Saturday, 29 September 2018

Gender Sorting and the Glass Ceiling in High-Tech Firms

Roberto M. Fernandez, Santiago Campero, First Published September 7, 2016 Research Article Article information Article has an altmetric score of 3 Free Access Abstract With few exceptions, studies have conceived of the glass ceiling as reflecting internal promotion biases. In this article, the authors argue that glass ceiling patterns can also be the result of external recruitment and hiring processes. Using data on people applying by means of the Internet for jobs at 441 small- and medium-sized high-tech firms, they find evidence that the glass ceiling is produced by both internal and external hiring processes. On the supply side, females are sorted into lower-level job queues than males. On the demand side, screening biases against women also are evident, but a series of “what if” simulations suggest that demand-side screening processes play a comparatively minor role in producing the glass ceiling pattern. These results suggest that bias remediation policies designed to equalize gender differences in hiring chances are likely to be less effective than recruitment and outreach policies designed to improve gender disparities in candidate pools. Keywords glass ceiling, economic inequality, gender inequality, gender discrimination, internal labor markets, hiring processes, recruitment Research aimed at understanding the organizational roots of inequality and stratification has burgeoned. Much of this research is motivated by the desire to suggest organizational policies that can be used to produce more equitable outcomes by gender and race (Tomaskovic-Devey 1993; Kalev, Dobbin, and Kelly 2006). Although research on gender inequality in organizational rewards is relatively plentiful (for reviews, see Blau, Brinton, and Grusky 2006; Ridgeway 2011), research has been less clear on the mechanisms that produce these patterns. Clarifying the sources of gender disparities remains a high research priority since the effectiveness of organizational policies seeking to reduce inequality depends vitally on an accurate understanding of the organizational mechanisms that produce inequality in rewards (Bielby 2000; Reskin 2000). In this article, we seek to advance our knowledge of how organizational processes contribute to gender inequality. Specifically, we offer a fresh perspective on the organizational roots of the “glass ceiling,” the phenomenon in which women disappear as one looks up through the levels of the organizational hierarchy. At a theoretical level, we seek to re-orient current understandings of the mechanisms that produce the glass ceiling. While much past research on gender and hierarchy documents that the proportion of women declines as one examines the upper levels of organizations (Ragins, Townsend, and Mattis 1998; Blau et al. 2006), with few exceptions, these studies have conceived of the glass ceiling as reflecting internal promotion biases. Although internal processes are clearly important, we argue that this focus on internal processes is unnecessarily limiting. To the extent that external market competition is gendered, then internally focused studies will not reflect the mechanisms that produce gender stratification and may yield misleading inferences about the nature of the organizational barriers to women’s advancement. We seek to correct this imbalance in the focus of prior research by highlighting the role of hitherto unexamined organizational mechanisms—specifically, external recruitment and hiring processes—and offer empirical evidence on how these processes strongly contribute to producing a glass ceiling pattern. In our view, the glass ceiling describes a vertical form of job sex segregation, and as such can be influenced by the same allocative processes (Petersen and Saporta 2004) that might produce other forms of gender segregation of jobs. Such processes can be both internal—as in promotion practices that allocate individuals to higher levels of the organization—and external, as reflected in hiring patterns. Especially in young, rapidly growing organizations, such external hiring is likely to be occurring at all levels of the organization (Bidwell and Keller 2014). To the degree that organizations depart from the Doeringer and Piore (1971) model of the internal labor market, and higher-level jobs are open to being filled from outside the firm, an internal focus will mis-specify the sets of people at risk for filling job openings, which can yield an incomplete and misleading picture of gender differences in allocation to jobs across the organizational hierarchy. These insights have important policy implications. Existing policies targeting gender biases in internal promotion will not effectively address the gender disparities originating in external recruitment processes. By contrast with policies aimed at ameliorating promotion disparities—for example, changing supervisors’ gender-biased internal assessment processes—external recruitment and hiring processes are often controlled by human resources professionals charged with reaching beyond the organizational boundary. At a minimum, our findings suggest that policy efforts specifically aimed at gender disparities in external recruitment are needed to make progress in overcoming the glass ceiling. Moreover, our data analyses and “what if” simulations of the processes at work in the external hiring interface provide further guidance for designing policies to address the glass ceiling. We show that, by themselves, policies designed to reduce gender discrimination in screening are likely to be of limited help in tackling the problem. By contrast, recruitment policies aimed at producing more gender-equitable candidate pools for jobs at various levels of the hierarchy are likely to pay the biggest dividends in ameliorating the glass ceiling. Hierarchy and Gender Inequality in Rewards The notion of a glass ceiling is a popular metaphor framing numerous studies of gender inequality in both sociology and economics (Morgan 1998; Cotter, Hermsen, Ovadia, and Vanneman 2001; Albrecht, Björklund, and Vroman 2003; Arulampalam, Booth, and Bryan 2007; Gorman and Kmec 2009). Whereas these and other studies have reported results consistent with the glass ceiling notion that gender inequality is more severe at the top of the reward distribution (e.g., Cotter et al. 2001), others have found evidence of “sticky floors,” wherein women’s disadvantages are most pronounced at lower levels of the reward hierarchy (e.g., Booth, Francesconi, and Frank 2003; Zeng 2011). Adding to the debate, various definitions and strategies have been used to investigate the phenomenon of the glass ceiling. For example, Cotter et al. (2001), Albrecht et al. (2003), and Arulampalam et al. (2007) studied gender differences in earnings across broad sectors of the labor market. Others examined earnings and levels within particular professions (Kay and Hagan 1995; Tanner, Cockerill, Barnsley, and Williams 1999; Noonan and Corcoran 2004; Gorman and Kmec 2009), labor market sectors (Cohen, Broshak, and Haveman 1998; Barnett, Baron, and Stuart 2000; Storvik and Schone 2008), and specific organizations (Petersen and Saporta 2004: 887–93; Yap and Konrad 2009), as well as within the context of the gender composition of “C-suite” (i.e., CEO, CFO, and other chief function officers) and board-level jobs (Bertrand and Hallock 2001; Hillman, Shropshire, and Cannella 2007; Smith, Smith, and Verner 2011, 2013). Although much research explores the relationship between hierarchy and gender inequality in rewards, studies in this area have not always been clear about the mechanisms that produce these patterns. As the phenomenon is likely to be the result of multiple, perhaps competing processes, distinguishing among these mechanisms is a high priority for current research on the glass ceiling. In its original conception (Hymowitz and Schellhardt 1986; Morrison, White, Van Velsor, and Center for Creative Leadership 1987), the glass ceiling was thought to be rooted in the ways employers sorted individuals within organizations. A significant branch of the research has followed this line of reasoning and studied organizational processes affecting women’s advancement within firms. Several studies specifically used the glass ceiling idea when examining gender differences in internal advancement. For example, Kay and Hagan (1995) framed their discussion of female–male differences in promotion to partner in law firms as indicative of there being glass ceilings in these firms. Using data on a large service organization, Petersen and Saporta (2004) looked at female–male differences across a broad range of organizational outcomes, including initial job level, turnover, and wages. The section of their article devoted to examining sex differences in promotion is titled “The Glass Ceiling” (ibid: 887–93). Other organizational studies, too, focused on internal advancement when addressing the glass ceiling (Powell and Butterfield 1994; Kalev 2009; Yap and Konrad 2009). Although we applaud the search for specific organizational mechanisms in this research, we believe there are significant and, to date, unrecognized limitations to this internal focus on gender inequality. Research on internal advancement has made the implicit assumption that jobs at higher levels of the organizational hierarchy are relatively closed to the external labor market, so that the set of people at risk of obtaining jobs is well defined. Depending on the setting, this assumption might be warranted. For example, the federal (DiPrete and Soule 1988; Yamagata, Yeh, Stewman, and Dodge 1997) and California (Barnett’s et al. 2000) bureaucracies are purposely structured along internal labor market lines such that higher-level jobs are sheltered from external competition and entry to the system is restricted to the bottom of the hierarchy. This assumption is also reasonable in the case of the relatively closed market for elite lawyers (Gorman and Kmec 2009). The key feature of these contexts is that the population of those at risk for obtaining jobs can be clearly identified, so that the gender composition of those at risk of filling positions can be compared with the gender of those who ultimately obtain those positions. Concerns arise, however, to the extent that the organizations under study depart from the Doeringer and Piore (1971) model of the internal labor market. If higher-level jobs are typically open to competition from the external labor market, then within-organization studies of promotion will obscure this part of the process.1 Indeed, numerous firm-level studies have questioned whether the internal labor market metaphor is currently (Cappelli 1999; Treble, van Gameren, Bridges, and Barmby 2001; Dohmen, Kriechel, and Pfann 2004), or has ever been (Baker, Gibbs, and Holmstrom 1994; Baker and Holmstrom 1995; but see Seltzer and Merrett 2000), a faithful depiction of most firms. Moreover, several recent studies have shown that many executive jobs are filled by external hires (Lazear and Oyer 2004; Cappelli and Hamori 2005; Hassink and Russo 2010) and that externally hired managers tend to be better paid than those promoted from within (Harris and Helfat 1997; Bidwell 2011; but see Hassink and Russo 2008). To the degree that both internal and external candidates are in competition for job openings (Bidwell and Keller 2014; Bidwell and Mollick 2014), studies of internal processes will not reflect the actual mechanisms producing gender stratification and may yield misleading inferences on the nature of the organizational barriers to women’s advancement. Failing to take into account that external candidates may be vying for the same openings as internal candidates can lead to a biased assessment of the gender composition of those at risk for filling jobs across levels of the organization (Fernandez and Abraham 2010, 2011). If external market competition is itself gendered, then gender biases in screening candidates entering the firm from the external labor market will also limit women’s organizational achievement. As a consequence, focusing exclusively on internal promotions runs the risk of wrongly attributing a glass ceiling pattern to internal processes when the observed result might instead be caused by external hiring. Indeed, previous work has proposed that females benefit less than men from external labor market transitions (Brett and Stroh 1997; Lyness and Judiesch 1999; but see Gorman and Kmec 2009). Hassink and Russo (2010), too, argued that a “glass door” mechanism exists through which women are excluded from higher level jobs in allocation processes from the external labor market. Evidence also suggests that external hiring alone can produce a glass ceiling pattern of increasing gender disparities at higher levels of the organization (Fernandez and Abraham 2010, 2011). Examining the external labor market interface is even more important when one considers the common argument that hiring is likely to be an important locus of discrimination. As a number of scholars have pointed out (Collinson, Knights, and Collinson 1990; Jencks 1992: 53; Petersen, Saporta, and Seidel 2000: 766; Petersen and Saporta 2004: 859–60), if employers are going to discriminate, they have the most opportunity to do so at the hiring interface (Petersen and Saporta 2004). Among other reasons, in the case of hiring discrimination, a potential complainant is not usually identifiable and available when the person is not hired, and the information on the people in the candidate pool who might have been hired is not easily obtained. Without such information, claims of discrimination are difficult to sustain. Employer audit studies, in which researchers submit fictitious résumés for well-matched candidates of distinct genders, often show evidence of gender-biased screening in hiring (Azmat and Petrongolo 2014). Despite its relevance as a potential source of gender inequality, external hiring across levels of the organization is not thoroughly understood. A major reason for our lack of understanding is the difficulty of obtaining adequate data to address these questions. To separate the internal and external barriers to women’s achievement within the organizational hierarchy, we need to obtain data on how the allocation process differs for males and females across levels of the hierarchy. And to do so requires properly identifying the pool of both internal and external candidates who are at risk of filling jobs across levels of the organization. For each position, we need information on who was competing for the job, who made it through each stage of the process, and who was eventually hired. Even when the focus is broadened to consider external hiring across levels (e.g., Cohen et al. 1998; Dreher and Cox 2000; Hassink and Russo 2010), studies have tended to select on the dependent variable, observing only the survivors of the hiring process (see Fernandez and Weinberg [1997] for a discussion of the logical problems of “start with hire” studies). Allocation studies of lower levels of the organizational hierarchy have demonstrated, however, that making accurate inferences about gender differences requires observing not only those who were hired but also those who were considered and not hired (see Fernandez and Sosa 2005; Petersen and Togstad 2006). With few exceptions,2 extant research has lacked adequate data—either internal or external—to study gender stratification in allocation processes by level of the organization. In this article, we seek to correct this imbalance. In contrast to most previous research, we properly specify the candidate pool—both internal and external—at risk of being hired across levels of the organization. We describe where in the organizational hierarchy gender imbalances in job composition occur and the degree to which those imbalances can be traced to external hiring processes. In contrast to past studies of relatively mature, single-firm hierarchies, we study a sample of small- and medium-sized firms in the high-tech sector of the economy. The organizations studied have relatively flat organizational structures and as such provide a conservative test for studying the glass ceiling. While we find elements of the internal labor market model in this setting, we find little evidence of gender biases in internal processes. The fact that external hiring across levels of the hierarchy is common in this sector provides an excellent window for viewing how external hiring dynamics contribute to the glass ceiling. Bolstering our argument that the internal focus of past research has been limiting, we find evidence of external processes contributing to the glass ceiling. Moreover, our findings suggest that policies addressing gender biases in internal promotion are not likely to be effective remedies to gender disparities in organizational level. We find some evidence of demand-side screening biases, but our results suggest that these processes play a comparatively minor role in producing the glass ceiling pattern. The largest determinant of the glass ceiling is that external candidate pools are themselves gendered, with women composing a lower percentage of the applicant pools for higher-level jobs, but a greater proportion of the pools for the lower-level jobs. Our results point to the importance of recruitment and outreach policies designed to improve gender disparities in the formation of candidate pools. Data and Setting The data for this study come from a sample of 441 small- and medium-sized firms in the technology sector that used a common applicant-tracking system. Firms posted job advertisements—sometimes for multiple openings of the same role—to the system. A few postings occurred in February 2008, but applications for these jobs did not begin to flow into the system until March 2008. We focus our analyses on the 50-month period from March 2008 to April 2012.3 This time frame corresponds to the period of recovery following the Great Recession—the jobs open during the early phase attracted many applications, with a gradual decline in the numbers of applicants in the later phase of the study window (Fernandez and Campero 2014). Applicants for positions applied by way of the company’s website by clicking on links to specific jobs. Short job descriptions and information about the company, but not salary information, were provided when applicants clicked on the links. Candidates could apply directly through companies’ websites or via click-through arrangements with Internet job boards. In addition to these external applications, job applications were recorded from internal candidates. Especially important in light of our focus on gender differences, candidates were asked to voluntarily self-identify their gender and race as a part of the application process.4 Our analyses focuses on 2,718 job openings that were filled during the study period. Of these, 53.9% were located in California, 11.6% were located overseas, 6.9% in New York, and the rest were distributed across various other states, with no other state accounting for more than 3% of the job queues. These openings attracted 251,561 applications, 23,738 of these candidates were interviewed, a total of 6,103 received job offers, and 5,055 were hired. These data were collected into the applicant tracking system and were anonymized before being provided to us for use in this study. For several reasons, the high-tech setting constitutes a strategic research site in which to address questions about the organizational roots of the glass ceiling. First, this sector has been traditionally male dominated (Koput and Gutek 2010) and has allegedly displayed glass ceiling barriers for women’s advancement (Blumenthal 2013). Second, similar to Yakubovich and Lup’s (2006) Internet-based recruitment setting, the highly formalized procedures in this setting mirror the suggestions of a number of scholars (e.g., Nelson and Bridges 1999; Bielby 2000; Reskin and McBrier 2000) about the diversity-enhancing benefits of limiting screeners’ discretion. Although the automated nature of the application process makes it less likely that screeners are explicitly steering candidates toward certain roles at the initial application step (Fernandez and Mors 2008), we conceive of the formation of applicant pools as reflecting both candidates’ job choices and firms’ outreach and recruiting activities. Third, as pointed out by Gorman and Kmec (2009: 1430), one challenge in studying patterns of women’s organizational achievement has been the difficulty of gaining access to data on samples of organizations with comparable hierarchies. The data in this context are well-suited to address this challenge. When using this applicant-tracking system, employers registered job openings as being one of six levels: 1) entry-level (e.g., client service representatives, sales associates, junior software engineers); 2) mid-level (e.g., account executives, client support specialists); 3) experienced (e.g., product managers, systems administrators, software engineers); 4) manager (e.g., directors, brand manager, senior manager); 5) executive (e.g., vice presidents and executive directors); and 6) senior executive (e.g., chief financial officer). As might be expected, these jobs are unevenly distributed across the hierarchy: Level 1 (entry-level) jobs make up 19.3% of the data set, level 2 (39.4%), level 3 (29.1%), level 4 (8.1%), level 5 (3.8%), level 6 (0.3%). (Because only seven jobs were filled at the senior executive level (level 6), we combine levels 5 and 6 in subsequent analyses.) Although employers might judge the level of job openings in their own way, they were asked to classify their jobs in the same relative hierarchy.5 Fourth, job applications to positions at these companies came from a number of sources. Most relevant for our purposes is that many applications to these job openings were from people already employed at the firm, that is, internal candidates. For any particular opening, however, external candidates might also have applied and found themselves in competition with internal candidates. As a consequence, within the limits of the study design (see footnote 5), we were able to properly specify both the internal and the external risk sets for jobs at distinct organizational levels and to measure the extent to which gender differences in level at the point of hire can be traced to both internal and external hiring processes. Last, as previously mentioned, past organizational studies of gender allocation by level (Fernandez and Mors 2008; Fernandez and Abraham 2010, 2011) have lacked controls for candidates’ skills and experience. By contrast, in the current setting we have controls for key information on human capital and other background characteristics. As part of the online application process, applicants’ résumés were parsed to capture their contact information and work experience. Career histories were coded based on dates of employment to obtain the number of years of experience. Applicants were also asked questions about their recruitment source. Distance (measured in air miles) between the location of the job to which the person is applying and the candidates’ home address was automatically calculated and recorded as part of the application process. This step allows us to control for well-known gender differences in commuting patterns (Madden 1981; Fernandez and Su 2004). We also control for the timing of the application (see, e.g., Fernandez and Weinberg 1997) by measuring the number of days between when the job first opened and when the candidate applied for the job. Also important in light of the evidence that jobs are often sex-typed (Correll 2001; Fernandez and Friedrich 2011), employers were asked to classify the job openings into functional areas. Thus, we are able to control for the functional area of the job opening by adding fixed effects for the following job functions: IT/engineering (29.4% of the jobs), production/operations (17.1%), marketing (14.1%), sales (11.5%), client service (8.5%), administration (6.6%), human resources (2.4%), and other (10.3%). Finally, although many candidates did not choose to do so, applicants were asked to voluntarily self-identify their racial background; we include a set of dummy variables for race as controls in the multivariate analyses. We believe these firms’ hiring practices are typical of other small- and medium-sized firms, as many of them use the Internet for recruiting (Autor 2001; Cappelli 2001; Kerka 2001). Our main goal in adopting this empirically grounded, case-study approach is to elucidate the workings of previously understudied external mechanisms that affect gender composition at various levels of the organizational hierarchy. The theoretical significance of this study is that it allows us to observe the operations of a set of processes that are normally hidden from view, especially when compared with extant single organization studies of the internal and external roots of the glass ceiling (e.g., Fernandez and Mors 2008; Fernandez and Abraham 2010, 2011). Analysis We argued that the currently dominant understanding of the glass ceiling as reflecting internal promotion barriers is unnecessarily limiting. To the extent that organizations depart from the internal labor market model (Doeringer and Piore 1971), gender differences in external recruitment might also contribute to gender inequality in organizational rank. We begin by presenting evidence on the degree to which these firms conform to the internal labor market model. Table 1 shows the degree to which external candidates (hereafter, “externals”) are present across four stages of the job allocation process by level of the organization. Broadly consistent with the internal labor market model, the data presented in the first column show that the highest percentage of external hires is found in the lowest rung of these organizations (83.09%), and the lowest percentage of external hires is found in the top tier (57.14%). We would be hard pressed to interpret this pattern as evidence of distinct “ports of entry” at the bottom of the firm, however, since nontrivial percentages of the job openings at each level are filled by externals. Indeed, at all levels of the hierarchy, internal moves account for the minority of jobs filled across these firms. Even at the top level, the clear majority of hires are coming from external sources (57.14%). Other studies, to varying degrees, have found the same pattern of many jobs being externally filled at higher organizational levels for large, mature organizations (Baker et al. 1994; Dohmen et al. 2004; Lazear and Oyer 2004; Hassink and Russo 2010; Bidwell and Keller 2014). These studies, however, observed only the outcome of the hiring process. To understand how the allocation process is structured in organizational hierarchies, we need to observe the set of candidates at risk of obtaining those jobs. Table Table 1. Percentage of External Candidates at Four Stages of the Screening Process, by Organizational Level Table 1. Percentage of External Candidates at Four Stages of the Screening Process, by Organizational Level View larger version Column (4) of Table 1 lists the external/internal composition of the candidates competing for job openings, by level. Irrespective of level, the vast majority—more than 90%—of the candidates considered for these jobs are externals. For each of the subsequent steps—interview, offer, and hire—the percentage of externals decreases compared to the applicant pool for all levels, indicating that internal candidates (hereafter, “internals”) are favored in obtaining access to jobs throughout the screening process (see Figure 1). While screeners are selecting internals at higher rates than externals at all levels, the across-level pattern in Figure 1 indicates an increasing preference for internals as one goes up the levels of these organizations. figure Figure 1. Percentage of External Candidates, by Level and Stage of Screening Process These findings have important implications for our understanding of the external bases of the glass ceiling. Unlike the ideal-typical imagery in the classical model of the internal labor market, internals and externals are in competition for jobs at all levels of these organizations. Although the fan-like pattern in Figure 1 indicates an increasing preference for internals as one rises through the levels, external allocation processes clearly still matter, even at the top of the hierarchy. An exclusively internal focus is especially questionable in this context, in which small- and medium-sized firms are not likely to contain enough people to meet their personnel needs by relying solely on internal promotion processes. These patterns suggest that internally focused studies of gendered promotion barriers are yielding an incomplete, and possibly distorted, view of how gender affects the allocation of people to jobs across organizational levels. Indeed, Table 2 shows a glass ceiling pattern of women disappearing as one ascends the levels of the organization, for both internals and externals. Considering first the hires (cf. columns (1) and (2)), we find that in both cases the percentage of female hires declines approximately 15 points from the lowest (entry-level) to the highest (executive) levels of these organizations. A very similar pattern of declining female representation across levels is evident for job offers as well (cf. columns (4) and (5)). These outcomes clearly show that the external job-matching process alone can produce a glass ceiling result. Table Table 2. Percentage of Female Hires, Job Offers, and Applications—Internal and External—by Organizational Level Table 2. Percentage of Female Hires, Job Offers, and Applications—Internal and External—by Organizational Level View larger version Figure 2 plots the percentage of female candidates for hires and job offers across levels. Two patterns emerge. First, for both internals and externals, the pattern of job offers closely tracks the pattern of hires. The one exception is the case of the internals at the executive level (level 5), for which females are more prevalent among those receiving job offers than among those hired. Second, the lines for both external job offers and hires are always above the lines for internal offers and hires. Females are always more prevalent among external candidates than among internal candidates. Interestingly, however, this external pattern is quite parallel to the internal pattern. Thus, a glass ceiling pattern is emerging on organizational entry for these small- and medium-sized firms. figure Figure 2. Percentage Female, by Organization Level We now consider screeners’ role in producing the glass ceiling pattern shown in Table 2 and Figure 2. We investigate the degree to which males are being allocated to job openings at higher rates than females are across levels of the organizational hierarchy.6 If screeners view males and females as being appropriate for distinct sets of jobs, then hiring agents may bias their hiring decisions against females who are applying for male-typed jobs, and vice versa. Indeed, some scholars (e.g., Cejka and Eagly 1999; Ridgeway 2011) argued that rank is itself gendered, so that higher-level jobs are likely to be male typed. This line of reasoning leads to an expectation that screening disadvantages for women should increase as one goes up organizational levels (Gorman and Kmec 2009). Columns (7) to (9) in Table 2 show the percentage of female job applications by level of the hierarchy for internal and external candidates. If gender-biased screening is present, we would expect a higher prevalence of females in the overall applicant pool relative to the pool of candidates with job offers. Further, if screeners are playing a role in producing the glass ceiling pattern, we would expect the under-selection of females in screening to increase as we move up the organizational hierarchy. Comparing the share of females in the applicant pool (Columns (7) to (9)) with the share of females with job offers (Columns (4) to (6)) shows no evidence of increasing under-selection of females by level, for either internal or external candidates. Next, using multivariate analysis, we analyze gender differences in screening odds. Table 3 presents descriptive statistics for the variables used to predict the rates of interview and job offer7 across organizational levels. Over all levels, and without controls, males are more likely than females to be interviewed (9.73% compared to 8.69%), receive job offers (2.60% compared to 2.11%), and be hired (2.11% compared to 1.82%). Males applied for jobs that are, on average, one-quarter of a level higher than those for which females applied. Males also have higher mean values for years of work experience and years of management experience when compared with women. Consistent with past research (Madden 1981), females are applying to jobs that are closer to home. Males tend to apply for job openings an average of one month later than do females. Gender differences are also apparent in the application source, with males constituting a higher percentage of internal candidates and referrals, but lower percentages of candidates utilizing external websites or job boards. With respect to race, we see that a higher proportion of males than females did not provide information on their racial background (40.04 compared to 33.37%), and that the percentage of African Americans is higher among females than males (7.40 compared to 4.13%). Table Table 3. Summary Statistics of Variables Used in Multivariate Analysis Table 3. Summary Statistics of Variables Used in Multivariate Analysis View larger version Finally, Table 3 also shows important gender differences in the pattern of applications by job function. A dramatically higher representation of males than females exists in the candidate pools for IT and engineering jobs (37.63% for males compared to 17.29% for females). In marked contrast, female candidates are more plentiful than males in the candidate pools for administrative (18.60% compared to 4.47%) and human resources functions (6.92% compared to 1.77%). (For a similar pattern, see Fernandez and Friedrich 2011.) It is important to recognize that such “application segregation” (Barbulescu and Bidwell 2013) corresponds to gender-stereotypic images of occupations (Cejka and Eagly 1999). If screeners view males and females as being appropriate for distinct jobs (e.g., Mun 2010), then screeners may further exacerbate and reinforce the initial gender segregation of applicant pools by biasing their interview and hiring decisions against people applying for gender-atypical jobs. Indeed, such gender-typed screening bias has often been found in well-designed audit studies of gender discrimination in the labor market (for a thorough review, see Azmat and Petrongolo 2014). If male-typed jobs tend to be higher in the organizational hierarchy, then gender-biased screening will contribute to the relative absence of women at higher levels of the organization. We explore this possibility by analyzing screeners’ interview decisions. Table 4 reports the results of logistic regression models predicting screeners’ interview decisions. We use logit models with robust errors clustered by candidate to account for those candidates being considered for multiple vacancies. To the degree that hiring agents statistically discriminate on the basis of gender stereotypes, these effects would likely be most evident at the interview stage, when screening is based on paper credentials. We present models of the entire population (Table 4, columns (1) and (2)), as well as models stratified by job function. In light of research suggesting that IT/engineering is male typed (Koput and Gutek 2010), we analyze the screening process for IT/engineering jobs separately (columns (3) and (4)). We also analyze the pooled set of female-typed human resources and administrative jobs (Roos and Manley 1996) (columns (5) and (6)). Finally, we combine the data for the remaining job functions and analyze them separately (columns (7) and (8)). Across all these models, we test for interactions between gender and organizational level, looking specifically for a pattern of increasing bias against women at higher levels of the organization. Table Table 4. Logit Models of Screeners’ Interview Decisions Table 4. Logit Models of Screeners’ Interview Decisions View larger version Considering first the entire sample without controls (Table 4, column (1)), we find the odds ratio for the main effect of being female is slightly greater than 1 (1.0224), but not statistically significant. Internals, however, are 4.4 times more likely to be interviewed than externals, but this effect does not differ for males and females (the 0.99 odds ratio for female × internal interaction term is not statistically different from 1).8 The dummy variables for levels show that compared to level 1, applicants of both genders have significantly higher chances of interview at levels 2 through 4. Finally, the female × level interactions specify the test of whether the female effect at each level is different from the overall female effect. These coefficients provide the key test of whether women are increasingly disadvantaged in screening at higher levels of the organization. Except for level 3, where females are 0.9120 times as likely to be interviewed than entry-level females, no significant level differences exist in the likelihood of females being interviewed as compared to males. In column (2), we repeat the model, adding the control variables listed in Table 3. Although a number of these controls are significantly related to the chances of interview in ways that might be expected (see Appendix Table A.1, column (1)),9 the key results with respect to gender are virtually identical to those in the model without controls. Subsequent columns in Table 4 repeat the analyses for separate job functions. Columns (3) and (4) report the results of logit models predicting the interview rates for the stereotypically male IT/engineering jobs, the context in which gender-biased screening against women is most likely to occur. Here, the coefficient for the main effect of gender shows that females have significantly lower chances of being granted an interview than do males across all levels of the organization (odds ratio = 0.6734). This effect is robust to controls (column (4)) (female/male odds of interview = 0.7047). Similar to models in columns (1) and (2), we find in columns (3) and (4) a strong tendency for internals to be interviewed (odds ratios of 4.00 and 3.53), but the coefficient on the female × internal interaction term reveals that this tendency is not weaker for women. With respect to the key female × job level interactions, we find no evidence of increasing disadvantage as one proceeds upward through the hierarchy. Indeed, for models with and without controls, the interaction with level 4 shows the opposite: chances of interview are significantly higher for females at level 4 than they are for females at level 1. Rather than a glass ceiling, this pattern of results is more consistent with that of a “sticky floor” (Booth et al. 2003; Zeng 2011) wherein gender biases are strongest for the lowest-level jobs. Columns (5) and (6) report the results for the stereotypical female HR and administrative jobs combined (substantively similar results are obtained when the two categories are analyzed separately). In the model without controls (column (5)), we find that, overall, women have better chances of being selected for an interview than do men. This effect becomes statistically insignificant, however, when controls are added to the model (column (6)). With respect to the other organizational levels, we find no evidence of statistically significant female × level interactions. The final two columns of Table 4 present the results for the other job functions considered together. Here, women, overall, have higher chances of being interviewed, and this effect is robust to controls. The lack of significance for the female × internal interaction effects shows that this female advantage does not differ between internals and externals. When examining the key female × level interactions, we find no statistically reliable evidence of an increasing screening disadvantage against women across levels. Moreover, these patterns remain even when we compare women and men who are competing for the same job opening.10 Table 5 reports the results of analyses designed to examine gender differences in job offer rates, conditional on the candidate being interviewed. Overall, we find that at this stage females are less likely than males to receive a job offer. But, here too, we find no statistically reliable evidence that the gender difference in job offer rates varies between internals and externals, or for the various levels of the organization. This finding holds true for models with and without controls (cf. columns (1) and (2) in Table 5). To understand how women fared across both of these steps combined, we re-estimated these models by predicting the job offer rate for the full sample, not conditional on the candidate having been interviewed. Across both stages, we find that females have lower odds of receiving job offers (odds ratio = 0.8587) net of controls, but this gender difference did not vary by internal status or level. Table Table 5. Odds of Job Offer Conditional on Interview Table 5. Odds of Job Offer Conditional on Interview View larger version Table 5, columns (3) and (4), predict job offers for candidates applying to IT/engineering jobs. In contrast with the interview step of the screening process, here we find no evidence of statistically reliable gender differences in job offer rates at any level of the organization. Re-estimating the model for job offer not conditional on interview for IT/engineering jobs, we find that across both stages, the odds ratio for females with controls is 0.8052, but this effect is not statistically reliable (z-value = −1.09). Nor were there any significant female × internal or female × level interactions. Here, too, these results are the same for within-job analyses (estimated with fixed effects for job queues), wherein men and women are in competition for the same job openings. In sum, across these analyses, we find some evidence of gender disparities in screening, but no evidence of screeners’ biases differing significantly by level. Consequently, screening disparities do not contribute to the glass ceiling, through either external or internal screening. If the very notion of high rank is stereotypically male, this notion is not being reflected in these screeners’ actions (cf. Gorman and Kmec 2009). As we reviewed previously, theories emphasizing the role of labor supply factors in hindering women’s organizational advancement argue that men and women pursue different kinds of positions. Although this research is mostly about horizontal segregation of males and females across occupations, to the degree that occupations with distinct gender associations are more prevalent at certain organizational levels, such segregation can also take on a vertical dimension. Table 6 presents important data in this respect, as it includes descriptive information on the percentage of female candidates by level within each job function. Looking down the columns for each of the job functions, the percentage of female candidates is considerably higher in entry-level jobs than the percentage of females seeking executive jobs. Although the decline in female representation across levels is not always smooth,11 a glass ceiling pattern of females disappearing as one goes up levels of the hierarchy appears within each of the functions. Further, looking across the rows of Table 6, we also see evidence of horizontal application segregation: female applicants are underrepresented in the stereotypically male IT/engineering function and overrepresented in administration and human resources jobs. This same pattern is repeated within each level. Together, these patterns suggest that the horizontal and vertical dimensions of gendered application patterns are relatively uncorrelated in these applications. The lack of correlation gives some initial insight into why we did not find demand-side screeners’ decisions for gender-stereotypical job functions significantly contributing to a vertical glass ceiling. Moreover, it suggests that the seeds of the glass ceiling are already evident in the initial candidate pools, even before hiring agents begin the screening process within job function. Table Table 6. Percentage of Female Applications, by Level and Department Table 6. Percentage of Female Applications, by Level and Department View larger version “What If” Simulations To assess the relative contribution of supply- and demand-side factors to the glass ceiling, we perform two sets of policy-relevant “what if” simulations based on our results. Because the processes are largely parallel, we focus on understanding job offer patterns for both internals and externals combined. For simplicity and to help understand the net effects of demand-side screening, we compress the interview and the job offer steps by focusing on job offers, unconditional on interview. We compare, using two kinds of scenarios, the observed gender distribution of job offers at each level. First, we simulate the percentage of female candidates among candidates receiving job offers, assuming no gender disparities at all in job applicant screening at any level. Addressing screening discrimination is currently the focus of most bias-mitigation efforts in organizations (Kalev et al. 2006). This simulation provides a picture of what the distribution of females across organizational levels might look like if we were to eliminate all gender bias in candidate screening processes. The mental experiment in this case corresponds to equalizing the chances of receiving a job offer for males and females; that is, we are fixing the gender odds ratios in the screening models to 1.0. Second, we simulate the gender distribution of job offers assuming screening disparities remain at observed levels, but equalize the gender distribution of the candidate pools by organizational rank. Specifically, for each level of the hierarchy, we set the distribution of females in the candidate pools to the percentage of female candidates across all job levels and use the observed point estimates derived from logistic regression models.12 In the case of the whole population, females represent, on average, 36.25% of the candidates for these jobs (see last row and column of Table 6). This percentage varies considerably by job level, however, from 45.53% for entry-level jobs to 27.57% for executive-level jobs. In this scenario, we fix the percentage of female candidates at 36.25% at each level, in essence redistributing women from job queues at the lower level to the higher level, and then applying the estimated odds ratios for all variables from the appropriate model. Figure 3 shows the simulated results for all candidates across all job functions. The observed gender composition of candidates receiving job offers follows a glass ceiling pattern of decline from 42.38% female composition at the entry level to 27.40% at the executive level. Perhaps surprisingly, full gender parity in screening at every level (scenario 1) would produce a very similar gender composition of the candidates receiving job offers: females would compose 45.53% at entry level, and this representation would decline to 27.57% at the executive level. These analyses suggest that, taken alone, policies that produce gender parity in screening would leave the glass ceiling firmly intact. figure Figure 3. Simulated Percentage of Female Job Offers—All Functions By contrast, scenario 2, which assumes an equal gender composition of the applicant pool at each level, produces a much more equitable gender distribution across organizational levels. At lower levels, fewer females would be present, but these women would be redistributed to higher organizational levels. The 95% confidence interval reveals considerable uncertainty in the estimates as the number of job offers used to estimate the screening odds become thinner at the higher levels of the hierarchy; but given these analyses, the most likely outcome is a considerably flatter distribution of women across all levels of these organizations. Next, using the same estimation strategy, we consider the case of IT/engineering jobs.13 Here, too, we simulated two scenarios (see Figure 4). The observed gender composition of candidates receiving offers for these jobs is highly male dominated. Further, the absence of females among those receiving offers for IT/engineering jobs becomes significantly more pronounced as one moves up the levels of the hierarchy: from 18.04% female at the entry level to 9.38% at the executive level. The results of scenario 1 suggest that for jobs at the three lower organizational levels, screening disparities play an important role in accounting for the absence of women with job offers. Eliminating screening bias in the competition for jobs at these levels would lead to an increase in female representation of between 6 and 8 percentage points. This increase is considerable given that on average females constitute only 13.4% of the candidates receiving job offers at these levels. The results of scenario 1 also suggest, however, that eliminating screening bias would have virtually no effect on the gender composition of candidates receiving job offers at the highest two levels. figure Figure 4. Simulated Percentage of Female Job Offers—IT/engineering With respect to scenario 2, the simulated line is very close to the observed line for the bottom levels of jobs. Thus, changing the gender supply of candidates alone would do nothing to address the decline in female representation across these three levels. At higher organizational levels, the results of scenario 2 suggest that the decline in female prevalence is driven by the relatively low percentage of females present in the applicant pools for these jobs. Here, too, the small number of females in applicant pools for upper-level IT/engineering jobs makes our estimates of their screening odds quite imprecise. The point estimates produced in scenario 2, however, suggest at a minimum that the absence of females in applicant pools to senior IT/engineering jobs is a bigger contributor to the glass ceiling in these fields than is biased screening. One final point with respect to the scenario 2 simulations is in order. As previously mentioned, setting the female representation in the applicant pools to the percentage of females across all jobs—36.68% for all job functions and 20.72% for IT/engineering jobs—in essence redistributes female candidates from lower-level job queues to higher-level queues while leaving the overall supply of female candidates constant. Although this redistribution eliminates the downward-sloping glass ceiling pattern in Figure 3, consider that the composition of females in these candidate pools is starting from a relatively low percentage. This condition is especially true in the case of the IT/engineering jobs. It is worth considering what would happen if the overall supply of female candidates were to increase, for example, as a result of outreach policies designed to attract more women to the high-tech sector overall, or into engineering jobs in particular. Our results imply that, ceteris paribus, these policies would have the effect of shifting the scenario 2 lines vertically upward across the board. Of course, demand-side screeners’ biases against women might also attenuate if such an overall rise in female representation were to be achieved. But at that point we are beyond the limits of the insights that might be gleaned from these simulations. Conclusion In this article, we offer a fresh perspective on the organizational roots of the glass ceiling. We suggest that past understandings of the phenomenon as reflecting internal promotion biases are unnecessarily limiting. We argue that many jobs are filled through external hiring, even at higher levels of the organization. To the extent that external processes are gendered, then internally focused studies can yield an incomplete and misleading picture of the organizational barriers to women’s advancement. For this reason, our research has highlighted the role of external recruitment and hiring processes. Using data gathered from 441 small- and medium-sized high-tech firms, we offer empirical evidence on how external processes strongly contribute to producing a glass ceiling pattern. Across levels of these firms’ hierarchies, we find some evidence of demand-side screening biases against women, but these disparities do not differ by level of the organization, and thus do not account for the increasing absence of females at the top levels of these organizations. Instead, we find that supply-side processes that lead to the formation of applicant pools at separate levels of the hierarchy play a powerful role in producing the glass ceiling. Gender disparities on the supply side may partly reflect the legacy of male-typing in the high-tech industry (Koput and Gutek 2010), in which relatively few females had opportunities to gain the necessary experience to pursue jobs at higher organizational levels. Our findings suggest, however, that efforts aimed at increasing the supply of female candidates—both by individual firms and industry-wide—may pay greater dividends than would efforts to address gender bias in screening. Although other researchers have also found a reliance on external hiring across levels of the organization, such external processes are especially relevant in the context of these young, rapidly growing high-tech organizations. Several studies of large, mature organizations have also found gendered external recruitment processes. Our study is the first, however, to document a glass ceiling being formed through the early influx of new personnel. These small firms might not yet have large pools of internal candidates to stoke an internal glass ceiling pattern, but their rapid growth generates many vacancies and promotion opportunities for candidates who are likely to be external. That these companies were resorting to large-scale recruiting over the Internet suggests a desire to expand their potential talent pool beyond their current workforce and direct network ties. For young, rapidly growing high-tech firms, focusing on external recruitment and hiring processes is likely to be the most effective way of tackling the glass ceiling. In this respect, our data analyses and “what if” simulations provide further guidance for designing policy efforts to address the glass ceiling. Our results suggest that, by themselves, bias-remediation policies designed to reduce gender discrimination in screening are likely to be of limited help in addressing the problem of the glass ceiling. By contrast, recruitment and outreach policies aimed at producing less-gendered candidate pools for jobs at various levels of the hierarchy are likely to pay the biggest dividends for organizations seeking to produce a more equitable distribution of men and women across their ranks. Appendix Table Table A.1. Full Logit Models Table A.1. Full Logit Models View larger version Table Table A.2. Replication of Models in Tables 4 and 5, within Racea Table A.2. Replication of Models in Tables 4 and 5, within Racea View larger version Additional results and copies of the computer programs used to generate the results presented in the article are available from the authors at or Notes 1Such concerns about the appropriateness of the risk set also apply to studies that examine gender patterns within and across organizations within a single market or system, as in the examples of Cohen and colleagues’ (1998) study of California savings and loan organizations. To the degree that hiring occurs from outside the system, then the risk set will be mis-specified, thus hindering inferences about gender differences in allocation to jobs across the system. 2See Fernandez and Mors (2008); Gorman and Kmec (2009); and Fernandez and Abraham (2010, 2011). In contrast with Kay and Hagan’s (1995) study of promotions in law firms, Gorman and Kmec (2009) specified both the internal and the external risk set when examining gender stratification within the relatively closed system of elite law firms. They found a pattern of increasing female disadvantage in internal promotion, but not in external hires. Fernandez and colleagues examined gender differences in both internal and external allocation within the context of three separate firms: a call-center (Fernandez and Mors 2008), a retail bank (Fernandez and Abraham 2010), and a BioPharma firm (Fernandez and Abraham 2011). All three of these latter studies lacked controls for human capital so caution is warranted. The descriptive evidence, however, suggests that external hiring processes alone can produce a glass ceiling pattern in which women become relatively scarcer at higher levels of the organizational hierarchy. 3Overall, the data set used includes 544,462 applications to 13,368 jobs at 1,270 companies received from December 2007 to March 2012. Of these 13,368 jobs, 581 were to student internship positions and were excluded from the analysis (yielding a sample of 517,981 applications to 12,787 full-time jobs). An additional 9,574 full-time jobs (253,307 applications) were excluded because of censoring (i.e., no hires occurred during the observation period). This led to a sample of 264,674 applications to 3,213 jobs. We also excluded 307 single-person jobs. For a portion of these queues (103), the candidate hired was internal, which suggests these were likely promotions for which a single internal candidate was considered. Excluding these jobs yields a sample of 2,906 jobs and 264,367 applications. Finally, we excluded 50 firms (190 jobs) in the data set that were not in the high-tech industry. Excluding these firms yielded a data set of 251,561 applications to 2,718 jobs at 441 firms. These 251,561 applications came from 234,011 individuals (1.07 applications per individual). 4Of our final sample of 251,561 applications, 44,177 (17.6%) did not include responses to questions about the applicants’ race or gender. In an additional 43,267 (17.2%) cases, the applicant selected “Decline to Identify” on the gender question, and in 52,638 (20.9%) cases, the applicant declined to identify his or her race. We were given access to the candidates’ first names and used the IBM InfoSphere Name Management Tool ( to score each name in terms of the likelihood of being female and were able to code gender for all but 3,488 (4.0%) of the 87,444 applicants who did not voluntarily provide their gender. Overall, we were able to identify gender for 248,073 cases, or 98.6% of these applications. 5Although our data include job openings across the full organizational hierarchy (i.e., from entry level to senior executive), firms may have also hired for other jobs without posting them on the system. In particular, firms may be less likely to post a job online when they have a strong internal contender to fill it. Given that we cannot rule out this possibility, our intent is not to discount the potential importance of internal promotion processes in contributing to the glass ceiling but rather to examine the extent to which external hiring processes can be an additional mechanism that leads to the relative scarcity of women at higher organizational levels. 6Note that unlike audit studies (Azmat and Petrongolo 2014), we do not have the benefit of random assignment when assessing screeners’ gender bias. We have, however, coded an unusually rich set of variables to control for some alternative explanations for our findings. One strength of our approach is that we study live hiring and examine both the interview and the job offer steps of the process. By contrast, audits are generally limited to examining the initial stages of contact with employers. Indeed, it may be ethically dubious and inherently infeasible to conduct a random assignment study of live hiring. Nevertheless, our analyses remain subject to the caveat that unobserved factors might explain our findings. 7Substantively similar results are obtained when hire is the outcome variable. 8In preliminary analyses, we found no evidence of a three-way interaction between female, internal, and level. 9In particular, years of work experience are positively and significantly related to interview chances, but this advantage is curvilinear and levels off since the associated squared term is also significant and negative. Years of management experience, however, works in the opposite way, with the linear term negatively related to interview chances, with a positive but insignificant squared term. Compared with candidates applying by way of the website, external referrals and those using “other” sources of application are also more likely to be interviewed. As expected, people living farther from work are less likely to be interviewed. Those applying later are slightly more likely to be interviewed than those applying earlier. Interestingly, the dummy variables for race indicate that racial minorities are less likely to be interviewed than are majority whites. While this is a focus of another article (Fernandez and Campero 2014), we checked whether race interacts with gender in the screening process by level in this context. Appendix Table A.2 documents that the gender differences by level we study here are also happening within race groups. The table shows no significant gender disparities in odds of interview for white, Asian, African American, or Hispanic applicants. Finally, in separate analysis, we find no evidence of a racial glass ceiling; although minorities are disadvantaged in screening, this disadvantage does not increase by level of the organization. 10The models in Table 4 pool candidates across job queues within each level of the organization. For this reason, it is possible that screeners’ biases might be masked by within-level heterogeneity in job queues. As a final check on whether gender-biased screening for interview is occurring at various levels of the organization, we further disaggregate all the analyses in Table 4 by estimating linear probability models with fixed effects for job queues, which purges all between-queue factors—both observed and unobserved—associated with job queues. We find the same pattern of findings as those in Table 4. 11Not all firms filled jobs at each level within each function and, as such, differences in gender composition by function and level can reflect differences between firms that posted jobs in each cell. For example, in the case of marketing, the share of females increases from the experienced level to the manager level. Firms that posted manager-level marketing jobs, however, are firms that attract more females overall. Although firm-level variation can generate unevenness in the pattern, the general trend in Table 6 shows fewer females in higher-level applicant pools within job function. 12To clarify the effects of gender at each level, we stratified the data by level and estimated the unconditional probability of job offer using all the controls used in Table 4, model 2. For the purposes of the simulations, we used level-specific point estimates, irrespective of whether the estimated odds were statistically significant. For the simulations corresponding to the whole population (Figure 3, scenario 2), the female/male odds of job offer by level are as follows: entry-level = 0.8916 (n.s.); mid-level = 0.8522 (p < .01); experienced = 0.8328 (p < .01); managers = 0.9512 (n.s.); executives = 1.0155 (n.s.). The results of the simulation show a very similar pattern if we set the odds ratios for the insignificant effects to 1.0. 13The estimated female/male odds of job offer by level are: entry-level = 0.8856 (n.s.); mid-level = 0.7322 (p < .01); experienced = 0.5538 (p < .01); manager = 1.3893 (n.s.); executive = 1.3413 (n.s.). The results of this simulation show a very similar pattern if we set the odds ratios for the insignificant effects to 1.0. References Albrecht, James, Björklund, Anders, Vroman, Susan. 2003. Is there a glass ceiling in Sweden? Journal of Labor Economics 21: 145–77. Google Scholar, Crossref, ISI Arulampalam, Wiji, Booth, Alison L., Bryan, Mark L. 2007. Is there a glass ceiling over Europe? Exploring the gender pay gap across the wage distribution. Industrial and Labor Relations Review 60:163–86. Google Scholar, SAGE Journals Autor, David H. 2001. Wiring the labor market. Journal of Economic Perspectives 15: 25–40. Google Scholar, Crossref, ISI Azmat, Ghazala, Petrongolo, Barbara. 2014. Gender and the labor market: What have we learned from field and lab experiments? Labour Economics 30: 32–40. Google Scholar, Crossref, ISI Baker, George, Holmstrom, Bengt. 1995. Internal labor markets: Too many theories, too little evidence. American Economic Review 85(2): 255–59. Google Scholar, ISI Baker, George, Gibbs, Michael, Holmstrom, Bengt. 1994. The internal economics of the firm: Evidence from personnel data. Quarterly Journal of Economics 109: 881–919. Google Scholar, Crossref, ISI Barbulescu, Roxana, Bidwell, Matthew. 2013. Do women choose different jobs from men? Mechanisms of application segregation in the market for managerial workers. Organization Science 24(3): 737–56. Google Scholar, Crossref, ISI Barnett, William P., Baron, James N., Stuart, Toby E. 2000. Avenues of attainment: Occupational demography and organizational careers in the California civil service. American Journal of Sociology 106: 88–144. Google Scholar, Crossref, ISI Bertrand, Marianne, Hallock, Kevin F. 2001. The gender gap in top corporate jobs. Industrial and Labor Relations Review 55: 3–21. Google Scholar, SAGE Journals Bidwell, Matthew . 2011. Paying more to get less: The effects of external hiring versus internal mobility. Administrative Science Quarterly 56(3): 369–407. Google Scholar, SAGE Journals, ISI Bidwell, Matthew, Keller, Joseph. 2014. Within or without? How firms combine internal and external labor markets to fill jobs. Academy of Management Journal 57(4): 1035–55. Google Scholar, Crossref, ISI Bidwell, Matthew, Mollick, Ethan. 2014. Shifts and ladders: Comparing the role of internal and external mobility in executive careers. Working paper. Philadelphia: Wharton School of Management of the University of Pennsylvania. Google Scholar Bielby, William T. 2000. Minimizing workplace gender and racial bias. Contemporary Sociology 29: 120–29. Google Scholar, Crossref, ISI Blau, Francine D., Brinton, Mary C., Grusky, David B. (Eds.). 2006. The Declining Significance of Gender? New York: Russell Sage Foundation. Google Scholar Blumenthal, Richard . 2013. Breaking Silicon Valley’s glass ceiling. Accessed at (April 26, 2015). Google Scholar Booth, Alison L., Francesconi, Marco, Frank, Jeff. 2003. A sticky floors model of promotion, pay, and gender. European Economic Review 47: 295–322. Google Scholar, Crossref, ISI Brett, Jeanne M., Stroh, Linda K. 1997. Jumping ship: Who benefits from an external labor market career strategy? Journal of Applied Psychology 82: 331–41. Google Scholar, Crossref, ISI Cappelli, Peter . 1999. Career jobs are dead. California Management Review 42: 146–67. Google Scholar, SAGE Journals, ISI Cappelli, Peter . 2001. Making the most of online recruiting. Harvard Business Review 79:139–46. Google Scholar, Medline, ISI Cappelli, Peter, Hamori, Monika. 2005. The new road to the top. Harvard Business Review 83: 25–32. Google Scholar, Medline, ISI Cejka, Mary Ann, Eagly, Alice H. 1999. Gender-stereotypic images of occupations correspond to the sex segregation of employment. Personality and Social Psychology Bulletin 25: 413–23. Google Scholar, SAGE Journals, ISI Cohen, Lisa E., Broshak, Joseph P., Haveman, Heather. 1998. And then there were more? The effect of organizational sex composition on the hiring and promotion of managers. American Sociological Review 63: 711–27. Google Scholar, Crossref, ISI Collinson, David L., Knights, David, Collinson, Margaret. 1990. Managing to Discriminate. London: Routledge. Google Scholar Correll, Shelley J. 2001. Gender and the career choice process: The role of biased self-assessments. American Journal of Sociology 106: 1691–1730. Google Scholar, Crossref, ISI Cotter, David A., Hermsen, Joan M., Ovadia, Seth, Vanneman, Reeve. 2001. The glass ceiling effect. Social Forces 80: 655–81. Google Scholar, Crossref, ISI DiPrete, Thomas, Soule, Whitman T. 1988. Gender and promotion in segmented job ladder systems. American Sociological Review 53: 26–39. Google Scholar, Crossref, ISI Doeringer, Peter B., Piore, Michael J. 1971. Internal Labor Markets and Manpower Analysis. Lexington, MA: Heath Lexington Books. Google Scholar Dohmen, Thomas J., Kriechel, Ben, Pfann, Gerard A. 2004. Monkey bars and ladders: The importance of lateral and vertical job mobility in internal labor market careers. Journal of Population Economics 17: 193–228. Google Scholar, Crossref, ISI Dreher, George F., Cox, Taylor H. 2000. Labor market mobility and cash compensation: The moderating effects of race and gender. Academy of Management Journal 43(5): 890–900. Google Scholar, Crossref, ISI Fernandez, Roberto M., Abraham, Mabel Bothelo. 2010. From metaphors to mechanisms: Gender sorting in(to) an organizational hierarchy. Paper presented at the 2010 annual meetings of the American Sociological Association, Atlanta. Accessed at Google Scholar Fernandez, Roberto M., Abraham, Mabel Bothelo. 2011. Glass ceilings and glass doors? Internal and external hiring in an organizational hierarchy. Paper presented at the 2011 annual meetings of the American Sociological Association, Las Vegas. Accessed at Google Scholar Fernandez, Roberto M., Campero, Santiago. 2014. Does competition drive out discrimination? Paper presented at the 2014 annual meetings of the American Sociological Association, San Francisco. Google Scholar Fernandez, Roberto M., Friedrich, Colette. 2011. Gender sorting at the application interface. Industrial Relations 50: 591–609. Google Scholar, Crossref, ISI Fernandez, Roberto M., Mors, Marie Louise. 2008. Competing for jobs: Labor queues and gender sorting in the hiring process. Social Science Research 37: 1061–80. Google Scholar, Crossref, Medline, ISI Fernandez, Roberto M., Sosa, M. Lourdes. 2005. Gendering the job: Networks and recruitment at a call center. American Journal of Sociology 111: 859–904. Google Scholar, Crossref, ISI Fernandez, Roberto M., Su, Celina. 2004. Space and the study of labor markets. Annual Review of Sociology 30: 545–69. Google Scholar, Crossref, ISI Fernandez, Roberto M., Weinberg, Nancy. 1997. Sifting and sorting: Personal contacts and hiring in a retail bank. American Sociological Review 62: 883–902. Google Scholar, Crossref, ISI Gorman, Elizabeth H., Kmec, Julie A. 2009. Hierarchical rank and women’s organizational mobility: Glass ceilings in corporate law firms. American Journal of Sociology 114: 1428–74. Google Scholar, Crossref, Medline, ISI Harris, Dawn, Helfat, Constance. 1997. Specificity of CEO human capital and compensation. Strategic Management Journal 18(11): 895–920. Google Scholar, Crossref, ISI Hassink, Wolter J., Russo, Giovanni. 2008. Wage differences between internal and external candidates. International Journal of Manpower 29(8): 715–30. Google Scholar, Crossref, ISI Hassink, Wolter J., Russo, Giovanni. 2010. The glass door: The gender composition of newly-hired workers across hierarchical job levels. IZA Discussion Paper No. 4858. Bonn, Germany: Institute for the Study of Labor. Google Scholar Hillman, Amy J., Shropshire, Christine, Cannella, Albert A. 2007. Organizational predictors of women on corporate boards. Academy of Management Journal 50: 941–52. Google Scholar, Crossref, ISI Hymowitz, Carol, Schellhardt, Timothy. 1986. The glass ceiling: Why women can’t seem to break the invisible barrier that blocks them from the top jobs. Wall Street Journal (March 24). Google Scholar Jencks, Christopher . 1992. Rethinking Social Policy: Race, Poverty, and the Underclass. Cambridge, MA: Harvard University Press. Google Scholar Kalev, Alexandra . 2009. Cracking the glass cages? Restructuring and ascriptive inequality at work. American Journal of Sociology 114(6): 1591–1643. Google Scholar, Crossref, ISI Kalev, Alexandra, Dobbin, Frank, Kelly, Erin. 2006. Best practice or best guesses? Diversity management and the remediation of inequality. American Sociological Review 71: 589–617. Google Scholar, SAGE Journals, ISI Kay, Fiona M., Hagan, John. 1995. The persistent glass ceiling: Gendered inequalities in the earnings of lawyers. British Journal of Sociology 46: 279–310. Google Scholar, Crossref, ISI Kerka, Sandra . 2001. Job searching in the 21st century. Myths and Realities No. 14. Center on Education and Training for Employment. Columbus: Ohio State University. Google Scholar Koput, Kenneth William, Gutek, Barbara A. 2010. Gender Stratification in the IT Industry: Sex, Status and Social Capital. Cheltenham, UK: Edward Elgar Publishing. Google Scholar, Crossref Lazear, Edward P., Oyer, Paul. 2004. The structure of wages and internal mobility. American Economic Review 94(2): 212–16. Google Scholar, Crossref, ISI Lyness, Karen S., Judiesch, Michael K. 1999. Are women more likely to be hired or promoted into management positions? Journal of Vocational Behavior 54: 158–73. Google Scholar, Crossref, ISI Madden, Janice Fanning . 1981. Why women work closer to home. Urban Studies 18: 18–94. Google Scholar, ISI Morgan, Laurie A. 1998. Glass-ceiling effect or cohort effect? A longitudinal study of the gender earnings gap for engineers, 1982 to 1989. American Sociological Review 63: 479–93. Google Scholar, Crossref, ISI Morrison, Ann M., White, Randall P., Van Velsor, Ellen Center for Creative Leadership . 1987. Breaking the Glass Ceiling: Can Women Reach the Top of America’s Largest Corporations? Reading, PA: Addison Wesley Longman Publishing. Google Scholar Mun, Eunmi . 2010. Sex typing of jobs in hiring: Evidence from Japan. Social Forces 88(5): 1999–2026. Google Scholar, Crossref, ISI Nelson, Robert L., Bridges, William P. 1999. Legalizing Gender Inequality: Courts, Markets and Unequal Pay for Women in America. Cambridge, UK: Cambridge University Press. Google Scholar, Crossref Noonan, Mary C., Corcoran, Mary E. 2004. The mommy track and partnership: Temporary delay or dead end? Annals of the American Academy of Political and Social Science 596: 130–50. Google Scholar, SAGE Journals, ISI Petersen, Trond, Saporta, Ishak. 2004. The opportunity structure for discrimination. American Journal of Sociology 109: 852–901. Google Scholar, Crossref, ISI Petersen, Trond, Saporta, Ishak, Seidel, Marc-David L. 2000. Offering a job: Meritocracy and social networks. American Journal of Sociology 106(3): 763–816. Google Scholar, Crossref, ISI Petersen, Trond, Togstad, Thea. 2006. Getting the offer: Sex discrimination in hiring. Research in Social Stratification and Mobility 24: 239–57. Google Scholar, Crossref Powell, Gary N., Butterfield, D. Anthony. 1994. Investigating the “glass ceiling” phenomenon: An empirical study of actual promotions to top management. Academy of Management Journal 37: 68–86. Google Scholar, Crossref, ISI Ragins, Belle Rose, Townsend, Bickley, Mattis, Mary. 1998. Gender gap in the executive suite: CEOs and female executives report on breaking the glass ceiling. Academy of Management Executive 12(1): 28–42. Google Scholar Reskin, Barbara F. 2000. The proximate causes of discrimination. Contemporary Sociology 29: 319–29. Google Scholar, Crossref, ISI Reskin, Barbara F., McBrier, Debra Branch. 2000. Why not ascription? Organizations’ employment of male and female managers. American Sociological Review 65(2): 210–33. Google Scholar, Crossref, ISI Ridgeway, Cecilia L. 2011. Framed by Gender: How Gender Inequality Persists in the Modern World. New York: Oxford University Press. Google Scholar, Crossref Roos, Patricia A., Manley, Joan E. 1996. Staffing personnel: Feminization and change in human resource management. Sociological Focus 29: 245–61. Google Scholar, Crossref Seltzer, Andrew, Merrett, David. 2000. Personnel policies at the Union Bank of Australia: Evidence from the 1888–1900 entry cohorts. Journal of Labor Economics 18: 573–613. Google Scholar, Crossref, ISI Smith, Nina, Smith, Valdemar, Verner, Mette. 2011. The gender pay gap in top corporate jobs in Denmark: Glass ceilings, sticky floors or both? International Journal of Manpower 32(2): 156–77. Google Scholar, Crossref, ISI Smith, Nina, Smith, Valdemar, Verner, Mette. 2013. Why are so few females promoted into CEO and vice president positions? Danish empirical evidence, 1997–2008. ILR Review 66: 381–408. Google Scholar, SAGE Journals, ISI Storvik, Aagoth Elise, Schone, Pål. 2008. In search of the glass ceiling: Gender and recruitment to management in Norway’s state bureaucracy. British Journal of Sociology 59: 729–53. Google Scholar, Crossref, Medline, ISI Tanner, Julia, Cockerill, Rhonda, Barnsley, Jan, Williams, A. Paul. 1999. Gender and income in pharmacy: Human capital and gender stratification theories revisited. British Journal of Sociology 50: 97–117. Google Scholar, Crossref, Medline, ISI Tomaskovic-Devey, Donald . 1993. Gender and Racial Inequality at Work. Ithaca, NY: ILR Press. Google Scholar Treble, John, van Gameren, Edwin, Bridges, Sarah, Barmby, Tim. 2001. The internal economics of the firm: Further evidence from personnel data. Labour Economics 8: 531–52. Google Scholar, Crossref, ISI Yakubovich, Valery, Lup, Daniela. 2006. Stages of the recruitment process and the referrer’s performance effect. Organization Science 17(6): 710–23. Google Scholar, Crossref, ISI Yamagata, Hisashi, Yeh, Kuang S., Stewman, Shelby, Dodge, Hiroko. 1997. Sex segregation and glass ceilings: A comparative statics model of women’s career opportunities in the federal government over a quarter century. American Journal of Sociology 103: 566–632. Google Scholar, Crossref, ISI Yap, Margaret, Konrad, Alison M. 2009. Gender and racial differentials in promotions: Is there a sticky floor, a mid-level bottleneck, or a glass ceiling? Relations Industrielles 64: 593–619. Google Scholar, Crossref, ISI Zeng, Zhen . 2011. The myth of the glass ceiling: Evidence From a stock-flow analysis of authority attainment. Social Science Research 40: 312–25. Google Scholar, Crossref, ISI