced by the negative effect of relationship costs to ensure that current relationship value reflected an accounting of both the benefits and costs of continuing the service relationship. The magnitude of relationship rewards and costs effects, together with the lack of any significant association between alternative relationship value and customer defection, implies that current relationship value drives customer decisions to maintain or end a faltering relationship. Specifically, the stay/go decision seems to be confined to an assessment of the absolute value that the current service relationship is capable of delivering, rather than a comparative evaluation of how that value equates with what could be obtained from an alternative service relationship. If, as these results suggest, the stay/go decision is based on an absolute judgment of the value of the current service relationship, service providers would seem to be in more control of their futures than many would have dared hope, and have the means

Maute and Forrester

134

to shape customer perceptions of that value by selecting service recoveries that correctly leverage relationship rewards, costs, and investments. CONCLUSIONS AND FUTURE RESEARCH: This study responds to the need for research that explains the psychological mechanism by which service recovery effects are translated into customer decisions to maintain or terminate service relationships in the aftermath of a service failure. The investment model of ongoing relationships was invoked to explain how background exchange-theoretical variables influence service recovery satisfaction and customer defection, measurement scales for the exchange-theoretical constructs were developed, and the predictive validity of model was affirmed. In the brief discussion that follows, we consider the managerial implications of these results and promising directions for future research. When crafting service recoveries, it appears that managers should concentrate their efforts on strengthening perceptions of the value of the current service relationship rather than focusing on defeating the appeal of alternative service providers. In so far as the goal of recovery is to influence customer judgments about service recovery satisfaction, relationship rewards seem to be the decisive factor. Rewards can take many forms, and when offerings from complementary service providers are included, the scope for service providers to create innovative and highly effective relationship rewards is almost unlimited. Relationship rewards also play an important role when the goal of service recovery is to deter customer defection, but in this instance, not to the exclusion of relationship costs and investments. Since recoveries will vary in their capacity to manipulate the way customers perceive service relationships, among the challenges facing managers is conceiving and executing recovery interventions that leverage those rewards, costs, and investments that shape perceptions of relationship value most effectively for those groups of customers most at risk to defect. Service providers seem unlikely to have this information at hand, so experimentation and market research may be prerequisites to the confident execution of effective, targeted recovery interventions. Future research should seek to improve the relationship cost and relationship investment measures and expand the generalizability of results to other populations, service industries, and recovery contexts. Research that helps to more fully explain how the psychological mechanism works would also be timely. For example, modeling exchange-theoretical constructs as mediators of recovery strategy effects on service failure responses may expose more fully how the psychological mechanism unfolds. With deeper knowledge of the mediational process in hand, applied research can begin to assess the effectiveness of different recovery strategies in varying industry, market, and competitive conditions. This will permit service managers to make informed decisions about the relationship rewards, costs, and investments to leverage with different service recovery strategies to achieve maximum effect on customer retention.


135

REFERENCES Allen, S. G. (1984). Trade Unions, Absenteeism, and Exit-Voice, Industrial and

Labour Relations Review, 27, 331-345. Bitner, M. J., B. H. Booms & M. S. Tetreault (1990). The Service Encounter: Diagnosing Favorable and Unfavorable Incidents, Journal of Marketing, 54, 71-84. Bolton, R. N., P. K. Kannan & M. D. Bramlett (1998). Implications of Loyalty Program Membership and Service Experiences for Customer Retention and Value, Academy of Marketing. Journal, 28(1), 95-108. Churchill, Gilbert A. (1979), A Paradigm for Developing Better Measures of MarketingConstructs, Journal of Marketing Research, Vol. 16(February), 64-73. Gwinner, Kevin P., Dwayne D. Gremler & May Jo Bitner (1998), Relational Benefits in Services Industries: The Customers� Perspective, Academy of Marketing Science. Journal, 28(2), 101-114. Hirschman, A. O. (1970). Exit, Voice and Loyalty: Response to Declines in Firms, Organizations, and States. Cambridge, Mass.: Harvard University Press. Johnston, Robert (1995), Service Failure and Recovery: Impact, Attributes, and Process. In Advances in Services Marketing and Management, Teresa A. Schwartz, David E. Bowen, and Stephen W. Brown, eds., vol. 4, Greenwich, CT: JAI, 211-28. Keveaney, S. M. (1995). Customer Switching Behavior in Service Industries: An Exploratory Study, Journal of Marketing, 59(2), 71-83. Kelley, H. H., & J. W. Thibaut (1978). Interpersonal Relations: A Theory of Interdependence. New York: Wiley. Levesque, T. J. and G. H.G. McDougall (2000). Service Problems and Recovery Strategies: An Experiment, Canadian Journal of Administrative Sciences, 17(1), 20- 37. Morgan, Robert W. & Shelby D. Hunt (1994). The Commitment-Trust Theory of Relationship Marketing, Journal of Marketing, 58, 20-38. Nunnally, J.C. (1977). Psychometric Theory, Second Edition, New York: McGraw- Hill. Book Company. Rosse, J. G. (1988). Relations Among Lateness, Absence, and Turnover: Is There a Progression of Withdrawl? Human Relations, 41, 517-531. Rusbult, C. E. (1980). Commitment and Satisfaction in Romantic Associations: A Test of the Investment Model, Journal of Experimental Social Psychology, 16, 172-186. __________, I. M. Zembrodt & L. K. Gunn (1982). Exit, Voice, Loyalty, and Neglect: Responses to Dissatisfaction in Romantic Involvements, Journal of Personality and Social Psychology, 43, 1230-1242. ___________, D. J. Johnson & G. D. Morrow (1986). Determinants and Consequences of Exit, Voice, Loyalty, and Neglect in Adult Romantic Involvements,� Human Relations, 39, 45-63. ___________, D. Farrell, G. Rogers & A. G. Mainous III (1988). Impact of Exchange Variables on Exit, Voice, Loyalty, and Neglect: An Integrative

Maute and Forrester

136

Model of Responses to Declining Job Satisfaction, Academy of Management Journal, 31, 599-627. Smith, A., R. M. Bolton & J. Wagner (1999). A Model of Customer Satisfaction with Service Encounters Involving Failure and Recovery, Journal of Marketing Research, Vol. 26 (August), 356-372. Spreng, R. A., G.D. Harrell & R.D. Mackoy (1995), Service Recovery: Impact on Satisfaction and Intentions, Journal of Services Marketing, 9(1), 15-23. Tax, S. S., S. W. Brown & M. Chandrashekaran (1998). Customer Evaluations of Service Complaint Experiences: Implications for Relationship Marketing, Journal of Marketing, 62(2), 60-77. APPENDIX 1: SERVICE RECOVERY AND CUSTOMER LOYALTY MANIPULATIONS Low Service Recovery Condition: Arriving at the airport to check in for a flight, you notice confusion at the focal airline check-in counter. When an airline employee finally appears, you are informed that the flight has been canceled because of a mechanical problem with the aircraft. The employee does not apologize for the cancellation and tells you he busy and can�t explain the nature of the problem to every single passenger. He advises you that it is against the focal airline's policy to compensate passengers for canceled flights. The employee suggests that you call back periodically to determine if the problem has been resolved and to book a seat on a later flight, if necessary.

High Service Recovery: Arriving at the airport to check in for a flight, you notice considerable activity at the focal airline's check-in counter. Almost immediately, you are approached by a focal airline employee who apologizes for the inconvenience and explains that the flight has been cancelled because of a mechanical problem with the aircraft. After describing what steps the airline is taking to correct the problem, the employee escorts you to a check-in counter where a ticketing agent gives you a meal and taxi voucher. The employee takes your telephone number and promises to call as soon as the mechanical problem is resolved. S/he offers to book a seat for you on the first available flight. Low Customer Loyalty Condition: Although you have flown with the focal airline infrequently, your experience has been generally dissatisfactory. The airline is frequently short staffed. Even simple things like getting fare and schedule information over the telephone can involve a ten-minute wait. Much the same pattern was repeated during flights. Flight attendants always seem to be very busy and have little time to make passengers comfortable and attend to special needs. The focal airline offers few incentives to loyal customers. Since the frequent flier program isn�t very rewarding, you never bothered to join. High Customer Loyalty: You have flown with the focal airline often and the experience has generally been very positive. Ticket agents were friendly and


137

courteous and promptly answered telephone inquiries about fares and schedules. Much the same pattern was repeated during flights. Flight attendants always seemed to go out of their way to make sure everyone was comfortable and provide for the special needs of passengers. The focal airline offers numerous incentives to loyal customers, including a generous frequent flier rewards program. You have accumulated almost enough miles to qualify for a free trip. APPENDIX 2: MEASUREMENT ITEMS COMPRISING THE EXCHANGE-THEORETICAL SUBSCALES Current Relationship Rewards Q1 You can trust the focal airline. Q2 The focal airline has done a good job of meeting your needs as a customer. Q3 It is rewarding to be a loyal customer of the focal airline. Q4 The focal airline is committed to your satisfaction as a customer. Current Relationship Costs Q5 You can�t count on the focal airline to keeps its promises. Q6 Your relationship with the focal airline has not always been smooth. Q7 You never know what to expect when you fly with the focal airline. Q8 Your relationship with the focal airline has more negative than positive spects. Current Relationship Investments Q9 Your relationship with the focal airline is longstanding. Q10 You have invested a lot in your relationship with the focal airline. Q11 Maintaining your relationship with the focal airline has involved sacrifices. Q12 You would think twice before ending your relationship with the focal airline. Alternative Relationship Value Q13 You can trust alternative airlines. Q14 Alternative airline would do a good job of meeting customer needs Q15 It could be rewarding to be a customer of an alternative airline. Q16 Alternative airlines are committed to your satisfaction as a customer.

138

CUSTOMER RELATIONSHIP MANAGEMENT AND CONSUMER CORRESPONDENCE:

A THEORETICAL APPROACH

Michael McCall Donald W. Eckrich

Ithaca College ABSTRACT: This paper examines the results of a field experiment where 120 major consumer-goods manufacturers received pairs of written communications associated with positive and negative consumer correspondence. The results are examined in the context of gain-and-loss theory and offer promising new directions for future relationship marketing research. Several implications for further study using the new theoretical framework are explored. INTRODUCTION: The notion that marketing is about relationships with customers is axiomatic. Indeed, probably one of the most frequently discussed topics into the new millennium has been the notion of relationship marketing (Baker, 1998; Petrof, 1997; Gruen , 1997; Evans & Laskin, 1994; Reichheld, 1993; Copulsky & Wolf, 1990). The most recent extensions have involved discussions of one-to-one marketing, customer relationship management (CRM), enterprise marketing automation (EMA), and even smart-marketing (Bresnahan, 1999). What is far less obvious is the manner in which marketing relationships should be managed to maximize their long-term potential. Actually preventing its premature death is the concern of one of the most referenced articles (Fournier, Dobscha, Mick, 1998). From a theoretical standpoint, consumer behavior textbooks discuss the notion of building relationships as a form of customer exchange (Vavra, 1992), often referring to the relationship process as �marriage-like.� For instance, Mowen & Minor (2001) suggest that brand loyalty and repeat purchases represent an ongoing relationship that serves as �the fundamental axiom of relationship marketing� (see also Sheth & Parvatiyar, 1995). Note however that the relationship between the customer and supplier can, like human interpersonal relationships, be harmed from a variety of sources, thereby leading to dissatisfaction. In the present paper we examine how a number of major companies respond to customer satisfaction and dissatisfaction within the context of relationship maintenance. Specifically, we consider organizational responses to written consumer compliments and complaints (i.e. correspondence handling). Like relationship marketing, methods and practices associated with customer correspondence handling have a long history of development in the marketing literature. Most prominent in this regard has been the focus on complaining behavior (i.e., Best & Andreasen, 1977 is the pioneering work in the field; Singh & Wilkes, 1996 contains an excellent literature review). Alternatively, fewer


139

studies have examined and compared organizational responses to complimentary correspondence (Pearsen, 1976; Martin & Smart, 1989; 1988). Among the conclusions, however, it has been noted that the theoretical connections between relationship marketing and both types of correspondence handling remain underdeveloped, leaving both gaps in the literature and opportunities for further study (Erickson & Eckrich, 2001). Thus, given the interest and capacity to consider buyer-customer interactions in human relationship terms we propose to examine these issues within the theoretical framework of gain-and-loss theory (Aronson & Linder, 1965). THEORETICAL DEVELOPMENT: Gain-and-loss theory proposes that those interpersonal interactions in which the outcome can be interpreted as a net gain or net loss will be valued more than those interactions in which the outcome maintains a status quo. For instance, assume that over a period of time your interactions with an attractive member of the opposite sex remain positive. According to traditional theories of liking, a consistently positive set of interactions should produce the greatest amount of liking. Imagine, however, an alternative possibility. Assume that the first few interactions with this attractive opposite-sex individual were somewhat negative, yet over time the interactions became more and more positive until they reached the same level as in the always positive condition. According to gain-and-loss theory, it is this second case in which you would be most attracted to the target. Indeed, Aronson and Linder (1965), demonstrated that feelings for another individual were more affected by an initial negative then positive sequence over the �always positive.� The logic of this argument is actually quite compelling. When negative impressions become positive over time the individual actually gains. Thus, a gain is more valuable than any alternative. Had the attractive other always felt positive towards us, their continued good will is important, yet to gain the approval of a once negative individual is preferable. Similarly the reverse is also true. That is, Aronson and Linder (1965) were also able to demonstrate the deleterious effects of a positive evaluation becoming negative. When another holds us in high regard and that opinion becomes negative, we suffer greater loss, than when the opinions of another are always negative. Here again, according to the gain-loss theory it is the loss that is felt more strongly than an evaluation that is always negative. Within the context of human relationships, Aronson and Linder (1965) argue that relationship maintenance can be best understood within the economic framework of gains and losses. The implications of gain-and-loss theory for customer relationship management present an intriguing theoretical framework. Indeed, assuming some degree of loyalty (or at least positive affinity) among the customer base, the potential loss of a loyal customer should be felt more severely than an expression of continued loyalty. In gain-loss theory terms, the negativity obviously associated with complaints should represent a more troublesome situation than the

McCall and Eckrich

continuing positive message associated with compliments. Stated simply, losses and gains would have more of a direct impact and therefore require a greater response. It is therefore predicted that customers who complain and are thought to be �loyals� would receive a more significant, timely, and favorable response than customers who express continued loyalty through compliments. METHOD: As part of a classroom assignment teams of two students were instructed to identify a �consumable product� (Powers, 1993), which had recently left them dissatisfied. Consumable products have been defined as typical low-risk, grocery and over-the-counter pharmaceutical products with relatively short repeat-purchase cycles, such as beverages, health and beauty aids, snack foods, etc. One member of the team was subsequently required to contact the consumer affairs department (or other appropriate organizational contact) at the company responsible in order to communicate their negative experience. Correspondingly, the other team member was required to send a letter of praise to the same contact regarding the same product. In all cases, to ensure that letters were never connected and thereby maintain the integrity of the experimental treatment, participant return addresses were required to be no closer than 50 miles apart. Thus, parallel outgoing letters were constructed to represent a complaint and a countervailing compliment regarding a specific consumable product. All correspondence was both judged and double-checked by independent observers to ensure suitability and general comparability (e.g. identical addresses, different stationery and fonts, etc.). In all, 522 letters were sent to 120 different companies representing 261 dyads (i.e. pairs of parallel letters). Neither the compliment nor the complaint letter requested a specific response, and no questions were asked, implied or otherwise, in the letter. Companies whose products were regularly chosen included Nabisco, Kelloggs, Kraft, Gillette, Lipton, Hershey, Proctor & Gamble, and Pepsico. See Table 1 for a list of the most frequently represented companies.

TABLE 1 Company Number of Dyads

Represented Nabisco 11 Kellogg 10 Kraft 8 Gillette 6 Lipton 5 Hershey 5 M&M Mars 5 P&G 5 Pepsico 5

Examples of products and their corresponding complaints included such things as �too few raisins in the Kellogg�s Raisin Bran cereal, a skin irritation from a Colgate-Palmolive Speed Stick deodorant, and Kraft Macaroni and Cheese


141

not containing enough cheese. Data were considered complete if responses to both letters were received. In all, 67% of the dyads were completed (N=175); 16% of the responses were for complaints only, 11% responded to the compliment only, and 7% of the companies failed to respond to either letter. Comparatively, these results match those of earlier studies by Pearson (1976; 1980) that reported 69% and 63% of communications received responses to both communications (dyads). This high degree of similarity with previous research suggests the experimental treatment probably did not alter companies� ordinary response predispositions, and thereby enhances the likelihood that actions suggested by gain-loss theory help explain any variance in the dependent variables. There were three key dependent variables examined in this study. The first was the total dollar value of the response. That is the monetary value of all enclosures included in the response (i.e. economic incentives such as cash refunds, coupons, etc.), and whatever the letter writer also received from the manufacturer (e.g. t-shirts, ball-point pens, bottle openers, refrigerator magnets, etc.). Note that in both the compliment and complaint conditions no specific requests or questions were permitted, either directly or implied. A second dependent variable, and one that also bears directly upon the gain-and-loss theory, was the elapsed response time. This was operationalized as the total elapsed postmark-to-postmark time, minus non-business weekend days. It was decided that even though mail delivery occurred on both Saturdays and Sundays in a few places in the U.S., very few organizations regularly obtained mail pick-up services over the weekend. So, while incoming mail might be deliverable on weekends, outgoing mail likely remained unprocessed by the recipient except from Monday through Friday. Participants were required to obtain official a Certificate-of-Mailing at the post office from which they sent their correspondence. These served to document both the outgoing postmark and mailing location. The third dependent variable was the readability (i.e. the degree of technical and verbal sophistication), associated with each company�s reply letter. Higher readability would indicate lengthier replies, more sophisticated and technical references, and greater effort by the responding company (all as suggested by gain-and-loss theory). To measure this, the Gunning Fog Index (GFI) was computed for each written reply received. The GFI produces a numerical score from a formula based on the following components:

GFI = ((ave. sentence length) + (% of 3+ syllable words)) x .4 = Approx Grade Level

The formula requires the literal counting of the number of words in a letter, the number of sentences, the number of paragraphs, and the number of three-syllable (or greater) words which appear in the reply. The resulting score approximates the number of years of education required to fully comprehend the message. Thus, a score of 9.0 would mean a required education of 9th grade, while a 14.5 score would mean at least two years of college. RESULTS AND DISCUSSION: To examine the differences in each manufacturer�s handling of pairs of complaints and compliments, t-tests with

McCall and Eckrich

142

paired samples were used. The objective was to isolate the impact suggested by gain-and-loss theory, and to compare correspondence handling within each company. As predicted by gain-and-loss theory, the total monetary value of a response to a complaint was significantly higher than to a compliment ($ 4.44 versus 1.98, t (174) = 5.02, p < .0001). Similarly, again as predicted, a complaint received a response in an average of 12.5 days whereas a compliment received a response averaging 13.5 days, t (172) = -1.69 p < .046 (one-tailed). Finally, responses to complaint letters exhibited a much higher educational requirement for readership with a complaint response averaging 12.1 years (i.e. beyond high school), and the compliment response averaging 10.4 years (a rising sophomore in high school) (p<.000). To illustrate the typical pattern a few of the dyads will be exhibited and briefly discussed. For instance, when writing to a major cereal manufacturer (see Tables 2-6), it was noted that the date of freshness had expired and the product was stale. The complaint was responded to in 11 days and received a $5.50 reimbursement; the compliment was responded to in 17 days and received no coupons or monetary return. Correspondingly, the GFI for the complaint indicated a degree of readability requiring almost 4 years of college (i.e. 14.8 years), while the compliment received a GFI score of 8.2 years (rising junior high school student). In the five cases exhibited, coupon values regularly exceed those of the compliments by factors of almost 10, and in only one instance did the response time for the compliment exceed that of the complaint. None of the cases exhibited were seen as remarkably different from the average response pattern across all dyads.

TABLE 2 Kellogg’s

Apple Jacks � Stale on date of purchase

Complaint Compliment Dollar Value $5.50 $ 0.00 Response Time 11 days 17 days GFI 14.8 8.2 Total # Words 159 16 # Sentences 9 2 #Hard Words 34 2


143

TABLE 3 Lipton

Noodles & Sauce � Maggots in mix

Complaint Compliment Dollar Value $ 9.95* $ 1.29 Response Time 8 days 6 days Gunning Fog 14.5 10.7 Total # Words 134 74 # Sentences 7 5 #Hard Words 23 9 *included 3 coupons for tally unrelated products � Lipton Tea, Cup-a-Soup, and Recipe Secrets Soup Mix

TABLE 4 The Golden Grain Company

Rice-A-Roni � Seasoning crystallized

Complaint Compliment Dollar Value $ 2.40* $ .25* Response Time 6 days 28 days Gunning Fog 10.0 7.37 Total # Words 122 65 # Sentences 8 6 #Hard Words 12 5 *Both received �What�s for Dinner� recipe books.

TABLE 5 Nissin Foods, Inc.

Oodles-of-Noodles � Bugs in package

Complaint Compliment Dollar Value $ 3.20 $ .40 Response Time 14 days 110 days Gunning Fog 13.2 9.0 Total # Words 3337 51 # Sentences 6 2 #Hard Words 44 4

TABLE 6 Hershey Foods

Kit Kat Bars � Mushy wafer

Complaint Compliment Dollar Value $ 4.00 $ 0.00 Response Time 6 days 7 days Gunning Fog 12.0 7.9 Total # Words 122 82 # Sentences 8 6 #Hard Words 18 5

McCall and Eckrich

144

While many companies offered coupons in response to both the complaint and the compliment it was the complaint that regularly received individual coupons with greater face value. Thus, not only did the larger total monetary value support the outcome predicted by gain-and-loss theory, but the fact that several organizations systematically produced and utilized coupons with greater face value is even greater support than originally anticipated. These data, while preliminary, are important for several reasons. First, it is clear that complaints represent a more significant concern for manufacturers due to the potential loss of business. Indeed, over 54% of those complainers whose complaints are satisfied repurchase the product. The higher total dollar value, quicker response time, and higher Gunning Fog Index clearly demonstrates the importance placed on this group of consumers, and provide evidence to support the hypotheses provided by gain-and-loss theory. A second interesting finding of this study is the lesser importance that seems to be given to consumer compliments. While it is clear why one needs to address customer complaints, there are important practical implications for rewarding customer compliments. Notably, all customers are not created equally, and the frequently cited Pareto Rule of “80/20” suggests that the favored 20% of our customers should be treated with great care. Relationship marketing speaks to customer satisfaction and the reasons for that satisfaction are of great importance to future sales. Finally, while the evidence appears to suggest that complainers are indeed winning, these data begin to suggest the importance of caring for those customers who support us. While there may not be immediate gains and losses, the long-term implications of customer follow-up and support are clear. Much like the person who hears negative feedback that over time becomes positive, there is potentially much to be gained from a once disgruntled, but now increasingly satisfied customer. Future research should articulate more clearly the exact benefits of customer satisfaction and the theoretical antecedents thereof, as well as consider the effects of different communication media, such as e-mail and �800� telephone numbers on the dependent measures examined in this research. REFERENCES Aronson, E., & Linder, D.E. (1965). Gain and loss of esteem as determinants of

interpersonal attractiveness. Journal of Experimental Social Psychology, 1, 156-171.

Baker, M.J. (1998) Relationship Marketing in Three Dimensions, Journal of Interactive Marketing (Autumn).

Best, A. & A.R. Andreason (1977), Consumer Response to Unsatisfactory Purchases: A Survey of Perceiving Defects, Voicing Complaints, and Obtaining Redress, Law and Society (Spring), 11,p. 701-742.

Bresnahan, J. (1998), Improving the Odds, CIO Enterprise Magazine (November). Copulsky, J.R., & Wolf, M.J. (1990). Relationship marketing: Positioning for the

future. Journal of Business Strategy, (July/August), 16-20.


145

Erickson, G. Scott & D.W. Eckrich (2001), Consumer Affairs Responses to Unsolicited Customer Compliments, Journal of Marketing Management, 27, Western Publishers Ltd.

Evans, J.R., & Laskin, R.L. (1994). The relationship marketing process: A conceptualization and application. Industrial Marketing Management, 23, 439-452.

Fournier, S., S. Dobsha, D.G. Mick (1998), Preventing the Premature Death of Relationship Marketing, Harvard Business Review (November-December).

Gruen, T. W., (1997), Relationship Marketing: The Route to Marketing Efficiency and Effectiveness, Business Horizons (November-December).

Martin, C.L. & D.T. Smart, (1988), Relationship Correspondence: Similarities and Differences in Business Response to Complimentary versus Complaining Consumers, Journal of Business Research, Vol. 17, No. 2 (September), p. 155-173.

Martin, C.L. and D.T. Smart (1989), Consumer Correspondence: An Exploratory Investigation of Consistency Between Business Policy and Practice, Journal of Consumer Affairs, Vol. 23, No. 2 (Winter), p. 364-382.

Mowen, J.C., & Minor, M.S. (2001). Consumer Behavior: A framework. New Jersey: Prentice-Hall

Pearson, M.M. (1976), A Note on Business Replies to Consumer Letters of Praise and Complaint, Journal of Business Research, (February), No. 4, p. 61-68.

Pearson, M.M. W.R. Hoskins, & G.M.Gazda (1980), The Use of Student Letters of Praise and Complaint As An Introduction to Marketing activity, Journal of Marketing Education, (Spring), p. 18-24.

Peterson, R.S. (1995), Relationship Marketing and the Consumer, Journal of the Academy of Marketing Science, Vol. 23, No. 4 (Fall), p. 255-271.

Petrof, J. V. (1997), Relationship Marketing: The Wheel Reinvented, Business Horizons, (November-December).

Power, C. (1993). A Smithsonian for stinkers. Business Week, (August), 82. Reichheld, F. F. (1993). Loyalty-Based Management. Harvard Business Review,

(March/April), 64-73. Sheth, J.N., & Parvatiyar, A. (1995). Relationship marketing in consumer

markets: antecedents and consequences. Journal of the Academy of Marketing Science 23, 255-271.

Singh, J. & R. E. Wilkes (1996), When Consumers Complain: A Path Analysis of the Key Antecedents of Consumer Complaint Response Estimates, Journal of the Academy of Marketing Science (Fall), Vol. 24, No. 4, p. 350-365.

Vavra, T.G. (1992). Aftermarketing: How to keep customers for life through relationship marketing. Homewood, II.: Irwin

146

AN INVESTIGATION OF VOLUNTARY DISCLOSURES, SIZE AND FOREIGN SALES OF IRISH FIRMS

Sinéad O’Connor

PricewaterhouseCoopers Robyn Lawrence

University of Scranton

ABSTRACT: This study examined the relationship between the voluntary disclosures made by firms listed on the Irish Stock Exchange in their annual reports and two firm specific characteristics. The annual reports of 39 firms were examined for voluntary disclosures. The frequency with which each voluntary item was disclosed was analyzed according to the firm's size and the percentage of the firm's sales made abroad.

This study is a response to the increasing importance of voluntary disclosures and the lack of research regarding this topic for firms in Ireland. The decision to include an item in the annual report is of great importance because voluntary disclosures result in direct costs and can potentially put the firm at a competitive disadvantage. Firms must balance these costs against the potential benefits. Two research questions were addressed: (1) Do larger firms listed on the Irish Stock Exchange make more voluntary disclosures in their annual reports than smaller firms? and (2) Do firms listed on the Irish Stock Exchange with a greater percentage of sales outside of Ireland make more voluntary disclosures? The results of this study found that those firms with turnover or total assets under IR50 million generally disclosed the items examined less frequently than firms of greater size. Those firms with a greater percentage of sales abroad tended to disclose more additional information. However, for both firm size and percentage of foreign sales, the relationships with voluntary disclosures did not appear to be strictly linear. INTRODUCTION: Numerous articles can be found in the accounting literature on research relating to voluntary disclosures in many countries, particularly the United States (US) and the United Kingdom (UK) (for example, see studies by Beattie and Jones (1997) and Gray, Radebaugh, and Roberts (1990)). However, very little information is available on voluntary disclosures of firms in the Republic of Ireland. Some may attribute this to the similarity of the standards in the UK and Ireland. However, the environmental influences on companies in each country are very different. While both countries are members of the European Union (EU), Ireland is adopting the Euro, whereas the UK has persistently decided to remain out of the currency. Also, since its formal separation from the London Stock Exchange (LSE) in 1995, an event that occurred as a result of requirements of the European Community Investment Services Directive, the Irish Stock Exchange (ISE) has, naturally, become more independent.


147

This study examines the relationship between two firm specific characteristics, the firm�s size and the percentage of foreign sales, and the extent of voluntary disclosures made in the annual reports by companies listed on the Irish Stock Exchange (ISE). An understanding of the factors influencing voluntary disclosures has implications for regulators, providers and users of information disclosures in annual reports. The sections that follow contain discussions of why firms might disclose information voluntarily, accounting in Ireland, previous research conducted in this area, the two research questions, the methodology used in this study, the results, and the conclusions, limitations and some extensions of this study.

REASONS FOR MAKING VOLUNTARY DISCLOSURES: Management will make voluntary disclosures in the annual reports only if they believe that the resulting benefits will be greater than the direct and indirect costs incurred. The primary benefit to a firm of making voluntary disclosures is a reduction in the perceived riskiness of the firm resulting in a lower cost of capital. Although there are efforts toward harmonizing financial reporting internationally (for example, the International Accounting Standards Committee and in the European Union), significant differences exist in the substance and enforcement of the financial reporting requirements from country to country. With the increase in transnational business and a growing internationalization of the financial markets and share ownership, users of the annual reports increasingly seek comparability of financial information. Effective voluntary disclosures can enhance comparability, improve public relations (with, for example, employees, trade unions, investors, customers, and the public in general) and ease pressure on governments for increased disclosure regulation.

When deciding what information to include in an annual report, management must identify the targeted user, and the range and depth of the information to be disclosed (Radebaugh and Gray, 1997). The direct costs include data collection, auditing, and the dissemination of the information. One of the main deterrents to voluntary disclosure is that of competitive disadvantage (Firth, 1979). Radebaugh and Gray (1997) define �competitive disadvantage� as �the use of additional information by competitors to the detriment of the corporation disclosing the information� (p. 210). According to Gray, Radebaugh, and Roberts (1990), this competitive disadvantage is associated mainly with future-oriented information and segment information where the segments are narrowly defined. FINANCIAL REPORTING IN IRELAND: The �true and fair� view requirement is an essential part of the accounting system in Ireland as expressed in Irish company law (the Companies Act of 1963 as amended by the Companies Acts of 1986 and 1990, and the European Community�s Group Accounts Regulations of 1992) and due to the inclusion of the requirement in almost every type of business regulation in Ireland. According to Mitchell Chartered Accountants �It requires the directors to make judgments and to select the

O’Connor and Lawrence

accounting policies that are most appropriate to the company�s circumstances� (source: www.mitchellsaccountant.co.Uk/atrueand.htm). According to the Companies Act of 1986, a company�s annual report should include the following items (Price Waterhouse, 1994):

1. The report of the directors on the company�s activities which includes: ♦ A review of the business ♦ The directors� dividend recommendations ♦ Any important events since the end of the financial year ♦ Any likely future developments in the business ♦ An indication of any research and development activities ♦ Information on the company�s policies on the safety, health, and

welfare of employees 2. The financial statements (group and parent company) for the financial

year which include: ♦ Profit and loss account ♦ Statement of total recognized gains and losses ♦ Balance sheet ♦ Cash flow statement

3. The report of the auditors on the financial statements. The parent company profit and loss account can be omitted if the

omission is disclosed in the notes to the statements (Coopers & Lybrand, 1993, I-50.) The format and content of these statements is set out under the European Community (now European Union) Council directive on the preparation of consolidated or group accounts (Deasy, 1994, 746). The historical cost basis is usually used in preparing financial statements, although there are certain situations where, according to the Company Act of 1986, alternative accounting bases can be used (Coopers & Lybrand, I-49). The going concern presumption, consistency with regards to the application of accounting policies from one year to the next, prudence when determining the amount of an item to be recorded in the financial statements, and the accrual basis of accounting must be applied (Ernst and Young, 1993, 72). Departures from measurement and reporting requirements are allowed with adequate reasons that must be disclosed, along with the effects of the departures on the profit and loss account, and the balance sheet (Deasy, 1994, 755).

PREVIOUS RESEARCH: Previous studies on voluntary disclosures in the annual reports have examined the impact of listing requirements by securities exchanges, the level of exports by a firm, the firm�s size, the firm�s auditor, the firm�s country of origin and the perceived costs of the disclosures. Buzby (1975) found that company size was positively associated with the amount of information voluntarily disclosed in the annual reports of United States (US) companies. He posited that the expense of data collection and information dissemination favored disclosures by larger firms. Buzby also found that the listing status (stock


149

exchange versus over-the-counter) did not appear to influence a firm�s decision to disclose additional information. Firth (1979) also found that the voluntary disclosures of a firm were influenced by its size. However, contrary to Buzby�s results, Firth found that firms listed on a US stock exchange were more likely to make voluntary disclosures versus those U.S. firms not listed on a stock exchange. In addition, Firth found that the firm�s auditor (�Big Eight� versus non-Big Eight) appeared to have no impact on the voluntary disclosures made by the firm. Saudagaran and Biddle (1975) looked at the relationships between the firm�s decision to list on a foreign securities exchange (US, Canada, UK, Netherlands, France, Japan, Germany, and Switzerland), and the securities exchange�s requirements and the level of exports to the country of the foreign exchange. Their study found evidence of an inverse relationship between the disclosure levels required by a stock exchange and a firm�s decision to be listed on that stock exchange. Their study also found a positive relationship between the level of a firm�s exports to a country and the firm�s decision to be listed on the stock exchange in that country.

McNally, Eng, and Hasseldine (1982), following up on the studies carried out by Firth (1979), Buzby, (1975), and others, examined the relationship between firm characteristics and the quality of voluntary disclosures made by firms in New Zealand. The results of their study confirmed the positive relationship between the size of a firm and the extent of the disclosures made in an annual report. Additionally, they concluded that while two groups of external users of financial information (stockbrokers and financial editors) placed great importance on the voluntary disclosures made by firms, there was a significant gap between the actual disclosures made by the firms and those disclosures preferred by the users.

Meek and Gray (1990) conducted a study that examined the voluntary disclosures made by European companies listed on the London Stock Exchange. They concluded that the competitive pressure to disclose additional information was far greater than the pressure to meet the minimal disclosure requirements of the London Stock Exchange. The study also noticed specific patterns with regards to disclosures made by firms from the same country. Most voluntary disclosures that were made were on segment information, forecast information, information pertaining to research and development, value-added information, and capital market information.

Gray, Radebaugh, and Roberts (1990), using a questionnaire administered to the chief financial officers of UK firms and US firms, studied the various costs associated with making different types of additional disclosures and some of the benefits that resulted from making these disclosures. They found that although firms were making voluntary disclosures, they generally believed that voluntary disclosures resulted in a net cost. While there were differences in the way that US and UK executives perceived the costs and the benefits associated with voluntary disclosures, they agreed that competitive disadvantage was the main constraint when making voluntary disclosures. The direct costs of making additional disclosures acted as a constraint on disclosing information such as narrowly defined segment information, quantified forecasts, and inflation adjusted profits.


150

RESEARCH QUESTIONS: This study examines the voluntary disclosures made by firms listed on the Irish Stock Exchange, and the relationship between these disclosures and two firm specific characteristics, firm size and percentage of foreign sales. The list of voluntary disclosures used in this study was created on the basis of previously published research and an initial examination of the reports. The voluntary disclosures examined in this study are:

1. Turnover (net sales), profit, net assets and number of employees by geographical segment;

2. Turnover (net sales), profit, net assets and number of employees by line of business;

3. Financial statements that are not required by company law: a. Value Added Statement; b. Reconciliation of Movement in Shareholder Funds; c. Reconciliation of Operating Profit to Net Cash Inflow from

Operations; d. Analysis of Changes in Cash and Cash Equivalents; e. Analysis of Changes in Financing; and f. Reconciliation of Net Cash Flow to Debt.

4. Capital market information 5. Historical information (financial data presented beyond the past two

years). 6. Percentage increases in Profit and Turnover.

Segment information is recommended by various Irish accounting

standards, but is considered voluntary information in this study because there are no statutory provisions defining a segment (Coopers & Lybrand, I-58) and because of the exemption clause in SSAP 25 and the 1986 Act. This exemption allows firms to refrain from disclosing certain items of information pertaining to segment (geographical and line of business) data if the firm feels that the disclosure of this information would be �prejudicial to the company�s interests� (MacLochlainn and O�Kane, 1994, 124). However, if the decision is made by the firm not to disclose turnover, net assets, and/or net profit for each geographical segment and/or each line of business, the firm must disclose this decision in its report and the reason why the information is not being included. The number of employees by location and line of business was frequently disclosed and was, therefore, included in this study. Firms also disclosed substantial amounts of other types of geographical and segment information. However, due to the wide variety of items disclosed and the relative infrequency of the disclosure of any particular item, these items were recorded under the heading of �other items�.

There are a number of statements that are recommended for disclosure in the Republic of Ireland but that only some firms disclose. One is the value added statement which provides information about the value generated by a firm�s business and the allocation of this added value to the various stakeholders. Increasingly, firms are disclosing this statement even though its disclosure is


151

perceived by some (see Meek and Gray (1990) and Gray, Radebaugh and Roberts (1990) for conflicting evidence) to leave firms susceptible to competitive disadvantage. Results of prior research on voluntary disclosures made by firms in their annual reports support the idea that the firm�s size has a significant impact on the level of its voluntary disclosures. The most common explanation for this is that larger firms have the required resources to gather the necessary information and to disseminate the findings. Another common explanation is that the competitive disadvantage is far greater for smaller firms making voluntary disclosures than for the bigger firms, with the result that smaller firms tend to refrain from making any disclosures that are not required.

In addition to firm size, the percentage of sales abroad was examined in this study. The percentage of sales abroad was examined to investigate the possibility that firms with more foreign involvement may make more voluntary disclosures.

Based on the preceding considerations the following research questions were constructed: 1. Do larger firms listed on the Irish Stock Exchange make more voluntary disclosures in their annual reports than smaller firms? 2. Do firms, listed on the Irish Stock Exchange, with a greater percentage of sales outside of Ireland, make more voluntary disclosures? METHODOLOGY: Following a request to over 70 of the companies currently listed on the Irish Stock Exchange, 39 annual reports were received and examined for this study. All requests for annual reports were sent to companies listed on the Main Market. All of those companies whose annual reports were used in this study are classified as large companies according to Statutory Instrument No. 396 of 1993, having total assets in excess of IR6 million and total turnover in excess of IR12 million. The reports used were those for the 1996 fiscal year.

The companies were grouped according to size by total assets and turnover as disclosed in the group balance sheets and the group profit and loss accounts, respectively. The groups used were similar to categories used by the AICPA in their annual survey of accounting practices, Accounting Trends and Techniques. (Due to the small number of firms that fell into the range, Groups A2 and S2 each consisted of a combination of two groups used by Accounting Trends and Techniques.) Similar categories were used to group firms in terms of turnover. Table 1 shows the ten size groups (five defined by total assets and five defined by turnover) used in this study. Total assets of the firms examined ranged from IR20.9 billion to IR1.65 million with a mean of IR1.1 billion. Turnover ranged from IR0 to IR19 billion, with a mean of IR806 million. As shown in Table 1, the firms also were assigned to one of four groups based upon their percentage of sales abroad. The percentage of sales abroad varied from 0% to 87%, with an average of 30.6%.


152

TABLE 1 � DEFINITION OF FIRM GROUPINGS Group Name

Number of Firms

Value Range for Group

A1 7 Total assets over IR2 billion A2 3 Total assets between IR500 million & IR2 billion A3 13 Total assets between IR100 million & IR500 million A4 9 Total assets between IR50 million & IR100 million A5 7 Total assets under IR50 million

39 S1 2 Total turnover over IR2 billion S2 5 Total turnover between IR500 million & IR2 billion S3 15 Total turnover between IR100 million & IR500 million S4 4 Total turnover between IR50 million & IR100 million S5 13 Total turnover under IR50 million

39 A 6 75% or more of sales are abroad B 5 Between 50% and 75% of sales are abroad C 9 Between 25% and 50% of sales are abroad D 8 Between 1% and 25% of sales are abroad 28

Once the firms were assigned to their appropriate groups, each annual report was examined for voluntary disclosures. The overall percentage of firms to disclose an item was calculated, as well as, a calculation of the percentage of firms to disclose each item in each of the groups constructed. Using the Shapiro-Wilk test for normality, total assets, total sales, and the percentage of sales abroad were not found to be normally distributed. Consequently, the Wilcoxon Ranked-Sums test was used. RESULTS: All of the reports examined disclosed some type of additional information. As can be seen from Table 2, 86% of the firms with two or more geographical segments disclosed some type of geographical information while only 62% of the firms with two or more business segments disclosed some type of information by line of business (LOB). Those firms that did not disclose any geographical or LOB information availed themselves of the exemption included in Section 6 of the 1986 Act and in SSAP 25. Turnover was the most frequently disclosed item both geographically and by LOB, followed by profit. This is significant considering that according to Gray, Radebaugh, and Roberts, (1990), LOB turnover, geographic profits and LOB profits ranked high in terms of resulting in net costs, competitive disadvantage and production costs. Note, however, that only 31% of the firms disclosed all three items (turnover, net profit, and net assets) geographically, and only 26% of the firms disclosed all three items by LOB. Net assets by geographic segments were disclosed by 34% of the firms,


153

while only 29% of the firms disclosed this item by LOB. The low disclosure of net assets by segments is somewhat surprising given that it was not ranked particularly high regarding net costs, competitive disadvantage, or production costs in the Gray, Radebaugh, and Roberts (1990) study.

Of the non-segmental items that were examined in this study (see Table 2), a reconciliation of the movement in shareholders funds was the one most frequently disclosed (by 87% of the firms). The next most frequently disclosed item was historical information. This ranged from a summary of key financial numbers for the last five years to condensed balance sheets and profit and loss accounts for the past ten years. Interestingly, the value added statement and capital market information were disclosed by only 10% and 5% of the firms, respectively.

TABLE 2 � OVERALL RESULTS OF INFORMATION DISCLOSURES Type of Information Number of Firms

Disclosing Information

Percentage of Firms Disclosing Information

Geographical Information: 30 86% Turnover 26 74% Profit 19 54% Net Assets 12 34% Turnover, Profits & Net Assets 11 31% Number of Employees 6 17% Line of Business Information: 21 62% Turnover 20 59% Profit 15 44% Net Assets 10 29% Turnover, Profits & Net Assets 9 26% Reconciliation of movement in shareholders funds

34 87%

Historical information 28 72% Percentage increase in profit 27 69% Percentage increase in turnover 26 67% Reconciliation of operating profit to net cash inflow from operations

24 62%

Analysis of changes in cash & cash equivalents

17 44%

Analysis of changes in financing 12 31% Reconciliation of net cash flow to debt 8 21% Value added statement 4 10% Capital Market Information 2 5%

The results relevant to whether larger firms listed on the Irish Stock

Exchange make more voluntary disclosures in their annual reports than smaller firms (research question number one) are disclosed in Tables 3 through 6. An examination of the disclosure percentages by groups for geographical information


154

reveals a possible non-linear relationship between firm size (defined by total assets and total net sales) and each of the voluntary information disclosures examined in this study. According to the information reported in Table 3, geographic information was disclosed by 86% of Group A1 firms (the largest firms), 100% of Group A2 firms, 92% of Group A3 firms, 86% of Group A4 firms and only 67% of Group A5 firms (the smallest firms). Geographic information on turnover and profit was most likely to be disclosed by Group A2 firms and least likely by Group A5 firms. Geographical net asset information was most likely to be disclosed by Group A4 firms and was not disclosed by Group A2 firms. Geographical employee information was most likely to be disclosed by Group A1 firms, with no disclosures by Group A2 and Group A5 firms. The results of the Wilcoxon Rank Sums tests, reported in Table 6, revealed that larger firms tended to disclose geographic turnover (p-value of .005) and profit (p-value of .020) more often than smaller firms as defined by total assets. A marginally significant relationship was found between employee information and size (p-value of .054). No statistically significant size relationship was found for geographic information disclosures of net assets.

TABLE 3 � GEOGRAPHICAL INFORMATION BY SIZE (TOTAL ASSETS AND TURNOVER)

Group Overall Turnover Profit Net Assets Employees A1 86% 86% 86% 43% 43% A2 100% 100% 100% 0% 0% A3 92% 77% 38% 31% 15% A4 86% 86% 57% 57% 14% A5 67% 33% 33% 17% 0% S1 100% 100% 100% 50% 50% S2 100% 100% 60% 20% 40% S3 86% 79% 57% 29% 21% S4 100% 100% 33% 67% 0% S5 73% 45% 45% 36% 0%

A nonlinear relationship between geographical disclosure and firm size as

defined by net sales was also evident. According to the information reported in Table 3, all of the firms in Groups S1 (the largest firms), S2 and S4 reported some geographical information. Eighty-six percent of Group S3 firms and 73% of Group S5 firms (the smallest firms) made some geographical disclosures. Geographical information on turnover was most likely to be disclosed by firms in Groups S1, S2 and S4, and least likely by Group S5 firms. Geographical profit information was most likely to be disclosed by Group S1 firms and least likely by Group S4 firms. Geographical net asset information was most likely to be disclosed by Group S4 firms and least likely by Group S2 firms. Geographical employee information was most likely to be disclosed by Group S1 employees and was not disclosed by Group S4 or Group S5 firms. The results of the Wilcoxon Rank Sums tests, reported in Table 6, were generally similar to those


155

reported for size based upon total assets. Disclosures of geographic information on turnover (p-value of .002) and employees (p-value of .034) were significantly related to the total sales of the firm; disclosures of geographical information on net assets were not. The relationship between total sales and the disclosure of geographical profit information was only marginally statistically significant (p-value of .0525).

Overall, LOB disclosures were made less frequently by the largest and the smallest firms as defined by total assets and turnover (total net sales), revealing, as with geographical disclosures, a possible non-linear relationship between firm size and the voluntary LOB disclosures examined in this study. According to Table 4, LOB information was disclosed by 29% of Group A1 firms, 100% of Group A2 firms, 91% of Group A3 firms, 63% of Group A4 firms and 17% of Group A5 firms. LOB turnover and profit information was most likely to be disclosed by Group A2 firms. LOB turnover was least likely to be disclosed by Group A1 firms and it was the only LOB item disclosed by any of the Group A5 firms. LOB net assets information was most likely to be disclosed by Group A2 firms. LOB employee information was most likely to be disclosed by Group A3 firms and was not disclosed at all by Groups A1, A2, and A5. The results of the Wilcoxon Rank Sums tests, reported in Table 6, show that larger firms disclosed LOB turnover (p-value of .032) and profit (p-value of .006) more frequently than smaller firms as defined by total assets. No statistically significant size relationship was found for LOB disclosures of net assets and employee information.

TABLE 4 � LOB INFORMATION BY SIZE (TOTAL ASSETS & TURNOVER) Group Overall Turnover Profit Net Assets Employees

A1 29% 14% 14% 14% 0% A2 100% 100% 100% 67% 0% A3 91% 91% 64% 36% 27% A4 63% 63% 50% 38% 25% A5 17% 17% 0% 0% 0% S1 0% 0% 0% 0% 0% S2 100% 80% 60% 40% 0% S3 67% 67% 53% 33% 20% S4 100% 100% 67% 67% 67% S5 33% 33% 22% 11% 0%

The results for LOB disclosures examined in terms of firm size according

to turnover revealed that, similar to the results using total assets, more LOB disclosures were made by the intermediate sized firms; fewer disclosures were made by larger firms and smaller firms. Overall, Group S4 firms, the second to the smallest firms in terms of turnover, disclosed the four LOB items most frequently. According to Table 4, LOB information was disclosed by none of the Group S1 firms (the largest firms), 100% of Group S2 firms, 67% of Group S3 firms, 100% of Group S4 firms, and 33% of Group S5 firms. In addition to Group


156

S1 firms, Group S2 firms and Group S5 firms did not disclose LOB employee information. The result of the Wilcoxon Rank Sums test for LOB disclosures found no statistically significant size relationship based on turnover.

Table 5 contains the results relevant to the other voluntary disclosures examined according to size. Group A1 firms disclosed only historical information, the percentage increase in profit, and capital market information most frequently. Group A2 firms disclosed four of the ten items most frequently, but did not disclose an analysis of changes in financing, an analysis of changes in cash and cash equivalents, or capital market information. Group A5 firms (the smallest firms in terms of total assets) disclosed seven of the ten items examined least frequently. Group A2 firms and Group A4 firms were the most likely to disclose a reconciliation of movement in shareholders funds. Group A2 firms were the most likely and Group A1 firms were the least likely, to disclose a reconciliation of operating profit to net cash inflow from operations. Group A4 and Group A5 firms were the most likely to disclose an analysis of changes in financing while none of the firms in Group A2 disclosed this item. Group A5, A4 and A3 firms were the most likely to disclose an analysis of changes in cash and cash equivalents, while none of the firms in Group A2 disclosed this item. However, Group A2 firms were the most likely to disclose a reconciliation of net cash inflow to debt, while Group A1 and Group A5 firms were the least likely to disclose this item. Group A2 firms were the most likely to disclose a value added statement while none of the firms in Groups A3 and A5 disclosed this item. Overall disclosure of historical information was high (43% to 86%). Group A1 and Group A3 firms were the most likely to disclose this information. Group A4 firms were the most likely firms to disclose the percentage increase in turnover while Group A5 firms were the least likely firms to disclose this information. Group A3 firms disclosed the percentage increase in profit most frequently, while Group A5 firms were the least likely to disclose this information. Only Group A1 and Group A3 firms disclosed capital market information.

The only items in Table 5 that revealed a statistically significant association with total assets, according to the Wilcoxon Rank Sums test (reported in Table 6), were the reconciliation of movement in shareholders funds (p-value of .005) and the disclosure of the percentage change in profit (p-value of .027). Although this suggests that larger firms disclose these two items more frequently than smaller firms, the association does not appear to be a strictly linear one. The results of the analysis defining size by turnover revealed that Groups S1 and S4 each disclosed five of the ten items most frequently. Group S5 firms (the smallest firms in terms of turnover) disclosed six of the ten items least frequently. This, again, suggests that, while, overall, smaller firms tend to make fewer voluntary disclosures, a non-linear association exists between firm size and voluntary disclosures. All of the firms in Groups S1, S2, and S4 disclosed a reconciliation of movement in shareholders funds. Group S5 firms were the least likely to disclose this item (69%). Group S4 firms were the most likely to disclose a reconciliation of operating profit to net cash inflow from operations while none of the firms in Group S1 disclosed this item. Group S4 firms were the most likely to


157

disclose a reconciliation of operating profit to net cash inflow from operations, an analysis of changes in financing and an analysis of changes in cash and cash equivalents. None of the firms in Group S1 disclosed these items. Group S3 firms were the most likely to disclose a value added statement. None of the Group S2 and S5 firms disclosed a value added statement. Group S2 firms, the second to the largest firms in terms of turnover, were the most likely to disclose historical information. All of the firms in Groups S1, S2, and S3 disclosed the percentage change in profit. Group S1 and Group S2 firms were the most likely to disclose the percentage increase in profit. Only firms in Groups S1 and S3 disclosed capital market information. Group S1 firms were the most likely firms to disclose this information.

TABLE 5 � OTHER VOLUNTARY DISCLOSURES BY SIZE (TOTAL ASSETS AND TURNOVER) Group Reconciliation

of Movement in Shareholder

Funds

Reconciliation of Operating Profit to Net Cash Inflow

From Operations

Analysis of Changes in Financing

Analysis of Changes in

Cash & Cash

Equivalents

Reconciliation of Net Cash

Flow to Debt

A1 85% 14% 14% 14% 14% A2 100% 100% 0% 0% 33% A3 92% 61% 31% 54% 23% A4 100% 89% 44% 56% 22% A5 57% 57% 43% 57% 14% S1 100% 0% 0% 0% 50% S2 100% 80% 20% 40% 0% S3 93% 16% 20% 33% 20% S4 100% 100% 50% 75% 25% S5 69% 69% 46% 54% 23%

Group Value Added Statement

Historical Information

% Increase in Sales

% Increase in Profit

Capital Market Information

A1 14% 86% 57% 86% 14% A2 33% 50% 50% 50% 0% A3 0% 85% 85% 92% 8% A4 22% 67% 89% 67% 0% A5 0% 43% 14% 14% 0% S1 50% 50% 100% 100% 50% S2 0% 100% 100% 100% 0% S3 67% 87% 80% 93% 7% S4 50% 75% 100% 75% 0% S5 0% 46% 31% 31% 0%

According to the results of the Wilcoxon Rank Sums tests, a significant

association (see Table 6) exists between firm size defined according to turnover, and the disclosure of an analysis of changes in financing (p-value of .034), historical information (p-value of .012), percentage change in turnover (p-value of


158

.012) and the percentage change in profit (p-value of .000). Disclosure of an analysis of changes in financing tended to be more prevalent for smaller firms, while the disclosures of the other three items tended to be more prevalent for larger firms.

TABLE 6 � WILCOXON RANK SUMS RESULTS Total Assets Total Turnover Percentage of Sales

Abroad

p-value

Mean (Yes),(No)

p-value

Mean (Yes),(No)

p-value

Mean (Yes),(No)

0.005 (23.6),(12.8) 0.002 (23.9),(12.1) 0.000 (25.9),(8.3)

Geographic Profit 0.020 (24.4),(15.8) 0.053 (23.7), 16.5) 0.006 (25.1),(15.1) Geographic Net Assets

0.513 (21.8),(19.2) 0.670 (21.2),(19.5) 0.230 (23.3),(18.5)

Geographic-Employees

0.054 (28.3),(18.5) 0.034 (29.2),(18.3) 0.012 (30.7),(18.1)

LOB Turnover 0.032 (23.9),(15.9) 0.067 (23.3),(16.6) 0.058 (23.4),(16.5) LOB Profit 0.006 (26.3),(16.0) 0.120 (23.6),(17.8) 0.039 (24.7),(17.0) LOB Net Assets 0.150 (24.5),(18.4) 0.230 (23.8),(18.7) 0.058 (25.6),(17.9) LOB # of Employees

0.850 (19),(20.1) 0.437 (23.8),(19.4) 0.565 (22.8),(19.6)

Reconciliation of Movement in Shareholder Funds

0.005 (22.0),(6.4) 0.059 (21.3),(10.9) 0.120 (21.1),(12.6)

Reconciliation of Operating Profit from Net Cash Inflow from Operations

0.400 (18.8),(22.0) 0.120 (17.7),(23.7) 0.800 (19.6),(20.6)

Analysis of Changes in Financing

0.190 (16.3),(21.6) 0.034 (14.2),(22.6) 0.032 (14.2),(22.6)

Analysis of Changes in Cash and Cash Equivalents

0.130 (16.8),(22.5) 0.090 (16.4),(22.8) 0.060 (16.1),(23.0)

Reconciliation of Net Cash Inflow to Debt

0.800 (21.0),(19.7) 0.700 (21.6),(19.6) 0.190 (24.8),(18.8)

Value Added Statement

0.560 (23.3),(19.6) 0.320 (25.5),(19.4) 0.023 (32.2),(18.6)

Historical Information

0.270 (21.3),(16.7) 0.012 (22.9),(12.6) 0.004 (23.3),(11.6)

% Change in Turnover

0.484 (20.9),(18.2) 0.012 (23.3),(13.5) 0.004 (23.7),(12.7)

% Change in Profit 0.027 (23.7),(13.9) 0.000 (24.4),(10.1) 0.002 (23.8),(11.4) Capital Market Information

0.265 (29.0),(19.5) 0.192 (30.5),(19.4) 0.039 (36.2),(19.1)

The results relevant to the second research question, examining whether

firms with a greater percentage of foreign sales are more likely to make more voluntary disclosures, are reported in Tables 6 and 7. An examination of the voluntary disclosures by firms grouped according to the percentage of their sales


159

made to customers outside of Ireland and the UK (see Table 7), revealed that firms with more sales abroad tended to disclose geographic turnover information, a reconciliation of movement in shareholder funds, a valued-added statement, the percentage increase in sales and capital market information more often than firms with less sales abroad. Evidence of greater disclosures by firms with fewer sales abroad were found for disclosures of geographical net assets, LOB employees, a reconciliation of operating profit to net cash inflow from operations, an analysis of changes in financing, and an analysis of changes in cash and cash equivalents. Group B firms (between 50% and 75% of sales abroad) dominated LOB disclosures. These firms also disclosed geographical turnover information, geographical profit information, a reconciliation of movement in shareholders funds, a reconciliation of net cash flow to debt, and historical information most frequently. Thus, non-linear relationships are suggested between the percentage of foreign sales and many of the voluntary disclosure items.

TABLE 7 � DISCLOSURES BY PERCENTAGE OF SALES ABROAD Group A

(75%-100%) Group B

(50%-75%) Group C

(25%-50%) Group D

(1%-25%) Geographic Information Overall 100% 100% 100% 88% Turnover 100% 100% 89% 88% Profit/loss 67% 75% 67% 63% Net assets 17% 50% 44% 63% Number of employees 50% 0% 33% 0% Line of Business Overall 50% 100% 78% 50% Turnover 90% 100% 67% 50% Profit/loss 33% 100% 44% 50% Net assets 17% 75% 33% 38% Number of employees 0% 25% 22% 25% Reconciliation of movement in shareholders� funds

100% 100% 100% 63%

Reconciliation of operating profit to net cash inflow from operations

50% 50% 67% 75%

Analysis of changes in financing 0% 25% 22% 38% Analysis of changes in cash & cash equivalents

0% 0% 67% 63%

Reconciliation of net cash flow to debt

33% 50% 0% 50%

Value added statement 33% 25% 11% 0% Historical information 83% 100% 100% 63% Percentage increase in sales 100% 75% 78% 75% Percentage increase in profit 100% 75% 89% 75% Capital market information 33% 0% 0% 0%

The results of the Wilcoxon Rank Sums tests, reported in Table 6, revealed that disclosures of the following segment information was more likely for firms with more sales abroad: geographical turnover (p-value of .000), geographical profits (p-value of .006), the geographical distribution of employees (p-value of .012), and LOB profit (p-value of .039). Additionally, disclosures of a value added statement (p-value of .023), historical information (p-value of .004), the


160

percentage change in turnover (p-value of .004), the percentage change in profits (p-value of .002), and capital market information (p-value of .039) were more likely for firms with more sales abroad. Disclosure of the analysis of changes in financing (p-value of .032) was less likely for firms with more sales abroad. CONCLUSIONS AND LIMITATIONS: Voluntary disclosures are an important part of a firm�s communications to external users of accounting information. Every firm in this study made at least one voluntary disclosure. A reconciliation of movement in shareholders funds (somewhat analogous to a statement of shareholders� equity in the US) was the most frequently disclosed item, followed by geographical turnover information. Evidence was found consistent with the presence of mitigating factors regarding a firm�s response to regulatory encouragement (versus mandate) for supplemental information. As mentioned previously, disclosure of a reconciliation of movement in shareholders� funds, encouraged by FRS #3, greatly exceeded disclosures of a reconciliation of operating profit to net cash inflow from operations (FRS #1) and a reconciliation of net cash flow to net debt (FRS #3). Consistent with concerns associated with competitive disadvantage, disclosures of segment turnover exceeded those of segment profit, and disclosures of geographical information exceeded those of line of business information, especially for turnover and profit.

Regarding the two research questions posited in this study, evidence was found that was consistent with a relationship between a firm�s size and its percentage of sales abroad, and the presence of certain voluntary disclosures in its annual report. Specifically, a firm�s size (defined by either total assets or turnover) and its percentage of sales abroad appear to be associated with disclosures of segment information (especially disclosures of segment turnover, segment profit information, and the geographical distribution of employees). The smallest firms and those firms with the smallest percentage of foreign sales appear to be less likely to make segment disclosures. These disclosures are likely to place smaller firms with fewer subsidiaries at a competitive disadvantage. This finding is consistent with those by Firth (1979), McNally, Eng and Hasseldine (1982), and Busby (1975).

The firm�s size (particularly as defined by turnover) and its percentage of sales abroad appear to be associated with disclosures of an analysis of changes in financing (negative relationship), historical information, the percentage change in turnover, and the percentage change in profit. The percentage of foreign sales also appears to be associated with the disclosure of the value added statement and capital market information. The finding that those firms with the smallest percentage of sales abroad tended to make fewer voluntary disclosures is consistent with the idea that the firms making the voluntary financial disclosures were targeting users beyond investors and creditors. Those firms with customers in foreign countries may perceive net benefits to making certain voluntary disclosures. The finding that the smaller firms tended to make more disclosures of an analysis of changes in financing points to the presence of mitigating factors.


161

Of particular interest in this study is the finding of a noticeable decline in voluntary disclosures for the largest firms (defined by either total assets or turnover). This finding is consistent with the existence of additional factors that mitigate the declining costs and/or rising benefits of voluntary disclosures (such as declining relative direct costs and/or less competitive disadvantage) enjoyed by larger firms. One such factor may be that larger firms are more likely to come under increased scrutiny from regulatory, political and/or employee representatives. Thus, these firms would have a higher cost threshold to meet when deciding whether to make voluntary disclosures. A similar argument could apply to firms making 75 percent or more of their sales abroad. Potential costs associated with additional regulatory and political scrutiny of firms with heavy foreign involvement may overcome the advantages of voluntary disclosures. The results of this study support the proposition that market forces may be most effective for encouraging voluntary disclosures by mid-sized firms and by firms with moderate foreign involvement. This has implications for regulators as they ponder requirements that target firms of particular sizes (for example, �big GAAP� versus �little GAAP� in the US), and requirements that apply to firms with foreign sales. There are several limitations associated with this study, as well as, a number of possible extensions. One limitation is that only 39 of the 90 firms listed on the ISE were examined. Also, the study is based on the annual reports of these 39 firms for just one year. Possible extensions of this study could include all or, at least, more of the firms listed on the Main Market and possibly the Unlisted Stock Market and the Third Stock Market, and annual reports from multiple fiscal periods for each firm. Replications of this study using firms from other countries would provide evidence concerning the ability to generalize these results to firms of other countries. The results of this study point to mitigating factors in the relationships between firm size and percentage of foreign sales, and certain voluntary disclosures. Studies to identify and model these mitigating factors should be undertaken. This study was exploratory in nature. More research is needed to explain why firm size and/or percentage of sales abroad appeared to be related to certain voluntary disclosures and not others. REFERENCES: American Institute of Certified Public Accountants (1996). AICPA Annual

Survey of Accounting Practices Followed in 600 Stockholders� Reports: Accounting Trends & Techniques. New Jersey: AICPA.

Beattie, V., & Jones, M.J. (1997). A Comparative Study of the Use of Financial Graphs in the Corporate Annual Reports of Major U.K. and U.S. Companies. Journal of International Financial Management and Accounting. 8, 33-69.

Buzby, S.L. (1975). Company Size, Listed versus Unlisted Stocks, and the Extent of Financial Disclosure. Journal of Accounting Research, Spring, 16-33. Choi, F.D.S. (1993). Handbook of International Accounting. New York: John Wiley & Sons Inc.


162

Coopers & Lybrand (1993). International Accounting Summaries. New York: John Wiley & Sons Inc.

Deasy, N.W. (1994). European Accounting Guide: Section on Republic of Ireland. London: Harcourt, Brace and Company.

Ernst & Young (1993). International Business Series: Doing Business in the Republic of Ireland. New York: Ernst & Young.

Firth, M. (1979). The Impact of Size, Stock Market Listing, and Auditors on Voluntary Disclosure in Corporate Annual Reports. Accounting and Business Research. Autumn, 273-279.

Gerrard, Scallan, & O�Brien (1995). Stock Exchange Listing Rules.� [Online] http://www.businessmonitor.co.uk/docs/disc3/stoex

Gray, S.J., Radebaugh, L.H., & Roberts, C.B., (1990). International Perceptions of Cost Constraints on Voluntary Information Disclosures: A Comparative Study of U.K. and U.S. Multinationals. Journal of International Business Studies. 4th Qtr., 597-621.

Guide to the Irish Stock Exchange. (1997). [Online] http:www//ncb.ie/Private/stockex.htm

MacLochlainn, N. & O�Kane, B. (1994). Accounting Standards. Dublin: Oak Tree Press.

McKechnie, P. (1996). Ireland. [Online] http:www//csom.urnn.edu/Newpages/courses/Acct/5310/Ireland.htm

McNally, G.M., Eng, L.H. & Hasseldine, C.R. (1982). Corporate Financial Reporting in New Zealand: An Analysis of User Preferences, Corporate Characteristics, and Disclosure, Practices for Discretionary Information.

Accounting and Business Research, Winter, 11-20. Meek, G.K., and Gray, S.J. (1989). Globalization of Stock Markets and Foreign

Listing Requirements: Voluntary Disclosures by Continental European Companies listed on the London Stock Exchange. Journal of International Business Studies. Summer, 315-335.

Mitchells (1997). A True and Fair View. [Online] http://www.mitchellsaccountant.co.uk/atrueand.htm

Price Waterhouse (1994). Doing Business in Ireland. New York: Price Waterhouse.

Radebaugh, L.H. & Gray, S.J. (1997). International Accounting and Multinational Enterprises. Toronto: John Wiley & Sons Inc.

Saudagaran, S.M. & Biddle, G.C. (1995). Foreign Listing Location: A Study of MNCs and Stock Exchanges in Eight Countries in Eight Countries. Journal of International Business Studies, 2nd Qtr, 319-340.

http://www.mitchellsaccountant.co/

j business and behavioral sciences - semantic · pdf filep.o. box 10367, san bernardino, ca...

Documents