Career mobility and job flocking

Career mobility and job flocking

Social Social Science Research 34 (2005) 800–820 Science RESEARCH www.elsevier.com/locate/ssresearch Career mobility and job flocking Stanislav D...

229KB Sizes 0 Downloads 116 Views

Social

Social Science Research 34 (2005) 800–820

Science

RESEARCH

www.elsevier.com/locate/ssresearch

Career mobility and job flocking Stanislav D. Dobrev Graduate School of Business, University of Chicago, Chicago, IL 60637, USA Available online 11 March 2005

Abstract Flocking is oftentimes used to metaphorically describe social behavior. Do people actually flock in a way that avian species do? This paper develops a purely ecological mechanism for explaining similarity in human behavior by distinguishing between social networks of personto-person ties and ecological networks of ties of observability. I test the ‘‘information center hypothesis’’ [Ibis 115 (1973) 517] from bioecology on the career mobility of professional managers who all graduated from the same university. In this case, spatial proximity in birds is replaced with sociodemographic homophily on an acquired status characteristic among persons. The results indicate that homophilous individuals exchange information about a favorable destination state by ecological ties of observability, that is, they follow the career mobility of fellow alumni and ‘‘flock’’ along with them.  2005 Elsevier Inc. All rights reserved. Keywords: Careers; Job change; Ecology; Contagion; Imitation; Flocking

1. Introduction There are, of course, a great many kinds of collective action; the most elementary and the most pervasive is undoubtedly mass migration. Bees swarm, birds migrate, and human beings rush madly hither and yon in search for some new El Dorado or in hope of achieving somewhere a new Utopia. Park ([1939] 1972, p. 120) E-mail address: [email protected]. 0049-089X/$ - see front matter  2005 Elsevier Inc. All rights reserved. doi:10.1016/j.ssresearch.2005.01.002

S.D. Dobrev / Social Science Research 34 (2005) 800–820

801

Sociologists have long pointed out that among human individuals similarity breeds connection (Bott, 1928; Hubbard, 1929; Wellman, 1929). The assertion is central to contemporary sociological theory and evident in the classic works of Blau (1970, 1977), Lazarsfeld and Merton (1954), Hawley (1950), and Durkheim ([1893] 1933). It has been formalized in McPhersonÕs (1983) Ôhomophily principleÕ which posits that people who are similar on one or more of a variety of relevant dimensions in sociodemographic space(e.g., race, gender, education, and socioeconomic status) are more likely to socialize with others like them than with those from whom they are different. Thus, social networks emerge and serve to facilitate and perpetuate many social divides. One of the mechanisms through which this process occurs is the transfer of information among members of a social network via the ties that connect them (Katz and Lazarsfeld, 1955). The empirical evidence amassed from research on social networks supports this argument and is usually summarized by the popular proverbial wisdom: ‘‘Birds of a feather flock together’’ (Lazarsfeld and Merton, 1954; McPherson et al., 2001). The purpose of this article is to investigate if the same mechanism that accounts for flocking behavior in avian species can be applied to information exchange in an ecological network of human individuals. My goal here is not to develop new theory in a sense of developing novel predictions, but to elaborate an ecological mechanism for the contagion processes often theorized in popular sociological accounts. My main contention is that although undoubtedly the presence of direct ties among individuals in a social network shapes information exchange, this observed effect may at least partly derive from a purely ecological dynamic of observation of alter behavior within a set of ecologically proximate peers. In other words, there is a meaningful distinction between a social network of person-to-person ties and an ecological network of ties of observability. If correct, my argument would suggest that network and institutional accounts of similarity in social behavior ought to carefully explain to what extent their predicted effects obtain simply as a function of ecological processes that do not require information to flow across direct interpersonal ties. The job flocking mechanism I propose here is more than an analogy or a metaphor—it is developed based on existing models in bioecology that predict the formation of flocks as a function of information exchange among avian species who, unlike humans, are incapable of social communication, i.e., cannot form direct social ties with one another. Also unlike humans, for birds the decision to join a flock, roost or colony is not at all a means of satisfying a need for sociability but a way of increasing foraging and survival chances (Perrins and Birkhaed, 1983). Members of a foraging group of avian species are not individually acquainted with one another and typically are not aware of each otherÕs existence unless the very reason that brings them together (the search for food, mating and breeding, or protection form predators) provides an opportunity. For the purpose of this research then and to adapt the concept to the study of flocking behavior in humans, I define flocking as the (almost) simultaneous movement across time and space of ecologically proximate individuals. It is this proximity that, in addition to the presence of direct interpersonal ties, may also produce similarity in observed behavior. And if members of an ecological network of ties of observability are not related by ties of acquaintanceship, the emergence of such ties

802

S.D. Dobrev / Social Science Research 34 (2005) 800–820

and the formation of social networks of person-to-person ties may in fact be a function (and not just a cause) of similarity in behavior driven by ecological contagion. Yet in most cases, social networks are likely to exist among subsets of actors belonging to a broader ecological network. Perhaps because the effect of person-to-person ties on social behavior is likely to be stronger than the purported ecological effect, the latter is invariably omitted from mainstream sociological analysis. In my view, this is an important omission that needs to be redressed. I set out to do so here by spelling out the mechanism that triggers the ecological effect. The idea of flocking (ecological contagion) can be developed further. One way to do so is to apply the ecological mechanism that biologists have theorized to account for flocking behavior among avian species, and to test if this mechanism holds for the member behavior of the specified human population. This is the purpose of my empirical investigation. The population I studied is the alumni population of a prestigious US business school: it provides a good context for the study because— although undoubtedly many of its members know and network with one another—the assumption that the reach of the personal network of any given member of this population does not spread the entire member population is reasonable. This assumption can also be justified by some experimental research that estimates the upper reach of an individualÕs network at 1526 persons (Killworth et al., 1990). By contrast, the total size of the alumni population that I surveyed is substantially larger and exceeds 12,200 individuals. A necessary assumption about the nature of the surveyed population pertains to the extent of homophily among its members. The assumption is required as a substitute for the spatial proximity condition that defines a flocking population of avian species as such. In bioecology research, the formation of a flock occurs as a function of avian individuals being able to observe the behavior of physically proximate others. In the population of human individuals that I investigate here, ‘‘observation’’ occurs among socially proximate persons who can be considered homophilous based on an acquired characteristic, namely the possession of a graduate business degree from an elite educational institution, a fact that introduces what sociologists refer to as ‘‘status homophily’’ (Lazarsfeld and Merton, 1954). To account for collective movement in time and space, I track the career progression of alumni and focus on the movement between states distinguished on the basis of different jobs in different industries. Collective movement then can be observed when multiple transitions to new jobs in a specific industry occur within a constrained time period.

2. An ecological mechanism for job flocking What mechanism could trigger this purported ‘‘job flocking’’ behavior? In line with the search for an ecological explanation to flocking behavior in humans, i.e., their aggregation in ‘‘population units which mere propinquity enforces’’ (Park, [1939] 1972, p. 130), I use a theoretical proposition that has received substantial support from field experiments and observations conducted by biologists studying avian ecology. Specifically, I use the information center hypothesis (Ward and Zahavi, 1973)

S.D. Dobrev / Social Science Research 34 (2005) 800–820

803

which argues that a variety of bird species assemble into large flocks for the benefit of exchanging information about food sources. Field tests of this hypothesis have revealed that foraging individuals do not leave the colony independently of each other and individuals foraging in groups have higher food intake than solitary birds (Krebs, 1974). In bird flocks, information exchange occurs not by direct communication among individuals but via observation, a fact that—unlike social networks—makes it difficult for the successful forager to prevent the relaying of information (Bertram, 1978). This condition is important because it highlights a process which is not facilitated by the individualÕs engagement in processes of direct communication based on personal ties but merely by its embeddedness in an ecological network formed on the basis of ties of observability. Note the following example from research in bioecology—a field study of the cliff swallow (Brown, 1986) revealed that birds inferred other birdsÕ foraging success by observing their direct flight to a different foraging location and proceeded to follow them in route to that location. Without the presence of direct interpersonal ties, individuals situated in spatial proximity benefited from a transfer of information. It is thus not necessary that individuals be organized in a social network of direct ties but simply in an ecologically proximate assemblage. I reckon that in sociodemographic space, homophily is analogous to ecological propinquity and applies to any relevant sociodemographic dimension that can reorder individuals along common characteristics. In my empirical application, I focus on one of several possible relevant dimensions—receiving elite business education from the same institution. Another important tenet of the information-transfer hypothesis is that it does not contain an assumption that individuals lack the necessary information to forage unaided. Rather, even if every member of the population is capable of finding food on its own, converging to an information center is valuable as insurance against eventual depletion of resources in the current foraging area (Ward and Zahavi, 1973). This condition relaxes the necessary assumption that socially advantaged individuals such as professional managers (the population studied here) are unable to make beneficial career decisions independently. Keeping in mind the assumptions specified above, it is plausible to translate the information transfer hypothesis in a way that can account for the job flocking behavior of members of the population of professional managers that I investigate: When some members of the population engage in transition to a more lucrative state (leave the current job to assume a new one in an opportune industry), other members of the population are likely to observe this behavior and engage in a job transition to the same destination state. Another condition needs to be resolved: Observation by individuals in a flock is unhampered and based on visual interpretation. This is obviously impossible among people whose proximity is based on status homophily (higher education) rather than on spatial propinquity (geography). But observation of career moves among same school alumni is facilitated and indeed promoted in a variety of ways. Monthly publications, electronic databases, school websites, capital campaigns, and the like help to keep members abreast of fellow alumniÕs career development. Thus, it is possible that a school alumna can learn about the career progression of fellow alumni

804

S.D. Dobrev / Social Science Research 34 (2005) 800–820

through ‘‘observation’’ and not via interpersonal ties (direct or through others) with any of her job-changing peers. More broadly, observation of behavior within a group of ecologically proximate but unconnected peers may be driven by communication flows that do not involve transmitting information through person-to-person networks(e.g., various mass and specialized media outlets or marketing campaigns). To approximate the lucrative destination states in the observed career transitions in a way that makes it analogous to the search for food in birds, I identify two industries in which the opportunities for financial rewards and professional advancement abound: the finance industry and the consulting industry. A prior study of the same alumni population reveals that on average, jobs in these two industries are associated with higher monetary compensation than jobs in alternate industries (Dobrev and Barnett, 2005). Although prior studies testing the information transfer hypothesis did not predict a functional form for the relationship between number of departing and number of following individuals, I think it unrealistic to assume that this relationship is linear, especially when applying the model to human individuals. I propose that increases in the low counts of members of the alumni population who transition to a new job in the consulting or finance industry will increase the visibility of the transition and transmit information to other alumni about the appeal of these destination states. Yet as the number of departing individuals continues to grow, this increase will signal crowding in the flock (i.e., the transitioning group), and its impact on inducing similar moves will likely diminish. This conjecture accords with GranovetterÕs (1978) threshold model of individual participation in collective action and is consistent with the observed limits in the size of bird flocks (Pulliam and Caraco, 1984). Though average flock size varies substantially depending on the presence of other relevant selective pressures, it is generally acknowledged that as the size of the flock grows, competition for food among its members increases while the benefit of the information transfer plateaus. For these reasons, I expect that the likelihood to join a ‘‘job flock’’ will increase at a decreasing rate as flock size rises. A question remains: what is the value of articulating a mechanism of ecological contagion? That the behavior of social actors is shaped by the actions of their relevant peers is a well-established sociological insight (Burt, 1987; Coleman et al., 1966; Katz and Lazarsfeld, 1955; Marsden and Friedkin, 1994; Weber, [1922] 1978). And in recent years, some of the most influential research in sociology has been focused on explaining how diffusion processes (including contagion, mimicry, social learning, dissemination, etc.) lead to the exchange of information between a source and an adopter (for a review, see Rogers, 1995; Strang and Soule, 1998). Consider, for example, BurtÕs (1987) study of the mechanism by which medical innovation diffuses among doctors. Building on Coleman et al. (1966) classic diffusion of innovation analysis, Burt showed that observation of the behavior of peers occupying the same position in the social structure(structural equivalence) rather than direct conversations with colleagues (cohesion) led to the adoption of a new drug. This argument appears strikingly similar to the flocking mechanism outlined above—both can be applied to explaining the operation of an information-driven contagion process. The key difference is that structural equivalence applies to a set of social actors occu-

S.D. Dobrev / Social Science Research 34 (2005) 800–820

805

pying the exact same structural position in a social network while the flocking argument relaxes the equivalence constraint and posits that proximity in the individualsÕ order on a relevant sociodemographic dimension is sufficient to generate observation-driven information exchange. That is, the flocking argument is purely ecological—all that is needed for information exchange to occur is observation among proximate individuals. To be sure, ecological arguments play an important role in explaining similarity in social behavior. Aside from spatial diffusion effects attributed to geographic proximity (Hedstro¨m, 1994; Land et al., 1991), a stream of research originating with McPhersonÕs (1983) study of ecologically induced affiliations, posits that proximity in sociodemographic space determines the relational embeddedness of social actors and consequently their position in the social structure. Structural equivalence in the position of two actors, thus, may at least partly result from the ecological proximity of these actors because: Homophily implies that the probability of two individuals sharing a network connection. . .is a direct function of their similarity in sociodemographic and spatial characteristics. That is, the more similar two people are, the more likely they are to share a network connection; in more informal terms, birds of a feather flock together (Rotolo and McPherson, 2001, p. 1001). In these terms, the flocking metaphor invariably implies a direct network connection, and it is such connections that homogenize (i.e., direct and limit) an individualÕs course of action; hence, the observed similarity in behavior among a set of socially proximate actors. My argument here simply aims to explicate the nature of network connections likely to be formed among homophilous individuals. I distinguish between social(e.g., friendship, acquaintance, professional, etc.) and ecological ties and argue that the latter are capable of producing similarity in social behavior, even if they do not lead to the formation of interpersonal ties, as McPhersonÕs theory predicts. Certainly, it would be a mistake to interpret this argument as an attempt to supplant a social network-based explanation with ecological reasoning, as in most cases it would be foolish to expect that a purely ecological account will suffice in explaining a complex social phenomenon. Yet, understanding the extent to which ecological arguments can explain interpersonal dynamics(e.g., information sharing, conferral of power, institution building, etc.) can provide a clear baseline against which social exchange and other theory can be developed. In that sense, my research follows in the footsteps of the human ecologists of the Chicago School, and particularly Robert Park, who masterfully elucidated the distinction between ecological processes and communication-based interaction (Park, [1939] 1972). This distinction, which in large part motivates the current study, rests on the premise that ‘‘society begins with a mere aggregation, i.e., a population unit’’ (Park, [1939] 1972, p. 128). That is, understanding the ecological organization of human life is the foundation necessary for explicating higher level forms of social (i.e., economic, political, or cultural) association. A research effort in this direction may be the only way to redress the

806

S.D. Dobrev / Social Science Research 34 (2005) 800–820

long-standing misconception that ecology is simply ‘‘an extension of economics to the whole world of life’’ (Wells et al., 1931, p. 961).

3. Completing the model Even if adapting the information transfer hypothesis from avian ecology might contribute to our understanding of similarity in social behavior, a natural skepticism pertains to the appropriateness of using a simple theory from bioecology to explain a social process. The complexity of social life undoubtedly far exceeds that in the world of avian species and makes for a primitive and poorly suited comparison with the primal search for food among non-humans. Obviously, decisions to change jobs are a function of multiple, often competing considerations based on individual preferences. Choices resulting in career moves are fraught with the opportunities and constraints that arise from the broader social context. The argument that the phenomenon of job flocking can be understood as an ecological process would be naive if it did not explicitly allow for a variety of other factors related to the individualÕs embeddedness in social space to also affect the outcome of the process. The flocking pattern that I seek to uncover may well be due to personal and contextual characteristics that push or pull the individual toward the observed job transition. Accordingly I consider several alternative explanations for why the effect I predict may obtain. First, individuals with inherently higher propensity to change jobs (those with higher-than-average drive for success, risk-taking propensity, etc. (McClelland, 1961)) are constantly at a higher risk of transition and this behavior may simply coincide in time and space with that of other individuals experiencing a job change. Such unobserved heterogeneity among individuals would result in observed spurious association between egoÕs and altersÕ rates of job changes (Heckman and Borjas, 1980). To avoid this fallacy, I will account for the disproportionate propensity of some individuals to ‘‘job-hop’’ by controlling for the serial effect in rates of job transitions. Similarly, the propensity to engage in career moves may be influenced by lifecourse developments and so may vary with stages of an individualÕs lifecycle. That experimentation and risk-taking are associated with earlier stages of the lifecycle while natural conservatism tends to develop as individuals grow old has been empirically documented by earlier research (Sanders and Nee, 1996). Second, a rich tradition of sociological research convincingly demonstrates that the likelihood of getting a job hinges on actorsÕ demographic and socioeconomic make-up which serves as a powerful sanctioning mechanism for whatever personal aspirations and preferences actors may have (Abbott, 1993; Granovetter, 1974). Though the sample of individuals whose career histories I analyze represents a privileged social stratum (holders of professional business degrees form a well reputed school), variations in gender, race, and ethnicity may still play a role and will be accounted for in the model. Additionally, access to information about new opportunities varies with the social capital of individuals (Burt, 1997), and the latter may increase as professional and business experience accumulates over the course of oneÕs

S.D. Dobrev / Social Science Research 34 (2005) 800–820

807

experience in the labor market (Carroll and Mosakowski, 1987; Phillips, 2002; Portes and Zhou, 1996). Third, properties of the job, related to the incentive system and nature of work in the current organization may exercise powerful effects on oneÕs likelihood to change jobs. Pecuniary rewards, position in the top tier of the organizational hierarchy, higher levels of personal commitment and various intrinsic rewards may serve as detracting forces in leaving oneÕs job. Additionally, clearly defined prospects for career progression and job security offered by firms with developed internal labor markets may have a similar effect even if current rewards are less satisfactory (Baron and Bielby, 1980; Baron et al., 1986). Finally, a purely economic demand explanation of the observed job transitions would suggest that managers transition to jobs in the finance or consulting industry in periods when these industries are experiencing an economic upturn. Since demand for labor is unevenly distributed across time, job openings and the necessary hires to fill them will be concentrated in discrete time intervals, a pattern that will likely result in same destination job transitions clustered within those intervals. Thus the job changes of ‘‘first movers’’ as well as of those who ‘‘follow’’ them will be a function of contemporaneous changes in the external structure of opportunity, and not of contagion as I contend. This exogenous effect that may be driving the observed similarity in career progression is a particular example of the general problem of confounding the contagion effect between a source and adopter to a third-party effect that may render contagion spurious (Van Den Bulte and Lilien, 2001). To the extent possible, I will account for these multi-level (individual, organization, and environment) processes in the model developed below.

4. Data and method I use data that were collected from a career survey administered to all MBA alumni of a prestigious US business school in 1997. Completed(or partially completed) surveys were received from 5283 individuals, a response rate of 43%. This compares favorably with the response rates obtained in prior organizational analyses based on the survey method: Kelly and Dobbin (1999) and Dobbin and Sutton (1998) report a 45% response rate, Milliken et al. (1998) report an 18% response rate, Lincoln and Kalleberg (1985) report a 35% response rate, and Blau et al. (1976) report a 36% response rate. Respondents were asked to provide a complete account of their career histories describing the core features not only of their previous positions, but also of the organizations where they were employed. Information about any job changes was also collected along with some general demographics. I excluded incomplete surveys whenever handling the missing values through interpolation was infeasible. Fortunately, a sensitivity analysis revealed that paring down the data to only those respondents for whom I had complete records (N = 2693) did not introduce a detectable bias in the final sample. In Table 1, I show comparative distributions of the basic demographic covariates before and after the exclusion of missing cases. As the frequencies of these variables indicate, the two sets of distributions are remarkably similar.

808

S.D. Dobrev / Social Science Research 34 (2005) 800–820

After an elaborate logical cleaning procedure, I restructured the data in an event history format where a single spell accounts for each job held by each respondent during his or her labor force experience since graduation. These spells were then divided into person-year segments in order to update independent variables. Particularly important for my research is that respondents were asked to indicate if and when they changed jobs and when such changes involved transitions between industries. I marked up years in which respondents reported to have assumed a new job in a new organization in the finance or in the consulting industry. To specify the risk-sets for the transitions to finance and consulting, I produced two data-files, each with person-years for individuals not employed in finance or consulting, respectively. Each data file contains records for people Ôat riskÕ of joining a flock—the year following the first observation of a transition to finance (1938) or to consulting (1944). The two resulting data files contain 44,226 person-year spells (finance flock) and 47,964 person-year spells(consulting flock). Using event history analysis as implemented in TDA 5.7 (Blossfeld and Rohwer, 1995), I estimated piecewise exponential hazard rate models with the general form: rj ðlj Þ ¼ rj ðlj Þ exp½aT ; where rj (lj) is the rate of job transition by person j and lj is jÕs tenure on the labor market at a given point in time since graduation. Based on exploratory analyses of the distribution of job transitions across labor market tenure (measured in years) and

Table 1 Comparative distributions of respondentsÕ demographic characteristics before and after data cleaning Percent in full sample (N = 5283)

Percent in cleaned sample (N = 2693)

Class cohorts: 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995

(Mean = 1976) 10.1 9.8 10.6 12.5 16.8

(Mean = 1975) 10.3 10.3 10.0 12.5 17.1

Birth cohorts: 1941–1945 1946–1950 1951–1955 1956–1960 1961–1965

(Mean = 1949) 11.5 11.2 11.1 12.9 15.5

(Mean = 1948) 12.1 11.5 11.1 12.4 15.8

Number work years: 11–15 16–20 21–25 26–30 31–35

(Mean = 20.3) 10.8 10.3 9.7 10.2 10.2

(Mean = 20.3) 10.4 10.4 10.1 10.3 11.1

Gender (male)

83.1

83.5

Ethnicity (white)

87.1

90.4

S.D. Dobrev / Social Science Research 34 (2005) 800–820

809

to maximize model fit, I established the duration of the tenure pieces along the following yearly breakpoints: 1, 3, 5, 7, 15, and 25 resulting in seven pieces within which the rate is constrained to be constant but varies otherwise. rj (lj)* is the baseline rate for person j estimated as a function of observables, T is the cumulative number of same destination transitions experienced by other alumni in the preceding calendar year and a is the parameter to be estimated. Consistent with the argument that the likelihood to join a job-flock plateaus as the flock size increases, I included a logarithmic transformation of T in the equations. If prior transitions by fellow alumni increase the likelihood that a person j will experience the same transition by transferring information about the industry in which their new jobs reside at a decreasing rate, then I expect 0 < a < 1. A variety of additional covariates were included to control for the alternative explanations discussed above. In addition to basic demographic characteristics of respondents such as age, gender, race, and citizenship, I also computed job-related measures for span of control (log number of reports), annual salary (inflation-adjusted in 1997 US Dollars), and a dummy for whether the job was in a family owned business. At the organization-level, I computed measures for the age and size of the organization in which a respondent holds a job. Finally, to control for rate-dependence, I computed a covariate that counts the number of prior job changes that each respondent has experienced. These variables, descriptive statistics for which are presented in Table 2, allow us to control for some of the complexities involved at each stage of an individualÕs career progression. To make the analysis tractable, I lagged the values of the two main explanatory covariates(the summed number of transitions by other alumni to jobs in the finance or in the consulting industry) by one year. To account for variation in the external structure of opportunity, it is necessary to control for the labor demand in the finance and consulting industries. Ideally, the data on job opportunities would tell what positions were available in exactly those finance and consulting industry segments to which I observe job transitions. Table 2 Descriptive statistics of variables in the full event history file (N = 52,519) Variable

Min

Max

Ave

SD

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

1 22 4.84 0 0 0 2.30 0 0 0 2.30 0 2.30 2.30

44 65 42.25 1 1 1 12.77 1061 1 396 13.82 12 3.89 3.93

15.90 41.72 18.46 0.91 0.95 0.95 1.95 26.30 0.16 41.72 6.42 1.82 2.88 2.68

10.40 10.28 9.07 0.28 0.22 0.22 2.75 40.49 0.37 42.03 3.47 1.85 0.85 1.22

Elapsed labor market tenure Individual age Individual age2 · 102 Male White US-born Ln(Number of reports in origin job + 0.1) Annual salary in origin job · 104 Origin job is a family business Age of origin job organization Ln(Size of origin job organization + 0.1) N prior job changes Ln(N alter job transitions to finance in prior year + 0.1) Ln(N alter job transitions to consulting in prior year + 0.1)

810

S.D. Dobrev / Social Science Research 34 (2005) 800–820

Although data on job openings is not available, I was able to obtain data (collected by the Department of Labor, Bureau of Labor Statistics and publicly available at www.bls.gov) on the total number of people employed precisely in those segments. The matching between industry segments to which respondents transitioned and industry segments for which total employment data are available is based on standard industry classification codes and include SIC codes 60-62, 67 for finance, and SIC code 8742 for consulting. I used these data to compute a control variable (for each industry) as the ratio of total number employed in a given year to that number in the previous year. I called this variable annual change in total number employed. Unfortunately, total employment data are available only for part of the time period covered by my analyses. Specifically, data on finance jobs go back to 1972, and data on consulting jobs have been collected from 1988 onward. At least, the data are monthly and seasonally adjusted so it is not necessary to lag them, as the value for January of each available year can be used. To make the most use of these data in attempting to eliminate a potentially spurious contagion effect, I used the annual change in total number employed variable even though it has non-zero values only in years for which the number of total jobs in the industries is available. To control for missing values, I added a dummy variable that marks up years for which industry jobs data are unavailable. This is admittedly an imperfect approach but much preferable to the only other alternative of severely left-truncating the data by excluding records with missing values.

5. Results The models for the two transitions (models 3a and 4a) are presented in the first columns of Tables 3 and 4. Most baseline effects are synchronous between the two transitions and are in line with earlier arguments from sociological theory. RespondentsÕ propensity to change jobs is nonmonotonic across their lifespan and its initial increase is followed by a decline. This effect agrees with lifecourse theoryÕs prediction that individualsÕ risk-taking propensity declines as they get older (Elder, 1975). Note that this effect is independent of the career progression effect, which suggests that a mangerÕs propensity to transition to either a consulting or a finance job monotonically declines with her labor market experience. The effects of demographic characteristics are mixed: the race effect is insignificant, while foreign-born respondents are significantly more likely to transition to a job in either industry. Finally, the gender effect diverges in the two transitions, perhaps reflecting personal preferences. Females are significantly more likely to transition to a job in consulting, while males are substantially more prone to assuming a job in the finance industry. As expected, job characteristics play an important role in oneÕs likelihood of leaving a job. A managerÕs span of control diminishes this likelihood, suggesting that a position in the higher echelon of the organizational hierarchy (the measure includes both direct and indirect reports) is an important reward that deters a possible job change. While the effect of monetary rewards (salary) predictably decreases the like-

S.D. Dobrev / Social Science Research 34 (2005) 800–820

811

Table 3 Piecewise exponential models of transitions to finance industry jobs Model 2a Labor market tenure (l) 0 25 Individual age Individual age2 · 102 Male White US-born Ln(Number of reports) in origin job Salary in origin job · 104 Origin job is a family business Age of origin job organization Ln(Size) of origin job organization Variation in job opportunity: Annual change in total N employed in finance (1972–1997) Dummy for period with no data (1939–1971) N prior job changes Log(N alter job transitions to finance) Log-likelihood

1.72 (1.87) 3.57** (1.88) 3.51** (1.90) 4.12* (1.92) 4.45* (1.94) 4.98* (1.95) 4.98* (1.91) 0.06 (0.05) 0.14* (0.06) 0.38* (0.11) 0.02 (0.13) 0.29* (0.14) 0.13* (0.02) 0.01* (0.00) 0.54* (0.14) 0.01* (0.00) 0.11* (0.01) 0.47 (1.51) 1.00 (1.56) 0.32* (0.03) 3271.0

Model 2b 3.56** (1.94) 5.41* (1.96) 5.35* (1.97) 5.94* (1.99) 6.25* (2.01) 6.78* (2.02) 6.78* (1.99) 0.04 (0.05) 0.13* (0.06) 0.41* (0.11) 0.06 (0.13) 0.28* (0.14) 0.13* (0.02) 0.01* (0.00) 0.53* (0.14) 0.01* (0.00) 0.11* (0.01) 0.82 (1.56) 0.65 (1.62) 0.32* (0.03) 0.24* (0.06) 3262.5

Note. Standard errors in parentheses; Number of person-years: 44,226; Number of individuals: 2702; Number of job transitions: 820. * Significant at the .05 level, two-tailed tests. ** Significant at the .10 level, two-tailed tests.

lihood of getting a new job in the consulting industry, it has the opposite effect in the transition to finance—the more money one makes in their current job, the more likely they are to transition to a job in the finance industry. Perhaps this effect is a corollary of the fact that most managers qualified to pursue a job in a finance industry probably have a background in the field of finance, and managers in positions that require expertise in finance are better compensated than their peers even if their positions are with organizations operating in other industries. Based on earlier research in self-employment (Carroll and Mosakowski, 1987; Portes and Zhou, 1996) and work commitment(Lincoln and Kalleberg, 1985), I interpreted the meaning of having a job in a family business as a proxy for high degree of personal commitment and job satisfaction and expected that it would significantly reduce the risk of leaving the job, regardless of the properties of the destination state. While this expectation is confirmed by the estimate in the transition to finance model, the effect is positive and significant in the transition to consulting model. I would not speculate on what may be driving this effect but suspect that the opportunity for self-employment(which also offers many of the intrinsic rewards

812

S.D. Dobrev / Social Science Research 34 (2005) 800–820

Table 4 Piecewise exponential models of transitions to consulting industry jobs Model 3a

Model 3b

Labor market tenure (l) 0 25 6.81* (1.84) 6.12* (1.84) Individual age 0.06 (0.04) 0.01 (0.04) Individual age2 · 102 0.10** (0.05) 0.04 (0.05) 0.27* (0.10) Male 0.39* (0.10) White 0.18 (0.13) 0.09 (0.13) US-born 0.37* (0.14) 0.36* (0.14) 0.16* (0.02) Ln(Number of reports) in origin job 0.16* (0.02)  * Salary in origin job · 10 4 0.003 (0.00) 0.003** (0.00) Origin job is a family business 0.46* (0.10) 0.48* (0.10) * Age of origin job organization 0.01 (0.00) 0.01* (0.00) 0.09* (0.01) Ln(Size) of origin job organization 0.10* (0.01) Variation in job opportunity: Annual change in total N employed in consulting (1988–1997) 1.56 (1.41) 0.60 (1.41) Dummy for period with no data (1945–1987) 1.65 (1.54) 0.99 (1.54) N prior job changes 0.29* (0.02) 0.29* (0.02) Log(N alter job transitions to consulting) 0.38* (0.05) Log-likelihood 3041.3 3007.7 Note. Standard errors in parentheses; N person-years: 47,964; N individuals: 2702; N job transitions: 752. * Significant at the .05 level, two-tailed tests. ** Significant at the .10 level, two-tailed tests.

associated with running a family business) in the field of consulting may be greater than that in finance. Following earlier arguments linking organizational age and size to the presence of developed internal labor markets (Dobrev and Barnett, 2005), I surmised that the likelihood to leave a current job will decrease for managers working in large and old organizations because the opportunities for career progression within the organization will countervail whatever ‘‘pull’’ effect a lucrative job in the finance or consulting industry may exert. The estimated negative organizational size effects agree with this prediction while the organizational age effects differ between the two transitions—respondents working in old firms are less likely to leave for a job in consulting but more likely to do so for a job in finance. Finally, variations in the sheer propensity of individuals to change jobs (between any organizations and industries) appear to strongly influence their likelihood to also take on jobs in consulting or finance. A manager who has changed jobs five times in the past is about five(four) times more likely than a person with no prior job change experience to transition to a job in the finance (consulting) industry. I am confident in the interpretation of this effect as capturing variations in personal propensity even

S.D. Dobrev / Social Science Research 34 (2005) 800–820

813

though it may arguably represent the effect of social capital accumulated as oneÕs career progresses, or may simply reflect a qualification effect, where one must gradually climb up the career ladder(involving various job changes) before her qualifications are appealing in a more attractive industry. But these important effects are already captured by the controls discussed above, namely, extensiveness of labor market tenure and job properties of the current position. The estimates that speak to the main argument of this thesis are presented in models 3b and 4b of Tables 3 and 4, which include the main effects of transitions to finance and consulting industry jobs by other alumni in the preceding year. In each transition, the main term effect is in the predicted direction and suggests that as fellow alumni transition to jobs in the finance (consulting) industry, a managerÕs own likelihood of getting a job in the finance(consulting) industry increases significantly at a decreasing rate. The estimates form these most complete models support the conjecture that the pattern of career mobility among professional managers resembles a flock-like behavior in both transitions, net of a variety of other effects. To help interpret the results, the estimated multipliers are plotted in Fig. 1 and suggest that when 49 alumni (the highest observed number of transitions in any given year) changed jobs from a non-finance to a finance industry, other alums became more than two and half times more likely to experience the same job transition in the following year {exp[0.24 * log(49)] = 2.55}. The maximum number of observed transitions to the consulting industry was 51 and it propelled other alumni to become almost four and a half times more likely to also get a consulting job in the subsequent year {exp[0.38 * log(51)] = 4.46}. These estimates are merely illustrative of

Fig. 1. Estimated job flocking effects among fellow alumni.

814

S.D. Dobrev / Social Science Research 34 (2005) 800–820

the coefficients and almost certainly differ from the actual number of job transitions in the total population, only a fraction of which is sampled in the data.

6. Discussion and conclusion The matching of people and jobs has been a central tenet of analysis in the social sciences (Granovetter, 1986). In economics, the main explanatory variables gravitate toward the distribution of human capital which affects the balance between supply and demand of labor (Nakamura and Nakamura, 1989). Sociologists often point to the demographic characteristics of the labor pool and the structural embeddedness of people in social contexts, attributing variations in levels of occupational mobility to race and gender discrimination, human and social capital, and socioeconomic status (Barnett et al., 2000). Because career mobility has been studied extensively in earlier research, my models included a variety of controls for factors that have been empirically shown to predict the propensity to change jobs. The frequency-dependence model specified to test the job-flocking behavior of professional managers shows strong statistically significant effects that improve the fit of the data even when controls for the occurrence of some other sort of career staging process (that could explain the transitions to consulting and finance jobs without recourse to jobflocking) are accounted for in estimating the rate of job transitions. The results presented here should be interpreted as a complement (and not an alternative) to social network-based approaches to studying economic mobility. Network theorists have shown that the formation of social groups, in general, and the transfer of information among members of these groups, in particular, hinge on the shared characteristics of individuals that make them more likely to associate with one another. Convincing evidence reveals that the valence (positive or negative) and intensity (weak or strong) of social ties (Granovetter, 1973), as well as the presence of positions that link previously unconnected nodes of the network (i.e., structural holes) obviously have much to do with how information flows among people (Burt, 1992). My findings suggest that a purely ecological mechanism may help to explain the same phenomenon in cases where(by assumption) simple ecological ties of observability connect members of a population characterized by similarity in social behavior. It is also appropriate to address the questions of why and how this study is different from any other study of social imitation which attributes similarity in behavior to the principle of social proof (Cialdini, 1984) and posits that the behavior of others constitutes a heuristic that helps actors infer what course of action may be appropriate. This and other similar arguments explain imitation either as a function of a natural human tendency (Le Bon ([1895] 1960); Tarde ([1895] 1962)) or as a function of individual preferences for conformity (Becker, 1991; Jones, 1984; Leibenstein, 1950). By contrast, the research presented here focuses on the process by which information about a certain course of action is acquired and does not dwell on the issue of whether and how decisions to act on newly acquired information are made (an area of interest for decision science analysts). The contribution of this research lies in

S.D. Dobrev / Social Science Research 34 (2005) 800–820

815

specifying conditions on the individualsÕ likelihood of obtaining information and, specifically, in highlighting the importance of interpersonal proximity in sociodemographic space for the transfer of information through a process driven by entirely ecological forces. Additionally, while research on imitation stresses the important role played by external agents in promulgating mimicking behavior among both individuals and organizations (DiMaggio and Powell, 1983), an ecological approach brings attention to the constraining effect of position in the social structure (Dobrev, 1999; Dobrev et al., 2001; Rotolo and McPherson, 2001, p. 1101). If dissemination of the signals relayed by external agents is modified by the proximity among actors in social space (based on a variety of relevant sociodemographic dimensions), further theoretical elaboration may be appropriate to predict which class of actors and under what conditions may be more or less susceptible to contagion-driven processes. This line of research may be particularly useful in light of persistent findings that for many social actors—collective and individual—the problem of identifying a specific course of action lies not with the lack of information but with the choice among options based on multiple(and sometimes) conflicting sources of information (Coleman et al., 1966; Van Den Bulte and Lilien, 2001). Ecological research in this direction may uncover, for example, that social proximity sorts out and thus mitigates the purported effect of exogenous third parties that facilitate contagion. This study has three critical limitations that warrant caution in assessing the validity and generalizability of the results. First, the central thesis of this paper has been that a population ecology approach to career dynamics is practical for investigating the process by which individuals acquire information about prospective career moves—a dynamic that sociologists typically attribute almost entirely to processes operating through social exchange. Yet the analyses presented here are based on data that do not cover the whole population of alumni: because the data are survey-based, they cover the career histories of about 22% of all alumni—those who returned the survey instrument fully complete. While exploratory analysis revealed that the data are representative and it is unlikely the results would change substantively if the entire population is analyzed, it is nevertheless hard to fully understand ecological dynamics (e.g., to estimate the size of the job-flock or the relative threshold levels at which individuals decide to join or abstain from joining the migrating group) if some potential ‘‘leaders’’ or ‘‘followers’’ are unaccounted for. This brings up the issue of the kind of data that future analysts should opt to collect. It seems that the best alternative would be to balance between the need to collect population data in its entirety and the need to target an aggregation of individuals who are not (all) connected to one another via social networks. Second, although I made every effort to control for the possibility that exogenous shifts in the opportunity structure of respondentsÕ environment are not driving the job transitions that I interpreted as evidence of job flocking, the data needed to categorically reject a labor demand based alternative hypothesis simply do not exist. The need for more complete external context measures and the theoretical primacy of the threat of unobserved third party effects point to an additional data collection tradeoff. These considerations are not unique to ecological analysis and speak to the

816

S.D. Dobrev / Social Science Research 34 (2005) 800–820

difficulties of constructing and estimating adequate multilevel model specifications of various social processes (DiPrete and Forristal, 1994). Finally, it is implausible to convincingly argue that the exchange of information occurs merely as a function of ecological observation and not simply through social networks without controlling precisely for how much of this exchange is in fact due to information circulated through interpersonal contacts. Ideally, one would have data on both extensive career histories of socially proximate individuals and on their complete networks. At present, such data have not been collected and the analysis presented here is intended merely as a guidepost for bringing attention to an ecology-based approach to explaining similarity in human behavior. The empirical analysis presented here is limited by the absence of network data and its findings should be interpreted with caution. However, just as I cannot rule out the likelihood that the effects I estimated are due simply to information circulated via interpersonal networks, so too many a network studies can be reevaluated by asking to what extent the estimated network effects in those studies were concealing other, perhaps purely ecological effects that were not directly controlled for. In fact, recent findings suggest that this may constitute a reasonable research agenda. In a reexamination of BurtÕs (1987) analysis of the diffusion of new drug adoption among medical doctors, Van den Bulte and Lilien concluded that ‘‘the data do not document that diffusion was driven by contagion operating over social networks’’ (2001, p. 1429). Network analysts should be aware that not attending explicitly to ecological dynamics may lead them to overspecify (i.e., specify in too strong a form) the impact of interpersonal relationships on various social outcomes. Although implicit, many contemporary network theories do rest on ecological assumptions. Consider, for example, GranovetterÕs (1973) argument where the strength of ties by definition will be weakened if the ability of two individuals to interact is limited by either social incompatibility or simply spatial distance. Such reasoning is informed by HawleyÕs (1950) well asserted proposition that ecological proximity breeds social interaction. What I have argued here is that ecological proximity can homogenize behavioral outcomes even when it does not produce social interaction. More broadly, if the conjectures developed here are correct, population-level dynamics may be an important part of many general-level sociological predictions that future analysts can ill afford to ignore. A distinction between a social network of person-to-person ties and an ecological network of ties of observability may prove useful in explicating discernable social patterns. One can find many analogues to job flocking that all exhibit the ecological contagion effect I proposed here. For example, people choose to live in a certain residential neighborhood because other people with similar socioeconomic status live there and not necessarily because they know people living in the same neighborhood. People may choose to go and see Michael MooreÕs ‘‘Fahrenheit 9/11’’ not necessarily because a friend told them they should see it but because they have heard/read in the media that the movie appeals to viewers like themselves who profess a liberal ideology. People visiting Chicago may go to a Cubs game not necessarily because their friends are going to the game nor even because they like baseball but because every

S.D. Dobrev / Social Science Research 34 (2005) 800–820

817

tourist guidebook and every radio-station in the city trumpets that this is ‘‘the thing to do’’ when visiting Chicago. Polish people who immigrate to America settle in Chicago not because they necessarily know other Polish people there but because they know(this is common trivia) that Chicago is one of the largest cities with an ethnic Polish population in the world. Among other things, ecological contagion may partly explain the persistence of ethnic industry enclaves in countries with large immigrant communities like the US (e.g., Vietnamese immigrants work in the grooming/beauty-salon industry, Romanian immigrants dominate the flooring and carpeting business, etc.). In all of these examples, individuals choose to do what others like them, even if complete strangers, are doing. So, it may be that similarity in behavior(engendered by ecological contagion) leads to the formation of personal ties, not only the other way around, as network ecologists (Mark, 1998; McPherson, 1983; Rotolo and McPherson, 2001) convincingly demonstrate. Clearly, more research and better data are needed to determine the contribution of the job flocking (ecological contagion) mechanism advanced here. But this can only be done if future research explicitly models a simple ecological dynamic, one by which individuals who may not be connected through interpersonal ties, are still able to exchange information via ecological ties of observability—a process facilitated by sociodemographic proximity. Such ecological models should not be developed on their own but as part of broader sociological theories whose explanatory power can predict the occurrence of complex social processes like labor market dynamics and career progression. Acknowledgment This research was supported by the Kauffman Foundation of Kansas City, MO and by the University of Chicago Graduate School of Business. I am grateful to Glenn Carroll, Toby Stuart, and Valery Yakubovich for critiquing an earlier draft of this paper, to Salih Zeki Ozdemir for valuable research assistance, and to Livia S. Dobrev for good ideas and editorial help. The flaws are all mine. References Abbott, A., 1993. The sociology of work and occupations. Annual Review of Sociology 19, 187–209. Barnett, W.P., Baron, J.N., Stuart, T., 2000. Avenues of attainment: occupational demography and organizational careers in the California Civil Service. American Journal of Sociology 106, 88. Baron, J.N., Bielby, W., 1980. Bringing the firms back in: stratification, segmentation, and the organization of work. American Sociological Review 45, 737–765. Baron, J.N., Davis-Blake, A., Bielby, W., 1986. The structure of opportunity: how promotion ladders vary within and among organizations. Administrative Science Quarterly 31, 248–273. Becker, G.S., 1991. A note on restaurant pricing and other examples of social influences on price. Journal of Political Economy 99, 1109. Bertram, B.C., 1978. Living in groups: predators and prey. In: Krebs, J.R., Davies, N.B. (Eds.), Behavioural Ecology: An Evolutionary Approach, first ed. Blackwell, Oxford, pp. 64–96. Bott, H., 1928. Observation of play activities in a nursery school. Genet Psychology Monographs 4, 44.

818

S.D. Dobrev / Social Science Research 34 (2005) 800–820

Blau, P.M., 1970. A formal theory of differentiation in organizations. American Sociological Review 35, 201–218. Blau, P.M., Falbe, C.M., McKinley, W., Tracy, P.K., 1976. Technology and organization in manufacturing. Administrative Science Quarterly 21, 20–40. Blau, P.M., 1977. A macrosociological theory of social structure. American Journal of Sociology 83, 26– 54. Blossfeld, H-P., Rohwer, G., 1995. Techniques of Event-history Analysis. Erlbaum, Mahwah, NJ. Brown, C.R., 1986. Cliff swallow colonies as information centers. Science 234, 83. Burt, R.S., 1987. Social contagion and innovation: cohesion versus structural equivalence. American Journal of Sociology 92, 1287–1335. Burt, R.S., 1992. Structural Holes. Harvard University Press, Cambridge, MA. Burt, R.S., 1997. The contingent value of social capital. Administrative Science Quarterly 42, 339–365. Carroll, G.R., Mosakowski, E., 1987. The career dynamics of self-employment. Administrative Science Quarterly 32, 570–589. Cialdini, R.B., 1984. Influence: the Psychology of Persuasion. Quill, New York. Coleman, J.S., Katz, E., Menzel, H., 1966. Medical Innovation: a Diffusion Study. Bobbs-Merrill, Indianapolis. DiMaggio, P.J., Powell, W.W., 1983. The iron cage revisited: institutional isomorphism and collective rationality in organizational fields. American Sociological Review 48, 147–160. DiPrete, T.A., Forristal, J.D., 1994. Multilevel models: methods and substance. Annual Review of Sociology 20, 331–357. Dobbin, F., Sutton, J., 1998. The strength of a weak state: the employment rights revolution and the rise of human resources management divisions. American Journal of Sociology 104, 441– 476. Dobrev, S.D., 1999. The dynamics of the Bulgarian newspaper industry in a period of transition: organizational adaptation, structural inertia and political change. Industrial and Corporate Change 8, 573–605. Dobrev, S.D., Barnett, W.P., 2005. Organizational roles and transition to entrepreneurship. Academy of Management Journal 48 (3), in press. Dobrev, S.D., Kim, T.-Y., Hannan, M.T., 2001. Dynamics of niche width and resource partitioning. American Journal of Sociology 106, 1299–1337. Durkheim, E., [1893] 1933. The Division of Labor in Society. Free Press, Glencoe, IL. Elder, G.H., 1975. Age differentiation and the life course. In: Inkeles, A., Coleman, J., Smelser, N. (Eds.), Annual Review of Sociology, vol. 1, pp. 165–190. Annual Reviews. Granovetter, M., 1973. The strength of weak ties. American Journal of Sociology 78, 1360– 1380. Granovetter, M., 1978. Threshold models of collective behavior. American Journal of Sociology 83, 1420– 1443. Granovetter, M., 1974. Getting a Job: a Study of Contacts and Careers. Harvard University Press, Cambridge, MA. Granovetter, M., 1986. Labor mobility, internal markets, and job matching: a comparison of the sociological and economic approaches. Research in Social Stratification and Mobility 5, 3– 39. Hawley, A.H., 1950. Human Ecology: a Theory of Community Structure. Ronald Press, New York. Heckman, J.J., Borjas, G., 1980. Does unemployment cause future unemployment? Definitions, questions and answers from a continuous time model of heterogeneity and state dependence. Economica 8, 247– 283. Hedstro¨m, P., 1994. Contagious collectives: on the spatial diffusion of Swedish trade unions, 1890–1940. American Journal of Sociology 99, 1157–1179. Hubbard, R.M., 1929. A method for studying spontaneous group formation. In: Thomas, D.S. (Ed.), Some New Techniques for Studying Social Behavior. Child Development Monographs, pp. 76–85. Jones, S.R., 1984. The Economics of Conformism. Blackwell, Oxford.

S.D. Dobrev / Social Science Research 34 (2005) 800–820

819

Katz, E., Lazarsfeld, P., 1955. Personal Influence. Free Press, New York. Kelly, E., Dobbin, F., 1999. Civil rights law at work: sex discrimination and the rise of maternity leave policies. American Journal of Sociology 105, 455–492. Killworth, P.D., Johnsen, E.C., Russell, B.H., Shelley, G.A., McCarty, C., 1990. Estimating the size of personal networks. Social Networks 12, 289–312. Krebs, J.R., 1974. Colonial nesting and social feeding as strategies for exploiting food resources in the Great Blue Heron (Ardea herodias). Behaviour 51, 99. Land, K.C., Deane, G., Blau, J.R., 1991. Religious pluralism and church membership: a spatial diffusion model. American Sociological Review 56, 237–249. Lazarsfeld, P.F., Merton, R.K., 1954. Friendship as a social process: a substantive and methodological analysis. In: Berger, M. (Ed.), Freedom and Control in Modern Society. Van Nostrand, New York, pp. 18–66. Le Bon, G., [1895] 1960. The Crowd. Viking, New York. Leibenstein, H., 1950. Bandwagon, snob, and Veblen effects in the theory of consumer demand. Quarterly Journal of Economics 64, 183. Lincoln, J.R., Kalleberg, A.L., 1985. Work organization and workforce commitment: a study of plants and employees in the US and Japan. American Sociological Review 50, 738–760. Mark, N., 1998. Birds of a feather sing together. Social Forces 77, 453–485. Marsden, P.V., Friedkin, N.E., 1994. Network studies of social influence. In: Wasserman, S., Galaskiewicz, J. (Eds.), Advances in Social Network Analysis: Research in the Social and Behavioral Sciences. Sage, Thousand Oaks, CA, pp. 3–25. McClelland, D.C., 1961. The Achieving Society. D. Van Nostrand, New York. McPherson, J.M., 1983. An ecology of affiliation. American Sociological Review 48, 519–535. McPherson, J.M., Smith-Lovin, L., Cook, J.M., 2001. Birds of a feather: homophily in social networks. Annual Review of Sociology 27, 415. Milliken, F.J., Martins, L., Morgan, H., 1998. Explaining organizational responsiveness to work-family issues: the role of human resource executives as issue interpreters. Academy of Management Journal 5, 580–592. Nakamura, A., Nakamura, M., 1989. Effects of excess supply on the wage rate of young women. In: Michael, R.T., Hartmann, H.I., OÕFarrell, B. (Eds.), Pay Equity: Empirical Inquiries. National Academy Press, Washington, pp. 70–90. Park, R.E., [1939] 1972. Symbiosis and socialization. In: Elsner, H. Jr. (Ed.), The Crowd and the Public and Other Essays. University of Chicago Press, Chicago, pp. 117–142. Perrins, C.M., Birkhaed, T.R., 1983. Avian Ecology. Chapman and Hall, New York. 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. Portes, A., Zhou, M., 1996. Self-employment and the earnings of immigrants. American Sociological Review 61, 219–230. Pulliam, H.R., Caraco, T., 1984. Living in groups: is there an optimal group size? In: Krebs, J.R., Davies, N.B. (Eds.), Behavioural Ecology: An Evolutionary Approach, second ed. Blackwell, Oxford, pp. 122– 147. Rogers, E.M., 1995. Diffusion of Innovations. Free Press, New York. Rotolo, T., McPherson, J.M., 2001. The system of occupations: modeling occupations in sociodemographic space. Social Forces 79, 1095–1130. Sanders, J.M., Nee, V., 1996. Immigrant self-employment: the family as social capital and the value of human capital. American Sociological Review 61, 231–249. Strang, D., Soule, S.A., 1998. Diffusion in organizations and social movements: from hybrid corn to poison pills. Annual Review of Sociology 24, 265–290. Tarde, G., [1895] 1962. The Laws of Imitation, second ed. Peter Smith, Gloucester. Van Den Bulte, C., Lilien, G.L., 2001. Medical innovation revisited: social contagion versus marketing effort. American Journal of Sociology 106, 1409–1435. Ward, P., Zahavi, A., 1973. The importance of certain assemblages of birds as ÔinformationÕ centers for food finding. Ibis 115, 517.

820

S.D. Dobrev / Social Science Research 34 (2005) 800–820

Weber, M., [1922] 1978. Economy and Society: an Outline of Interpretive Sociology. Bedmeister, New York. Wellman, B., 1929. The schoolÕs child choice of companions. Journal of Education Research 14, 126. Wells, H.G., Huxley, J.S., Wells, G.P., 1931. The Science of Life. Doubleday, New York.