Cutting Your Teeth: Building on the Micro-Foundations for Dynamic Capabilities
Charles E. Eesley
Stanford University
Department of Management Science & Engineering
380 Panama Way
Stanford, CA 94305
Tel: (740) 236 4653
[email protected]
Edward B. Roberts
Massachusetts Institute of Technology
50 Memorial Drive
Cambridge, MA 02142
Tel: (617) 253 4934
[email protected]

We gratefully acknowledge funding from the Ewing Marion Kauffman Foundation and the M.I.T.
Entrepreneurship Center. The authors would like to thank James Utterback, Mark Kennedy and participants at the
West Coast Research Symposium (WCRS) for helpful comments.
2
Cutting Your Teeth: Building on the Micro-Foundations for Dynamic Capabilities
Abstract:
Theorizing on the sources of dynamic capabilities has been rare, particularly on the micro-level
determinants of capabilities. We look at evidence for a model where cross-functional experience
(in this case, founding a firm) results in more accurate prediction of industry trends. This paper
investigates whether cross-functional experience improves entrepreneurial firm performance via
more accurate cognitive representations of the competitive landscape. We show that a major
industry disruption (in the form of the dotcom boom) resulted in the loss of benefit from
cognitive maps developed in the prior environment. The results provide evidence consistent with
our model of cognitive maps as a source of capability and counter to several alternative accounts.
3
The question of why some firms develop more valuable resources and capabilities than
others has been a central one for strategy researchers. The resource-based view (RBV) of the
firm has been a central theoretical framework for explaining heterogeneity in firm performance
(Peteraf, 1993; Wernerfelt, 1984). There is a growing consensus that differences in firm
performance are driven by higher order capabilities to develop or reconfigure bundles of
resources that will generate value over time (Teece, Pisano, & Shuen, 1997; Winter, 2003).
However, we still do not understand the source of dynamic capabilities and why some
managers are able to develop more valuable resources than others. A growing stream of literature
examines how managers may play an active, cognitive role in making bets and predicting which
resources and capabilities will be more valuable than others. Managers face numerous cognitive
challenges as they attempt to understand the competitive environment and its trends in enough
detail to search for a cluster of resources that will provide competitive advantage (Gavetti &
Rivkin, 2007). Cognitive representations have been identified by prior research as important in
shaping decision-making as well as focus and interpretation (Fiol & Huff, 1992; Simon, 1991;
Walsh, 1995; Weick, 1995). Yet, existing work has focused on the content of mental models
rather than on their use for certain capabilities, such as predicting what resources will be more
valuable in the near future.
The purpose of this paper is to address the gap in our understanding of the sources of
dynamic capabilities. We show that one initial source of firm-level dynamic capabilities is the
cognitive representations of its founder(s) that result from their careers. Conceptually linking the
individual and firm levels, we extend the line of work that recognizes that strategy arises from
managers’ cognitive theories about the world (Gavetti & Rivkin, 2007; Porac, Thomas, &
Baden-Fuller, 1989). We demonstrate that improved cognitive maps of industry trends and
4
causal relationships come from specific types of industry experience and guide the successful
recombination of resources and capabilities. We respond to calls in the literature to build the
behavioral foundations of our understanding of the evolution of capabilities (Gavetti, 2005).
THEORY AND HYPOTHESES
Introduction. Drawing on evolutionary economics, the existing conceptions of the
sources of dynamic capabilities have been guided by relatively passive, inertial mechanisms of
incremental, local search (Teece, Pisano, & Shuen, 1997; Zott, 2003). This literature has argued
that firms inherit rule-based mechanisms for choosing a local search method and that firm
productivity is a result of the performance of these inherited search routines. Evolutionary
theorists have proposed higher order routines that guide history-dependent, stable, local search
processes for more valuable bundles of capabilities and resources (Winter, 2003). Yet, routinebased theories of managerial behavior have long been thought by scholars to be entwined with
more cognitive logics (Gavetti & Levinthal, 2000; March & Olsen, 1976). A strictly inheritancebased account neglects the active role that managers also play in analyzing the landscape to
make strategic decisions to develop some resources and capabilities rather than others (Gavetti &
Levinthal, 2000). This paper emphasizes cognitive representations that guide search and firm
performance.
Industry-level beliefs or cognitive representations influence action and decision-making
(Levitt & March, 1988). Common cognitive representations can translate into norms that
perpetuate the competitive environment (Bogner & Barr, 2000) or that result in mistaken
assumptions about changes in technology and errors in predicted technological performance
limitations (Henderson, 1995). Numerous studies have explored the sociocognitive factors
influencing competitor definition, both within firms and across organizations (Porac, Thomas, &
5
Baden-Fuller, 1989; Porac & Thomas, 1990). Other work within this growing body of literature
has examined the extent of cognitive inertia (or resistance to change) in connection with
cognitive representations of the competitive environment (Hodgkinson, 1997; Reger & Palmer,
1996). Much of this work centers on the idea that managers view and communicate with a subset
of firms, the strategic group, as their competition and that the social and cognitive factors
defining this subset also frame the firm’s strategic choices. One of the main insights arising from
the stream of work on cognitive representations of competition is the idea that managers become
locked into ways of thinking about how to compete on the basis of how they define their
competitors. Even broader than the strategic group, other studies have shown that there are
shared cognitive representations across the industry (Phillips, 1994; Hodgkinson, 1997).
Organizations in the same industry have been shown to share metaphors or worldviews on
customer preferences, future demand, or regulation (Huff, 1982, p. 125), develop “industry
recipes” for looking at situations (Spender, 1989 p.188) and create shared causal beliefs (Porac,
Thomas, & Baden-Fuller, 1989). In addition, common prior industry experience has been shown
to influence product feature choices as well (Benner & Tripsas, 2009). In the knitwear industry,
“industry models” are shown to represent shared models of the fundamental attributes used to
categorize competitors (Porac, Thomas, Wilson, Paton, & Kanfer, 1995). These models are
thought to form through repeated social interactions with customers, suppliers and other
stakeholders along the supply chain, together with managers of rival firms and perpetuate
cognitive biases, resulting in cognitive inertia (Chattopadhyay, Glick, Miller, & Huber, 1999;
Porac, Thomas, Wilson, Paton, & Kanfer, 1995)Lant & Baum, 1995; Reger & Huff, 1993). The
forces of cognitive inertia and shared cognitive representations of the competitive landscape
cause firms to neglect changes in the environment and are forces leading to greater continuity
6
and homogeneity in strategy and performance. Taking cognitive inertia and firm inheritance of
routines into consideration, firms should converge over time on common search routines and
strategies. Drawing on this stream of work, it is possible to argue that success among a group of
related firms breeds a bias toward the tried and trusted. Institutional theory has also documented
that similar institutional environments lead to isomorphism in firm practices and imitation of
structures and behaviors (Dimaggio & Powell, 1983).
However, our understanding of how and why certain managers, in the face of cognitive
inertia are able to uniquely identify resources and capabilities that will lead to future
differentiation, higher performance and competitive advantage is much more limited.
Researchers have not focused as much attention on cognitive representations of the competitive
landscape (outside of competitor definition) or specifically on the uses and implications of
cognitive representations for variation in firm-level performance. From the entrepreneur’s
perspective, or even the manager developing a new product line or diversifying into a new
market, the problem is to make a bet or a prediction on which bundle of resources, capabilities or
product features will be most valuable at some point multiple months or years in the future.
Successfully making this prediction requires a cognitive representation of the industry that
allows for causal models and beliefs about trends occurring in the industry that will make some
bundles yield a firm performance advantage. The inability to notice key trends in the industry
due to cognitive biases in framing may inhibit the entrepreneur or manager’s ability to predict
which resources and capabilities to develop.
Our aim is to contribute to the literature on strategic capability and resource development
by building on the micro-level behavioral foundations. In addressing the gap in our
understanding of how a manager’s cognitive representations guide strategic capability
7
development, we build on work that has investigated the extent to which managers’
representations of the competitive landscape vary with variation in functional background and
experience (Hodgkinson & Johnson, 1994). A growing line of work by strategy and
organizational theorists demonstrates that firms can benefit from the careers of their founders
and employees (Beckman & Burton, 2008). In particular, we are interested in whether variation
in career experience results in variation in the usefulness of cognitive representations for the
purpose of predicting which strategic resources and capabilities will drive performance. This
analysis will allow us to address a puzzle in the literature: When organizations are in the act of
assembling new bundles of routines and capabilities, what determines who puts together the
more valuable and difficult to imitate bundles? We further examine when cognitive
representations are especially important and when they are less important or even a disadvantage.
Psychological theory offers mechanisms that help build our understanding of how
differences in heuristics and biases at the individual-level psychological level might shape the
firm-level resources and assets that are perceived to match industry trends. Cognitive
representations can be thought of in terms of groupings of causally related beliefs or subjective
probabilities that individuals give to various events or the relationships between actions and the
likelihood of specific outcomes. These representations can be inaccurate, overly simplified, or of
varied usefulness for the purpose of deciding which resources and capabilities to develop.
Most research on cognitive biases in the strategy field has followed the heuristics and
biases tradition of behavioral decision theory and advanced the notion that managers are marked
by a tendency to attribute failure to a range of externalities beyond their personal control while
attributing success to their own actions and capabilities (or those of their own organizations)
(Barr, Stimpert & Huff, 1992; Clapham & Schwenk, 1991; Das & Teng, 1999; Lant, Milliken &
8
Batra, 1992; Schwenk, 1986, 1988). Less work has focused on what types of experiences lead to
less (or more) biased judgments and decision-making or the characteristics of careers and
experience that lead to more useful cognitive representations.
Impact of career experience on cognitive bias. We know from work on the availability
heuristic that instances that can be easily recalled or scenarios that can be constructed with ease
tend to bias the judged probabilities of such events upward from reality (Tversky & Kahneman,
1973). Many terms and a long line of work has described a similar and persistent phenomenon.
People with less domain knowledge tend to rely on the ease of recollecting instances, whereas
domain experts tend to be biased as well by the number of cases that come to mind (Dubé-Rioux
& Russo, 1988; Fischhoff, Slovic, & Lichtenstein, 1978; Ofir, 2000; Russo & Kolzow, 1994).
The availability heuristic has been used to explain the “pruning bias” where individuals are given
a set of categories containing reasons why, for instance, a car might fail to start (or why a
business might go bankrupt). When the researchers “prune” categories and absorb them into the
residual category, respondents appear to distribute the probability from the pruned categories
across the remaining ones rather than add it to the residual category. Pruning bias can also be
explained more broadly by partition dependence (Fox & Clemen, 2005). Partition dependence is
where assessed probabilities can vary substantially with the way that the researcher partitions the
categories. The phenomenon is similar to framing, where the way in which alternatives are
presented influences decisions (Tversky & Kahneman, 1986). In the same way that social and
cognitive factors limit perceived competitors to a strategic group shaping the range of strategic
options considered and framing failures as being due to external forces (Porac, Thomas, &
Baden-Fuller, 1989), certain career experiences may partition an individual’s cognitive
representation of trends in the industry. Given the prior research in psychology cited above it
9
seems likely that if different work experiences lead to variation in mental partitioning of the
industry, some managers may hold more accurate cognitive maps than others
A rapidly growing area of literature examines prior career experience as an individuallevel asset carried with the individual to a new organization (Beckman & Burton, 2008; Boeker,
1989). A large section of this work has been interested in the knowledge, aspirations, skills, or
routines that employees inherit from their firms (Agarwal, Echambadi, Franco, & Sarkar, 2004).
Other evidence shows that resources accrue to individuals and firms based at least partly on the
structural positions of their former employers (Burton, Sørensen, & Beckman, 2002). However,
employees may gain more than just task knowledge, skills or routines from their firms, they may
also acquire beliefs and cognitive representations of the competitive environment which have
varying degrees of usefulness or accuracy (Dearborn & Simon, 1958; Hambrick & Mason,
1984). We have reason to believe that variation in career experiences might give rise to variation
in the usefulness of cognitive representations. Case studies and theoretical work shows that
cognitive framing and prior industry experience can lead to inertia in commercializing new
capabilities (Tripsas, 2009) and can have different impacts beyond those predicted by economic
or organizational models at various stages in the technology life cycle (Kaplan & Tripsas, 2008).
Prior work has shown that alongside industry maturity and firm inertia, the location of an
individual within an organizational hierarchy can shape the accuracy of cognitive representations
and the choice of search mechanisms (Gavetti & Rivkin, 2007). While many have examined the
functional diversity of the top management team, recently, a stream of research has developed
the notion of intrapersonal functional diversity – the number of different functions an individual
has worked in (Bunderson & Sutcliffe, 2002). A large-scale study found a negative relationship
between intrapersonal functional diversity and commitment to the strategic status quo
10
(Geletkanycz & Hambrick, 1997). Individuals were more likely to consider alternative strategies
if they had intrapersonal diversity and were more less likely to consider alternatives when their
work experience was in a single function.
Intrapersonal functional diversity may allow consideration of alternative strategies due to
effects that functional diversity in work experience have on cognition and the ability to correctly
see industry trends and make predictions. Recent work experience that is heavily focused in a
single, narrow functional domain can result in more focused attention on a subset of the
competitive environment (for example, technology, to the neglect of shifting markets). The result
is more detailed partitioning of that sub-area while other functional areas of the landscape
become more grossly partitioned and less detailed in the cognitive representation. The other
areas are less available to the mind in seeing trends and decision-making over what capabilities
will result in competitive advantage in the future. More detailed partitioning of one sub-area
leads to underweighting the probability and importance of trends, constraints or negative events
from other functional areas. In this way, strategic search and the bundles of capabilities a
manager chooses to develop are likely to be optimized for one domain or miss important changes
in other domains. The individual for whom cognitive partitions are more detailed in one area and
constraints related to that function are more accessible will be resistant to strategic options that
may fit constraints elsewhere but are less ideal in the domain of their own work experience.
However, for an individual who has responsibilities and decision rights that span multiple
functions, the scope of scanning and monitoring the relationships between actions and outcomes
in the industry is much wider. While depth may be lost, a sense of the landscape and changes in
the competitive environment through multiple functional lenses (i.e., marketing, technology,
finance, etc.) develops. Work experience that includes responsibilities simultaneously spanning
11
multiple functional areas is particularly important for the capacity to make predictions on the
future value of asset and capability bundles. Prior theoretical work has argued that managers
with more functionally diverse career histories may contribute the most to organizational
performance (Gupta, 1984). For the purposes of using cognitive maps to search for bundles of
resources that will be valuable in the coming years, current beliefs and knowledge of industry
trends is necessary. An individual in a position to perceive industry trends and constraints across
diverse areas through multiple functional lenses in an up-to-date fashion will have more accurate
cognitive representations of trends and a greater likelihood of identifying capabilities that can be
developed to profit from those trends. The above discussion leads to the following prediction:
Hypothesis 1: Individuals with more cross-functional experience will develop greater
predictive ability for the future value of resources and capabilities and found firms with
higher performance.
Improved cognitive representations are likely to be more beneficial in certain
environments. In psychology, a transfer effect is the beneficial impact of a prior event on the
performance of a subsequent event (Finkelstein & Haleblian, 2002). The similarity between the
characteristics of events influences the probability of positive outcomes from transfer (Argote &
Ingram, 2000). The probability of negative outcomes can increase when events are dissimilar yet
lessons from prior experience are applied anyway (Finkelstein & Haleblian, 2002). We suggest
that a distinct but related mechanism to transfer effects occurs between individual representations
and the organizational levels of analysis. Typically transfer effects involve learning a skill in one
context but with a loss of performance when the skill is wrongly applied in a different context.
However, different from learning a skill, cognitive maps honed on one competitive landscape
will lead to errors when used to guide firm resource development in a new landscape.
12
Previous literature argued that an individual can bring transfer effects to the benefit (or
detriment) of a new organization’s performance, depending on the similarity of industrial
contexts (Cohen & Bacdayan, 1994; Zander & Kogut, 1995). Inferences may be misapplied or
the wrong inferences may be drawn from the beginning (Finkelstein & Haleblian, 2002). Along
these lines, we argue that cognitive representations developed in one context will not aid
strategic search in a substantially different context. Furthermore, examining the impact of the
similarity of experience will allow us an additional test of whether higher performance for
subsequent firms is a result of improved representations or of higher skill levels. Unless higher
skill individuals tend to remain in the same industrial context, better performance for those with
experience in a similar industry compared to those with prior experience in a different industry
supports better cognitive maps as the correct mechanism. These arguments yield two predictions:
Hypothesis 2a: Cross-functional experience will lead to higher performance for those
remaining in the same industry.
Cognitive representations and industry disruptions. For the same reasons that when an individual
moves to a different industry environment, beliefs and cognitive representations built in the old
environment are less helpful, major disruptions to an industry have a similar impact. When a
significant disruption occurs in an industry then cognitive representations developed in the
previous industry environment will no longer be relevant. In the new, more fluid period,
individuals will again have to struggle to find the right causal relationships and models to fit the
changed environment (Kaplan & Tripsas, 2008).
Hypothesis 2b: After a significant shift in the industry, improved cognitive representations
honed in the previous industry environment through cross-functional experience will become
a liability.
Success, hubris and cognitive representations. Certain individuals may benefit more from crossfunctional experience. Whether an event turns out as a success may influence what knowledge
13
about the competitive landscape the individual takes away from a previous experience and how
she or he applies that knowledge to future situations (Cyert & March, 1963). Prior success can
show a path forward, but it may not spur much additional thought about why the success
occurred. Failure can have positive benefits by increasing the search for new opportunities
(McGrath, 1999) and can create greater variety in actions (March, 1991). Extensive literatures on
over-confidence, hindsight bias, illusion of control (Das & Teng, 1999; Schwenk, 1986) and
managerial hubris (Hiller & Hambrick, 2005) show that managers may wrongly believe that they
have acquired surperior knowledge and misapply lessons from prior experience. However, this
long line of research has also shown that certain individuals are more likely to suffer from the
downsides of hubris than others and a certain level of self-confidence aids high levels of
performance (Hiller & Hambrick, 2005). Success does not universally come with the downsides
of hubris and over-confidence. The literature appears somewhat mixed on whether more
knowledge is gained from success or failure (Starbuck WH, 2001). However, prior arguments
have neglected one important mechanism that yields improved cognitive representations from
success experience, but not from failure. When an individual experiences significant success
others often come to that person to seek their advice or involvement in new opportunities. This is
less true for those who have failed. The flow of other individuals proposing new ideas not only
increases the number of trends and opportunities examined but may increase the detail and
accuracy of representations through a sharp increase in the signals of trends in technology,
competitors, and/or markets. It is clear that individuals do learn from failure as well as success
and that learning from failure can enhance survival (Kim & Miner, 2007). Yet those who have
failed lack this rich source of information on industry trends that may be more important for
identifying new capabilities.
14
Hypothesis 3: Cross-functional experience will lead individuals who have experienced success
to continue higher performance.
A variety of studies show that the degree to which an individual has certain demographic
characteristics predicts his or her interpretations and receptivity to change (Staw & Ross, 1980;
Stevens, Beyer, & Harrison M. Trice, 1978). Training and background experiences influence an
individual’s cognitive base (Cyert & March, 1963) and demographic variables are known to
predict beliefs and values (Walsh, 1988), and the ability to integrate new information (Taylor,
1975). Demographic characteristics have certain advantages even when the underlying,
intervening constructs are mental processes (Hambrick & Mason, 1984; Pfeffer, 1983).
An individual’s cognitive skills and ability are reflected in the level of education attained.
The capacity to classify a variety of stimuli and greater ability for information process are
associated with high levels of education (Schroder, Driver, & Steufert, 1967). The ability to
integrate complex information, engage in boundary spanning and tolerate ambiguity have been
associated with more highly educated individuals. Education has also been associated with being
receptive to innovation (Bantel, K. A. and Jackson, S. E., 1989; Becker, 1970; Kimberly &
Evanisko, 1981). Higher education levels in the top management team have also been shown to
result in a greater likelihood of changing corporate strategy (Margarethe F. Wiersema & Bantel,
1992). The evaluation of acquisition candidates is influenced by educational characteristics as
well (Hitt & Tyler, 1991). Assuming cognitive ability is correlated with education level, an
individual’s ability to translate experiences and information into cognitive representations with
sufficient detail to make predictions of valuable resources and capabilities should be associated
with higher levels of education. The correlation between education and cognitive abilities
suggests that more highly educated individuals should be better able to use cross-functional
experiences to make accurate choices on which resources to develop.
15
Hypothesis 4: More highly educated individuals will benefit more from cross-functional
experience.
The impact of cross-functional work experiences on the ability to see important industry
trends and predict which resources and capabilities will be more valuable to develop should
decline with time. The more time that has passed since the cross-functional experience, the less
relevant it will be for creating useful cognitive representations and predictions of sources of
future value. Interference from other tasks has been shown to cause forgetting and also not only
does forgetting occur with delays (Anderson, 2003; Argote, Beckman, & Epple, 1990) but also
the trend information and causal models are less likely to be relevant.
Hypothesis 5: More recent cross-functional experience matters more to develop greater
predictive ability for the future value of resources and capabilities.
Empirical Strategy and Setting
Finding a setting where we can empirically isolate the impact of cognition separately
from the inheritance of routines or resources is challenging. Attempting to test our hypotheses in
a sample of existing large firms would be problematic for several reasons. One is that the
original founders who guided the initial conditions and strategic search when inertia was lower
may no longer be around to respond to questions. Second, the selection mechanisms may be far
from clear as to which managers are assigned to which projects and how much of the
performance of those projects is due to the formal or informal control of managers in specific
locations in the hierarchy. Using existing public firms only also introduces survival bias. Finally,
transfer of resources or existing capabilities from other business units in the firm makes it
challenging to identify the effects that interest us. However, a novel survey allows us to make
empirical progress, particularly on individual-level data and multiple measures of firm
performance where prior work has largely relied on simulation analysis. At the founding of a
16
new firm, resources are limited and existing firm capabilities are non-existent, allowing us to
better isolate the impact of cognitive representations developed from prior work experience. The
founders are a key input at the beginning stages of a firm and other work shows the importance
of founders in framing the initial conditions for strategic search (Gavetti & Rivkin, 2007; Kaplan
& Tripsas, 2008; Tripsas & Gavetti, 2000).
The idea of organizational inertia is that elements of the organization are surprisingly
resistant to change once established early in the life of the organization (Baron, Hannan, &
Burton, 1999). The closely related idea of imprinting is that both environmental conditions at
founding and strategic decisions made by the founders early on are not easily reversible and
leave a lasting imprint on the firm’s subsequent development (Romanelli, 1991; Stinchcombe,
1965). Founders bring blueprints or models that then shape the future directions of the firm
(Baron, Burton, & Hannan, 1999; Beckman & Burton, 2008).
Cognitive representations guide search mechanisms and become important initial
conditions that imprint the future directions of development for the company. Furthermore,
organizations are known to become less plastic as they age (Hannan & Freeman, 1977) so the
cognitive maps available to managers in the beginning stages of developing a new business line
or new venture are vital since hiring individuals with higher quality cognitive representations
after inertia has set in will have less of a payoff. If more accurate cognitive representations of the
competitive landscape function as dynamic capabilities that improve firm performance, then this
is a source of competitive advantage which other firms clearly would not be able to imitate. It is
impossible for competitors to go back in time and replace the founders with others who had
different experiences and beliefs about the competitive landscape.
Data and Methods
17
Our identification strategy begins with the selection of the setting of company founders
as a more narrow and ideal empirical sample to test our hypotheses because of the factors
discussed above. In particular, it better allows us to isolate the impact of cognition from the
complicating factors of existing firm routines, capabilities and resources that would otherwise
more strongly impair the identification of the effects we hypothesize about in other settings.
We take advantage of both differences between those who have and have not had the
experience of founding a firm as an indicator of cross-functional experience. Those who have
worked in established companies are much more likely to have worked in a single functional
domain at a time, whereas we know that those who were founders almost surely had work
experience that crossed functional boundaries while founding the firm. Those individuals with
more cross-functional experience compared to those with less, offer another identification
strategy. If an episode of cross-functional experience is thought of as the treatment effect, then
we look both at differences between the treated and untreated as well as difference between those
individuals with more episodes of the treatment effect compared to those with fewer.
We use a novel survey administered in 2001 to all 105,928 alumni from a prestigious
research- and technology-based university in the United States to generate a sample of firms
where we have detailed information on founders as well as on firm performance. This survey
generated 43,668 responses. Out of 7,798 alumni who had indicated that they had founded a
company, 2,111 founders completed more detailed surveys in 2003, representing a response rate
of 25.6%.1
Examining the firm names and founding years, we identified and dropped 44
duplicate observations where multiple cofounders reported on the same firm. Industries covered
include aerospace, architecture, biomedical, chemicals, consumer products, consulting,
1
In previously published work the authors (2007) show t-tests of the null hypothesis that the average
(observed) characteristics of the responders and non-responders are the same statistically, for both the
2001 and 2003 surveys.
18
electronics, energy, finance, law, machine tools, publishing, software, telecommunications, other
services, as well as other manufacturing. Each founder reported information on up to five firms
which he or she had founded up to the date of the survey, yielding a total of 3,698 firm
observations. On average, 1.79 firms were founded per individual, or 3.85 firms per individual
who founded more than 1 firm. The founders were also asked for the total number of firms they
had attempted to found over the course of their career and 80 indicated having founded more
than 5 firms (up to 11). The average number of firms per individual by this measure is 2.13 so
we appear to have captured data on the vast majority of firms. To provide still more information
about these companies including current sales, employment, industry category and location, this
new database was further updated to 2006 data from the records of Compustat (for public
companies) and Dun & Bradstreet (private companies).2 For consistency in the country and
institutional context, the 1,121 firms that were founded outside of the U.S. were dropped from
the analyses reported in this paper. Information on revenues was adjusted for inflation.
Although teams of multiple co-founders are more likely to start a new firm, as well as be
more successful in their firms (Roberts, 1991), we only have complete founder information on
prior startup experience for one entrepreneur from each team. Previous findings of strong
homophily among founding teams indicate that the prior founding experience of one
entrepreneur may be a good proxy for that of the team (Ruef, Aldrich, & Carter, 2003) and the
results are robust to using only the single-founder teams and to using all co-founded teams. To
eliminate concerns of biases in the Dun & Bradstreet data, the analyses are also run on only the
subset of firms for which the founders provided more detailed revenues and employee data (each
multi-company founder chose only one firm to provide more detailed data). Although we lose
2
Successful matches were found for 80% of the company names in the D&B database. A firm is included
in the Dun & Bradstreet database when it needs to obtain a credit rating.
19
the panel structure, this sub-sample also provides us with more detailed information on control
variables and increases the confidence in our results. Due to skipped survey items and missing
data we allow the final number of observations to vary by the analysis in order to utilize all
available data. Meaningful numbers of foundings begin in the 1950s, therefore we restrict our
analysis here to firms founded from 1950-2003. A key feature of this dataset is its long time
horizon allowing us to analyze almost entire careers.
Dependent Variables. Because our focus is on measuring the performance effects of
variation in the cognitive representations that come from work experience, we use revenues and
exits in the form of acquisition or initial public offerings (IPO) as the dependent variables. No
single outcome measure is ideal. Using acquisitions has the drawback of not observing the
valuation of the acquisition as compared to the valuation at the time of funding. Similarly, using
IPOs does not identify the valuation of the firm at the time of the IPO, or the post-IPO
performance of the stock, or the personal financial benefits to the founders or the initial
investors. Both IPOs and acquisitions apply only to a subset of foundings, not to all of them,
whereas revenues are a common goal of all companies. The limitation of using the fact of IPO or
acquisition is that both of these are sensitive to the industry, the economic environment, and the
founders’ desire to retain control. It is best to consider multiple performance measures, which is
why we look for (and find) robustness with various measures. The variable log revenues is the
revenue for the most recent fiscal year in operation as reported by the entrepreneur. We adjust
for inflation (2001 $) and take the natural log of this measure for our dependent variable. Out of
2,111 firms, 1,370 survey respondents reported revenues for their firms ranging from $0 to $2.56
billion (mean = $34.6 million, median = $1.12 million).3
3
To alleviate concerns of response bias where defunct firms might be non-responders, we examine the
proportions of firms “in operation”, “acquired”, and “out of operation” in the group reporting revenues
20
Independent Variables. The key independent variable is our proxy for cross-functional
experience, which is coded as a one if the individual has founded at least one previous start-up
and zero otherwise.4 We argue that the main difference is that founders who have had prior
experience in a small, entrepreneurial firm are more likely to have utilized general skills and to
have had a wider range of responsibilities compared to individuals in larger firms who typically
have had more narrow functional roles. Prior work has already shown that entrepreneurs and
leaders appear to invest in more general human capital and diverse experiences relative to others
(Lazear, 2004). We expect that those who have been CEOs would similarly build more balanced
cognitive maps as a result of their cross-functional experience. However, CEOs also have much
higher opportunity costs relative to most employees so that would simultaneously influence the
performance of the projects they select to pursue. Because firms develop at different paces in
different industries and the wisdom of early decisions is often not known until the founders have
experienced some sort of exit event (if ever), we propose that cross-functional experience, rather
than years in a cross-functional role, is a more suitable proxy for the amount of cross-functional
experience. Another problem with using the number of years of experience is that it implicitly
penalizes an entrepreneur who quickly took a firm successfully to acquisition or IPO. Similarly,
with focus on the repetitive task, pilots might be expected to learn from the number of flights or
take-offs and landings, not from the number of miles flown. Managers cannot truly gauge the
ultimate success of their actions until the final outcome is known.
(1,424 observations) and the group of non-responders (687 observations) to this question. Our concerns
are alleviated in finding that the proportions are roughly equivalent with 68.5% of those reporting
revenues still in operation and 62.3% of the non-responders still in operation. 10.9% of the reporting
firms were out of operation whereas that number is 18.8% for the non-responders. 19.7% of the reporting
firms had been acquired, whereas 18.8% of the non-responders had been acquired.
4
A total of 3,156 alumni indicated that they had started multiple companies, of whom 960 completed the
survey for a multi-founder response rate of 30.4%. A total of 1,107 single-firm founders responded to the
survey giving a 21.8% response rate out of the 5,086 single-firm alumni founders.
21
In recent years there has been an explosion of scholarly interest in the links between
managers’ mental representations and their background characteristics. Research on personal,
cognitive characteristics and organizational performance has taken two approaches to
measurement. Both methods have their advantages and their limitations (Markoczy, 1997;
Hodgkinson et al., 2004). The direct assessment approach has been used in relatively small
samples, such as laboratory studies and CEO studies (Miller, Kets de Vries and Toulouse, 1982).
A large stream of work has examined the content of belief or knowledge structures, but very few
studies have examined the origins or consequences of these structures (Walsh, 1995). Markoczy
(1995) finds that managers’ functional area explains more of the similarity in business beliefs
than nationality. Other work has shown that beliefs are not fixed and inert, but rather executives
appear to socially influence one another’s beliefs (Chattopadhyay et al., 1999). The other
approach is the use of proxy variables such as demographic characteristics or work experiences
under the assumption that these are related to cognitive abilities, knowledge structures, or beliefs.
This demographic approach has typically been used in the literature on top management teams
(Bantel and Jackson, 1989) and in the strategy literature (Kaplan, 2008; Benner and Tripsas,
2009). This study uses the latter approach with the common disadvantage that such proxy
variables should not be expected to perfectly covary with cognitive attributes (Hambrick and
Mason, 1984; Michel and Hambrick, 1992). However, the advantage is that they are more
practically implemented in large enough samples to statistically assess performance implications.
As a measure of success experience, we use Prior IPOs, the number of previous IPOs for
an entrepreneur’s previous firms. The variable Prior Acquisitions is a count of the number of a
founder’s prior firms which have been acquired. Capturing the similarity of the industrial
context, Same 2-digit SIC code is a count of the number of prior startups that have the same 2-
22
digit SIC code as the current firm. Different 2-digit SIC code is a count of the number of prior
startups with a different 2-digit SIC code as the current firm. SIC and VEIC codes were matched
from the Dun & Bradstreet Million Dollar database and from VentureXpert. VEIC codes were
converted to SIC codes with a previously used matching scheme (Dushnitsky and Lenox, 2005).
Control Variables. We use a set of industry dummies as controls for the coarse industry
segment within which the firm competes (such as biotech, software, and electronics). Individuals
also differ in their starting human capital and in particular in the number of years of education
they have received. We include dummy variables indicating whether the individual’s highest
degree was a Bachelor’s degree or Master’s degree. The variable founder age is the
entrepreneur’s age when the firm was founded. We control for the age of the startup, as
measured by firm age from founding to the year for which revenues are reported. A set of year
dummies, one for each year from 1950-2003, captures temporal changes in the economy. Initial
Capital is the natural log of the amount of initial capital raised (adjusted to 2001 dollars, defined
as capital raised within the first year after founding). VC Funded is equal to 1 if the individual
reported raising funds from venture capital firms.
Results
--------------------------------------Insert Tables 1 and 2 about here
--------------------------------------Descriptive statistics and variable definitions are presented in Table 1. Table 2 is the
pairwise correlation table. Table 3 shows the univariate difference in means tests for log
revenues and exited conditional on cross-functional experience and other key independent
variables. Significance tests are reported for differences in conditional means from left to right in
the columns. Panel A shows that cross-functional experience is associated with significantly
higher revenues and probability of exit. The results in Panel B show support for the hypothesis
23
that the effect is stronger for those remaining in similar industry contexts. Panel C and Panel D
restrict the sample to software firms and show that while cross-functional experience was a
benefit prior to a major industry disruption, in the post-dotcom boom period it decreased the
likelihood of exit. Panels E and F show some support for the results being stronger for those with
a shorter lag from the cross-functional experience and for Master’s degree holders.
Multivariate Regressions on Firm Performance
Table 3 shows initial support for our hypotheses, but we turn to Table 4 to test whether
the results hold once controls have been introduced. We begin with a baseline model. To test the
influence on firm performance of prior cross-functional experience we use a variant of a
production function modified to better fit the case of younger firms. The Cobb-Douglas
production function is widely used, but in the case of private firms, output and capital in
particular are extremely difficult to measure for a number of reasons. Cognitive representations
of the industry resulting from varied prior experience of the founders are considered an “input”
into the firm formation process in the sense that a more accurate cognitive representation
increases performance by acting as a dynamic capability. The specification of the regression
model is as follows:
yit = (’x it)
(2)
where yit is a measure of firm performance, and the vector xit includes our demographic and firm
level variables including cross-functional experience. Subscripts indicate a founding year and 2digit SIC code. We include industry controls in all models and founding year controls in models
using revenues as the dependent variable. When using exits (acquisition or IPO) as the dependent
variable we worry about right-side censoring and use a Cox hazard rate model (Cox, 1972) to
take the timing of events into account. In these models, firms begin to be “at risk” of an exit at
24
the time of their founding (thus we do not control for founding year). Looking first across the
columns at the controls, we find that a larger cofounding team is significantly associated with
higher revenues and VC funding is associated with significantly greater likelihood of exit. Teams
with greater functional diversity were also more likely to experience an exit. Firm age, Master’s
degree and log initial capital were also correlated with higher performance. In Model (4-1),
cross-functional experience is significantly associated with higher revenues and a longer lag
from the cross-functional experience to the firm founding is associated with significantly lower
revenues. Model (4-2) also shows that cross-functional experience is positively and significantly
related to the likelihood that the current firm has an exit. The coefficient on log lag is below one,
indicating a lower probability of exit with a longer lag, however it does not reach statistical
significance. Model (4-3) and (4-4) show that firms are significantly higher performing when the
founder has experienced an acquisition with a prior firm, but not IPO (there were only 84 prior
IPOs linked to 72 different founders). Again, we find that a longer lag is associated with lower
revenues. In models (4-5) and (4-6) the results show that starting a new firm in the same U.S.
state associated with higher revenues. Since communication falls off with distance (Allen, 1977),
this provides evidence that some of the effect may be operating through social networks.5
Looking at models (4-7) and (4-8), we find that a prior cross-functional experience in the same
SIC as a measure of similarity of industry context is associated with significantly higher revenues
(but not higher likelihood of exit). In models (4-9) and (4-10) we do not find that Master’s
degree holders benefit significantly more from cross-functional experience.
-----------------------------Insert Table 3 and 4 about here
------------------------------Effects of a Major Industry Disruption
5
The total number of founders shifting states between the current startup and the prior founding is 172 and the
number that we know remained in the same state is 382.
25
If more accurate cognitive representations of the industry are the mechanism driving our
results, then after a change significant enough to disrupt many causal relationships in an industry
then we should see that the benefits are lost. We use the major disruption to the software industry
of the internet boom during the late 1990s to test this idea. Narrowing the sample to only
software firms, Table 5 tests whether the effect of cross-functional experience remains after the
internet boom. First, model (5-1) narrows the sample further to only individuals with at least one
cross-functional experience and breaks the lag into quantiles. We find that individuals with a
shorter lag had a significantly higher likelihood of exit while this result reversed for those with
the longest lag, further supporting hypothesis 5. Models (5-2) and (5-3) include an interaction
term between a post-1997 dummy variable and cross-functional experience. The negative
coefficient indicates that after the dotcom boom, cross-functional experience was associated with
lower performance. The coefficient is significant (p<0.05) when exited is the dependent variable.
Models (5-4) and (5-5) use the subsample of software firms before the dotcom boom and show
that cross-functional experience is associated with significantly higher performance. However,
this result goes away when models (5-6) and (5-7) use the subsample of firms founded after the
dotcom boom. These results are consistent with the mechanism of cross-functional experience
leading to more accurate cognitive representations, which are disrupted by a major industry shift.
------------------------------Insert Table 5 about here
------------------------------The real effects implied by these results are strong. The results imply that for each crossfunctional experience, revenues increase by 54.3 percent on average and the probability of an
acquistion or IPO increases by 49.3 percent. A successful cross functional experience (in the
form of an acquistion) similarly increases revenues by 53.6 percent and the likelihood of exit by
44.9 percent. A prior experience in the same industry increases revenues by over six times on
26
average and the likelihood of exit by 13.8 percent (though the latter coefficient does not reach
statistical significance). For a one standard deviation increase in the time since the crossfunctional experience, we expect a 9.2 percent decrease in expected revenues. In the software
industry, a lag in the highest quartile represents a firm over six times less likely to exit. A lag in
the second quartile represents a 145 percent increase in the likelihood of exit. In the software
industry after the dotcom boom, cross-functional experience is associated with a 221 percent
lower likelihood of exit and 35.2 percent lower revenues. However, before the dotcom
disruption, cross-functional experience increased expected revenues by 367 percent on average
and a prior acquistion increased revenues by over four times.
Robustness and Limitations
Controls for Individual Effects. Time-invariant individual level differences may be
correlated both with the likelihood of founding additional firms and with performance. An
alternative explanation for why performance might appear to improve with prior cross-functional
experience is that those who choose to start a second firm have higher skill levels than those who
choose to start only a single firm (Gompers, Kovner, Lerner, & Scharfstein, 2009). If on average
those who start multiple firms are also more persistent or more talented than those who start only
one firm, then we would also observe average performance improvements as lower skill
individuals return to wage employment. Chen (2009) finds evidence for both selection on innate
ability and learning by doing. The evidence for learning by doing is consistent with our account
and provides reassurance that selection alone does not drive the results. However we provide a
more specific mechanism at work for the learning by doing effect. We use the panel structure of
data on firm foundings over time to shed light on the alternative explanation of a selection effect
rather than a treatment effect. Data on entire founding histories allows the use of individual fixed
27
effects for those who have founded more than one firm.6 We exploit the panel structure of the
data, which includes observations of multiple firm foundings for many individuals, to implement
a regression including individual fixed effects to control for time-invariant factors from the
individual influencing performance. Results available from the authors drop the unchanging
individual characteristics for education and instead exploit the multiple observations on
individuals to include individual fixed effects that capture time-invariant differences in
individuals which may include higher underlying skill, persistence, family wealth, or preferences
for variety, all of which are likely confounding the earlier estimates. The main results remain
consistent in these models. We also re-ran the results conditioning on at least one firm founding.
The results remained significant, reassuring us that underlying skill or persistence are not driving
the results. If there is an improvement in the accuracy of cognitive maps leading to competitive
advantages, then conditioning on high persistence (more than one firm founding) as expected, we
continue to observe higher performance with additional cross-functional experience.
Starting a firm in a recessionary market can reasonably be expected to be a different
context than starting a firm during a boom time. As a further robustness check, (available from
the authors) we show results are consistent with the idea that representations of the industry built
during unusual economic conditions may be less helpful when the environment shifts to a more
stable condition. The results show that the subsequent firms have lower revenues if the first firm
was started during a recession (using the National Bureau of Economic Research index).
There are a few alternative accounts worth noting. Empirical work has begun to explore
learning from rare events with intriguing, yet mixed results (Kim & Miner, 2007; Zollo & Singh,
6
The individual fixed effects allow us to rule out the selection effect (time-invariant individual
characteristics) in going from fewer episodes of treatment to more, but not in going from no treatment to
treatment (no cross-functional experience to one prior founding). A Heckman selection model might help
with the latter concern, however we lack the requisite exclusion restriction (something correlated with
entrepreneurship but not with performance) required for this identification approach.
28
2004). Some have suggested that learning rates may be higher in slightly heterogeneous settings
(Haunschild & Sullivan, 2002). However, it seems unlikely that the results are explained entirely
by individuals learning the generic set of skills required for founding a firm. If this were the case,
then we would expect those skills to transfer to different industries and across states. We would
also expect that conditioning on one founding experience, individuals would have picked up
most of these entrepreneurship-specific skills and would benefit less from subsequent
experiences. Gompers, Kovner, Lerner, and Scharfstein (2009) argue that a large component of
success in entrepreneurship and venture capital can be attributed to skill rather than luck and
show that entrepreneurs with a track record of success are more likely to succeed than first time
entrepreneurs. However, the Gompers et al. sample is limited to founders who received venture
capital financing. Thus the authors lack data on the much larger proportion of prior foundings
that were not VC funded. Furthermore, many more successful start-ups undergo acquisition
rather than IPO as opportunities to go public vary with the economic environment and by
industry even more than do the opportunities to be acquired. Therefore, the Gompers et al.
(2009) analysis may be missing many actual prior successes which would tend to bias their
estimates. A skill-based story where skill is constant over time requires explanation for why
revenues appear to continue to increase with the number of prior cross-functional experiences
(successful or not) even when conditioning on at least one prior start-up. However, we cannot
eliminate the possibility that more accurate representations result from failure, yet other
mechanisms such as a tarnished reputation affect performance via impact upon potential recruits,
financiers and even suppliers and customers. Our data do not come from a random sample from
the entire population. Nonetheless, the fact that all the respondents are alumni of a prestigious
research- and technology-based university reduces the concern that there are large differences in
29
wealth, skill, or initial human capital.
Conclusion
We extend the microfoundations of strategy (Gavetti, 2005) and seek to answer the
question: When organizations are putting together new bundles of routines and capabilities, what
determines who puts together the more valuable and difficult to imitate bundles? Specifically
this paper is interested in whether firms gain a competitive advantage through dynamic
capabilities in the form of improved cognitive representations that the founders acquired via
prior work experience? The results support the main thesis of the paper that the answer is yes.
We suggest a novel mechanism through which dynamic capability arises from improved
cognitive maps of the competitive landscape. We examine the micro-foundations of resource and
capability development by theorizing how cross-functional experience results in variation in
cognition due to the well-known psychological phenomenon of partition dependence. We extend
and build on prior research that has suggested that differences in managerial cognition lead to
variation in strategic decisions (Adner & Helfat, 2003; Porac, Thomas, & Baden-Fuller, 1989;
Tripsas & Gavetti, 2000). Our primary proposition, Hypothesis 1, that individuals with crossfunctional experience will build more accurate cognitive representations, resulting in assembling
more valuable combinations of resources/capabilities was supported. Our analysis shows both
environmental and individual-level contingencies when improved cognitive representations are
expected to be of greater importance for making predictions on trends and the future value of
resource positions. Hypothesis 2a was also supported in that those remaining in the same
industry context experienced higher performance. Hypothesis 2b was that after a significant
industry disruption, cognitive representations built up through prior cross-functional experience
would become a liability. The results show that after the major industry disruption of the dotcom
30
boom, the impact of cross-functional experience became negative and significant as individuals
who had built up cognitive representations before the internet boom mistakenly attempted to
apply them after the disruption to the industry. Similarly, when we separately ran the analysis on
the pre-dotcom boom time period and post-dotcom time period, the coefficients on crossfunctional experience are positive and significant for the pre- time period and are significantly
lower (indistinguishable from zero) for the post-dotcom time period. The data tended to support
hypothesis 3, that cross-functional experience leads those who experienced success to higher
performance.7 After initial results in Table 3 showed that more highly educated individuals
appeared to benefit more, once additional controls were introduced, we failed to continue to find
support for hypothesis 4. The results appear to support the idea that a more recent crossfunctional experience is more useful and that the benefit fades with time, supporting hypothesis
5. This result supports the notion that for predicting trends, a more recent cross-functional
experience is quite important. The overall pattern of results, under a number of different
specifications, appears to provide robust evidence consistent with an account where experience
with a broader set of responsibilities and functional decision rights leads to more balanced
cognitive representations to guide strategic decisions.
This study extends and challenges existing work on the resource-based view and
particularly on the ability to restructuring and identify more valuable assets and capability
bundles. Prior work on the sources of performance-enhancing resources and capabilities has
focused on the inheritance of search routines or superior information. While this moves our
understanding forward, we still lack a psychological foundation for identifying where differences
arise in cognition or beliefs about the future value of routines, resources or capabilities. Rather
7
Prior IPOs or acquisitions may be viewed as successes, and are by many other authors, though this
largely depends on the valuations achieved.
31
than asking what forces constrain the strategic options under consideration or factors that lead to
greater homogeneity and mimicry in strategic choices, we examine how certain managers and
firms uniquely make predictions leading to differentiation and higher performance. We build on
work showing the strategic importance of framing and individual level cognition (Kaplan &
Tripsas, 2008; Tripsas & Gavetti, 2000). We examine a different cognitive spillover mechanism:
where individuals appear to transfer to a subsequent firm more accurate representations as a
result of the prior founding experience. We build on the theoretical framework for the
psychological foundations of capabilities’ development laid out around cognition and the
integration of knowledge (Gavetti, 2005; Grant, 1996) by extending the theory to propose
mechanisms by which differences in cognition emerge around variation in work experiences.
Similarly, we provide previously missing theoretical account for when the managers of firms
notice trends and change the bundles of routines and capabilities they hold and for how they
know in what directions to change them.
Extending earlier efforts to disaggregate the influence on business-unit profits of
industry, corporate-parent and market share effects, scholars have examined the influence of
firm-level effects (Schmalensee, 1985). Examining lower levels of analysis shows that industrylevel effects are approximately half as important as business effects in determining business-unit
profits (McGahan & Porter, 1997; Rumelt, 1991). Yet, with the exception of work on top
management teams and entrepreneurship, much less work has looked at the influence of even
lower levels of analysis (including individuals) on performance (Higgins & Gulati, 2006). Why
do some firms outperform others even when in the same industry? Our contribution is to use
psychological foundations to show how cognitive representations specific to the competitive
32
landscape embedded in individuals can function as a difficult to imitate dynamic capability,
guiding firms to build competitive advantages.
Much prior work has focused on the content of cognitive representations, but
significantly less is known about their origins and the usefulness of certain types of
representations (Walsh, 1995). We provide evidence that certain types of work experience can
lead to representations that have a higher likelihood of breaking out of the strategic status quo
and predicting the development of resources leading to greater performance. However, we also
show the downside of relying on these models when a very significant industry change occurs. In
a very fluid phase in the industry, cognitive flexibility appears to be more important (Furr, 2009).
We still have unanswered questions for future research about the relationship between the
content of a cognitive representation and the information environment it represents (Walsh,
1995). Others have shown that there are differences between accuracy and the usefulness of
cognitive representations when simplifying and screening out unnecessary information may be of
greater importance (Starbuck & Milliken, 1988).
The findings also have implications for the entrepreneurship literature. Our account
provides a complementary theory of why managers and entrepreneurs may leave and go outside
of their firms to work for others or to start their own ventures. Most organizations tend to
become more rigid over time. If individuals within the same firm can develop substantially
different cognitive representations of the competitive landscape due to differences in types of
work experience and the accumulation of psychological biases, then individuals can begin to
have fundamental disagreements with the organization. Furthermore, due to the inertia and
rigidity of the existing bundles of capabilities, it becomes increasingly difficult to move the
organization to a new set of resources and capabilities that individuals perceive will provide
33
competitive advantage according to their representations of the competitive landscape. Some
individuals leave for new organizations due to increased rigidity in bundles combined with
cognitive maps drifting away from the increasingly small set of recombinations possible within
the firm or from an inability to convince others to select their bundles of capabilities. Differences
in the accuracy of cognitive representations, particularly concerning trends and shifts in the
competitive landscape may also be a reason why individuals voluntarily leave firms. Individuals
may choose to move to other firms or start their own new ventures when disagreements arise,
when they see opportunities others do not, or when the existing firm seems incapable of
exploiting quickly due to inertia or loss of plasticity (Klepper, 2007).
In conclusion, we have developed and tested a model that links psychological theory to
dynamic capabilities and heterogeneity in firm performance. Variation in career experiences
leads to variation in the extent of known psychological biases such as partitioning dependence,
and then variation in the extent of these biases results in differences in cognitive representations
that function as a dynamic capability providing a map to future bundles of resources that will
provide a performance advantage.
TABLE 1 Descriptive statistics
VARIABLE
Log revenues
Acquired
Exited
Cross Functional Experience
Prior Acquisitions
Prior IPOs
Same SIC
Different SIC
Same State
Diff State
Log(lag)
Age Founded
Bachelor's degree
Master's degree
Firm Age
Num. Cofounders
VC
Log(initial capital)
Functional Diversity
Year founded
Industry sector
DEFINITION
Mean
Std. Dev.
Min
Max
Natural log of the revenues for the most recent fiscal year
14.05
3.08
0.03
21.66
= 1 if the venture was acquired
0.19
0.39
0
1
= 1 if venture was acquired or had an initial public offering
0.28
0.45
0
1
= 1 if the individual had founded a prior firm
1.61
1.30
0
1
Count of the number of prior founded firms resulting in an
acquisition
0.13
0.42
0
3
Count of the number of prior founded firms resulting in an
IPO
0.04
0.23
0
3
Count of the number of prior founded firms in the same 2
digit SIC code
0.02
0.14
0
2
Count of the number of prior founded firms in a different
2 digit SIC code
0.02
0.18
0
3
Count of the number of prior founded firms in the same U.S.
state
0.38
0.90
0
8
Count of the number of prior founded firms in a different
U.S. state
0.23
0.79
0
7
= years from closing the previous venture to founding the
next
12.89
9.63
0
51
Age of the founder/respondent
39.65
10.59
18
83
= 1 if the founder/respondent’s highest degree is a Bachelor’s
0.43
0.49
0
1
= 1 if the founder/respondent’s highest degree is a Master’s
0.42
0.49
0
1
= number of years in operation
14.34
11.30
0
74
Count of the number of cofounders
2.02
1.79
0
4
= 1 if the firm received venture capital (VC) funding
0.25
0.43
0
1
Natural log of the amount of capital raised in the first year
11.91
2.72
0.28
21.02
Count of the number of functions (engineering, finance,
sales and marketing) present on the founding team (max 3)
1.22
0.46
1
3
Year in which the venture was founded
1985
10.88
1950
2003
Dummy variables for industries: aerospace, machinery and other manufacturing, biomedical and
chemicals, consumer products, electronics and telecommunications, energy, finance and consulting,
publishing, software, and other services
TABLE 2
Pair-wise correlations
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Cross-functional
experience
Prior acquired
Prior IPO
Same SIC
Diff. SIC
Log(lag)
Num. cofound
Founder Age
Bachelor’s
Master’s
Firm Age
Log(initial cap)
VC
Func diversity
Found year
1
0.49
0.26
0.62
0.70
-0.21
0.15
0.22
-0.02
0.04
-0.14
0.09
0.03
0.06
0.22
2
1
0.31
0.42
0.52
-0.09
0.11
0.11
-0.02
0.02
-0.01
0.14
0.05
0.04
0.12
3
1
0.33
0.35
-0.05
0.06
0.12
-0.03
0.04
-0.05
0.15
0.07
0.03
0.06
4
1
-0.02
-0.11
0.05
0.05
-0.02
0.04
-0.08
0.12
0.15
-0.04
0.05
5
1
-0.06
0.09
0.08
-0.06
0.07
-0.02
0.16
0.16
0.06
0.06
6
1
-0.08
0.48
0.12
-0.14
-0.01
0.00
-0.05
-0.03
-0.02
7
1
-0.08
-0.06
0.05
-0.01
0.25
0.20
0.56
0.28
8
1
-0.01
-0.05
-0.16
-0.03
-0.09
-0.05
0.19
9
1
-0.72
0.00
-0.06
-0.05
-0.05
-0.01
10
1
0.00
0.05
0.05
0.06
0.01
11
1
-0.04
-0.13
0.00
-0.88
12
1
0.54
0.13
0.06
13
1
0.11
0.10
14
1
0.14
15
1
Panel A
Revenues
Exits
Panel B
Revenues
Exits
Panel C+
Revenues
Exits
Panel D
Revenues
Exits
Panel E
Revenues
Exits
Panel F
Revenues
Exits
Panel G
Revenues
Exits
TABLE 3
Univariate difference in means tests with and
without cross-functional experience (CFE)
No cross-func. experience
At least 1 cross-func. experience
(CFE)
13.957
14.219*
0.205
0.240**
No cross-func. exp. in same
Cross-func. exper. in same
industry
industry
12.095
15.907**
0.205
0.421***
No CFE before dotcom boom CFE before dotcom boom
14.614
15.227*
0.275
0.391*
No CFE after dotcom boom
CFE after dotcom boom
12.810
13.375
0.182
0.061**
Lag from CFE to founding Lag from CFE to founding
less than the median
greater than the median
14.155
14.273
0.212
0.262*
MA degree, no CFE
MA degree, at least 1 CFE
13.898
14.316*
0.195
0.279***
Bachelor’s degree, no CFE
Bachelor’s degree, at least 1 CFE
14.068
14.136
0.214
0.233
Statistical tests are with respect to the column to the left. ***, **, and * indicate statistical
significance at the 1%, 5%, and 10% levels, respectively; +Includes software firms only.
Ln(Rev)
(4-1)
Cross-func. exper.
0.434**
(0.206)
Pr(Acq)
Ln(Rev)
(4-2)
(4-3)
Pr(Acq)
(4-4)
TABLE 4
Ln(Rev)
(4-5)
Pr(Acq)
(4-6)
Ln(Rev)
(4-7)
Pr(Acq)
(4-8)
1.493**
(0.248)
Prior acquired
0.429**
(0.201)
0.357
(0.343)
Prior IPO
0.206**
(0.101)
0.058
(0.116)
Diff state
1.813***
(0.652)
0.721
(0.630)
Diff SIC
Func diversity
Industry F.E.
Founding Yr. F.E.
Constant
Observations
R-squared
-0.171**
(0.081)
0.948
(0.051)
-0.197***
(0.076)
0.922
(0.047)
-0.188**
(0.079)
0.910*
(0.047)
-0.107
(0.129)
0.914
(0.073)
0.251***
(0.084)
0.344
(0.292)
-0.059
(0.184)
Y
Y
8.708***
(2.658)
874
0.398
0.970
(0.062)
1.903***
(0.373)
1.314**
(0.179)
Y
N
0.259***
(0.084)
0.368
(0.292)
-0.058
(0.183)
Y
Y
8.841***
(2.656)
874
0.400
0.976
(0.062)
1.882***
(0.365)
1.346**
(0.184)
Y
N
0.259***
(0.084)
0.350
(0.293)
-0.083
(0.184)
Y
Y
8.696***
(2.664)
874
0.398
0.976
(0.062)
1.859***
(0.363)
1.286*
(0.175)
Y
N
0.352***
(0.107)
0.621*
(0.360)
0.018
(0.231)
Y
Y
8.814***
(2.871)
606
0.420
0.918
(0.072)
1.899***
(0.447)
1.581***
(0.255)
Y
N
1063
1.845***
(0.383)
0.417
(0.363)
-0.165**
(0.081)
0.624
(0.181)
0.938
(0.051)
0.258***
(0.085)
0.342
(0.292)
-0.063
(0.184)
Y
Y
8.837***
(2.660)
874
0.399
0.957
(0.061)
1.878***
(0.368)
1.319**
(0.180)
Y
N
1.138
(0.514)
1.230
(0.452)
Master’s*CFE
VC
(4-10)
0.253
(0.260)
1.019
(0.099)
1.082
(0.107)
Same SIC
Controls
Num. cofounders
(4-9)
Pr(Acq)
1.449***
(0.186)
1.161
(0.257)
Same state
Log(lag)
Ln(Rev)
1063
1063
729
1063
***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Hazard ratio and standard errors reported; 1063 firms, 234 events and
16,068 years at risk. Controls for the firm age, the highest degree earned (Master’s and Bachelor’s degrees) as well as the founder age and log initial capital were
included but the coefficients are not shown to save space.
38
TABLE 5
Major Industry Disruption
VARIABLES
Cross-functional Exper.
Pr(Exited)
Pr(Exited) Ln(Revenues) Ln(Revenues)
At least 1 CFE
Before Dotcom
(5-1)
(5-2)
(5-3)
(5-4)
0.407
1.333**
1.301*
(0.287)
(0.596)
(0.731)
Post-1997*Cross Func.
Experience
-1.214**
(0.522)
Prior IPO
Lag 50-75th quartile
Lag 75th+ quartile
Age founded
2.453**
(1.085)
1.633
(1.407)
-6.608***
(2.549)
-0.015
(0.038)
0.315
(0.259)
0.750***
(0.231)
-
Ln(Revenues)
After Dotcom
(5-6)
0.690
(0.532)
Ln(Revenues)
After Dotcom
(5-7)
-0.302
(0.840)
-0.456
(1.217)
1.470**
(0.713)
Prior acquired
Lag 25-50th quartile
Ln(Revenues)
Before Dotcom
(5-5)
1.288
(1.043)
0.396
(0.527)
-0.021*
-0.029
-0.081**
-0.091**
0.012
0.013
(0.012)
(0.022)
(0.036)
(0.038)
(0.029)
(0.029)
Num. cofounders
0.266***
0.383**
0.322
0.283
0.005
-0.071
(0.090)
(0.180)
(0.245)
(0.246)
(0.271)
(0.276)
Log(initial capital)
0.182***
0.214**
0.339**
0.306**
0.371***
0.367***
(0.049)
(0.087)
(0.127)
(0.131)
(0.129)
(0.130)
Post-1997
0.900
-7.316
N
N
Y
Y
(0.937)
(2.989)
Constant
-10.221**
-2.551**
16.370***
11.98***
13.27***
7.182***
7.854***
(4.232)
(1.170)
(2.767)
(3.569)
(3.752)
(1.998)
(2.052)
Founding Year F.E.
YES
YES
YES
YES
YES
YES
YES
R-squared
0.593
0.263
0.518
0.639
0.647
0.391
0.398
Observations
64
203
205
82
82
80
80
Only software firms founded after 1985 are included. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.
Model 3-3 contains additional year fixed effects for the most recent operating year that revenues were reported on.
39
References
Adner, R., & Helfat, C. 2003. Corporate effects and dynamic managerial capabilities. Strategic
Management Journal, 24: 1011-1025.
Agarwal, R., Echambadi, R., Franco, A. M., & Sarkar, M. B. 2004. Knowledge transfer through
inheritance: Spin-out generation, development, and survival. Academy of Management
Journal, 47(4): 501.
Anderson, M. C. 2003. Rethinking interference theory: Executive control and the mechanisms of
forgetting. Journal of Memory and Language, 49(4): 415-445.
Argote, L., Beckman, S. L., & Epple, D. 1990. The persistence and transfer of learning in
industrial settings. Management Science, 36(2): 140-154.
Argote, L., & Ingram, P. 2000. Knowledge transfer: A basis for competitive advantage in firms.
Organizational Behavior and Human Decision Processes, 82(1): 150-169.
Bantel, K. A. and Jackson, S. E. 1989. Top management and innovations in banking: Does the
composition of the top team make a difference. Strategic Management Journal, 10(1):
107-124.
Baron, J. N., Hannan, M. T., & Burton, M. D. 1999. Building the iron cage: Determinants of
managerial intensity in the early years of organizations. American Sociological Review,
64(4): 527-547.
Baron, J., Burton, M., & Hannan, M. 1999. Engineering bureaucracy: The genesis of formal
policies, positions, and structures in high-technology firms. Journal of Law, Economics,
and Organization, 15(1): 1-41.
Becker, M. H. 1970. Sociometric location and innovativeness: Reformulation and extension of
the diffusion model. American Sociological Review, 35(2): 267-282.
Beckman, C. M., & Burton, M. D. 2008. Founding the future: Path dependence in the evolution
of top management teams from founding to IPO. Organization Science, 19(1): 3-24.
Boeker, W. 1989. Strategic change: The effects of founding and history. Academy of
Management Journal, 32: 489-515.
Bunderson, J. S., & Sutcliffe, K. M. 2002. Comparing alternative conceptualizations of
functional diversity in management teams: Process and performance effects. The
Academy of Management Journal, 45(5): 875-893.
Burton, M. D., Sørensen, J. B., & Beckman, C. M. 2002. Coming from good stock: Career
histories and new venture formation. In M. Lounsbury & M. Ventresca (Ed.), Research
in the sociology of organizations: 229-262. Oxford, U.K.: Elsevier (JAI Press).
Chattopadhyay, P., Glick, W. H., Miller, C. C., & Huber, G. P. 1999. Determinants of executive
beliefs: Comparing functional conditioning and social influence. Strategic Management
Journal, 20: 763-789.
Chen, Jing. 2009. Selection and Serial Entrepreneurs. Working paper.
Cohen, M. D., & Bacdayan, P. 1994. Organizational routines are stored as procedural memory:
Evidence from a laboratory study. Organization Science, 5(4): 554-568.
40
Cyert, R., & March, J. 1963. A behavioral theory of the firm. Englewood Cliffs, NJ: PrenticeHall.
Das, T. K., & Teng, B. S. 1999. Cognitive biases and strategic decision processes. Journal of
Management Studies, 36: 757-778.
Dearborn, D. C., & Simon, H. A. 1958. Selective perception: A note on the departmental
identifications of executives. Sociometry, 21(2): 140-144.
Dimaggio, P., & Powell, W. 1983. The iron cage revisited: Institutional isomorphism and
collective rationality on organizational fields. American Sociological Review, 48: 147160.
Dubé-Rioux, L., & Russo, J. E. 1988. An availability bias in professional judgment. . J.
Behavioral Decision Making, 1: 223-237.
Finkelstein, S., & Haleblian, J. 2002. Understanding acquistion performance: The role of transfer
effects. Organization Science, 13(1): 36.
Fiol, C. M., & Huff, A. S. 1992. Maps for managers: Where are we? where do we go from here?
Journal of Management Studies, 29: 267-285.
Fischhoff, B., Slovic, P., & Lichtenstein, S. 1978. Fault trees: Sensitivity of estimated failure
probabilities to problem representation. Journal of Experimental Psychology, 4: 330344.
Fox, C. R., & Clemen, R. T. 2005. Subjective probability assessment in decision analysis:
Partition dependence and bias toward the ignorance prior. Management Science, 51(9):
1417-1432.
Furr, N. 2009. Cognitive Flexibility: The Adaptive Reality of Concrete Organization Change.
Working paper.
Gavetti, G. 2005. Cognition and hierarchy: Rethinking the microfoundations of capabilities'
development. Organization Science, 16(6): 599-617.
Gavetti, G., & Rivkin, J. W. 2007. On the origin of strategy: Action and cognition over time.
Organization Science, 18(3): 420-439.
Gavetti, G., & Levinthal, D. 2000. Looking forward and looking backward: Cognitive and
experiential search. Administrative Science Quarterly, 45(1): 113-137.
Geletkanycz, M. A. and Hambrick, D. C. 1997. The external ties of top executives: Implications
for strategic choice and performance. Administrative Science Quarterly, 42(4): 654-681.
Gompers, P. A., Kovner, A., Lerner, J., & Scharfstein, D. S. 2009. Persistence in performance:
Evidence from serial entrepreneurs. SSRN: http://ssrn.com/abstract=933932.
Grant, R. M. 1996. Prospering in dynamically-competitive environments: Organizational
capability as knowledge integration. Organization Science, 7(4): 375-387.
Gupta, A. K. 1984. Contingency linkages between strategy and general manager characteristics:
A conceptual examination. The Academy of Management Review, 9(3): 399-412.
Hambrick, D. C., & Mason, P. A. 1984. Upper echelons: The organization as a reflection of its
top managers. The Academy of Management Review, 9(2): 193-206.
41
Hannan, M. T., & Freeman, J. 1977. The population ecology of organizations. American
Journal of Sociology, 82: 929-964.
Haunschild, P. R., & Sullivan, B. N. 2002. Learning from complexity: Effects of prior accidents
and incidents on airlines' learning. Administrative Science Quarterly, 47(4): 609-643.
Henderson, R. M. 1995. Of life cycles real and imaginary: The unexpectedly long old age of
optical lithography. Research Policy, 24: 631-643.
Higgins, M. C., & Gulati, R. 2006. Stacking the deck: The effects of top management
backgrounds on investor decisions. Strategic Management Journal, 27(1): 1.
Hiller, N. J., & Hambrick, D. C. 2005. Conceptualizing executive hubris: The role of (hyper-)
core self-evaluations in strategic decision-making. Strategic Management Journal, 26:
297-319.
Hitt, M. A., & Tyler, B. B. 1991. Strategic decision models: Integrating different perspectives.
Strategic Management Journal, 12(5): 327-351.
Hsu, D. H., Roberts, E. B., & Eesley, C. E. 2007. Entrepreneurs from technology-based
universities: Evidence from MIT. Research Policy, 36: 768-788.
Kaplan, S., & Tripsas, M. 2008. Thinking about technology: Applying a cognitive lens to
technical change. Research Policy, 37(5): 790-805.
Kim, J., & Miner, A. 2007. Vicarious learning from the failures and near-failures of others:
Evidence from the U.S. commercial banking industry. Academy of Management
Journal, 50(3): 687-714.
Kimberly, J. R., & Evanisko, M. J. 1981. Organizational innovation: The influence of individual,
organizational, and contextual factors on hospital adoption of technological and
administrative innovations. The Academy of Management Journal, 24(4): 689-713.
Klepper, S. 2007. Disagreements, spinoffs, and the evolution of detroit as the capital of the U.S.
automobile industry. Management Science, 53(4): 616-631.
Levitt, B., & March, J. G. 1988. Organizational learning. Annual Review of Sociology, 14: 319340.
March, J. G., & Olsen, J. P. 1976. Ambiguity and choice in organizations. Boston, MA:
Scandinavian University Press.
March, J. G. 1991. Exploration and exploitation in organizational learning. Organization
Science, 2(1, Special Issue: Organizational Learning: Papers in Honor of (and by) James
G. March): 71-87.
Wiersema, M.F. & Bantel, K. A. 1992. Top management team demography and corporate
strategic change. The Academy of Management Journal, 35(1): 91-121.
McGahan, A. M., & Porter, M. E. 1997. How much does industry matter, really? 18(Summer
Special Issue): 15.
McGrath, R. G. 1999. Falling forward: Real options reasoning and entrepreneurial failure.
Academy of Management.The Academy of Management Review, 24(1): 13.
42
Ofir, C. 2000. Ease of recall vs. recalled evidence in judgment: Experts vs. laymen. Organ.
Behavior Human Decision Processes, 81: 28-42.
Peteraf, M. 1993. The cornerstone of competitive advantage: A resource based view. Strategic
Management Journal, 14: 179-191.
Pfeffer, J., Cummings, L. L., & Staw, B. M. 1983. Research in organizational behavior JAI.
Porac, J. F., Thomas, H., Wilson, F., Paton, D., & Kanfer, A. 1995. Rivalry and the industry
model of scottish knitwear producers. Administrative Science Quarterly, 40(2): 203-227.
Porac, J. F., & Thomas, H. 1990. Taxonomic mental models in competitor definition. Academy
of Management Review, 15: 224-240.
Porac, J. F., Thomas, H., & Baden-Fuller, C. 1989. Competitive groups as cognitive
communities: The case of scottish knitwear manufacturers. Journal of Management
Studies, 26: 397-416.
Roberts, E. B. 1991. Entrepreneurs in high technology: Lessons from MIT and beyond. New
York, NY: Oxford University Press.
Romanelli, E. 1991. The evolution of new organizational forms. Annual Review of Sociology,
17(1): 79-103.
Ruef, M., Aldrich, H. E., & Carter, N. M. 2003. The structure of founding teams: Homophily,
strong ties, and isolation among U.S. entrepreneurs. American Sociological Review, 68:
195-222.
Rumelt, R. 1991. How much does industry matter. Strategic Management Journal, 12(3): 167185.
Russo, J. E., & Kolzow, K. J. 1994. Where is the fault in fault trees? J. Experiment. Psych.:
Human Perception Performance, 20: 17-32.
Schmalensee, R. 1985. Do markets differ much? American Economic Review, 75: 341-351.
Schroder, H. M., Driver, M. J., & Steufert, S. 1967. Human information processing. New York:
McGraw-Hill.
Schwenk, C. R. 1986. Information, cognitive biases, and commitment to a course of action. The
Academy of Management Review, 11(2): 298-310.
Simon, H. A. 1991. Bounded rationality and organizational learning. Organization Science, 2(1,
Special Issue: Organizational Learning: Papers in Honor of (and by) James G. March):
125-134.
Starbuck WH, H. B. 2001. How organizations learn from success and failure. In M. Dierkes, A.
Berthoin Antal, J. Child, & I. Nonaka (Ed.), Handbook of organizational learning and
knowledge: 327-350. New York: Oxford University Press.
Starbuck, W. H., & Milliken, F. J. 1988. Executives' perceptual filters: What they notice and
how they make sense. In D. C. Hambrick (Ed.), The executive effect: Concepts and
methods for studying top managers: 35-65. Greenwich, CT: JAI Press.
43
Staw, B. M., & Ross, J. 1980. Commitment in an experimenting society:A study of the
attribution of leadership from administrative scenarios. Journal of Applied Psychology,
65: 249-260.
Stevens, J. M., Beyer, J. M., & Harrison M. Trice. 1978. Assessing personal, role, and
organizational predictors of managerial commitment. The Academy of Management
Journal, 21(3): 380-396.
Stinchcombe, A. L. 1965. Social structure of organizations. In J. March (Ed.), Handbook of
organizations: 142-193. Chicago, IL: Rand-McNally.
Taylor, R. N. 1975. Age and experience as determinants of managerial information processing
and decision making performance. The Academy of Management Journal, 18(1): 74-81.
Teece, D., Pisano, G., & Shuen, A. 1997. Dynamic capabilities and strategic management.
Strategic Management Journal, 18(7): 509-533.
Tripsas, M. 2009. Technology, identity, and inertia through the lens of "the digital photography
company". Organization Science, 20(2): 441-460.
Tripsas, M., & Gavetti, G. 2000. Capabilities, cognition, and inertia: Evidence from digital
imaging. Strategic Management Journal, 21(10/11, Special Issue: The Evolution of
Firm Capabilities): 1147-1161.
Tversky, A., & Kahneman, D. 1973. Availability: A heuristic for judging frequency and
probability. Cognitive Psychology, 4: 207-232.
Tversky, A., & Kahneman, D. 1986. Rational choice and the framing of decisions. Journal of
Business, 59: S251-S278.
Walsh, J. P. 1995. Managerial and organizational cognition: Notes from a trip down memory
lane. Organization Science, 6: 280-321.
Walsh, J. P. 1988. Selectivity and selective perception: An investigation of managers' belief
structures and information processing. The Academy of Management Journal, 31(4):
873-896.
Weick, K. E. 1995. Sensemaking in organizations. Thousand Oaks, CA: Sage.
Wernerfelt, B. 1984. A resource-based view of the firm. Strategic Management Journal, 5(2):
171.
Winter, S. G. 2003. Understanding dynamic capabilities. Strategic Management Journal, 24:
991.
Zander, U., & Kogut, B. 1995. Knowledge and the speed of the transfer and imitation of
organizational capabilities: An empirical test. Organization Science, 6(1, Focused Issue:
European Perspective on Organization Theory): 76-92.
Zollo, M., & Singh, H. 2004. Deliberate learning in corporate acquisitions: Post-acquisition
strategies and integration capability in U.S. bank mergers. Strategic Management
Journal, 25(13): 1233.
Zott, C. 2003. Dynamic capabilities and the emergence of intra-industry differential firm
performance: Insights from a simulation study. Strategic Management Journal, 24: 97125.