pay dispersion and workforce performance: … · 2006. 2. 1. · yield these beneﬁts. dispersion...

Strategic Management JournalStrat. Mgmt. J., 23: 491–512 (2002)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.235

PAY DISPERSION AND WORKFORCE PERFORMANCE:MODERATING EFFECTS OF INCENTIVES ANDINTERDEPENDENCE

JASON D. SHAW,1* NINA GUPTA2 and JOHN E. DELERY2

1 Gatton College of Business and Economics, University of Kentucky, Lexington,Kentucky, U.S.A.2 Sam M. Walton College of Business Administration, University of Arkansas,Fayetteville, Arkansas, U.S.A.

The compensation literature is replete with arguments, but lacking in empirical tests, regardingthe effects of pay dispersion on organizational outcomes. Pay dispersion may increase effortand provide incentives for high workforce performance levels, but may also inhibit cooperationand goal orientation among employees. Drawing on several theoretical perspectives (individualmotivation, institutional theory, organizational justice, and neoclassical economics), this studypredicts that pay dispersion will be associated with higher levels of workforce performancewhen accompanied by formal individual incentive systems and independent work, while paycompression is desirable in the absence of individual incentive systems and when work isinterdependent. Survey research studies in two industrial sectors (the motor carrier and concretepipe industries) were conducted to address these issues. Interactive regression results weregenerally supportive of the predictions across several measures of workforce performance(accident rates, safety violations, and productivity). Implications of these studies for strategyimplementation in terms of compensation theory and practice are addressed. Copyright 2002John Wiley & Sons, Ltd.

Effective strategy implementation demands a syn-ergistic amalgam of organizational processes andsystems; i.e., congruence is necessary to ensurethat organizational elements work together to pro-mote strategic goals. Most theories extol the virtuesof congruence; few empiricists explore the issuesystematically. An organization’s compensationsystem is arguably the most significant humanresource management system for effective strategyimplementation (Montemayor, 1996), particularlyin view of the enormous costs that a compen-sation system entails. Compensation researchersand practitioners alike posit that pay structures,particularly pay dispersion vs. compression, havecritical implications for strategy implementationand organizational performance (e.g., see Lawler,

Key words: compensation; organizational performance;dispersion*Correspondence to: Jason D. Shaw, Gatton College of Businessand Economics, University of Kentucky, Lexington, KY, 40506-0034, U.S.A.

1981). But is pay more strategically valuable ormore effective when it is widely dispersed orcompressed? This question spurs theoretical dis-agreement (Bloom, 1999; Pfeffer and Langton,1993), providing little resolution to the condi-tions under which widely dispersed or tightly com-pressed compensation systems result in superiorstrategy implementation. This study is designedto begin answering the question theoretically andempirically.

Specifically, this research addresses severalnotable omissions in the pay dispersion/strategyimplementation literature. One, we view pay dis-persion as effective, not across the board, butunder certain organizational contingencies. Thus,we define theoretically the conditions under whicheffective strategy implementation is advanced orimpeded by wage dispersion. Two, we addresspay dispersion at the organizational level, incontrast to the individual focus of much priorresearch (Becker and Huselid, 1992; Ehrenberg

Copyright 2002 John Wiley & Sons, Ltd. Received 1 November 1999Final revision received 25 October 2001

492 J. D. Shaw, N. Gupta and J. E. Delery

and Bognanno, 1990; Pfeffer and Langton, 1993;but see Bloom, 1999, and Cowherd and Levine,1992, for exceptions). Three, we address the issueof horizontal wage dispersion, i.e., dispersion ofpay within the same organizational echelon orcore group of employees, whereas prior researchprimarily concerned vertical pay dispersion, i.e.,the spread of pay across organizational echelons(Cowherd and Levine, 1992). Based on social iden-tity and equity considerations (Baron and Pfeffer,1994; Bloom, 1999; Lazear, 1989; Oldham et al.,1982), it is reasonable to infer that intraclass dis-persion evokes stronger employee reactions thaninterclass dispersion. Four, we examine the impactof wage dispersion on intermediate organizationaloutcomes. Improved financial performance is, ofcourse, the ultimate strategic goal of most organi-zations, but intermediate outcomes are more proxi-mal indicators of strategic success. The logical the-oretical progression moves from strategic compen-sation decisions through employee performancebehaviors (Lawler and Jenkins, 1992) to finan-cial performance. In other words, employee per-formance is the intermediate outcome or the paththrough which compensation strategies affect orga-nizational performance. Thus, this study exam-ines the relationship between pay dispersion andintermediate workforce performance, focusing onpay dispersion within a core employee group,and outlining specific contingencies that determinewhether pay dispersion is beneficial. These dynam-ics are explored in two different industrial settingsto enhance generalizability.

BACKGROUND AND THEORY

Despite implicit universality—pay dispersion iseither good or bad—the literature reveals at leasttwo tacit moderators or conditions for pay disper-sion to be an effective element of strategy imple-mentation. When pay dispersion results from theuse of individual financial incentive systems, andwhen it is used in the context of independent work,it is more likely to be an effective implementationtool than otherwise. These contingency dynamicsare outlined below.

Positive effects of dispersion: the role ofindividual incentives

Bishop (1987) outlined three primary benefits ofdispersed pay structures: they provide incentives

for higher employee effort, they attract a highercaliber of workforce (Freeman, 1977), and theyreduce attrition of good performers for better jobselsewhere. These arguments assume that dispersionoccurs for legitimate reasons, and that highly val-ued human capital receives higher pay than lessvalued human capital. But when pay dispersionoccurs for illegitimate reasons, it is unlikely toyield these benefits. Dispersion due to dysfunc-tional procedures, a lack of formal procedures,game-playing, or politics is unlikely to be effec-tive (Gupta and Jenkins, 1996). That is, the pre-sumed benefits of pay dispersion are attributable,not to highly dispersed pay per se, but to nor-matively accepted dispersion-creating practices.Indeed, internal labor market (e.g., Lazear, 1989)and similar theories of pay dispersion assume thathigher-performing individuals receive higher pay,an assumption that is valid only when proce-dures are in place to ensure this outcome. It isnot surprising that organizations that are unableto defend outcomes with ‘acceptable legitimizedaccounts of their activities . . . are more vulner-able to claims that they are negligent, irrational,or unnecessary’ (Meyer and Rowan, 1991: 50).These arguments are consistent with Suchman’s(1995) descriptions of moral legitimacy, or thenormative evaluation of an organization’s activ-ities. Suchman outlined how outcomes of deci-sions are perceived as legitimate when organiza-tions embrace ‘socially accepted techniques andprocedures’ to demonstrate that they are ‘mak-ing a good-faith effort to achieve valued . . . ends’(Suchman, 1995 : 580). In essence, the motivatingaspects of pay dispersion will be realized onlywhen accompanied by legitimate or normativelyaccepted factors.

More than 90 percent of U.S. organizations usesome kind of individual incentives; this speaks tothe normative or societal acceptance of such sys-tems (Heneman, 1992; Lawler and Jenkins, 1992).Paying for individual performance ‘is so widelyaccepted that almost every organization says itpays for performance’ (Lawler and Jenkins, 1992:1026). In view of the popularity of these sys-tems, it is likely that pay dispersion is perceivedpositively, and behavioral reactions to pay disper-sion are more favorable, when individual incen-tives are used. Several theoretical perspectives(e.g., expectancy, goal-setting, operant condition-ing, and justice theories at the individual leveland institutional theory at the organizational level)

Copyright 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 491–512 (2002)

Pay Dispersion 493

support the argument that pay dispersion enhancesworkforce performance when formal individualincentive systems are used. The use of financialincentives is consistently related to individual per-formance levels (Jenkins et al., 1998). But individ-ual incentives are not necessarily effective withoutcorresponding perceptible pay differentials amongemployees. Individual-level research (e.g., Mitra,Gupta, and Jenkins, 1997; Rambo and Pinto, 1989)validates arguments that meaningful pay differenti-ations, i.e., highly dispersed pay levels, are neces-sary for individuals to be motivated ‘to obtain theprize of high pay’ (Bloom, 1999: 28). An impres-sive array of individual-level tournament compen-sation research shows that performance improvesas the spread of pay increases (Ehrenberg andBognanno, 1990). Beyond employee performance,Becker and Huselid (1992) reported that driversafety in automobile racing also improved as thespread of pay increased.

Organizational justice arguments (Sheppard,Lewicki, and Minton, 1992) also support thebenefits of pay dispersion resulting from theuse of individual incentives. Linking individualincentives and pay dispersion to organizationallydesired outcomes involves a consideration of theappearance of justice (or pseudo-justice; Sheppardet al., 1992). It also provides a strong linkto behavioral reactions resulting from justice(improved workforce performance) and injustice(effort reduction, retaliation, skepticism, andsabotage) (e.g., Bishop, 1987; Skarlicki andFolger, 1997). Organizations continually manageimpressions of the systems or procedures usedto make decisions, in particular, decisions aboutallocation of outcomes among organizationalmembers. In doing so, organizations must establishthe first critical component of perceived fairness:consistency (Leventhal, Karuza, and Fry, 1980).Formal procedures, such as individual incentivesystems, reduce the suspicion that justice ruleshave been violated and the probability ofunexpected results (Gergen, Greenberg, and Willis,1980), and hence the perceived likelihood thatfavoritism, nepotism, and other political factorscan account for the dispersion in pay amongindividuals (Gupta and Jenkins, 1996). Moreover,reliance on incentives conveys a greater senseof perceived control over the outcome, i.e., thedispersion of pay. Such perceptions of control arean essential component of organizational justice(Folger and Greenberg, 1985).

To summarize, a variety of theoretical perspec-tives supports the argument that pay dispersion ismotivation enhancing and results in higher levelsof workforce performance, but these benefits areconcomitant with the use of individual incentives.Pay dispersion in the absence of individual incen-tives is likely to weaken performance–reward link-ages and valences (individual motivation theories),diminish the perceived legitimacy of the system(institutional theory), and violate rules of consis-tency and control (organizational justice theory).By contrast, dispersed pay combined with indi-vidual incentives is likely to enhance workforceperformance including productivity (e.g., Freemanand Kleiner, 1998; Lind and Tyler, 1988) andsafety (e.g., Becker and Huselid, 1992). Thus, thefollowing formal hypothesis is proposed:

Hypothesis 1: The interaction of pay dispersionand individual incentives will be significantlyrelated to dimensions of workforce performance.The form of the predicted interaction is suchthat the relationship between pay dispersion andperformance will be positive when individualincentives are high and negative when individualincentives are low.

Negative effects of dispersion: the role of workinterdependence

The idea that a low level of pay dispersion—paycompression—is a desirable organizational out-come gained popularity in recent years. In a listof ‘best’ HR practices, Pfeffer (1995) includedpay compression as something for which organi-zations should strive. But a contingency factor isagain couched in this seemingly universal argu-ment. Pfeffer (1995) notes that pay compression iseffective in enhancing organizational performanceby reducing interpersonal competition and enhanc-ing cooperation, but only when work is interde-pendent. These arguments can be traced to earliersuggestions (e.g., Hicks, 1963) that the distribu-tion of pay must be set against its impact on thesocial fabric of organizations. More recent socio-logical (e.g., Deutsch, 1985) and economic (e.g.,Levine, 1991) works extend this line of reason-ing. Deutsch (1985) argued that pay dispersiondiminishes performance when work is interdepen-dent by reducing cooperation among employees.He later noted that any movement toward com-pression in interdependent situations increases effi-ciency (Deutsch, 1988; cited in Levine, 1991). The



difficulties ‘created by a given degree of earningsdispersion in a work group increase with the degreeof contact that occurs between the group’s mem-bers’ (Frank, 1984: 564). Levine (1991) also notedthat in participative organizations compression fos-ters cohesiveness, promotes an atmosphere of trustand confidence, and increases the likelihood thatgroup norms will be followed.

Thus, work interdependence is the implicit keyto the effectiveness of pay compression. These the-oretical propositions found empirical support inBloom (1999). The author reported a negative rela-tionship between pay dispersion and performanceamong a sample of professional baseball teams,a context where work interdependence is presum-ably high and fairly constant across organizations.This relationship was evident in terms of behav-ioral performance (e.g., on-field performance suchas winning percentage and finishing position), sim-ilar in nature to the intermediate workforce perfor-mance outcomes at issue here.

In short, sociological and economic theories sug-gest that pay compression enhances cohesiveness,fosters cooperation, and coincides with higher lev-els of workforce performance when work interde-pendencies are high, and some empirical researchsupports this. Interestingly, this theoretical deriva-tion provides no foundation for the positive effectsof dispersion, but rather suggests that the nega-tive effects of dispersion will be more pronouncedwhen work interdependence is high. The followinginteractive hypothesis is therefore proposed:

Hypothesis 2: The interaction of pay dispersionand work interdependence will be significantlyrelated to dimensions of workforce performance.The form of the predicted interaction is such thatthe relationship between pay dispersion and per-formance will be more strongly negative whenwork interdependence is high.

Theoretical integration

The arguments outlined above suggest a morecomplex interactive picture than that painted byabsolute approaches to pay dispersion and perfor-mance. Drawing from individual motivation andjustice theories, we predicted that individual incen-tives would enhance the efficacy of pay dispersionin terms of performance. Drawing from sociologi-cal and economic efficiency perspectives, we pre-dicted that the negative effects of pay dispersion

would be more pronounced when work interdepen-dence was high. These seemingly divergent per-spectives can be integrated to develop a more bal-anced picture of the relationship between pay dis-persion and performance. That is, the perspectivesin toto provide the theoretical context for predict-ing a three-way interaction among pay dispersion,individual incentives, and work interdependence.

When employees do not have to rely on oneanother to accomplish work, individual incentivesprovide the motivational impetus for increasingeffort and reaching higher levels in the pay dis-tribution. Independent work also reduces the needfor employees to engage in dysfunctional compe-tition with their colleagues (Gupta and Jenkins,1996; Lazear, 1989; 1995) to receive higher pay.When work interdependence is low, pay disper-sion should be associated with better workforceperformance among firms using individual incen-tives; i.e., the individual motivation and justicearguments (Hypothesis 1) should prevail underthese conditions. Individual incentives provide theimpetus for higher individual performance (andhigher pay); increases in individual effort con-tribute independently to overall workforce perfor-mance and do not detract from the performance ofothers.

By contrast, when work interdependence is high,a prevailing negative relationship between paydispersion and performance should be evident(Hypothesis 2), but the negative relationship shouldbe stronger among firms also using individualincentives. The potential benefits of pay com-pression, i.e., higher workforce performance lev-els through cooperation and cohesiveness, are lesslikely among firms also using individual incen-tives, since the simultaneous use of individualincentives and interdependent work sends a mixedmessage to employees. Bloom (although there wasno direct test of this argument) notes that in suchsituations individuals ‘will concentrate only ontheir own performance—to the exclusion of orga-nizational goals—since their own performance iswhat matters for moving up in the pay distri-bution’ (Bloom, 1999: 28). Thus, integrating theperspectives allows for a resolution of the theo-retical dilemma. When work interdependence islow, the relationship between pay dispersion andperformance should be positive when individualincentives are used. When work interdependenceis high, the use of individual incentives shouldexacerbate the negative relationship between pay


Pay Dispersion 495

dispersion and performance. Thus, the followinghypothesis is proposed:

Hypothesis 3: A three-way interaction amongpay dispersion, individual incentives, and workinterdependence will be significantly related todimensions of workforce performance. Whenwork interdependence is low, the relationshipbetween pay dispersion and performance willbe strongly positive when individual incentivesare high. When work interdependence is high,the relationship between pay dispersion and per-formance will be more strongly negative whenindividual incentives are high.

STUDY 1 METHOD

Context

Study 1 was conducted in the trucking indus-try. Motor carrier organizations provide a relevant,but interestingly constrained, context in which tostudy pay dispersion. The key employees in truck-ing organizations are, obviously, the truck driversthemselves. Drivers, on average, comprised about73 percent of total employees in the organiza-tions in our sample and therefore compensationmanipulations for drivers have significant strate-gic implications (Bloom, 1999). Driver or work-force performance (i.e., intermediate outcomes) isof critical concern in the industry but, by defini-tion, the primary tasks that drivers engage in arealmost completely independent in nature. Thus,this industry provides a unique opportunity toexamine Hypothesis 1 in a setting isolated fromthe influence of interdependent work. By contrast,Bloom (1999) conducted his ambitious study ina setting (professional baseball) where high workinterdependence was an assumed constant. TheStudy 1 setting constrains interdependence in theconverse fashion, thereby enabling a direct test ofHypothesis 1.

Sample

The population for this study consisted of 3104trucking organizations that reported informationto the Interstate Commerce Commission (ICC)and were included in the 1993–94 TTS BlueBook of Trucking Companies (Blue Book ). Sincemotor carrier companies are generally small, it was

imperative that the sampled organizations have asufficient number of employees to have establishedformal policies for drivers and other employees.Thus, included companies were required to have atleast 30 total employees in either the 1991, 1992,or 1993 Blue Book data, be listed in the most recentversion (1993 calendar year), use company drivers(rather than ‘owner-operators’ exclusively), andstill be in business when contacted. The 1072 com-panies that met all relevant criteria constituted thefinal sample for the study. A 24-page questionnairewas mailed to the highest human resource man-ager in each company, and completed responseswere returned by 379, yielding a 36% (379/1072)response rate. All compensation measures in thequestionnaire refer to driver compensation prac-tices. Because of missing data, the analysis samplesizes varied and are reported in the tables.

Data sources

Three data sources were used: the key infor-mant questionnaire noted above and two archivalsources. The first archival source was the BlueBook. The Blue Book database is publicly avail-able and reports organizational and financial infor-mation that motor carriers file with the federalgovernment. The second archival source was theSAFER database. SAFER reports 2-year runningtotals of accident, inspection violations, and otherinformation that motor carriers file with the FederalHighway Administration—Office of Motor Car-riers (FHWA). Safety information on motor car-riers is publicly available from the FHWA. BlueBook information was obtained for the 1994 (theyear corresponding to the questionnaire adminis-tration) and 1995 (the year following) calendaryears, and SAFER information was obtained tocorrespond precisely to this 2-year time frame (i.e.,for 1994–95).

Measures: independent variables

Data for all independent variables were obtainedfrom the questionnaire.

Pay dispersion

Respondents reported how much (a) a new driverearned per year and (b) a senior driver earnedper year. The difference between the two numbersforms the measure.



Individual incentives

A list of four types of individual incentives, com-mon in the trucking industry and derived throughpilot testing and discussions with industry officials,was used for this measure. Respondents reportedthe extent to which these incentive systems (indi-vidual incentives tied to individual performance,on-the-spot bonuses for exceptional performance,lump-sum bonuses, and merit pay systems) wereused. Responses were obtained on a scale from 1(Not at all) to 5 (To a very great extent). The meanof the items forms the measure.

Measures: dependent variables

All three data sources—the questionnaire, the BlueBook and SAFER —were used to obtain informa-tion on the dependent variables.

Accident frequency ratio

This is the traditional bellwether safety measure inthe industry. It was operationalized as the numberof accidents per million miles driven. The measurewas calculated using information from the BlueBook and SAFER. Total highway miles drivenwere reported in the Blue Book for 1994 and 1995and were summed. Total accidents for 1994 and1995 (combined) were reported in SAFER. Totalaccidents for these years, divided by total highwaymiles for these years, and multiplied by 1,000,000yields the measure.

Out of service percentage

This measure was obtained from SAFER. Motorcarriers are subject to various random and regularinspections by federal agencies (e.g., U.S. Depart-ment of Transportation) and state transportationauthorities. Violators of federal or state regula-tions pertaining to equipment or drivers (e.g., logbook violations) are often removed from serviceuntil the problem is corrected. Our measure isspecific to driver performance, i.e., the frequencyof violations attributable to driver mistakes, anddoes not include factors beyond the driver’s con-trol (e.g., dilapidated equipment). It was opera-tionalized as the number of trucks/drivers takenout of service due to driver fault, divided bythe total number of inspections, and multipliedby 100.

Perceptual performance

This variable was operationalized as the extentto which driver performance in the focal com-pany compared to the performance of other driversin the industry. Performance was compared on14 dimensions in the questionnaire. These dimen-sions concerned controllable aspects of driverperformance: on-time deliveries, on-time pick-ups, consistent transit times, drivers’ friendli-ness to customers, drivers’ helpfulness to cus-tomers, drivers’ willingness to accommodate spe-cial customer needs, adherence to special ship-ping instructions, customer complaints concern-ing drivers, loss/damage history, ‘logging com-pliance,’ driver accident rates, fuel consumption,speed limit compliance, and traffic safety rulescompliance. Responses were obtained using 7-point response options ranging from ‘Ours aremuch worse’ (1) to ‘Ours are much better’ (7).The mean of the items forms the measure.

This perceptual measure of the aggregate perfor-mance of the driver workforce was derived throughextensive pilot testing and intensive discussionswith high-ranking industry executives. Items wereselected if drivers themselves, not factors beyondthe drivers’ control, were primarily responsiblefor performance variations. The measure is bench-marked against the performance of other driversin the industry, strengthening its validity (Delaneyand Huselid, 1996). Such key informant perfor-mance reports are often positively (moderately tostrongly) correlated with corresponding archivalmeasures (e.g., Dollinger and Golden, 1992; Pow-ell, 1992). It is particularly useful in the presentstudy since it is not susceptible to contaminationby factors beyond driver control (e.g., asset utiliza-tion, operating capacity, technology).

Measures: control variables

Interorganizational studies require control of fac-tors that may affect independent or dependentvariables. We controlled for several such vari-ables based on the literature dealing with per-formance, compensation, and the transportationindustry. Organizational characteristics—size(Terborg and Lee, 1984), age (Arthur, 1994), andunionization (Freeman and Medoff, 1984)—mayinfluence pay dispersion, compensation practices,and performance. An industry-specific control was


Pay Dispersion 497

carrier type, since pay dispersion and other com-pensation practices may vary across industry seg-ments (Perser, 1994).

Given the focus of the present study, several spe-cific controls were also necessary. One is the aver-age tenure of drivers. Human capital differences(e.g., training and experience; see Powell, Mont-gomery, and Cosgrove, 1994) could well influencepay dispersion creating a need to control for tenuredifferences across firms. Three compensation prac-tices—pay level, benefits level, and pay systemcommunication—were included as controls sincethey are related to performance and may influ-ence perceptions of compensation decisions (e.g.,Lawler, 1971). In addition, since our focus wasassessing incentive pay dynamics, it was also nec-essary to control for pay differentials as a result ofseniority (Lazear, 1981).

Of the organizational controls, size was opera-tionalized as the log of total number of employ-ees in the company, as reported by the respon-dent. Organizational age was operationalized asthe log of 1994 minus the founding year. Informa-tion on founding year was obtained from the BlueBook. Unionization was measured as the percentof drivers covered under a collective bargainingagreement, as reported by the respondent. Car-rier type was coded 1 for truckload firms and 0for less-than-truckload (LTL) firms. Tenure wasoperationalized as the percent of drivers who hadworked for the company for more than 24 months.Of the compensation controls, pay level was mea-sured as the average annual pay for a typicaldriver in the company. Benefits level was oper-ationalized as the mean percent of health insur-ance, disability, and life insurance premiums paidby the company. Pay system communication wasmeasured as a mean of two items with 7-pointagree/disagree response options. The items were‘Drivers know exactly what they need to do to getpay raises’ and ‘We make it clear to our driverswhat they must do to get more pay.’ Seniority-based pay was assessed with a single item with fiveresponse options concerning the extent to whichdifferences in pay rates across drivers were theresult of seniority.

Response bias check

It was possible to use archival data from the BlueBook and SAFER to compare responding and non-responding organizations. We chose a sample of

organizational and operating characteristics fromthe Blue Book (number of drivers, total fringe ben-efits cost, total highway miles driven, total wagespaid, average haul (in miles), total insurance costs,current assets, company age, tons per mile, andaverage load (in tons)) and the safety measuresavailable in SAFER as comparison variables. Fol-lowing Osterman (1994), we used logistic regres-sion to test for differences, coding a dichotomousdependent variable ‘1’ if a usable questionnairewas returned and ‘0’ if not. No independent vari-able was significant in the equation. Since missingdata reduced our analysis sample to 258, we alsoconducted a logistic regression comparing analy-sis cases (1) to all other cases (0) on the variablesnoted above. Again no independent variable wassignificant, indicating that response bias should notaffect our results.

STUDY 1 RESULTS

Table 1 shows descriptive statistics and correla-tions for all variables. Coefficient α reliabilitiesare shown in the main diagonal of Table 1 whereappropriate. As Table 1 shows, the use of indi-vidual incentives is moderately and significantlyrelated to pay dispersion (r = 0.23, p < 0.01),providing a level of confidence that the individualincentives create some pay dispersion.

The results in Table 2 contain the tests ofHypothesis 1. The first block in the hierarchicalregression consisted of control variables, the sec-ond block contained pay dispersion and individualincentives, and the last block contained the pay dis-persion by individual incentives product term. Theinteraction term was significant in each equation,but support for Hypothesis 1 is evident with onlya specific pattern of interaction. To investigate thepattern, the interactions were plotted for each ofthe three dependent variables using the standard-ized β’s and values of plus one (+1) and minusone (−1) standard deviation. The resulting inter-action patterns are shown in Figures 1–3.

Figure 1 depicts the interaction between paydispersion and individual incentives in predictingaccident frequency ratio. The figure is generallysupportive of Hypothesis 1. When the use of indi-vidual incentives is low, there is a strong positiverelationship between pay dispersion and accidentfrequency ratio (recall that a higher ratio indicates



Tabl

e1.

Stud

y1:

corr

elat

ions

and

desc

ript

ive

stat

istic

sfo

ral

lva

riab

les

Mea

nS.

D.

12

34

56

78

910

1112

1314

Con

trol

vari

able

s1.

Org

aniz

atio

nal

Size

4.72

1.06

#2.

Org

aniz

atio

nal

Age

3.57

0.89

−0.0

9#

3.U

nion

izat

ion

21.8

139

.64

−0.0

80.

18∗∗

#4.

Car

rier

Type

0.85

0.35

−0.1

9∗∗−0

.15∗

−0.2

8∗∗#

5.Te

nure

60.2

424

.99

−0.2

9∗∗0.

15∗

0.26

∗∗−0

.21∗∗

#6.

Pay

Lev

el34

882

5616

0.22

−0.0

10.

16∗

−0.1

4∗0.

13∗

#7.

Ben

efits

Lev

el68

.88

26.0

70.

010.

050.

30∗∗

−0.1

20.

18∗∗

0.11

#8.

Pay

Syst

emC

omm

unic

atio

n4.

771.

420.

09−0

.15∗

−0.2

0∗∗0.

11−0

.20∗∗

0.05

−0.1

5∗(0

.71)

9.Se

nior

ity-b

ased

Pay

2.92

1.57

0.22

∗∗−0

.08

−0.0

9−0

.05

−0.2

8∗∗−0

.01

−0.1

4∗0.

26∗∗

#

Inde

pend

ent

vari

able

s10

.Pa

yD

ispe

rsio

n10

410

6646

0.13

∗−0

.03

0.05

0.08

−0.0

30.

29∗∗

0.07

0.21

∗∗0.

31∗∗

#11

.In

divi

dual

Ince

ntiv

es1.

920.

770.

05−0

.09

−0.3

2∗∗0.

19∗∗

−0.1

5∗0.

03−0

.15∗

0.24

∗∗0.

22∗∗

0.23

∗∗(.

73)

Wor

kfor

cepe

rfor

man

ceva

riab

les

12.

Acc

iden

tFr

eque

ncy

Rat

io0.

950.

740.

05−0

.13∗

−0.0

8−0

.03

−0.0

8−0

.07

0.03

0.01

0.00

0.09

−0.0

2#

13.

Out

ofSe

rvic

ePe

rcen

tage

5.08

4.32

−0.0

5−0

.11

−0.2

6∗∗0.

29∗∗

−0.1

9∗∗−0

.09

−0.2

0∗∗0.

050.

090.

010.

070.

09#

14.

Perc

eptu

alPe

rfor

man

ce5.

220.

70−0

.03

0.16

∗0.

03−0

.08

0.23

∗∗0.

20∗∗

0.10

0.13

∗−0

.01

−0.0

20.

07−0

.06

−0.2

7∗∗(0

.90)

∗∗p

<0.

01;

∗ p<

0.05

.N

=25

8.C

oeffi

cien

tα

relia

bilit

yes

timat

esar

ere

port

edin

the

diag

onal

whe

reap

prop

riat

e.


Pay Dispersion 499

Table 2. Study 1: hierarchical regression results of the relationship of individual incentives andpay dispersion of performance

Block Accidentfrequency

ratio

Out-of-service

percentage

Perceptualperformance

Control Variables † † †

Independent variablesPay Dispersion 0.12 −0.02 −0.19∗∗

Individual Incentives −0.01 0.02 −0.01

InteractionPay Dispersion × individual incentives −0.15∗ −0.15∗ 0.13∗

Total R2 0.084∗ 0.194∗∗ 0.238∗∗

�R2 —interaction step 0.019∗ 0.017∗ 0.013∗

† Regression coefficients for control variables omitted.∗∗ p < 0.01; ∗p < 0.05. N = 258.Final equation standardized regression coefficients are reported in the columns.

Figure 1. Study 1: interaction between pay dispersionand individual incentives in predicting accident frequency

ratio

poorer performance). By contrast, the relationshipis negative when the use of individual incentivesis high. Notable in this context is that the poorerlevels of workforce performance, i.e., the highestaccident rates, are seen when pay is dispersed, butindividual incentives are low, or when pay disper-sion is low and individual incentives are stronglyused. A similarly supportive pattern of findings isevident in Figure 2 for out-of-service percentage.In this case, there is an almost symmetric cross-ing pattern where poorer workforce performanceoccurs at the low incentives–high dispersion andhigh incentives–low dispersion plots points. Con-versely, more favorable performance levels occurwhen pay dispersion is accompanied by individual

Figure 2. Study 1: interaction between pay dispersionand individual incentives in predicting out-of-service

percentage

incentives or when neither is present. Figure 3shows the pattern of the interaction for perceptualperformance. In this case, there is a prevailing neg-ative relationship between pay dispersion and per-formance regardless of the use of individual incen-tives. Consistent with expectations, the relationshipis more strongly negative when individual incen-tives use is low. Inconsistent with expectations, therelationship between pay dispersion and percep-tual performance is slightly negative when the useof individual incentives is high. Taken together,the results show strong support for Hypothesis1 for accident frequency ratio and out-of-servicepercentage and partial support with respect to per-ceptual performance.



Figure 3. Study 1: interaction between pay dispersionand individual incentives predicting perceptual

performance

STUDY 2 METHOD

Context, extensions, and methodologicalimprovements

Study 2 was conducted in the concrete pipe indus-try among member organizations of the AmericanConcrete Pipe Association (ACPA). The popula-tion for the study was the membership of theACPA, currently about 65 organizations represent-ing about 202 different concrete pipe plants acrossthe United States and Canada. Production workerscomprise the key employee group in this industry,accounting for roughly 70 percent of total employ-ees in our sample.

This setting allowed for a more thorough inves-tigation of the pay dispersion issue. In particular,plants varied on pay dispersion, individual incen-tives, and work interdependence, which made pos-sible a test of all three hypotheses. In addition,two aspects of Study 1 in particular warrantedmethodological improvements that could be ad-dressed in Study 2. First, the measure of pay dis-persion in Study 1 was based on only two datapoints per organization; i.e., it did not provideinformation on the actual distribution, but ratherthe range, of pay. Three different measures of paydispersion, two of which estimate the actual dis-tribution of pay, could be gleaned from the Study2 database. Also, Study 1 used perceived work-force performance as an outcome, but a commonproductivity metric (labor hours per ton; Arthur,1994) was also available in the concrete pipe study.

Sample

After extensive prior contacts, plant managers ofeach of the 202 plants were mailed a 44-page

survey dealing with general management, humanresource management, production and operation,and effectiveness issues. The plant manager wasinstructed to complete the survey himself/herselfor to have the most knowledgeable people inthe plant complete relevant sections. Completedresponses were obtained from 141 plant managers,representing a 71 percent response rate (141/202).The plant survey constitutes the sole data sourcein this study.

Measures: independent variables

Pay dispersion

Three alternative operationalizations of pay disper-sion were available in this data set. The analyseswere repeated using each of these three measures.Since it was practically impossible to obtain a com-plete pay distribution from each organization, plantmanagers reported instead the starting hourly rate,the hourly rate at the middle of the pay range,and the top hourly rate for production employ-ees in the facility. Respondents also reported theamount of overtime that production employees typ-ically worked per week. The pay rates (includ-ing estimated overtime pay) were annualized tocreate an estimated annual pay for each of thethree categories (minimum, middle, and maxi-mum). Respondents also reported the approximatenumber of production employees earning each ofthese three rates. On average, respondents reportedthe estimated pay of about 90 percent of theirtotal production employees, building confidencethat the derived estimates accurately reflect the paydistribution.

The first operationalization (high minus low )corresponds to the measure used in Study 1. Itis simply defined as the highest pay level minusthe lowest (or starting) pay level. The second isthe coefficient of variation (Pfeffer and Langton,1993), defined as the standard deviation dividedby the mean. In a review of measures of disper-sion, Allison (1978) reported that the coefficientof variation was generally preferable to the highminus low measure. The third measure was anestimated gini coefficient, the commonly used mea-sure of income inequality in economics (Bloom,1999; Donaldson and Weymark, 1980). The max-imum gini coefficient is 1.0, indicating absoluteinequality of pay; the minimum is zero, indicatingcomplete pay compression. As noted, the precise


Pay Dispersion 501

distribution of pay was unknown and could onlybe estimated by the number of individuals reportedto be making near the minimum, middle, and high-est pay levels. As a consequence, the coefficient ofvariation and gini coefficient estimates are under-estimated in this study.1

Individual incentives

Pilot tests and discussions with industry officialswere again used to develop a list of individualincentive practices applicable to the sample. Themean of seven items—individual incentives tied toindividual performance, on-the-spot incentives forexceptional performance, merit pay systems, lump-sum salary increases, cash incentives for qualityimprovement suggestions, individual pay increasesbased on individual quality performance or out-comes, and bonuses or lump-sum payments basedon quality—was used. Responses were obtainedon a 5-point scale ranging from 1 (Not at all) to 5(To a very large extent).

Work interdependence

The production of concrete pipe must be accom-plished sequentially: mixing, casting, and cur-ing. Since sequential interdependence is thereforeconstant across the industry, a measure of workinterdependence that captured interdependence lev-els beyond the production sequence was needed.To achieve this, we assessed the extent of use ofself-managed teams. Key characteristics of self-managed work teams include groups of individ-uals that: (1) regulate behavior on interdepen-dent tasks; (2) engage in face-to-face interaction;and (3) work on interrelated tasks to make a prod-uct or deliver a service (Cohen and Ledford,1994). Thus, use of self-managing work teams cap-tures incremental interdependence yielded throughself-regulation and control over a changing workenvironment. The item assessed the percent ofproduction employees currently involved in self-managing work teams. Response options were1 (None), 2 (Almost none, 1–20%), 3 (Some,21–40%), 4 (About half, 41–60%), 5 (Most,

1 A fourth measure of pay dispersion, the ratio of the minimumpay to maximum pay, was also used by Bloom (1999). In thepresent study the correlation between the ratio measure and thehigh minus low measure was 0.96. The reported results weresubstantively identical using either measure. To conserve space,only the high minus low results are reported.

61–80%), 6 (Almost all, 81–99%), and 7 (All,100%).

Measures: dependent variables

Labor hours per ton

This variable was operationalized as the numberof labor hours worked by production employees inthe focal year divided by the tons of concrete prod-ucts produced in the same year. It was derived bymultiplying the average number of full-time pro-duction employees in the plant by 2008 hours (thenumber of working hours in a typical year; alsoadjusted for average overtime hours reported bythe respondent) and dividing by the tons of con-crete products produced. With this operationaliza-tion, fewer labor hours per ton is indicative of highproductivity. Labor hours per ton is a commonproductivity measure in heavy manufacturing, andalthough an independent or archival measure wasnot available, previous research demonstrates thatkey informant reports of labor hours per ton cor-respond well with archival measures (e.g., Arthur,1994; Hogan, 1987). Since the key informant inthis study was the plant manager, presumably witha vested interest in this information, the reports arelikely to be accurate estimates.

Lost-time accidents

This variable was operationalized as the naturallog of the number of lost-time accidents in theplant in the last 5 years.2 As with labor hoursper ton, fewer lost-time accidents indicates betterworkforce performance.

Perceptual performance

The perceptual performance measure was designedto parallel the measure used in Study 1, i.e.,one which assessed controllable aspects of pro-duction employee performance in the industry. It

2 The use of a 5-year time frame introduces inconsistency into thetiming of the independent and dependent variables. The decisionto include a longer time frame was the result of consultationwith industry experts. The relatively small size of concrete pipeplants results in a few number of accidents per year. Becauseaccidents are rare, a 1-year accident total is likely to yielda biased estimate of workforce performance. A 5-year periodallows for the averaging out of a peculiar year and, accordingto knowledgeable industry insiders, provides a better indicationof safety performance.



was developed after extensive pilot testing, severalsite visits, and extensive contacts with industryofficials. The resulting 5-item scale assessed thepercent of pipe meeting technical specifications,raw material waste, percent of pipe needing to bereworked, percent of production scrapped, and pro-duction efficiency. Responses were obtained using7-point response options ranging from ‘Ours aremuch worse’ (1) to ‘Ours are much better’ (7).The mean of the items forms the measure. As withthe Study 1 measure, responses were benchmarkedagainst the industry.

Measures: control variables

As in Study 1, several organizational and indus-try controls were used. Size (natural log of totalemployees), age (natural log of years since found-ing date), and unionization (1 = unionized plant, 0= no union) were included as organizational con-trols. An important consideration in the concretepipe industry is corporate dependence, which caninfluence compensation policies and performance(Dean and Snell, 1991; Majchrzak, 1988). Corpo-rate dependence was controlled by coding plantsthat were part of multiplant corporations 1 andfree-standing plants 0. An average tenure variablewas not available among these data. As a proxy,the total turnover rate was included. Including theturnover rate (as the converse of average tenure)should control for human capital and performancedifferentials. Total turnover rate was calculated asthe number of production workers who quit orwere fired divided by the total number of produc-tion workers. Pay level (average annualized payfor production workers), benefits level (mean per-cent of health insurance, disability, and life insur-ance premiums paid by the company), and themean of a three-item pay system communicationsscale (e.g., ‘We make it clear to our employeeswhat they must do to get more pay’) were alsoincluded. Seniority-based pay was assessed with asingle item with five response options concerningthe extent to which differences in pay rates acrossdrivers were the result of seniority.

STUDY 2 RESULTS

Table 3 shows descriptive statistics and correla-tions for all variables. Coefficient α reliabilitiesare shown in the main diagonal of Table 3 where

appropriate. The table shows a strong correspon-dence among the pay dispersion measures. Highminus low is strongly related to the coefficient ofvariation (r = 0.58, p < 0.01) and gini coefficient(r = 0.61, p < 0.01), while the coefficient of vari-ation and gini coefficient are even more highly cor-related (r = 0.75, p < 0.01). These findings cor-respond to previous research (e.g., coefficient ofvariation and gini coefficient were correlated 0.71in the Bloom, 1999, study) and provide evidenceof the construct validity of the measure of dis-persion used in Study 1. The individual incentivesvariable is significantly related to the coefficient ofvariation (r = 0.21, p < 0.01) and gini coefficient(r = 0.19, p < 0.05) and marginally related to thehigh minus low measure (r = 0.17, p < 0.10).Individual incentives and work independence arealso significantly related (r = 0.32, p < 0.01).

The first series of analyses, shown in Table 4,was designed to test Hypotheses 1 and 2. Inthese regressions, the two-way interaction terms(pay dispersion × individual incentives and paydispersion × work interdependence) were enteredalternatively in the final steps after controls and themain effects of the substantive variables. Hypoth-esis 3 was tested in the second set of analyses(Table 5). Here, the three-way interaction term wasentered in the final step after controls, main effects,and the two-way terms. Changes in R2 and stan-dardized regression coefficients (β) were exam-ined. These analyses were repeated for each ofthree measures of pay dispersion (high minus low,coefficient of variation, and gini coefficient).

As Table 4 shows, the pay dispersion × indivi-dual incentives interaction term (test of Hypothesis1) is significant in two of the labor hours per tonand all three of the lost-time accidents equations,but fails to reach significance in any of the percep-tual performance equations. The pattern of theseinteractions generally conforms to the predictedform (and the results from Study 1), i.e., a cross-ing pattern is evident and the highest level of laborhours per ton and lost-time accidents (poorer per-formance) is found when pay dispersion is highbut use of incentives is low. Overall, the resultsin Table 4 provide moderate to strong support forHypothesis 1.

The pay dispersion × work interdependence in-teraction step (Hypothesis 2) was also consistentlysignificant across all three outcomes variables. Theinteraction term is significant with labor hours perton as the dependent variable and coefficient of


Pay Dispersion 503

Tabl

e3.

Stud

y2:

corr

elat

ions

and

desc

ript

ive

stat

istic

sfo

ral

lva

riab

les

Mea

nS.

D.

12

34

56

78

910

1112

1314

1516

17

Con

trol

vari

able

s1.

Org

aniz

atio

nal

Size

3.22

0.62

#2.

Org

aniz

atio

nal

Age

3.45

1.04

−0.0

6#

3.U

nion

izat

ion

0.36

0.49

0.23

∗∗0.

14#

4.C

orpo

rate

Dep

ende

nce

0.87

0.34

−0.1

5−0

.01

0.00

#

5.To

tal

Tur

nove

rR

ate

0.18

0.23

−0.0

2−0

.11

−0.2

5∗∗0.

10#

6.Pa

yL

evel

1896

842

850.

17∗

0.08

0.60

∗∗−0

.03

−0.2

4∗∗#

7.B

enefi

tsL

evel

74.3

443

.58

−0.0

90.

11−0

.22∗∗

−0.0

40.

13−0

.09

#8.

Pay

Syst

emC

omm

unic

atio

n4.

171.

280.

00−0

.08

−0.3

2∗∗0.

120.

22∗∗

−0.3

7∗∗0.

26∗∗

(0.7

2)

9.Se

nior

ity-b

ased

pay

2.73

1.33

−0.0

2−0

.17∗

−0.1

5−0

.10

0.08

−0.2

6∗∗0.

090.

10#

Pay

disp

ersi

onva

riab

les

10.

Hig

hM

inus

Low

4799

3112

0.14

−0.0

3−0

.01

−0.0

5−0

.05

0.25

∗∗0.

100.

080.

02#

11.

Coe

ffici

ent

ofV

aria

tion

0.07

0.03

−0.3

2∗∗−0

.02

−0.3

0∗∗0.

010.

10−0

.20∗∗

0.18

∗0.

23∗∗

0.08

0.58

∗∗#

12.

Gin

iC

oeffi

cien

t0.

110.

040.

11−0

.03

−0.2

2∗∗−0

.08

0.12

−0.1

10.

100.

25∗∗

0.10

0.61

∗∗0.

75∗∗

#

Ince

ntiv

esan

din

terd

epen

denc

e13

.In

divi

dual

Ince

ntiv

es1.

860.

600.

110.

03−0

.27∗∗

−0.0

40.

20∗∗

−0.2

3∗∗0.

30∗∗

0.31

∗∗0.

080.

170.

21∗∗

0.19

∗(.

75)

14.

Wor

kIn

terd

epen

denc

e1.

450.

840.

110.

24∗∗

−0.0

4−0

.04

−0.0

9−0

.04

0.08

0.18

∗0.

040.

060.

080.

32∗∗

−0.0

5#

Wor

kfor

cepe

rfor

man

ceva

riab

les

15.

Lab

orH

ours

per

Ton

1.43

0.62

0.28

∗∗−0

.05

0.31

∗∗−0

.05

−0.1

00.

11−0

.17

−0.1

70.

060.

04−0

.12

0.07

−0.1

4−0

.05

#

16.

Los

t-tim

eA

ccid

ents

1.56

1.06

0.50

∗∗−0

.03

0.27

∗∗−0

.24∗∗

0.05

0.16

0.11

0.08

0.01

0.05

−0.1

30.

01−0

.15

0.08

0.26

∗∗#

17.

Perc

eptu

alPe

rfor

man

ce4.

670.

790.

03−0

.07

0.02

−0.0

30.

000.

12−0

.04

0.17

−0.1

00.

13−0

.02

0.01

0.12

−0.2

3∗∗−0

.04

−0.0

2(0

.90)

∗∗p

<0.

01;

∗ p<

0.05

;N

=11

3.C

oeffi

cien

tα

relia

bilit

yes

timat

esar

ere

port

edin

the

diag

onal

whe

reap

prop

riat

e.



variation as the dispersion measure, with all threedispersion measures in the lost-time accidents, andfor the high minus low and gini coefficient mea-sures when perceptual performance is the depen-dent variable. Again, plots of these interactionsgenerally conform to expectations; i.e., the poorestperformance (highest labor hours per ton, highestlost-time accidents, and lowest perceptual perfor-mance) is evident when pay dispersion and workinterdependence are both high. Moderately strongsupport therefore is found for Hypothesis 2.

The critical test, of course, is that of Hypothesis3, which is essentially a synthesis of the some-what contradictory predictions from Hypotheses 1and 2. Table 5 shows the results for these tests.The three-way interaction among pay dispersion,individual incentives, and work interdependence issignificant in four of the nine equations. In partic-ular, the three-way interaction is significant in allthree labor hours per ton equations and, in the caseof accidents, when high minus low was the pay dis-persion measure. Where the three-way interactionsare significant, these effects supersede the signifi-cant two-way interactions in Table 4.

To support Hypothesis 3 fully, the pattern ofthe interactions must be consistent with expecta-tions. Plots of the significant three-way interac-tions (Figures 4 and 5) provide substantial sup-port for the rationale grounding Hypothesis 3,but they also reveal a more complex pattern thanexpected. As an example, the three-way interac-tion among pay dispersion (gini coefficient), indi-vidual incentives, and work interdependence isshown in Figure 4. The plot was split by lev-els (low vs. high) of work interdependence. Theleft-hand panel of Figure 4 depicts the interactiverelationship between pay dispersion and individ-ual incentives when work interdependence is low.The expected fully crossing pattern of results, i.e.,similar to the patterns found in Study 1, failedto materialize in these analyses. Instead, there isessentially no relationship between pay dispersionand labor hours per ton when work interdepen-dence is low. By contrast, the right-hand panelshows the relationships when work is highly inter-dependent. Here, a prevailing positive relationshipbetween pay dispersion and labor hours per ton(recall that higher labor hours per ton indicatespoorer performance) is apparent. Interestingly, thisrelationship is stronger when the use of individualincentives is low, and is attenuated somewhat bythe use of individual incentives.

Figure 5 shows the significant three-way inter-action with accidents as the dependent variable.A variation of the expected pattern emerges whenwork interdependence is low. That is, pay disper-sion (measured as high minus low) is strongly andpositively associated with accidents, i.e., poorerperformance, in the absence of individual incen-tives. When incentive use is high, there is a slightlynegative, but nonsignificant, relationship betweenpay dispersion and accidents. On the other hand,when work interdependence is high, there is a pre-vailing, but weak, negative pattern between paydispersion and accidents and no significant differ-ential effects of individual incentives.

To summarize, Hypothesis 3, the integration ofpreviously distinct theoretical predictions, receivesmoderate support in Study 2, but the results var-ied according to a more complex pattern thanthat predicted. In particular, for labor hour perton (see Figure 4), the expected pattern of poorerperformance emerged when pay dispersion andwork interdependence were high; the observa-tion that individual incentives attenuate the paydispersion → performance relationship in situa-tions of high interdependence is unexpected, asis an observation of the lack of strongly differ-ential relationships when interdependence is low.With respect to accidents, differential effects wereevident when work interdependence was low andwere generally consistent with expectations, butnot when work interdependence was high.

DISCUSSION

This study focused on the specific dynamics of akey element of strategy implementation, attempt-ing to reconcile seemingly conflicting perspectivesabout the effects of pay dispersion on organiza-tional performance. We argued that dispersion perse is neither functional nor dysfunctional; rather,situational contingencies determine the strategiceffectiveness of dispersion (or lack thereof). Wepredicted that dispersion is more likely to be ben-eficial when individual incentives are used in con-junction with it; conversely, dispersion is likely tobe ineffective when used in the context of highwork interdependence. Tests of these predictionsin two industrial settings using a range of perfor-mance measures provides interesting insights aboutthe dispersion phenomenon. Our results show thatdispersion is indeed more effective when it exists


Pay Dispersion 505

Tabl

e4.

Stud

y2:

hier

arch

ical

regr

essi

onan

alys

esw

ithal

tern

ativ

em

easu

res

ofpa

ydi

sper

sion

—te

sts

ofH

ypot

hese

s1

and

2

Blo

ckL

abor

hour

spe

rto

nL

ost-

time

acci

dent

sPe

rcep

tual

perf

orm

ance

Hig

hm

inus

low

C.

ofva

riat

ion

Gin

ico

effic

ient

Hig

hm

inus

low

C.

ofva

riat

ion

Gin

ico

effic

ient

Hig

hm

inus

low

C.

ofva

riat

ion

Gin

ico

effic

ient

Con

trol

vari

able

s†

††

††

††

††

††

††

††

††

†

Inde

pend

ent

vari

able

sPa

yD

ispe

rsio

n—

−0.0

50.

020.

17∗

0.05

0.07

0.06

Hig

hM

inus

Low

Pay

Dis

pers

ion

—0.

25∗∗

0.03

0.13

0.13

−0.0

5−0

.03

Coe

ffici

ent

ofV

aria

tion

Pay

Dis

pers

ion

—0.

21∗

0.16

0.01

0.06

−0.0

6−0

.05

Gin

iC

oeffi

cien

tIn

divi

dual

Ince

ntiv

es−0

.17

−0.0

80.

120.

090.

140.

03−0

.23∗

−0.2

1∗−0

.29∗∗

−0.2

0∗−0

.29∗∗

−0.2

7∗∗0.

04−0

.03

0.02

0.03

0.08

0.01

Wor

k Inte

rdep

ende

nce

−0.0

40.

010.

06−0

.06

0.05

−0.0

30.

030.

120.

19∗

−0.0

70.

24∗∗

−0.0

3−0

.23∗

−0.1

5−0

.24∗∗

−0.2

8∗∗−0

.21∗

−0.1

2

Two-

way

inte

ract

ion

Pay

Dis

pers

ion

×−0

.05

−0.2

5∗∗−0

.20∗

−0.1

7∗−0

.16∗

−0.1

8∗0.

090.

130.

11In

divi

dual

Ince

ntiv

es(H

1)Pa

yD

ispe

rsio

n×

0.07

0.21

∗0.

140.

25∗∗

0.19

∗0.

24∗

−0.2

1∗−0

.08

−0.1

8∗

Wor

kIn

terd

epen

denc

e(H

2)

Tota

lR

20.

194∗

0.20

0∗0.

278∗∗

0.27

3∗∗0.

261∗∗

0.23

6∗∗0.

338∗∗

0.35

3∗∗0.

312∗∗

0.31

3∗∗0.

339∗∗

0.34

6∗∗0.

212∗∗

0.23

1∗∗0.

224∗

0.21

4∗0.

192∗

0.20

5∗

�R

2—

inte

ract

ion

step

0.00

80.

014

0.05

6∗∗0.

051∗∗

0.04

7∗∗0.

022

0.02

4∗0.

039∗∗

0.02

4∗0.

025∗

0.02

4∗0.

031∗

0.01

00.

029∗

0.01

90.

009

0.01

00.

023∗

†R

egre

ssio

nco

effic

ient

sfo

rco

ntro

lva

riab

les

omitt

ed.

∗∗p

<0.

01;

∗p

<0.

05.N

=11

0.Fi

nal

equa

tion

stan

dard

ized

regr

essi

onco

effic

ient

sar

ere

port

edin

the

colu

mns

.T

hefir

stco

lum

nin

each

set

repo

rts

the

resu

ltsw

ithin

divi

dual

ince

ntiv

es,

and

the

seco

ndco

lum

nth

ere

sults

with

inte

rdep

ende

nce.



Tabl

e5.

Stud

y2:

hier

arch

ical

regr

essi

onan

alys

esw

ithin

cent

ives

,in

terd

epen

denc

e,an

dal

tern

ativ

em

easu

res

ofpa

ydi

sper

sion

—te

sts

ofH

ypot

hesi

s3

Blo

ckL

abor

hour

spe

rto

nL

ost-

time

acci

dent

sPe

rcep

tual

perf

orm

ance

Hig

hm

inus

low

Coe

ffici

ent

ofva

riat

ion

Gin

ico

effic

ient

Hig

hm

inus

low

Coe

ffici

ent

ofva

riat

ion

Gin

ico

effic

ient

Hig

hm

inus

low

Coe

ffici

ent

ofva

riat

ion

Gin

ico

effic

ient

Con

trol

vari

able

s†

††

††

††

††

Inde

pend

ent

vari

able

sPa

yD

ispe

rsio

n—

0.16

0.07

0.02

Hig

hM

inus

Low

Pay

Dis

pers

ion

—0.

28∗∗

0.22

∗−0

.08

Coe

ffici

ent

ofV

aria

tion

Pay

Dis

pers

ion

—0.

34∗∗

0.03

−0.1

1G

ini

Coe

ffici

ent

Indi

vidu

alIn

cent

ives

−0.1

20.

090.

09−0

.20

−0.3

5∗∗−0

.24∗

0.09

0.13

0.12

Wor

kIn

terd

epen

denc

e0.

090.

010.

010.

060.

170.

05−0

.31∗∗

−0.3

1∗∗−0

.18

Two-

way

inte

ract

ions

Pay

Dis

pers

ion

×0.

04−0

.18∗

−0.2

6∗∗0.

010.

060.

010.

050.

160.

15In

divi

dual

Ince

ntiv

esPa

yD

ispe

rsio

n×

0.18

0.36

∗∗0.

42∗∗

−0.2

9∗∗−0

.10

−0.1

2−0

.21∗

−0.0

9−0

.22∗

Wor

kIn

terd

epen

denc

eIn

divi

dual

Ince

ntiv

es×

−0.0

8−0

.03

−0.0

40.

19∗

0.10

0.18

0.02

0.05

−0.0

7W

ork

Inte

rdep

ende

nce

Thr

ee-w

ayin

tera

ctio

nPa

yD

ispe

rsio

n×

−0.2

3∗−0

.28∗∗

−0.3

0∗∗0.

22∗

0.04

0.07

0.06

−0.1

4−0

.02

Indi

vidu

alIn

cent

ives

×W

ork

Inte

rdep

ende

nce

(H3)

Tota

lR

20.

244∗

0.46

0∗∗0.

419∗∗

0.41

4∗∗0.

310∗∗

0.35

6∗∗0.

248∗∗

0.25

6∗∗0.

248∗∗

�R

2—

two-

way

inte

ract

ion

step

0.02

70.

184∗∗

0.14

9∗∗0.

071∗∗

0.02

10.

037

0.04

40.

038

0.06

6∗

�R

2—

thre

e-w

ayin

tera

ctio

nst

ep0.

031∗

0.05

4∗∗0.

056∗∗

0.02

9∗0.

001

0.00

40.

002

0.01

30.

000

†R

egre

ssio

nco

effic

ient

sfo

rco

ntro

lva

riab

les

omitt

ed.

∗∗p

<0.

01;

∗p

<0.

05.N

=11

0.Fi

nal

equa

tion

stan

dard

ized

regr

essi

onco

effic

ient

sar

ere

port

edin

the

colu

mns

.


Pay Dispersion 507

Figure 4. Study 2: Three-way interaction among pay dispersion (gini coefficient), individual incentives, and workinterdependence in predicting labor hours per ton

Figure 5. Study 2: Three-way interaction among pay dispersion (high minus low), individual incentives, and workinterdependence in predicting accidents

in conjunction with individual incentives and that,at least for some performance measures, disper-sion is less effective when used in interdepen-dent settings. In addition, the results indicate arather complex pattern of interrelationships amongpay dispersion, individual incentives, work interde-pendence, and organizational effectiveness. Theseissues are discussed below.

Both studies supported the hypothesized interac-tion between the use of individual incentives andpay dispersion in predicting measures of effective-ness. This finding spanned measures of dispersion(in Study 2) and measures of organizational perfor-mance (in both studies). We argued that dispersionis functional only when it is attributable to legiti-mate sources, and that basing pay on performance

is normatively viewed as legitimate. This argumentis supported. Performance is lower when pay dis-persion occurs in the absence of incentives thanwhen it occurs in the presence of incentives. Theconsistency of this effect underscores the dan-gers of widely dispersed pay without normativelyacceptable reasons. The problems with politicallybased pay policies have been discussed before(e.g., Gupta and Jenkins, 1996). As Pfeffer andLangton note, pay dispersion ‘does not necessarilyimply that a pay-for-performance scheme exists.Dispersed wages can result from the operation offavoritism or rewards for other things than per-formance’ (Pfeffer and Langton, 1993: 387). Thepresent results suggest that illegitimate sources ofpay dispersion may hurt not just employee attitudes



but organizational performance as well. We couldnot test this proposition directly, but it is a fruitfularea of inquiry.

The second hypothesis predicted that pay disper-sion in interdependent work contexts is dysfunc-tional. This hypothesis received moderately strongsupport in Study 2 across all three measures of per-formance and with varying degrees of significanceacross three alternative measures of pay dispersion.These findings lend credence to the argument thatdispersion can foster competition and lack of coop-eration—dynamics that are diametrically opposedto the goals of interdependence.

The moderating effects of individual incentivesand work interdependence when taken in isola-tion are illuminating. But arguably even moreimportant are the results of Hypothesis 3 tests,which explored the interaction of both variableswith pay dispersion simultaneously in predictingwork performance. In several cases, especiallyin the labor hour per ton equations, the signif-icant three-way interaction superseded the two-way interaction effects. Overall, our predictionsin Hypothesis 3 are supported to some extent, buta more complex picture than predicted emergesfrom the results. Particularly interesting in thisregard is the right panel of Figure 4. When inter-dependence is high, pay dispersion is negativelyrelated to performance, but this negative effectis weaker when individual incentives are usedthan when they are not! This effect occurs despitethe fact that the use of individual incentives isinconsistent with the goals of interdependence(e.g., Shaw, Duffy, and Stark, 2000). These resultsunderscore the importance of basing pay disper-sion on legitimate grounds. As argued above it islikely, that individual incentives supply a legiti-mate justification for dispersion. The inconsistencybetween interdependence and individual incentivesmay be less corrosive for performance than is dis-persion created through illegitimate avenues. Thatis, a good justification for dispersion, despite itsincongruence, may be more acceptable than incon-sistency accompanied by no legitimate justifica-tion. An individual incentive system may ame-liorate deleterious effects when work is interde-pendent by providing a rationale for pay dif-ferentials. This explanation is particularly com-pelling in light of previous results showing broad-ranging and severe dysfunctional effects of ille-gitimate or unjust systems (e.g., Sheppard et al.,1992; Skarlicki and Folger, 1997). Nonetheless,

this explanation is speculative, and should beaddressed systematically in the future.

The complex dynamics regarding Hypothesis3 also highlight a critical dilemma for managersas they attempt effective strategy implementation.Managerially, a seemingly valid justification (useof individual incentives) for an organizational out-come (pay dispersion) is better than no justifica-tion, even when the justification is inconsistentwith work design, or more broadly, the objectivesof the organization. Since pay practices do notalways accompany work design changes hand inhand (see Snell and Dean, 1994), it is likely thatstrategic decision-makers face such dilemmas ona regular basis. At a conceptual level, researcherstend to view congruence issues in isolation (e.g.,certain pay practices are more or less congruentwith different work designs). They rarely con-sider the decision-making constraints that man-agers face on a daily basis and the satisfying,but not completely satisfying, solutions that oftenresult. It quickly becomes a particularly complextheoretical task, then, to identify when inconsis-tent practices buffer (and when they exacerbate)the negative effects of already difficult situations.But complexity may be necessary to provide anadequate description of organizational dynamics,as is the use of multitheoretical approaches to dis-entangle such complexities; one theory in isolationis unlikely to explain the gamut of effects (Lewisand Grimes, 1999).

The implications of this complexity for strat-egy implementation are further highlighted whenour results are set against those of Bloom (1999).In Bloom’s sample of professional baseball teams,work interdependence was high and relatively con-stant. The study reported a negative relationshipbetween pay dispersion and workforce perfor-mance. Likewise, in our research, Study 2 showedany main significant main effects of pay dispersionon performance to be in the same direction: poorerperformance was evident when dispersion washigh. In fact, Study 1, where work interdependencewas low, also showed a negative main effect on oneperformance dimension. Do our results, combinedwith those of Bloom (1999), then suggest a largelynegative relationship between dispersion and per-formance? Actually not. Overall, the pattern of oursignificant main effects was inconsistent, indicat-ing a more complex higher-order relationship thanwould be inferred from simple main effects. TheStudy 1 results, combined with tests of Hypothesis


Pay Dispersion 509

3, clarify the issue. The use of individual incentivesand dispersed wages is associated with better per-formance in low interdependence settings. In highinterdependence settings, wage dispersion is likelyto have negative effects, but these negative effectsare buffered by the use of legitimate dispersion-creating practices such as individual incentives. Inessence, Bloom’s (1999) findings are reinforced byour results. At the same time, his findings containonly a partial picture; our results bring the pictureinto clearer focus.

This study is rare in that it focused on hori-zontal, lateral, or intraclass pay dispersion. Albeitargued otherwise elsewhere (e.g., Cowherd andLevine, 1992), there is strong justification for thefact that horizontal pay dispersion has the mostbroad-ranging effects on organizational perfor-mance (Baron and Pfeffer, 1994). Horizontal paydistributions hold constant many potentially con-founding factors (e.g., differences in status, socialclass, job titles) that could reasonably explainvariations in pay levels. They highlight perfor-mance, justice, and legitimacy issues among rel-evant comparison individuals. As Bloom (1999)notes, individuals may tolerate vertical dispersionmore readily than horizontal dispersion since dif-ferences in job requirements, prestige, social sta-tus, etc., are acceptable reasons for the higherpay of higher-ranking individuals. But differencesin pay among individuals at the same organiza-tional level hit closer to home. As suggested above,justifiable differences (especially when work isindependent) facilitate higher performance, butunplanned or illegitimate differences endanger per-formance and outcomes. A useful extension of thisresearch is a comparison of the benefits and diffi-culties of vertical and horizontal dispersion.

Our results reinforce the idea that consistencybetween an organization’s management and humanresource management philosophy on the one handand the structure of the pay system on the otheris of vital importance.3 Organizational design isnot necessarily consistent with the pay systeman organization uses; the consequent mismatchesare detrimental to workforce performance in manyinstances. Of course, this study focused only onpay system dynamics. Still, it has broader impli-cations. The strategic human resource manage-ment literature (e.g., Arthur, 1994; Delery andDoty, 1996; Huselid, 1995; Ichniowski, Shaw, and

3 We thank an anonymous reviewer for these suggestions.

Prennushi, 1997; Youndt et al., 1996) emphasizesthe value of human resource ‘bundles’ in pro-moting organizational outcomes. The implicit orexplicit assumption is that the bundles must beinternally consistent, and this approach is sup-ported by our results to a large extent. Higherperformance can be achieved through appropriatecombinations of practices. For instance, Figure 2shows high performance under high dispersion/high incentive or low dispersion/low incentive con-ditions. In other words, organizations may possiblyperform well by following a ‘high road’ or a ‘lowroad’ (both internally consistent) human resourcemanagement strategy.

In addition, however, the results also provideinsights about inconsistent human resource man-agement practices. They imply that inconsistencycan be useful in some circumstances—an inconsis-tent pay system (individual incentives) can bufferdetrimental effects. This finding, if upheld inother studies, poses new dilemmas for the strate-gic human resource management area. On theone hand, consistency is necessary; on the otherhand, some inconsistencies may be beneficial.Researchers already struggling with the theoreticaland empirical issues regarding appropriate combi-nations of human resource management practices(Delery and Shaw, 2001) must also contend withthese inconsistent nuances as they develop con-ceptual frameworks of strategic human resourcemanagement.

This study suggests several other interestingavenues for research, particularly those related tothe unexpected findings in Study 2. A valuable pro-gram of research would start with dispersion issuesat the individual level and build up to the strate-gic level. To begin with, this program would focuson individual reactions to pay structures encom-passing legitimate and illegitimate sources of vari-ation in independent and interdependent settings.Such studies would have the ability to support orfalsify our speculations. The second step in thisresearch program would entail the developmentof organizational-level theories examining strate-gic compensation frameworks. Our focus on hori-zontal dispersion is particularly significant in thiscontext due to the increased prevalence of flat-ter rather than taller organizational designs. Ashierarchical levels are removed, more individualspursue technical (as opposed to managerial) careerpaths, and a ‘broad-banding’ approach to com-pensation design is adopted. The result is fewer



job titles, fewer status differentials, and wider paybands. Horizontal pay dispersion becomes a muchmore critical concern in these settings since moreindividuals occupy jobs at the same hierarchicallevel. The importance of this issue is further under-scored by the finding (in testing Hypothesis 3) thatinconsistent pay dynamics may buffer deleteriouseffects. The interplay among organizational struc-ture, sociotechnical systems, and pay dispersion inaffecting critical organizational outcomes assumescritical proportions in this regard. For example, thebenefits of an egalitarianism organizational sys-tem may be jeopardized without careful consid-eration of the complexities inherent in designingand administering, not merely consistent, but alsojust and legitimate compensation systems.

No study is perfect, and ours is no exception.The first issue concerns pay measures. Our mea-sures of pay dispersion are estimates and thusopen to challenge, but the high overlap amongalternative measures in Study 2 ameliorates thisconcern. Our measures of individual incentivescan be questioned. The incentive measures weresignificantly correlated with dispersion, but themagnitude of these correlations was not extraor-dinarily high. That the measures were moderatelycorrelated speaks to their validity as dispersion-creating practices. That they did not account formore variance in pay dispersion likely speaks tothe variety of dispersion-creating factors found inorganizations. We controlled for a multitude of fac-tors that could affect dispersion (e.g., size, age,tenure), but many other legitimate and illegiti-mate factors (politics, nepotism, favoritism) alsoaffect dispersion levels (Bartol and Martin, 1990;Gupta and Jenkins, 1996). Moreover, pay disper-sion and individual incentives are not synonymousbut distinct constructs (Pfeffer and Langton, 1993).They should be moderately, not perfectly, corre-lated. In fact, simultaneous use of pay compressionand individual incentives is advocated on occasion(Pfeffer, 1995)—something that would be impos-sible if the two were perfectly (but negatively)related. In addition, our measures of pay dispersionin Study 2 relied on hourly rates rather than totalannual pay. This reliance may not have allowedus to capture all the variation in dispersion createdby individual incentives since lump-sum bonusesand other incentives may not be included in hourlypay rates. We conducted some analyses post hocusing single incentive items (e.g., merit pay) and

found similar results across the items, but this issuewarrants further research.

Our outcome variables in Study 2 were basedexclusively on key informant reports—somethingresearchers have challenged in the past. On theother hand, key informants are commonly usedin organizational research. The perceptual perfor-mance equations in Study 1, and all equations inStudy 2, were estimated using data from the samesource, raising common method variance concerns.But common method variance should not supportsubstantive predictions across the level of a third(or the in three-way interaction cases, a fourth)variable, ameliorating these concerns. Also, ourdata are essentially cross-sectional in nature andcausal conclusions cannot be drawn. Since theresearch was confined to two industry settings andfocused on two key jobs, the generalizability ofour results to other industries and other jobs isnot certain. In addition, the study was limited totwo potential moderators (incentives and interde-pendence), but other contingencies likely to berelevant should also be examined in the future. Ourfocus by design on mid-level outcomes (workforceperformance) also makes it imperative for futureresearch to examine the applicability of our resultsto other outcomes (such as operating ratio).

Assuaging these concerns is the fact that stronginteractive results were obtained across three de-pendent variables derived from three distinct datasources in Study 1, and the fact that the strongestinteractions in Study 2 were observed with ‘hard’(labor hours per ton and accidents), rather thanpurely perceptual, dependent variables. Further-more, common method variance is an unlikelysubstantive explanation of theoretically derivedhigher-order interaction effects. The support (albeitmoderate) across two settings and across differentmeasures of both independent and dependent vari-ables further points to the robustness of the results.

Confidence in the substantive validity of theresults is also bolstered by other aspects of ourdesign. We conducted organizational-level analy-ses of a large sample of organizations within oneindustry (trucking) and a sample comprising morethan half of the total population of plants in another(concrete pipe). We held the focal job constant inboth studies and controlled for a variety of factorsthat could contaminate observed relationships. InStudy 1, we obtained data from key informants andtwo distinct archival sources, eliminating concernsabout common method variance. In Study 2, we


Pay Dispersion 511

examined two ‘hard’ performance measures (laborhours per ton and accidents) and found signifi-cant interaction results. Furthermore, the sample inStudy 2 consisted primarily of small companies,whereas the sample in Study 1 covered a broadspectrum of organizational size.

Taken together, the results clearly demonstratethe complexity of pay dispersion effects on work-force performance. In the final analysis, the sourcesof pay dispersion, the way work is designed, andthe way compensation dynamics are perceived arecritical to organizational effectiveness. These con-textual factors must be incorporated in systematicresearch on the effects of pay dispersion.

ACKNOWLEDGEMENTS

This research was supported by grants from theMack-Blackwell Rural Transportation ResearchCenter, created and sponsored by the United StatesDepartment of Transportation, the National Sci-ence Foundation, and the Society for HumanResource Management (SHRM) Foundation. Theauthors wish to thank Associate Editor Ed Zajacand two anonymous reviewers for constructivecomments on previous drafts.

REFERENCES

Allison PD. 1978. Measures of inequality. AmericanSociological Review 43: 865–880.

Arthur JB. 1994. Effects of human resource systems onmanufacturing performance and turnover. Academy ofManagement Journal 37: 670–687.

Baron JN, Pfeffer J. 1994. The social psychologyof organizations and inequality. Social PsychologyQuarterly 57(3): 190–209.

Bartol KM, Martin DC. 1990. When politics pays: fac-tors influencing managerial compensation decisions.Personnel Psychology 43: 599–614.

Becker BE, Huselid MA. 1992. The incentive effectsof tournament compensation systems. AdministrativeScience Quarterly 37: 336–351.

Bishop J. 1987. The recognition and reward of employeeperformance. Journal of Labor Economics 5: s36–s56.

Bloom M. 1999. The performance effects of paydispersion on individuals and organizations. Academyof Management Journal 42: 25–40.

Cohen SG, Ledford GE. 1994. The effectiveness ofself-managing teams: a quasi-experiment. HumanRelations 47: 13–31.

Cowherd DM, Levine DI. 1992. Product quality andpay equity between lower-level employees andtop management: an investigation of distributive

justice theory. Administrative Science Quarterly 37:302–321.

Dean JW Jr, Snell SA. 1991. Integrated manufactur-ing and job design: moderating effects of organiza-tional inertia. Academy of Management Journal 34:776–804.

Delaney JT, Huselid MA. 1996. The impact of humanresource management practices on perceptions oforganizational performance. Academy of ManagementJournal 39: 949–969.

Delery JE, Doty DH. 1996. Modes of theorizingin strategic human resource management: testsof universalistic, contingency, and configurationalperformance predictions. Academy of ManagementJournal 39: 802–835.

Delery JE, Shaw JD. 2001. The strategic managementof people in work organizations: review, synthesis,and extension. In Research in Personnel and HumanResources Management, Vol. 20, Ferris GR (ed.). JAIPress: Greenwich, CT; 165–197.

Deutsch M. 1985. Distributive Justice: A Social Psy-chological Perspective. Yale University Press: NewHaven, CT.

Deutsch M. 1988. Is there a tradeoff between economicefficiency and equity? Remarks to the AsilomarConference on Organizations.

Dollinger MJ, Golden PA. 1992. Interorganizational andcollective strategies in small firms: environmentaleffects and performance. Journal of Management 18:695–715.

Donaldson D, Weymark JA. 1980. A single-parametergeneralization of the gini indices of inequality. Journalof Economic Theory 22: 147–149.

Ehrenberg RG, Bognanno ML. 1990. Do tournamentshave incentive effects? Journal of Political Economy98: 1307–1324.

Folger R, Greenberg J. 1985. Procedural justice:an interpretative analysis of personnel systems.In Research in Personnel and Human ResourceManagement, Vol. 3, Ferris GR (ed.). JAI Press:Greenwich, CT; 141–183.

Frank RH. 1984. Are workers paid their marginalproducts? American Economic Review 74: 549–571.

Freeman S. 1977. Wage trends as performance displaysproductivity potential: a model and application toacademic early retirement. Bell Journal of Economics8: 419–443.

Freeman RB, Kleiner MM. 1998. From piece ratesto time rates: surviving global competition. Paperpresented at the Industrial Relations ResearchAssociation Meetings. Chicago, IL.

Freeman R, Medoff JL. 1984. What do Unions Do? BasicBooks: New York.

Gergen KJ, Greenberg MS, Willis RH. 1980. SocialExchange: Advances in Theory and Research . PlenumPress: New York.

Gupta N, Jenkins GD Jr. 1996. The politics of pay.Compensation and Benefits Review , March/April:23–30.

Heneman RL. 1992. Merit Pay . Addison-Wesley: Read-ing, MA.



Hicks JR. 1963. The Theory of Wages (2nd edn).Macmillan: London.

Hogan WT. 1987. Minimills and Integrated Mills: AComparison of Steelmaking in the United States .Heath: Lexington, MA.

Huselid MA. 1995. The impact of human resourcemanagement on practices, on turnover, productivity,and corporate financial performance. Academy ofManagement Journal 38: 291–313.

Ichniowski C, Shaw K, Prennushi G. 1997. The effects ofhuman resource management practices on productiv-ity: a study of steel finishing lines. American EconomicReview 87: 291–313.

Jenkins GD Jr, Mitra A, Gupta N, Shaw JD. 1998. Arefinancial incentives related to performance? A meta-analytic review of empirical research. Journal ofApplied Psychology 83: 777–787.

Lawler EE III. 1971. Pay and Organizational Effective-ness: A Psychological View . McGraw-Hill: New York.

Lawler EE III. 1973. Motivation in Work Organizations .Brooks/Cole: Monterery, CA.

Lawler EE III. 1981. Pay and Organizational Develop-ment . Addison-Wesley: Reading, MA.

Lawler EE III, Jenkins GD Jr.. 1992. Strategic rewardsystems. In Handbook of Industrial and Orga-nizational Psychology (3rd edn). Dunnette MD,Hough LM (eds). Consulting Psychologists Press:Palo Alto, CA; 1009–1055.

Lazear EP. 1981. Agency, earning profiles, productivityand hours restriction. American Economic Review 71:606–620.

Lazear EP. 1989. Pay equality and industrial politics.Journal of Political Economy 97: 561–580.

Lazear EP. 1995. Personnel Economics . MIT Press:Cambridge, MA.

Leventhal GS, Karuza J, Fry WR. 1980. Beyond fairness:a theory of allocation preferences. In Justice andSocial Interaction , Mikula G (ed.). Springer: NewYork; 167–218.

Levine DI. 1991. Cohesiveness, productivity, and wagedispersion. Journal of Economic Behavior andOrganization 15: 237–255.

Lewis MW, Grimes AJ. 1999. Metatriangulation: build-ing theory from multiple paradigms. Academy of Man-agement Review 24: 672–710.

Lind EA, Tyler TR. 1988. The Social Psychology ofProcedural Justice. Plenum: New York.

Majchrzak A. 1988. The Human Side of FactoryAutomation . Jossey-Bass: San Francisco, CA.

Meyer JW, Rowan B. 1991. Institutionalized organiza-tions: formal structure as myth and ceremony. In TheNew Institutionalism in Organizational Analysis , Pow-ell WW, DiMaggio PJ (eds). University of ChicagoPress: Chicago, IL; 41–62.

Mitra A, Gupta N, Jenkins GD Jr. 1997. A drop in thebucket: when is a pay raise a pay raise? Journal ofOrganizational Behavior 18: 117–137.

Montemayor EF. 1996. Congruence between pay policyand competitive strategy in high-performing firms.Journal of Management 22: 889–908.

Oldham GR, Nottenburg G, Kassner MW, Ferris G,Fedor D, Masters M. 1982. The selection andconsequences of job comparisons. OrganizationalBehavior and Human Performance 29: 84–111.

Osterman P. 1994. How common is workplace transfor-mation and who adopts it? Industrial and Labor Rela-tions Review 47: 173–188.

Perser JH. 1994. Lumping, long hours, absence fromhome continue to fuel truckload driver shortage.Traffic World 14(2): 31–33.

Pfeffer J. 1995. Producing sustainable competitiveadvantage through the effective management ofpeople. Academy of Management Executive 9(1):55–72.

Pfeffer J, Langton N. 1993. The effect of wage dispersionon satisfaction, productivity, and working collabora-tively: evidence from college and university faculty.Administrative Science Quarterly 38: 382–407.

Powell CT. 1992. Organizational alignment as competi-tive advantage. Strategic Management Journal 13(2):119–134.

Powell I, Montgomery M, Cosgrove J. 1994. Compensa-tion structure and establishment quit and fire rates.Industrial Relations 33: 229–248.

Rambo WW, Pinto JN. 1989. Employee’s perception ofpay increases. Journal of Occupational Psychology62: 135–145.

Shaw JD, Duffy MK, Stark EM. 2000. Interdependenceand preference for group work: main and congruenceeffects on the satisfaction and performance of groupmembers. Journal of Management 26: 259–279.

Sheppard BH, Lewicki RJ, Minton JW. 1992. Organiza-tional Justice. Lexington: New York.

Skarlicki DP, Folger R. 1997. Retaliation in theworkplace: the roles of distributive, procedural, andinteractional justice. Journal of Applied Psychology82: 434–443.

Snell SA, Dean JW Jr. 1994. Strategic compensationfor integrated manufacturing: the moderating effectsof jobs and organizational inertia. Academy ofManagement Journal 37: 1109–1140.

Suchman MC. 1995. Managing legitimacy: strategic andinstitutional approaches. Academy of ManagementReview 20: 571–610.

Terborg JR, Lee TW. 1984. A predictive study of organi-zational turnover rates. Academy of Management Jour-nal 27: 793–810.

Youndt MA, Snell SA, Dean JW Jr, Lepak DP. 1996.Human resource management, manufacturing strategy,and firm performance. Academy of ManagementJournal 39: 836–866.


pay dispersion and workforce performance: … · 2006. 2. 1. · yield these beneﬁts. dispersion...

Documents