Strategizing with Others Misperceptions of Luck in Extreme Performances

Chengwei Liu
Chengwei.Liu@wbs.ac.uk
Warwick Business School, UK

Abstract
How performance is perceived and attributed has important implications for strategizing. But
decades of research in cognitive and social sciences suggests that people tend to be fooled by
randomness and mistake luck for skill, particularly when evaluating extreme performances. I
argue that these predictable mistakes can be translated into a source of competitive advantage:
informed managers can exploit others misperceptions of the luck involved in extreme
performances, such as by arbitraging in strategic factor markets. I then discuss limits to this
arbitrage strategy due to social constraints: employees or stakeholders may not be able to
accurately evaluate performances and may therefore discount atypical strategic activities and
attribute them to managerial incompetence, suggesting that only managers who are less
sensitive to this lemon problem can take advantage of the resulting arbitrage opportunities.
I conclude with a flowchart about whether one should pursue an arbitrage opportunity
resulted from misperceptions of luck. More generally, this paper suggests an alternative
source of strategic opportunity by turning the biases documented in the literature on their
head and instead seeing behavioral strategy as providing ways to identify and exploit these
opportunities with social savvy.

Keywords: luck, biases, strategic factor market, arbitrage, behavioral strategy

I am grateful for comments from Jim March, Phanish Puranam, Loizos Heracleous, Arkadiy
Sakhartov, Jerker Denrell, Christina Fang and participants at the 2015 Behavioral Strategy
conference at New York University. This research was partially supported by the British
Academy and Leverhulme Trust [Grant SRG 34963]. Please do not cite or circulate this
working paper without the authors permission.

1
1. INTRODUCTION
How performances are perceived and attributed have important implications for strategizing.
For example, if a success is attributed to a strategy, the manager who formulated this strategy
is likely to be rewarded and promoted, and the strategy will be replicated and diffused (Cyert
& March, 1963; Haunschild & Miner, 1997; Strang & Macy, 2001). However, decades of
research in cognitive and social sciences suggest that people tend to make predictable
mistakes in attributions (Jones, 1979; Kelley, 1971; Ross, 1977). For example, the
Fundamental Attribution Error suggests that outcomes are usually over-attributed to the
person (e.g., the traits or skills of the strategist) rather than to the situation (e.g., being at the
right time and right place) (Ross & Nisbett, 1991). This suggests that people systematically
underestimate how events could have unfolded differently due to factors beyond individuals
control, implying that managers can be praised (blamed) for good (bad) luck.

The way people mistake luck for skill is particularly problematic when evaluating extreme
performances. Recent studies demonstrate that exceptional performances tend to happen in
exceptional circumstances rather than being a reliable indicator of exceptional skill (Denrell
& Fang, 2010; Denrell, Fang, & Liu, 2015; Denrell & Liu, 2012). It follows that extreme
performances should be attributed more to the situations the actors happen to be in (Liu & de
Rond, 2016). For example, exceptional successes are unlikely to occur without strong
performance reinforcements (Denrell, Fang, & Zhao, 2013; Denrell & Liu, 2012). A mediocre
actor can get lucky early and her initial fortune reinforced to such an extent that she becomes
the eventual star who receives all the fame and credits, whereas her more competent
counterparts become obscure due to early bad luck (Arthur, 1989; Frank & Cook, 1995;
Merton, 1968; Salganik, Dodds, & Watts, 2006). Similarly, research on disaster dynamics
suggests that many failures should be attributed to circumstances such as tightly coupled,
interdependent systems (Perrow, 1984). Even skilled operators can fail to stop a cascade of
small errors becoming something more serious once a tipping point is passed (Rudolph &
Repenning, 2002). The theoretical implication is that a non-monotonic pattern between
performance and the expected level of skill may exist, as illustrated in Figure 1A. Higher
performances indicate higher skill only in the region of moderate performances. Top
performers may not have the highest level of expected skill, while the worst performers may
not have the lowest level of expected skill (see the appendix for the mechanism that generates
this pattern). Their extreme performances reflect a strong, enduring impact of luck so one
should regress more to the mean when evaluating these outliers (Denrell & Liu, 2012).
(A) modelling result, 10 million simulations (B) Major League Baseball, 2000-10 (C) USA firms,1980-2010
1.5 0.65
100
ROA percentile at year t+1

1 0.6
winning percentage

75
Expected skill

2nd half season

0.5 0.55

0 0.5 50

-0.5 0.45
25
-1 0.4

-1.5 0.35 1
-5 0 5 0 0.2 0.4 0.6 0.8 1 1 25 50 75 100
Performance winning percentage ROA percentile at year t
1st half season

Figure 1. Three illustrations of the non-monotonic associations between performance and

expected skill (or future performances) when extreme performances indicate unreliability. See
the appendix for the method and robustness analyses.

Importantly, the pattern in Figure 1A is not merely a theoretical possibility. Figure 1B and 1C
illustrate its empirical plausibility using data from sport and firm performances. Figure 1B is
based on data of 2,980 game results during 2000-2010 in Major League Baseball to illustrate

2
that extreme team winning percentages in the first half of a season is likely to regress
(downward for high and upward for low) more to the mean (i.e., 50% winning percentage) in
the second half of the same season than their less extreme counterpart, i.e., the second
best/worst performances in the first half of the season. The firm performance data is based on
return on assets (ROA) of 7,147 public firms in the United States during 1980-2010. Figure
1C is an illustration based on year 2004-2005 pair with 3,216 firms survived both years. In
particular, the top 5% performing firms (i.e., the best 200 firms) during year t perform
systematically worse than the top 5%-10% (i.e., the second best) during year t+1.1 I
examined 660 combinations of different years and different intervals binning. In 608 of these
examinations (92.1%) the top performing group during year t does not have the highest
expected performance ranks at year t+1 (see the appendix for more details about how the data
were collected and analyzed and their robustness). Overall, these empirical illustrations
suggest that extreme performances are neither a reliable indicator of the underlying traits of
the actors involved (e.g., teams, firms) nor a reliable predictor for the future performances of
the current top performers.

Nevertheless, people tend to romanticize and attribute extreme performances more to the
persons involved than to the situations and predict that these extreme performances are likely
to continue due to these persons (Meindl, Ehrlich, & Dukerich, 1985; Ross & Nisbett, 1991),
exactly the opposite to the above normative account. These misattributions of luck can have
severe consequences. For example, competent but unlucky managers can be fired after
failures, e.g., managers of Major League Baseball teams who lose most of the games in the
first half of a season, as illustrated in Figure 1B (see also Feiler & Taylor, 2013). The
organization may develop false confidence from this decision because the performance is
likely to improve after a new manager joins. But the performance changes may have nothing
to do with the new manager but instead reflect regressing upward to the mean after an
extremely low performance, implying a version of superstitious learning (Levitt & March,
1988; Liu & de Rond, 2016; Zollo, 2009). If this suboptimal learning occurs in an
interdependent system like nuclear plants, the system may remain tightly coupled and fragile,
awaiting the next normal accident (Perrow, 1984). Importantly, even with clear cues that
extreme performance cannot happen without extreme luck, experiments show that people still
tend to believe the extreme performers must have done something right (or wrong) to deserve
the extreme reward or punishment (Denrell & Liu, 2012). The implication is that peoples
tendency of mistaking luck for skill is the most salient and consequential when evaluating
extreme performances.

I argue that the predictable misperceptions of luck about extreme performances can be
translated into a source of competitive advantage. Conventional approaches to this problem
focus on how to more accurately evaluate performances through de-biasing the evaluators
(Bazerman & Moore, 2009; Kahneman, 2011; Kahneman, Lovallo, & Sibony, 2011). Here I
focus on how informed managers can take advantage of others predictable mistakes in order
to gain an advantage (Gavetti, 2012). For example, informed managers can arbitrage in a
strategic factor market (Barney, 1986) by hiring managers who have been unfairly blamed for
1
An interesting observation in Figure 1C is that the non-monotonic pattern also exists for poorer performances.
But it occurs not at the extreme as our theoretical model predicts but around the bottom 20%-30% firms. A
modification of our model can reproduce this pattern. Consider firms whose performances at period t are so low
that threaten their survival (March & Shapira, 1992) but not low enough to give up hopes. If we assume that
these firms (e.g., bottom 20%-30%) are more likely to take more risk at period t+1 but their outcomes are likely
variable and binomial, i.e., either very high or very low (Denrell, 2008; Denrell et al., 2015). This reproduces the
pattern of regressing upward to the mean around the 25th percentile performance, i.e., some of these low
performing firms future performances improve due to the favorable outcome of their risk-taking.

3
failure (i.e., buying low) or selling exceptionally performing business units to nave buyers
who ignore regression to the mean (i.e., selling high). The key is that peoples misperceptions
of luck about extreme performances can lead to systematic over- or under-estimation in
strategic factor markets that enable informed managers to identify and arbitrage the market
inefficiencies.

However, there is a limit to this arbitrage strategy due to a lemon problem in markets
(Benner & Zenger, 2016). Financial market can be inefficient because traders can lose capital
support from their funder if the formers trading strategy is too sophisticated for the latter,
resulting in fewer arbitrage activities than it needs for eliminating market inefficiencies
(Shleifer & Vishny, 1997). This analogy can be applied to strategic factor markets (Barney,
1986; Denrell, Fang, & Winter, 2003): managers can lose both capital and social support if
stakeholders fail to comprehend their strategy. Exploiting others misperceptions of luck
implies that managers often have to act against the crowds misperceptions of luck. Arbitrage
strategies such as hiring the fired or dismissing the stars are likely to be discounted due to
their uniqueness and atypicality (Correll, Ridgeway, Zuckerman, & Jank, 2012; Litov,
Moreton, & Zenger, 2012). My theory predicts that managers who are less sensitive to social
and stakeholders constraints (e.g., private equity, see Benner and Zenger, 2016) are more
likely to engage in such arbitrage. The implication is that fortune may favor the less sensitive
because there is a limit to arbitrage for the more sensitive ones even when they are smart
enough to identify the misevaluation opportunities.

More generally, this paper suggests an alternative source of strategic opportunity by turning
the biases documented in the literature on their head. Decades of research in Organization
Behavior/Theory, Judgment and Decision-making and Strategy have documented numerous
suboptimal organizational processes such as decision biases, learning myopia and
organizational inertia (Bazerman & Moore, 2009; Hannan & Freeman, 1984; Levinthal &
March, 1993). These biases are counterproductive for the focal company, but I argue that they
in fact illuminate ways for informed managers and entrepreneurs to gain an advantage by
exploiting these predictable mistakes. This paper focuses on one particular bias, e.g.,
misperceptions of luck about extreme performances. Strategy researchers and practitioners
can extend this approach by examining the extent to which other biases can be translated into
an advantage and their scope conditions. Overall this approach suggests an additional branch
of Behavioral Strategy (Gavetti, 2012; Powell, Lovallo, & Fox, 2011) as providing ways to
identify and exploit arbitrage opportunities in strategic factor markets with social savvy.

The paper is structured as follows. In the next section, I firstly define what luck is. I then
review the cognitive and social reasons for mistaking luck for skill, before emphasizing that
these biases are mostly problematic when evaluating extreme performances. Next, I outline
how these predictable biases can be translated into advantage and discuss the caveats due to
limits to arbitrage. I conclude the paper by presenting a decision flowchart about whether one
should pursue an arbitrage opportunity resulted from misperceptions of luck, and discussing
the implications of this arbitrage strategy for theories and practices, particularly the idea of
behavioral strategy as (overcoming limits to) arbitrage in strategic factor markets.

2. COGNITIVE AND SOCIAL REASONS FOR THE MISPERCEPTIONS OF LUCK

2.1 What Is Luck?
A discussion of the misperceptions of luck requires a definition of luck. Most observed
outcomes are a combination of skill and luck (Kahneman, 2011; Makridakis, Hogarth, &
Gaba, 2009; Weiner et al., 1971). Skill is referring to the dispositional factors of the actors

4
involved in an outcome, such as individual trait, effort or organizational routines. Luck is
referring to the situational factors that contribute to an outcome, such as uncontrollable
factors, unintended consequences or chance (de Rond & Thietart, 2007). Some outcomes are
almost entirely about skill. For example, one cannot win an Olympic 100 meter gold medal by
luck alone. Other outcomes are almost entirely about luck. For example, one cannot
intentionally lose in a game of chance such as roulette or lotteries (Mauboussin, 2012). Most
outcomes in management are between the two extremes above. The question is whether an
outcome is more about skill or more about luck and the extent to which we can accurately
evaluate their relative roles in an outcome.2 A recent article reviewed the most common
applications of luck in the management literature and its implication for management (Liu &
de Rond, 2016). I follow the same approach and define luck as a psychological attribution for
the persons or organizations that significantly benefit (or suffer) from events that are beyond
their control or foresight. Liu and de Rond (2016) also emphasized a common feature of luck
attributions: they are changeable and not always accurate. As I will elaborate below, prior
studies have documented many reasons why people tend to systemically underestimate the
role of luck and instead mistake luck for skill.

2.2 Cognitive Reasons for the Misperceptions of Luck

Luck as a theoretical construct was first introduced in Attribution Theory (Kelley, 1971;
Weiner et al., 1971). Attribution Theory suggests that people tend to attribute outcomes to
four factors. Two of them are internal or dispositional factors: skill and effort and the other
two factors and external or situational ones: task difficulties and luck. One shared
characteristic of research on attributions is that people have various attribution biases. For
example, people tend to make self-serving attributions (Miller & Ross, 1975): own successes
are over-attributed to own skill and effort while own failures are over-attributed to bad luck or
timing. This is to mistake good luck for superior skill when evaluating successes and to
mistake poor skill for bad luck when evaluating failures. The implication is that people can
become overconfident and develop illusion of control (Langer, 1975) after successes while
people can repeat unnecessary failures due to a self-serving attribution bias.

There are three additional attributions biases that may lead people to systematically mistake
luck for skill. The first is the Fundamental Attribution Error (Ross & Nisbett, 1991): people
tend to underestimate the impact of situational factors when evaluating observed outcomes.
Second, people tend to judge the quality of a decision based on its outcome rather than its
process an Outcome Bias (Baron & Hershey, 1988). Third is a Hindsight Bias (Fischhoff,
1975): people tend to exaggerate the probability of a realized event after the fact even when
the event could have unfolded differently. If history could be repeated, the successful (or
failed) person could have been replaced by others with slight changes in the situational
factors. For example, twentieth century technology development will likely enable a
software empire to emerge, but its founder could have been another lucky person other than
Bill Gates, as he admitted in his own autobiography (Gates, 1995). The implication of these
biases is that we may give too much credit (and blame) for the observed extreme successes
2
Some could argue that this position omits an important aspect of luck, e.g., how skill and luck interacts. For
example, many people believe that one makes her own luck. I agree with this conventional wisdom as long as
it is applied to moderate rather than extreme performances. For extreme performances, the impact of situational
factors usually overwhelms the impact of the persons involved. Moreover, I am not arguing that all exceptional
performers are unskilled. As illustrated all three graphs in Figure 1, the top performers tend to have high skill,
but, importantly, not the highest skill. This is because an actor/firm without the highest skill can get lucky and
sometimes acquire top performance due to the enduring impact of luck. If this can occur, exceptional
performance is an unreliable indicator of skill and attributing them to exceptional luck is more sensible. For
more detailed discussion see Denrell et al., (2015) and Liu and de Rond (2016).

5
(and failures) to the persons involved when they may just be at the right time and right place
(or wrong time and wrong place).

Another damaging bias is peoples ignorance of regression to the mean. Higher performances
may indicate higher levels of skill, but also a stronger degree of luck. Since luck is the
uncontrollable factor that does not tend to persist, a more extreme performance is likely to be
followed by a less extreme performance, regressing towards the mean. Importantly, regression
to the mean is a statistical phenomenon that requires no causal explanation for the systematic
changes in performances. Nevertheless, people often attempt to invent a causal process to
explain away the performance changes,3 failing to recognize regression to the mean (Harrison
& March, 1984).

Our misperceptions of randomness suggest another reason for mistaking luck for skill.
Randomness in structured environments tends to produce systematic patterns. For example,
chance events do not even out when they are added rather than averaged. Good or bad luck
can persist for longer than might be expected (Feller, 1957), but research on the gamblers
fallacy suggests that people tend to believe that luck will even out (Ayton & Fischer, 2004).
More importantly, people believe that extreme performances cannot be explained by pure
randomness or luck (Denrell, 2004). This suggests that non-luck explanations are often
exaggerated, making people fooled by randomness (Taleb, 2001).

Overall, a common feature of the attribution and cognitive biases above is that people tend to
underestimate the role of luck in outcomes and these biases may lead to inaccurate
interpretations of the world. Although the consequences are not always undesirable, as they
can have motivational functions in evolution, the misperceptions of luck about extreme
performances can lead to disasters in modern societies (Liu & de Rond, 2016). For example,
managers at NASA mistook good luck in the prior near misses for superior skill and
continued the missions until the Columbia Space Shuttle Disaster (Dillon & Tinsley, 2008;
Tinsley, Dillon, & Cronin, 2012). The implication is that misperceptions of luck can lead to
predictable mistakes that have dire consequences.

2.3 Social Reasons for the Misperceptions of Luck

Interactive social dynamics can also lead people to develop systematic misperceptions of
luck. Cognitive biases can be overcome if people learn from experiences (including their
mistakes) over time. Nevertheless, interactive social dynamics often direct us in a way that
prevent us properly learning from experiences, sometimes through influencing our sampling
strategy (Denrell, 2005; Denrell & Le Mens, 2007) and sometimes through social herding
processes (Lorenz, Rauhut, Schweitzer, & Helbing, 2011; Salganik et al., 2006).

3
One technique to communicate regression to the mean (Mauboussin, 2012) is first showing how extreme past
performances (e.g., the heights of fathers or firm performances at period t) are likely to be followed by less
extreme future performances (the heights of sons or firm performances at period t+1). The audiences are likely
to provide alternative, causal explanations for these systematic changes in performances such as taller males are
likely to marry shorter females or executives at top performing firms become lazy or overconfident. Now show
the audiences a time-reversed version of regression to the mean based on the same data, i.e., how extreme
future performances (e.g., the heights of sons or firm performances at period t+1) are likely to be associated
with less extreme past performances (the heights of fathers or firm performances at period t). Since events in
the future cannot cause events in the past, the audience are more likely to understand that (a) regression the mean
is a statistical phenomenon that does not require causal explanations; and (b) for the alternative, causal
explanations provided for the first set of result to hold their effect size has to be at least larger than the base-line
result suggested by regression to the mean (Denrell et al., 2015; Liu & de Rond, 2016; Schwab, Abrahamson,
Starbuck, & Fidler, 2011; Starbuck, 1994).

6
For example, organizations often develop certain criteria or routines when hiring new talent.
Candidates whose performances are above the criteria threshold are hired and those below are
rejected. When candidates performances are not a perfectly reliable indicator of their skill
and/or the recruiters judgment are not perfectly reliable (Thorngate, Dawes, & Foddy, 2008),
the above practice will inevitably lead to two possible errors: (a) a false negative error:
competent candidates falsely rejected due to bad luck; and (b) a false positive error:
incompetent candidates falsely hired due to good luck. The problem is that we often fail to
correct for these two errors and can become overconfident in the evaluations criteria or
routine. On one hand, false negative errors are rarely revisited in organizations they are
invisible errors because additional samples that could correct for the original errors are
censored (Feiler, Tong, & Larrick, 2013). On the other hand, false positive errors are often
amended because the lucky hires may receive training and resources after joining the
organization, allowing them to appear competent enough due to a self-fulfilling prophecy
process (Merton, 1948). Moreover, evaluators may be motivated to cover these errors even
when they are aware of these possible mistakes, censoring the samples that could help to
correct for the mistakes. The implication is that organizations can systematically hire the
lucky ones and dismiss the unlucky ones, thus allowing these suboptimal practices to emerge
and persist (Gilovich, 2008).

Another important source of mistaking luck for skill is social herding (Asch, 1951; Banerjee,
1992). Skill (or quality for objects such as culture products) is difficult to observe and
evaluate directly so people often rely on others choices to infer the level of skill or quality.
This interdependence often leads to an augmentation of judgment errors. An experiment with
music downloads demonstrates how such augmentation from interactive social dynamics can
lead to enduring consequences from earlier instances of luck (Salganik et al., 2006). They first
provide a more reliable indicator of quality, as measured by the song downloads in an
independent condition where participants judge and decide which songs to download
without the information of others choices. In the rest of the social influence conditions,
information about others downloads were made salient to examine how social influences
impact the outcomes. A mediocre song (ranked 31st among the 48 songs in the independent
condition) manages to gain the highest market share in one of the social influence conditions
because it happened to be downloaded by some participants in an early stage of the
experiment. This small advantage from being popular in the early stages then attracted many
subsequent participants to download this song to such an extent that this song became the
most popular song in this condition whereas it only attracted moderate downloads in all other
social influence conditions. Their results suggest that stronger social herding will make the
performance distribution more unequal and unpredictable (Salganik et al., 2006): a few
winners will dominate the market but these winners do not necessarily win in other conditions
with slight difference in the initial performances. The implication is that the winners
exceptional performances do not necessarily mean that they have top quality they are
merely likely to be the ones that enjoyed early luck and crowd out potentially superior ones.

2.4 Summary
I reviewed some of the most damaging cognitive and social processes that can systematically
lead people to mistake luck for skill. One shared implication is that these processes can distort
quality ranking (Lynn, Podolny, & Tao, 2009). Performance is usually an imperfect indicator
of actors or objects inherent quality. The cognitive and social processes reviewed above can
further augment the initial distortion to such an extent that the actors with the highest quality

7
may not end up with the top rank whereas the actors with the lowest quality may not end up
with the lowest rank.

Still, such a distortion effect may not be problematic if the overall association between
performance and skill is monotonic. That is, if higher performances indicate higher levels of
the expected skill, the distortion effect may indicate that a few lucky (unlucky) outliers do not
deserve the exceptional credit (blame) they receive. But overall, judging the higher
performers to be superior, and thus rewarding and imitating the most successful, may be a
good enough heuristic to save time and energy (Richerson & Boyd, 2005).

In the next section, I will challenge this heuristic by referring to recent studies that
demonstrate a non-monotonic association between performance and the expected level of
skill, particularly for extreme performances. In these cases, there is a more fundamental
decoupling between performance and skill and the biases reviewed above will lead not only to
systematic but also consequential errors.

3. WHEN HIGHER PERFORMANCES DO NOT INDICATE SUPERIOR SKILL

Many prior studies emphasized the impact of luck in performances (Arthur, 1989; Barabasi &
Albert, 1999; Carroll & Harrison, 1994; Lynn et al., 2009). However, none of them considers
the possibility of a non-monotonic association between performance and the expected level of
skill (Denrell & Liu, 2012). Recent studies formally and empirically challenge the idea that
higher performers are better and also demonstrate that higher performers can be systemically
worse under some circumstances, suggesting a non-monotonic association between
performance and expected skill. I categorize these studies into three different versions.

A weak version. The first version is a weak account of the role of luck played in extreme
performances, which is consistent with most prior studies on the impact of luck. Winning or
having higher performances do indicate better skill but also a larger portion of luck. For
example, Frank (2016) formally demonstrates that top performers are increasingly likely to be
the luckier ones when there are more competitors and that the weight on skill is smaller. This
suggests that winners in these conditions should deserve proportionally less reward and
attention because the skill between the top and those close to top performers are small. If we
plot the association between performance and skill (similar to Figure 1A), the association will
likely become flat above a certain point (or becoming flat below a certain point for the lowest
performances) the skills among the highest performers (or among the lowest performers) are
indistinguishable (Fitza, 2014, forthcoming; Henderson, Raynor, & Ahmed, 2012). The
implication is that higher performances do not always indicate higher skill sometimes it
indicates luck by making the performance differences uninformative about skill differences.

This is a weak version of the impact of luck because the highest level performers are still
better than the rest even when the skill differences are indistinguishable among these high
performers. The implications according to this weak account are more conservative. For
example, one should still reward and imitate the top performer, but not in a way that is so
different from their underperforming counterparts even when their differences in
performances differ by orders of magnitude. This implication is qualitatively different from
the semi-strong and the strong versions below that suggest we should reward top performers
and punish the worst performers less.

A semi-strong version The second, semi-strong version of the impact of luck focuses on a
strategy paradox (Raynor, 2007). Raynor (2007) argues that exceptional successes and

8
extreme failures are two sides of the same coin. The reason is that realizing exceptional
success is usually built on two actions: (a) predicting an unlikely scenario that more sensible
strategists are likely to avoid; and (b) committing all resources to fully exploit this scenario if
it comes true. If this scenario does occur, all committed resources will enable an exceptional
success because of first mover advantage and, more importantly, less intense competition.
However, in most cases this unlikely scenario would not occur, suggesting that extreme
failure will follow because all the committed resources will likely become valueless. This also
suggests that sensible strategies such as real option strategies (Raynor, 2007) are likely to lead
to moderate performances at best (e.g., second best performers in Figure 1A who have the
highest level of expected skill) because they have to allocate (and thus dilute) resources for
developing strategies for different scenarios. Top performers are likely to indicate excessive
risk taking, unreasonable commitment, and exceptional luck. This is also supported by an
empirical study on Wall Street analysts (Denrell & Fang, 2010). Exceptional forecasters who
successfully predict the next black swan (Taleb, 2007) tend to be those who bet on unlikely
scenarios that forecasters with better track records tend to dismiss.

Research on firm risk-taking also supports this semi-strong version. For example, March
(1991) proposes a model where all firms performances follow a normal distribution with
mean equals zero and variance equals one except one reference firm whose performance also
follows a normal distribution but the mean and variance can change. He examines the means
and the variances required for this reference firm to outcompete all other N firms. March
(1991) finds that having higher means (e.g., having superior firm traits) or having higher
variances (e.g., taking more risk) both enhance the probability for this reference firm to win.
Importantly, when the number of firm (N) increases, having higher variance is increasingly
important for winning and as N goes to infinity the mean performance of the reference firm
becomes irrelevant. This suggests that when N is large, firms with low means can win if they
have extremely high variances. These firms can be considered as the lucky winners - they
would have lost dramatically in most counterfactual cases when their risk-taking led to the
opposite, extremely unfavorable outcomes.

This is a semi-strong version of the impact of luck. Top performers are likely the luckiest
ones because they take excessive risk that happens to produce favorable outcomes or they
commit to an unlikely scenario that happens to come true. In most counterfactual cases they
would have failed dramatically. But we tend to under-sample these failures (Denrell, 2003).
The fact that they succeed in the realized world indicates greater luck than prudent
strategizing. Sustainable performance by a firm requires more robust strategies, such as
scenario planning or real options, and more evenly allocating resources, but this helps to
guarantee that these more prudent managers and firms will not be able to become the top
winner who takes it all (Raynor, 2007). The implication is similar to the pattern in Figure 1A:
top performers can be systematically worse in terms of having a less robust strategy
formulation process and relying their performances more on luck.

A strong version. The strong version of the impact of luck formalizes a condition where
higher performances can indicate lower expected skill a non-monotonic pattern in the
association between performance and expected skill, as illustrated in Figure 1A (Denrell et al.,
2015; Denrell & Liu, 2012). The formal condition is: when the heterogeneity in noise is
greater than the heterogeneity in skill, extremely high (or low) performances are more likely
to indicate extreme noise than extremely high (or low) skill. The question becomes, when is
the heterogeneity in skill smaller than that of the noise?

9
Several studies have proposed mechanisms that imply the heterogeneity of skill can be
systematically less than that of noise. In particular, selection mechanisms can compress the
skill distribution, making the impact of noise (or luck) increase over time (March & March,
1977). The better candidates are selected for the next level and are likely to have superior
skills than the non-selected. But there is a side effect to this: the ones being selected are more
similar compared to all possible candidates, implying that the heterogeneity in skill (or traits
in general) of the selected would become smaller (March & March, 1977; Thorngate, 1988).
Unless the distribution of noise decreases at a faster rate than the decrease in the skill
distribution, the implication is that the performances differences of the selected will depend
increasingly on noise or luck. On the other hand, many studies suggest that the distribution of
noise is likely to be very wide. As suggested in the semi-strong version, some managers can
take excessive risk and happen to win big. Moreover, small initial difference can be
augmented by a rich-get-richer dynamics, or the Matthew Effect (Merton, 1968). The
decoupling effect can be so strong that the early winners continue to gain more resources and
training to boost their initial fortune, to such an extent that they become the top performers.
Taking these studies together, the implication is that the heterogeneity in skill can be
systematically smaller than that of the noise, particularly for the highest level of executives
who have survived multiple rounds of selections (March & March, 1977). Exceptional
performances not only fail to indicate higher skill, as the weak and semi-strong versions
suggest, but actually a lower level of skill because skills of this group are indistinguishable,
meaning one should attribute their extreme performances more to the situations than the
person, implying a non-monotonic pattern as illustrated in Figure 1A. This non-monotonic
pattern also suggests that the heuristic of higher performances indicating higher skill is
systematically wrong.

One possible criticism is that many exceptional performers gained their exceptional skill
through hard work and exceptional motivation so they do not deserve to receive lower reward
and praise. Some even suggested that there is a magic number for greatness, i.e., a 10,000
hour rule (Ericsson, Prietula, & Cokely, 2007; Gladwell, 2008): many exceptional performers
acquired their exceptional skill through persistent practices. However, detailed analyses of the
case studies of experts often suggest that some situational factors beyond the control of these
exceptional performers also played an important role. For example, three national champions
in table tennis came from the same street in a small suburb of one town in England. It turns
out that there was a world class coach of table tennis happened to retire in this particular
suburb (Syed, 2010). Many kids on that street were attracted to this sport because of this
coach and three of them realized the 10,000 hour rule and performed exceptionally well,
including winning national championship. Their talent and efforts were of course essential for
realizing their exceptional performances. But without their early luck (i.e., having a world
class coach and supportive families), just by practicing 10,000 hours without proper feedback
would not normally lead to a randomly picked kid becoming a national champion in table
tennis. We could also imagine a child with superior talent in table tennis suffering from early
bad luck (e.g., not having a capable coach or being in a country where being an athlete was
not considered to be a promising career) and never having a chance to realize her potential.
The implication is that the more exceptional a performance is, the more likely the situational
factors play a crucial role. So, one should attribute the more extreme performances to the
situations than to the persons, consistent with the non-monotonic pattern.

Now consider failures. The idea that major disasters can be disproportionally influenced by
external events is consistent with organizational research on the causes of disaster.
Organizational researchers have argued that blaming specific factors for failure is misplaced

10
when the fundamental causes of failures are tightly coupled systems which make
organizations sensitive to external shocks (Hollnagel, 2004; Leveson, 2011; Perrow, 1984;
Rudolph & Repenning, 2002). In such systems, disaster seldom occurs as a result of
individual mistakes. Mistakes can often be corrected and will not, by themselves, lead to
catastrophic failure. Rather, disaster requires bad luck. In particular, disaster requires that a
cascade of failure is set in motion. If a cascade is set in motion, failures will occur at an
increasing rate and operators will not have time (Rudolph & Repenning, 2002) or sufficient
insight into the underlying complex system (Perrow, 1984) to correct a system-wide failure
before it triggers other failures. If the triggers of a cascade of failures occurs is largely beyond
the control of individuals perhaps because not even the most skilled operator can understand
a complex system - it follows that system failure is not diagnostic of individual traits. The
implication is that we should attribute more extreme failures less to the persons involved and
more to the situations, such as tightly coupled, interdependent systems. Nevertheless, low (but
not extreme) performance can still be diagnostic of individual traits they are likely the less
skilled who were lucky enough to have near-misses but not disasters. To avoid recurrent
normal accidents (Perrow, 1984), these second worst performers should attract more
scrutiny and training.

The above discussion suggests that the top performers may be systematically worse than the
second best performers, whereas the worst performing actors may be the unluckiest and the
second worst performers (who may only narrowly avoid a disaster) should be the least skilled
or least mindful ones (Weick & Sutcliffe, 2006). The question is whether people can learn
these non-monotonic patterns (i.e., the best is not the best and the worst is not the worst)?
Recent experiments examined this question and suggested that, even with sufficient input
about how extreme performances cannot occur without luck, most participants still tend to
rate the top performers as the most skilled and the worst performers the least skilled, again
mistaking luck for skill (Denrell & Liu, 2012; Liu & de Rond, 2016). This is also consistent
with the research on the romance of leadership (Meindl et al., 1985). People tend to attribute
extreme performances more to leaders than circumstances, contrary to the normative account
discussed above. The implication is that peoples tendency to mistake luck for skill is the
strongest and the most consequential when evaluating extreme performances.

Prior studies on the misperceptions of luck have suggested ways to de-bias the evaluators
(Bazerman & Moore, 2009; Kahneman, 2011; Kahneman et al., 2011). The focus here is
different if these biases are predictable, informed managers can take advantage of others
mistakes. More importantly, the predictable biases about luck suggest an opportunity to
arbitrage because those who suffer from the misperceptions of luck will systematically
misevaluate the top and the worst performing individuals and organizations. I outline these
possibilities in the next section.

4. ARBITRAGING OTHERS PREDICTABLE MISPERCEPTIONS OF LUCK

Arbitrage is about taking advantage of price differences. Whenever price differences for a
transaction target exist (e.g., products, bonds, currencies), an arbitrager can buy it in one
market (where its price is lower) and sell it in another (where its price is higher), profiting
from a temporary difference. In a strategic context, arbitrage opportunity is primarily about
acquiring undervalued resources in strategic factor markets and/or selling the outputs in
product markets (preferably to buyers who overestimate their values). In short, arbitrage is
about buy low, sell high so heterogeneity in the evaluations of resources are necessary for
arbitrage opportunities to exist (Barney, 1986; Denrell et al., 2003).

11
The ways people mistake luck for skill as discussed in previous sections suggest that people
are likely to systematically misevaluate the values of high and low performers, which
provides opportunities for arbitrage. In particular, the gap between the expectations and the
actual value is the greatest for extreme performers, as illustrated by a non-monotonic
association between performance and skill. For example, top performers can have lower
expected skill (and thus lower future performance) than the second best performers, while
nave evaluators may believe that they will continue to perform at the top (and thus be willing
to pay more than they should for acquiring this top performer). More importantly, arbitrage
opportunities are mostly profitable when many people are evaluating in a similar but
predictably biased way. I discuss how to arbitrage the misperceptions about good luck (for
successes) and bad luck (for failures) below.

4.1 When Some Successes Are Too Good to Be True

Some successes are exceptional but what makes these successes exceptional is usually down
to exceptional circumstances rather than exceptional dispositional factors. So, one should
attribute these successes more to the situations that enable them to be exceptional rather than
to the persons that happen to be involved. I am not arguing that these exceptional performers
are entirely unskilled. They must be able to recognize and grasp the opportunity and then
skilled enough to realize the potential. The problem is that people tend to over attribute the
successes to the dispositional factors of the successful, thus giving them too much credit and
reward. More importantly, nave evaluators are likely to overestimate the values of the
successful and the likelihood that they could repeat their exceptional successes in other
contexts. In these cases, informed managers can take advantage of others misevaluations of
the successful and below I outline four possible ways to arbitrage others systematic
misperceptions of good luck in exceptional successes.

Selling your star employees or business units. Exceptionally performing employees or

business units are usually considered valuable assets. Sometimes their exceptional
performances can continue if the context remains unchanged, e.g., a rich-get-richer dynamics
is still present to boost the advantage. However, research has suggested that one should not
expect an exceptional performance to continue for at least three reasons: (a) the actors who
achieved the exceptional success will likely become overconfident due to self-serving
attribution and hindsight biases. These biases make them more likely to make self-defeating
errors (Camerer & Lovallo, 1999; Moore, Oesch, & Zietsma, 2007); (b) superior profit will
likely increase competition (Porter, 1979); and (c) an exceptional performance is generally
expected to regress downward to the mean (Greve, 2003). Since nave evaluators are unlikely
to recognize the above three processes they are more likely to romanticize the successful
individuals, to neglect possible competitions and to ignore regression to the mean. Thus,
nave evaluators are likely to predict that an exceptional performer is going to continue to
perform exceptionally. This implies an overestimation of the value of a star employee, an
organizational routine or business unit. An informed manager can do the opposite by selling
these stars to the less informed counterpart to benefit from the misevaluation.

One well known example is how Billy Beane, the manager of the Oakland As in Major
League Baseball, often sold his top performing players to other teams (Lewis, 2003). These
players are likely skilled, but their exceptional performance was reliant on a range of
circumstances which were not being factored into the evaluation by other teams. The Oakland
As were able to benefit from this selling high strategy by trading their stars to other teams
who had an unwarranted high expectation of these stars future performance. Moreover,
Figure 1B is based on team performances but suggest a similar strategy to the Oakland As

12
may work one could exploit this systematic pattern by betting against the teams with long
winning streaks in the first half of a season. The odd is likely favorable as most bidders are
likely to predict that these teams will continue their exceptional performance in the second
half of the same season instead of regressing downward to the mean.4

Selling the false ideas that imitating the top could become great. There are systematic ways
for a poor performer to improve and become a better performer. But there are no rules for
becoming the richest: to move from good to great depends more on being at the right place
and the right time, such as grasping a rare opportunity and benefiting from the Matthew
Effect to boost the initial fortune. But people often imitate the most successful with the
assumption that they are likely to be the most skilled (Strang & Macy, 2001). Imitating the
most successful could have been useful for our ancestors when the highest performances were
not so extreme that they could have provided a reliable indicator of high skill (Richerson &
Boyd, 2005). Modern technologies and globalization, however, could enable exceptional
performances that are so extreme that our heuristic of imitating the most successful fails
systematically.

Instead of de-biasing and educating people to be cautious when evaluating top performers,
one can gain advantage by selling the false idea that imitating and learning from the top
performers is beneficial. For example, consultants can promote the practices of the top
performing firms as best practices to their clients (Strang & Macy, 2001); head hunters can
promote star executives to save underperforming firms (Groysberg, 2010); authors can write
books on how to learn from exceptional performers to move from good to great (e.g., Collins,
2001; Peters & Waterman, 1982). But few of the best practices, star executives or principles
of success will enable the clients/readers to achieve exceptional performances. The gap
between clients unwarranted high expectations and the actual performances of the best
practices, star executives and principles for becoming great may explain recurrent
management fashions, the short tenure of star executives and the bankruptcies of the case
companies in the management bestsellers that teach people how to move from good to great
(Liu & de Rond, 2016). However, since peoples intuition that top performers are the best
remains so heavily favored, there is likely a market for the more informed, such as
consultants, headhunters or writers to continue exploiting peoples success bias. In fact,
alternative approaches such as benchmarking the second best are unlikely to attract many
clients.

Note that I am not saying the best practices, star executives or principles of becoming great
introduced in management bestsellers are all without merit or useless. After all, to become
great, one needs to become good first, consistent with the conventional wisdom that chance
favors the prepared mind. But luck still plays a more important role in moving from good to
great. The dispositional factors of the outliers may have little to do with the realizations of
their exceptional performances. Imitating their dispositional factors can lead to unwarranted
high expectation and predictable disappointments (Harrison & March, 1984).

4
One suspicion is that the author should not share this insight if it really works. The author confirms that this
worked when he tried this betting strategy in 2011. But the profit made was not enough to justify quitting
academia unless a strong capital support was acquired because this strategy, like many arbitrage strategies,
guarantees to generate a profit only in the long run but has short-term earning variability. Without sufficient
capital support bankruptcies are likely before realizing the expected profit. In a way the author is constrained by
own theory of limits to arbitrage (i.e., capital and family responsibilities constraints) and decided to share this
insight with those who may be more fortunate than him.

13
Imitating and rewarding the second best to gain sustainable advantage. If one should not
imitate the top performers then who should one learn from? I suggest three possible
heuristics: (a) one should ignore the global best performer and imitate a better performing
network neighbour, even if their performance is mediocre in a global sense; (b) the benefit of
imitating better neighbours increases with population size; (c) one should aspire to the top
performer but should not imitate their practices when dissatisfied. When the population size
increases, extreme performances are more likely to occur (Denrell & Liu, 2012; March,
1991). In this case, more is often less: the more exceptional the performance is, the less
one could learn from these outliers, because they are more likely to achieve their exceptional
performance due to irreplicable factors such as luck, risk taking, or untransferable rich-get-
richer dynamics (Denrell et al., 2015). In such cases, imitating a better performing network
neighbour is desirable because the contextual factors are more likely to be controlled for and
the information about the practices of better performing neighbours is more reliable. Aspiring
only to the neighbours is not optimal, however, because one could prematurely stop searching
for better practices (Cyert & March, 1963; March, 1991). Thus, aspiring to the global best but
imitating a better performing neighbour is a simple heuristic that is most effective in
environments where actors in local networks are similar and performances can be highly
variable due to luck. More importantly, their less extreme performances may make these
strategic resources cheaper to acquire on the market, further enhancing the gap to profit from
these resources.

Randomly selecting the next top-level executive. This may sound a proposal for anarchy but
it is in fact an old, well-known concept known as demarchy. In ancient Greece and the
Venetian Republic, cities selected their leaders through a lottery system (Stone, 2009;
Zeitoun, Osterloh, & Frey, 2014). I am not suggesting firms should always randomly select
their top executives. Random selection makes sense when performance depends more on luck
than on skill. This is likely to be the case in many large businesses. Firstly, although top
executives are very skilled, the difference in skill among them tends to be minimal many of
them are educated at the same elite universities, share similar networks and mental models
(Belliveau, O'Reilly, & Wade, 1996; Geletkanycz & Hambrick, 1997; Westphal, 1999;
Westphal, Seidel, & Stewart, 2001). A randomly selected executive may well perform
similarly to a carefully selected one because their skills are likely indistinguishable at such a
high level. In fact, many executives may be selected not because of their superior skill but
possible biases of the evaluators, such as fitting into a stereotype of competent executives
(Thorngate et al., 2008). Random selection is better because blind luck helps to trump human
biases. Secondly, there are large differences in performances between businesses. This means
that the differences in performance are more likely to result from factors unrelated to the
executives' skill, i.e., being at the right time and right place. Random selection is also
beneficial because it is associated with higher levels of perceived fairness, resistance to
corruption, and to more robust outcomes than other approaches. Firms may create conflicts of
interest when selecting high-level executives through traditional evaluations of performance.
Randomly selecting a leader can avoid such conflict - people do not need to work against each
other in order to stand out, nor undermine the performance of their competitors, which can be
detrimental for the firm (Stone, 2011).

4.2 When Some Failures Are beyond Incompetence

Case studies of disaster have demonstrated that failures in organizations often result from a
cascade of errors triggered by minor mistakes occurring in interdependent systems (Dorner,
1996; Perrow, 1984; Rudolph & Repenning, 2002). Organizational researchers have argued
that blaming the individuals in charge is usually misplaced in such cases because the

14
fundamental causes of failures are tightly coupled systems which make organizations
sensitive to even trivial mistakes. Still, research shows that the people in charge are usually
blamed for failure (Coughlan & Schmidt, 1985; Denis & Kruse, 2000; Groysberg, Nanda, &
Nohria, 2004). Below I outline three specific heuristics to exploit the ways people mistake
bad luck for poor skill in these cases of misattributed blame.

Hire the fired. This discrepancy between the normative and descriptive accounts of responses
to failures implies that inefficiencies in the labor market exist: competent actors can be
dismissed for bad luck (Feiler & Taylor, 2013). The possibility of such inefficiency, in turn,
suggests an arbitrage opportunity: smart managers could hire, at a low wage, the individuals
associated with failure. The fact that they failed perhaps indicates that they could have been
more careful and cautious. But the extreme failures they experienced are unlikely to have
happened by their mindlessness alone. However, these unlucky managers are likely to suffer
from stigmatization in the markets (Wiesenfeld, Wurthmann, & Hambrick, 2008). This
suggests that informed managers could hire these underdogs with a lower wage than their
intrinsic value.

Give the failed another chance. Managers who are associated with failures should be
retained instead of fired for two reasons. First of all, they are likely the unluckiest rather than
least skilled. Consider a skilled employee who happens to perform poorly in the first task
assigned to him. This failure may leave the boss a negative impression against this employee
and reduce future interactions with him, making the negative impression (and early bad luck)
to persist (Denrell, 2005; Liu, Eubanks, & Chater, 2015). Unless there are procedures to
create interactions independent to attitudes (Denrell & Le Mens, 2007), failures may be an
unreliable indicator of low skill because competent actors can suffer from the enduring impact
of bad luck: they may never have a chance to prove the audience wrong. Also, the failed are
likely to have had the opportunity to learn valuable lessons leading to the possibility of them
providing better outcomes in future. More importantly, if the company fires this manager and
hires a new one, the firm is likely to engage in superstitious learning as outlined in the
introduction. Encouraging the failed managers to exercise the lesson learned and to make
systems less tightly coupled can provide a more constructive way to improve performances
and to avoid the next normal accident (Perrow, 1984). Secondly, failure can indicate your
employee is trying something new that may be worthy of encouragement. People who fail can
suffer from others wanting to reduce their interactions with those associated with the failure.
But since good decisions can sometimes lead to failures and bad ones to successes, this is
unlikely to be good practice. Companies will likely end up with the mediocre if they only
select the successful because candidates with impeccable records are likely to indicate that
they never tried anything new but instead have largely remained in their comfort zone.
Overcoming the success bias and giving the failed another chance may provide the
undervalued hidden gems for rejuvenating an organization.

Pay attention to the least popular. Popularity often has a self-reinforcing feature, particular
after salient events. If an emerging industry happens to suffer a major bankruptcy, this failure
can be so salient that investors and entrepreneurs have an excessively negative impression,
leading them to avoid further investment in this industry which might otherwise be profitable.
Recent research suggests that this is exactly the timing to pay attention to those who enter this
industry (Pontikes & Barnett, Forthcoming). Those who entered might be the ones who are
viable enough to overcome the negative evaluation and the competition will be less intense
for them to realize superior performance. In contrast, most entrepreneurs and investors are
attracted to industries where salient successes occur, but those who jump on the bandwagon

15
are actually less likely to succeed (Pontikes & Barnett, Forthcoming). The implication is that
an arbitrager should pay attention to the events people tend to overreact to and then the
arbitrage opportunity is likely to arise from a contrarian approach: avoiding the most popular
individual or firms and embracing the least popular ones.

4.3 Summary
Arbitrage opportunities in strategic factor markets depend on different, or even better, biased
perceptions or interpretations of the same resources. This paper focuses on the misevaluations
of extreme performances where managers who follow their intuitions are likely to overvalue
the top performers and to undervalue the worst performers. These predictable mistakes give
informed managers an opportunity to arbitrage by selling high and buying low. I outlined
some heuristics about how these misperceptions of luck about extreme performances could be
exploited. However, implementing these suggestions is likely to be risky because it requires
the arbitrager to act against the crowd and their biased intuitions and evaluations. If a
strategist is sensitive to other stakeholders evaluations (such as the stock market or unions),
she might adopt a more conventional strategy and give up the arbitrage opportunity identified.
This is the lemon problem in strategic factor markets: valuable but unique strategies may be
discounted and thus dismissed whereas conventional but less valuable strategies may become
dominant an adverse selection process (Benner & Zenger, 2016). This poses a challenge for
managers to exploit the arbitrage opportunities outlined above and I will discuss these limits
to arbitrage and ways to overcome them in the next section.

5. OVERCOMING THE LIMITS TO ARBITRAGING OTHERS MISPERCEPTIONS

OF LUCK
It is not easy to arbitrage strategic opportunity resulted from others misperceptions of luck
because the arbitrager usually has to act against the crowd and their biases. If one is sensitive
to others (mis)perceptions, implementing an arbitrage strategy may be counterproductive. I
first discuss the hurdles of arbitraging others misperceptions of luck before discussing
possible solutions to overcome these limits to arbitrage.

The first problem is how to identify the circumstance where a non-monotonic pattern occurs.
Luck is difficult to measure, particularly when the focus is on extreme performances that are
often samples of one or fewer (March, Sproull, & Tamuz, 1991). Some recent research
applies sports data that may be able to overcome this hurdle (Feiler & Taylor, 2013; Kilduff,
Crossland, Tsai, & Bowers, 2016). For example, it has been demonstrated that the role of bad
luck is significant in contexts such as Formula One racing: the worst performing drivers may
be the unluckiest rather than the least skilled, but they are still more likely to be fired due to
bad luck (Denrell, Liu, & Maslach). These patterns can be demonstrated because sports data
is detailed and less ambiguous (e.g., Bothner, Kang, & Stuart, 2007; Castellucci & Ertug,
2010), but the role of luck may be more difficult to measure in a business context. This does
not mean that the impact of luck is any weaker in business context, instead this predicts that a
non-monotonic pattern is more likely to occur due to correlated social influences. However,
empirically identifying when the top performers are not the best is essential for implementing
an effective arbitrage strategy and failure to provide this identification may be the first (and
fatal) constraint for implementing an arbitrage strategy.

I suggest three possible approaches to address this problem. The non-monotonic patterns hold
when the non-skill factors, i.e., luck can substantially impact performances, implying that low
or medium skilled actors could obtain extremely high performances. This is possible in some
tasks where outcomes are influenced by numerous unpredictable factors beyond the control of

16
actors (Denrell et al., 2015). Consider trading: it is possible that an individual with moderate
skill can obtain a really high return from trading during one year. One might argue that
repeated successes then exclude the explanation of luck. Nevertheless, if there are a
sufficiently large number of traders, it is likely to observe a few traders having extremely high
return several years in a row even when they do not have higher skill (Barney, 1997).

This also suggests that there are clearly situations where the impact of luck on performance is
trivial. Consider the 100-meter dash. An unskilled individual who runs 100 meters on 25
seconds, on average, will not, by luck, be able to run below 10 seconds. A heuristic as the first
approach for testing whether a task is more about skill or about luck is the following: consider
whether an actor can intentionally fail the task (Mauboussin, 2012). If she can (for tasks such
as chess or running), this is likely to indicate that she can exercise her skill to control the
outcome, implying that outcomes are more about skill; if she cannot (for tasks such as roulette
wheel or investment), this is likely to indicate her lack of control, implying that outcomes are
more about luck.

The second approach is to measure how similar the actors dispositional factors are. The more
similar they are (e.g., in terms of shared ties or education background), the more likely that
they will be exposed to similar information and hence similar judgment and decisions
(Denrell & Le Mens, 2007; McPherson, Smith-Lovin, & Cook, 2001; Page, 2008). If this is
the case, it follows that their performance differences should be attributed more to the
situational factors. Take CEO performances for example. Recent studies empirically showed
that a large portion of firms performance differences that are often attributed to CEOs may in
fact be due to luck (Fitza, 2014, forthcoming). Importantly, this does not mean that CEOs are
low skilled. Rather, top executives are likely all high skilled but the differences in their skills
are indistinguishable (March & March, 1977). For example, many CEOs graduated from the
same schools, shared similar social networks and likely think and act similarly (Khurana,
2002). Since the variance in CEO skill is low, large differences in firm performances are
unlikely to be explained by CEO skills alone but by factors such as some firms being at the
right time and right place.

The third approach is to measure the extent of regression to the mean between performances
at consecutive periods like as illustrated in Figure 1B and 1C. If extreme performances are
likely followed by less extreme performances and that the ranking at the highest (and the
lowest) ends are systematically reversed, this implies that performance-skill link in this
context is likely non-monotonic as well due to the enduring impact of luck. The implication is
that one should discount the expected skill more for the more extreme performances, a
statistic adjustment technique also known as Steins estimation rule (Efron & Morris, 1973;
Efron & Morris, 1977).

The second problem is whether a non-monotonic pattern, if existed, promises a profitable

arbitrage opportunity. Competition dynamics is the most important consideration for this
problem. For example, one should ask how many competitors are informed enough to
implement a similar arbitrage strategy. The success of an arbitrage opportunity is likely to be
in line with the uniqueness of the insight and a few smart arbitragers are usually sufficient to
eliminate financial market inefficiencies (Shleifer & Vishny, 1997). This may not be the case
in a strategic factor market but the scope for profitability is still associated with how many
strategists share the same insight. For example, research on relative age effect (Musch &
Grondin, 2001; Pierson, Addona, & Yates, 2014) may imply an opportunity. Older kids in an
age group enjoy advantages in physical and IQ development so parents with this knowledge

17
can postpone their childs entry year to school to realize such an advantage. But this
knowledge quickly diffused thanks to the book Outlier by Malcolm Gladwell (2008). Many
parents exercised this knowledge and postponed the start of their childrens school entry to
such an extent that many regulators had to reduce the flexibility of school entry year
(Dustmann, Fitzenberger, & Machin, 2008), making the market more efficient again. Kids
whose parents who were less informed would be disadvantaged but parents who owned and
exercised this knowledge can at best gain competitive parity.

One approach to overcome this problem is to look for the more complicated arbitrage
opportunity rather than the low hanging fruit. The relative age effect and its practical
implication are easy to understand so the advantage is arbitraged away soon. One should look
for arbitrage opportunities that are ambiguous, difficult to measure and hard to understand so
competition will be less intense (Denrell et al., 2003; Gavetti, 2012).

Another approach is to sell this idea to others instead of exploiting them yourself. For
example, one can become a service provider to help parents to go through the process of
postponing their kids school entry. This was what one of the first persons who learned the
discovery of gold in California in 1848 did. Instead of joining some 300,000 competitors for
the Gold Rush, Samuel Brannan shared this news with a lot of people and quickly set up
stores to sell gold prospecting supplies and became the first millionaire thanks not to the gold
per se but to the crowds who believe in their luck in finding the gold.5 Another example is
how Michael Lewis wrote the bestseller Moneyball (2003) instead of exploiting the market
inefficiencies in baseball himself. This example also highlights another limit to arbitrage: can
you keep your mouth shut after successfully exploiting arbitrage opportunities and also
prevent others from noticing your success? The Oakland As were not cautious enough and
allowed Michael Lewis to write a book about their arbitrage strategies and their strategies
became widely imitated after the publication of the book, meaning the mispricing was
corrected for and the Oakland As advantage evaporated (Hakes & Sauer, 2006). Unless an
arbitrager could not profit from an opportunity without many people also recognizing it (e.g.,
Samuel Brannans example in the California Gold Rush), successful arbitragers (e.g., Oakland
As) should not share the insight with others - just say you were lucky.

The third, and most difficult constraint of arbitrage strategies is the lemon problem in
markets (Benner & Zenger, 2016). It is well-known that even if decision-makers have the
cognitive abilities to draw the correct inference, this does not mean that they should act
upon it. In some settings, being accurate is less important than acting upon a belief widely
shared by others (Asch, 1951; Keynes, 1936; Schelling, 1960). As Keynes emphasized,
someone with accurate insight into the fundamental value of a stock will not necessarily be
able to benefit from such an insight and insight into how stocks are evaluated by others may
be more valuable (Froot, Scharfstein, & Stein, 1992). More generally, whenever the payoff of
an action depends on the actions of others, acting upon a correct but infrequent inference may
not be wise (Litov et al., 2012). In such settings people often refrain from acting upon a
correct inference when they are uncertain whether others have drawn the same conclusion.
Even if all individuals involved are rational and able to draw the correct inference, a situation
resembling pluralistic ignorance (Kuran, 1997) may develop in which no one acts upon the
correct inference.

5
Unfortunately, Samuel Brannans exceptional performance did not last he died relatively poor and in relative
obscurity in 1889.

18
I argue that performance arbitragers may hesitate to act upon a correct inference about the
role of skill for similar reasons. Performance evaluation seldom occurs in a social vacuum;
rather evaluators often care about and thus have to anticipate the reactions of others.
Evaluators partly care about the evaluations of competitors because such evaluations impact
wages. A manager might have unique insight into how skilled an employee is but cannot
benefit from this insight if others overestimate her skill and are willing to pay a high wage.
Evaluators also care about the evaluations of stakeholders, such as customers, suppliers, and
board of directors, because such stakeholders evaluate their actions. If customers of a service
firm incorrectly believe that employees associated with failure are unskilled, it may be
difficult to ignore this belief. Finally, actors care about the evaluations of others simply
because they may feel the pressure to conform (Durand & Kremp, 2016; Phillips &
Zuckerman, 2001).

Because managers care about how stakeholders evaluate the people they employ, they may
hesitate to retain or hire people associated with poor performance not because they believe
that these managers have poor skills, but because they are uncertain about whether others
attribute the failures of these managers to low skill. Consider what would happen if a firm
hires an executive from a recently failed firm. The hiring firm might be convinced that the
failure should be attributed to bad luck rather than to low skill, but the firm cannot be sure
that others think the same. If there is some chance that customers or employees will blame the
manager for the failure, hiring the manager becomes increasingly risky (Khurana, 2002).
Resistance from customers or employees might undermine the effectiveness of the new
manager. Moreover, if the manager is hired and performance of the hiring firm deteriorates,
the failure will be attributed to a poor hiring choice. Thus, even if a decision-maker has
privileged information or more data than others, which enables them to more accurately
identify the skills of a manager, the decision-maker may not be able to benefit unless they can
convince others of the validity of their actions (Froot et al., 1992; Litov et al., 2012) and in a
setting without repeated interaction, this is likely to prove difficult (Zuckerman, 1999).

Hiring actors associated with failure or dismissing the stars is especially risky if the correct
inference about skill from observed performance violates common sense judgments. For
example, consider a system in which a non-monotonic patter occurs: actors with the lowest
observed performance do not have the lowest expected skill. Such a less-is-more effect
violates the expectation that higher performance indicates higher skill (Milgrom, 1981). More
generally, it violates the default belief that two variables are linearly related (Brehmer, 1980;
Dawes & Corrigan, 1974). Actors who draw the correct inferences about performance in a
system with a less-is-more effect may be uncertain whether others have drawn the same
inference. Acting conventionally in line with common-sense ideas, i.e., firing instead of
retaining managers associated with poor performance, is often preferred in such ambiguous
situations. Acting unconventionally, i.e., hiring managers associated with failure, remains
risky.

Who can ignore the pressure to act conventionally and retain atypical strategic factors? Only
actors who have a sufficiently secure social or economic position, or who are independent of
important stakeholders, can ignore the pressure to act conventionally (Durand & Kremp,
2015; Hollander, 1958; Phillips, Turco, & Zuckerman, 2013; Phillips & Zuckerman, 2001;
Piazza & Castellucci, 2014). Actors who are independent of stakeholders do not have to care
about or anticipate what others think. For example, independently held family firms can
ignore stock market analysts reactions when hiring a CEO. Actors who are well established
and have high status can afford to act unconventionally because they have built up an

19
idiosyncrasy credit (Hollander, 1958) that allows them to deviate from norms without anyone
questioning their competence or loyalty. High status actors do not just get the freedom to go
against convention but also to violate norms that are typical within a particular group and to
set the new standards of behavior for acting appropriately within the group. Even though
these actions may be atypical for now, others will have confidence that these atypical
behaviors from high status actors will result in higher expected performance in the future
(Podolny, 1993).

Actors with high status can also afford to act unconventionally because their actions are
interpreted differently from the actions of actors with low status and the high status actors
may have a wider range of choices available to them. Availability implies that the audience
infers that some compensating factor must have motivated a choice of an unusual and
seemingly inferior option. Consider what happens when a high status academic department
hires an academic with an undistinguished record. Outsiders infer that the academic had
unobserved talent, but if a department with low prestige hired the same academic, outsiders
are likely to conclude it was a hiring mistake. Finally, actors with high status have greater
security in their position to act unconventionally because those who are dependent on them
will be less likely to protest, and research shows that actors recognize their power are more
likely to violate social norms (Keltner, Gruenfeld, & Anderson, 2003). The implication is that
fortune may favor the insensitive (e.g., having high status or power) because there is a limit to
arbitrage for the rest due to coordination challenges, even when they are smart enough to
identify the opportunities.

More generally, people sometimes care more about ego (e.g., boosted status or following
norm to gain legitimacy) than money and this means that opportunities may be left
unexploited even when they are obvious to many. Opportunities thus favor those who dare to
act against the norm or possibly just are nave and unaware of the norms. One potential
approach to exploit arbitrage opportunities due to social constraints is to make oneself more
isolated. By reducing exposure to the norms and other stakeholders, a strategist is more likely
to judge an opportunity based on its merit and put less weight on social dynamics. But this is
a lesson difficult to swallow, particularly for those who already invested resources in building
reputations or status. It may not be surprising that dominant incumbents tend to be
overthrown by the unknown or the peripherals because the insiders are likely to be bounded
by status or norms and fail to see alternative way of doing things so being on the periphery
can be a blessing for exploiting arbitrage opportunities. This also implies that arbitrage
opportunities may be a bad news for the incumbents but good news for less resourceful
entrepreneurs who are forced to adopt an alternative approach to compete.

To summarize the discussions in this section, Figure 2 presents a flowchart for evaluating
whether an arbitrage opportunity resulted from misperceptions of luck exists and whether one
should be exploit it. The first question is about whether there exists a non-monotonic pattern
between performance and expected skill. I suggested three heuristics to examine this question:
luck is likely to have greater impacts in a particular context (and hence the association
between performance and expected skill are likely to be non-monotonic) when (a) an actor
can intentionally fail in this context; (b) actors in this context tend to share similar traits such
as backgrounds, networks or mental models; and (c) regression to the mean effect is strong in
this context. If the answer to Question 1 is negative, the association between performance and
expected skill is likely monotonic, implying that higher performances are likely a reliable
indicator of higher skill. Arbitrage opportunity resulted from the misperceptions about
extreme performances are not likely to exist.

20
Figure 2. A flowchart of three questions about whether one should pursue an arbitrage
opportunity resulted from misperceptions of luck about extreme performances

If the answer to the first question is positive, one should ask the second question about
whether the arbitrage opportunity resulted from misevaluations of extreme performances are
sufficiently profitable. Three heuristics for examining this question: this arbitrage opportunity
is likely to be profitable when (a) very few, if any, competitors are aware of this arbitrage
opportunity; (b) the arbitrage opportunity is not low hanging fruit; and (c) you can profit from
selling this opportunity to others (particularly when this opportunity is easy to understand). If
the answer to Question 2 is negative, an arbitrage opportunity may exist but it may not be
profitable enough for you to exploit. For example, profitability will quickly be competed
away once your success story diffuses (e.g., the case of Moneyball). In this case, arbitrage
opportunity may imply temporary competitive advantage at best and it may not be desirable
overall when taking the social cost into account, which leads to the third question.

If the answer to the second question is positive, one should ask the third question about
whether you are in a position to defend the possible negative responses when you exploit an
atypical, arbitrage opportunity such as dismissing the stars or recruiting the failed. You are
more likely to fully realize the expected profit of this arbitrage opportunity if you are less
sensitive to the reactions of your important stakeholders, such as by having high status or high
power. Or, under some circumstances you can be insensitive because you are the underdog or
the newcomers who have little to lose. Last, you can be insensitive to others reactions simply
because you do not care. If this is not the case, the social constraints imposed on you imply

21
that you are likely to let go this profitable opportunity identified. If the answer to the third
question is still positive, it is more likely that you can successfully overcome the hurdles of
identifying and exploiting this arbitrage opportunity. Fortune favors arbitragers like you as
being both the insightful and the insensitive.

6. CONCLUDING REMARKS: BEHAVIORAL STRATEGY AS OVERCOMING

LIMITS TO ARBITRAGE
I outlined an alternative source of strategic opportunity by arbitraging strategic factor market
inefficiencies resulted from peoples systematic misperceptions of luck when evaluating
extreme performances. I also discussed the caveats of identifying and exploiting such
arbitrage opportunities due to hurdles such as the lemon problem. Misperception of luck is
just one of the predictable mistakes identified in the literature. There are other biases such as
overconfidence, learning myopia, organizational inertia or the hot stove effect that can lead
managers make systematic mistakes. Future research can examine the scope conditions of
these biases and whether they can also become a source of strategic opportunity.

An analogy of the framework proposed in this paper is how Michael Porter turned Industrial
Economics on its head and created his Five Force framework that contributes to competitive
advantage (Porter, 1979; Porter, 1980). Few ideas in his Five Force framework were novel in
Industrial Economics, but the focus then was on how to avoid monopoly and make the market
more in line with perfectly competitive markets. Michael Porter focused instead on how a
firm can become a monopoly suggesting that all constraining factors against perfectly
competitive markets could turn into enabling factors for a firm to maximize its competitive
advantage.

Similarly, most of the biases discussed in this paper are well known in the literature. The
main contribution of this paper is how these biases can be turned on their head by suggesting
how informed managers can gain advantage by exploiting others predictable mistakes. The
limits to arbitrage also suggest that money may be left on the table and a strategic opportunity
may continue to be unexploited. Some people may not see an opportunity because of decision
biases or learning myopia. Some people may be able to see an opportunity but fail to act on it
due to social constraints. Fortune may favor the insensitive and the nave because others will
not be able to fully realize the strategic opportunity due to interactive cognitive and social
dynamics. This paper thus suggests a new branch of Behavioral Strategy for illuminating
ways to overcome the limits to arbitrage strategic factor market inefficiencies with social
savvy.

References
Arthur, W. B. 1989. Competing technologies, increasing returns, and lock-in by historical
events. The Economic Journal, 99(394): 116-131.
Asch, S. E. 1951. Effects of group pressure upon the modification and distortion of
judgments. In H. Guetzkow (Ed.), Groups, Leadership, and Men: 222-236.
Pittsburgh, PA: Carnegie Press.
Ayton, P., & Fischer, I. 2004. The hot hand fallacy and the gambler's fallacy: Two faces of
subjective randomness? Memory & Cognition, 32(8): 1369-1378.
Banerjee, A. V. 1992. A simple model of herd behavior. The Quarterly Journal of
Economics, 107(3): 797-817.
Barabasi, A. L., & Albert, R. 1999. Emergence of scaling in random networks. Science,
286(5439): 509-512.

22
Barney, J. B. 1986. Strategic factor markets: Expectations, luck, and business strategy.
Management Science, 32(10): 1231-1241.
Barney, J. B. 1997. On flipping coins and making technology decisions: Luck on an
explanation of technological foresight and oversight. In R. Garud, P. R. Nayyar, & Z.
B. Shapira (Eds.), Technological Innovation Oversights and Foresights: 13-19. New
York, NY: Cambridge University Press.
Baron, J., & Hershey, J. C. 1988. Outcome bias in decision evaluation. Journal of
Personality and Social Psychology, 54(4): 569-579.
Bazerman, M. H., & Moore, D. A. 2009. Judgment in Managerial Decision Making (7th
ed.). New York, NY: Wiley
Belliveau, M. A., O'Reilly, C. A., & Wade, J. B. 1996. Social capital at the top: Effects of
social similarity and status on CEO compensation. Academy of management Journal,
39(6): 1568-1593.
Benner, M. J., & Zenger, T. 2016. The lemons problem in markets for strategy. Strategy
Science, 1(2): 71-89.
Bothner, M. S., Kang, J., & Stuart, T. E. 2007. Competitive crowding and risk taking in a
tournament: Evidence from NASCAR racing. Administrative Science Quarterly,
52(2): 208-247.
Brehmer, B. 1980. In one word: Not from experience. Acta Psychologica, 45(1): 223-241.
Camerer, C., & Lovallo, D. A. N. 1999. Overconfidence and excess entry: An experimental
approach. American Economic Review, 89(1): 306-318.
Carroll, G. R., & Harrison, J. R. 1994. On the historical efficiency of competition between
organizational populations. American Journal of Sociology, 100(3): 720-749.
Castellucci, F., & Ertug, G. 2010. What's in it for them? Advantages of higher-status partners
in exchange relationships. Academy of Management Journal, 53(1): 149-166.
Collins, J. C. 2001. Good to Great: Why Some Companies Make the Leap--and Others
Don't. New York, NY: Harper.
Correll, S. J., Ridgeway, C. L., Zuckerman, E. W., & Jank, S. 2012. Its the conventional
thought that counts: How third-order inference produces status advantage. Working
Paper. Stanford University.
Coughlan, A. T., & Schmidt, R. 1985. Executive compensation, management turnover, and
firm performance. Journal of Accounting and Economics, 7(1-3): 43-66.
Cyert, R. M., & March, J. G. 1963. A Behavioral Theory of the Firm. Malden, MA:
Blackwell Publishers Inc.
Dawes, R. M., & Corrigan, B. 1974. Linear models in decision making. Psychological
Bulletin, 81(2): 95-106.
de Rond, M., & Thietart, R.-A. 2007. Choice, chance, and inevitability in strategy. Strategic
Management Journal, 28(5): 535-551.
Denis, D. J., & Kruse, T. A. 2000. Managerial discipline and corporate restructuring
following performance declines. Journal of Financial Economics, 55(3): 391-424.
Denrell, J. 2003. Vicarious learning, undersampling of failure, and the myths of management.
Organization Science, 14(3): 227-243.
Denrell, J. 2004. Random walks and sustained competitive advantage. Management Science,
50(7): 922-934.
Denrell, J. 2005. Why most people disapprove of me: Experience sampling in impression
formation. Psychological Review, 112(4): 951-978.
Denrell, J. 2008. Organizational risk taking: Adaptation versus variable risk preferences.
Industrial & Corporate Change, 17(3): 427-466.
Denrell, J., & Fang, C. 2010. Predicting the next big thing: Success as a signal of poor
judgment. Management Science, 56(10): 1653-1667.

23
Denrell, J., Fang, C., & Liu, C. 2015. Chance explanations in the management sciences.
Organization Science, 26(3): 923-940.
Denrell, J., Fang, C., & Winter, S. G. 2003. The economics of strategic opportunity. Strategic
Management Journal, 24(10): 977-990.
Denrell, J., Fang, C., & Zhao, Z. 2013. Inferring superior capabilities from sustained superior
performance: A Bayesian analysis. Strategic Management Journal, 34(2): 182-196.
Denrell, J., & Le Mens, G. 2007. Interdependent sampling and social influence.
Psychological Review, 114(2): 398-422.
Denrell, J., & Liu, C. 2012. Top performers are not the most impressive when extreme
performance indicates unreliability. Proceedings of the National Academy of
Sciences, 109(24): 9331-9336.
Denrell, J., Liu, C., & Maslach, D. 2016. Bad luck or bad management: How uncertainty
about others' inferences moderate responses to failures. Paper under R&R.
Dillon, R. L., & Tinsley, C. H. 2008. How near-misses influence decision making under risk:
A missed opportunity for learning. Management Science, 54(8): 1425-1440.
Dorner, D. 1996. The Logic of Failure: Recognizing and Avoiding Error in Complex
Situations. New York, NY: Metropolitan Books.
Durand, R., & Kremp, P.-A. 2015. Classical deviation: Organizational and individual status as
antecedents of conformity. Academy of Management Journal, 59(1): 65-89.
Durand, R., & Kremp, P.-A. 2016. Classical deviation: Organizational and individual status as
antecedents of conformity. Academy of Management Journal, 59(1): 65-89.
Dustmann, C., Fitzenberger, B., & Machin, S. 2008. The Economics of Education and
Training. London, UK: Springer Science & Business Media.
Efron, B., & Morris, C. 1973. Stein's estimation rule and its competitorsan empirical Bayes
approach. Journal of the American Statistical Association, 68(341): 117-130.
Efron, B., & Morris, C. N. 1977. Stein's paradox in statistics. Scientific American.
Ericsson, K. A., Prietula, M. J., & Cokely, E. T. 2007. The Making of an Expert. (Cover
story). Harvard Business Review, 85(7/8): 114-121.
Feiler, D., & Taylor, E. J. 2013. The hot seat: Over-attribution to leaders and dismissal for
bad luck. Academy of Management Proceedings, 1: 12371.
Feiler, D. C., Tong, J. D., & Larrick, R. P. 2013. Biased judgment in censored environments.
Management Science, 59(3): 573-591.
Feller, W. 1957. An Introduction to Probability Theory and its Applications, Vol. I. New
York, NY: Wiley.
Fischhoff, B. 1975. Hindsight not equal to foresight: Effect of outcome knowledge on
judgment under uncertainty. Journal of Experimental Psychology-Human
Perception and Performance, 1(3): 288-299.
Fitza, M. A. 2014. The use of variance decomposition in the investigation of CEO effects:
How large must the CEO effect be to rule out chance? Strategic Management
Journal, 31(11): 1839-1852.
Fitza, M. A. forthcoming. How much do CEOs really matter? Reaffirming that the CEO
effect is mostly due to chance. Strategic Management Journal.
Frank, R. H., & Cook, P. J. 1995. The Winner-Take-All Society. New York, NY: Free Press.
Froot, K. A., Scharfstein, D. S., & Stein, J. C. 1992. Herd on the street: Informational
inefficiencies in a market with shortterm speculation. The Journal of Finance,
47(4): 1461-1484.
Gates, B. 1995. The Road Ahead. New York, NY: Viking.
Gavetti, G. 2012. Toward a behavioral theory of strategy. Organization Science, 23(1): 267-
285.

24
Geletkanycz, M. A., & Hambrick, D. C. 1997. The external ties of top executives:
Implications for strategic choice and performance. Administrative Science Quarterly,
42(4): 654-681.
Gilovich, T. 2008. How We Know What Isn't So. New York, NY: Simonand Schuster.
Gladwell, M. 2008. Outliers: The Story of Success. London, UK: Allen Lane.
Greve, H. R. 2003. Organizational Learning from Performance Feedback: A Behavioral
Perspective on Innovation and Change. Cambridge, England: Cambridge University
Press.
Groysberg, B. 2010. Chasing Stars: The Myth of Talent and the Portability of Performance.
Princeton, New Jersey: Princeton University Press.
Groysberg, B., Nanda, A., & Nohria, N. 2004. The risky business of hiring stars. Harvard
Business Review, 82(5): 92-101.
Hakes, J. K., & Sauer, R. D. 2006. An economic evaluation of the Moneyball hypothesis. The
Journal of Economic Perspectives, 20(3): 173-185.
Hannan, M. T., & Freeman, J. 1984. Structural inertia and organizational change. American
Sociological Review, 49(2): 149-164.
Harrison, J. R., & March, J. G. 1984. Decision making and postdecision surprises.
Administrative Science Quarterly, 29(1): 26-42.
Haunschild, P. R., & Miner, A. S. 1997. Modes of interorganizational imitation: The effects
of outcome salience and uncertainty. Administrative Science Quarterly, 42(3): 472-
500.
Henderson, A. D., Raynor, M. E., & Ahmed, M. 2012. How long must a firm be great to rule
out chance? Benchmarking sustained superior performance without being fooled by
randomness. Strategic Management Journal, 33(4): 387-406.
Hollander, E. P. 1958. Conformity, status, and idiosyncrasy credit. Psychological Review,
65(2): 117-127.
Hollnagel, E. 2004. Barriers and Accident Prevention. Aldershot, England: Ashgate
Publishing, Ltd.
Jones, E. E. 1979. The rocky road from acts to dispositions. American Psychologist, 34(2):
107-117.
Kahneman, D. 2011. Thinking, Fast and Slow. London, UK: Penguin.
Kahneman, D., Lovallo, D., & Sibony, O. 2011. Before you make that big decision. Harvard
business review, 89(6): 50-60.
Kelley, H. H. 1971. Attribution in Social Interaction. New York, NY: General Learning
Corporation.
Keltner, D., Gruenfeld, D. H., & Anderson, C. 2003. Power, approach, and inhibition.
Psychological Review, 110(2): 265-284.
Keynes, J. M. 1936. General Theory of Employment, Interest and Money: New York, NY:
Harcourt Brace and Co.
Khurana, R. 2002. Searching for a Corporate Savior: The Irrational Quest for Charismatic
CEOs. Princeton, NJ: Princeton University Press.
Kilduff, M., Crossland, C., Tsai, W., & Bowers, M. T. 2016. Magnification and correction of
the acolyte effect: Initial benefits and ex post settling up in NFL coaching careers.
Academy of Management Journal, 59(1): 352-375.
Kuran, T. 1997. Private truths, public lies: The social consequences of preference
falsification. Cambridge, MA: Harvard University Press.
Langer, E. J. 1975. Illusion of control. Journal of Personality and Social Psychology, 32(2):
311-328.
Leveson, N. 2011. Engineering a Safer World: Systems Thinking Applied to Safety.
Cambridge, MA: MIT Press.

25
Levinthal, D. A., & March, J. G. 1993. The myopia of learning. Strategic Management
Journal, 14(8): 95-112.
Levitt, B., & March, J. G. 1988. Organizational learning. Annual Review of Sociology, 14:
319-340.
Lewis, M. M. 2003. Moneyball: The Art of Winning an Unfair Game. New York, NY: WW
Norton & Company.
Litov, L. P., Moreton, P., & Zenger, T. R. 2012. Corporate strategy, analyst coverage, and the
uniqueness paradox. Management Science, 58(10): 1797-1815.
Liu, C., & de Rond, M. 2016. Good night and good luck: Perspectives on luck in management
scholarship. Academy of Management Annals, 10(1): 409-451.
Liu, C., Eubanks, D., & Chater, N. 2015. Weakness of strong ties: Decision biases, social ties
and nepotism in family business succession. Leadership Quarterly, 26(1): 419-435.
Lorenz, J., Rauhut, H., Schweitzer, F., & Helbing, D. 2011. How social influence can
undermine the wisdom of crowd effect. Proceedings of the National Academy of
Sciences, 108(22): 9020-9025.
Lynn, F. B., Podolny, J. M., & Tao, L. 2009. A sociological (de)construction of the
relationship between status and quality. American Journal of Sociology, 115(3): 755-
804.
Makridakis, S., Hogarth, R., & Gaba, A. 2009. Dance with Chance: Making Luck Work for
You. London, England: Oneworld.
March, J. C., & March, J. G. 1977. Almost random careers: The Wisconsin school
superintendency, 1940-1972. Administrative Science Quarterly, 22(3): 377-409.
March, J. G. 1991. Exploration and exploitation in organizational learning. Organization
Science, 2(1): 71-87.
March, J. G., & Shapira, Z. 1992. Variable risk preferences and the focus of attention.
Psychological Review, 99(1): 172-183.
March, J. G., Sproull, L. S., & Tamuz, M. 1991. Learning from samples of one or fewer.
Organization Science, 2(1): 1-13.
Mauboussin, M. J. 2012. The Success Equation: Untangling Skill and Luck in Business,
Sports and Investing. Cambridge, MA: Harvard Business School Press.
McPherson, M., Smith-Lovin, L., & Cook, J. M. 2001. Birds of a feather: Homophily in
social networks. Annual Review of Sociology, 27(1): 415-444.
Meindl, J. R., Ehrlich, S. B., & Dukerich, J. M. 1985. The romance of leadership.
Administrative Science Quarterly, 30(1): 78-102.
Merton, R. K. 1948. The self-fulfilling prophecy. The Antioch Review, 8(2): 193-210.
Merton, R. K. 1968. The Matthew Effect in science: The reward and communication systems
of science. Science, 159(3810): 55-63.
Milgrom, P. R. 1981. Good news and bad news: Representation theorems and applications.
The Bell Journal of Economics, 12(2): 380-391.
Miller, D. T., & Ross, M. 1975. Self-serving biases in the attribution of causality: Fact or
fiction? Psychological Bulletin, 82(2): 213-225.
Moore, D. A., Oesch, J. M., & Zietsma, C. 2007. What Competition? Myopic Self-Focus in
Market-Entry Decisions. Organization Science, 18(3): 440-454.
Musch, J., & Grondin, S. 2001. Unequal competition as an impediment to personal
development: A review of the relative age effect in sport. Developmental Review,
21(2): 147-167.
Page, S. E. 2008. The Difference: How the Power of Diversity Creates Better Groups,
Firms, Schools, and Societies. Princeton, NJ: Princeton University Press.
Perrow, C. 1984. Normal Accidents: Living with High Risk Technologies. New York, NY:
Basic Books.

26
Peters, T. J., & Waterman, R. H. 1982. In Search of Excellence: Lessons from America's
Best-run Companies. London, England: Harper.
Phillips, D. J., Turco, C. J., & Zuckerman, E. W. 2013. Betrayal as market barrier: Identity-
based limits to diversification among high-status corporate law firms. American
Journal of Sociology, 118(4): 1023-1054.
Phillips, D. J., & Zuckerman, E. W. 2001. Middle-status conformity: Theoretical restatement
and empirical demonstration in two markets. American Journal of Sociology, 107(2):
379-429.
Piazza, A., & Castellucci, F. 2014. Status in organization and management theory. Journal of
Management, 40(1): 287-315.
Pierson, K., Addona, V., & Yates, P. 2014. A behavioural dynamic model of the relative age
effect. Journal of sports sciences, 32(8): 776-784.
Podolny, J. M. 1993. A status-based model of market competition. American Journal of
Sociology, 98(4): 829-872.
Pontikes, E. G., & Barnett, W. P. Forthcoming. The Non-consensus Entrepreneur
Organizational Responses to Vital Events. Administrative Science Quarterly: 1-39.
Porter, M. E. 1979. How competitive forces shape strategy. Harvard Business Review, 57(2):
137-145.
Porter, M. E. 1980. Competitive Strategies: Techniques for Analyzing Industries and
Competitors. New York, NY: Free Press.
Powell, T. C., Lovallo, D., & Fox, C. R. 2011. Behavioral strategy. Strategic Management
Journal, 32(13): 1369-1386.
Raynor, M. E. 2007. The Strategy Paradox: Why committing to success leads to failure (and
what to do about it). New York, NY: Random House.
Richerson, P. J., & Boyd, R. 2005. Not by Genes Alone: How Culture Transformed Human
Evolution. Chicago, IL: University of Chicago Press.
Ross, L. 1977. The intuitive psychologist and his shortcomings: Distortions in the attribution
process. Advances in Experimental Social Psychology, 10: 173-220.
Ross, L., & Nisbett, R. E. 1991. The Person and the Situation: Perspectives of Social
Psychology. London, UK: McGraw-Hill.
Rudolph, J. W., & Repenning, N. P. 2002. Disaster dynamics: Understanding the role of
quantity in organizational collapse. Administrative Science Quarterly, 47(1): 1-30.
Salganik, M. J., Dodds, P. S., & Watts, D. J. 2006. Experimental study of inequality and
unpredictability in an artificial cultural market. Science, 311(5762): 854-856.
Schelling, T. C. 1960. The Strategy of Conict. Cambridge, MA: Harvard University Press.
Schwab, A., Abrahamson, E., Starbuck, W. H., & Fidler, F. 2011. PerspectiveResearchers
Should Make Thoughtful Assessments Instead of Null-Hypothesis Significance Tests.
Organization Science, 22(4): 1105-1120.
Shleifer, A., & Vishny, R. W. 1997. The limits of arbitrage. The Journal of Finance, 52(1):
35-55.
Starbuck, W. H. 1994. On behalf of naivete. In J. Baum, & J. Singh (Eds.), Evolutionary
Dynamics of Organizations: 205-220. New York, NY: Oxford University Press.
Stone, P. 2009. The logic of random selection. Political Theory, 37(3): 375-397.
Stone, P. 2011. The Luck of the Draw: The Role of Lotteries in Decision Making. Oxford,
England: Oxford University Press.
Strang, D., & Macy, M. W. 2001. In search of excellence: Fads, success stories, and adaptive
emulation. American Journal of Sociology, 107(1): 147-182.
Syed, M. 2010. Bounce: The Myth of Talent and the Power of Practice. London, UK:
Harper Collins.

27
Taleb, N. 2007. The Black Swan: The Impact of the Highly Improbable. New York, NY:
Random House.
Taleb, N. N. 2001. Fooled by Randomness: The Hidden Role of Chance in Life and in the
Markets. London, England: Penguin.
Thorngate, W. 1988. On the evolution of adjudicated contests and the principle of invidious.
Journal of Behavioral Decision Making, 1(1): 5-15.
Thorngate, W., Dawes, R., & Foddy, M. 2008. Judging Merit. New York, NY: Psychology
Press.
Tinsley, C. H., Dillon, R. L., & Cronin, M. A. 2012. How near-miss events amplify or
attenuate risky decision making. Management Science, 58(9): 1596-1613.
Weick, K. E., & Sutcliffe, B. T. 2006. Managing the Unexpected: Assuring High
Performance in an Age of Complexity. London, UK: John Wiley & Sons.
Weiner, B., Frieze, I., Kukla, A., Reed, L., Rest, S., & Rosenbaum, R. M. 1971. Perceiving
the Cause of Success and Failure. New York, NY: General Learning Press.
Westphal, J. D. 1999. Collaboration in the boardroom: Behavioral and performance
consequences of CEO-board social ties. Academy of Management Journal, 42(1): 7-
24.
Westphal, J. D., Seidel, M.-D. L., & Stewart, K. J. 2001. Second-order Imitation: Uncovering
Latent Effects of Board Network Ties. Administrative Science Quarterly, 46(4): 717-
747.
Wiesenfeld, B. M., Wurthmann, K. A., & Hambrick, D. C. 2008. The stigmatization and
devaluation of elites associated with corporate failures: A process model. Academy of
Management Review, 33(1): 231-251.
Zeitoun, H., Osterloh, M., & Frey, B. S. 2014. Learning from ancient Athens: Demarchy and
corporate governance. The Academy of Management Perspectives, 28(1): 1-14.
Zollo, M. 2009. Superstitious learning with rare strategic decisions: theory and evidence from
corporate acquisitions. Organization Science, 20(5): 894-908.
Zuckerman, E. W. 1999. The categorical imperative: Securities analysts and the illegitimacy
discount. American Journal of Sociology, 104(5): 1398-1438.

28
APPENDIX
This appendix documents how the three graphs in Figure 1 were constructed. Figure 1A is
based on a computational model; Figure 1B is based on an analysis of data from sport, i.e.,
game results in Major League Baseball during 2000-2010; Figure 1C is based on data from
firm performances, i.e., public firms in the United States during 1980-2010. A shared feature
of the three analyses is that they all demonstrate a non-monotonic pattern between
performance and expected skill (or future performances).

How Figure 1A Was Constructed

The purpose of this illustration in Figure 1A is to examine when higher performances do not
indicate higher skill. The formal model is an extension of Model 2 in Denrell and Liu (2012)
which assumed that Pi,t=Ui,t+Ei,t where Ui,t is the skill of actor i at time t, Ei,t is a noise term
for actor i at time t, which is drawn from a normal distribution with mean zero. Pi,t is the actor
is performance at time t. In the model each actor starts with some level of skill (Ui,t) and risk
(the standard deviation of the noise term). The initial value of skill and risk is determined as
follows: (a) skill is drawn from a normal distribution with mean zero and standard deviation
equal to one; (b) the standard deviation of the noise term is drawn from a gamma distribution
with parameters (1,1). This implies that the standard deviation of noise has a mean of one but
a few actors can have very high values of noise due to the heterogeneity of risk.

An important observation is that heterogeneity in Ei,t implies that the aggregate distributions
of noise will have fatter tails than skill which is drawn from a standardized normal
distribution. Denrell and Liu (2012) shows that when this is the case, the expected skill, E[ui
|Pi], is not an increasing, but rather a non-monotonic function of the observed performance,
Pi. Figure 1A provides an illustration: expected skill is not increasing in the observed level of
performance. The reason is that high levels of performance are less informative about skills
than lower levels of performance. The explanation is that a very high performance indicates
that it was achieved in a context where there is substantial noise (i.e., high level of risk drawn
from a right skewed gamma distribution). In particular, it is decreasingly likely that an
extreme performance (e.g., greater than 3 or smaller than -3) results from an extreme value of
skill because 99.7% of the values of a standardized normal distribution are within +3 and -3.
The implication is that the highest (lowest) performances do not have the highest (lowest)
expected skill when the noise distribution is more fat-tailed than the skill distribution.

How Figure 1B Was Constructed

The purpose of this illustration in Figure 1B is to empirically examine whether our theoretical
prediction in Figure 1A holds using sport data. I downloaded the game results from Major
Baseball League (MLB) during 2000-10. I examined how the team winning percentages for
the first half of the season are associated with the team winning percentages during the second
half of the same season. The two observations of performance are nearly independent in the
sense that performance during the second half is unlikely to be substantially influenced by
performance during the first half. There are possible exceptions to this - players might be
influenced by what happened in the past - but there is at least less dependence within seasons
than across seasons (because players are more likely to be traded between seasons).

The data analysis proceeded as follows: First, the raw data of the MLB game log was
downloaded from Retrosheet.com. Second, the teams that played each other in both halves of
a season were extracted and the results of these games for both the first and the second half of
the season were calculated. Thus, the focus was on teams that repeatedly played against each
other (which is common). Limiting the analysis to such observations provides a way to

29
control for differences in the skills of the opponent. Third, the results in the first half of the
season were divided into eleven sections from 0 to 1 with an interval of 0.1. The average of
the game results in the second half of the season, given their first half season game results
was then calculated. The results are plotted in Figure 1B.

Figure 1B illustrates the phenomenon of regression to the mean: all of the results in the first
half of the seasons regress downward to the mean (supposing the mean for baseball
matchups to be 0.5) in the second half of seasons. In addition to such regression to the mean
we find that the relationship of the game results between the first and second half of the
seasons to be non-monotonic: the highest performers in the first half of the seasons performed
worse than the second best group (similarly for the lowest end). Stated differently, the data
shows that the highest performers in the first half of the season do not have the highest
average performance in the second half. This is consistent with our model (and pattern in
Figure 1A) and provides an illustration that the non-monotonicity predicted by our model can
emerge in sport data.

How Figure 1C Was Constructed

The purpose of this illustration in Figure 1C is to empirically examine whether our theoretical
prediction in Figure 1A holds using firm performance data. I chose return on assets (ROA), a
typical firm performance measure for this analysis. The data was downloaded from
Thompson Reuters database and included all active US public firms (excluding ADRs) from
1980-2010 and there are 7,147 firms in total during these 31 years. I computed the ranks for
the firms by their ROA for each year as their performance measures, primarily because using
ranks as performance measure is more robust than using the averages. Similarly to the
analysis in Figure 1B, I first separated the firm performance ranks into groups at year t. I then
computed the expected rank for each of these groups at year t+1. Figure 1C is based on the
result of 2004-2005 year pair.

The result in Figure 1C is robust I replicated this pattern in most other formulations, such as
year, the interval binning or selection. In particular, I examined 660 combinations that consist
of different years (30 year-pair), different intervals binning (including different number of
firms per interval, 11 variations) and different selections (either including all firms or only
firms that survived all 31 years, 2 variations). In 608 of the combinations I examined (92.1%),
the top-performing group at year t does not have the highest performance in year t+1. This
suggests a robust non-monotonic association between past and future performances.

30