An Intra-organizational Ecology of Individual Attainment*
Christopher C. Liu
Rotman School of Management-University of Toronto
105 St. George Street
Toronto, Ontario M5S 3E6 Canada
chris.liu@rotman.utoronto.ca
Sameer B. Srivastava
Haas School of Business, University of California-Berkeley
545 Student Services, #1900
Berkeley, CA 94720-1900
srivastava@haas.berkeley.edu
Toby E. Stuart**
Haas School of Business, University of California-Berkeley
545 Student Services, #1900
Berkeley, CA 94720-1900
tstuart@haas.berkeley.edu
Word Count: 11,129
July 2012
*The authors would like to thank ISCO and BTCO for their support of this project. The nature and
sensitivity of the data they have provided precludes their identity. The usual disclaimers apply.
Authorship on this paper is alphabetical.
** Corresponding author.
1
An Intra-organizational Ecology of Individual Attainment
Abstract
Extending niche theory in ecological perspectives, this article develops an intra-organizational
conceptualization of the niche that is grounded in the activities of organizational members.
Niches are constructed from mapping individuals to formal and informal activities. Because the
many activities within organizations are difficult to observe, we propose a novel empirical
strategy to characterize niches: we exploit the complete roster of memberships in electronic
mailing lists. We characterize niches along four dimensions: competitive crowding, status,
diversity, and typicality, and we develop theoretical propositions about the resources that accrue
to occupants of niches that vary on these dimensions. Propositions are tested in two, disparate
empirical settings: the R&D laboratory of a biopharmaceutical company and an information
services firm. Results indicate that, across both settings, people in competitively crowded niches
are less central in the communication network and achieve lower levels of attainment, whereas
those in high status and diverse niches are more central in the communication network and
achieve higher levels of attainment. We find suggestive evidence that employees who deviate
from the identity blueprint for their job role also garner fewer rewards.
July, 2012
2
Introduction
Niche theory has been enormously influential in the literature on interorganizational dynamics. Indeed,
the niche is the theoretical cornerstone of organizational ecology (Freeman and Hannan 1983; Hannan,
Carroll, and Polos 2003; McPherson 1983; cf. Popielarz and Neal 2007), and it is also foundational in
other, seminal sociological approaches to the study of economic markets (e.g., White 1981; DiMaggio
1986; Burt 1992). This work has established that population dynamics, including organizational births
(Hannan and Freeman 1987), growth and mortality rates (Barron, West, and Hannan 1994), resource
partitioning processes (Carroll 1985; Dobrev, Kim, and Hannan 2001), and status differences (Podolny,
Stuart, and Hannan 1996), depend on multiple aspects of the articulation of the niche spaces that host
organizational populations. Niche theory has also been extended to a wide range of social phenomena,
including interaction among occupations (Rotolo and McPherson 2001; Abbott 1988), the emergence of
forms in an institutional identity space (Ruef 2000), and tastes in music (Mark 2003).
In this article, we bring niche theory to intra-organizational analysis. Just as the organizations in a
population interact competitively and symbiotically within a confined resource space, so too do
employees within an intraorganizational ecology cooperate and compete under conditions of resource
scarcity. In population ecology, organizations of a given form vie for resources such as customers,
employees, financial capital, and legitimacy. Similarly, inside an organization, individuals compete and
collaborate to attract resources such as salary, bonus, budget allocation, attention from senior leaders, and
the legitimacy of their work roles. Likewise, just as the finiteness of resources available to support any
given organizational form creates a carrying capacity that shapes organizational vital rates, so too are the
resources available to employees in organizations subject to constraints. Finally, just as aspects of
organizations’ realized niches affect organization-specific outcomes, we will demonstrate that features of
employees’ intraorganizational niches influence their attainment and other career outcomes.
Although the analogies are numerous, the operative question is how to conceptualize and measure
employees’ niches in an intraorganizational ecology. We believe that any valid approach to the concept
and measurement of an intraorganizational niche must consider individuals’ positions in both the formal
3
and informal structure of an organization. There is too much theory and evidence that informal structure
matters to exclusively rely on the formal structure to construct an intraorganizational niche space. The
solution we propose is to locate employees in a resource space that is defined by the multitude of
recurrent “activities” (broadly defined) in the organization. This emphasis on activities is reminiscent of
Elton’s (1927; p. 63) classic definition of a niche, “a term to describe the status of an animal in its
community, to indicate what it is doing … .” In most organizations, some activities will exactly parallel
the roles prescribed by the formal reporting structure; however, many other activities will stem from
informal social and interest groups, as well as less formal and sometimes crisscrossing modes of
organization, such as task forces, committees, and so forth. Thus, in our framework, intraorganizational
niches arise from the pattern of recurrent activities that embed the employees in the formal and the
informal organization.
Specifically, we construct the niche space from a mapping between individuals and activities,
which forms a dual mode network. Conceptually, this activity-focused affiliation network encapsulates
multiple dimensions of an organization’s social structure, but pragmatically, it is very difficult to observe.
The solution we propose to this problem relies on a novel data source that has potentially broad
application in organizational analysis: the complete set of electronic mailing lists in an organization. If we
think of each mailing list as a membership roster for an “activity,” the full set of electronic mailing lists is
a dual-mode network with two disjoint sets of elements: employees and activities.
Within an organization, many e-mail lists correspond to department, office, and functional
memberships and therefore trace the formal organizational structure. However, lists also offer a window
into the complex ways in which work actually gets done. In a typical organization, for example, mailing
lists exist for standing cross-functional teams, ad hoc task forces, the “kitchen cabinets” of organizational
leaders, and many professional interest groups. Likewise, lists also reveal social groups, such as the
softball league or communities of employees in common ethnic groups. Using these data, it is possible to
identify employee-level niches, and in conjunction with other information such as compensation,
performance ratings, promotions, and turnover, to relate niche characteristics to career outcomes.
4
In the analysis to follow, we demonstrate the utility of the framework and the data source using a
few empirical strategies. First, we measure four properties of niches, competitive crowding, status, niche
diversity, and typicality, and show that these characteristics affect employee attainment in exactly the
directions that theory would predict. Second, we construct intraorganizational ecologies using data from
two very different organizations: a biopharmaceutical laboratory and a large information services
company. We find remarkable consistency in results across the two organizations. Third, we test our
hypotheses exploiting two, quite different dependent variables. The first is individual attainment
(measured by annual bonus or performance rating), and the second is centrality in the internal
communication network. Once again, we find strong concordance in the results between these two
outcome variables. Taken together, the findings offer support for the proposed conceptualization and
measurement of intraorganizational niches.
Niches in an Intraorganizational Ecology
Our theory of intraorganizational niches extensively borrows from organizational ecology. Ecological
theories begin with a distinction between the fundamental and the realized niche. Drawing on Hutchison
(1957), Hannan and Freeman (1989) define the fundamental niche of an organizational population as the
region of a resource space in which the population will experience a non-negative rate of growth. The
realized niche of an organizational population is the subset of the fundamental niche that is actually
occupied, given the presence of competitive interactions with rivalrous populations.
As research in organizational ecology has progressed, much of the empirical work on ecological
dynamics has examined the structure of realized niches. We build on this vein of the literature. Two
conceptual aspects of this work guide our development of a theory of intraorganizational niches. First,
like many theories of social structure (Simmel 1902), ours rests on an analytical distinction between actor
and position—positions or niches can be characterized in general terms and in a manner that is
independent of any specific occupant, and positional characteristics have a primary influence on the
allocation of opportunities. Second, implicit in the concept of a realized niche is the relational nature of
its definition: realized niches are elaborated through ongoing interactions among the occupants of the
5
space. Put another way, the structure of intraorganizational niches arises from a process akin to
endogenous population structuring (Hannan and Freeman 1977). Like the niche in Carroll’s (1985)
resource partitioning theory or in models of size-localized competition in interorganizational ecology
(Baum and Mezias 1992), positions in the intraorganizational niche space emerge from interactions inside
the organization.
Many studies have linked organizations’ positions in a niche space to their life chances.
McPherson’s (1983) formulation of a competition matrix for a group of organizations represents a
seminal contribution to this area. McPherson (1983) defined niches of voluntary organizations in terms of
their locations in a resource space comprising the distribution of possible organizational members across
sociodemographic categories, such as age and sex (see also Baum and Singh 1994; Popielarz and
McPherson 1995). Much of the follow-on work, however, implicitly has examined niches in some form
of an activity space. For instance, Podolny, Stuart and Hannan (1996) defined niches based on patent
citation patterns in a group of semiconductor firms. This creates an affiliation network defined by firms’
choices to become active in some technical areas, but not in others. The properties of organizationspecific niches are then derived from the relational structure of the affiliation network: two firms compete
insofar as their participation in technical activities overlap. Therefore, the network is defined by the
mapping of companies to fine-grained areas of technology. A similar conception of niche is presented in
Dobrev, Kim and Hannan (2001). In their article, technological niches are constructed from the range of
engine sizes that automobile manufacturers produce (see also Dobrev, Kim, and Carroll 2002).
In this article, we distill properties of intraorganizational niches in the recurrent activities of the
organization. We use the term “activity” broadly: it includes work-related tasks, but also interactions
within ad hoc work groups, social groups, standing committees, and so forth. Why focus on employee
niches in a broadly defined activities space? We have two primary rationales. First and most generally, we
believe that activities are the intraorganizational occurrences in which collaborative and competitive
interactions translate into heterogeneous career outcomes. Put simply, activities are the what and where of
the allocation of individuals’ time in organizations. Therefore, when they can be observed, niche
6
characteristics will provide empirical leverage in analyses of pertinent outcomes. Other things equal, by
shaping the level of competition individuals face for scarce resources, by structuring their opportunities to
be of strategic value to the organization, or simply by determining their visibility to actors in positions of
power, the characteristics of employees’ positions in the activities space will affect their prospects for
obtaining resources.
Second, beyond a certain level of organizational size and complexity, no single actor has the
capacity to take part in more than a fraction of the ongoing activities in the organization. This implies a
heterogeneous pattern of individual engagement with activities that, in turn, insures variety in niche
positions. Moreover, the fact that niches equate to positions in a dual mode network implies that, like
other relational conceptions of social positions, person-specific intraorganizational niches also are
relational; they are defined in juxtaposition to one another.
We describe an individual’s intraorganizational niche in four dimensions: competitive crowding,
status, diversity, and typicality. Competitive crowding refers to the density of actors who occupy similar
niche positions. The status of a niche varies with the extent to which it offers access to colleagues who are
in positions of authority in the organizational hierarchy. A niche’s diversity gauges the extent to which it
provides access to employees in different functional areas and hierarchical levels of the organization.
Finally, typicality refers to the extent to which an individual’s niche conforms to the blueprint for her
role: she is typical insofar as her position in the activity space converges on the mean position vector for
others in her specific job role, which we assume represents the legitimated mix of roles for that job.
Hypotheses
Niche Crowding
Organizational conceptions of competitive crowding lie at the intersection of organizational ecology and
social network theory. Ecologists have gauged variation in competitive crowding across niches in terms
of “overlap densities” (McPherson 1983), or the count of organizations that participate in a given niche
(Baum and Singh 1994). When overlap density is high, organizational life chances are compromised.
Beginning with DiMaggio (1986), scholars of social networks have noted a conceptual similarity between
7
measures of structural equivalence in a network and ecological niche overlap (Burt 1987; Podolny, Stuart,
and Hannan 1996). The parallel arises because structural equivalence precisely is a measure of overlap;
two structurally equivalent elements in a network are, by construction, perfect substitutes: one element
may replace the second without consequence to the network’s shape. Thus, when niches can be defined as
positions in a network that connects actors to objects, network-based similarity measures are a means to
quantify the competitive intensity of niches. A literature has since evolved in which the intensity of
competition between actors is a function of the similarity between them in some type of resource space,
such as recruitment patterns in a labor market (Sørensen 1999; Bidwell and Fernandez-Mateo 2010), a
supplier-buyer network (Burt 1982), a geographic area (Baum and Haveman 1997; Sorenson and Audia
2000), a technology space (Podolny, Stuart, and Hannan 1996), or a product features space (Dobrev, Kim,
and Hannan 2001).
We assess the competitive crowding of positions in the intraorganizational activity space. The
general idea that competition may have adverse consequences for individuals’ attainment is a longrunning theme in research on careers. Studies of workforce demography (Barnett and Miner 1992;
McCain, O'Reilly, and Pfeffer 1983; Stewman and Konda 1983), for instance, have found that employees
who enter organizations in large cohorts may face greater competition for senior-level jobs. Under
conditions of high crowding, mobility rates decrease as individuals’ advancement paths encounter
bottlenecks (Reed 1978). Likewise, there is evidence that crowding reduces individuals’ ability to realize
the potential gains from opportunities for brokerage (Burt 1997). Human capital theorists have noted too
that women frequently enter occupations requiring general skills that are robust to temporary exits from
the labor market. The result of this process is the crowding of workers into such roles and a reduction in
the wages attached to such jobs (Barnett, Baron, and Stuart 2000; Bergmann 1986).
There are both general theories of the effects of competitive crowding of niches and labor-market
specific theories of competition in work roles that lead us to anticipate that occupants of crowded
intraorganizational niches will experience relatively less desirable career outcomes. In particular, we
argue that occupants of crowded niches will be less effective in garnering organizational attention and
8
will therefore occupy less central positions in the communication network within an organization.
Moreover, individuals in crowded niches have less leverage with which to secure rewards and managerial
recognition. Our work extends the existing literature, however, because we conceptualize and measure the
competitive intensity of niches defined by aspects of formal job roles and of individuals’ position in the
informal structure of the organization. In accounting for employees positions in the formal workflow and
the informal structure, we can generate a nuanced gauge of the competitive intensity of employees’
niches. We hypothesize: Employees in competitively crowded intraorganizational niches will
experience lower levels of attainment and will occupy less central positions in the
intraorganizational communication network than will employees in niches that are not
competitively crowded (H1).
Niche Status
Individuals in crowded niches lack differentiation from alters because there are many who could
substitute for them in the activities in which they participate in the organization. Crowding, therefore,
concerns the level of competitive differentiation between the actors in a social system. Status, conversely,
references their social standing in hierarchical orderings. A robust finding in the sociological literature is
that there is a positive return to positions of prestige. For instance, high status actors garner greater
recognition for a given quality of product (Merton 1968; Podolny 1993); they obtain the broadest range of
choice among potential collaborators (Stuart 1998); the experience accelerated rates of growth (Podolny
and Phillips 1996); and status facilitates entry into new market segments (Jensen 2003).
One of the defining features of status is that it “leaks” (Podolny 2005). Within an organization,
for example, an individual’s, a work group’s, or even a full department’s reputation is a function of those
who associate with that person or collectivity. This occurs because other community members infer status
from affiliations. As a few among many examples in the literature, graduate students derive status from
their affiliations with particular departments and universities (Merton 1968); law firms accrue status from
the prestige of the universities their staff attended (Phillips and Zuckerman 2001); and early stage
companies derive status from their investors and strategic alliance partners (Stuart, Hoang, and Hybels
9
1999). Likewise, in an intraorganizational ecology, certain niches are high status because the activities
that define them confer privileged access to those in positions of power and prestige in the organization.
We hypothesize that occupants of these niches derive career benefits because they have superior access to
valuable resources. Specifically, Employees in niches that confer access to high status actors will
experience higher levels of attainment and will occupy more central positions in the communication
network than will employees in niches that do not confer access to high status actors (H2).
Niche Diversity
Much of the seminal research on social networks, including Granovetter’s (1973) theory of weak ties and
Burt’s (1992) theory of structural holes, concerns how the shape of the network determines the
dissemination of information. In fundamental ways, the arcs of the social network in a community
determine the distribution of information within it. This particularly is true for tacit or proprietary
information that is not easily or willingly transferred through public broadcast channels.
There are two very different types of arguments about the potential advantages that accrue to
occupants of niches that provide access to a diverse array of contacts. First, strategic theories of control in
social networks, most notably Burt (1992), describe the potential gains from intermediating transactions
among disconnected alters. Second, information-based theories of opportunity posit that individuals with
broad contact networks gain exposure to more varied streams of ideas and information, and therefore they
are ideally positioned to identify untapped opportunities or to experience the creative spark that comes
from the recombination of synergistic knowledge inputs arriving from alters in a diverse contact network
(e.g., Brass 1995; Burt 2004; Fleming, Mingo, and Chen 2007; Tortoriello, Reagans, and McEvily 2012).
We argue that the breadth of access to diverse contacts also distinguishes intraorganizational
niches. Some niches limit their occupants to relatively closed networks, in the sense that homogenous
groups of individuals engage in the activities that define the niche. Others, by contrast, gather a diverse
array of participants and therefore serve as gateways to a range of information and talent. We therefore
hypothesize: Employees in niches that confer access to a diverse range of contacts will experience
10
higher levels of attainment and will occupy more central positions in the communication network
than will employees in niches that do not confer access to a diverse range of contacts (H3).
Identity Blueprints
There has been a groundswell of interest in categorization processes and their implications for marketbased outcomes. A category emerges when the members of an audience achieve a certain measure of
consensus regarding the label they apply to the elements in a common set. From an ontological
perspective, the recent focus on categories, spurred by Zuckerman (1999), represents a shift from
conventional conceptions of social structures to cognitive structures of meaning. This constructivist shift
suggests that actors’ level of access to resources also will depend on how they are positioned in the
meaning systems audience members craft. Thus, the meaning system itself is a niche space, in which
characteristics of the positions of its constituents potentially are antecedent to outcomes.
The literature on identities in category spaces has considered how a variety of types of audience
members contribute to the formation of identities that then legitimate and constrain organizational action,
including consumers (Carroll and Swaminathan 2000), employees (Baron 2004), and professional critics
(Zuckerman 1999; Hsu and Podolny 2005). It also has documented an identity-based penalty experienced
by actors who deviate from the expectations of audience members regarding the socially acceptable form
for a given role. When actors deviate from the properties that audience members perceive to be legitimate
for a given identity, these “code violations” (Polos, Hannan, and Carroll 2002) are sanctioned through
some form of devaluation of the focal actor. For instance, in research on generalists versus specialists,
actors who attempt to span too many categories often may experience evaluative sanctions if they are
perceived as poor fits in any individual category, which imposes a limit on the variety of identities that a
given actor may profitably assume (Zuckerman 2005). This has been shown to be true in a number of
different market contexts (Zuckerman et al. 2003; Hsu 2006; Hsu, Hannan, and Kocak 2009; Kovács and
Hannan 2010).
What are the implications of this new line of work for a theory of intra-organizational niches?
The relevance may be in the formation of identity-based expectations that attach a specific or prototypical
11
set of activities to clearly defined work roles within organizations. If individuals in an organization
associate a certain set of activities with a given job, audience members—fellow employees and
supervisors within the organization—may then form concrete assumptions about the legitimate
configuration of activities that map to the job. And assuming that such prototypical configurations exist
for job categories within organizations, individuals who deviate from the activity profile for his or her
job-specific prototype may find that their contributions to the organization are devalued by colleagues.
We hypothesize: Employees who are distant from the identity blueprint for their specific job role
will experience lower levels of attainment than will employees who are proximate to the job-specific
identity blueprint (H4).
Method
Research Sites and Study Population
We tested these four hypotheses in two, quite different organizational settings: a biopharmaceutical
company, which we label BTCO, and an information services provider, which we label ISCO. We
collected two datasets to build confidence in the findings if they are replicable.
The study population at BTCO included almost 1,000 employees in the research division. This
unit of the company conducts basic and applied scientific research to supply the company’s drug
development pipeline. Employees in research worked across a range of biological disciplines and
methods. The division was modeled after an academic research center, with senior scientists directing the
work of junior researchers and staff in a decentralized, laboratory setting. Given the scientific nature of its
purpose, the workforce at BTCO was very highly educated, with many doctoral degree holders on staff.
The global information services provider, ISCO, was the largest business unit of a conglomerate.
At the time of data collection, ISCO had a workforce of 10,000 in over 100 offices and generated over $4
billion in revenue. Our study population included all employees based in the U.S. (nearly 5,000
employees). The company’s operations in the U.S. were structured primarily along functional lines
including product development, marketing, sales, finance, legal, and human resources. In addition, the
12
company had recently introduced a small number of integrated units that combined functional resources
and were accountable for the profitability of an entire line of products and associated services.
Data
We collected three types of data from both BTCO and ISCO: email distribution lists, the server logs that
contained all internal email exchanges, and human resource records. Email distribution lists included all
email addresses associated with the list. In both organizations, email addresses were encrypted to preserve
employee privacy. In addition, ISCO (but not BTCO) encrypted distribution list names before sharing the
data with us. Therefore, we could identify the nature of all email distribution lists in one of the
organizations, BTCO, but not from our second research site.
We collected email distribution lists from both organizations at frequent intervals for several
months. Although there was little variation in the list composition or members over this short window of
time, we selected a period early on in our data collection from each organization for our analyses.1 In both
organizations, distribution lists were used to facilitate communication among groups of employees who
interact frequently. Based on interviews and our own review of the list names in the biopharmaceutical
company, we identified three general categories of lists: (a) formal organizational or geographic work
units (b) social groups (e.g., the running club); and (c) work-related teams or professional interest groups
(e.g., cross-functional project teams, interest groups for specific molecules under investigation or disease
areas). As we describe subsequently, the majority of email distribution lists at BTCO were for workrelated activities, but it is clear from their spread across categories that these lists represent a mapping of
individuals to formally prescribed work tasks and to a wide variety of memberships in informal work and
social groups.
In addition to distribution lists, we collected email logs. The email logs followed the time period
in which we extracted distribution lists, so niche characteristics were measured in temporal lags to the
outcome variables. Again to preserve employee privacy, both companies stripped message content before
1
In ISCO, there was high overlap (88%) between the distribution lists in use over a six month period. In BTCO, the
overlap was 82%. Given the limited amount of change in the lists, we perform our analyses in the cross section. Of
course, this will limit our ability to make any causal claims from the analyses to follow.
13
providing the data to us. After removing message content, the email logs included an encrypted identifier
both for senders and receivers, as well as time and message stamps. In interviews with senior leaders, we
repeatedly were told that email was the primary means of communication in both organizations.
In using the email, we restricted the data to messages sent to a single recipient (i.e., we excluded
mass emails and messages sent to distribution lists). We imposed this screen to eliminate the possibility of
a tautological relationship between the distribution-list-based characteristics of employee niches and the
email-based outcome variable. This screen insures that none of a person’s incoming communication
activity originated from messages addressed to email lists. At BTCO, we were able to obtain email logs
for all employees in the research division; however, ISCO only provided email logs for a senior cadre of
employees, the North American members of the company’s extended leadership team. Therefore, at ISCO
analyses that require email data were restricted to this group of senior managers.
Finally, we collected encrypted employee records from the human resource systems of both
companies. We used the same encryption algorithm across all three data sets (distribution lists, electronic
mail logs, and human resource records) so the datasets could be readily merged at the person-level via a
common, hashed identifier. For both organizations, we were able to gather employees’ sex, tenure,
hierarchical rank, organizational affiliation subunit, supervisor identification number, annual performance
rating, and target bonus paid (the latter only at BTCO). Because annual performance ratings and bonuses
reflect contributions made in the prior year of service, we collected the human resource data in the year
after the one from which we drew email and distribution list data.
Measures
Our measures of employee attainment slightly differed between the two organizations. At BTCO, we used
a measure of the bonus paid to an employee. In BTCO’s compensation plan, the HR department used a
formula to provide each supervisor with a target bonus for each direct report. The supervisor then had
discretion to adjust this target bonus based on his or her evaluation of the employee’s performance in the
prior year. We report the ratio of actual bonus divided by target bonus. This is formulaically centered near
one based on the firm’s compensation policy: the “average” performer hits the target.
14
At ISCO, our measure of employee attainment was the performance rating given to an employee
by his or her supervisor. Ratings ranged from 1 (does not meet expectations) to 5 (exceeds all
expectations). We would have preferred a direct measure of compensation as we obtained from BTCO;
however, these data were not available from ISCO. We note, however, that BTCO also assessed
employees on a 5-point scale, and employee rating was highly correlated with target bonus paid.
We also hypothesized that characteristics of intraorganizational niches will be correlated with
centrality within the organization. We assessed centrality using the person-to-person, direct tie (i.e., no
mass mailing) email network. In constructing this measure, we eliminated all messages sent to
distribution lists. We then measured the number of colleagues sending one-to-one messages to the focal
actor in a given week. 2
To measure characteristics of intraorganizational niches, we first create, for each organization, a
person-by-distribution list, two-mode matrix, A i,k :
𝑎𝑖,𝑘 = 1 𝑖𝑓 𝑖 ∈ 𝑘; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
where i indexes actors and k indexes lists. To define niche crowding, we start by converting the two-mode
matrix, A i,k , into a one-mode matrix, O i,j , of overlapping distribution lists between actors, i, and alters, j.
The entries of O i,j are given by:
𝑜𝑖,𝑗 = � 𝑎𝑖,𝑘 ∗ 𝑎𝑗,𝑘
𝑘
Next we define a matrix of supervisor overlap, S i,j , between i and j. The entries of S i,j are given by:
𝑠𝑖,𝑗 = 1 𝑖𝑓 𝑖, 𝑗 𝑠ℎ𝑎𝑟𝑒 𝑠𝑎𝑚𝑒 𝑠𝑢𝑝𝑒𝑟𝑣𝑖𝑠𝑜𝑟, 𝑧; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Competitive crowding of the niche occupied by actor i is then defined:
𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑣𝑒 𝐶𝑟𝑜𝑤𝑑𝑖𝑛𝑔𝑖 =
∑𝑗≠𝑖
𝑜𝑖,𝑗 ∗ 𝑠𝑖,𝑗
∑𝑘 𝑎𝑖,𝑘 + ∑𝑘 𝑎𝑗,𝑘
∑𝑗≠𝑖 𝑐𝑖,𝑗
2
Our results were robust to the use of longer time periods and alternative centrality measures, such as the volume of
one-on-one messages received in a given week (versus the count of senders). We also obtained similar results when
we chose a different time period for the cross-sectional analysis.
15
In this equation, the numerator represents the number of email lists shared between actor i and a
given alter j, conditional on both reporting to a common supervisor, z. This measure is then weighted by
the sum of distribution lists to which i and j each belong, and summed across all alters, j. Finally, we
weight the resulting measure by the number of number of alters, j, with whom i shares a supervisor.
Competitive crowding rises when colleagues who report to actor i’s supervisor are structurally equivalent
in the activity space. To flexibly identify people occupying highly crowded niches, we create an indicator,
set to 1 for people in the top quartile of the distribution of competitive crowding and to 0 otherwise.3
We measure niche status based on an individual’s exposure, through list co-memberships, to
high-ranking colleagues. In ISCO, a high-ranking employee is defined as someone in an executive-level
salary band. These individuals represented the top 5% of employees in ISCO. In BTCO, a high-ranking
employee is defined as a laboratory-head. These individuals, who were all in managerial and executive
positions, comprised the top 9% of employees. Using these definitions, we define a vector, E:
𝑒𝑖 = 1 𝑖𝑓 𝑖 𝑖𝑠 ℎ𝑖𝑔ℎ 𝑟𝑎𝑛𝑘𝑖𝑛𝑔; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Next, we calculate for each list, k, the proportion, p k , of high-ranking individuals on the list:
𝑝𝑘 =
∑𝑖 𝑎𝑖,𝑘∗ 𝑒𝑖
∑𝑖 𝑎𝑖,𝑘
The status of the niche occupied by actor, i, is then given by:
𝑆𝑡𝑎𝑡𝑢𝑠𝑖 =
∑𝑘 𝑝𝑘
∑𝑘 𝑎𝑖,𝑘
That is, the status of an actor’s niche in the activity space increases in the mean proportion of
high-ranking colleagues across all lists to which the actor belongs. We again create an indicator, which is
set to 1 for people in the top quartile of this distribution, and to 0 otherwise.
3
This simple spline offers greater flexibility than a linear effect and does not introduce the correlation of a quadratic
specification. In all of the analyses, we use the 75th percentile as the cut point. Although this choice is necessarily
arbitrary, we have replicated all regressions using the 90th percentile as the cut point for each covariate, and we have
also run the regressions with linear effects. There are very modest differences across these specifications, and the
substantive conclusions are unchanged.
16
Activities vary in the extent to which they bring together individuals who are otherwise unlikely
to interact. We constructed two measures of the diversity of employees’ niches. The first reflects an
individual’s exposure, through list co-membership, to colleagues from different organizational units. The
second gauges exposure to colleagues at different hierarchical levels of the organization. In other work on
intraorganizational communication patterns, it has been clearly shown that there is very limited, direct
communication between individuals in different divisions, functions, and organizational levels. Therefore,
both measures reflect the extent to which the activities in which individuals participate expose them to a
broad cross section of organizational members, which they are otherwise unlikely to be in communication
with (Kleinbaum and Stuart 2012; Kleinbaum, Stuart and Tushman 2012). We derived these measures
based on the “Blau index” of heterogeneity (Blau 1977a; Blau 1977b). For each list, k, we calculate the
proportion of members, 𝑝𝑢 , across all, U, organizational units or hierarchical levels. The resulting
measure is:
"𝐵𝑙𝑎𝑢" 𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖 =
∑𝑘
1 − ∑𝑢 𝑝𝑢2
∑𝑖 𝑎𝑖,𝑘
∑𝑘 𝑎𝑖,𝑘
That is, for each distribution list, we sum the squares of proportions of members from each
organizational unit or hierarchical level, which we then subtract from one. We then divide this quantity by
the size of the distribution list. We chose to denominate by the size of the list because we believe that that
larger lists, such as the one that encompasses all employees in an entire division of the organization,
represent less meaningful social groups than smaller, more intimate ones. In other words, we assume that
high diversity in a smaller group is more likely to convey meaningful exposure to diverse actors than is a
wide range of participants in a large group. 4 Finally, we compute the mean of this measure across all lists
of which person i was a member. We again create indicators based on these measures, set to 1 for people
in the top quartile of the distribution, and to 0 otherwise.
4
Our operative assumption is that in large groups, there is ample opportunity for members with common
backgrounds and organizational affiliation to band together, so that relatively homogenous sub-groups may form in
large activities. When diverse participants engage in a small group, however, we anticipate that a greater amount of
mixing will occur.
17
To measure an individual’s distance from the identity blueprint for a given job role, we calculate
the distance between the position vector for each actor, i, and the “prototypical” actor in the job role that i
occupies. We begin by selecting common job roles in each organization, which define the set of
individuals for whom a prototypical actor was identified. For ISCO, which had a large number of distinct
job roles, we followed this procedure for the ten most commonly occurring job roles. Examples include
Sales Territory Manager and Software Engineer. For BTCO, we also applied this procedure across ten job
roles, as indicated by salary-grades. 5 For each of these job roles, we define a vector, l, the elements of
which were defined as:
𝑙𝑘 = 𝑝𝑟𝑜𝑝𝑜𝑟𝑡𝑖𝑜𝑛 𝑜𝑓 𝑖𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑠 𝑖𝑛 𝑗𝑜𝑏 𝑟𝑜𝑙𝑒 𝑤ℎ𝑜 𝑏𝑒𝑙𝑜𝑛𝑔 𝑡𝑜 𝑙𝑖𝑠𝑡, 𝑘
We then compute the distance between the vector, a, representing the distribution lists to which
actor, i, belongs and the vector, l. The result is a measure of each actor’s distance from the identity
blueprint for the job role:
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑟𝑜𝑚 𝐵𝑙𝑢𝑒𝑝𝑟𝑖𝑛𝑡𝑖 = �
𝑘
(𝑎𝑘 − 𝑙𝑘 )2
𝑙𝑘
We create an indicator, set to 1 for people in the top quartile of this distribution, and to 0 otherwise.
To interpret the effects of individuals’ niches, it is necessary to control for employees’ overall
level of engagement with the activities in the firm. To account for the skewed distribution of this measure,
we created indicators for individuals in each quartile of the distribution of list memberships. We include
this flexible specification to take special care to guard against the possibility that the coefficients for the
niche characteristics just pick-up a misspecification of the functional form of the relationship between the
volume of work in which employees participate and the outcome variables.
In addition, we included an indicator for sex (set to 1 for females) and multiple dummy variables
for employee rank and for job function. (The number of dummies varied between the two organizations
based on differences in their organizational structures). We also controlled for employee tenure using
5
Individuals in the job roles for which we were able to measure identity blueprints at ISCO do not correspond to
those for whom we were able to obtain e-mail archives. Therefore, we were unable to examine the association
between the measure of atypicality and indegree at ISCO.
18
linear and quadratic terms, following the typical specification in studies that estimate earnings equations.
Finally, in all regressions using the BTCO data, we included indicators for employees’ highest
educational degree attained. (Educational attainment was not available for employees of ISCO.)
Estimations
We modeled the relationships among employee niche characteristics and three dependent variables. For
analyses of attainment in BTCO, the response variable, the proportion of target bonus actually paid, was
normally distributed. We therefore estimated models of target bonuses using OLS regressions with robust
standard errors, which we clustered at the laboratory level because the lab head was the primary decisionmaker in the setting of individual bonuses. For analyses of attainment in ISCO, the dependent variable
was employee performance rating on a 5-point scale. We used ordered logistic regression with robust
standard errors. Finally, in the analyses of indegree centrality in both organizations, we used Poisson
Quasi-Maximum Likelihood regressions with robust standard errors (Wooldridge 1997). 6
Results
Table 1 reports descriptive statistics. Of the 4,661 employees in the data from ISCO, 53.2% were female.
The mean tenure was 8.5 years. The mean performance rating was 3.5 on a 5-point scale. Of the 979
employees in BTCO’s research division, 50.3% were female. The mean tenure at BTCO, 6.5 years, was
slightly lower than ISCO. Despite the many differences between the two organizations, we find surprising
consistency in employees’ levels of participation in the activities space. At ISCO, the average employee
was a member of 12.2 distribution lists; at BTCO, the typical staff member belonged to 12.1 lists. In both
organizations, the variance in list memberships was approximately half of the mean, reflecting significant
heterogeneity in employees’ levels of engagement in the activity spaces of each enterprise.
- Table 1 about here Because these data have not been previously used in organizational analysis, we describe in Table
2 the composition of the mailing lists in BTCO in some detail. (Recall that the names of the mailing lists
6
QML standard errors are consistent even if the underlying data-generating process is not Poisson. The PQML
estimator can be used for any non-negative dependent variable, whether integer or continuous, and it does not
depend on the assumption of the equivalence of the mean and the variance of the event count.
19
were encrypted in the ISCO data, which means that we only are able to report this information for
BTCO.) As previously noted, based on the titles of the distribution lists at BTCO, we categorize them into
three broad types, which we label social, organizational, and workflow.7 Social lists ranged from
recreational activities to carpooling. Organizational lists included same-building members, as well as
formal organizational units. Workflow lists at BTCO often were linked to specific drug targets and
molecular pathways. There were vivid differences in the membership rosters across these three categories.
On average, organizational and social list membership rosters were much larger than those of workflow
lists. The mean size of workflow groups, 8.9 members, was less than half that of organization and social
groups. Not surprisingly, organizational lists matched the overall demographics of the research division.
In contrast, social lists comprised younger, ethnically diverse, and less-well-educated members.
Workflow lists, on the other hand, were male- and White-dominated, with an older membership, a higher
fraction of doctoral degree holders and a high degree of functional and rank diversity.
- Table 2 about here Turning to a test of our hypotheses, the first prediction posits a negative association between the
competitive crowding of employees’ intraorganizational niches and the level of individual attainment and
indegree centrality in the two companies. We report these results in Table 3. Models 1-4 are for ISCO,
while models 5-8 are for BTCO. In the baseline regression for individual attainment in ISCO (Model 1),
women have higher performance ratings and employee tenure exhibits the typical, non-monotonic pattern.
In our spline specification of the number of distribution lists in which an employee participates, we find a
monotonic increase in performance for individuals who were deeply immersed in the organization’s
activity space. The odds of receiving a higher, rather than lower, performance rating are 1.52 times higher
for someone who is in the top quartile of the count of email distribution list memberships relative to
someone in the lowest quartile. Model 2 adds the competitive crowding of an employee’s niche. The
7
In many instances, the title of a mailing list did not definitively indicate its type. For instance, some lists had
uninterruptable titles like, “6789-f”. We were conservative in assigning lists to categories: we assigned them only
when we were virtually certain of their type. Based on conversations with the company, however, we believe that
the majority of the unclassified lists were in the workflow category.
20
parameter estimate is negative and statistically significant. The odds of receiving a higher, rather than
lower, performance rating were multiplied by 0.865 for someone in the top quartile of competitive
crowding relative to the rest of the distribution.
Turning to the second outcome variable, indegree in the intra-company direct-tie email network,
Model 3 presents the baseline results for ISCO. (This model has many fewer observations than the
performance evaluation regressions because we were only able to gather email data for a subset of senior
employees at ISCO.) Model 3 shows that the number of distribution lists to which an employee belongs is
also positively associated with an employee’s indegree in the internal email network. In interpreting this
result, recall that the measure of centrality excludes all list-based and mass messages, so there is no
tautological relationship between list memberships and indegress centrality. A person in the top quartile
of email distribution list memberships has a 32% higher predicted indegree than a person in the bottom
quartile of distribution lists. Model 4 adds the indicator for competitive crowding, and we again obtain a
negative, statistically significant coefficient. Those in the top quartile of the distribution of competitive
crowding have a 52% lower predicted indegree than those not in the top quartile.
We find a similar pattern of results with respect to competitive crowding in BTCO. In the
baseline regression of annual bonus, the only significant covariate is the number of distribution lists. As
in ISCO, employees in the third and fourth quartiles of number of distribution list memberships have
higher annual bonuses (Models 5 and 6) and indegree centralities (Models 7 and 8). Relative to people in
the bottom quartile of email distribution lists, those in the top quartile are predicted to have a 20.4%
higher ratio of actual-to-target-bonus paid and a 112% higher predicted indegree. Models 6 and 8 include
the competitive crowding indicator. Model 6 shows that the coefficient is negative and significant at the
10-percent level in the target bonus regression: individuals in niches in the top quartile of the competitive
crowding distribution garner a ratio of actual-to-target bonus that is 9% lower than those not in the top
quartile. Model 8 shows that, like at ISCO, employees in crowded niches at BTCO are less central in the
intra-company email network. Those in the top quartile of the competitive crowding distribution have a
predicted indegree that is 34% lower than those not in the top quartile. Taken together, these results
21
support the hypothesis that high niche crowding correlates with decreased attainment and less central
positions in the communication network.
- Table 3 about here Hypothesis 2 proposes that individuals in high status intraorganizational niches will experience
higher attainment levels and degree centrality. Table 4 reports these results. Because Table 3 includes
results from the baseline model, we do not replicate them in this or subsequent tables. Model 9 adds the
indicator for niche status to the baseline performance model for ISCO. The coefficient is positive and
significant. The odds of receiving a higher rather than lower performance rating grow by 1.42 for those in
the top quartile of the niche status distribution relative to those not in the top quartile. Model 10 reports
the ISCO indegree regression, and here too we find a very strong effect of the status of the niche on the
level of centrality. The parameter estimate suggests that those in the top quartile of the distribution of
niche status have a predicted indegree that is 132% greater than those not in the top quartile.
The findings for BTCO also strongly support hypothesis 2. Model 11 includes the status indicator
in the regression of target bonus. The status indicator is positive and statistically significant. Lastly,
Model 12, which adds the status indicator to the regression of indegree centrality, is also positive at the
p<.001 significance level. At BTCO, those in top quartile of the distribution of status have a predicted
ratio of actual-to-target bonus that is 17% higher than those not in the top quartile. Likewise, they
experience a predicted indegree that is 89% higher. Taken together, the findings from both organizations
provide strong support for Hypothesis 2.
- Table 4 about here The third hypothesis posits that it is advantageous to individuals to occupy a functional- or rankdiverse niche in the intraorganizational activities space. We address this hypothesis in Table 5. In Model
13, we add function and rank diversity indicators to the baseline attainment model for ISCO. 8 The
coefficients for both are positive and strongly significant (p<.001). The odds ratios associated with these
8
For ISCO, including just function or rank diversity alone does not change our results. For BTCO, including rank
diversity alone results in a positive, significant coefficient.
22
coefficients are 1.313 (function diversity) and 1.237 (rank diversity). In Model 14, we add the function
and rank diversity indicators to the baseline regressions for indegree centrality at ISCO. Once again, both
coefficients are positive and highly significant (p<.001). Those in the top quartile of function and rank
diversity have a predicted indegree that is 62% and 59% higher, respectively, than those not in the top
quartile. Comparable results were obtained in the BTCO data. In Model 17, we add the function and rank
diversity indicators to the baseline target bonus model. Those in the top quartile of the function diversity
distribution had an actual-to-target bonus payout that is predicted to be 20%, respectively, than those not
in the top quartile. We did not observe significance for rank diversity. Model 18 adds the function and
rank diversity indicators to the baseline indegree centrality model. We find that those in top quartile of the
distribution of function or rank diversity have an estimated indegree that is predicted to be 94% and 109%
higher, respectively, than those not in the top quartile. Taken together, findings from both organizations
provide strong support for Hypothesis 3. Models 15, 16, 19, and 20 present results of regressions
designed to test the first three hypotheses simultaneously. Although the coefficients for competitive
crowding and rank diversity are not significant in some specifications, the results are largely consistent
with those reported above. 9
- Table 5 about here The fourth hypothesis proposes that people who deviate from the identity blueprint for their job
role will achieve lower levels of attainment. (Recall that data limitations preclude us from testing this
hypothesis using the e-mail centrality measures.) Table 6 reports results for this hypothesis. In Model 21,
which adds the indicator for distance from the role identity blueprint to the regression of employee
performance ratings at ISCO, the parameter estimate for this measure is negative and statistically
significant. The odds of receiving a higher performance rating were multiplied by 0.721 for someone in
the top quartile of distance from the identity blueprint for his or her job role, relative to someone not in
the top quartile. Model 22 replicates the analysis for BTCO. In this case, the indicator for distance from
9
We do not include typicality in these combined models because doing so would result in a different – and
significantly smaller – sample size than in the models reported above.
23
the role identity blueprint is not statistically significant. Therefore, taken together, the evidence for the
argument that there is a benefit to niche typicality is suggestive but inconclusive.
- Table 6 about here Discussion: Concerns about Causality
Although the results we report broadly are consistent with the theory we develop, the limitations of the
data create a number of empirical concerns. Most significantly, because the number and memberships of
distribution lists in both organizations changed only modestly over the observation window, our
regressions were run in the cross-section. Therefore, we cannot discount the possibility that unobserved
individual differences influence both the niches individuals occupy in the emergent role space and their
attainment levels. In fact, the coefficients in many of the regressions are large, and it would not be a
surprise to discover that the magnitudes are partly attributable to unobserved correlates. With these data,
we cannot identify causal effects, but future research designs can improve upon our current, suggestive
results. First, datasets that span longer periods of time will enable the inclusion of employee fixed effects,
so that findings are based on within-person changes in niche characteristics that arise as the dual mode,
employee-to-list matrices change. Modeling the data with person-specific intercepts would have the
obvious benefit of accounting for all time-constant employee characteristics.
Second, we also can envision a series of studies that exploit the fact that distribution lists lend
themselves to analysis of the emergence and articulation of the niche space itself. As we consider
alternative interpretations for the findings in this paper, we believe that the most uncertain result concerns
the effect of niche status. This finding is questionable because we know that high status groups preserve
their status by maintaining exclusive membership criteria. Oftentimes, only individuals who stand in the
good graces of the elites in the organization are asked to participate in the activities that create high status
niches. In fact, it would be entirely consistent with our general understanding of status processes if much
of the effect of memberships in high status lists is due to an underlying, unobserved assignment process
that matches individuals with certain, desirable characteristics to activities that include high status
members.
24
With longitudinal data, it may be possible to study the assignment / matching of individuals to
mailing lists, and to exploit the understandings that emerge to estimate a causal effect of list memberships
on subsequent outcomes. Although we cannot undertake these analyses with the data available to us, there
is one step that may bring us nearer to a true estimate. In our framework, changes in person-specific niche
characteristics result from a few different processes. They occur when new distribution lists are created
and when no-longer-used ones are culled; they result from changes to the membership rosters of existing
lists; and they occur when there are changes to the employment statuses (promotions, departmental
reassignments, etc.) of the members of pre-existing lists. We believe that empirical support for a causal
effect of niche status is stronger if the econometric identification is solely based on variation in niche
status that occurs through promotions among members of pre-existing lists. This empirical strategy
excludes all variation that arises when individuals are assigned to high status lists for reasons we cannot
observe, and therefore represents a more conservative test of the prediction.
The challenge in implementing this empirical strategy is, once again, the short time frame
spanned by the distribution list data, and therefore the limited number of promotions that occur within the
observation window. Nonetheless, when we conduct this analysis (for brevity, table not reported), we find
in both organizations that when promotions to positions of high status are earned by co-participants in the
activities in which a focal individual engages, that person experiences higher performance appraisals.
While admittedly still short of causal evidence, this result lends credence to the test of hypothesis 2.
Extension—The Sociodemographic Composition of Intraorganizational Niches
The approach in the paper opens up several lines of inquiry. One that particularly intrigues us is the
opportunity to examine the extent to which intraorganizational niches are sex segregated and, if so, what
are the consequences for inequality within organizations? A descriptive analysis of the two datasets yields
some very interesting, preliminary results. First, despite relatively equal gender representation in the
overall samples in both organizations—ISCO and BTCO are 53% and 50% female, respectively—many
activities fall at the two extremes of the sex composition distribution. While the overall mean sex
compositions of list memberships almost exactly parallels the gender distribution of the two datasets
25
(implying that men and women participate in activities in the two organizations at approximately equal
rates), it is nevertheless the case that 100% male or 100% female lists are common in both organizations.
Figure 1 reports histograms of proportion male memberships for email distribution lists in both
organizations. In BTCO, more than 27% of the email distribution lists are single sex: these activities
either are 100% male or 100% female. In ISCO, the numbers also are substantial: more than a fifth of the
activities in ISCO are single sex. Therefore, like previous research that shows higher degrees of sex
segregation as the analytical lens is moved from the level of an overall organization down to the
occupancy of specific roles within organizations (cf. Baron 1984), it appears that there is significant sex
segregation in the activity space, as well as in the occupancy of specific job titles.
- Figure 1 about here We suspect as well that sex segregation in the space of enacted roles may affect levels of
attainment and, more generally, access to organizational resources. As preliminary evidence, we
computed the gender composition of lists of varying degrees of status. To illustrate, we classified lists
into those that are high status and those that are not. We defined high status lists to be those in the top
quartile of the distribution of “percent high status members.” We then looked at gender composition
across the status divide. In both organizations, there is a large, statistically very significant (p<0.001)
difference in gender composition across the high / low status dichotomy: high status lists have a much
higher proportion of male members. Moreover, this difference is not simply driven by the fact that men
dominate the ranks of the high status members of both organizations. To verify this, we calculated the
percent of male members on all lists, excluding all high status individuals. In other words, we calculated
the gender composition of low status members of all email lists. In this comparison too, we find that, for
high status lists, a much greater percentage of the non-high-status members are male.
Conclusion
The goal of this article has been to focus an ecological lens on intraorganizational attainment. Drawing on
ecological insights about how niche characteristics script competitive and symbiotic dynamics in
populations of organizations, we argue that individuals within firms occupy niche spaces that influence
26
the resources they obtain. These intra-organizational niches are defined by positioning individuals in
recurring organizational activities, which sometimes correspond to the organization’s formal structure,
but also derive from cross-functional project teams, task forces, and informal work and social groups that
crisscross the formal structure. We identify four niche characteristics and theorize about their relationship
to individual attainment. These features are the extent to which a person-specific niche is, (1)
competitively crowded, (2) a gateway to high status actors in the organization, (3) a conduit to individuals
in possession of dissimilar knowledge, skills, and talents, and (4) proximate to a prototypical
representation for the individual’s work role. We then propose a novel solution to the empirical challenge
of this line of inquiry: How does one observe the myriad forms of recurring activity that occur in
organizations? Our approach is to characterize niche positions based on an affiliation network derived
from the complete roster of electronic mailing lists within organizations.
The findings in the paper are remarkably consistent across two, quite-different empirical settings,
and two, quite different measures of performance. In both organizations and for both dependent variables,
we find that people who occupy competitively crowded niches occupy less central positions in the
communication network and achieve lower levels of attainment, whereas those who occupy high status
occupy more central positions in the communication network and receive more positive evaluations.
Likewise, using two different measures of niche diversity, we find that individuals who are exposed to a
broader cross section of members of their organization’s rank hierarchy and landscape of organizational
units are more central in the network and evaluated more positively.
Our empirical analysis reveals only partial support for the hypothesis that departing from the
identity blueprint for job roles correlates with lower levels of attainment. We find this result in the
information services company, but not in the biopharmaceutical lab. Of course, one possibility is that this
hypothesis simply is not borne out in the BTCO data. However, we acknowledge that of the four
measured niche characteristics, we are least confident in our ability to accurately gauge typicality. This is
because the concept is perceptual; a person is typical or not insofar as she conforms to audience members’
expectations of a given role. We should only anticipate a penalty for deviating from a blueprint when
27
there is clarity and consensus in perceptions of the typical profile of the occupant of the role (cf. Baron
2004; Ruef and Patterson 2009; Kovács and Hannan 2010). Absent clarity and consensus, identity
elements scatter and we would not expect sanctions for deviating from an ill-defined norm. It is likely that
some of the job titles in the two organizations have sharper identity elements attached to them, while
others are broad to the point that there are few widely held assumptions about appropriate roles for that
particular job title. It is difficult for us to know whether or not the jobs we can identify in the data are
those for which role expectations are narrowly and clearly defined.
The article makes three primary contributions. First, it builds our understanding of ecological
processes in intra-organizational settings. Whereas prior research in this tradition has considered the
interplay of ecology and organizational change (e.g., Amburgey, Kelly, and Barnett 1993), our study
demonstrates that the same niche characteristics that influence the outcomes of resource competition in
populations of organizations also affect the attainment of individuals within particular organizations. It
also offers partial support for the notion that conformance to identity blueprints, which figures
prominently in contemporary ecological research, may influence intraorganizational attainment.
Second, the study adds to our understanding of the organizational bases of inequality. In
particular, characterizing an individual’s niche in the web of recurring organizational activities provides a
new window into the organization’s stratification system.
Finally, we introduce a novel data source that has broad applicability in organizational research.
In particular, email distribution lists may provide a means for researchers to “dust the fingerprints of
informal organization” (Nickerson and Silverman 2009: 538). They help uncover, and can be used to
position individuals within, the less formal and the informal organizational structure, including work
groups, task forces, and social clubs. Moreover, they provide an efficient and unobtrusive means to
collect such data. In sum, this study lays the conceptual and empirical foundation for future research on
the effects of intraorganizational niches and for deeper investigation into their origins and dynamics.
28
References
Abbott, Andrew. 1988. The System of Professions. Chicago: The University of Chicago Press.
Amburgey, Terry L., Dawn Kelly, and William P. Barnett. 1993. "Resetting the Clock: The
Dynamics of Organizational Change." Administrative Science Quarterly 38:51.
Barnett, William P., James N. Baron, and Toby E. Stuart. 2000. "Avenues of Attainment:
Occupational Demography and Organizational Careers in the California Civil Service."
American Journal of Sociology 106:88-144.
Barnett, William P. and Anne S. Miner. 1992. "Standing on the Shoulders of Others: Career
Interdependence in Job Mobility." Administrative Science Quarterly 37:262-281.
Baron, James N. 1984. "Organizational Perspectives on Stratification." Annual Review of
Sociology 10:37-69.
—. 2004. "Employing Identities in Organizational Ecology." Industrial and Corporate Change
13:3-32.
Barron, David D., Elizabeth West, and Michael T. Hannan. 1994. "A Time to Grow and a Time
to Die: Growth and Mortality of Credit Unions, 1914-1990." American Journal of
Sociology 100:381-421.
Baum, Joel A. C. and Heather A. Haveman. 1997. "Love Thy Neighbor? Differentiation and
Agglomeration in the Manhattan Hotel Industry, 1898-1990." Administrative Science
Quarterly 42:304-338.
Baum, Joel A. C. and Stephen J. Mezias. 1992. "Localized Competition and Organizational
Failure in the Manhattan Hotel Industry, 1898-1990." Administrative Science Quarterly
34:580-604.
Baum, Joel A. C. and Jitendra V. Singh. 1994. "Organizational Niches and the Dynamics of
Organizational Mortality." American Journal of Sociology 100:346-380.
Bergmann, Barbara R. 1986. The Economic Emergence of Women. New York: Basic Books, Inc.
Bidwell, Matthew and Isabel Fernandez-Mateo. 2010. “Relationship Duration and Returns to
Brokerage in the Staffing Sector.” Organization Science 21:1141-1158.
Blau, Peter M. 1977a. Inequality and Heterogeneity: A Primitive Theory of Social Structure.
New York: Free Press.
—. 1977b. "A Macrosociological Theory of Social Structure." American Journal of Sociology
83:26-54.
Brass, Daniel J. 1995. "A Social Network Perspective on Human Resources Management."
Research in Personnel and Human Resources Management 13:39-79.
Burt, Ronald S. 1987. "Social Contagion and Innovation: Cohesion Versus Structural
Equivalence." American Journal of Sociology 92:1287-1335.
—. 1992. Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard
University Press.
—. 1997. "The Contingent Value of Social Capital." Administrative Science Quarterly 42:339365.
—. 2004. "Structural Holes and Good Ideas." American Journal of Sociology 110:349-399.
Carroll, Glenn R. 1985. "Concentration and Specialization: Dynamics of Niche Width in
Populations of Organizations." American Journal of Sociology 90: 1262-1283.
Carroll, Glenn R. and Anand Swaminathan. 2000. "Why the Microbrewery Movement?
Organizational Dynamics of Resource Partitioning in the U.S. Brewing Industry."
American Journal of Sociology 106:715-762.
29
DiMaggio, Paul. 1986. "Structural Analysis of Organizational Fields: A Blockmodel Approach."
Research in Organizational Behavior 8:335-370.
Dobrev, Stanislav D. , Tai-Young Kim, and Glenn R. Carroll. 2002. "The Evolution of
Organizational Niches: U.S. Automobile Manufacturers, 1885-1981." Administrative
Science Quarterly 47:233-264.
Dobrev, Stanislav D., Tai-Young Kim, and Michael T. Hannan. 2001. "Dynamics of Niche
Width and Resource Partitioning." American Journal of Sociology 106:1299-1337.
Fleming, Lee, Santiago Mingo, and David Chen. 2007. "Collaborative Brokerage, Generative
Creativity, and Creative Success." Administrative Science Quarterly 52:443-475.
Freeman, John and Michael T. Hannan. 1983. "Niche Width and the Dynamics of Organizational
Populations." American Journal of Sociology 88:1116-1145.
Granovetter, Mark S. 1973. "The Strength of Weak Ties." American Journal of Sociology
78:1360-1380.
Hannan, Michael T., Glenn R. Carroll, and László Pólos. 2003. "The Organizational Niche."
Sociological Theory 21:309-340.
Hannan, Michael T. and John Freeman. 1977. "The Population Ecology of Organizations." The
American Journal of Sociology 82:929-964.
—. 1987. "The Ecology of Organizational Founding: American Labor Unions, 1836-1985."
American Journal of Sociology 92:910-943.
—. 1989. Organizational Ecology. Cambridge, MA: Harvard University Press.
Hsu, Greta. 2006. "Jacks of All Trades and Masters of None: Audiences' Reactions to Spanning
Genres in Feature Film Production." Administrative Science Quarterly 51:420-450.
Hsu, Greta, Michael T. Hannan, and Özgecan Koçak. 2009. "Multiple Category Memberships in
Markets: An Integrative Theory and Two Empirical Tests." American Sociological
Review 74:150-169.
Hsu, Greta and Joel M. Podolny. 2005. "Critiquing the Critics: An Approach for the
Comparative Evaluation of Critical Schemas." Social Science Research 34:189-214.
Hutchinson, G. Evelyn. 1957. "Concluding Remarks." Cold Spring Harbor Symposium on
Quantitative Biology 22:415-427.
Jensen, Michael. 2003. "The Role of Network Resources in Market Entry: Commercial Banks'
Entry into Investment Banking, 1991-1997." Administrative Science Quarterly 48:466497.
Kleinbaum, Adam and Toby E. Stuart, “Inside the Black Box of the Corporate Staff: Social
Networks and the Implementation of Corporate Strategy” Tuck School of Business
Working Paper No. 2011-90.
Kleinbaum, Adam, Toby E. Stuart and Michael Tushman, “Discretion Within the Constraints of
Opportunity: Gender Homophily and Structure in a Formal Organization” Tuck School of
Business Working Paper No. 2011-95.
Kovács, Balázs and Michael T. Hannan. 2010. "The Consequences of Category Spanning
Depend on Contrast." Pp. 175-201 in Categories in Markets: Origins and Evolution
(Research in the Sociology of Organizations), vol. 31, edited by G. Hsu, G. Negro, and
Ö. Koçak.
Mark, Noah P. 2003. "Culture and Competition: Homophily and Distancing Explanations for
Cultural Niches." American Sociological Review 68:319-345.
30
McCain, Bruce E., Charles A. O'Reilly, III, and Jeffrey Pfeffer. 1983. "The Effects of
Departmental Demography on Turnover: The Case of a University." Academy of
Management Journal 26:626-641.
McPherson, J. Miller. 1983. "An Ecology of Affiliation." American Sociological Review 48:519532.
Merton, Robert K. 1968. Social Theory and Social Structure. New York: Free Press.
Nickerson, Jackson A. and Brian S. Silverman. 2009. "New Frontiers in Strategic Management
of Organizational Change." in Economic Institutions of Strategy, edited by J. A.
Nickerson and B. S. Silverman. Bingley, England: Emerald Jai.
Phillips, Damon J. and Ezra Zuckerman. 2001. "Middle Status Conformity: Theoretical
Restatement and Empirical Demonstration in Two Markets." American Journal of
Sociology 107:379-429.
Podolny, Joel M. 1993. "A Status-Based Model of Market Competition." American Journal of
Sociology 98:829-872.
—. 2005. Status Signals: A Sociological Study of Market Competition Princeton, NJ: Princeton
University Press.
Podolny, Joel M. and Damon J. Phillips. 1996. “The Dynamics of Organizational Status.”
Industrial and Corporate Change 5: 453-471.
Podolny, Joel M., Toby E. Stuart, and Michael T. Hannan. 1996. "Networks, Knowledge, and
Niches: Competition in the Worldwide Semiconductor Industry, 1984-1991." American
Journal of Sociology 102:659-689.
Popielarz, Pamela A. and Miller J. McPherson. 1995. “On the Edge or In Between: Niche
Position, Niche Overlap, and the Duration of Voluntary Association Memberships.”
American Journal of Sociology 101: 698-720.
Popielarz, Pamela A. and Zachary P. Neal. 2007. "The Niche as a Theoretical Tool." Annual
Review of Sociology 33:65-84.
Reed, Theodore L. 1978. "Organizational Change in the American Foreign Service, 1925-1965:
The Utility of Cohort Analysis." American Sociological Review 43:404-421.
Ruef, Martin. 2000. "The Emergence of Organizational Forms: A Community Ecology
Approach." American Journal of Sociology 106:658-714.
Ruef, Martin and Kelly Patterson. 2009. "Credit and Classification: The Impact of Industry
Boundaries in Nineteenth-century America." Administrative Science Quarterly 54: 486520.
Rotolo, Thomas and J. Miller McPherson. 2001. "The System of Occupations: Modeling
Occupations in Sociodemographic Space." Social Forces 79:1095-1130.
Simmel, Georg. 1902. "The Number of Members as Determining the Sociological Forum of the
Group." American Journal of Sociology 8:1-46.
Sørensen, Jesper B. 1999. "The Ecology of Organizational Demography: Managerial Tenure
Distributions and Organizational Competition." Industrial and Corporate Change 8:713744.
Sorenson, Olav and Pino Audia. 2000. "The Social Structure of Entrepreneurial Activity:
Geographic Concentration of Footwear Production in the United States, 1940–1989."
American Journal of Sociology 106:424-462.
Sorenson, Olav and Toby E. Stuart. 2008. "Bringing the Context Back In: Settings and the
Search for Syndicate Partners in Venture Capital Investment Networks." Administrative
Science Quarterly 53:266-294.
31
Stewman, Shelby and Suresh L. Konda. 1983. "Careers and Organizational Labor Markets:
Demographic Models of Organizational Behavior " American Journal of Sociology
88:637-685.
Stuart, Toby E. 1998. "Network Positions and Propensities to Collaborate: An Investigation of
Strategic Alliance Formation in a High-Technology Industry." Administrative Science
Quarterly 43:668-698.
Stuart, Toby E., Ha Hoang, and Ralph C. Hybels. 1999. "Interorganizational Endorsements and
the Performance of Entrepreneurial Ventures." Administrative Science Quarterly 44:315349.
Tortoriello, Marco, Ray Reagans, and Bill McEvily. 2012 “Bridging the Knowledge Gap: The
Influence of Strong Ties, Network Cohesion, and Network Range on the Transfer of
Knowledge Between Organizational Units.” Organization Science 23:1024-1039.
White, Harrison C. 1981. "Where Do Markets Come From?" American Journal of Sociology
87:517-547.
Wooldridge, Jeffrey M. 1997. "Quasi-Likelihood Methods for Count Data." in Handbook of
Applied Econometrics, edited by M. H. Pesaran and P. Schmidt. Oxford, UK: Blackwell.
Zuckerman, Ezra W. 1999. "The Categorical Imperative: Securities Analysts and the Illegitimacy
Discount." American Journal of Sociology 104:1398-1438.
—. 2005. "Typecasting and Generalism in Firm and Market: Genre-Based Career Concentration
in the Feature Film Industry, 1933–1995." Research in the Sociology of Organizations
23:173-216.
Zuckerman, Ezra W., Tai-Young Kim, Kalinda Ukanwa, and James von Rittmann. 2003.
"Robust Identities or Non-Entities? Typecasting in the Feature Film Labor Market."
American Journal of Sociology 108:1018-1075.
32
Figure 1: Proportion of Male Membership on Distribution Lists
Panel A: ISCO distribution lists.
Panel B: BTCO distribution lists.
Note: 2,245 ISCO lists shown.
Note: 455 BTCO lists shown.
33
Table 1: Descriptive Statistics
ISCO
BTCO
Std.
Std.
Mean
Min
Max
Mean
Min
Dev
Dev
*
Performance
3.502
0.715
1
5
1.054
0.266
0
**
Indegree
15.0
13.5
1
59
20.1
15.0
0
Competitive Crowding – Same Supervisor***
0.062
0.098
0
1
0.183
0.209
0
Status – Seniority
0.229
0.131
0
1
0.093
0.053
0.027
Diversity – Function
0.051
0.111
0
2.843
0.142
0.232
0.001
Diversity – Level
0.178
0.249
0
5.664
0.238
0.320
0.001
Distance from Role Identity Blueprint****
2.144
16.870
0
291.0
231.5
257.6
6.8
Female
0.532
0.499
0
1
0.503
0.500
0
Number of Lists
12.220
6.786
2
65
12.128
6.524
2
Tenure
8.549
6.611
0.523
42.3
6.474
5.930
1
N=4661 for ISCO and 1028 for BTCO, unless otherwise noted; * N=964 for BTCO performance; ** N=156 for ISCO
Indegree; *** N=979 for BTCO Crowding; *** N=386 for ISCO
Max
2.5
116
1
0.415
2.095
2.723
2485.2
1
47
31
Table 2: Distribution List Descriptive Statistics
Social Lists
(e.g., Sports, Commuting)
Organizational Lists
Workflow Lists
(e.g., Depts, Labs,
(e.g.,Molecules, Molecular
Buildings)
Pathways)
Avg. # of list members
20.8 (31.2)
21.3 (29.1)
8.9 (12.7)
Mean age
36.0 (4.6)
38.8 (3.7)
42.3 (5.3)
Mean tenure
4.7 (2.4)
5.5 (2.7)
7.4 (5.3)
Mean % married
48 (24.2)
64 (22.7)
74 (27.3)
Mean %female
50 (17.4)
53 (23.4)
35 (29.8)
Mean %White
47 (30.4)
54 (23.9)
64 (32.9)
Mean %PhD
33 (19.7)
51 (31.0)
74 (26.3)
Mean Status – Senior Colleagues
.006 (.013)
.006 (.016)
.060 (.105)
Mean Functional Diversity
.065 (.045)
.062 (.093)
.085 (.091)
Mean Rank Diversity
.068 (.048)
.084 (.078)
.135 (.090)
Total # of unique individuals
278
802
342
Total # of lists
18
84
71
Note: Where indicated, list means are shown, with standard deviations in parentheses. There are no statistics for crowding, as this is not a list-level
measure.
34
Table 3: Regressions of Performance and Indegree on Covariates, Including Competitive Crowding
(1)
(2)
(4)
(5)
(6)
(7)
(8)
BTCO
PoissonOLS
OLS
PoissonPoissonQML
QML
QML
Dependent Variable
Indegree
Performance Performance
Indegree
Indegree
Crowding – Same Supv.
-0.538*
-0.027†
-0.138†
(Top 25%)
(0.244)
(0.014)
(0.073)
0.356***
-0.067
-0.056
-0.004
-0.004
0.035
0.039
Female
(0.058)
(0.140)
(0.140)
(0.017)
(0.017)
(0.044)
(0.044)
0.023
0.093
0.053
0.026
0.024
0.009
0.005
No. Lists – 25-50%
(0.091)
(0.235)
(0.231)
(0.018)
(0.017)
(0.083)
(0.086)
0.242*
0.463*
0.449*
0.045**
0.040**
0.249**
0.230**
No. Lists –50-75%
(0.096)
(0.227)
(0.226)
(0.015)
(0.015)
(0.089)
(0.094)
0.422***
0.671*
0.645*
0.175***
0.168***
0.617***
0.590***
No. Lists–75-100%
(0.105)
(0.259)
(0.254)
(0.024)
(0.024)
(0.095)
(0.103)
0.026†
-0.059†
-0.068*
0.007
0.006
0.029*
0.027*
Tenure
(0.014)
(0.033)
(0.032)
(0.005)
(0.005)
(0.013)
(0.013)
-0.001*
0.001†
0.002*
-0.0003
-0.0003
-0.001
-0.001
Tenure-squared
(0.000)
(0.001)
(0.001)
(0.0002)
(0.0002)
(0.000)
(0.000)
2.736***
2.823***
0.923***
0.940***
2.402***
2.480***
Constant
See notes
See notes
(0.158)
(0.172)
(0.030)
(0.031)
(0.086)
(0.101)
Observations
4661
4661
156
156
915
915
979
979
-4955
-4953
-1100
-1091
-5263
-5239
Log-pseudolikelihood
Note: Robust standard errors (clustered by laboratory for BTCO). Coefficients for level, education, and function indicators are included, but not
shown. Cut-points for ordered logit models are not reported. 147 error clusters in Models 5 & 6; 148 in Models 7 & 8.
†
p < .10; * p < .05; ** p < .01; *** p < .001 (two-tailed tests).
Dataset
Model
Ordered
Logit
Performance
ISCO
Ordered
Logit
Performance
-0.145*
(0.069)
0.358***
(0.058)
0.032
(0.091)
0.241*
(0.096)
0.416***
(0.104)
0.021
(0.014)
-0.001*
(0.000)
(3)
PoissonQML
Indegree
35
Table 4: Regressions of Performance and Indegree Centrality on Covariates, Including
Status
(9)
(10)
(11)
(12)
ISCO
BTCO
Model
Ordered Logit Poisson-QML
OLS
Poisson-QML
Dependent Variable
Performance
Indegree
Performance
Indegree
0.350***
0.769***
0.093**
0.323***
Status – Senior Colleagues (Top 25%)
(0.074)
(0.143)
(0.029)
(0.067)
0.384***
0.129
-0.009
0.039
Female
(0.059)
(0.138)
(0.017)
(0.045)
0.008
0.055
0.030†
0.034
Number of Lists – 25-50%
(0.091)
(0.218)
(0.018)
(0.077)
0.198*
0.181
0.038*
0.242**
Number of Lists – 50-75%
(0.097)
(0.212)
(0.016)
(0.082)
0.293**
0.263
0.160***
0.578***
Number of Lists – 75-100%
(0.108)
(0.223)
(0.026)
(0.082)
†
0.025
-0.026
0.006
0.021†
Tenure
(0.014)
(0.029)
(0.005)
(0.012)
-0.001**
0.001
-0.0003
-0.0005
Tenure-squared
(0.000)
(0.001)
(0.0002)
(0.0004)
2.484***
0.932***
2.404***
Constant
See notes
(0.172)
(0.029)
(0.086)
Observations
4661
156
964
1028
Log-pseudolikelihood
-4944
-978
-5395
Note: Robust standard errors (clustered by laboratory for BTCO). Coefficients for rank, education, and
function indicators and cut-points for ordered logit not reported. 147 error clusters in Model 11; 148 error
clusters in Model 12.
†
p < .10; * p < .05; ** p < .01; *** p < .001 (two-tailed tests).
Dataset
36
Table 5: Regressions of Performance and Indegree Centrality on Covariates, Including Diversity
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)
ISCO
BTCO
PoissonPoissonPoissonModel
OLS
OLS
QML
QML
QML
Dependent Variable
Indegree
Performance
Indegree
Performance
Indegree
Crowding – Same Supv.
-0.684**
-0.018
-0.101
(Top 25%)
(0.217)
(0.016)
(0.075)
Status – Senior
0.579***
0.076*
0.294***
Colleagues (Top 25%)
(0.139)
(0.029)
(0.066)
Diversity – Function
0.272***
0.386**
0.243*
0.069*
0.196**
0.065*
0.166**
(Top 25%)
(0.073)
(0.138)
(0.122)
(0.029)
(0.057)
(0.030)
(0.057)
Diversity – Rank (Top
0.212**
0.497***
0.363**
0.030
0.194**
0.035
0.208**
25%)
(0.079)
(0.137)
(0.134)
(0.029)
(0.063)
(0.029)
(0.060)
0.371***
-0.033
0.125
-0.005
0.037
-0.004
0.039
Female
(0.059)
(0.131)
(0.139)
(0.017)
(0.044)
(0.017)
(0.044)
-0.037
-0.047
-0.071
0.022
-0.015
0.023
-0.001
No. Lists–25-50%
(0.091)
(0.220)
(0.205)
(0.018)
(0.079)
(0.018)
(0.084)
0.137
0.235
0.098
0.026
0.016†
0.020
0.142
No. Lists–50-75%
(0.099)
(0.237)
(0.214)
(0.016)
(0.082)
(0.017)
(0.086)
0.155
0.208
0.011
0.103***
0.327***
0.086**
0.272**
No. Lists–75-100%
(0.120)
(0.258)
(0.231)
(0.029)
(0.084)
(0.029)
(0.085)
0.032*
-0.061*
-0.050
0.005
0.022†
0.003
0.014
Tenure
(0.014)
(0.029)
(0.027)
(0.005)
(0.012)
(0.005)
(.086)
-0.001**
0.002*
0.002
-0.0002
-0.0004
-0.000
-0.0003
Tenure-squared
(0.000)
(0.001)
(0.001)
(0.0002)
(0.0002)
(0.000)
(0.0004)
2.736***
2.665***
0.930***
2.439***
0.948**
2.450***
Constant
See Notes
See Notes
(0.154)
(0.173)
(0.031)
(0.081)
(0.033)
(.100)
Observations
4661
156
4661
156
916
980
916
980
Log-pseudo-likelihood
-4928
-1005
-4929
-932
-5129
-5004
Note: Robust standard errors (clustered by laboratory for BTCO). Coefficients for rank, education, and function indicators and cut-points for
ordered logit not reported. 148 error clusters in Models 17 & 19; 149 in Models 18 & 20.
†
p < .10; * p < .05; ** p < .01; *** p < .001 (two-tailed tests).
Dataset
Ordered
Logit
Performance
PoissonQML
Indegree
Ordered
Logit
Performance
-0.087
(0.070)
0.213**
(0.079)
0.205**
(0.076)
0.205**
(0.079)
0.414***
(0.059)
-0.032
(0.091)
0.078
(0.100)
0.017
(0.124)
0.029*
(0.014)
-0.001**
(0.000)
37
Table 6: Regressions of Performance on Covariates, Including Distance from
Role Identity Blueprint
(21)
(22)
ISCO
BTCO
Model
Ordered Logit
OLS
Dependent Variable
Performance Performance
-0.328*
0.042
Distant from Role Identity Blueprint (Top 25%)
(0.157)
(0.026)
-0.067
0.001
Female
(0.152)
(0.016)
0.386
0.017
Number of Lists – 25-50%
(0.252)
(0.017)
0.366
0.031*
Number of Lists – 50-75%
(0.270)
(0.016)
0.657*
0.121**
Number of Lists – 75-100%
(0.282)
(0.025)
†
0.056
0.002
Tenure
(0.031)
(0.005)
-0.002†
-0.0003
Tenure-squared
(0.001)
(0.0002)
0.973**
Constant
See notes
(0.027)
Observations
843
964
-916
Log-pseudolikelihood
Note: Robust standard errors (clustered by laboratory for BTCO). Coefficients for rank, education, and
function indicators and cut-points for ordered logit not reported. 147 error clusters in Model 22.
†
p < .10; * p < .05; ** p < .01; *** p < .001 (two-tailed tests).
Dataset
38