Jacobs 2015 - New

Accepted Manuscript
Title: Performance Effects of Early and Late Six Sigma

Adoptions
Author: Brian W. Jacobs Morgan Swink Kevin W. Linderman
PII:
DOI:
Reference:
S0272-6963(15)00005-4
http://dx.doi.org/doi:10.1016/j.jom.2015.01.002
OPEMAN 896
To appear in:
OPEMAN
Received date:
Revised date:
Accepted date:
21-7-2014
2-12-2014
15-1-2015
Please cite this article as: Jacobs, B.W., Swink, M., Linderman, K.W.,Performance
Effects of Early and Late Six Sigma Adoptions, Journal of Operations Management
(2015), http://dx.doi.org/10.1016/j.jom.2015.01.002
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
us
cr
Brian W. Jacobs
Michigan State University
Eli Broad College of Business
632 Bogue Street N370
East Lansing, Michigan 48824
Tel: (517) 884-6370
Fax: (517) 432-1112
e-mail: jacobsb@broad.msu.edu
ip
t
PERFORMANCE EFFECTS OF EARLY AND LATE SIX SIGMA ADOPTIONS
an
Morgan Swink
Texas Christian University
Neeley Business School
PO Box 298530
Fort Worth, Texas 76129
Tel: (817) 257-7463
Fax: (817) 257-7227
e-mail: m.swink@tcu.edu
Ac
ce
p
te
Kevin W. Linderman
University of Minnesota
Carlson School of Management
321 19th Ave. S.
Minneapolis, Minnesota 55455
Tel: (612) 626-8632
Fax: (612) 624-8804
e-mail: linde037@umn.edu
Corresponding author
Page 1 of 35
PERFORMANCE EFFECTS OF EARLY AND LATE SIX SIGMA ADOPTIONS
ABSTRACT
Operations managers confront the challenge of deciding when to implement various administrative
ip
t
innovations such as Six Sigma, ISO 9000, and Lean. This research examines the operating performance
effects of early versus late adoption of Six Sigma process improvement. Using theories of organizational
learning and knowledge transfer, we develop hypotheses describing the advantages of late adoption, and
cr
factors that affect a firms ability to benefit from Six Sigma either as an early or late adopter. We test our
hypotheses using an event study methodology. The empirical results show that, on average, late adopters
us
in our sample enjoy significantly greater performance gains than early adopters. However, the analysis
also shows that the advantages of late adopters tend to be moderated by certain environmental and
structural characteristics of a firm. Specifically, late adoption has been favorable when firms operate in
an
low-velocity industries, when they primarily sell in business-to-business markets, when they have good
financial performance prior to adoption, and when they are large. Conversely, when adopters operate in
results.
conditions that have the opposite characteristics, then early adoption appears to have produced better
Understanding the effects of these factors can enhance managers abilities to determine
appropriate adoption timing to increase performance.
Ac
ce
p
1. INTRODUCTION
te
Keywords: Six Sigma; quality management; administrative innovation; adoption timing; organizational
learning; empirical research
Managers continuously search for administrative innovations that can potentially improve their business
processes and enhance operating performance. Defined by Teece (1980, p. 464) as improvements in
administrative techniques and in the organization of economic activity, administrative innovations
include practices such as quality management, matrix management, and zero-based budgeting (Westphal
et al. 1997), and certifications such as ISO 9000, ISO 14000, and C-TPAT (Ritchie and Melnyk 2012). In
this study, we focus on Six Sigma as an administrative innovation. While managers search for new
administrative innovations like Six Sigma to improve performance, they face advertisements and
proposals from vendors, business consultants, and other purveyors of such improvement methodologies.
In a fast moving world, managers have to decide whether, when, and how they should adopt new
innovations. Such decisions can significantly impact an organizations competitive standing, because
they affect the organizations ability to maintain parity or even create advantages over competitors. This
study examines the important question of when to adopt administrative innovations by studying the
impacts of early and late adoption of a widely diffused innovative program, Six Sigma.
1
Page 2 of 35
Administrative innovations like Six Sigma typically require major reassignments of tasks and
responsibilities across an organization; they can therefore be expensive and disruptive to adopt (Teece
1980). Accordingly, researchers have examined whether such innovation adoptions lead to operating
performance improvements that exceed these investments. Examples include studies of the performance
ip
t
effects of TQM (Yeung et al. 2006), ISO 9000 (Corbett et al. 2005), JIT (Kinney and Wempe 2002), and
Six Sigma (Swink and Jacobs 2012; Shafer and Moeller 2012). On the whole, these studies found that the
decision to adopt of administrative innovations yields significant performance benefits. However, the
cr
benefits may vary depending on when organizations decide to adopt the administrative innovation.
Importantly, the ease and effectiveness of implementation may be influenced by the adoption timing.
us
Administrative innovations may be more difficult to implement during early stages of diffusion, since
little experience or knowledge has accumulated to guide adoptions.
Without such knowledge,
organizations rely on learning-by-doing, which proceeds at a slower pace (Aldrich and Ruef 2006). A
an
substantial portion of the literature describing the diffusion of innovations contrasts the motivations of
early and late adopters of such practices (e.g., Rogers 2003). However, few studies have directly
examined the operating performance effects of early versus late adoption. In this article, we extend the
arguments for early versus late adoption of administrative innovations in order to develop and test
improvement programs.
hypotheses that describe performance effects associated with the adoption timing of Six Sigma process
te
Administrative innovations such as JIT manufacturing and Six Sigma have no externally defined
standards and no external agencies that certify adopting organizations to established norms or standards.
They are typically diffused through the
Ac
ce
p
Such administrative innovations are not strictly defined.
documented experiences of early and exemplary adopters. Organizations therefore have great latitude in
how they adopt and implement the innovation. Innovation implementation examples, such as those
provided by early Six Sigma adopters, can guide later adopters by serving both as referents and
persuaders (Jensen and Szulanski 2007). Referents permit later adopters to observe implementation
details and nuances. Persuaders help to convince managers and employees of late adopters by providing
physical proof of adoption benefits. As Williams (2007) explains, successful knowledge transfer requires
both replication due to the ambiguity of the knowledge, and adaptation because knowledge is context
dependent.
Accordingly, adopters will customize their adoptions (if possible) to best fit their
organizational contexts.
Early adopters are typically thought to enjoy certain first mover advantages which produce superior
financial outcomes (Lieberman and Montgomery 1988). Such advantages include early access to limited
information, head starts on experience curves, buyer lock-ins due to switching costs, and brand equity
derived from positive customer perceptions of the firms technological leadership (Lieberman and
2
Page 3 of 35
Montgomery 1988; Kerin et al. 1992; Frynas et al. 2006). In addition, early adopters might enjoy greater
benefits since they likely adopt for efficiency reasons rather than the normative, mimetic, and coercive
pressures that late adopters are more subject to (Westphal et al. 1997, Guler et al. 2002, Benner and
Veloso 2008). On the other hand, early adopters face greater uncertainties about the applicability of the
ip
t
innovation, and have less knowledge on how to effectively implement it. Because late adopters of Six
Sigma have access to more knowledge and experience that has accrued for the innovation, they can make
more appropriate adoption decisions, and better determine which components of the innovation to
cr
replicate and which to adapt to their local contexts. Thus, we expect that late adopters should experience
late mover advantages over early adopters. Importantly, environmental and structural factors may limit
us
or enhance late adopters abilities to effectively implement innovative processes. For example, late
adopters in stable environments might be more successful at implementing the innovation than late
adopters in dynamic environments. Firms in a stable environment can better apply the knowledge and
an
experience gained from earlier periods to their more predictable setting, thus making improved adoption,
replication, and adaptation decisions.
Six Sigma has diffused to a wide variety of industries, and a large literature documents its growth.
Six Sigma adoptions often involve large investments in training, consulting support, reorganizations, and
associated information systems. A number of case studies, anecdotes, and small sample studies describe
adoption success factors (Goh et al. 2003; Zu et al. 2008; Gutirrez et al. 2009; Braunscheidel et al.
te
2011), but rigorous studies of performance effects are only now emerging (Swink and Jacobs 2012;
Shafer and Moeller 2012). Given the evidence that Six Sigma adoption is beneficial to operating
Ac
ce
p
performance, our study provides further insights on benefits of adoption timing, and contingencies that
moderate these benefits.
In a larger theoretical sense, the findings of our study show that, on average, late adopters of Six
Sigma enjoy significantly greater performance gains than early adopters, suggesting that the growth of
supportive knowledge resources over time confers advantages upon late adopters. However, the analysis
also shows that the advantages of late adopters tend to be moderated by certain environmental and
structural characteristics of a firm. Namely, late adoption has been favorable when firms operate in lowvelocity industries, when they primarily sell in business-to-business markets, when they have good
financial performance prior to adoption, and when they are large. Conversely, when adopters operate in
conditions that have the opposite characteristics, early adoption appears to have produced better results.
We explain how understanding the effects of these factors may serve to enhance managers abilities to
determine appropriate adoption timing to increase performance benefits.
3
Page 4 of 35
2. SIX SIGMA ADOPTION AND TIMING

2.1 Six Sigma Adoption
Schroeder et al. (2008, p. 540) defines Six Sigma as an organized, parallel-meso structure used to reduce
ip
t
variation in organizational processes by employing improvement specialists, a structured method, and

customer-oriented performance metrics with the aim of achieving strategic objectives. Schroeder et al.
(2008, p. 540) note that companies may choose variations of this base definition when implementing Six
cr
Sigma in order to customize it to their situation. Implementing Six Sigma typically involves the creation
of an authority structure, dispersed and specialized training efforts, and a cross-functional project
us
execution hierarchy. Its core elements include a structured method, a focus on variance reduction, and a
combination of both general and specific performance metrics oriented to customers.
an
Many organizations have adopted Six Sigma, including most Fortune 500 companies (Nakhai and
Neves 2009). High profile firms including Motorola, General Electric, and Honeywell helped to promote
and legitimize the approach. The growth of Six Sigma has spawned a large literature and supporting
knowledge infrastructure for late adopters of the practice. Dozens of books are devoted to the topic, and a
large consulting base has emerged. Adoptions usually involve large resource expenditures, and they can
provide significant returns. As an example, General Electric purportedly spent upwards of $1.6 billion on
Six Sigma during 1996-1999 (Feng 2008). Related training costs as high as $50,000 for each trained
te
worker have been reported (Antony 2006; Fahmy 2006). Numerous successes and failures have been
recorded. A recent study of 214 adopting firms documented an average 4-year post-adoption increase in
Ac
ce
p
abnormal ROA of 0.83% (Swink and Jacobs 2012).

Administrative innovations such as Six Sigma affect the ways that organizational members conduct
their routine work (Sinha and Van de Ven 2005). Several researchers note that studies of the diffusion of
administrative innovations have tended to treat them as being homogeneous, neglecting practice variation
over time (Cool et al. 1997; Ansari et al. 2010). In reality, practices evolve over time as various
organizations adapt and reconfigure them to meet their specific needs and contexts (Robertson et al. 1996;
Rogers 2003; Strang and Kim 2004). As noted above, successful knowledge transfer comprises both
replication of discrete elements and adaptation of context dependent elements (Williams 2007). While
the foundational elements of the Six Sigma template are identifiable, distinct, and might be considered
discrete, the literature provides numerous examples of firms that have adapted Six Sigma to their business
contexts (Yu and Zaheer 2010; Zbaracki 1998; Schroeder et al. 2008; Shah et al. 2008; Wang and Chen
2010). Differences in adoptions include modifications to role structures (e.g., usage of Black Belt roles),
project performance metrics (e.g., customer satisfaction versus dollar savings), and combinations with
4
Page 5 of 35
other administrative innovations (e.g., Lean Six Sigma, Honeywells Six Sigma Plus, integration with
TRIZ) (Shah et al. 2008; Wang and Chen 2010).
2.2 Early Adopters versus Late Adopters
Early adopters of innovations typically enjoy first mover advantages due to novelty and to the
ip
t
development of advanced capabilities (Lieberman and Montgomery 1988). In addition, institutional

theory suggests that early adopters are more likely to adopt for efficiency reasons, rather than the
normative, mimetic, and coercive pressures that often drive late adoption (Westphal et al. 1997, Guler et
cr
al. 2002, Benner and Veloso 2008). These quasi-monopoly and motivational advantages suggest that
early adopters of innovations such as Six Sigma would gain greater benefits than late adopters. However,
us
early adopters also face greater uncertainties about the utility and characteristics of such processes. These
adopters confront more challenges since they have little to no knowledge about the characteristics and
benefits of a practice (Meyer et al. 2009). With few existing examples to observe, they will more likely
an
discover the need to adapt the practice only during implementation. Much experimentation, trial, and
error may be needed to resolve important unknowns about how new processes will fit with the technical,
cultural, and political characteristics of the adopting organization. Such activities promote learning, yet at
the same time they raise the total investment costs and time of implementation, and increase the level of
returns required to make the practice adoption financially positive.
Moreover, organizations that
encounter many difficulties in the practice implementation are more likely to abandon them.
te
Conversely, late adopters face fewer uncertainties about the utility and characteristics of an
administrative innovation, as knowledge stocks grow over time from the documented decisions and
Potential adopters can observe
Ac
ce
p
experiential accounts of early adopters (Terlaak and Gong 2008).
predecessors with similar traits, and assess the benefits of the practice. The more information they have
about the practice, the better decisions they can make about implementing the practice. Thus, they can
better understand the technical, cultural, and political characteristics of implementing Six Sigma. In
addition, late adopters can use the experiences of others, often captured in an existing knowledge
infrastructure (e.g., consultants, literature), to design more efficient implementation processes, thus
lowering implementation costs and maximizing positive impacts of the practice (Ritchie and Melnyk
2012). Bloom and Van Reenen (2010) discuss how consulting firms transfer management practices
across firms. More generally, research shows that organizations not only learn from direct experience, but
also from the experience of other organizations (Huber 1991).
Managers who are equipped with more information regarding administrative innovations such as Six
Sigma are likely to make more effective adoption decisions, and to develop efficient implementation
plans.
Early adopters have less information about the practice characteristics, which makes the
innovation more difficult to assess and implement. However, over time information and knowledge about
5
Page 6 of 35
the practice will accumulate from the experience of early adopters. This knowledge can spill over, or
transfer, from one organization to another (Henderson and Cockburn 1996). As a result, late adopters
have access to more knowledge about the practice, helping them to better understand and implement the
innovation.
greater than that of early adopters.
2.3 Factors Affecting an Organizations Ability to Successfully Adopt
ip
t
Hypothesis 1. The post-adoption operating performance of late adopters of Six Sigma is significantly
cr
Late adopters stand to benefit from the greater availability of information and knowledge about
administrative innovation characteristics. They can incorporate and apply the successes and failures of
us
early adopters. However, these benefits may be moderated by factors that either limit or enhance the
adopters abilities to successfully transfer and apply the growing knowledge stocks generated by earlier
adopters.
In general, we posit first that environmental factors may raise (or lower) the levels of
an
uncertainty regarding how to best implement Six Sigma practices, and second that structural factors may
also raise (or lower) the organizations ability to effectively implement Six Sigma.
2.3.1 Technological Velocity. An organizations ability to effectively adopt Six Sigma is affected by the
organizations learning capabilities, as well as its environmental dynamism and uncertainty. High-tech
firms and other organizations compete in uncertain, rapidly changing, fast clockspeed environments (Fine
2000). As a result, these high-velocity organizations typically possess capabilities that enable them to
te
quickly learn and adapt. McCarthy et al. (2010, p. 609) define technological velocity as the rate and
direction of change in the production processes and component technologies that underlie a specific
Ac
ce
p
industrial context Aldrich and Ruef (2006) describe how organizations learn by quickly cycling
through idea variation-selection-retention processes. Variation occurs when performance efforts fail to
meet targets; selection is the choice of variations that best meet targets; and retention is the routinization
of the selected variations. Because high-velocity organizations develop processes with quicker feedbacks
and shorter delays, they learn and adapt faster. In contrast, low-velocity organizations compete in more
stable and predictable, slower clockspeed environments. As a result, they cycle through variationselection-retention processes at slower rates. Sterman et al. (1997) note that low-velocity organizations
experience longer feedbacks and delays when implementing new innovations. Because high-velocity
companies are more experienced in embracing new technologies, they typically have more robust
implementation processes. For example, Gowen and Tallon (2005) found that high-tech Six Sigma
adopters had more comprehensive and thorough implementation planning processes, whereas low-tech
adopters were less likely to consider design factors related to competitive advantage. Because of these
differences in internal learning capabilities, early adopters in high-velocity industries are likely to develop
greater differential benefits than early adopters in low-velocity industries. In the early stage of an
6
Page 7 of 35
innovations life, there is little available knowledge, and successful implementation is more dependent on
a firms internal learning capabilities.
At the same time, high levels of uncertainty in high-velocity industries can make the future value of
new process capabilities hard to predict, while also lessening the value of available knowledge. Late
ip
t
adopters of Six Sigma who operate in highly uncertain and rapidly changing environments are likely to
find that existing knowledge stocks from earlier periods are not clearly applicable to their current and
continually changing environmental characteristics. In addition, because these organizations are adept at
cr
rapid learning, they are likely less dependent on learning from the experiences of early adopters. Thus,
we would expect that late adopters in high-velocity environments have few advantages over early
us
adopters in those same environments. In contrast, late adopters in more stable and predictable lowvelocity environments can more easily identify and apply specific knowledge that is needed to
successfully fit Six Sigma to their well-understood characteristics and business needs. And because they
an
are slower learners, late adopters in low-velocity environments are also likely to find greater benefits from
existing knowledge stocks that describe the experiences and lessons learned from foregoing early
adoptions. Hence, late adopters in stable, low-velocity industries are likely to have advantages over late
adopters in high-velocity industries, because they can more effectively transfer and apply existing
knowledge. Also, late adopters in low-velocity industries are likely to have advantages over early
adopters in low-velocity industries, because no such knowledge exists for early adopters. Collectively,
te
these arguments describe an interaction effect:
Hypothesis 2. The differential operating performance associated with late adoption of Six Sigma is
Ac
ce
p
greater for organizations in low-velocity industries than for those in high-velocity industries.
2.3.2 Business-to-Business and Business-to-Consumer Markets. The literature describes differences in
the competitive forces found in business-to-business (B2B) versus business-to-consumer (B2C) industrial
settings (e.g., Homburg and Frst 2005; Srinivasan et al. 2011). These differences are thought to affect an
organizations ability to effectively adopt an administrative innovation. For several reasons, B2B firms
have clearer customer requirements than B2C firms. First, B2B firms typically deal more intimately with
a much smaller number of clearly delineated individual business customers, as opposed to the thousands
or even millions of individual consumers that a B2C firm faces. Second, business transactions in a B2B
setting tend to be evaluated using more objective and explicitly specified criteria. As a result, B2B firms
are more likely able to clearly relate available knowledge regarding an administrative innovation like Six
Sigma to their immediate competitive and market contexts. They have less diverse, more predictable
customers, and better guidance on how to apply specific practices to their business needs. For example,
Sarkees (2011) found that B2B firms take better advantage of technological opportunities than B2C firms
do.
Hence, as greater knowledge emerges about the benefits and limitations of an administrative
7
Page 8 of 35
innovation like Six Sigma, late adopters in B2B market settings can more easily apply this knowledge to
their situation. Conversely, because B2C markets are more complex, late-adopting B2C firms likely have
less clarity about how to put such knowledge to proper use.
An additional factor favors late adopters of Six Sigma in B2B markets. Because connections between
ip
t
customers and suppliers are fewer, more interactive, and more objectively documented in B2B markets,
customers have greater insights into the inner workings of their suppliers, and they are likely to attach
greater value to suppliers adoptions of process improvement programs. Again, this distinction is salient
cr
for industrial customers as opposed to end consumers. Lemke et al. (2011, p. 846) found that in B2B
contexts, customers place a greater emphasis on firm practices that focus on understanding and delivering
us
value-in-use than is generally the case in B2C contexts. Industrial customers, using objective criteria,
are more likely to evaluate the innovative process adoptions of their suppliers directly. In contrast, the
customers of B2C firms evaluate firms more on the basis of indirect signals of quality, like brand image,
an
rather than actual practices (Glynn 2012). Indeed, industrial customers are more likely to be familiar with
Six Sigma programs, whereas B2C consumers may be entirely ignorant of them.
For these reasons, we expect that late adopters in B2B markets will have greater abilities to clearly
identify implementation concerns of Six Sigma, perhaps even with the help of their primary business
customers who may possess relevant knowledge, as they may themselves be Six Sigma adopters. In
addition, late adoptions are more likely to be appreciated by customers in B2B markets, therefore opening
te
up greater business opportunities. In later stages of the innovations life, enough time has passed for
institutional forces to create positive perceptions of adopters. Again this effect would be stronger in B2B
Ac
ce
p
markets, because business customers are more likely to be aware of Six Sigma, and to care more about the
inner workings of their suppliers than individual end consumers do. On the other hand, late adopters in
more complex B2C markets will have less guidance and less clarity in connecting Six Sigma adoptions to
competitive requirements.
And, because their customers (individual end consumers) are less
knowledgeable about innovations such as Six Sigma, they are likely to be of little help, and are likely to
attach little relevance to the adoption.
greater for organizations in B2B markets than for those in B2C markets.
2.3.3 Prior Performance. Over fifty years ago, Cyert and March (1963) first speculated about the role of
an organizations performance in either motivating or inhibiting innovation. They posited that poor
performance can be a strong motivator for organizations to seek new solutions, whereas good
performance might cause organizations to become complacent and less interested in change. At the same
time, Cyert and March (1963) noted that those high performing firms that did choose to innovate might
8
Page 9 of 35
have greater cushions of slack resources to enable greater commitment to experimentation, thus producing
more comprehensive innovation efforts.
Extending these notions to our research context leads to the next hypothesis. Organizations with
higher than average prior financial performance have more time and money to effectively implement new
ip
t
practices (Cyert and March 1963). Research shows that slack resources helps organizations learn and
better adapt (Wiersma 2007). Organizations with slack resources are not subject to the urgency of
remedying poor performance; they are less likely to be looking for quick solutions. For example, one of
cr
the authors worked with a high performing, internationally recognized health care organization that
recently sought to implement Six Sigma. This organization did considerable benchmarking and extensive
us
pilot testing, choosing to include some projects that were likely to succeed and some that were likely to
fail. The CEO led the effort by being among the first group to go through training, and by helping to
diagnose the pilot projects to determine how to best implement Six Sigma for the firms organizational
an
context. Organizations with this level of resource commitment can engage in more deliberate learning.
Late adopting low performers have fewer resources needed to avail themselves of emergent
knowledge resources. In contrast, late adopters that have enjoyed extra-ordinary financial performance
are better enabled to make maximal use of existing information. They can spend time and money
required to take full advantage of codified adoption experiences by other organizations, consulting, and
other available knowledge stocks. High performing late adopters also have greater wherewithal with
te
which to exploit external resources (e.g., consultants, advanced technologies) in implementing Six Sigma.
Early adopting high performers may also have similar levels of wherewithal, but they have fewer
Ac
ce
p
available knowledge resources on which to spend it. Again, these complementary issues argue for the
presence of an interaction effect:
greater for organizations with high pre-adoption financial performance than for those with low preadoption performance.
2.3.4 Organization Size. Our arguments for hypothesis 4 identify the value of available material
resources, as they may be applied toward the incorporation of externally available knowledge stocks for
Six Sigma. In addition to generating more resources through profitable operations, an organization may
have a relative abundance of applicable resources by virtue of its size. Larger firms are likely to have
more slack resources that can be applied to implementation programs. Even more important, larger firms
are likely to have more specialized resources needed to implement the new practice. Large firms
typically create corporate-level organizations that are dedicated to process innovation, optimization, and
standardization. For example, many large firms establish centers of excellence, or other organizational
units that perform technological scanning and knowledge repository services. Researchers have noted
9
Page 10 of 35
that such groups are useful sources of expertise in the implementation of Six Sigma (e.g., Sinha and Van
de Ven 2005), and that improvement experts or sensei are critical to implementing Lean techniques (e.g.,
Liker 2004). Larger organizations typically have dedicated assets, both human and other, that can help
them acquire, interpret and apply knowledge stocks that develop in later stages of an administrative
late adoption that small adopters will not.
ip
t
innovations life. For these reasons, we expect that large adopters of Six Sigma will have advantages in
greater for large organizations than for small organizations.

3. RESEARCH METHOD
us
3.1 Data Sample
cr
Given that both Swink and Jacobs (2012) and Shafer and Moeller (2012) found that Six Sigma adoption
positively impacts operating performance, our interest is not in replicating those results but instead
an
discerning the differential performance effects of adoption timing. To test our hypotheses, the data set
used is an expansion of the sample collected by Swink and Jacobs (2012). In addition, our study
examines several independent variables that act as contingencies on timing effects.

We used the same methods employed by Swink and Jacobs (2012) to expand their list of 214 Six
Sigma adopters. Varied sources practitioner journals, web searches, and academic journals were used
to tentatively identify additional firms that adopted Six Sigma. We verified the additional firms that
te
actually had adopted Six Sigma, and determined the years of their adoptions. We employed key words
including Six Sigma together with adoption or history to systematically search public documents
Ac
ce
p
(e.g., Factiva database publications, corporate websites, the Internet, company 10-K reports) for each of
the added firms. To enhance accuracy, two researchers independently determined the adoption dates. As
validation of their adoption date determinations, Swink and Jacobs (2012) surveyed the identified
adopting firms; 49 of 58 respondents were able to identify a specific adoption year, and 46 of those
(93.8%) concurred that the adoption year identified by the researchers was accurate within one year.
Please see Swink and Jacobs (2012) for further details. In our final sample, we kept only publicly traded
firms that adopted Six Sigma in or before 2007, to have enough data to estimate performance after
adoption. With these sources, we established definite adoption years prior to 2008 for an additional 34
firms, expanding the sample by 16% to 248 total firms.
Given that administrative innovations, including Six Sigma, have the greatest impact when widely
deployed throughout an organization, survey respondents were asked whether Six Sigma was deployed
companywide and, if not, what percentage of sales were represented by units adopting Six Sigma. Of the
58 respondents, 33 (56.9%) stated that Six Sigma was adopted companywide. For the 25 respondents that
did not report companywide adoption, the average response was that units representing 47.8% of
10
Page 11 of 35
company sales adopted Six Sigma in the first year, increasing to 66.8% of sales by the third year
following adoption. The survey data thus indicates a significant (majority) penetration across revenuegenerating units. Importantly, no significant differences in deployment percentages were noted across
time; i.e., deployment scope appears to be similar between early and late adopters.
ip
t
We present the number of adopting firms by year in Table 1, and match them to the widely-cited
categorization of Rogers (2003) as innovators, early adopters, early majority, late majority, or laggards.
The adoption years in our sample begin in 1986, and 2000 is the adoption year that occurs most
Given the continued importance of Six Sigma as indicated by current curriculums and
cr
frequently.
textbooks in business schools, the marked drop in post-2001 Six Sigma adoptions included in our sample
us
likely indicates either non-newsworthiness or saturation. In other words, Six Sigma is in advanced stages
of diffusion and has become business-as-usual, similar to TQM or Lean.
an
---------------------------Insert Table 1 about here

---------------------------The sample firms comprise 47 unique 2-digit SIC codes and 105 unique 3-digit SIC codes. The most
frequently occurring 2-digit SIC codes are: 1) chemicals (SIC28, 11.3% of the sample); 2) electrical and
electronic devices (SIC36, 11.3%); and 3) machinery (SIC35, 10.1%). The most frequently occurring 3digit SIC codes are: 1) motor vehicles (SIC371, 5.6% of the sample); 2) banks (SIC602, 4.8%); and 3)
computers (SIC357, 4.4%). Although most firms are in manufacturing industries, roughly one-third of
te
the sample firms are in service industries.
We present descriptive statistics for the firms in our sample in Table 2 Panel A. The statistics use
Ac
ce
p
data from the 2000 fiscal year, the most frequently occurring Six Sigma adoption year. The median
sample firm has a market value of $5.5B, total assets equal to $7.7B, sales of $6.5B, operating income
equal to $1.0B, and 28,300 employees. As a comparison, we present descriptive statistics in Table 2
Panel B for 173 suspected, publicly-traded Six Sigma adopters which we identified in our initial search of
journals and websites but for which we could not identify a precise adoption year. Comparisons of the
two samples revealed no statistically significant differences. In short, our sample represents a varied set
of industries, and appears to be unbiased in that there are no significant differences from the firms
suspected of adopting Six Sigma for which we did not identify a precise adoption year.
---------------------------3.2 Event Study Methodology
We followed the event study method recommended by Barber and Lyon (1996) regarding the timeframe
to measure the Six Sigma adoption effects on operating performance, the estimation approach for the
performance effects, and the analyses carried out to determine significance.
11
Page 12 of 35
Six Sigma and other process management programs often are implemented over several years.
Correspondingly, we examine operating performance effects using both year-to-year changes, and
changes over multiple-year periods. To pool data for the firms, each firms fiscal years were translated
into event years. Year 0 is defined as the fiscal year of Six Sigma adoption. The subsequent fiscal year is
ip
t
defined as year +1, and the previous fiscal year is defined as year 1. We analyze data starting with the
change in performance from year 1 to year 0 and concluding with the change in performance from year
+3 to year +4. Thus, our analysis timeframe comprises the six event years from year 1 to year +4. To
cr
isolate the performance effects of Six Sigma adoption from the effects of external factors including
economic conditions, we examine the abnormal operating performance for each adopting firm. As in
us
Swink and Jacobs (2012), abnormal operating performance is defined as:

Abnormal operating performance = Actual operating performance
Expected operating performance if Six Sigma adoption had not occurred
(1)
an
We focus on abnormal ROA as our dependent variable both because it is recommended by Barber
and Lyon (1996) as the best overall measure of operating performance, and because it is the primary
measure employed by Swink and Jacobs (2012) and Shafer and Moeller (2012) to estimate the operating
performance effects of Six Sigma adoption. To establish the expected operating performance in equation
(1), the conventional approach is to use firms similar to the adopting firms as controls. As Barber and
Lyon (1996) recommend, we match firms based on industry, prior performance, and size. In this paper,
te
we report only the results of the most conservative matching method employed by Swink and Jacobs
(2012), the median-performance-size-industry matching method. For each sample firm, the three steps
Ac
ce
p
in this matching method are:
Step 1. We identify all firms in the same 2-digit SIC code as each sample firm that have median ROA
in years 2, 3, and 4 (the years just before the study period) within 10% of the sample firms median
ROA, and that have median total assets in years 2, 3, and 4 within a factor of 25 of the sample firms
median total assets.
Step 2. If no firms are found in Step 1, we identify all firms in the same single-digit SIC code that are
within the 10% performance range and size within a factor of 25.
Step 3. If no firms are found in Step 2, we identify all firms regardless of SIC code that have the
10% performance range and factor of 25 size range.
To determine Six Sigma adoption effects, we exclude from our control groups any firms suspected of
Six Sigma adoption identified in our initial search. Given that it was not possible to conclusively
determine that all control firms had not adopted Six Sigma during the studied time frame, at least some of
the control firms might be adopters. Thus, the abnormal performance that we estimate should be
considered conservative. We also note that given the diffusion of Six Sigma over time, the conservatism
12
Page 13 of 35
of our abnormal performance estimates is likely greater for later adopters. To reduce the effects of
adopters who might be present in the control groups, our matching criteria are set so as to increase the
number of firms in each group, and thus limit the number of one-to-one comparisons.
Of the 248 firms in our sample, we dropped firms that had insufficient Compustat data in years 2,
ip
t
3, and 4 to calculate the baseline size or ROA metrics. We thus dropped ten firms from our analyses,
reducing our sample size to 238 firms. The most matches were achieved in Step 1 with the strictest
matching criteria; 229 of the 238 control groups (96.2%) were generated using Step 1. Eight control
cr
groups were generated via Step 2, and the remaining single control group was generated via Step 3. For
the control groups, the mean (median) size was 14.5 (9.0) firms, and only nine control groups consisted of
us
a single firm.
Expected operating performance for a sample firm if Six Sigma adoption had not occurred is the sum
of the firms actual performance in the base year, plus the median operating performance change for the
an
control group firms over the period of interest (as in equation (1)). Abnormal operating performance for a
sample firm with Six Sigma adoption is then equal to the sample firms actual performance (with Six
Sigma adoption) minus the sample firms expected performance (without Six Sigma adoption). For
example, if a sample firms control group experiences a 2.0% median ROA increase from years 1 to 0,
the sample firm would be expected to also increase its ROA by 2.0%. However, if the sample firm
actually experiences a 3.0% ROA increase from years 1 to 0, we would compute the sample firms
te
abnormal performance resulting from Six Sigma adoption to be 1.0%.

Using abnormal performance calculations is especially important in our analyses since we define
Ac
ce
p
early and late adoption as calendar time periods. Otherwise, macro-economic and industry factors would
be confounded with the early and late adoption periods. Since we obtain abnormal performance by
comparing a sample firm relative to a group of control firms from the same industry and the same time
period, we control for such exogenous factors. Further, we included measures of industry dynamism,
munificence, and competition as controls for abnormal performance in our regression analyses.
To compensate for non-normality and extreme values, we trimmed the data for all analyses at the
1.0% level in each tail. In light of non-normality, we emphasize nonparametric statistics and tests
although we also report means and changes in means using t-statistics. The Wilcoxon signed-rank test is
used to determine if the median of a single group is significantly different from zero, and the MannWhitney U-test is used to test for significant differences between group medians. We use the binomial
sign test to detect whether the percentage of sample firms with positive operating performance is
significantly different from the null of 50%, and the two-sample binomial sign test calculates the
significance of proportion differences between groups. Since we hypothesize increases in operating
performance due to late Six Sigma adoption, we report one-tailed significance tests except where noted.
13
Page 14 of 35
3.3 Measures
Employing multiple sources, we collect data for the factors presented in hypotheses 1-5. First, to test
hypothesis 1, we use the adoption year (year 0) to designate whether the firm is an early or late Six Sigma
ip
t
adopter. We define early and late adopters by aggregating the innovators, early adopters, and early
majority categories of Rogers (2003) as early adopters, and the late majority and laggards categories as
late adopters. See Table 1. This is also consistent with Ritchie and Melnyk (2012) who denote firms as
cr
early (late) adopters if their year 0 occurs before (after) the median adoption year. For robustness in our
classification of early and late adopters, we also present analyses excluding firms that adopted Six Sigma
us
in 1998-2000, the early majority category of Rogers (2003). To test the interactions of the contextual
factors in hypotheses 2-5 with early/late Six Sigma adopters, we use a dummy variable, Late. The Late
an
variable has a value of 1 if the firm is a late adopter, and a value of 0 if the firm is an early adopter.
Second, to determine whether firms are in high-velocity industries for testing hypothesis 2, we
consider the technology level and clockspeed of each industry in which the firm participates. Industries
are designated as high-velocity if they appear either as High-Technology in Kile and Phillips (2009,
Table 5) or as Fast Clockspeed in Nadkarni and Narayanan (2007, Table 1). For each sample firms
reported business segments, we use both the primary and secondary SIC codes reported in Compustat.
te
Primary SIC
Secondary SIC
High-velocity
High-velocity
Low-velocity
Low-velocity
High-velocity
Low-velocity
High-velocity
Low-velocity
Ac
ce
p
following table:
Portions of each segment are then designated high-velocity or low-velocity in accordance with the
% Highvelocity
100%
75%
25%
0%
% Lowvelocity
0%
25%
75%
100%
For some segments, only a primary SIC code is reported and, in those cases, it is the sole determinant of
100% high-velocity or 100% low-velocity.
We combine the segment data to compute the overall
percentage sales attributable to low-velocity and high-velocity industries for each firm:
HiVelocity_percent
LoVelocity_percent = 1 HiVelocity_percent
where salei is the firm sales reported for segment i in the Six Sigma adoption year; k is the total number of
the firms reported segments; and HiVelocityi is a variable with the value 1.00, 0.75, 0.25, or 0.00 as
described in the table above. To permit more weight on the primary focus of the firm (low-velocity or
high-velocity), we use the calculated percentages to compute a concentration ratio, denoted as Velocity:
14
Page 15 of 35
Velocity =
Velocity is continuous in the range [0, 1]; for Velocity = 1.00, all reported segments are classified as highvelocity; and for Velocity = 0.00, all reported segments are classified as low-velocity. As an example, in
its adoption year, General Electric reported operating in 10 business segments; 10.8% of its revenues
ip
t
were generated in segments rated as high-velocity, resulting in a Velocity concentration ratio of 0.015.
Third, to test the effect of B2B versus B2C orientation (hypothesis 3), we employ variable B2B_B2C.
cr
To compute B2B_B2C, we measure the extent to which firms primary customers are/are not endconsumers or wholesalers/retailers that sell to end-consumers. In other words, B2C-oriented firms are
us
producing products or services that are end items, that is, items that are in final (or near final) form for
consumers, whether or not they actually manage the sales transaction to consumers. In a similar manner
to Velocity, the variable B2B_B2C is computed as a concentration ratio based on percentage sales
an
attributable to segments that are primarily B2B. Three graduate students, independent of the research
team, separately determined the B2B or B2C designation for each sample firm segment; the Fleiss' kappa
for the three raters is 0.723, indicating acceptable inter-rater agreement. Disputes were resolved by the
majority opinion of the raters. As a calculation example, 47.1% of General Electrics adoption year
revenues were generated in business segments rated as B2B; its B2B_B2C concentration ratio is thus
0.443.
Fourth, to examine the effect of past performance, we computed industry-adjusted ROA for each
te
adopting firm as its ROA in year 2 minus the industry median ROA in year 2. Industry is defined as
all firms in the same 3-digit SIC code. We denote this proxy for prior financial performance as PFP.
Ac
ce
p
More positive PFP indicates better past performance relative to industry peers.
Last, we seek to gauge the influence of organization size on adoption performance. We use the
natural logarithm of total assets at the end of year 0 as our measure of size, and denote it as Size. For
robustness, we also operationalize Size using the natural logarithms of sales and employees in year 0.
We add both firm- and industry-level control variables in our regression analyses. Hendricks and
Singhal (2001) hypothesized and demonstrated a positive link between the benefits of quality
improvement programs and the firms labor intensity.
The inherent variability of labor-intensive
processes likely offers greater opportunities for variance reduction. Swink and Jacobs (2012) found that
labor intensity significantly influenced the operating performance benefits of Six Sigma adoption. Thus,
we include Labor_Intensity as a control variable; we measure it as the number of firm employees divided
by firm sales in year 0. Additional firm-level controls include R&D_Intensity, measured as the firms
R&D expenses divided by firm sales in year 0, and Ad_Intensity, measured as the firms advertising
expenses divided by firm sales in year 0. To control for the firms overall maturity, we include Age,
15
Page 16 of 35
defined as the firm age in years at the time of Six Sigma adoption. We also control for the firms quality
maturity using ISO9000_Experience as a proxy, defined as in Swink and Jacobs (2012) as the number of
years between year 0 and initial certification to ISO 9000; firms that were not ISO 9000 certified at the
time of Six Sigma adoption were coded as 1.
ip
t
Last, we control for the business environment of the firm by including industry-level controls for
dynamism, competition, and munificence. We again define the industry of the adopting firm as all firms
in the same 3-digit SIC code. As in Dess and Beard (1984), Munificence and Dynamism are defined as
cr
the slope coefficient and the standard error of the slope coefficient, respectively, obtained by regressing
sales over the five-year period prior to year 0, and dividing by the mean sales in that period. Competition
us
is measured via the Herfindahl Index.

4. RESULTS
We discuss our results beginning with a focus on the Six Sigma adoption timing effects on ROA, both on
an
a year-to-year basis and for multiple-year periods. We then discuss how our empirical results are
contingent on the environmental factors of technological velocity and B2B/B2C markets, and the
structural factors of prior performance and organization size.

4.1 Overall Timing Effects of Six Sigma Adoption
We present the results in Table 3 for abnormal changes in ROA on a year-to-year basis for the early
adopters, late adopters, and the differences between early and late adopters. Since our baseline period to
te
establish control groups is years 2, 3, and 4, we present the five yearly changes in abnormal
performance starting with the change from years 1 to 0 and continuing through the change from years +3
Ac
ce
p
to +4. For aggregate measures, we compute the changes in abnormal performance for three multiple-year
periods: from years 1 to +4; years 0 to +4; and years +1 to +4. By considering results across both
single- and multiple-year periods, we can see if the timing impacts of Six Sigma adoption begin at year 0
or are lagged, and if the effects persist. Strong significant results are indicated by general agreement in
sign and significance across the mean, median, and percent positive measures for a particular time period.
---------------------------Our first observation is that the year-to-year changes in abnormal ROA for the early adopters are
significantly positive only from years 1 to 0. The median (mean) improvement in ROA from years 1 to
0 is 0.515% (0.752%) for early adopters, significant at the 5% (1%) level. 60.40% of early adopters
experience positive changes from years 1 to 0, significantly greater than 50% at the 5% level. For late
adopters, the 0.374% median change in ROA from years +1 to +2 is significant at the 10% level. From
years +2 to +3, the median and mean changes per late adopter are 0.319% and 0.583%, respectively, both
significant at the 5% level, and 56.90% of firms experience positive changes, a value significantly greater
16
Page 17 of 35
than 50% at the 10% level. Similarly, from years +3 to +4, the median (mean) change per late adopter is
0.640% (0.403%), significant at the 10% (10%) level, and 58.8% of firms experience positive changes,
significant at the 5% level. The evidence suggests that early Six Sigma adopters experience significant
benefits from pre-adoption efforts while late Six Sigma adopters experience significant benefits beginning
ip
t
two years after adoption.

Considering the differences between early and late Six Sigma adopters, we see that differences
between the groups are significant for the last two annual changes in our study period. From years +2 to
cr
+3, the median and mean differential benefits for late adopters versus early adopters are 0.482% and
0.776%, respectively, both significant at the 5% level, and 11.54% more late adopters experience positive
us
changes than early adopters, a difference significant at the 5% level. From years +3 to +4, the median
(mean) differential benefits for late adopters is 0.642% (0.734%), significant at the 5% (10%) level, and
9.40% more late adopters experience positive changes than early adopters, a difference significant at the
an
10% level.
We next focus on the empirical results for the three multiple-year periods. For early Six Sigma
adopters, none of the median, mean, or percent positive changes over the multiple-year periods are
significantly positive. However, for late Six Sigma adopters, we find significantly positive results for all
three multiple-year periods. Focusing on the differences between early and late Six Sigma adopters over
the multiple-year periods, we see no significant difference for the 6-year period from years 1 to +4. This
te
is due to the significantly positive pre-adoption (years 1 to 0) performance of the early adopters noted
above. However, the median and mean changes for late adopters versus early adopters for the 5-year
Ac
ce
p
period from years 0 to +4 are both significantly different, and significantly more late adopters experience
positive changes compared to early adopters. For the 4-year period from year +1 to year +4, the median
and mean differences are both significant, as is the percentage of late adopters that experience positive
changes.

---------------------------As a robustness check, Table 4 presents data in the same format as Table 3, but excluding the sample
Six Sigma adopters in 1998-2000, the late majority category of Rogers (2003). This gives us a clear
separation between early and late adoption periods. Early adopters are redefined as those firms adopting
in 1997 or before, and late adopters remain as those firms adopting in 2001 or later. The sample of
matched firms is reduced from 238 to 163. We note that the pattern of results is largely similar to that of
Table 3, but with less significance, as expected from the smaller sample size. This additional analysis
increases our confidence that the greater benefits of late adoption are not an artifice of the sample
partition method. In summary, the annual and multiple-year changes, and the two early-versus-late
17
Page 18 of 35
partitions displayed in Tables 3 and 4 indicate that, on average, Six Sigma adoption generates a
significantly greater positive effect on ROA for late adopters compared to early adopters. These results
provide support for hypothesis 1.
Some might ask whether the increased ROA enjoyed by the late adopters is worth the wait. In other
ip
t
words, although late adopters experience greater abnormal performance, they are only able to accumulate
the increased performance for a shorter period, potentially resulting in overall lesser cumulative
performance over the early and late periods as compared to the early adopters. However, given the
cr
common assumption that firms and managers are profit maximizing, firms that did not adopt Six Sigma
during the early period likely invested in alternative initiatives or projects with similar or greater returns
us
than Six Sigma. Thus, the late adopters are able to enjoy similar performance to early adopters during the
early adoption period, plus the increased performance from investing in Six Sigma later.
4.2 Relating Timing Effects of Six Sigma Adoption to Contextual Factors
an
Before we examine the adoption period differential impacts of the contextual factors presented in
hypotheses 2-5, we first consider the impact of the factors main effects. Combining the hypothesized
variables with the control factors yields the following model:
Abnormal_ROA = + 1Late + 2Velocity + 3B2B_B2C + 4PFP + 5Size + 6Labor_Intensity

+ 7R&D_Intensity + 8Ad_Intensity + 9Age + 10ISO9000_Experience
(2)
+ 11Dynamism + 12Competition + 13Munificence +
where Late has the value 1 if firms adopted Six Sigma in 2001 or later, and the value 0 if firms adopted in
te
2000 or earlier. The other independent variables are as defined in section 3.4, and the dependent variable
Abnormal_ROA is the abnormal ROA change for the 5-year period from years 0 to +4. We note that
Ac
ce
p
analyses of the sample partitioned as early and late adopters shows no significant differences between the
groups in terms of Velocity, B2B_B2C, PFP, or Size.
Table 5 provides descriptive statistics and a correlation matrix for the independent and control
variables. The only significant correlation between the adoption time period variable Late and the other
hypothesized factors is with PFP.
Further we note that PFP is significantly correlated with the other
independent variables as well as several control factors.
Thus, we check for the influence of
multicollinearity by examining the Variance Inflation Factors (VIFs) of our variables; maximum VIFs less
than 10 are an indication that multicollinearity is not overly influencing the results (Kutner et al. 2005).
---------------------------To test the adoption period differential impacts proposed in hypotheses 2-5, we introduce the
interaction terms between the contextual factors and the Late variable. The expanded model is:
Abnormal_ROA = + 1Late + 2Velocity + 2aVelocityLate + 3B2B_B2C + 3aB2B_B2CLate +
4PFP + 4aPFPLate + 5Size+ 5aSizeLate + 6Labor_Intensity +
18
Page 19 of 35
7R&D_Intensity + 8Ad_Intensity + 9Age + 10ISO9000_Experience +

11Dynamism + 12Competition + 13Munificence +
(3)
Table 6 presents the estimated regression coefficients (standardized, with t-statistics in parentheses)
for models (2) and (3). We first consider the results for model (2). The F-value for model (2) is 1.851,
significant at the 5% level, and the adjusted R2 is 5.15%. We note that although the total portion of
ip
t
variance explained is low, it is in line with other cross-sectional analyses of abnormal financial
performance results (e.g., Hendricks and Singhal 2008). The maximum VIF for model (2) is 2.05 for
As hypothesized, the
cr
B2B_B2C, indicating that multicollinearity is not a concern in the model.
coefficient for Late is significantly positive (at the 1% level), indicating that Six Sigma adoption benefits
us
are greater for late adopters. The main effects of the other hypothesized factors are all insignificant.
Similar to the findings of Swink and Jacobs (2012), the control variable Labor_Intensity is positive and
significant (at the 5% level), indicating that Six Sigma is more beneficial in labor-intensive firms. The
an
control variable R&D_Intensity is also positive and significant (at the 5% level), suggesting that Six
Sigma is more beneficial in R&D-intensive firms.

---------------------------For model (3), the F-value is 3.017, significant at the 1% level, and the adjusted R2 is 14.39%. The
maximum VIF in model (3) is 3.47 for Size, again indicating that multicollinearity is not a concern in the
model. As in model (2), Labor_Intensity and R&D_Intensity are positive and significant (both at the 5%
te
level). Interestingly, Late remains significant (at the 1% level) as a main effect but its direction becomes
negative; Late also interacts significantly with each of the hypothesized variables. These results indicate
Ac
ce
p
that the performance effects of late or early adoption depend on the other moderating factors.
The simple effect of Velocity is insignificant, but the interaction of Velocity with Late is negative and
significant (at the 5% level). This indicates that the positive performance difference of late adopters over
early adopters is greater for low-velocity firms than for high-velocity firms. This finding supports
hypothesis 2. Figure 1a provides a plot of the interaction.
---------------------------Insert Figure 1 about here
---------------------------In model (3), the simple effect of B2B_B2C is negative and significant (at the 5% level) and the
interaction coefficient with Late is positive and significant (at the 1% level). This indicates that early
adopting B2C firms experienced greater performance impacts from Six Sigma than early adopting B2B
firms, but the improved performance of late versus early adopters is significantly greater for B2B firms
than for B2C firms. This finding supports hypothesis 3. Figure 1b illustrates the significantly greater
early-to-late adoption performance slope for B2B firms.
19
Page 20 of 35
The simple effect of PFP is insignificant, but its interaction with Late is positive and significant (at
the 1% level). These coefficients indicate that the improved performance of late versus early adopters is
greater for firms with greater pre-adoption financial performance. See Figure 1c for a plot of this
interaction. This finding supports hypothesis 4.
ip
t
The simple effect of Size in model (3) is negative and marginally significant at the 10% level,
indicating that in the early adoption period, small firms experienced greater performance impacts from
Six Sigma than large firms did. However, the SizeLate interaction is positive and significant (at the 1%
This positive interaction supports hypothesis 5, indicating that early-to-late performance
cr
level).
differences are greater for larger organizations. Figure 1d provides the interaction plot. We note that
us
using the natural logarithms of sales or employees as alternative operationalizations for Size in model (3)
yields no substantive differences.
Although we find the VIF for Size to be within acceptable limits, we analyze a third model as a
an
robustness check by deleting Size and the SizeLate interaction from model (3). The results are presented
as model (4) in Table 6. For model (4), the F-value is 2.585, significant at the 1% level, and the adjusted
R2 is 10.44%. The resulting coefficients and statistical significances for the remaining variables and
interactions are substantively similar to model (3). The exceptions are that the simple negative effects of
Late and B2B_B2C become less significant (now both at the 10% level) and the positive effects of the
te
5. DISCUSSION
B2B_B2CLate and PFPLate interactions become less significant (now both at the 5% level).
When coupled with the results of foregoing research, our findings suggest that the theories of
Ac
ce
p
organizational learning and knowledge transfer provide a useful explanation of adoption timing outcomes
for administrative innovations such as Six Sigma. The replication and application of practices required
for successful knowledge transfer are easier when more examples exist to serve as referents and
persuaders (Jensen and Szulanski 2007). This is particularly important for administrative innovations like
Six Sigma, since no external standards or common templates for the innovation exist. On the other hand,
innovations like ISO 9000 are themselves standards, with the explicit objective of assuring a common
process across organizations. This difference could explain why our findings differ from those of Guler
et al. (2002) and Benner and Veloso (2008), who both found that early adopters of ISO 9000
outperformed late adopters; they interpreted the finding as evidence for the notion that late adopters
experience more societal pressures to conform, and are thus less motivated to achieve the performance
benefits of their predecessors. Similarly, Gray et al. (2014) found that process compliance in the medical
device industry was greater for early ISO 9000 adopters, and they argued that the success of early
adopters was due to their internal motivation to adopt rather than adopting in response to the institutional
pressures that late adopters face. In a different study, Westphal et al. (1997) found in the context of the
20
Page 21 of 35
hospital industry that late TQM adopters largely replicated the normative model of TQM implementation,
did not adapt TQM to fit their specific needs, and consequently experienced less benefit from TQM.
While these foregoing studies focused on normative and institutional factors, our findings suggest that it
would be good for researchers to also consider environmental and structural contexts of the innovation
ip
t
adoptions, as these may also explain differences in performance across early and late adopters. Our
general conclusion is that the benefits of early or late adoption at least partially depend on the context in
which adoption is attempted, because context influences organizational learning.
cr
Quite interestingly, our results indicate that some early adopters in our sample did experience greater
financial benefits from Six Sigma than late adopters, but only if they operated primarily in high-velocity
us
industries, in a B2C market setting, had lower prior financial performance, and/or were small in size.
However, as an aggregated group, our sample of late adopters experienced a mean (median) 4-year ROA
gain of 1.323% (0.699%), whereas early adopters experienced no significant gains. In fact, the mean and
an
median ROA changes for early adopters as a group are nominally negative. This finding is interesting
given the hype that surrounded the reported successes of the earliest adopters of Six Sigma (e.g.,
Motorola, GE, Allied Signal). It suggests that institutional influences such as bandwagon effects may
actually have been stronger in the early phases of Six Sigmas life. Again, this finding is consistent with
our primary thesis; early adopters find implementation of Six Sigma more arduous due to the lack of
available guiding knowledge stocks and infrastructures.
On the other hand, if environmental and
te
structural conditions (velocity, customer diversity, lack of resources) also make implementation arduous
for late adopters, then the benefits of first mover advantages for early adopters are likely to prevail.
Ac
ce
p
Future research should investigate whether consistent effects are evident for other administrative
innovations.
Although the main effects of early versus late adoption are noteworthy, our greater contribution is in
considering how the hypothesized factors moderate the effects of Six Sigma adoption timing. Our
empirical results identify environmental and structural factors that amplify the potential advantages of late
adoption. Considering the industrial environment of the sample firms, we find that both technological
velocity and market orientation (B2B/B2C) moderate the effects of late adoption. Low-velocity and B2B
firms manifested more positive early-to-late performance differences than their counterparts in highvelocity and B2C firms did. As we discussed in our development of hypothesis 2, we expect that lowvelocity organizations can make better use of the additional information available in later periods of Six
Sigmas life cycle, because they operate in more stable and predictable technological environments.
Organizational needs are easier to identify and achieve when the technological future is easier to reliably
predict.
Stability also enables low-velocity organizations to more readily exploit implementation
templates that are developed over time (Szulanski and Jensen 2006).
For example, needed
21
Page 22 of 35
implementation refinements offered by an experienced consultant are likely to be easier to recognize and
apply. Hence, low-velocity organizations are better able to leverage the experience of their predecessors
to craft more complete and stable implementation regimes. Another way of viewing this result is that
low-velocity organizations are more reliant on the emerging knowledge stocks due to their relative
ip
t
inexperience in developing fast and efficient learning cycles. Low-velocity organizations are slower and
presumably less efficient in making needed enhancements to an adopted innovation like Six Sigma, thus
they depend more heavily on the knowledge provided by outside experts.
cr
We find that B2B firms experienced greater performance effects associated with late adoption than
did their B2C counterparts (see Figure 1b). With fewer customers, less market diversity, and a more
us
intimate understanding of their customers needs, B2B firms can more effectively use the additional
information gained from late adoption. It is also possible that, over time, B2B firms industrial customers
gained appreciation for Six Sigma as diffusion progressed. Conversely, early-adopting B2C firms appear
an
to have outperformed late adopting B2C counterparts. We attribute this to first mover advantages,
because advantages to late adoption are minimal for organizations operating in these diverse customer
markets.
Surprisingly, in the early adoption period B2C firms manifested significantly better adoption
performance gains than B2B firms did. We had no expectation of such an effect, and an explanation is
not immediately obvious. One possibility is that we may again see evidence of a bandwagon effect here.
te
In the early stages of Six Sigma diffusion, successful B2C adopters may have strongly encouraged
adoption by their suppliers. For example, early adopters including Motorola, General Electric, and Allied
Ac
ce
p
Signal are noted to have required adoption by their suppliers (Jaffe 1990). It may be that such coercive
pressure on B2B firms, rather than their rational choices, motivated their early adoption decisions,
resulting in poor performance relative to B2C firms that presumably were not subject to such coercive
pressures. Future research might explore this finding with a focus on possible differences in the nature of
Six Sigma benefits for B2C and B2B firms.
Consistent with hypothesis 4, the results suggest that positive prior financial performance does have a
synergistic effect on performance with late adoption (see Figure 1c). As we argued earlier, a firms
positive prior financial performance generates disposable income that can be converted into supportive
resources for implementation of Six Sigma. Researchers have also argued for an opposite effect based on
motivational elements. Kotter (1995), for example, suggests that poor prior financial performance serves
to create a more fertile environment for implementation of an administrative innovation, as managers are
likely to be more receptive to change, and more willing to challenge existing organizational policies and
procedures. A post hoc exploration of our data indicates that the correlation between prior performance
22
Page 23 of 35
and post-adoption performance is uniformly positive across the range of prior performance. We leave it
to future research to explore possible discontinuities in this relationship.
Similar to our finding for hypothesis 4, the findings for organization size provide strong evidence for
resource effects in improving late adoption success. We posited that larger organizations possess more,
ip
t
and more specialized, implementation resources. Such resources become more valuable as they have
more available codified knowledge and practical experience that they can apply. For example, Immaneni
et al. (2007) describe how Capital One, a Six Sigma adopter in 2005, recruited and hired available Six
cr
Sigma resources from other firms, and availed existing employees of extant Six Sigma training. Although
such training and knowledgeable potential employees were available to all late adopters, Capital One had
us
the required resources to pursue this strategy. Larger organizations also achieve minimum scales required
to justify specialization. Hence, larger organizations are more likely to have dedicated personnel who
give their full attention to implementation efforts. In particular, Six Sigma implementations are known to
an
be reliant on highly trained, dedicated specialists known as Black Belts. Smaller firms may view such
specialization as being prohibitively expensive.
On the other hand, our data indicate that in the early adoption period smaller firms experienced
marginally significant greater performance improvements from adopting Six Sigma than larger firms.
Again this is an unexpected finding. We conjecture that this could be due to the greater challenges during
early innovation diffusion that we outlined in our development of hypothesis 1, coupled with the greater
te
need to adapt to varied contexts in larger firms. Larger firms tend to be more diverse, hurting them when
customization is difficult and guidance from external knowledge sources has not yet evolved. The
Ac
ce
p
benefits of specialized resources in larger firms are negated in the early period, as there are few existing
knowledge stocks for them to exploit. Also, larger firms are more likely victims of their own past
successes, and hence more resistant to change, particularly during early stages of innovation diffusion
(e.g., Tushman and OReilly 2002). Further research is required to investigate these potential causes.
6. CONCLUSIONS AND LIMITATIONS
Operations managers often face the challenge of deciding when to implement administrative innovations
like Six Sigma. A comparison of our findings against prior research leads us to conclude that late
adopters of Six Sigma fare better, but the benefits of late adoption can be amplified or can disappear,
depending on the organizational context. Specifically, late Six Sigma adopters tend to experience greater
differential operating performance gains if they are in industries with low technological velocity and/or
B2B market orientation, if they have good pre-adoption financial performance, and if they are large.
Conversely, organizations in the opposite circumstances find little advantage to late adoption. First
mover advantages accruing to early adopters tend to dominate operating performance outcomes in these
circumstances.
23
Page 24 of 35
More broadly, this research highlights the importance of deciding when to adopt an administrative
innovation. Scholars have studied the benefits of implementing these practices, but few have evaluated
performance effects of adoption timing. Contrary to some of the literature on first mover advantages
(e.g., Lieberman and Montgomery 1988), sometimes the early bird does not get the worm. If later
ip
t
adopters can better understand the innovation and apply it to their business context, the late birds might
gather more worms.
A limitation of our study is its focus on Six Sigma. As evidenced by our discussion of the traits and
cr
characteristics of Six Sigma (see Section 3.1), we believe it is an important management practice, and
provides a useful representation of an administrative innovation. Nevertheless, future research should
us
investigate other innovations with the aim of comparing Six Sigma and other innovations directly. As
with any administrative innovation, there are particular aspects of Six Sigma that are somewhat unique.
For example, as a process management strategy, Six Sigma might adversely affect the exploratory efforts
an
of high tech firms (Benner and Tushman 2002, 2003). Further research examining the empirical evidence
offered by other administrative innovations, particularly non-process management strategies (e.g.,
digitization), will help to clarify this.
A further limitation of our research is our inability to conclusively determine that all control firms did
not adopt Six Sigma during the early and late adoption periods. This is of special concern for the late
adopter sample given the saturation level of Six Sigma adoption we expect during the advanced stages of
te
diffusion. If Six Sigma adoption is non-newsworthy during the late adoption period, our sample might
suffer from selection bias as only notable adoptions might be announced. However, our use of multiple
Ac
ce
p
sources (including third-party sources) protects us from self-promoting firms. Further, any selection bias
would manifest itself in the main effect of late adoption, and not in the hypothesized interaction effects
that our results demonstrate.
While our hypotheses are based on the premise that firms will seek to learn and understand Six Sigma
processes to fit their particular needs, this study does not address the specific ways that Six Sigma or
other administrative innovations should be implemented. Future research should seek to specify how Six
Sigma (and other administrative innovation) processes can be most effectively tailored to particular
technical, environmental, structural, and cultural characteristics of a firm. Future studies might also
examine a firms motivation for adoption, along with other environmental and structural variables that
potentially moderate the effects of adoption timing on performance. Important variables could include
market and competitive conditions, the organizations strategic focus, its degree of product
diversification, and its prior related knowledge. Although we characterize Six Sigma as an administrative
innovation, we note that certain components of Six Sigma such as the application of advanced statistical
analyses are more technical in nature than administrative.
Examining the relative importance and
24
Page 25 of 35
contributions of technical versus administrative components of Six Sigma is fruitful future research.
Successful adoption of Six Sigma can also result in technical innovations that improve the firms products
and/or processes. Studying the relationship between Six Sigma adoption and any resulting technical
innovations, and the potential mediating role of those technical innovations between Six Sigma and
ip
t
financial performance is also important future work. Studies of these effects would likely require
collection of primary data.
Our sample is made up of mostly large firms, all of which are publicly-traded and based in the U.S.,
cr
though many are global. Given the significant effects we find for firm size related to Six Sigma adoption
timing, it would be fruitful future research to examine adoption timing in small and privately-held firms.
us
Finally, we also neglect issues surrounding adoption of practices across international borders. There is
debate over the merits of heavy adaptation of practices that are transferred from one national setting to
another, pitting the value of fit to local context against the loss of comparability with the original template
an
implementations of the practice (Szulanski and Jensen 2006; Yu and Zaheer 2010). It would be useful for
future research to examine this trade-off.
References
te
Aldrich, H.E., Ruef, M. 2006. Organizations Evolving, 2nd ed. Sage, London.
Ansari, S.M., Fiss, P.C., Zajac, E.J. 2010. Made to fit: How practices vary as they diffuse. Academy of
Management Review 35(1), 67-92.
Ac
ce
p
Antony, J. 2006. Six Sigma for service processes. Business Process Management Journal 12(2), 234-248.
Barber, B.M., Lyon, J.D. 1996. Detecting abnormal operating performance: The empirical power and
specification of test statistics. Journal of Financial Economics 41(3), 359-399.
Benner, M. J., Tushman, M.L. 2002. Process management and technological innovation: A longitudinal
study of the photography and paint industries. Administrative Science Quarterly 47(4), 676-706.
Benner, M. J., Tushman, M.L. 2003. Exploitation, exploration, and process management: The
productivity dilemma revisited. Academy of Management Review 28(2), 238-256.
Benner, M.J., Veloso, F.M. 2008. ISO 9000 practices and financial performance: A technology coherence
perspective. Journal of Operations Management 26(5), 611-629.
Bloom, N., Van Reenen, J. 2010. Why do management practices differ across firms and countries?.
Journal of Economic Perspectives 24(1), 203-224.
Braunscheidel, M.J., Hamister, J.W., Nallan, C.S., Star, H. 2011. An institutional theory perspective on
Six Sigma adoption. International Journal of Operations & Production Management 31(4), 423-451.
Cool, K.O., Dierickx, I., Szulanski, G. 1997. Diffusion of innovations within organizations: Electronic
switching in the Bell system, 1971-1982. Organization Science 8(5), 543-559.
25
Page 26 of 35
Corbett, C.J., Montes-Sancho, M.J., Kirsch, D.A. 2005. The financial impact of ISO 9000 certification in
the United States: An empirical analysis. Management Science 51(7), 1046-1059.
Cyert, R.M., March, J.G. 1963. A Behavioral Theory of the Firm. Prentice-Hall, Englewood Cliffs CA.
Dess, G.G., Beard, D.W. 1984. Dimensions of organizational task environments. Administrative Science
Quarterly 29(1), 52-73.
ip
t
Fahmy, D. 2006. When one black belt is enough. Treasury and Risk Management 16(3), 55-59.
Feng, Q. 2008. Six Sigma: Continuous improvement toward excellence. A.K. Kamrani, E.A. Nasr, eds.
Collaborative Engineering: Theory and Practice. Springer, New York, 43-60.
cr
Fine, C.H. 2000. Clockspeed based strategies for supply chain design. Production and Operations
Management 9(3), 213-221.
us
Frynas, J. G., Mellahi, K., Pigman, G.A. 2006. First mover advantages in international business and firmspecific political resources. Strategic Management Journal 27(4), 321-345.
Glynn, M.S. 2012. Analyzing the B2B brand value chain. M.S. Glynn, A.G. Woodside, eds., Business-toBusiness Marketing Management: Strategies, Cases and Solutions. Emerald, Bingley UK, 245-261.
an
Goh, T.N., Low, P.C., Tsui, K.L., Xie,M. 2003. Impact of Six Sigma implementation on stock price
performance. TQM and Business Excellence 14(7), 753-763.
Gowen III, C.R., Tallon, W.J. 2005. Effect of technological intensity on the relationships among Six
Sigma design, electronic-business and competitive advantage: a dynamic capabilities model study.
Journal of High Technology Management Research 16(1), 59-87.
Gray, J.V., Anand, G., Roth, A.V. 2014. The influence of ISO 9000 certification on process compliance.
Production and Operations Management, doi 10.1111/poms.12252.
te
Gutirrez, L.J.G., Llorns-Montes, F.J., Snchez, .F.B. 2009. Six Sigma: From a goal-theoretic
perspective to shared-vision development. International Journal of Operations & Production
Management 29(2), 151-169.
Ac
ce
p
Guler, I., Guilln, M.F., Macpherson, J.M. 2002. Global competition, institutions, and the diffusion of
organizational practices: The international spread of the ISO 9000 quality standards. Administrative
Science Quarterly 47(2), 207-232.
Henderson, R., Cockburn, I. 1996. Scale, scope, and spillovers: The determinants of research productivity
in drug discovery. RAND Journal of Economics 27(1), 32-59.
Hendricks, K.B., Singhal, V.R. 2001. Firm characteristics, Total Quality Management, and financial
performance. Journal of Operations Management 19(3), 269-285.
Hendricks, K.B., Singhal, V.R. 2008. The effect of product introduction delays on operating performance.
Management Science 54(5), 878-892.
Homburg, C., Frst, A. 2005. How organizational complaint handling drives customer loyalty: An
analysis of the mechanistic and the organic approach. Journal of Marketing 69(1), 95-114.
Huber, G.P. 1991. Organizational learning: The contributing processes and the literatures. Organization
Science 2(1), 88-115.
Immaneni, A., McCombs, A., Cheatham, G., Andrews R. 2007. Capital One banks on Six Sigma for
strategy execution and culture transformation. Global Business and Organizational Excellence 26(6),
43-54.
Jaffe, C.A. 1990. Some resent being pushed into quality. The Morning Call, March 25.
26
Page 27 of 35
Jensen, R.J., G. Szulanski. 2007. Template use and the effectiveness of knowledge transfer. Management
Science 53(11), 1716-1730.
Kerin, R.A., Varadarajan, P., Rajan, P., Peterson, R.A. 1992. First-mover advantage: A synthesis,
conceptual framework, and research propositions. Journal of Marketing 56(4), 3352.
ip
t
Kile, C.O., Phillips, M.E. 2009. Using industry classification codes to sample high-technology firms:
Analysis and recommendations. Journal of Accounting, Auditing, and Finance 24(1), 35-58.
Kinney, M.R., Wempe, W.F. 2002. Further evidence on the extent and origins of JITs profitability
effects. The Accounting Review 77(1), 203-225.
cr
Kotter, J.P. 1995. Leading change: Why transformation efforts fail. Harvard Business Review 73(2), 5967.
us
Kutner, M.H., Nachtsheim, C.J., Neter, J., Li, W. 2005. Applied Linear Statistical Models, 5th ed.
McGraw-Hill Irwin, New York.
an
Lemke, F., Clark, M., Wilson, H. 2011. Customer experience quality: An exploration in business and
consumer contexts using repertory grid technique. Journal of the Academy of Marketing Science
39(6), 846-869.
Lieberman, M.B., Montgomery, D.B. 1988. First-mover advantages. Strategic Management Journal
9(S1), 41-58.
Liker, J. 2004. The Toyota Way. McGraw-Hill, New York.
McCarthy, I.P., Lawrence, T.B.,Wixted, B., Gordon, B.R. 2010. A multidimensional conceptualization of
environmental velocity. Academy of Management Review 35(4), 604-626.
Meyer, K.E., Estrin, S., Bhaumik, S.K., Peng, M.W. 2009. Institutions, resources and entry strategies in
emerging economics. Strategic Management Journal 30(1), 61-80.
te
Nakhai, B., Neves, J.S. 2009. The challenges of Six Sigma in improving service quality. International
Journal of Quality & Reliability Management 25(7), 663-684.
Ac
ce
p
Ritchie, W.J., Melnyk, S.A. 2012. The impact of emerging institutional norms on adoption timing
decisions: Evidence from C-TPAT A government antiterrorism initiative. Strategic Management
Journal 33(7), 860-870.
Robertson, M., Swan, J., Newell, S. 1996. The role of networks in the diffusion of technological
innovation. Journal of Management Studies 33(3), 333-359.
Rogers, E.M. 2003. Diffusion of Innovations, 5th ed. Free Press, New York.
Sarkees, M. 2011. Understanding the links between technological opportunism, marketing emphasis and
firm performance: Implications for B2B. Industrial Marketing Management 40(5), 785-795.
Schroeder, R.G., Linderman, K., Liedtke, C., Choo, A.S. 2008. Six Sigma: Definition and underlying
theory. Journal of Operations Management 26(4), 536-554.
Shafer, S.M., Moeller, S.B. 2012. The effects of Six Sigma on corporate performance: An empirical
investigation. Journal of Operations Management 30(7-8), 521532.
Shah, R., Chandrasekaran, A., Linderman, K. 2008. In pursuit of implementation patterns: The context of
Lean and Six Sigma. International Journal of Production Research 46(23), 6679-6699.
Sinha, K.K., Van de Ven, A.H. 2005. Designing work within and between organizations. Organization
Science 16(4), 389-408.
27
Page 28 of 35
Srinivasan, R., Lilien, G., Sridhar, S. 2011. Should firms spend more on research and development and
advertising during recessions? Journal of Marketing 75(3), 49-65.
Sterman, J.D., Repenning, N.P., Kofman, F. 1997. Unanticipated side eects of successful quality
ip
t
programs: Exploring a paradox of organizational improvement. Management Science 43(4), 503-521.
cr
Strang, D., Kim,Y. 2004. Diffusion and domestication of managerial innovations: The spread of scientific
management, quality circles, and TQM between the United States and Japan. S. Ackroyd, P.
Thompson, P. Tolbert, R. Batt, eds. The Oxford Handbook of Work and Organization. Oxford
University Press, Oxford, 177-199.
us
Swink, M., Jacobs, B.W. 2012. Six Sigma adoption: Operating performance impacts and contextual
drivers of success. Journal of Operations Management 30(6), 437-453.
an
Szulanski, G., Jensen, R.J. 2006. Presumptive adaptation and the effectiveness of knowledge transfer.
Strategic Management Journal 27(10), 937-957.
Teece, D.J. 1980. The diffusion of an administrative innovation. Management Science 26(5), 464-470.
Terlaak. A., Gong, Y. 2008. Vicarious learning and inferential accuracy in adoption processes. Academy
of Management Review 33(4), 846-868.
Tushman, M.L., OReilly, C.A. 2002. Winning through innovation: A practical guide to leading
organizational change and renewal. HBS Publishing, Boston.
Wang, F.K., Chen, K.S. 2010. Applying Lean Six Sigma and TRIZ in banking services. TQM and
Business Excellence 21(3), 301-315.
te
Westphal, J.D., Gulati, R., Shortell, S.M. 1997. Customization or conformity? An institutional and
network perspective on the content and consequences of TQM adoption. Administrative Science
Quarterly 42(2), 366-394.
Ac
ce
p
Wiersma, E. 2007. Conditions that shape the learning curve: Factors that increase the ability and
opportunity to learn. Management Science 53(12), 1903-1915.
Williams, C. 2007. Transfer in context: Replication and adaptation in knowledge transfer relationships.
Strategic Management Journal 28(9), 867-889.
Yeung, A.C.L., Cheng, T.C.E., Lai, K. 2006. An operational and institutional perspective on Total
Quality Management. Production and Operations Management 15(1), 156-170.
Yu, J., Zaheer, S. 2010. Building a process model of local adaptation of practices: A study of Six Sigma
implementation in Korean and US firms. Journal of International Business Studies 41(3), 475-499.
Zbaracki, M. 1998. The rhetoric and reality of TQM. Administrative Science Quarterly 43(3), 602-636.
Zu, Z., Frendall, L.D., Douglas, T.J. 2008. The evolving theory of quality management: The role of Six
Sigma. Journal of Operations Management 26(5), 630-650.
28
Page 29 of 35
Table 1: Adoption year frequency for the 248 sample firms with Six Sigma adoption years prior to 2008
ip
t
cr
10.08%
Innovators
(2.5%)
Early Adopters
(13.5%)
us
2.02%
2.42%
2.82%
3.23%
3.23%
3.23%
4.03%
5.65%
12.50%
18.55%
26.61%
44.76%
62.50%
70.56%
78.23%
86.69%
an
13
16
4
5
6
7
8
8
8
10
14
31
46
66
111
155
175
194
215
32.26%
Early Majority
(34%)
41.94%
Late Majority
(34%)
228
244
248
13.31%
Laggards
(16%)
te
2005
2006
2007
1
1
1
1
0
0
2
4
17
15
20
45
44
20
19
21
Rogers (2003)
Group
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Frequency of Six Sigma adoption years

Group %
Frequency Cum
Cum %
2
2
0.81%
0
2
0.81%
2.42%
2
4
1.61%
Year
91.94%
98.39%
100.00%
Ac
ce
p
Table 2. Descriptive statistics for fiscal year 2000, the most frequent sample Six Sigma adoption year
Panel A: Sample N = 248)

firms (
Market
Total
Operating
Employee
Sales
Value
Assets
Income
s (000s)
($M)
($M)
($M)
($M)
Media
5,537.9
7,697.0
6,526.0
1,001.0
28.3
n
Mean
28,678.5
34,928.9
13,797.3
2,672.9
51.6
Std
65,113.7
94,208.5
20,740.1
4,869.9
66.6
Dev
Max
476,115.5
715,348.0
170,064.0
37,342.0
416.5
Min
22.1
30.0
42.9
(257.3)
0.4
Panel B: Suspected Six Sigma adopting firms without known
N = 173)
adoption years
(
Market
Total
Operating
Employee
Sales
Media
n
Mean
Std
Dev
Max
Min
Value
Assets
($M)
($M)
3,318.9
3,544.5
20,082.0
21,284.8
43,798.1
66,568.0
301,238.4
673,814.0
0.4
0.2
Income
s (000s)
($M)
($M)
3,646.2
458.5
15.4
11,297.7
1,712.7
46.9
25,460.5
3,640.7
118.4
206,083.0
33,309.0
1,244.0
< 0.01
(19.0)
< 0.01
29
Page 30 of 35
Table 3. Annual abnormal changes in ROA for year 1 through year +4 for all sample firms
Year-to-Year
- 1 to 0
0 to 1
Single-Year Periods
1 to 2
2 to 3
3 to 4
- 1 to 4
Multi-Year Periods
0 to 4
1 to 4
Early Adopters (2000 or earlier)

N
101
100
98
97
95
96
95
95
-1.113%
-2.146
-0.541%
-0.955
-1.116%
-2.047
-0.119%
-0.248
38.95%
-2.155
46.32%
-0.718
110
110
0.515%
2.298 **
-0.055%
0.016
-0.081%
-0.074
-0.163%
-1.068
-0.002%
-0.780
0.060%
0.587
Mean
t -statistic
0.752%
2.423 ***
0.007%
0.023
0.110%
0.357
-0.193%
-0.761
-0.331%
-1.009
0.671%
1.064
% positive
Z -statistic
60.40%
2.090 **
48.00%
-0.400
46.94%
-0.606
45.36%
-0.914
49.47%
-0.103
50.00%
0.000
125
126
123
116
107
110
Median
Z -statistic
-0.500%
-1.266
0.178%
1.252
0.374%
1.534 *
0.319%
1.986 **
0.640%
1.622 *
0.348%
1.385 *
1.592%
2.566 ***
0.699%
2.229 **
Mean
t -statistic
-0.483%
-1.388
0.317%
1.121
0.366%
1.090
0.583%
2.060 **
0.403%
1.310 *
0.869%
1.402 *
1.597%
2.599 ***
1.323%
2.329 **
% positive
Z -statistic
42.40%
-1.699
53.17%
0.713
54.47%
0.992
56.90%
1.486 *
58.88%
1.837 **
54.55%
0.953
60.91%
2.288 **
57.27%
1.526 *
0.482%
2.197 **
0.642%
1.684 **
0.288%
0.534
2.705%
3.327 ***
1.240%
2.203 **
0.776%
2.006 **
0.734%
1.633 *
0.199%
0.224
2.713%
3.257 ***
1.442%
1.903 **
11.54%
1.677 **
9.40%
1.339 *
4.55%
0.652
21.96%
3.136 ***
10.96%
1.566 *
-1.015%
-2.524
0.233%
0.821
0.455%
1.127
Mean
t -statistic
-1.235%
-2.588
0.310%
0.728
0.256%
0.550
% positive
Z -statistic
-18.00%
-2.691
5.17%
0.773
7.53%
1.113
te
Median
Z -statistic
an
Differences in groups (late minus early)
cr
us
Late Adopters (2001 or later)
ip
t
Median
Z -statistic
Ac
ce
p
All samples trimmed at 1.0% each tail; Significance is one-tailed: * p = .10; ** p = .05; *** p = .01
Z -statistics for group medians are obtained using Wilcoxon Signed-Rank tests; Z -statistics for differences between group medians are obtained
using Mann-Whitney U -test
Z -statistics for % positive are obtained using Binomial Sign tests; Z -statistics for differences between group % positive are obtained using twoproportion Binomial sign tests
30
Page 31 of 35
Table 4. Annual abnormal changes in ROA for year 1 through year +4, excluding sample firms with
adoption years 1998 through 2000 to match categorization of Rogers (2003)
1 to 0
0 to 1
Single-Year Periods
1 to 2
2 to 3
3 to 4
1 to 4
Innovators and Early Adopters (1997 or earlier)

29
29
28
27
27
Median
Z -statistic
0.329%
1.195
0.214%
0.632
-0.173%
-0.189
-0.038%
-0.780
1.014%
1.147
0.621%
0.811
Mean
t -statistic
0.556%
0.893
0.507%
0.794
-0.035%
-0.049
-0.655%
-1.473
0.776%
0.933
% positive
Z -statistic
58.62%
0.928
51.72%
0.186
41.38%
-0.928
50.00%
0.000
123
116
Late Majority and Laggards (2001 or later)
27
-0.222%
-0.502
0.189%
-0.318
1.465%
1.228
-0.463%
-0.417
-0.147%
-0.151
59.26%
0.962
51.85%
0.192
46.15%
-0.392
51.85%
0.192
107
110
110
110
126
Median
Z -statistic
-0.500%
-1.266
0.178%
1.252
0.374%
1.534 *
0.319%
1.986 **
0.640%
1.622 *
0.348%
1.385 *
1.592%
2.566 ***
0.699%
2.229 **
Mean
t -statistic
-0.483%
-1.388
0.317%
1.121
0.366%
1.090
0.583%
2.060 **
0.403%
1.310 *
0.869%
1.402 *
1.597%
2.599 ***
1.323%
2.329 **
% positive
Z -statistic
42.40%
-1.699
53.17%
0.713
54.47%
0.992
56.90%
1.486 *
58.88%
1.837 **
54.55%
0.953
60.91%
2.288 **
57.27%
1.526 *
Differences in groups (late minus early)
an
125
1 to 4
26
cr
29
us
Multi-Year Periods
0 to 4
ip
t
Year-to-Year
-0.830%
-1.601
-0.036%
-0.050
0.546%
0.593
0.357%
1.913 **
-0.374%
-0.469
-0.273%
-0.157
1.814%
1.610 *
0.510%
1.301 *
Mean
t -statistic
-1.039%
-1.328
-0.189%
-0.284
0.400%
0.519
1.238%
2.007 **
-0.374%
-0.421
-0.595%
-0.430
2.061%
1.498 *
1.470%
1.181
% positive
Z -statistic
-16.22%
-1.574
1.45%
0.141
-0.38%
-0.036
2.69%
0.251
14.76%
1.357 *
5.42%
0.507
te
Median
Z -statistic
Ac
ce
p
13.09%
1.270
6.90%
0.657
All samples trimmed at 1.0% each tail; Significance is one-tailed: * p .10; ** p .05; *** p .01
Z -statistics for group medians are obtained using Wilcoxon Signed-Rank tests; Z -statistics for differences between group medians are obtained using MannWhitney U -test
Z -statistics for % positive are obtained using Binomial Sign tests; Z -statistics for differences between group % positive are obtained using two-proportion
Binomial sign tests
31
Page 32 of 35
ip
t
cr
Mean Std Dev
Min
Max
(1)
(2)
Late
(1)
0.530
0.500
Velocity
(2)
0.213
0.390
.034
B2B_B2C
(3)
0.630
0.461
-.054
(4)
0.057
0.220
Size
(5)
8.914
1.734
Labor_Intensity
(6)
0.005
0.003
-0.366
1.356
.184
-.091
R&D_Intensity
(7)
0.032
0.051
Ad_Intensity
(8)
0.010
0.021
(9)
81.5
48.7
ISO9000_Experience (10)
4.1
4.2
-1
0.024
0.016
0.003
Dynamism
(11)
0.380
-.061
0.131
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
***
.396
-.033
-.058
-.406
-.136
**
.174
***
***
.084
.279
***
.490
.020
-.265
***
**
-.423
.157
***
.309
***
.165
.187
-.533
**
**
-0.179
***
-.112
***
***
***
-.141
**
.135
***
-.134
**
.106
-.125
.010
***
1
*
314
-.139
.269
-.056
-.043
.069
13
.251
.214
.300
.246
-.034
.078
.330
-.084
.088
0.094
.065
.000
.109
.011
-.065
-.016
-.099
-.034
-.036
.037
ce
pt
Age
0.027
.084
***
4.060 13.820
0.000
(4)
ed
PFP
(3)
M
an
Variables
us
Table 5. Descriptive statistics and Spearman correlation matrix of the independent and control variables for the 205 observation points used in
regression analyses
***
***
***
Competition
(12)
0.231
0.184
0.021
.077
-.061
.134
.235
-.184
Munificence
(13)
0.071
0.083
-0.116
0.405
-.112
.105
-.028
-.008
.116
***
***
.324
-.126
.128
.023
.030
-.076
-.088
-.186
1
***
.025
***
.096
.287
***
-.186
1
***
-.204
Ac
Significance is two-tailed: * p .10; ** p .05; *** p .01
33
Page 33 of 35
Table 6. Estimated coefficients (standardized, t-Statistics in parentheses) from regressions of abnormal

ROA change from year 0 to year +4
Intercept
H1. Late
Velocity
-0.060
(-2.899) **
1 if Six Sigma adoption year 2001;
0 if Six Sigma adoption year 2000
Concentration ratio of percentage sales in high
velocity segments; continuous [0,1]
0.227
(3.056) ***
-0.138
(-1.508)
B2B_B2C
Concentration ratio of percentage sales in B2B

segments; continuous [0,1]
PFP
Industry-adjusted ROA
0.107
(1.287)
H4. PFP Late

ln(Total Assets)
H5. Size Late
Size
Employees / Sales
R&D_Intensity
R&D Expenses / Sales
Ac
ce
p
te
Labor_Intensity
-0.024
(-0.242)
an
H3. B2B_B2C Late
Model (4) without Size
-0.008
(-0.349)
-0.016
(-0.653)
-0.542
(-2.646) ***
-0.365
(-1.809) *
0.093
(0.698)
0.033
(0.369)
0.102
(0.747)
-0.487
(-2.426) **
-0.458
(-2.237) **
-0.276
(-2.357) **
-0.198
(-1.836) *
us
H2. Velocity Late
Model (3)
Interactions
with Late
ip
t
Model (2)
Main effects
only
Operationalization
cr
Independent Variable
0.536
(3.545) ***
-0.220
(-1.596)
0.416
(3.126) ***
0.321
(2.304) **
-0.196
(-1.405)
0.347
(2.584) **
-0.210
(-1.746) *
0.369
(3.201) ***
0.187
(2.552) **
0.164
(2.322) **
0.160
(2.290) **
0.181
(2.144) **
0.161
(2.007) **
0.172
(2.088) **
Ad_Intensity
Advertising Expenses / Sales
0.010
(0.125)
0.026
(0.334)
-0.002
(-0.023)
Age
Firm age at adoption year
0.071
(0.901)
0.063
(0.843)
0.094
(1.269)
-0.081
(-0.958)
-0.092
(-1.112)
-0.086
(-1.040)
ISO9000_Experience Adoption year minus 1st ISO9000

certification
Dynamism
Std error of beta from regressed sales

over 5 years
0.007
(0.099)
0.012
(0.181)
0.005
(0.071)
Competition
Herfindahl Index
0.042
(0.576)
0.085
(1.215)
0.060
(0.846)
Munificence
Beta from regressed sales over 5 years
0.088
(1.168)
0.041
(0.559)
0.046
(0.625)
205
1.851 **
11.19%
5.15%
205
3.017 ***
21.52%
14.39%
205
2.585 ***
17.02%
10.44%
Number of observations
Model F value
2
R
2
Adjusted R
2
F value for change in R from Model (3)
6.156 ***
--
5.365 ***
All samples trimmed at 1.0% each tail; Significance is two-tailed: * p .10; ** p .05; *** p .01
34
Page 34 of 35
ip
t
cr
us
an
Ac
ce
p
te
Figure 1. Significant interaction effects between hypothesized variables and period of Six Sigma
adoption on abnormal changes in ROA (%) for year 0 through year +4 for all sample firms,
trimmed at 1.0% in each tail. Plots are for model (3) with High/Low settings for variables of
interest at either 1/0 (for variables Late, B2B_B2C, and Velocity) or mean 1 (for variables PFP
and Size) with all other significant variables held constant at mean values.
35
Page 35 of 35

Jacobs 2015 - New

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jacobs 2015 - New

Uploaded by

Copyright:

Available Formats

Accepted Manuscript

Title: Performance Effects of Early and Late Six Sigma

PERFORMANCE EFFECTS OF EARLY AND LATE SIX SIGMA ADOPTIONS

PERFORMANCE EFFECTS OF EARLY AND LATE SIX SIGMA ADOPTIONS

appropriate adoption timing to increase performance.

Without such knowledge,

Such administrative innovations are not strictly defined.

2. SIX SIGMA ADOPTION AND TIMING

variation in organizational processes by employing improvement specialists, a structured method, and

abnormal ROA of 0.83% (Swink and Jacobs 2012).

development of advanced capabilities (Lieberman and Montgomery 1988). In addition, institutional

Moreover, organizations that

experiential accounts of early adopters (Terlaak and Gong 2008).

these arguments describe an interaction effect:

And, because their customers (individual end consumers) are less

greater for large organizations than for small organizations.

3.1 Data Sample

examines several independent variables that act as contingencies on timing effects.

---------------------------Insert Table 1 about here

the sample firms are in service industries.

Swink and Jacobs (2012), abnormal operating performance is defined as:

Expected operating performance if Six Sigma adoption had not occurred

in this matching method are:

abnormal performance resulting from Six Sigma adoption to be 1.0%.

We combine the segment data to compute the overall

The inherent variability of labor-intensive

is measured via the Herfindahl Index.

structural factors of prior performance and organization size.

two years after adoption.

---------------------------Insert Table 4 about here

variables with the control factors yields the following model:

Abnormal_ROA = + 1Late + 2Velocity + 3B2B_B2C + 4PFP + 5Size + 6Labor_Intensity

Further we note that PFP is significantly correlated with the other

independent variables as well as several control factors.

Thus, we check for the influence of

7R&D_Intensity + 8Ad_Intensity + 9Age + 10ISO9000_Experience +

B2B_B2C, indicating that multicollinearity is not a concern in the model.

---------------------------Insert Table 6 about here

available guiding knowledge stocks and infrastructures.

On the other hand, if environmental and

Stability also enables low-velocity organizations to more readily exploit implementation

For example, needed

digitization), will help to clarify this.

Examining the relative importance and

future research to examine this trade-off.

Liker, J. 2004. The Toyota Way. McGraw-Hill, New York.

programs: Exploring a paradox of organizational improvement. Management Science 43(4), 503-521.

Frequency of Six Sigma adoption years

Panel A: Sample N = 248)

Early Adopters (2000 or earlier)

Differences in groups (late minus early)

Late Adopters (2001 or later)

Innovators and Early Adopters (1997 or earlier)

Late Majority and Laggards (2001 or later)

Differences in groups (late minus early)

Mean Std Dev

Significance is two-tailed: * p .10; ** p .05; *** p .01

Table 6. Estimated coefficients (standardized, t-Statistics in parentheses) from regressions of abnormal

Concentration ratio of percentage sales in B2B

H4. PFP Late

H5. Size Late

R&D Expenses / Sales

H3. B2B_B2C Late

Significance is two-tailed: * p .10; p .05; * p .01