tema 11 i ca six sigma success

7/30/2019 Tema 11 i CA Six Sigma Success

1/17

Journal of Operations Management 30 (2012) 437453

Contents lists available at SciVerse ScienceDirect

Journal ofOperations Management

j ournal homepage: www.elsevier .com/ locate / jom

Six Sigma adoption: Operating performance impacts and contextual

drivers ofsuccess

Morgan Swink a,, Brian W.Jacobs b,1

a Neeley Business School, TCU, PO Box298530, Fort Worth,TX 76129,UnitedStatesb Broad College of Business, Michigan State University, N370 North Business Complex, East Lansing, MI 48824-1122, United States

a r t i c l e i n f o

Article history:

Received 18 October 2011

Received in revised form 24 February 2012

Accepted 25 May 2012

Available online 4 June 2012

Keywords:

Six sigma

Process innovation

Operating performance

Event study

a b s t r a c t

We assess the operational impacts ofSix Sigma program adoptions through an event study methodology,

comparing financial data for 200 Six Sigma adopting firms against data for matched firms, which serve as

control groups for the analyses. We employ various matching procedures using different combinations

of pre-adoption return on assets (ROA), industry, and size as matching criteria. By comparing perfor-

mance outcomes across a hierarchy of operating metrics, we establish a pattern of Six Sigma adoption

effects that provides strong evidence of a positive impact on ROA. Interestingly, these ROA improve-

ments arise mostly from significant reductions in indirect costs; significant improvements in direct costs

and asset productivity are not evident. We also find small improvements in sales growth due to Six

Sigma adoption. Cross-sectional analyses ofthe performance results reveal that distinctions in Six Sigma

impacts across manufacturingand service firms are negligible. Interestingly,we find that the performance

impact ofSix Sigma adoption is negatively correlated to the firms quality system maturity (indicated by

prior ISO 9000 certification). Further analyses ofmanufacturing and service firms reveals that Six Sigma

benefits are significantly correlated with intensity in manufacturing, and with financial performance

before adoption in services. We discuss the implications of these findings for practice and for future

research.

2012 Published by Elsevier B.V.

1. Introduction

Since its origins in the mid-1980s, the Six Sigma program

for process improvement has become widely embraced. One

report suggests that many Fortune 500 firms have adopted Six

Sigma (Nakhai and Neves, 2009). Early successes in high pro-

file companies such as Motorola, Allied Signal (now Honeywell),

and General Electric helped to both popularize and legitimize

the approach, and dozens of books have been devoted to the

topic.

The practitioner literature documents substantial cost savings

and other benefits from Six Sigma program adoptions (Pande

et al., 2000; Harry and Schroeder, 2000). However, some stillquestion whether these benefits sufficiently exceed the costs of

adoption. Corporate-wide adoption of Six Sigma often involves

considerable investments in consulting support, training, organi-

zational restructurings, and associated information and reporting

systems. For example, over a four year period (19961999)

Corresponding author. Tel.: +1 817 257 7463.

E-mail addresses:[email protected] (M. Swink),[email protected]

(B.W. Jacobs).1 Tel.: +1 517 8846370.

General Electric reportedly spent more than $1.6 billion on Six

Sigma investments. Researchers report that training costs are typi-

cally as much as $50,000 pertrained worker (Antony, 2006; Fahmy,

2006). The net operating effects of these types of investments

have not been rigorously examined. Most scholarly work to date

involves perceptual data from surveys, or financial studies of a

few select companies (Goh et al., 2003; Zu et al., 2008; Gutierrez

et al., 2009; Braunscheidel et al., 2011). In fact, some writers have

even questioned the validity and originality of Six Sigma, calling it

repackaging, a fad, and a PR ploy (Clifford, 2001; Rowlands,

2003).

Other questions pertain to the types of benefits provided by

Six Sigma, and their limitations. A number of researchers discussthe potential for capability gains in one area of performance to be

offset by added constraints or losses in another. In particular, Six

Sigma potentially creates a trade-off between gains in efficiency

versus growth. Several important studies suggest that process

improvement regimes can stifle innovative exploration in favor

of exploitation, thus impeding sales growth (Abernathy, 1978;

Tushman and OReilly, 1996; Benner and Tushman, 2002, 2003;

Naveh and Erez, 2004). Moreover, recent anecdotes from compa-

nies like General Electric and 3 M indicate that managers believe

Six Sigma practices may severely constrain innovation needed to

drive growth (Brady, 2005; Hindo, 2007).

0272-6963/$ seefrontmatter. 2012 Published by Elsevier B.V.

http://dx.doi.org/10.1016/j.jom.2012.05.001
http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jom.2012.05.001http://www.sciencedirect.com/science/journal/02726963http://www.elsevier.com/locate/jommailto:[email protected]:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jom.2012.05.001http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jom.2012.05.001mailto:[email protected]:[email protected]://www.elsevier.com/locate/jomhttp://www.sciencedirect.com/science/journal/02726963http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jom.2012.05.001


2/17

438 M. Swink, B.W. Jacobs/ Journal of Operations Management 30 (2012) 437453

Limitations might also stem from the context within which Six

Sigma is adopted. Like many process improvement programs, Six

Sigmaoriginated in manufacturingfirms; manyof its principlesand

tenets were developed in a setting of asset-intensive, repeatable

processes. The name itself, Six Sigma, refers to limits in mea-

surable variations of outputs that were established in Motorolas

manufacturing processes. In addition, researchers maintain that

a firm must possess certain resources and make certain commit-

ments in order to make Six Sigma successful (Antony et al., 2008;

Schroeder et al., 2008). Hence, Six Sigma methods and tools may

be more or less effective in certain technological and operational

contexts.

In thisarticle,we examine the operatingperformance impacts of

SixSigma adoptions. Thestudy seeksanswers tothe followingthree

research questions. First, does Six Sigmaadoption consistently pro-

duce a significant net effect on operating performance? Given the

widespreadadoptionand continued popularity of this program, we

consider this a very important question. A sizable literature on the

efficacy of other process management strategies exists, providing

mixed results. However, researchers argue that Six Sigma is differ-

ent fromother process management approaches; it is distinguished

by its requisite organizational structures, structured methods, and

emphasis on customer-oriented metrics (Linderman et al., 2003;

Sinha and Van de Ven, 2005; Schroeder et al., 2008). Given these

proposed distinctions, it is important to determine whether or not

managers should have reason to expect that Six Sigma will provide

benefits that exceed alternative programs for improvement.

Our second research question addresses the nature of Six

Sigmas impacts. Whattypes of beneficialimpacts are manifestedin

theoperating data of SixSigmaadopters? By examiningthe compo-

nents of both profit and growth-oriented financial outcomes of Six

Sigma adopters, we develop insights into the types of impacts pro-

vided by theprogram.These results serve to informthe debateover

the roles of process management programs in creating competitive

advantages for their adopters; they also point to some interesting

propositions for future research.

Our third research question is: are Six Sigma impacts related to

operating contexts? As Six Sigma adoptions have grown to includea wider scope of businesses, researchers have begun to question

the applicability and effectiveness of related tools and techniques

in certain contexts. In addition, case studies and anecdotal evi-

dence is suggestive of factors that may be critical to successful

implementation. We study differences in Six Sigma success asso-

ciated with industry (manufacturing or service), labor intensity,

R&D intensity, prior operating performance, and quality maturity.

Our examination of these factors provides insights into the sources

of, and constraints on, process improvements emerging from Six

Sigma adoption.

We address the foregoing questions through an event study

methodology, comparing financial data for about 200 Six Sigma

adopting firms against data for matched firms, which provide

control groups for the analyses. We employ various matching pro-cedures using different combinations of pre-adoption operating

performance (measured by return on assets (ROA)), industry, and

size as matching criteria. By comparing performance outcomes

across a hierarchy of operating metrics, we establish a pattern

of Six Sigma adoption effects that provides strong evidence of a

positive impact on ROA. Interestingly, these ROA improvements

arise mostly from significant reductions in indirect costs. Improve-

ments in direct costs and asset productivities are not evident. We

also find small improvements in sales growth due to Six Sigma

adoption.From cross-sectional analyses, we determine that perfor-

mance improvement due to Six Sigma adoption is not significantly

related to industry (manufacturing or service) or R&D intensity.

However, changes in performance are significantly correlated with

the quality maturityof the adopting firms. Interestingly, firms with

greater quality experience (as indicated by ISO 9000 certification)

appearto benefitlessfromSix Sigma. Forfirmsin service industries,

operating performance before Six Sigma adoption is a significant

determinant of performance changes. However, labor intensity is

the most significant driver of performance benefits in manufactur-

ing firms.

In the next section, we formulate hypotheses relating Six Sigma

adoption to operating performance by drawing uponthe literatures

on process improvement in general, and Six Sigma in particular.

Section 3 describes the sample data and event study method. Sec-

tion 4 presents the results. Section 5 discusses the findings and their

implications. Section 6 summarizesthe conclusions and limitations

of the study, and identifies opportunities for future research.

2. Theory development and hypotheses

Researchers have placed Six Sigma in the realm of operational

improvement programs that are oriented toward improvements in

quality or variability of process outcomes (Zu et al., 2008). There

are several scholarly studies of the impacts of process improve-

ment programs, yet none provide a rigorous examination of Six

Sigma adoptions. The existing literature can be classified into

three streams addressing the performance impacts of: (1) gen-eral process management strategies (Ittner and Larcker, 1997;

Schmenner, 1991), (2) Total Quality Management (TQM) imple-

mentations (Hendricks and Singhal, 1996, 1997, 2001a,b; Ittner

et al., 2001; Powell, 1995; Sila, 2007; York and Miree, 2004; Nair,

2006), and (3) ISO 9000 and other quality certifications (Corbett

et al., 2005; Martinez-Costa et al., 2009; Westphal et al., 1997;

Yeung et al., 2006; Naveh and Erez, 2004; Benner and Tushman,

2002; Benner and Veloso, 2008; Levine and Toffel, 2010). These

research streams provide an overall positive, though mixed, set

of conclusions regarding the effectiveness of respective process

improvement programs. Importantly, however, researchers have

argued that Six Sigma is distinguished from these other programs

by several characteristics.

2.1. The distinctive characteristics of Six Sigma

Researchers describe Six Sigma as a data driven approach to

problem solving, as a business process, as a disciplined statistical

approach, and as a management strategy (Blakeslee, 1999; Hahn

et al., 1999; Harry and Schroeder, 2000;Braunscheidel et al., 2011).

While these monikers have been applied to other process improve-

ment strategies as well, proponents and researchers argue that

Six Sigma is different than other process improvement programs

because it is exclusively a customer-driven and data-defined sys-

tem(Breyfogle, 2003). Schroederet al. (2008)suggestthat SixSigma

must be different by virtue of the fact that it has been adopted by

many firms that had already possessed quite mature quality man-

agement processes (e.g., 3M, Ford, Honeywell, American Express).Perhaps more compellingly, Schroeder et al. (2008) and Zu

et al. (2008) argue that, while Six Sigma shares some philosoph-

ical underpinnings and techniques with other quality and process

management approaches, it is distinguished by four attributes

of its unique organizational approach. Schroeder et al. (2008, p.

540) define Six Sigma as an organized, parallel-meso structure

used to reduce variation in organizational processes by employing

improvement specialists, a structured method, and customer-

oriented performance metrics with the aim of achieving strategic

objectives. The typical parallel-meso structure for Six Sigma

includes a centralized office within the firm that oversees a dis-

persed training and project execution hierarchy. The central office

has several purposes. It creates an authoritystructurethat acquires,

develops, and assigns resources for training and improvement


3/17

M. Swink, B.W. Jacobs/ Journal of OperationsManagement 30 (2012) 437453 439

projects.It alsousually assemblesan executive team(or teams) that

sets criteria and guides improvement project selection (Carnell,

2003; Snee and Hoerl, 2003). In addition, this structure enables

top management engagement and status reviews (Schroeder et al.,

2008). By engaging high level managers in a centralized way, Six

Sigmaprojectsare thought to beless myopicand more aligned with

business strategy. Finally, the structure purportedly affords more

effective diffusion of lessons learned from projects, thus creating

greater multi-level understanding (Sinha and Van de Ven, 2005).

Six Sigma programs involve a variety of both part-time and

full-time improvement specialists, including champions (execu-

tive project sponsors), master black belts, black belts, green belts,

and lower level designations. The different belts denote different

levels of training and experience with Six Sigma methods. Mas-

ter black belts are typically full-time trainers and project mentors,

while black belts and green belts are workers who may apply Six

Sigmaconcepts andtoolsto drive improvementsin their respective

areas of functional responsibility. Black belts often have the same

leadership characteristics as heavyweight project managers (Clark

and Fujimoto, 1991). This hierarchy of improvement specialists

is thought to enhance the coordination of work across organiza-

tional levels (Sinha and Van de Ven, 2005; Barney, 2002), and again

to ensure the matching of tactical tasks with business strategy

(Henderson and Evans, 2000; Linderman et al., 2003).

Six Sigma improvement projects follow a structured method

which has been recognized as a variation of the Plan, Do, Check,

Act (PDCA) cycle (Shewhart, 1931; Schroeder et al., 2008). The

Six Sigma method includes five steps known as Define, Measure,

Analyze, Improve, and Control (DMAIC). A variant of the method

used in design-oriented processes is Define, Measure, Analyze,

Design, and Verify (DMADV). Choo et al. (2007) suggest that the Six

Sigma method provides an effective learning framework to guide

knowledge acquisition and to ensure that project team members

execute a more complete search of problem solving alternatives.

It also provides a common language enabling workers to effec-

tively communicate project status and to make comparisons across

improvement efforts.

Finally, Six Sigma projects are guided and assessed by a mix-ture of common and unique performance metrics. In addition

to using typical financial and operational project metrics, Six

Sigma applies unique measures including process sigma, critical-

to-quality (CTQ) attributes, and defects per million opportunities

(DPMO). Researchers argue that such metrics establish challeng-

ing goals and guidance for project teams (Linderman et al., 2003;

Pande et al., 2000), that they focus on objective data which tend to

mitigate political agendas (Brewer, 2004), that they embody out-

comes at business, process, and project levels, and ultimately that

they prioritize a customer focus.

2.2. Six Sigmas operating performance impact

While the popular press contains examples of both positiveand negative Six Sigma impacts on performance (e.g., Pande et al.,

2000; Harry and Schroeder, 2000; Chakravorty, 2010), few rigor-

ous studies exist. A study of twenty firms by Goh et al. (2003)

indicates that announcements of Six Sigma adoption produce no

significant changes in stock returns on announcement day. A fur-

ther analysis of six of the firms shows that their long run stock

performance is not better than the S&P 500. Zu et al. (2008, p.

643) find that Six Sigma practices and traditional QM practices

work together to generate improved quality performance, which

then leads to higher business performance. However, they base

these conclusions on self reported, perceptual measures of quality

and performance. Also using a survey with perceptual measures,

Gutierrezet al. (2009) find that SixSigma improves shared vision,

but relationships to self-reported organizational performance are

not significant. Braunscheidel et al. (2011) conducts case studies

of seven firms and concludes that Six Sigma leads to documented

savings and perceived innovation benefits.

The fundamental argument for net positive financial impacts

of Six Sigma adoption is that it creates new learning and adap-

tation capabilities within the firm. In short, Six Sigma is thought

to both provide a structure and promote a culture that fosters

problem/opportunity identification, process analysis, and the cre-

ation of sustained improvements. Gowen and Tallon (2005) use

the dynamic capabilities theoretical perspective to describe the

learning and adaptation capabilities associated with Six Sigma

adoption. This perspective emphasizes the need for organizations

to dynamically align their processes with changes in the busi-

ness environment. Dynamic capability is defined as a learned and

stable pattern of collective activity through which the organiza-

tion systematically generates and modifies its operating routines

in pursuit of improved effectiveness (Zollo and Winter, 2002,

p. 340). Dynamic capabilities enable organizations to efficiently

adapt, integrate, and reconfigure resources (Teece et al., 1997).

Gowen and Tallon (2005) argue that, by addressing both techni-

cal designs and human resources, the structured approach of Six

Sigma imbues the adopting organization with greater dynamic

capabilities. In essence, the structured attributes of best practice

identification, customer focus, and disciplined project selection

and execution provide organizationalarchitecture neededfor these

capabilities. Gowen andTallon (2005)further suggest that effective

Six Sigmaimplementations embody the value, rareness, inimitabil-

ity, and non-substitutability (VRIN) characteristics associated with

resources that provide competitive advantage, as specified in the

resource-based view (Barney, 2002). Theirsurvey dataindicatethat

managers perceive this to be true; they findsignificant correlations

between various Six Sigma program design factors and each of the

VRIN elements.

More generally, Anand et al. (2009) describe infrastructural ele-

ments of continuous improvement programs that foster dynamic

capabilities. Indeed, they represent continuous improvement itself

as a dynamic capability, when it is embedded in a comprehen-

sive organizational context (p. 445). They further identify andstudy Six Sigma as a particular continuous improvement ini-

tiative that provides such a context. Their case study analyses

indicate that infrastructural elements such as balanced innova-

tion and improvement, a constant change culture, standardized

improvement processes, and training are important enablers of

organizational learning and dynamic capabilities.

Consistent with the above arguments, other researchers also

suggest that structured improvement methods lead to better orga-

nizationallearningand knowledge transfer (Ittneret al., 2001;Choo

et al., 2007; Molina et al., 2007), as well as overall improved job

quality (Levine and Toffel, 2010). Linderman et al. (2003, 2006)

demonstrate that the interaction of the structured method and

rigorous goal setting of Six Sigma explains its impact on the perfor-

manceof specific projects. Other researchersargue thatadvantagesfrom process improvement programs derive mainly from social

aspects (Powell, 1995), including a supportive learning culture

(Detert et al., 2000; Schroeder et al., 2008; Naor et al., 2008) and

cooperative values (Kull and Narasimhan, 2010). Gutierrez et al.

(2009) maintain that Six Sigma adoption creates a shared vision

that impels team members to work together to achieve common

goals.

An integration of these arguments and findings suggests that

Six Sigma adoption provides a structured approach to organi-

zational learning that creates dynamic capabilities, specifically,

capabilities to consistently improve current processes (Ittner and

Larcker, 1997), thereby raising quality and lowering costs. Teece

(2007) maintains that such capabilities are critical for business

success; due to the increasing pace and complexity of business


4/17


environments, organizations no longer compete on processes, but

on the ability to continually improve processes.

In order to test this proposition, our starting point is to test the

hypothesis that Six Sigmaadoption positively impacts overall prof-

itability, as commonly measured by ROA (e.g., Corbett et al., 2005;

Hendricks and Singhal, 1997), and defined as operating income per

total assets, assessed before depreciation.

Hypothesis 1. Six Sigma adoption produces a significant positive

effect on profitability (ROA).

Profitability ( ROA) is determined by return on sales (ROS,

defined as operating income/sales) and asset turnover (ATO,

defined as sales/assets). In turn, both ROS and ATO can also be

broken down into their components, including cost of goods sold

(COGS), sales,general,and administrativeexpenses (SG&A), current

andfixed assets, etc. Consistent with our second research question,

should hypothesis H1 be supported, we plan to examine significant

differences in the constituent elements of ROA in order to develop

insights into the nature of contributions attributable to Six Sigma

adoption.

In addition to profit, companies commonly seek growth, often

measured by year-on-yearincreases in sales.A debatehas emerged

over the effects of process management programs on the growth

of adopting firms. Researchers suggest that the disciplined struc-

ture of process management programs tends to crowd out growth

oriented innovation in favor of exploitation (Benner and Tushman,

2002, 2003). Emphases on waste reduction, standardization, and

continuous improvement are sometimes considered incompatible

with the slack resources, flexibility, and risk taking propensities

needed to support more explorative efforts. At least two stud-

ies provide evidence of such effects from ISO 9000 certification.

Benner and Tushman (2002) posit that an emphasis on process

management biases innovation project selection toward incre-

mental improvements and away from more exploratory efforts.

In a longitudinal study of patent activity in the photography and

paint industries, they document an increased share of the firms

total innovations that are exploitative and build upon existing firmknowledge, post ISO 9000 certification. Similarly, Naveh and Erez

(2004) find that ISO 9000 certification is positively associated with

greater process control, but negatively associated with innovation

outcomes.

As a process management strategy, Six Sigma has been criti-

cized as being narrowly designed to improve existing processes,

and not being helpful in the development of new products or dis-

ruptive technologies needed to drive sales growth (Morris, 2006).

Hence, the dynamic capabilities stemming from Six Sigma adop-

tion could be considered to be limited to continuous improvement,

rather than also applying to more radical changes (Anand et al.,

2009). Indeed, reports fromwell-known companiessuch as General

Electric and3M document managersfeelingsthatthey need topro-

tect growth-oriented functions in the firm (e.g., marketing, R&D)from possible strictures imposed by Six Sigmas disciplined struc-

ture (Brady, 2005; Hindo, 2007; Parast, 2011). However, Schroeder

et al. (2008) counter these concerns by maintaining that Six Sigma

provides switching structures that simultaneously promote the

conflicting demands of exploration and control in the improve-

ment effort (p. 537). Further, they identify Six Sigma black belts as

boundary spanning actors, who integrate strategic concerns with

tactical improvement efforts, thus facilitating exploration (Manev

and Stevenson, 2001). Finally, the active engagement of top man-

agersin project selection andmonitoring are thought to foster more

strategicand exploratory efforts. Theseconflicting viewsleave open

thequestionof whether SixSigmaadoptionaids growthfrom inno-

vation, whether it stifles it, or whether it has no significant effect

at all.

Six Sigma adoption may produce other effects on sales growth

that are independent of its effects on innovation and exploration.

Six Sigma adoption can also serve as a signal to customer markets

of improved product quality. First, as Six Sigma has gained notori-

ety (Gowen and Tallon, 2005), the reputational effects of adoption

have potentially created greater pricing power for adopting firms.

Second, real increases in product quality can be expected to lead to

higher customer satisfaction. As a result, customers may be willing

to pay more or to buy more; both outcomes increase sales revenue.

These expectations are consistent with the findings ofHendricks

and Singhal (1997) and Corbettet al. (2005); thesetwostudiesshow

significant increases in sales for firms that won quality awards and

obtained ISO 9000 certifications, respectively.

In sum, the foregoing discussion describes three mechanisms

through which Six Sigma adoption might foster greater sales

growth: (1) through supporting product innovation (though this

is contested in the literature), (2) through reputational enhance-

ments that improve product and brand image, and (3) through

process improvements that create better product quality (fewer

defects). All three effects presumably lead to improved customer

satisfaction and associated growth in sales.

Hypothesis 2. Six Sigma adoption produces a significant positive

effect on sales growth.

2.3. Contextual drivers of Six Sigma implementation success

Our final research question addresses potential relationships

between Six Sigma impacts and the operating context of adop-

tion. The nature of our study affords us the opportunity to examine

a number of contextual factors that can enhance or impede an

adopting firms abilities to extract real benefits from Six Sigma

implementation.

2.3.1. Manufacturing versus service

Of particular interest are differences in adoptions by primarily

manufacturing versus primarily service firms. Case studies doc-

ument Six Sigma adoptions in a wide variety of service firms

including hospitals, government, banks and financial services, util-

ities, fitness clubs, retailers, and so on. Some researchers study Six

Sigma programs and projects in both manufacturing and service

firms as if they are universal (Schroeder et al., 2008; Nair et al.,

2011). However, others identify limitations and challenges of Six

Sigma methodology and tools in service settings. Some argue that

it is harder to measure outcomes, collect reliable data, and control

service processes (Hensley and Dobie, 2005; Antony et al., 2007).

The ambiguous and customer-specific nature of critical-to-quality

service features can make them difficult to define, such that Six

Sigma metrics such as DPMO are unnecessarily stringent and dif-

ficult to apply (Nakhai and Neves, 2009). Moreover, common Six

Sigma tools and training topics may not adequately address differ-

ences between service customers expectations and perceptions.

For example, Six Sigma does not typically address marketing com-munications or other influencers of customers expectations. In

addition, Six Sigma rarely addresses softer dimensions of service

quality such as empathy (Nakhai and Neves, 2009). A survey by

Antony et al. (2007) indicates that core Six Sigma methods such as

statistical tools, process capability, and design of experiments are

among the least commonly used tools in services. Their research

also suggests that service process improvements are more depen-

dent on organizationalchanges than manufacturing improvements

are, and service organizations are at the same time more resistant

to change because of the higher personal involvement of workers.

Accordingly, we posit:

Hypothesis 3. The positive effect on profitability from Six Sigma

adoption is greater for manufacturing firms than for service firms.


5/17


2.3.2. Labor intensity

Rather than drawing contrasts between manufacturing and ser-

vice firms, the more salient differentiator might be whether firms

in heavier, more asset-intensive industries experience different

results from Six Sigma adoption than those of firms in more labor-

intensive industries. Hendricks and Singhal (2001b) argue that

labor-intense firms provide more fertile ground for quality pro-

cess improvements because they have more process options and

because they depend more on training and skills. Because labor-

intensive processes are inherently more variable, they likely offer

a greater range of variance reduction opportunities. Indeed, vari-

ance reduction is the essential value in the Six Sigma paradigm. On

theotherhand, highlyautomated processestend to beless variable.

Furthermore, because of theirhigh fixedcosts, processes employing

heavy and automated equipment typically already receive spe-

cialized attention by highly trained process and manufacturing

engineers, even in the absence of Six Sigma. Therefore, the impacts

of Six Sigma adoption in these contexts may be muted by compar-

ison.

Hypothesis4. The effect on profitability from Six Sigma adoption

is positively associated with the firms labor intensity.

2.3.3. R&D intensityGiven the debate over the effects of Six Sigma adoption on inno-

vation and growth, we are motivated to evaluate the relationship

between R&D intensity and the performance effects of Six Sigma

adoption. As noted in the discussion for hypothesis H2, a stream of

research indicates that process management programs such as Six

Sigma can exert a stifling effect on more exploratory innovation

efforts. We would expect this effect to be particularly damaging

to overall firm performance if the firm has strategically positioned

itself as a leading innovator. As was the case documented at 3M

(Hindo, 2007), if advanced R&D is the firms lifeblood, then Six

Sigmas highly structured approach to continuous improvement

might be regarded negatively by employees, potentially raising

resistance to change and hampering implementation. On the other

hand, if Six Sigma is embraced, organizational units might beat least implicitly encouraged to favor incremental exploitation

projects over more radical exploratory efforts, thus destroying the

innovative firms competitive advantage.

Similarly, the benefits of Six Sigmas program structure might

be heightened or lessened by the importance of staying techno-

logically current. Again, if a firm positions itself as a technology

leader, danger is associated with becoming too focused on the sta-

tus quo. In such a context, continuous improvement efforts that

focus on process refinement might be less important to success

than efforts aimed at uncovering replacement technologies and

entirelynew opportunities.As mentionedearlier,researchers argue

that Six Sigma builds dynamic capabilities and an organizational

learning infrastructure that enables adopting firms to adapt more

readily to changing environmental conditions (Gowen and Tallon,

2005; Anand et al., 2009). The central question becomes whether

these abilities are limited to incremental changes, or whether they

also apply to more dynamic, technologically intensive contexts.

We forward the following hypothesis as a frame for testing these

competing propositions.


is associated with the firms R&D intensity.

2.3.4. Prior financial performance

Following the logic ofHendricks and Singhal (2008), we rec-

ognize that prior operating performance can potentially affect the

impacts of Six Sigma on performance, in two different ways. First,

implementation resources can be a function of the firms pre-

adoption profitability. Highly profitable firms likely have greater

reserves of cash and other needed resources to invest in process

management infrastructural changes, training, and administration

(Hendricks and Singhal, 2008). Therefore, such firms are likely to

affect broader and more complete implementations. In addition,

available resources enable adoption on a larger scale. For example,

profitable firms will likely have the capital needed to fund a larger

number of simultaneous improvement projects. In these ways,

profitable firms can attain greater leverage from Six Sigma adop-

tions, thus leading to greater abnormal operating performance.

On the other hand, poor performing firms may be better posi-

tioned for the changes required by Six Sigma adoption. Poor

profitability can be a source of motivation; i.e., employees in a

loss-making firm might have a greater sense of urgency needed

to implement organization changes like Six Sigma. Kotter (1995)

suggests that poor business results can increase the probability of

successful implementation of organizational changes, because the

need for change is more apparent, and consequently the urgency

and motivation required for successful implementations is more

readily found in poor performers. As a result, levels of top manage-

ment and organizational commitment may be higher, leading to

more aggressive goal setting and a more effective implementation.

Numerous writers highlight the importance of such commitment

to Six Sigma success (e.g., Chakravorty, 2009; Antony et al., 2008;

Linderman et al., 2006; Kumar and Antony, 2009)


is associated with the firms prior financial performance.

2.3.5. Quality maturity

Schroeder et al. (2008) note that some firms adopting Six Sigma

already have quite mature quality management processes. This

prompts the question of whether the impacts of Six Sigma on

firm performance are contingent upon the firms prior quality

management knowledge. If Six Sigma simply replicates the capa-

bilities engendered by other quality management programs, then

we might expect little additional performance gain from Six Sigmaadoption. If, on the other hand, Six Sigmas program attributes are

trulydistinctive,as a number of researchersassert (Breyfogle, 2003;

Schroederet al., 2008;Zu et al., 2008), then we might expectunique

performance gains.

Operating on the premise that Six Sigma is related to, but truly

distinctive from, other programs, an absorptive capacity perspec-

tive would suggest that more quality mature firms possess greater

abilities to acquire, evaluate, assimilate, and exploit Six Sigma pro-

cess knowledge (Zahra and George, 2002). Cohen and Levinthal

(1990) describe absorptive capacity as an organizational ability to

embrace and exploit new knowledge. Further, they argue that this

ability depends on priorknowledge andexperience.Relatedknowl-

edge and experience provides a foundation for new knowledge

absorption, as it creates familiarity and lessens causal ambiguities.For example, organizations experienced in quality management

programs would likely speak much of the language of Six Sigma,

even before adoption. In addition, employees who have gone

through similar organizational transformations are likely to feel

less threatened by Six Sigma-driven changes. For these reasons,

we expect that firms experienced with quality oriented process

managementprograms willimplementSix Sigmafaster,more com-

pletely, andperhaps more effectively.As a result, they shouldenjoy

greater performance benefits from Six Sigma program adoption

than their less experienced counterparts.

Hypothesis 7. The effect on profitability from Six Sigma adop-

tion is positively associated with the firms quality maturity (prior

quality program experience).


6/17


3. Researchmethod

3.1. Sample collection and description

We used multiple sources web searches, books, practitioner

journals, and academic journals to identify a preliminary list of

over 600 companies named as adopters of the Six Sigma philoso-

phy and methodologies. While the list is certainly not exhaustive,

it appears to be fairly representative, as it includes a wide range

of industries, firm sizes, and adoption years. Of the identified

firms, 421 are publicly traded companies with financial data in

Compustat.

To corroborate whether the identified firms actually adopted

Six Sigma, and to determine their Six Sigma adoption dates, we

used key words such as Six Sigma in conjunction with history

or adoption to thoroughly search publicly available documents

(e.g., all publication sources in the Factiva database, corporate 10-

K reports, corporate websites, the Internet) for each of the 421

publicly traded firms. We retained in the sample only firms that

had adopted Six Sigma in 2007 or earlier, in order to have suf-

ficient data to establish post-adoption performance. Using these

sources, we found definite, pre-2008 adoption years for 214 of

the 421 firms (50.8%). For 143 firms (34.0%), we could not deter-

mine a specific adoption year, but we found enough evidence

of Six Sigma activity to establish a late bound for adoption. For

the remaining 64 firms (15.2%), we did not find sufficient infor-

mation to establish either an adoption date or a late bound for

adoption.

Because the available public information sometimes required

interpretation and/or was conflicting from different sources, each

firm was researched independently by two members of our

research team. The two independent researchers agreed on 351 of

421 findings (83.4%) of adoption dates, late bounds, or no adoption

dates. For the remaining 70 firms with disputed adoption dates or

late bounds, the mean (median) difference in designated adoption

years was0.6(1.0)years. Toresolve thedifferences, wesupplied the

datasourcesdiscoveredby the two researchersto a thirdresearcher

whoindependently weighed theevidence anddetermined thespe-cific adoption years for use in our analyses.

Early on in this research, we sent a survey to each publicly

held Six Sigma adopter for which we could identify a credible

contact person. The survey asked about adoption date and extent

of adoption of the aforementioned practices that are distinct to

Six Sigma (centralized team structure, improvement specialists,

structured methods/DMAIC, and customer-focused metrics). We

secured survey responses from 58 of the 214 publicly traded firms

with identified adoption dates (23.8% of our sample). Of the 58

single respondents: 38 (65.5%) agreed with our identified adop-

tion years; 9 (17.6%) were unable to provide a specific adoption

year; 7 (12.1%) supplied an adoption date one year earlier than our

finding; 3 (5.9%) supplied an adoption date more than one year

later than our finding; and 1 (1.7%) supplied an adoption date oneyear later than our finding. We note that all three respondents

with adoption dates greater than one year later from our finding

were reporting only for their division within the overall firm. To be

conservative, we usedthe earliest adoptionyear identified. Further-

more, the survey data indicate a remarkably uniform application

of Six Sigma practices across the respondents. For example, over

90% of the respondents indicatedthat they employed a black/green

belt structure, and over 95% designated that DMAIC and other Six

Sigma tools were used on at least 80% of improvement projects.

These results reinforce our overall confidence in the accuracy of

our estimates for both the timing and extent of adoptions in our

sample firms.

For the 214 firms with specific adoption years, Panel A of

Table 1 presents the number of adopting firms by year. The earliest

Table 1

Sampledescriptionfor 214 firms with specific Six Sigma adoptionyears.

Panel A: Frequency of Six Sigma adoption years

Year Frequency Year Frequency

1986 2 1997 16

1987 0 1998 12

1988 2 1999 18

1989 1 2000 38

1990 1 2001 40

1991 1 2002 18

1992 1 2003 18

1993 0 2004 15

1994 0 2005 9

1995 2 2006 13

1996 3 2007 4

Panel B: Occurrence (percentage) of most-frequent SIC codes

2-Digit code Frequency 3-Digit code Frequency

35 24 (11.2%) 371 12 (5.6%)

36 23 (10.7%) 602 11 (5.1%)

28 21 (9.8%) 357 10 (4.7%)

37 21 (9.8%) 283 6 (2.8%)

38 14 (6.5%) 367 6 (2.8%)

adoption year in oursample is 1986 andthe most frequently occur-

ring adoption year is 2001. We note the drop-off in Six Sigma

adoptions in our sample post-2001. Given the continued interest

and relevance of Six Sigma,as evidenced by academic publications,

current business school textbooks and curriculums, and practi-

tioner seminar offerings, we suspect that the drop-off of Six Sigma

adoptions in our sample is indicative of non-newsworthiness. In

other words, Six Sigma has become an accepted part of everyday

business, much like TQM or Lean. This highlights the importance

of rigorously studying the impact Six Sigma adoption on operating

performance.

Table 1 Panel B presents themostfrequently occurringSIC codes

within thesample firms. The sample contains firms from 47 unique

two-digit SIC codes and 101 unique three-digit SIC codes. Thoughthe majority of firms represent manufacturing industries, about

one-third of the firms are services. Table 1 provides more informa-

tion on the most frequently represented industries. Table 2 Panel

A presents descriptive statistics for our sample based on the 2001

fiscal year, the most common Six Sigma adoption year in our sam-

ple. The median observation in the sample represents a firm with

$5.6B in market value, $7.5B in total assets, $6.2B in annual sales,

$0.8B in annual operating income, and28,300 employees. Forcom-

parison, Table 2 Panel B presents descriptive statistics for the 207

suspected Six Sigma adopters for which we could not determine a

specific adoptionyear. In addition,Table 2 Panel C presents descrip-

tive statistics for all firms listed in the New York Stock Exchange

(NYSE), also for the 2001 fiscal year. In summary, our sample rep-

resents a wide variety of industries,and is not significantly differentfrom the suspected Six Sigma adopters for which we could not

determinea specific adoption year. However, when compared to all

NYSE firms, oursample is notrepresentative of smaller enterprises.

This outcome raises a question regarding thegeneralizability of our

findings, as the cause of the difference is not known. Research indi-

cates that small and medium sized firms are less likely to adopt Six

Sigma, mainly because theylack requisiteresourcesand knowledge

(Antony et al., 2008; Kumar and Antony, 2009). Thus, our sample

firms might be larger because of sampling bias (i.e., larger firms

are more likely to be identified by our sources), but the sample

firms might also be larger because they truly represent the pop-

ulation (i.e., larger firms are more likely to adopt Six Sigma). We

note that large-firm bias is common in OM research. We discuss

this limitation further in Section 6.


7/17


Table 2

Descriptivestatistics for2001,the most frequent sample Six Sigma adoption year.

Market value ($M) Total assets ($M) Sales ($M) Operating income ($M) Employees (000s)

Panel A: Samplefirms (N= 214)

Median 5616.4 7477.1 6204.9 760.2 28.3

Mean 19,394.2 38,370.3 14,205.1 2337.4 51.9

Std Dev 42,211.0 102,806.9 21,799.3 4695.6 69.2

Max 397,831.6 693,575.0 162,412.0 37,966.0 395.0

Min 0.1 44.2 49.1 (5062.0) 0.4

Panel B: Suspected Six Sigma adopting firms without known adoption years (N= 207)Median 4764.1 4684.2 3986.3 494.0 17.8

Mean 23,233.4 24,763.0 11,625.6 1754.2 47.6

Std Dev 52,321.7 76,197.4 24,409.1 3429.2 119.6

Max 392,959.0 695,877.0 218,529.0 29,602.0 1383.0

Min 0.4 0.2


8/17


Table 3

Matching process and benchmark group statistics.

Matching method

Performance in

year 2a, and

industry

Performance in

years 2,3,4b,

and industry

Performance and size in

years2,3,4c, and

industry

Step 1 matchesd 203 200 198

Step 2 matchese 6 6 7

Step 3 matchesf 0 0 1

Total firms matched 209 206 206

Mean group size 34.3 34.1 14.9

Median group size 18.0 21.5 9.0

Maximum group size 470 423 86

No. of groups with a single firm 6 2 8

a Performance definedas ROA in year 2; matching rangeis 90110%.b Performance definedas median ROAin years2, 3, and 4;datarequired in at least two years.c Size defined as medianTotal Assets in years2,3,and4;datarequired in atleast two years; matching rangeis withinfactor of 25.d All firms within thesame two-digit SICcode as thesample firm, and whoseperformanceand/or size arewithin the specified rangeof thesample firm.e All firms within thesame one-digit SICcode as thesample firm, and whoseperformanceand/or size arewithin the specified rangeof thesample firm.f All firms whoseperformance and/or size arewithin thespecifiedrangeof thesample firmregardless of SICcode.

Table 4

Median descriptive statistics for the matching period (years2, 3, and 4) prior to Six Sigma adoption.

ROA Total assets ($M) Market value ($M) Sales ($M) Operating income ($M) Employees (000s)

Panel A: Samplefirms (N= 214)

Mean 0.1362 12,932.2 23,086.6 10,450.9 1882.1 47.9

Median 0.1333 4303.1 5356.9 4889.7 761.9 25.5

Std Dev 0.0755 23,901.5 52,784.0 16,311.7 3672.0 64.5

Max 0.4627 250,138.5 303,989.0 146,991.0 32,291.0 413.0

Min (0.2629) 22.1 8.3 15.7 (22.2) 0.3

Panel B: Benchmark firms (obtained frommedian-performance-size-industrymatching method (N= 3077)

Mean 0.1313 9869.5 3981.7 2928.1 625.2 11.1

Median 0.1307 915.1 735.9 738.1 122.0 3.2

Std Dev 0.0605 44,271.1 20,646.7 8732.1 2014.6 28.4

Max 0.4331 933,559.1 911,494.2 174,694.0 31,750.0 775.1

Min (0.2736) 5.1 1.5 0.2 (9.8)


9/17


10/17


Table 5

Annual abnormalchanges in ROA forsample firms foryear1 through year +4.

From year: N Median Z-Statistic Mean t-Statistic % Positive Z-Statistic

Panel A: Performance-industry matching

1 to 0 197 0.121% 0.674 0.001% 0.004 52.28% 0.641

0 to +1 194 0.469% 2.741*** 0.600% 2.884*** 59.28% 2.585***

+1 to +2 191 0.180% 1.040 0.129% 0.633 53.40% 0.941

+2 to +3 181 0.113% 0.578 0.115% 0.491 51.93% 0.520

+3 to +4 166 0.423% 2.192** 0.447% 1.947** 58.43% 2.173**

1 to +4 167 0.355% 2.390***

1.214% 2.729***

55.69% 1.470*

0 to +4 167 0.077% 1.639* 0.994% 2.314** 52.10% 0.542

+1 to +4 167 0.030% 0.917 0.540% 1.406* 50.30% 0.077

Panel B:Median-performance-industrymatching

1 to 0 194 0.256% 1.002 0.130% 0.568 55.67% 1.580*

0 to +1 191 0.378% 2.208** 0.391% 1.904** 58.64% 2.388***

+1 to +2 189 0.241% 2.120** 0.367% 1.834** 56.61% 1.818**

+2 to +3 181 0.099% 0.634 0.034% 0.170 48.07% 0.520

+3 to +4 164 0.552% 1.679** 0.336% 1.497* 57.93% 2.030**

1 to +4 164 0.703% 2.632*** 1.294% 3.108*** 56.10% 1.562*

0 to +4 164 0.530% 2.019** 0.901% 2.262** 56.71% 1.718**

+1 to +4 164 0.411% 1.541* 0.559% 1.730** 55.49% 1.406*

Panel C:Median-performance-size-industrymatching

1 to 0 194 0.055% 0.415 0.049% 0.190 51.03% 0.287

0 to +1 191 0.080% 0.482 0.021% 0.103 50.79% 0.217

+1 to +2 188 0.061% 0.989 0.223% 1.071 51.60% 0.438

+2 to +3 179 0.000% 0.172 0.017% 0.086 50.28% 0.075

+3 to +4 163 0.465% 1.995** 0.471% 1.897** 59.51% 2.428***

1 to +4 163 0.433% 1.900** 1.029% 2.322** 53.99% 1.018

0 to +4 163 0.479% 1.569* 0.914% 2.097** 52.76% 0.705

+1 to +4 163 0.395% 1.696** 0.826% 2.223** 53.99% 1.018

Panel D: One-on-one median-performance-size-industry matching using only early adopters matched against samplefirms that adopted at least fiveyears later

1 to 0 34 0.150% 0.244 0.111% 0.222 55.88% 0.686

0 to +1 34 0.592% 1.362* 0.932% 1.488* 62.16% 1.480*

+1 to +2 34 0.172% 0.904 0.717% 1.244 61.54% 1.441*

+2 to +3 34 0.413% 0.962 0.280% 0.588 41.67% 1.000

+3 to +4 34 1.197% 2.039** 1.220% 2.370*** 58.82% 1.029

1 to +4 34 1.383% 1.406* 1.784% 1.628* 55.88% 0.686

0 to +4 34 1.740% 1.863** 2.025% 2.220** 62.86% 1.521*

+1 to +4 34 1.759% 2.101** 2.211% 2.362*** 66.67% 2.000**

Allsamples trimmed at 2.5% each tail.

Z-Statistics for medians are obtained using Wilcoxon Signed-Rank tests.

Z-Statistics for% positive areobtained using Binomial Sign tests.* Significance is one-tailed:p .10

**

Significance is one-tailed:p .05.*** Significance is one-tailed:p .01.

The change per firm over the 5-yearperiod (from year 0 to+4) is

also significantly positive for both the mean and median, using all

three matching methods. The % firms with positive 5-year changes

are significantly greater than 50% using the median-performance-

industry matching method. As expected, the magnitudes of the 5-

year changes are generally less than those of the 6-year changes.

The change per firm over the 4-year period (from year +1 to +4)

follows a similar pattern and is again generally lesser in magnitude

and significance than the 5-year changes.

As we noted in Section 3.2, a limitation of our study is that we

could notdefinitivelydetermine that all benchmark firms were not

also Six Sigma-adopters during thesampling time frame.If many ofthe benchmark firms were truly adopters, then estimates of abnor-

mal performance would be muted, making significant differences

difficult to detect. The fact that we do find significant differences

suggests that, in the worst case, our findings of abnormal perfor-

mance due to Six Sigma adoption are conservative.

To address this limitation, we would ideally match known

adopters only with known non-adopters. Such a pure comparison

would produce a more reliable estimate of expected performance

improvement from Six Sigma adoption. To approximate this pure

comparison, we identified knownnon-adoptersas firmsfrom our

sample that adopted Six Sigma at least five years later than earlier

sample adopter firms. The five-year delay allows us to consider

operating performance impacts to the early adopters during the

post-adoption period through year +4. Given that we are limited to

only the 214 firms in our sample, and to permit the greatest num-

ber of comparisons, we matched each adopter against only a single

firm, and did not use data from any one firm more than once; this

method permitted 41 matches using the ROA, assets, and industry

criteria from median-performance-size-industry matching. Table 5

Panel D presents the results for abnormal changes in ROA for year

1 through year +4.TheROA improvementsaresignificant andgen-

erally stronger than in our other matching methods. These results

should be regarded as somewhat tentative, given the small sample

size,andtheconfoundingwithtimeduetothismatchingmethodol-

ogy (the adopters all adopted prior to 2003). However, the results

do strongly reinforce the conclusion that the estimated improve-ments from our larger analyses are real, albeit conservative.

Considering boththe annual changes and multiple-year changes

indicates that Six Sigma adoption produces an immediate and per-

sistent positive effect on ROA. These results provide support for

hypothesis H1.

4.2. Decomposition of ROA effects

For brevity in presenting and discussing the remainder of

our results, we concentrate on our most conservative matching

method,median-performance-size-industry, noting that the pattern

of results is similar regardless of matching method. Table 6 Panel

A presents the results for the abnormal changes in the level of ROS

on an annual basis and for multiple-year periods. For the 6-year


11/17


Table 6

Annual abnormal changes in ROS, COGS/sales, SG&A/sales, and ATO for sample firms for year 1 through year +4 using the median-performance-size-industrymatching

method.


Panel A: Changes in abnormal ROS

1 to 0 194 0.117% 0.320 0.324% 1.406* 51.55% 0.431

0 to +1 191 0.048% 0.009 0.117% 0.492 50.79% 0.217

+1 to +2 188 0.127% 1.151 0.328% 1.449* 51.60% 0.438

+2 to +3 179 0.167% 0.249 0.065% 0.287 52.51% 0.673

+3 to +4 163 0.482% 1.955**

0.364% 1.639*

59.51% 2.428***

1 to +4 163 0.899% 0.937 0.342% 0.774 52.76% 0.705

0 to +4 163 0.632% 1.363* 0.490% 1.114 56.44% 1.645*

+1 to +4 163 0.892% 1.761** 0.596% 1.563* 56.44% 1.645*

+1to +4effecta 163 0.528% 2.063** 0.645% 1.976** 56.44% 1.645*

Panel B: Changes in abnormal COGS/sales

1 to 0 160 0.260% 0.106 0.222% 0.987 45.00% 1.265

0 to +1 159 0.214% 0.193 0.006% 0.026 45.91% 1.031

+1 to +2 157 0.200% 1.927** 0.447% 2.346*** 45.86% 1.038

+2 to +3 149 0.173% 0.344 0.086% 0.407 46.98% 0.737

+3 to +4 135 0.095% 0.738 0.211% 0.952 48.89% 0.258

1 to +4 132 0.533% 0.374 0.281% 0.539 53.03% 0.696

0 to +4 133 0.336% 0.225 0.150% 0.317 51.13% 0.260

+1 to +4 132 0.438% 0.018 0.033% 0.087 46.97% 0.696

Panel C: Changes in abnormal SG&A/sales

1 to 0 160 0.007% 0.319 0.051% 0.370 49.38% 0.158

0 to +1 159 0.096% 0.061 0.123% 0.855 55.35% 1.348*

+1 to +2 157 0.155% 0.762 0.023% 0.156 47.13% 0.718

+2 to +3 149 0.149% 2.277** 0.342% 2.639*** 41.61% 2.048**

+3 to +4 135 0.219% 1.979** 0.297% 2.190** 40.00% 2.324**

1 to +4 132 0.842% 2.846*** 1.094% 2.832*** 38.64% 2.611***

0 to +4 133 0.639% 2.140** 0.771% 2.108** 42.86% 1.648**

+1 to +4 132 0.604% 2.476*** 0.747% 2.707*** 38.64% 2.611***

Panel D: Changes in abnormal ATO

1 to 0 194 0.295% 0.882 0.200% 0.203 44.33% 1.580*

0 to +1 191 0.055% 0.512 0.866% 1.112 49.74% 0.072

+1 to +2 188 0.633% 1.299* 0.680% 0.795 44.68% 1.459*

+2 to +3 179 0.605% 0.626 0.164% 0.184 45.25% 1.271

+3 to +4 163 0.130% 0.802 1.067% 1.199 51.53% 0.392

1 to +4 163 0.004% 0.715 1.944% 0.992 49.69% 0.078

0 to +4 163 0.027% 0.367 1.120% 0.634 49.08% 0.235

+1 to +4 163 0.253% 0.095 0.026% 0.018 48.47% 0.392



Z-Statistics for% positive areobtained using Binomial Sign tests.a Effect onROA computed per firm asROS Change+1 to +4 Firm ATO in year0.

* Significance is one-tailed:p .10.** Significance is one-tailed:p .05.

*** Significance is one-tailed:p .01.

periodfrom year1 to +4, the median and meanchange per adopt-

ing firm, and the % of sample firms experiencing positive change,

are all positive but insignificant. For the 5-year period from year 0

to +4,the median change is 0.632%, significantlypositive at the 10%

level. The mean change is 0.490%, positive but insignificant, and

56.44% of firms experience positive changes, significantly greater

than 50% at the 10% level. For the 4-year period from year +1 to +4,

the median and mean changes are 0.528% and0.645%, respectively,

both significantlypositiveat the 5% level, and56.44% of firms expe-

rience positive changes, significantly greater than 50% at the 10%level.

To determine whether the improvement in ROS from year +1

to +4 contributes significantly to the improvement in ROA, we

employ the method ofKinney and Wempe (2002). For each adopt-

ing firm, we compute the effect of ROS on ROA by multiplying the

change in abnormal ROS from year +1 to +4 with the firms ATO at

adoption (year 0). The median and mean ROS effects over the 4-

year period are 0.528% and 0.645%, respectively, both significantly

positive at the5% level, and56.44% of samplefirms experience pos-

itive ROS effects, significantly greater than 50% at the 10% level.

These results indicate that the ROS improvement from Six Sigma

adoption significantly contributes to the overall improvement

in ROA.

The primary cost components that impact operating income

(and hence, ROS and ROA) are COGS and SG&A. To determine the

contributions of bothcomponents, we examined abnormalchanges

in COGS/sales and SG&A/sales. Given that all firms do not consis-

tently report COGS and SG&A separately each year, we included

only adopting firms and benchmark firms in our analyses that did

report both accounts. This eliminated approximately 35 adopting

firms from oursample. Table 6 Panels B andC present theresults for

the abnormal changes in the level of COGS/sales and SG&A/sales,

respectively, on an annual basis and for the multiple-year periodsof interest. Although the median annual changes in COGS/sales are

all negative, only the change from year +1 to +2 is significant. The

mean (median) abnormal change from year +1 to +2 is 0.200%

(0.444%), significant at the 5% (1%) level. For the 6-year period

(year 1 to +4) and the 5-year period (year 0 to +4), the median,

mean, and % positive changes are all positive but insignificant. For

the 4-year period (year +1 to +4), the change statistics are negative

but insignificant. The evidence provides no support that Six Sigma

adoption produces significant reductions in COGS.

Considering the annual abnormal changes in SG&A/sales for the

sample firms, we see that they are consistently negative. Four of

5 (5 of 5) median (mean) annual SG&A/sales changes are negative,

and the % firms with positive changes is less than 50% for 4 of 5


12/17


Table 7

Annual abnormalchanges in sales growthfor samplefirms foryear1 through year +4.


Panel A: Performance-industry matching

1 to 0 197 1.398% 1.693 ** 0.555% 0.589 43.65% 1.781**

0 to +1 194 0.646% 0.570 1.148% 1.284* 46.91% 0.862

+1 to +2 191 0.664% 0.753 0.498% 0.501 47.64% 0.651

+2 to +3 181 0.305% 1.278 1.613% 1.536* 47.51% 0.669

+3 to +4 166 0.689% 0.250 0.749% 0.853 46.39% 0.931

1 to +4 167 5.718% 0.229 3.179% 0.704 43.11% 1.780**

0 to +4 167 0.094% 0.103 1.985% 0.545 49.70% 0.077

+1 to +4 167 1.533% 0.378 0.034% 0.012 48.50% 0.387

Panel B:Median-performance-industrymatching

1 to 0 194 0.710% 0.260 0.957% 0.950 47.94% 0.574

0 to +1 191 0.532% 0.225 1.101% 1.182 46.60% 0.941

+1 to +2 189 0.486% 0.856 0.417% 0.411 48.15% 0.509

+2 to +3 181 0.030% 0.396 0.208% 0.217 49.72% 0.074

+3 to +4 164 1.540% 2.681*** 3.138% 3.296*** 57.32% 1.874**

1 to +4 164 2.463% 1.725** 11.897% 2.803*** 52.44% 0.625

0 to +4 164 3.485% 1.876** 8.118% 2.416*** 54.88% 1.249

+1 to +4 164 5.631% 1.952** 5.344% 2.028** 56.71% 1.718**

Panel C:Median-performance-size-industrymatching

1 to 0 194 0.008% 0.575 2.127% 1.973** 50.00% 0.000

0 to +1 191 1.485% 0.326 0.774% 0.859 46.07% 1.085

+1 to +2 188 0.271% 0.158 0.237% 0.234 51.06% 0.292

+2 to +3 179 0.030% 0.046 0.071% 0.072 49.72% 0.075

+3 to +4 163 1.259% 1.410* 2.098% 2.345*** 53.37% 0.862

1 to +4 163 0.695% 1.483* 11.022% 2.485*** 50.92% 0.235

0 to +4 163 3.502% 1.078 5.006% 1.422* 53.99% 1.018

+1 to +4 163 6.182% 1.403* 3.759% 1.441* 55.83% 1.488*



Z-Statistics for% positive areobtained using Binomial Sign tests.* Significance is one-tailed:p .10.

** Significance is one-tailed:p .05.*** Significance is one-tailed:p .01.

annual changes. Two of the 5 annual changes (from year +2 to +3,

and year +3 to +4) are statistically significant in all three tests for

median, mean, and% positive.For the6-year periodfromyear1 to+4, the median (mean) change is0.842% (1.094%), significantly

negative atthe 1% (1%) level,and 38.64% of samplefirmsexperience

positive median annual changes, significantly less than 50% at the

1% level. For the 5-year period from year 0 to +4, the median and

mean changes are0.639% and0.771%, respectively, both signifi-

cant at the 5% level. 42.86% of firms experience positive changes in

SG&A/sales for the 5-year period, significantly less than 50% at the

5% level. Similarly, the mean, median, and % positive changes for

the 4-year periodfrom year +1 to +4 are allnegative andsignificant

at the 1% level. These results suggest that Six Sigma improvements

significantly and persistently reduce SG&A costs.

Table 6 Panel D presents the results for the abnormal changes

in the level of ATO. Interestingly, the annual changes in abnormal

ATO for the sample firms are generally negative. Five of 5 (2 of 5) median (mean) annual changes are negative, and the % firms

with positive changes is less than 50% for 4 of 5 annual changes.

However, only the change from year +1 to +2 is marginally sig-

nificant. For the multiple-year periods, the median and % positive

changes are allnegative butinsignificant. The results thereforeindi-

cate no significant relationship between Six Sigma adoption and

ATO.

4.3. Sales growth effects

In order to test hypothesis H2, we next consider the effect of Six

Sigma adoption on changes in sales. Table 7 presents the results

for abnormal sales growth on an annual basis, for each of our three

matching methods. We note that, more so than for any other of

our performance measures, the results for abnormal sales growth

appear somewhat sensitive to the matching method employed.

There are at least two plausible reasons for this discrepancy: (1)sales growth data are inherently noisier than our other measures

since they are percentage changes in absolute numbers rather than

differencesin ratiosand(2) changes insalesare correlated with firm

size, so the size control added inmedian-performance-size-industry

matching has a greater effect. Accordingly, we again concentrate

our discussion on the results from the most conservative match-

ing method, median-performance-size-industry, presented in Panel

C. The annual abnormal % changes in sales for the sample firms are

neither consistently positive nor negative. Two of5 (5of 5)median

(mean) annual changes are positive, and the % firms with positive

changes is greater than50% for 3 of 5 annualchanges. Onlythe pos-

itive changefrom year +3 to +4 is statistically significant in thetests

for median and mean. For the 6-year period from year1 to+4,the

median (mean) change is 0.695% (11.022%), significantly differentfrom zero at the 10% (1%) level; 50.92% of sample firms experience

positive changes, insignificantly different from 50%. For the 5-year

period from year 0 to +4, the median and % positive changes are

positive but insignificant; the mean change is 5.006%, significant at

the 10% level. For the 4-year period from year 0 to +4, the median,

mean, and % positive changes are all significantly positive at the

10% level. We conclude that the results provide only limited sup-

port for Hypothesis 2, that Six Sigma adoption positively impacts

sales growth.

The foregoing findings collectively indicate that significantly

improved ROA in adopting firms is primarily due to indirect cost

reductions(SG&A),and perhaps mildly dueto positive salesgrowth.

Both of these changesare reflected in improved ROS, rather than to

improvements in ATO.


13/17


Table 8

Estimatedcoefficients (standardized,t-Statistics in parentheses) fromregressions ofabnormal ROAchange fromyear1to year +4usingthemedian-performance-size-industry

matching method.

Independent variables Operationalization Model 1

manufacturing and

Services

Model 2 services

only

Model 3

manufacturing

only

Model 4

manufacturing

only

Intercept 3.875 7.646 3.830 5.589

(1.203) (1.565) (0.907) (1.175)

Manufacturing o r services 1 i f manufacturing,

0 if services

0.074

(0.790)Labor intensity

R&D intensity

Employees/sales

R&D/sales

0.242

(2.932)***0.128

(0.852)

0.343

(3.199)***0.325

(2.661)***

0.042

(0.383)

Po sitive fi nan cial pe rf ormance I ndus tr y- adju sted

ROA if positive,

0 otherwise

0.059

(0.725)

0.358

(2.483)***0.081

(0.831)

0.115

(0.934)

Negativefinancialperformance Industry-adjusted

ROA if negative,

0 otherwise

0.136

(1.664)*0.374

(2.499)**0.103

(1.083)

0.118

(1.131)

ISO9000 e xperience Six S igma a doption

year minus 1st

ISO9000

certification

0.254

(2.551)**0.201

(1.278)

0.207

(1.755)*0.264

(2.024)**

Firm size ln(market value)a 0.027

(0.317)

0.192

(1.307)

0.000

(1.000)

0.067

(0.542)

Adoption year YearO 0.113

(1.200)

0.219

(1.553)

0.110

(0.903)

0.168

(1.172)

New CEO 1 if new CEO in

years 0,1,or 2,

0 otherwise

0.049

(0.612)

0.134

(0.954)

0.037

(0.385)

0.023

(0.220)1

Number of

observations

156 47 109 97

Model Fvalue 2.418** 2.346** 2.758** 2.181**

R2 11.63% 29.63% 16.05% 16.55%

AdjustedR2 6.82% 17.01% 10.23% 8.96%

Allsamples trimmed at 2.5% each tail.a Alternative operationalizations of firm size [ln(Sales), ln(Employees), ln(Total Assets)] yield substantively similar results.* Significance is two-tailed:p .10.

** Significance is two-tailed:p .05.*** Significance is two-tailed:p .01.

4.4. Relating abnormal ROA performance to Six Sigma adoptercontext

To examine the data for support ofhypotheses H3H7, which

address the potential roles of contextual factors in Six Sigma adop-

tions, we perform cross-sectional analyses. To assess the impacts

to operating performance over the entire study period, we use

the abnormal ROA change from year1 to +4, obtained using the

median-performance-industry-sizematching method, as our depen-

dent variable.

Table 8 shows the results for regressions of the abnormal per-

formance of Six Sigma adopters on the contextual variables for the

entire sample (Model 1),the service andmanufacturingsubsamples

(Models 2 and 3, respectively), and the manufacturing subsample

with firms reporting R&D intensity (Model 4). We note that all four

models are significant at the 5% level. We review the results in

the order of our hypotheses and discuss them further in the next

section.

In Model 1, the coefficient for the dummy indicator of manufac-

turing or service industry is not statistically significant, indicating

that the ROA benefits of Six Sigma adoption are not signifi-

cantly greater for manufacturing firms than for service firms.

This fails to support hypothesis H3. Despite the lack of sig-

nificant difference in the overall benefit of Six Sigma adoption

between manufacturing and service firms, we examine Models

2, 3, and 4 to determine whether the contextual factors that can

impact an adopting firms abilities to extract benefits from Six

Sigma implementation differ between firms in manufacturing or

services.

A significant association indicated in the results presented inTable 8 pertains to labor intensity. The labor intensity coefficient is

positive and significantly different from zero at the 1% level for the

total sample and manufacturing subsamples (Models 1, 3, and 4).

Labor intensity is positive but insignificant for the service industry

subsample (Model 2). Thus, hypothesis H4 is supported, but only

for manufacturing firms.

As noted previously, the impact of R&D intensity can only be

evaluated for our manufacturing subsample as most services firms

do not report R&D expenses. The results from Model 4 indicate no

significant effect of R&D intensity on abnormal ROA changes. Thus,

hypothesis H5 is not supported.

The results from Model 1 indicate that pre-adoption profitabil-

ity is significantly correlated with abnormal ROA from Six Sigma

adoption by sample firms, but only at a marginal level (p0.10)

and only for firms with negative financial performance. Given that

the values of the negative financial performance variable are non-

positive by definition, the negative coefficient indicates greater

benefits for Six Sigma adopters with more negative prior financial

performance. The results from Model 2 demonstrate that both pos-

itiveand negative pre-adoption financial performance is associated

with greater abnormal performance for service firms. These results

suggest that, while overall effects of Six Sigma adoption are typi-

cally positive in service firms, they may be amplified by either the

financial strength or weakness of the firm at the time of adoption.

Interestingly, these findings are not significant for manufacturing

firms. Thus, hypothesis H6 is supported, but only for service firms.

The cross-sectional analyses yield significant results for the ISO

9000 experience variable. The ISO 9000 experience coefficients for


14/17


thetotal sample and manufacturingsubsamples (Models 1, 3, and4)

are significantly negative at the 5%, 10%, and 5% levels, respectively.

To understand this counter-intuitive result better, we examined

the raw data. For the 21 firms that were certified to ISO 9000 after

Six Sigma adoption, the median and mean abnormal ROA changes

from year 1 to year +4 are 2.615% and 3.579%, respectively, both

significantlygreater than zero at the1% level. Forthe 124firms that

were ISO9000certifiedprior to SixSigmaadoption, themedian and

mean abnormal ROAchanges are0.501% and0.471%,neither signif-

icantly different from zero. The differences in medians and means

between thesetwo groups are bothsignificantly differentfrom zero

at the 1% level. This suggests that firms with greater quality matu-

rity benefit less from Six Sigma adoption, a finding counter to our

hypothesis H7.

We note that none of our control factors firm size, adoption

year, new CEO significantly impact the benefits from Six Sigma

adoption.

5. Discussion of the results

Overall, the results indicate that the benefits of Six Sigma

adoption tend to more than compensate for associated costs and

required investments. Recalling that our estimates are conserva-

tive, SixSigmaadopters shouldexpect an addition to ROAof at least

0.20.3 percentage points each year on average. These magnitudes

of change are both statistically and economically significant. The

results from the median-performance-size-industrymethod, which

are the most conservative, indicate that the sample firms abnor-

mal ROA increased on average by 1.029 percentage points in total

over the 6-year period from year1 to +4, or an average of 0.206

percentage points improvement per year. Given that the median

samplefirmhadanROAof13.22%inyear2, thischange represents

a 7.8% improvement relative to non-adopting firms.For themedian

sample firm with $7.0B in assets in year2, such an improvement

translates into roughly $220M in additional operating income over

the 6-year period for the same asset base.

While quite significant, this ROA boost nevertheless appears tobe modest in comparison to results indicatedin other process man-

agement event studies. Corbett et al. (2005) find an average yearly

ROA increase of 0.89 percentage points associated with ISO 9000

certification. Hendricks and Singhal (1997) find an average yearly

ROAincrease of5.01percentage pointsin years1to+3forwinners

of quality awards. We hesitate to conclude that Six Sigma actually

offers less of an impact than these other programs, however. As we

noted earlier, estimates from our study are conservative, given the

possibility thatsome firmspopulating our benchmark groups could

have also adopted Six Sigma before or during the sampled time

horizon. Indeed, the average ROAboost indicated from our one-on-

one matching process (Table 5 Panel D) is nearly double that of the

more conservative matching process. While Hendricks and Singhal

(1997) cite similar potential pollution of benchmark groups as alimitation oftheirstudy, itis importantto note that they studyqual-

ity award winners, i.e., successful purveyors of quality initiatives.

Thus, their sampleis upwardly biased. There is no reason to believe

that our sample is similarly upward biased; our results may repre-

senta morevariedand realisticsuccess ratein process management

implementation. Also important, our performance matching crite-

rion (3-year median) is stricter than the criteria used by either of

the other two studies.

Six Sigma is a relative newcomer to the ranks of process

improvement programs. It is likely that many Six Sigma adopters

have already put TQM, lean, or other strategies in place prior to

Six Sigma. In these firms, manager may have already targeted

low hanging fruit in previous process improvements, and we

should therefore not be surprised by the relatively small operating

performance improvements associated with Six Sigma reflected in

our overall sample. Indeed, our finding regarding quality maturity

calls into question the argument that Six Sigma is truly distinctive

from other quality oriented management processes. Using the

absorptive capacity perspective, we argued that firms with greater

experience in quality management should benefit more from Six

Sigma adoption, because Six Sigma is at the same time similar to,

yet distinct from, prior quality management programs. Since the

empirical evidence demonstrates less benefit for firms with greater

quality maturity, a more likely conclusion is that the capabilities

stemming fromSix Sigmaadoption addlittle measurable valueover

and above those that emanate from ISO 9000 certification. Thus,

while Six Sigma may entail distinctive organizational structures,

problem solving tools, and metrics, these attributes appear to be

less importantfor already experiencedfirms. The managerial impli-

cations of this finding for firms with and without mature quality

systems could be far-reaching, and thus require further research.

Our findings support the theorythat Six Sigma structure engen-

ders the development of dynamic learning capabilities. However,

one might still question whether the positive returns from these

capability developments justify risks and opportunity costs asso-

ciated with Six Sigma adoption. By using matched, presumed

non-adopting firms as benchmarks in our analysis, we have con-

structed proxies that incorporate such risks and opportunity costs.

However, a given manager considering Six Sigma adoption will

want to consider the specific risks and foregone opportunities that

are relevant to his/her firm. For example, the profits projected by

ourstudymightnot clear thedesired rate of return(hurdle rate) for

a given firm. Moreover, for a given firm a Six Sigma program might

be an inferior investment to a new distribution center, an ERP sys-

tem, a more fuel efficient fleet of trucks, or other such investments

if they have more certain orfaster paybacks.2 Ourresults do suggest

that positive returns from Six Sigma adoption take time. The most

significant driver of ROA in our data, savings in SG&A, first materi-

alizes significantly only in years +2 through +4 (see Table 6 Panel

C). Similarly, with one exception, significant improvements in sales

growthemergeonlyinyears+3and+4(see Table7). Thus,it appears

that dynamic capabilities emerge gradually, or at least take time tobe manifested in operating performance improvements. It is also

possible that Six Sigma is rolled out sequentially across divisions,

extendingthe time forimpactsto manifest in overall corporateper-

formance. Though these findings should be regarded as tentative,

they do suggest that managers shouldbe willing to wait at least 23

yearsbefore netimpacts of Six Sigmaadoption become significantly

positive.

Other insights into thenature of SixSigmas operational impacts

provided by our study are also quite interesting, and somewhat

surprising.Six Sigmais widely recognizedas a methodology for cut-

ting costs and eliminating defects (Byrne and Norris, 2003; Pande

et al., 2000). Six Sigma focuses organizational efforts on process

improvement, especially through reducing variance for outputs of

product (or service) features that are deemed to critically influ-ence customers perceptions of quality. At its core, the DMAIC

method aims to measure and analyze the deviation of a given pro-

cess from its critical-to-quality goals so that workers can install

preventive measures that eliminate the root cause of defects. Such

preventive measures involve the implementation of training, pro-

cedures, monitoring and control systems, tools, technologies, and

product redesign. One would expect that this focus on structural

control would yield improvements in process efficiencies. Reduc-

tions in variation and associated defects are known to create cost

savings in areas of internal product rework, inventories, capacity

2

We thank oneof thereviewers foroffering this observation.


15/17


buffers, warranties, and repair work. These also include material

costs savings from reduced scrap and labor savings from reduced

appraisal, material handling, and supervision. We expected that

such efficiency improvements would be reflected in overall low-

ered product or service costs and associated higher margins.

Surprisingly, our results clearly show that efficiencies gained

from Six Sigma adoption are reflected more strongly in indirect

cost savings (SG&A), as opposed to savings in direct operating costs

(COGS). We note that COGS includes direct purchase, labor, and

operating expenses, while SG&A captures indirect expenses asso-

ciated with governance, logistics, advertising, overhead, and other

indirect activities. It is important to note that most SG&A processes

in manufacturing firms are in fact, repeatable service processes

(e.g., customer service, billing, transportation, etc.). Nakhai and

Neves(2009) identify a number of non-manufacturing applications

of Six Sigma inside manufacturing firms. Such processes tend to be

labor-intensive and repetitive. A related finding in our data is that

Six Sigma benefits are strongly correlated with labor-intensity in

manufacturing firms, yet this same correlation is not significant

in service firms. Taken together, both findings suggest that labor-

intensive, repeatable processes offer the greatest opportunities for

applications of Six Sigma methods.

We offer several explanations for this finding, which at first

glance may seem somewhat counter-intuitive. First, Hendricks and

Singhal (2001b) argue that labor-intense firms provide more fer-

tile grounds for quality process improvements because they have

more process options and depend more on training and skills. As

we explained in our motivation for hypothesis H4, processes that

are more automated and less labor-intensive tend to be inherently

less variable.As a result, these processesprovide less overall oppor-

tunityfor improvement from Six Sigma structured methods, which

are mostly aimed at variance reduction.

Second, back-office (SG&A) operationstend to be less influenced

by specific customer requirements and idiosyncrasies, i.e., they are

more repeatable. Consequently, they may present more attractive

targets for Six Sigma projects. Indeed, many of the service firm

examples of Six Sigma applications described in the literature are

actually in back-office or business-to-business contexts (Nakhaiand Neves, 2009; Antony et al., 2007). Given the supposed chal-

lenges of implementing Six Sigma in highly personalized services,

future studies that directly compare Six Sigma implementations

in back-office versus more direct personal service contexts could

reveal important differences in how Six Sigma concepts are opera-

tionalized.

Finally, there is logic suggesting that process variance reduc-

tions will reduce indirect costs, perhaps even more strongly than

direct costs. Schmenner (1988, 1991) notes that overhead costs

often exceed direct costs. Further, he argues that slow mov-

ing and highly variable process flows are the primary drivers

of indirect overhead costs. For example, variable process flows

create requirements for many transactions (purchasing, inven-

tory control, production control, quality control) as well as otheradded overheads (inventory, space, material handling, manage-

ment attention). These arguments are echoed in swift-even

tema 11 i ca six sigma success

Documents