the survival chances of new firms in new industries
TRANSCRIPT
The Survival Chances of New Firms in New Industries: Legitimacy Vacuum,
Structural Imprinting, and the First-Mover Disadvantage
Stanislav D. Dobrev and Aleksios Gotsopoulos
University of Chicago GSB, 5807 S. Woodlawn Ave. Chicago, IL 60637
Tel. 773.834.5965
The Survival Chances of New Firms in New Industries: Legitimacy Vacuum,
Structural Imprinting, and the First-Mover Disadvantage
(abstract)
Unfavorable conditions at founding may result in consistently lower survival chances for new
firms. We focus on the effect of industry entry in the early years of an industry when clarity about
the form and function of a new category of firms is lacking. This population-level legitimacy
vacuum effect is persistent and adversely affects new entrants in an emerging industry for the
duration of their lifetimes. These ideas are integrated with received theory in organizational
ecology and help to extend broader notions of structural imprinting in organization theory. They
also help to explicate some of the inconclusiveness surrounding the metaphor of first-mover
advantages. Results from the analysis of survival rates of firms in the U.S. auto industry from its
inception in 1885 until 1981 support our conjectures and demonstrate the value of employing an
ecological approach to studying the role of history in organizational evolution.
0
INTRODUCTION
It is widely acknowledged that new ventures, for a variety of reasons—insufficient
capital, poor design, inexperienced management, etc...—have much higher failure rates than more
established firms. For a relatively small and clearly distinct set of entrepreneurial firms, however,
the primary reason behind early failure is quite different. At least anecdotally, firms who enter an
industry in its very early stages are often said to have been “ahead of their time,” or “too fast to
market” (Tellis and Golder, 1996), and to have suffered a “first-mover disadvantage” (Lieberman
and Montgomery, 1998). One of our two objectives in this paper is to build theory which
transcends metaphors and case-based insight and offers an empirically testable mechanism to
explain the failure of new firms in new industries. We argue that organizations are precariously
close to failure due to their inability to evoke institutionalized claims in pursuit of resources or to
invest resources optimally when they operate under conditions of legitimacy vacuum—conditions
in which clarity about the form and function of these organizations collectively as a socially
familiar categorical type or a blueprint is lacking.
Our second objective is to extend arguments of structural imprinting from organizational
sociology (Stinchcombe, 1965) by theorizing the causal forces that put new firms in new
industries at a persistent disadvantage relative to later entrants. We claim that firms that belong to
the first cohort of entrants in an emerging market have consistently worse survival chances than
firms from later cohorts and that this difference persists throughout their lifetimes. Initial
exposure to legitimacy vacuum is imprinted in the structure and processes of new firms and its
deleterious effect on survival is generally irreversible.
In pursuing our two research objectives, we advocate and employ an ecological approach
for two reasons. First, the density delay theory (Carroll and Hannan, 1989) in organizational
ecology represents one systematic way for modeling the persistent effect of founding conditions
on organizations’ life chances. This theory specifies two mechanisms, resource scarcity and tight
niche packing, that explain how competitive crowding at the time of founding (high population
density) leads to persistently higher death rates for organizations founded under such conditions.
We extend the density delay argument by proposing that even if resources are not scarce or
highly contested, securing access to them is jeopardized under conditions of legitimacy vacuum
when constitutive paradigms for linking organizational means and ends are non-existent.
Moreover, we claim that even if early industry entrants are able to acquire resources, their
chances of investing them in a manner consistent with what would later prove to be the successful
track record of an industry is contemporaneously impossible at the industry’s embryonic stage.
So we claim that the imprinting of the legitimacy vacuum effect results from both insufficient
resources and the uncertainty about their appropriate long-term investment.
Second, employing an ecological framework for the analysis of imprinting effects is
useful because it forces a disciplined analytical framework on an intuitively appealing but largely
underdeveloped theoretical argument. The notion of structural imprinting is popular in
organization theory but it has rarely been challenged on theoretical grounds or subjected to
rigorous empirical scrutiny. At least partly, the problem is excess generality. It is not known, for
example, what exactly accounts for the imprinting process in the sense that direct mechanisms for
explaining what features of social structure get imprinted into which organizations are scant.
Surely, no one has argued that all features of the social structure get imprinted into every new
organization founded at any given time. The ecological approach mandates that the imprinting
effects are confined to a specific set of environmental conditions (e.g., an emergent industry), and
that they neither apply similarly to all social actors exposed to these environmental conditions
(e.g., investors, producers, suppliers, analysts, etc…), nor do they apply differently to each social
actor when actors share homogeneity in resource dependence (e.g., members of an organizational
2
population). Instead, the ecological approach to imprinting underscores how resource-overlapping
entities acquire “particular social object preferences” (Immelman, 1975: 21) by considering the
time-constraint on the occurrence of imprinting (a brief sensitivity period after birth), the fact that
its effects are irreversible, and that imprinting is inherently a supra-individual, population-level
phenomenon (Lorenz 1935). Consistent with this approach, we consider imprinting to occur at the
time of a firm’s founding, to influence its survival chances significantly, and to affect similarly all
organizations in a population exposed to the same initial conditions in their environment.
In the next section, we develop an argument for the detrimental effect of legitimacy
vacuum at the time of organizational founding on survival chances and integrate it with the
original predictions of density delay theory. In the methods section, we justify the use of
secondary data and offer a modeling strategy that takes into account the specificity of the
historical contexts and the predictions of arguments that intuitively contradict our theory. In the
results section we interpret our findings and evaluate them in light of the first-mover metaphor
which they overtly contradict. We conclude by stating the advantages of an ecological approach
utilizing simple and generalizable models with flexible specifications to studying the impact of
historical contingencies on organizational evolution.
THEORY
Legitimacy Vacuum and Structural Imprinting
Holding the level of available resources constant, what are the factors that determine
whether new organizations can mobilize these resources? In addressing this question,
Stinchcombe (1965: 161) surmised that “…an organization must have an elite structure of such a
form and character that those people in the society who control resources essential to the
organization’s success will be satisfied that their interests are represented in the goal-setting
3
apparatus of the enterprise…” In short, an organization has to meet the expectations of external
constituents not just with respect to its goals and objectives but also in terms of the ways in which
resources will be deployed to pursue and accomplish these goals (Pólos, Hannan, and Carroll,
2002). The social consent that nascent entities must elicit to secure resources hinges on the
credibility of the claims that they make and to demonstrate such credibility, organizations
develop structures in support of the legitimacy of their means and ends. Given their audience’s
limited attention and likely varied opportunity to distribute resources, claims of legitimacy are
easier to make within a set of comparability with other similar social actors (White, 2001).
Legitimacy thus emerges at the level of categorically similar organizations and when
accomplished as a collective good, it leads to the emergence of a new form whose blueprint
serves as a model of organizing for other nascent entities. These blueprints attain a taken-for-
granted status with both entrepreneurs and gatekeepers of resources and often become inveigled
in institutional myths of rationality only loosely tied to objective features of organizational
structure (Meyer and Rowan, 1977).
Recasting the vast literature on constitutive legitimacy in organizational sociology for the
purpose of our theory raises two questions. First, given the importance of gaining an institutional
standing of taken-for-grantedness for all organizations in a population, is it useful to emphasize
its particular importance for new organizations? Second, if it is, should the particular liability of
new organizations to lack of legitimacy not be explained with recourse to received theory on
liability of newness (Stinchcombe, 1965; Hannan and Freeman 1989; Hannan 1998)? Our
answers to these questions follow.
There is no shortage of argument and evidence that link the improved survival chances of
all organizations bearing the same form to the presence of established normative expectations of
what these organizations do, how they do it, and what social boundaries they must observe. When
4
such expectations are lacking, survival chances are low. As legitimacy builds, access to resources
improves, and so do the survival chances of all organizations. But organizations founded in
legitimacy vacuum continue to be at a disadvantage relative to their peers because exposure to a
weak institutional environment during the nascent stage imprints in the highly resilient initial
structures that new firms develop. At the time of founding these structures are especially
malleable and influenced by external conditions. And when external conditions are unfavorable,
as the imprinting argument goes, the effect is irreversible.
Of course, the hazardous position of young organizations, including lack of legitimacy, is
often understood with recourse to their own age as in Stinchcombe’s (1965) liability of newness
story. New organizations find it hard to substantiate their claims for resources because external
constituents lack evidence in the form of past experience based on which to evaluate the
credibility of their claims. New organizations need time to develop routinized activities, to build
trust between members and relationships with outsiders, to learn how to coordinate between tasks
and how to monitor and incentivize members. For all these reasons, young organizations improve
their survival chances as they mature. And those that survive eventually become better fit than the
at large pool of potential entrants that have not yet faced the test of trial-by-fire (Swaminathan,
1996; Lomi and Larsen, 1998). In contrast to the liability of newness story, we surmise that when
lack of legitimacy is understood as a property of the environment, rather than of individual
organizations (an approach defined by the ecological framework of analysis), it is likely to leave
a lasting mark on entities founded under such conditions. A new organization that represents a
widely taken-for-granted form can rely on the form’s constitutive legitimacy to make claims
about its own right of passage, an ability that helps it to outgrow its own liability of newness. By
contrast, a new firm founded in an institutionally unrecognized social space is overwhelmed by
the compounded disadvantage of its own age and of the illegitimacy of its social position. In our
5
view, recovering from such a disadvantage is unlikely. While liability of newness is transitory
and passes with age, the imprinting effect of legitimacy vacuum is permanent. We suggest three
reasons to substantiate this claim—organizational founders’ finite resources, obsolescence, and
uncertainty—all of which relate the imprinting effect of legitimacy vacuum to low organizational
survival chances.
Ecological Mechanisms for Imprinting
Our goal in this section is to develop mechanisms that can substantively account for the
imprinting effect of legitimacy vacuum on survival chances of organizations in an industry.
Density delay theory (Carroll and Hannan, 1989) argues that firms founded in crowded markets
suffer persistently reduced competitive fitness for two reasons. First, organizations founded under
conditions of crowding when available inputs are highly contested tend to develop structures
designed to take advantage of insufficient resources that emphasize the ability to make do under
conditions of deprivation. By contrast, organizations experiencing resource affluence at founding
are able to devote attention and resources to organizing for growth, a likely advantage in latter
stage organizational experience. Initial exposure to scarce resources thus generates a lasting
liability that is difficult to overcome because it engrains in the evolving structure of the fledgling
entity. The second mechanism behind the density delay argument is that when the resource space
is tightly packed at the time of an organization’s initial entry, its choice of location is inflexible.
Incumbents occupy the resource rich areas of the market and newcomers are invariably pushed to
peripheral locations with sparse resources. Selection of initial market position is consequential
because alignment with local conditions where resources are embedded imprints into the nascent
entity’s organization. The persistence of structural arrangements made in the early period of
existence thus confines an organization to its initial location or likely exacerbates survival
6
chances when position moves are attempted. As long as resource distributions within an industry
remain stable over time, positional inertia disadvantages firms whose initial positions (forged by
the nature of the competitive environment faced at founding) are on the periphery.
Both arguments in density delay theory deal with a firm’s early experience in acquiring
resources and specifically, with the deleterious effect of crowding which limits resource
accessibility. We think that there is another reason—the illegitimacy of deploying resources in a
manner that is socially unrecognized. For example, perceptions of usury as deeply unethical
complicated the early rise of banking in Renaissance Florence (Parks, 2005) while in the U.S., the
inexperienced public’s irrational ambivalence about electric batteries stalled early
experimentation with electric cars in the nascent automobile industry (Kirsch, 2000). While both
industries eventually prospered, few of the early entrants shared in this prosperity. In short, the
idea of legitimacy vacuum summarizes features of the environment not merely reducible to
variation in resource levels and competitive crowding. In fact, it highlights the important
consequences of social experience in a firm’s formative period (i.e., the strife for acceptance and
recognition), an argument that perhaps most closely reflects Stinchcombe’s (1965) original claim
of social structural imprinting.
Founder roles and imprinting. Consistent with the resource scarcity argument in density
delay theory, we expect that insufficient resources adversely affect the design of a nascent
organization. But the mechanism we articulate is different—developing an organizational
structure inapt to support future growth results not only from stiff resource competition but from
the need for founders to invest a disproportionately higher amount of time and effort in acquiring
resources under conditions of legitimacy vacuum, a requirement which disrupts the balance
between the internal and external organizational roles entailed in the founder’s position.
7
It is often argued that even when organizations are brand new, they rarely constitute a
tabula rasa (Mitchell, 1989; Levinthal and Myatt, 1994). Type and duration of prior work
experience, position in social networks, ability and pertinent expertise and other social and human
capital characteristics of founders imbue a nascent organization with resources and opportunities.
In fact, perhaps the most developed line of theory on organizational imprinting to date examines
the influence of founders on the structure of their organizations (Kimberly, 1979, Freeman, 1986;
Boeker, 1988; Baron, Hannan, and Burton, 2001; Phillips, 2002). This research shows that
founders often have a clear idea of how they want to shape and model their creation, and that,
when instilled in the early years, such initial blueprints tend to persist even after the founder’s
departure (Hannan, Burton, and Baron, 1996). Research on the organizational roles of founders
posits that in the formative period of their ventures, founders are fully immersed both in the
internal workings of the organization and in mediating between it and the environment (Dobrev
and Barnett, 2005). The balance between the two activities is crucial to structuring an
organization that will not only materialize the founder’s idea but that will also gain support from
and fit well with its environment. Under conditions of legitimacy vacuum, the role of the founder
is unstable because the lack of institutional consent for her creation means that a
disproportionately higher fraction of the founder’s finite resources needs to go towards building
and maintaining claims to the external audience related to the shape and function of the new
entity and to developing the logic and rationale for supplying it with necessary resources.
Although every new organization lacks an established social standing, following a widely
accepted blueprint for organizing helps to convey a robust identity fitting with the prevalent
categorical schemas for the role and place of the new entity (Hsu and Hannan, 2005). When such
blueprints are undeveloped or non-existent, the task of imparting the organization to the external
audience as a legitimate social actor falls squarely on the founder’s shoulders. The more time and
8
effort the founder spends managing external expectations, the less of each of it can she devote to
managing the internal organization of the fledgling entity. Crucial decisions about key
architectural arrangements in the new firm are given secondary priority. This means that the
overall design of the firm may be poor and may not support its key strategic and technological
assets. To the extent that fundamental transformations in design are costly and difficult, an initial
internal misalignment between strategy and structure may persist and continuously diminish the
firm’s competitive strengths.
External misalignment and imprinting. The jeopardy of operating in unrecognized,
undeveloped markets also pertains to a process different from the one that gives rise to external
misalignment when position shifts are attempted. Long-term repercussions of an initial position
in unfavorable environment on organizational fitness can partly be understood in terms of
location, but this is only a special case (perhaps an especially simple one) of this relationship.
Unlike the positional inertia argument in density delay theory which assumes that location
characteristics are stable (e.g., peripheral location is associated with permanently limited
resources), features of the social structure vary in both time and space. In fact, assessing the
plausibility of a claim that firms founded in legitimacy vacuum have persistently lower
competitive fitness is only possible if the market in which they operate eventually develops and
gains legitimacy. This is consistent with the key theoretical insight in the imprinting story in
which organizational structure tends to persist even after the external conditions that once defined
it have shifted. As Stinchcombe (1965) observed, national fraternities whose charters did not
explicitly disavow racial segregation tended to be founded mostly in the South but from a
historical perspective, the meaningful predictor of variance in fraternities’ social organization
proved to be their time of founding (e.g., in the aftermath of the Civil War) rather than location.
9
Again, our prediction agrees with the expectations of density delay theory that lower
survival chances result from misalignment between a firm’s position and its environment. But,
while the positional inertia argument assumes that this misalignment is triggered when a firm
positioned in a resource scarce market location attempts to move, we contend that the
misalignment is generated by shifts in the environment itself without the need for a firm to
attempt a position shift. Specifically, organizations founded under conditions of legitimacy
vacuum develop routines and processes needed to deal with this environmental deficiency and
which become obsolete once the new organizational form they embody becomes institutionalized
(i.e., when legitimacy vacuum dissipates).
A focus on gaining external endorsements during the stage of legitimacy vacuum leads to
building organizational processes specifically geared to “sell” the firm’s model to gatekeepers of
resources who have the capacity to disperse resources despite the lack of broad social consent on
how they are to be used. While this may be helpful for the firm’s immediate survival it invariably
produces a long-term liability because success in a focal market is impossible unless the market
eventually gains legitimacy. When that happens, organizational structures primarily designed to
support a fit with an institutionally weak environment become obsolete and must be replaced. In
short, core structural change is likely mandated of firms founded in legitimacy vacuum because
the founder’s success at aligning the firm with an environment initially lacking constitutive
legitimacy produces strong misalignment pressures once the new market gains social acceptance.
And since routines established in the early years are likely to persist (Nelson and Winter, 1982)
and early organizational blueprints are more resilient to change than other features (Baron,
Hannan, and Burton, 2001), firms founded under legitimacy vacuum are especially susceptible to
inertia when attempting fundamental shifts in design. As a result, they face higher opportunity
and influence costs and ultimately lower competitive fitness than peer firms founded at a later
10
time.1 So by prompting the adoption of stable organizational structures and processes that
eventually become obsolete, legitimacy vacuum at founding leads to elevated failure hazards.
Uncertainty and imprinting. The arguments about poor internal design and obsolescence
in external fit both suggest indirect effects of legitimacy vacuum—through devising means to
facilitate resource acquisition whether by redefining the founder’s role or by a strong alignment
with a shifting, transitory environment—on structural imprinting and diminished survival
chances. We also conjecture that there is a direct effect, one that does not have a clear analogue in
density delay theory. This direct effect pertains to the role of uncertainty in the founders’
decisions about how to invest the resource endowments they have successfully solicited for their
firms. Even if founders are able to successfully resolve the resource acquisition problem under
conditions of legitimacy vacuum, investing these resources in a way that is conducive to future
growth and improved survival chances under such conditions may be unrealistic. Periods of
legitimacy vacuum in the environment are likely to coincide with the formative years of a new
market, industry or technology, periods during which it is unknown what evolutionary course the
system will take. And legitimacy is not only about what organizations should look like to obtain
resources but about how these resources should be deployed. Absent such set expectations,
organizations are likely to experiment with their investments thereby shaping the variance of
outcomes that sets the stage for the future course of the system. Organizational members in such
periods face the improbable task of having to guess the future of the market or industry.
While most organizational decision-makers confront this challenge, those working in new
organizations in a nascent industry have to do it from the standpoint of its most unpredictable
1 Of course, firms founded at any time during an industry’s evolution can become subjected to pressures of external misalignment and suffer elevated failure hazards. Unlike firms founded during legitimacy vacuum, however, firms founded during institutional regimes that are being swept by social change can point to the legacy of their past to justify future contributions. Stark (1996) provides several excellent examples of Hungarian state socialist firms’ claims for occupying a central position in the rapidly shifting post-socialist environment.
11
state. Historically, they are most likely to be proven wrong in their choice of resource
commitments and to be exposed to its consequences. As David (2000: 14) explains, “…The point
is not that these folks ought to have seen the shape of the future. Rather, it is that the shape of the
larger system that evolved was built upon their work…” Even if adequate resources are secured,
their ‘appropriate’ investment during a period of uncertainty is elusively difficult not the least
because it is only definable in retrospect once the evolutionary course of the industry has
unfolded.
To summarize, we argue that legitimacy vacuum has both direct and indirect effects on
the imprinting process within organizations in a population. The direct effect emanates from the
uncertainty that invariably permeates an environment where the means by which resources should
be deployed are contested and the value of new technologies is unclear. The indirect effects arise
either because the need to deal with the consequences of legitimacy vacuum requires of
organizational founders to prioritize and disproportionately allocate their limited resources to
processes and tasks geared to secure external endorsements (likely at the expense of improving
the internal fit between strategy and structure) or because building an organization well
positioned to garner resources during legitimacy vacuum leads to obsolescence and increasingly
deteriorating external fit once the organizational form becomes institutionalized.
Population density, survival chances, and imprinting. Ideas about constitutive legitimation
are central in organizational ecology and are an integral part of the density dependence model of
organizational evolution (Hannan and Freeman, 1977, 1989; Hannan and Carroll, 1992). The
theory holds that organizational proliferation within a population drives both the legitimation of a
new form and diffuse competition for resources. But it does so at different rates, such that
increases at low levels of density drive legitimation more than competition, and increases at high
levels of density drive competition but not legitimation (which has a theorized ceiling effect).
12
Empirically, the model is specified by estimating a non-monotonic effect of the contemporaneous
number of organizations in a population on their failure hazards predicting that the process of
building legitimacy reduces failure rates and intensified resource competition increases the
hazard. By contrast, in considering the role of density at founding for organizational evolution,
the density delay model specifies an imprinting effect of crowding and competition on survival
but not of legitimacy or its absence. Our theory suggests that there is a good reason to revise the
original model. The empirical prediction that emerges from the combination of density delay
theory and our conjectures about the imprinting effects of legitimacy vacuum is that the effect of
density at founding on the hazard of failure is nonmonotonic and has a U-shape: low-level
increases in density at founding contribute more to building legitimacy than to raising
competitive pressures. Accordingly, at low counts, density at founding decreases the failure
hazard. By contrast, high-level increases in density at founding only drive competitive crowding
because once an organizational form has attained a taken-for-granted status and the legitimacy
vacuum has been filled, the constitutive diffusion process (driven by organizational proliferation)
that helps the audience learn and become accustomed to a new category of social actors becomes
mute. Once something is learned, teaching it is unhelpful. At high counts, density at founding
only intensifies crowding and its effect on the hazard is positive.
Importantly, this prediction only applies to organizations operating at the dawn of a new
industry. Low density in a developed industry may reflect consolidation, intensified competition,
or stricter regulation, none of which are likely to affect the constitutive legitimacy of an
established organizational form. So the first-order density-at-founding effect should not hold for
firms founded at low density counts past the formative stage of the industry.
DATA AND METHODS
13
Empirical Analysis
Our choice of empirical application is guided by an effort to integrate the specificity of
historical context with the construction of general models. Below we report results of the
empirical analysis of survival rates of U.S. automobile manufacturers between 1885 and 1981.
Earlier analysis of this organizational population (Hannan et al., 1998) reports fully specified
density effects on organizational survival including a test of density dependence theory. The
internal validity of our results would be strengthened if we can show that they hold over and
above the baseline density effects. Building directly on an established density specification also
provides a conservative test of our theory because the margin between improving model fit by
introducing an additional density variable and overspecifying density effects is particularly small
when the baseline models are elaborate.
The form of automobile manufacturers was legitimated very quickly and mostly through
the widespread fascination with and interest in the novelty of the product (Flink, 1988). This
historical uniqueness of the population has to bear on the empirical specifications testing our
theory. This approach allows us to take note of earlier critiques of ahistoricism in ecological
analysis (Baum and Powell, 1995) while preserving the simplicity and generalizability of
ecological theories (Hannan and Carroll, 1995).
Data and measures. The U.S. automobile data were constructed from reports of
automobile historians and collectors, and include entries for all car producers ever known to
operate in the U.S. between 1885 and 1981. The end of the observation period reflects the last
year covered in the most comprehensive data source, a multi-volume encyclopedic book that
provides thorough authoritative coverage: The Standard Catalogue of American Cars (Flamang
1989; Kimes and Clark 1989; Gunnell, Schrimpf, and Buttolph 1992; Kimes et al. 1996; Kowalke
1997). Data on more recent periods were obtained from Kutner (1974) and Automotive News
14
(1993). The collection effort discovered an abundance of small, short-lived, and obscure firms
that frequently introduced peculiar automobile designs and production schemes. The data are
structured in event-history format so that each record corresponds to a single firm-year age spell
with the initial year split into two half-year periods to accommodate missing entry and exit data
for short-lived firms. The failure event is defined as exit to another industry, dissolution of the
firm, or if the firm drops out from the industry for an unknown reason. Merger and acquisition
cases are coded as (uninformatively) censored on the right. The years of the Second World War
are excluded from the data because car production was almost entirely devoted to military needs
during this period.
Our baseline model builds on Hannan et al.’s (1998) model and includes the same
measures and variables. Prior existence marks up firms that conducted other activities prior to
entering the automobile market.2 Contemporaneous organizational density (N) is specified
nonmonotonically, consistent with established theory and findings in organization ecology
(Carroll and Hannan, 2000), to include a linear and second-order term of annual counts of the
number of producer organizations. Industry age (t) is measured as the number of years elapsed
since the inception of the industry in 1885. Organizational size is measured as the natural
logarithm of the firm's annual production of automobiles, Ln(Size). Relative size of a firm is the
ratio of each firm's size to the size of the largest firm in the population at the time. A dummy
variable for size less than or equal to an annual production of 50 cars captures the compounded
absolute-small-size disadvantage, (Size ≤ 50). To control for socioeconomic and environmental
2 Of course, our arguments are about founding effects and clearly only apply to new firms, not to lateral entrants. We include these firms in our data (and control for their baseline effect) because the evidence in the historical record is not sufficiently clear on whether these were successful firms that transitioned from other industries or firms that failed in their operation in another industry and were completely reorganized for entry into automobile production (i.e., were de facto new firms). As a conservative precaution, we estimated all the models reported below (available from the authors) by excluding firms with prior experience from the data. The results remained unchanged. The results also did not change when we interacted the indicator variable marking up these firms with density at founding and the additional effect was not significant.
15
factors our specifications include effects of economic depression (Depression year dummies), the
level of the gross domestic product (GDP) adjusted for inflation (taken from Maddison, 1991),
and dummy period effects representing industry regimes (Mass production, Product
differentiation, JIT/TQC) as defined by Womack, Jones, and Roos (1990).
Methods
We estimate our models in terms of the instantaneous rate of failure (Tuma and Hannan,
1984; Blossfeld and Rohwer, 1995). This approach requires specifying a functional form of
duration dependence. We represented variation in organizational tenure (u) as a stochastic
piecewise-exponential function where the breakpoints for the pieces are denoted as
Assuming that 1 20 .≤ τ ≤ τ ≤ ≤ τP +1 ,τ = ∞P
P
t
0iu(N , t),
there are P periods:
We specify that the disbanding/exit rate is a function of
tenure in the industry ( ) and industry age ( ). The general class of models we estimate has the
form:
1I {u | u }+= τ ≤ ≤ τp p
i iμ (u, t) μ (u,=
, 1,..., .=p p
u
t) * exp (m )p
iμ
u I⋅ ψ ∈ p
where is the baseline rate for organization i estimated as a function of observables and
denotes a set of tenure-specific effects. The functions for assessing the imprinting arguments
relate to density at founding and its interaction with industry age (t), denoted by .
The u-subscript in indicates that we allow the density-at-founding effects to vary with
organizational tenure. Our model implies a non-monotonic effect of population density at
founding on the hazard of failure. Organizations founded at very low density have high failure
rates that decrease as the number of organizations at founding increases. The turning point in the
effect occurs when constitutive legitimacy is accomplished and further increases in density at
iμ (u, t)*
mp
0(N ) 0iu(N , t)ψ
0iuN
16
founding only intensify crowding. Specifying the exact functional form of the relationship
depends on context-specific historical developments as we explain below.
The automobile industry in the U.S. began in 1885, experienced an enormous boom
around the turn of the 20th century and by the time of the Great Depression was already highly
consolidated. The onset of competition was triggered by the quick legitimation of the new form,
which in turn was driven largely by the public’s fascination with automobility (Flink, 1988).
Proliferation of producers led to proliferation of car marques and models which quickly entered
the social discourse through three product-related mechanisms: the frequent rallies and car races,
the extensive reporting by the media, and the automobile fairs for new car models created an
institutional environment which itself precipitated the constitutive acceptance of the new
organizational form (Rao, 1994). Figure 1, which plots the historical density and number of
foundings in the population, illustrates these developments. The early rise in density somewhat
deviates from the typical pattern observed by ecologists in which the formative years of a
population initially witness a slow and gradual increase in density followed by a steep rise
(Hannan and Carroll, 1992). But in this case, the first phase of the process is almost non-existent
as density quickly shifted from close to zero to several hundred in the decade surrounding the
turn of the century. If legitimacy vacuum at founding was a factor, it would be confined to a very
short period. To reflect this brevity in the institutionalization of the blueprint of an automobile
producing firm and the onset of competitive crowding, we emphasize the variance over the very
low counts of population density. Accordingly, we transform density at founding into its natural
logarithm and enter it as a model covariate. In a second-order polynomial specification the simple
term reflects the build-up of constitutive legitimacy and is expected to have a negative effect on
the failure rate; the squared term captures crowding and should increase the hazard. Assuming
17
that density-at-founding effects are confined to the formative years of the industry, we expect:
0iu 0iu 0iu 0iu
2 2(lnN ) < 0; (lnN ) > 0; (lnN t) > 0; (lnN t) < 0.1 2 3 4ψ ψ ψ ⋅ ψ ⋅
[Figure 1 about here]
FINDINGS
Descriptive statistics and correlation coefficients are presented in Table1 and the
inferential statistics are presented in Table 2. We begin with a simple model (model 2.1) which
includes all the control variables and the main term of density at founding; its effect is positive
and significant, consistent with Hannan et al.’s (1998) estimates and supports the original density
delay argument. In the next model (model 2.2) we test for nonmonotonicity in the density at
founding effect and the estimates confirm our expectation. The first-order effect is negative and
the second order effect is positive3. We use the estimates from this model to plot the multiplier
of the hazard across the observed range of density in Figure 2a. Note that with only a singl
organization present at founding, failure chances are almost as high as when crowding is near its
peak. But as legitimacy builds, failure rates decline and at the point which by our estimates
coincides with the ceiling effect of legitimacy, the rate is reduced by nearly one half of what it is
at the height of legitimacy vacuum. Although model 2.2 does show the most general specification
e
3 We use two logarithmic functions of density at founding (N0) because we expect the overall effect of that variable to diminish as its values increase. This specification is in contrast to traditional density effects specifications where at least the squared term of density is non-logged to allow for the competitive effect to increase at an increasing rate. Our model takes into account arguments that as the auto industry quickly consolidated and partitioned into segments, processes of both mutualism and competition came to be driven more by ecological and social forces emanating from position in one of the industry’s three main market segments rather than by population level density. Nevertheless, we compared our model to a traditional non-logged specification of N0 effects. We found that in that specification the N0 effect is monotonically positive (obviously because it does not capture variance in the very low counts of N0) and that the overall model fit is inferior to the fit of our specification. To further assuage concerns of ad hoc modeling, we estimated a model with a simple step-function of N0 (instead of the logged N0 functions) which confirmed that the effect of N0 is nonmonotonic and reversing direction at low counts.
18
for modeling the predictions of our theory, it does not represent a complete test. Two additional
considerations remain to be addressed.
[Tables 1 and 2 and Figure 2a about here]
First, an argument that challenges the imprinting component of density delay theory holds
that, assuming heterogeneity in firm-level frailty, competitive crowding at founding may act as a
catalyst for the failure of frail firms. The survivors, however, emerge from this trial-by-fire
experience as stronger competitors and over time exhibit higher survival chances (Swaminathan,
1996; Lomi and Larsen, 1998). The empirical expectation is that density at founding will show a
positive effect on failure across the early tenure segment but that this effect will reverse to
negative across the high range of the tenure distribution.
A similar alternative explanation relates to the legitimacy-vacuum effect. Firms founded
in a period of high uncertainty may have initially high exit rates but those that survive that
precarious stage extract a strong learning benefit both from the failures of their pioneering peers
and from their own experience. Vicarious and experiential learning turn new firms in new
industries into stronger survivors if they outlast the formative stage of the industry. In empirical
terms, this conjecture predicts that the disadvantage of early entrants manifested in higher hazards
of failure when density at founding is very low will transform into an advantage with firm tenure
so that low density at founding for experienced firms will be associated with lower failure
chances.
To test this alternative hypothesis, in our next model (model 3.1 in table 3), we allow the
density at founding effects to vary with organizational tenure in the industry. If the alternative is
supported, we will see that the baseline effects of density at founding (negative and positive for
the main and squared terms, respectively) will reverse as firm tenure increases. The estimates
from our model clearly show that this is not the case: neither the linear nor the squared effect is
19
significant past the third-year tenure segment. More importantly, the likelihood ratio test of model
fit shows that allowing the density-at-founding effects to vary with tenure is not justified
(ΔLL=3.98, Δd.f.=2, p>.10) so it ought to be rejected in favor of the tenure-invariant
specification in the original model. Substantively, this finding confirms that both the legitimacy
vacuum and the crowding effects experienced at birth are persistent and irreversible. Of course,
there are strong learning effects as evidenced by the negative age dependence evidenced by the
estimates for the tenure piece constants but these effects occur independently of the imprinted
disadvantage of operating under legitimacy vacuum at founding.
[Table 3 about here]
A second consideration in interpreting our initial results is the need to allow for the
density-at-founding effects to vary with industry evolution. As we explained, the legitimacy
vacuum effect only obtains among early entrants at the start of the industry so we should observe
that as the industry matures, the legitimacy effect will disappear. Similarly, as competition
embeds in the evolving industry structure, it becomes less driven by mere ecological crowding
and more so by the competitive relationships leading to direct rivalry (Hannan, 1997). A general
way to model these expectations involves interacting the two density-at-founding terms with
industry age. If the initial density-at-founding effects wear off with time, we expect to see the
interaction effects work in the opposite direction, thus reducing the main effects. We test this
specification in our next model (model 3.2) which reveals that while the main and squared terms
of density at founding continue to be negative and positive (respectively), their interactions with
industry age show smaller coefficients that are in the opposite direction—positive and negative.
All density-at-founding effects are significant.4
4 We also estimated all of our models by including Hannan et al.’s (1998) extended population inertia specification of contemporaneous density effects in which the quadratic specification is interacted with population age. Our density-
20
DISCUSSION
Our goal here was to develop and test mechanisms that explain the relationship between
legitimacy vacuum, imprinting, and survival chances. We articulated three such mechanisms that
account for the lower survival chances of organizations founded under conditions of legitimacy
vacuum—the disproportionate investment of founders’ finite resources to manage the external
presentation of her venture as a legitimate social actor, the likelihood of such investments to
produce stable organizational routines that become obsolete and generate external misalignment
once legitimacy vacuum dissipates, and the likelihood of ineffective long-term investment of
resources available at birth during a period of uncertainty when the future evolutionary course of
the industry is unknowable. We argued that these three factors emanate from lack of constitutive
legitimacy at the level of the population to which an organization belongs and that, as ecological
and institutional theories predict, constitutive legitimacy is driven by increasing population
density. We combined these ideas with the logic of an established theory, the density delay model
in organizational ecology, which posits that as density at founding increases, organizational
survival chances deteriorate. In the revised version of the theory that we advanced here, the effect
of density at founding is non-monotonic—it improves survival chances as density at founding
increases at low counts (based on the arguments we introduced earlier) but failure hazards rise as
density at founding increases at high counts (based on the crowding argument in original density
delay theory).
At a broad level, our arguments are essentially about the role of history for organizational
evolution so we strived to depart from the common approach in economic and organizational
at-founding effects are unchanged when added to this specification but we do not present it here because it greatly complicates the interpretation of our effects.
21
analysis in which this role is reduced to a factor that is uniformly viewed as important (and
controlled for in most empirical specifications) but that does not lend itself to theoretical
generalizations. Not generalizing from historical occurrences is certainly warranted in many cases
in which path-dependence leads to realities sui generis with idiosyncratic outcomes. But we also
emphasized the importance of imprinting effects that may parallel any other relevant path-
dependent sequence. Bridging ideas from structural imprinting theory (Stinchcombe, 1965) and
the density delay theory in organizational ecology, both of which highlight the importance of
environmental characteristics at birth, we proposed one way to systematically examine and
summarize environmental conditions at the time of organizational founding related to the
difficulty of securing and investing resources during an industry’s early years. We recognize that
these conditions may vary greatly between markets or industries but conjecture that a measure of
population density summarizes these conditions reasonably well, even across different contexts.
The obvious advantage of a single, simple measure is not only that it makes
generalizations possible but that its application can be adapted flexibly to contexts with extended
degrees of specificity. We aspired to demonstrate this by selecting for our analysis an
organizational population with a rather unique history. Legitimation in the U.S. automobile
industry was only an indirect function of density. The early automobile offered by industry
pioneers generated sensational excitement among the public and quickly became inextricably
enmeshed in the social discourse. The numerous car rallies and fairs fueled this excitement, and
the media extensively catered to it (Flink, 1988). The consequence of such product-driven
legitimation was that it did not require as many firms as the typical diffusion of a new form
usually does. Accordingly, we used a simple logarithmic transformation of density at founding to
account for the early institutionalization of the nascent form and the impending shift to
competition for resources. The results strongly support the logic of our theory. After all, if the
22
poor survival chances of early entrants could be explained without recourse to our theory of
legitimacy vacuum, then their failure hazards would not have improved dramatically, as our
results demonstrated, with increases in density-dependent constitutive legitimation.
The first-mover (dis)advantage. Our theory of legitimacy vacuum and structural
imprinting appears to directly contradict notions of first-mover advantages (FMA) that are
popular in the strategic management literature. Given the conceptual haze surrounding the FMA
metaphor (Lieberman and Montgomery, 1988: 52) and the overwhelming sample selection bias in
most empirical studies in the FMA literature (i.e., a retrospective approach to the fates of
dominant survivors whose success is identified exclusively post hoc), we would not have
considered this contradiction had it not been for the fact that one of the strongest claims of
evidence in support of FMA comes from the analysis of the U.S. auto industry. Klepper (2002),
who used a different and more limited dataset of the industry, showed that first movers enjoyed
persistent survival advantages over later entrants. However, Klepper’s dataset traced the origin of
the industry to 1895 (a full decade later than Hannan et al.’s (1998) data that we used here), thus
entirely omitting the period of legitimacy vacuum. Every indication from the historical record is
that by 1895, the automobile was a social fact, not a curious novelty and that automobile
producers as an organizational form have by that time achieved a legitimate constitutive
standing—the first flagship industry publication Horseless Age published its inaugural issue in
December of 1895. Moreover, the subsequent few years provided clear evidence that the notion
of the automobile had become widespread—the convention of introducing new models through
the media began in 1897 (Flink, 1988), and it was also in that year that the New York Times
declared: “The new mechanical wagon with the awful name automobile has come to stay...”
Technically, our results agree with Klepper’s in that we too showed firms entering the
market when legitimacy vacuum had dissipated yet competitive crowding was still low to have
23
had the best chances of long-term success. But we also emphasized that these firms were not the
first movers. Instead, these firms, whether fortuitously or by great vision, entered the industry at
the point at which a sufficient number of incumbent firms had served to socially disseminate the
image and features of the new organizational form yet this number was insufficiently high to
generate crowding-out and resource scarcity effects. In terms of our model, this point is the
turning point in the U-shaped relationship between density at founding and failure hazards.
It appears then that our model may help to explicate some of the inconclusiveness
surrounding research on FMA once this framework explicitly allows for the legitimacy vacuum
effects that we theorized and evidenced here. And this would not be entirely at odds with current
conjectures. For example, Lieberman and Montgomery (1998: 1112) acknowledge that
“…pioneers often miss the best opportunities, which are obscured by technological and market
uncertainties,” and Tellis and Golder’s (1996: 66) descriptive analysis concludes that “early
leaders enter an average of thirteen years after pioneers yet are much more successful.” Given the
evidence we provided in support of the imprinting effect of legitimacy vacuum—a strong and
persistent first-mover disadvantage—perhaps it is useful to think of the mechanisms generating
FMA that Klepper (2002) articulated (e.g., economies of scale in R&D) as accruing to firms that
are first movers in a second wave of industry entrants.
CONCLUSION
We think that our model of the survival rates of new firms in new industries, which relies
on a single, simple explanatory measure, provides an empirically and theoretically tractable way
to model the impact of structural imprinting on organizations’ life chances. Unlike measures of
specific historical periods (typically marked up by indicator variables, as in our analysis here), the
density at founding measure and its effects are generalizable across contexts. Of course, testing
24
the theory in different contexts requires knowledge of historical conditions so the model can be
appropriately specified. If it is not, the danger is not that the predicted empirical relationships will
not be confirmed but that they might be even if the logic of the theory suggests otherwise
(Denrell and Kovács, 2006). This highlights perhaps the greatest weakness of ecological
models—their reliance on relatively simple measures makes them prone to misspecification.
Certainly, the mechanical application of density models can be completely atheoretical. But,
theoretical insight is at least as likely to derive from the historically informed (and accordingly
modified) application of simple ecological models as from models with more complicated
measures which may be harder to replicate and whose effects may be less clear to interpret. The
argument and evidence presented here lend credence to this claim.
25
REFERENCES
Automotive News. 1993. “America on the Wheel: 100 Years of the Automobile in America”. 4.
September, Special Issue.
Baron, J. N., M. T. Hannan, and M. D. Burton. 2001. “Labor pains: Change in organizational
models and employee turnover in young, high-tech firms.” American Journal of Sociology
106: 960-1012.
Baum, J. A. C., and W. W. Powell. 1995. "Cultivating an Institutional Ecology of Organizations:
Comment on Hannan, Carroll, Dundon, and Torres." American Sociological Review 60:
529-538.
Blossfeld, H. P., and G. Rohwer. 1995. Techniques of Event-History Analysis. Mahwah, NJ:
Erlbaum.
Boeker, W. 1988. “Organizational origins: Entrepreneurial and environmental imprinting at time
of founding.” in G. R. Carroll (ed.) Ecological models of organizations. Cambridge, MA:
Ballinger.
Carroll, G. R. and M. T. Hannan. 1989. “Density delay in the evolution of populations of
newspaper organizations: A model and five empirical tests.” Administrative Science
Quarterly 34: 411-30.
Carroll, Glenn R. and Michael T. Hannan. 2000. The Demography of Corporations and
Industries. Princeton, NJ: Princeton University Press.
David, P. A. 2000. “Path dependence, its critics and the quest for ‘historical economics’.” in P.
Garrouste and S. Ioannides (eds.) Evolution and Path Dependence in Economic Ideas:
Past and Present. Cheltenham, England: Edward Elgar Publishing.
Denrell, J. and B. Kovács. 2006. “Population Selection Bias: The Case of Density Dependence.”
Working Paper, Stanford Graduate School of Business.
26
Dobrev, S. D. and W. P. Barnett. 2005. “Organizational Roles and Transitions to
Entrepreneurship.” Academy of Management Journal, 48, 3: 433-449.
Flamang, J. M. 1989. Standard Catalog of American Cars 1976-1986. 2nd edn. Krause, Iola,
Wisconsin.
Flink, J. J. 1988. The Automobile Age. The MIT Press.
Freeman, J. 1986. Entrepreneurs as Organizational Products: Semiconductor Firms and Venture
Capital Firms. In Gary Libecap (Ed.) Advances in the Study of Entrepreneurship,
Innovation, and Economic Growth. Greenwich, CT: JAI Press.
Gunnell J. A., Schrimpf, D., and Buttolph, K. 1987. Standard Catalog of American Cars 1946-
1975. 2nd edn. Krause, Iola, Wisconsin.
Hannan, M. T. 1997. "Inertia, Density and the Structure of Organizational Populations: Entries in
European Automobile Industries, 1886-1981". Organization Studies. 18: 193-228.
Hannan, M. T. 1998. "Rethinking Age Dependence in Organizational Mortality: Logical
Formalizations." American Journal of Sociology 47:126-164.
Hannan, M. T., D. M. Burton, and J. N. Baron. 1996. “Inertia and change in the early years:
Employment relations in young, high technology firms.” Industrial and Corporate
Change 5: 503-536.
Hannan, M. T., and G. R. Carroll. 1992. Dynamics of organizational populations: Density,
legitimation and competition: New York and Oxford: Oxford University Press.
Hannan, M. T., and G. R. Carroll. 1995. “Theory Building and Cheap Talk about Legitimation:
Reply to Baum and Powell.” American Sociological Review 60: 539-44.
Hannan, M. T., G. R. Carroll, S D. Dobrev, J. Han, and J C. Torres. 1998. “Organizational
Mortality in European and American Automobile Industries, Part II: Coupled Clocks.”
European Sociological Review 14: 303-313.
27
Hannan, M. T., and J. Freeman. 1977. "Population Ecology of Organizations." American Journal
of Sociology 82:929-964.
Hannan, M. T., and J. Freeman 1989. Organizational Ecology. Cambridge, MA: Harvard
University Press.
Hsu, G., and M. T. Hannan. 2005. “Identities, genres, and organizational forms.” Organization
Science 16: 474-90.
Immelman, K. 1975. “Ecological Significance of Imprinting and Early Learning”. Ann. Rev. Ecol.
And Syst. 6:15-37.
Kimberly, J. R. 1979. “Issues in the Creation of Organizations: Initiation, Innovation, and
Institutionalizations.” The Academy of Management Journal 22: 437-457.
Kimes B.R. 1996. Standard Catalog of American Cars, 1805-1942. 3rd edition. Iola, WI: Krause
Publications.
Kimes, B. R. and H. A. Clark. 1989. Standard Catalog of American Cars 1805-1942. 2nd edn.
Krause, Iola, Wisconsin.
Kirsch, David. 2000. The Electric Vehicle and the Burden of History. Rutgers University Press
Klepper, Steven. 2002. “Firm Survival and the Evolution of Oligopoly”. The Rand Journal of
Economics. Vol. 33, No. 1. (Spring, 2002), pp. 37-61.
Kowalke, R. (ed). 1997. Standard Catalog of American Cars 1946-1975. 4th edn. Krause, Iola,
Wisconsin.
Kutner, R. M. 1974. The Complete Guide to Kit Cars, Auto Parts and Accessories. Auto Logic,
Wilmington, Delaware.
Levinthal, D., and J. Mayatt. 1994. “Co-Evolution of Capabilities and Industry: The Evolution of
Mutual Fund Processing.” Strategic Management Journal 15: 45-62.
28
Lieberman, Marvin B., and David B. Montgomery. 1998. “First Mover (Dis)Advantages:
Retrospective and Link with the Resource-Based View”. Strategic Management Journal.
19: 1111-1125.
Lomi A., Larsen E. 1998. “Density Delay and Organizational Survival: Computational Models
and Empirical Comparisons”. Computational and Mathematical Organizational Theory,
3(4): 219-247.
Lorenz, Konrad. 1935. “Der Kumpan in der Umwelt des Vogels. Der Artgenosse als auslösendes
Moment sozialer Verhaltensweisen“. Journal für Ornithologie 83, 137-215, 289-413.
Maddison, A. 1991. Dynamic Forces in Capitalist Development. New York: Oxford University
Press.
Meyer, J. W., and B Rowan. 1977. "Institutionalized Organizations - Formal-Structure as Myth
and Ceremony." American Journal of Sociology 83:340-363.
Mitchell, W. 1989. “Whether and When? Probability and Timing of Incumbents' Entry into
Emerging Industrial Subfields.” Administrative Science Quarterly 34: 208-230.
Nelson, R. R., and S. G. Winter. 1982. An Evolutionary Theory of Economic Change. Cambridge,
MA: Belknap Press of Harvard University Press.
Parks, Tim. 2005. Medici Money: Banking, Metaphysics, and Art in Fifteenth-Century Florence.
New York: W. W. Norton & Co, Inc.
Phillips, D. J. 2002. “A genealogical approach to organizational life chances: The parent-progeny
transfer among Silicon Valley law firms, 1946-1996.” Administrative Science Quarterly
47: 474-506.
Pólos, L., M. T. Hannan, and G. R. Carroll. 2002. “Foundations of a theory of social forms”
Industrial and Corporate Change 11: 85-115.
29
30
Rao, H. 1994. “The Social Construction of Reputation: Certification Contests, Legitimation, and
the Survival of Organizations in the American Automobile Industry: 1895-1912.”
Strategic Management Journal 15: 29-44.
Stark, David. 1996. "Recombinant Property in East European Capitalism." American Journal of
Sociology. vol. 101, no. 4 (January 1996), pp. 993-1027.
Stinchcombe, A. L. 1965. Organizations and Social Structure. In: J. G. March (Ed.), Handbook of
Organizations (pp. 153-193). Chicago: Rand McNally.
Swaminathan, A. 1996. “Environmental Conditions at Founding and Organizational Mortality: A
Trial-by-Fire Model.” The Academy of Management Journal 39: 1350-1377.
Tellis, Gerard J., and Peter N. Golder. 1996. “First to Market, First to Fail? Real Causes of
Enduring Market Leadership”. Sloan Management Review. 37: 65-75.
Tuma, Nancy Brandon, and Michael T. Hannan. 1984. Social Dynamics:Models and Methods.
New York: Academic.
White, H. C. 2001. Markets from Networks: Socioeconomic Models of Production. Princeton, NJ:
Princeton University Press.
Womack, J. P., D. T. Jones, and D. Roos. 1990. The Machine that Changed the World. New
York: Rawson Associates.
Table 1: Descriptive statistics and correlation coefficients for variables in the life-history spell file of U.S. autofirms Variab le n. x. an D. 1 2 3 4 5 6 7 8 9 10 11 12Mi Ma Me S. 1. Industry Tenure (u) 0.00 78.00 4.85 9.15 2. Mass Production 0.00 1.00 0.90 0.30 0.15 3. Production Differentiation 0.00 1.00 0.14 0.35 0.22 0.13 4. JIT / TQC 0.00 1.00 0.08 0.27 0.13 0.10 0.73 5. Prior existence 0.00 1.00 0.57 0.50 0.09 -0.01 -0.29 -0.26 6. Relative size 0.00 5284.50 35.55 284.97 0.59 0.04 0.21 0.13 -0.05 7. ln(size): -2.30 15.48 3.12 3.48 0.65 0.21 0.14 0.10 0.10 0.40 8. Size ≤ 50 0.00 1.00 0.68 0.47 -0.48 -0.17 0.01 0.04 -0.16 -0.18 -0.879. Depression year 0.00 1.00 0.17 0.38 0.04 0.06 -0.19 -0.14 0.04 -0.02 0.06 -0.0710. GDP 42.40 977.10 200.75 221.07 0.23 0.19 0.91 0.90 -0.29 0.18 0.19 -0.01 -0.15 11. Industry age (t) 1.00 96.00 35.15 22.06 0.33 0.31 0.90 0.76 -0.27 0.21 0.30 -0.12 -0.10 0.94 12. Density (n)
×1.00 345.00 204.90 109.55 -0.32 0.29 -0.61 -0.42 0.18 -0.18 -0.25 0.10 -0.08 -0.58 -0.65
13. n2 ( 10-3) 0.00 11.90 5.40 4.01 -0.29 0.30 -0.53 -0.38 0.16 -0.15 -0.24 0.10 -0.10 -0.52 -0.59 0.9714. Ln[Density at founding (no)] 0.10 5.84 5.17 0.88 0.04 0.22 -0.62 -0.44 0.22 0.07 0.03 -0.09 0.03 -0.56 -0.56 0.6915. Ln(no)2 0.01 34.15 27.50 7.86 0.05 0.22 -0.63 -0.46 0.22 0.07 0.03 -0.09 0.04 -0.58 -0.58 0.7016. Ln(no) t × 0.10 558.47 170.95 82.24 0.52 0.41 0.75 0.67 -0.20 0.36 0.44 -0.22 -0.06 0.84 0.92 -0.5217. Ln(no)2 t × 0.01 3248.86 866.73 384.93 0.65 0.45 0.44 0.43 -0.08 0.47 0.52 -0.31 -0.01 0.55 0.65 -0.28 Variable 13 14 15 16 14 Ln[Density at founding (no)] 0.61 15. Ln(no)2 0.63 0.99 16. Ln(no) t × -0.48 -0.24 -0.2617. Ln(no)2 × t -0.27 0.16 0.15 0.90
Table 2: ML estimates of density at founding effects on the failure rate of U.S. autofirms Model 2.1 Model 2.2 Tenure (u):
u < 0.5 -1.09** (0.28) -0.04 (0.45) 0.5 ≤ u < 1 -1.20** (0.28) -0.14 (0.46) 1 ≤ u < 3 -1.71** (0.28) -0.65 (0.46) 3 ≤ u < 7 -1.95** (0.28) -0.89* (0.46) u ≥ 7 -2.04** (0.29) -1.00* (0.45)
Mass Production 0.79** (0.11) 0.78** (0.11) Production Differentiation 0.00 (0.21) 0.00 (0.21) JIT / TQC -0.61* (0.34) -0.66* (0.34) Prior existence -0.12** (0.05) -0.12** (0.05) Relative size -0.02** (0.01) -0.03** (0.01) ln(size): -0.13** (0.02) -0.13** (0.02) Size ≤ 50 0.53** (0.15) 0.54** (0.15) Depression year -0.27** (0.07) -0.28** (0.07) GDP 0.002** (0.001) 0.002** (0.001) Density (n) -0.01** (0.00) -0.01** (0.00) n2 (× 10-3) 0.06* (0.03) 0.05* (0.03) Ln(no) 0.18** (0.06) -0.38* (0.21) Ln(no)2 0.07** (0.03) Log-likelihood -3709.11 -3705.99 LR test vs. model (d.f.) 6.24 vs. Model 2.1 (1) Numbers in parentheses are standard errors; Number of spells:8892; Number of events: 2051. * significant at .05, ** significant at .01, one-tailed tests
Table 3: ML estimates of density at founding effects (by tenure and industry age) on the failure rate of U.S. autofirms Model 3.1 Model 3.2 Tenure (u):
u < 0.5 0.86 (0.61) 0.65 (0.44) 0.5 ≤ u < 1 0.76 (0.61) 0.54 (0.44) 1 ≤ u < 3 0.26 (0.61) 0.04 (0.44) 3 ≤ u < 7 -1.74* (0.76) -0.20 (0.44) u ≥ 7 -1.85** (0.76) -0.35 (0.44)
Mass Production 0.76** (0.11) 0.28** (0.14) Production Differentiation 0.01 (0.21) 0.10 (0.21) JIT / TQC -0.66* (0.34) 0.69* (0.40) Prior existence -0.12** (0.05) -0.11** (0.05) Relative size -0.03** (0.01) -0.02** (0.01) ln(size): -0.13** (0.02) -0.15** (0.02) Size ≤ 50 0.55** (0.15) 0.48** (0.15) Depression year -0.29** (0.07) -0.26** (0.07) GDP 0.002** (0.001) -0.01** (0.001) Density (n) -0.01** (0.002) -0.003** (0.001) n2 (× 10-3) 0.04 (0.03) 0.03 (0.03) Ln(no) [all u] -0.99** (0.24)
u < 3 -0.82** (0.29) u ≥ 3 0.04 (0.35)
Ln(no)2 [all u] 0.15** (0.03) u < 3 0.12** (0.03) u ≥ 3 0.02 (0.04)
Ln(no) × t 0.02** (0.005) Ln(no)2 × t -0.003** (0.001) Log-likelihood -3704.00 -3685.60 LR test vs. model (d.f.) 3.98 vs. Model 2.2 (2) 40.78 vs. Model 2.2 (2) Numbers in parentheses are t-statistics; Number of spells:8892; Number of events: 2051. * significant at .05, ** significant at .01, one-tailed tests
33
35