Skew and heavy‐tail effects on firm performance
| Author | Christine M. Chan,Shige Makino |
| Published date | 01 August 2017 |
| Date | 01 August 2017 |
| DOI | http://doi.org/10.1002/smj.2632 |
Strategic Management Journal
Strat. Mgmt. J.,38: 1721–1740 (2017)
Published online EarlyView 9 March 2017 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2632
Received 5 October 2015;Final revision received15 December 2016
SKEW AND HEAVY-TAIL EFFECTS ON FIRM
PERFORMANCE
SHIGE MAKINO1*and CHRISTINE M. CHAN2
1Department of Management, The Chinese University of Hong Kong, Shatin,
Hong Kong
2School of Business, The University of Hong Kong, Pokfulam,
Hong Kong
Research summary: Most strategic management studies adopt an average-centered view that
uses the central tendency to explain between-group variation in performance (i.e., performance
differences between business units, rms, industries, and countries). In this study, we explain
within-group variation using a variance-centered viewthat focuses on the peripheral characteris-
tics of performance distributions as dened by skew and heavy tails (i.e., variance and kurtosis).
Drawing on performance feedback theory, we hypothesize that successful rms tend to develop
a positive skew in their performance distributions, which we call a “positive skew effect” in this
study, and that heavy tails moderate this effect.Our analysis of the performance of a group of for-
eign afliates provides general support for our hypotheses at both the rm and segment (industry
and country) levels.
Managerial summary: Managers of multi-business rms use various approaches to improve the
aggregate performance of their business units. Some expand the range of upper performance
outliers (exploration) or reduce the range of loweroutliers (downsizing); others improve the per-
formance of current business units (exploitation). We nd that rms with superior performance
tend to have a balanced mix of the three approaches. We also nd that segments (countries and
industries) with higher mean performances provide environments that facilitate the entry of pro-
ductive rms and the exit of unproductive rms and provide environments in which incumbents
can further improve their performance by learning from others. We observe that successful rms
and segments have a positive skew in their performance distributions, which we call a “positive
skew effect.” Copyright © 2016 John Wiley & Sons, Ltd.
INTRODUCTION
The primary purpose of strategic management
research is to identify the sources of variation in
rm performance. Drawing on industrial organi-
zation theory and resource- and institution-based
views, previous studies have identied a variety of
Keywords: analysis of outliers; heavy-tail effects;
non-normal distribution; skew effect; variance-centered
view
*Correspondence to: Shige Makino, Department of Management,
Chinese University of Hong Kong, Shatin, N.T., Hong Kong.
E-mail: makino@baf.cuhk.edu.hk
Copyright © 2016 John Wiley & Sons, Ltd.
“conditions” (e.g., industry structure, rm resources
and capabilities, and institutions) that explain per-
formance variation. However, these “conditions”
do not sufciently explain performance varia-
tion; studies examining the decomposition of
the effects of business units, rms, industries,
sub-national regions, and countries on performance
variation have revealed that these effects explain
at most 50% of the observed variation (Chan,
Makino, and Isobe, 2010; Ma, Tong, and Fitza,
2013; Makino, Isobe, and Chan, 2004; McGahan
and Porter, 1997; McGahan and Victer, 2010;
Rumelt, 1991).
1722 S. Makino and C. M. Chan
Underlying the conventional condition-based
explanations is the assumption that once a theo-
retically dened condition is met, all of the rms
will behave in a coherent manner (i.e., similar to
average rms) and achieve a similar level of rm
performance. We call this conventional perspective
the average-centered view. This view implicitly
assumes that average events (behavior or perfor-
mance) in a given condition represent the essential
characteristics of all of the events in the condition
and that events that deviate from the average are
“errors in theory.” Researchers have attempted to
decrease such errors by identifying omitted condi-
tions (i.e., developing new theories or combining
current theories). Although such attempts have
certainly advanced our knowledge of strategic man-
agement, they often result in “an undue emphasis
on the development of theory at the expense of
research which observes and reports actual facts”
(Hambrick, 2007: 1346).
The average-centered view, which is deeply
embedded in strategic management studies, also
makes the assumption that events are independent
and will generate Gaussian (normal) distribu-
tion (McKelvey and Andriani, 2005). As the
average-centered view focuses on the central
(average) tendencies of the observed events, it
considers outliers or residual values that deviate
from the average as unimportant sources of infor-
mation or even as noise in the descriptions of the
characteristics of the observed events (Andriani
and McKelvey,2007, 2009). For example, a typical
OLS imposes assumptions on the errors, including
homoscedasticity (variance of the errors is the
same across all levels of explanatory variables),
no autocorrelation (the errors are independent
of one another), and normality (the errors are
normally distributed) with the zero average value.
However, according to the behavioral tradition of
management research, these assumptions are often
not appropriate for strategic management practices,
as the behavior of rms is typically characterized
as interdependent, self-selected, and biased. There
are three reasons for this characterization. First, a
rm’s strategic behavior can hardly be considered
random or independent as it is (a) inuenced by
the rm’s past behavior and the choices made
by other rms (Cyert and March, 1963), (b)
inherently “endogenous” to the rm’s expected
performance outcomes (Hamilton and Nickerson,
2003; Shaver, 1998), and (c) subject to bias and
bounded rationality (Kahneman and Tversky,1979;
Simon, 1965) (i.e., a violation of the assumption of
independence). Second, rms do not only choose
behavior that follows the behavior of average rms;
their behavior may deviate from that of average
rms. Firms have different orientations in learning
(i.e., exploitation and exploration) that result in
variations in strategic behavior and hence in perfor-
mance outcomes, even between rms in the same
situations (March, 1991) (i.e., a violation of the
assumption of homoscedasticity). Third, rms are
biased toward behavior that produces superior per-
formance. Firms have a natural incentive to retain
high performing business units and improve or
terminate underperforming business units (Porter,
1991), resulting in a non-symmetrical distribution
of performance among the retained business units
around the mean (i.e., a violation of the assumption
of normality).
Although the average-centered view can explain
how mean performance varies between conditions
and can identify the conditions that lead to dif-
ferences in the mean performance, which we call
the effects on the mean, it fails to explain how
the shape of performance distribution varies within
conditions (i.e., the range and conguration of per-
formance variation within conditions), which we
call the effects on the variance. To achieve a pre-
cise understanding of the sources of strategic behav-
ior and performance variation, the conventional
average-centered view should be complemented by
a view that explains effects on the variance, which
we call the variance-centered view.
The variance-centered view can be extended in
two directions to examine two different aspects
of strategic management that the average-centered
view cannot fully explain. First, it can be extended
to identify factors that explain effects on the
variance. This approach has been widely adopted
by researchers examining inequities or disparities
in social and economic behavior and outcomes
(i.e., unequal regional economic growth within
and between countries; and income, consumption,
and expense disparities between individuals).1
Second, it can be extended to explain the possible
associations between effects on the mean and
effects on the variance. The average-centered view
1There has been some methodological development in this
approach. Western and Bloome (2009), for example, developed
a method of examining V(w), called “variance function regres-
sion,” which estimates the effects of covariates on both the mean
and variance (i.e., conditional variance measured by residual dis-
persion) of a dependent variable.
Copyright © 2016 John Wiley & Sons, Ltd. Strat. Mgmt. J.,38: 1721–1740 (2017)
DOI: 10.1002/smj
To continue reading
Request your trial