ASRXXX10.1177/0003122416630982American Sociological ReviewPedulla 6309822016

American Sociological Review 28–­1 Penalized or Protected? Gender © American Sociological Association 2016 DOI: 10.1177/0003122416630982 and the Consequences of http://asr.sagepub.com Nonstandard and Mismatched Employment Histories

David S. Pedullaa

Abstract Millions of workers are employed in positions that deviate from the full-time, standard employment relationship or work in jobs that are mismatched with their skills, education, or experience. Yet, little is known about how employers evaluate workers who have experienced these employment arrangements, limiting our knowledge about how part-time work, temporary agency employment, and skills underutilization affect workers’ labor market opportunities. Drawing on original field and survey experiment data, I examine three questions: (1) What are the consequences of having a nonstandard or mismatched employment history for workers’ labor market opportunities? (2) Are the effects of nonstandard or mismatched employment histories different for men and women? and (3) What are the mechanisms linking nonstandard or mismatched employment histories to labor market outcomes? The field experiment shows that skills underutilization is as scarring for workers as a year of unemployment, but that there are limited penalties for workers with histories of temporary agency employment. Additionally, although men are penalized for part-time employment histories, women face no penalty for part-time work. The survey experiment reveals that employers’ perceptions of workers’ competence and commitment mediate these effects. These findings shed light on the consequences of changing employment relations for the distribution of labor market opportunities in the “new economy.”

Keywords social stratification, gender, work and occupations, employment relations

Millions of workers are employed in positions are increasingly filling vacancies with job can- that deviate from the full-time, standard didates from the external labor market (Cap- employment relationship or work in jobs that pelli 2001; Hollister 2011), candidates about are mismatched with their skills, education, or whom they have less direct information than if experience (Bureau of Labor Statistics 2005, 2013; Kalleberg 2000, 2007; Smith 1997). Working in part-time positions, through tem- aUniversity of Texas at Austin porary help agencies, and at jobs below one’s skill level have become common experiences Corresponding Author: David Pedulla, Department of Sociology, for U.S. workers. At the same time, the conse- University of Texas at Austin, 305 E. 23rd Street, quences of these employment arrangements A1700, Austin, TX 78712 may be more significant than ever. Employers E-mail: [email protected] American Sociological Review 2 they were promoting workers from within their mismatched position—should improve a organizations (Leung 2014). These changes in worker’s future hiring outcomes over remain- employment, often considered key aspects of ing unemployed and, potentially, limit the the “new economy,” mean that workers’ expe- penalties a worker may face for not having a riences, including histories of nonstandard or full-time, standard employment history at his mismatched employment, are increasingly or her skill level. If this is the case, there may important during the hiring process. However, be an important role for nonstandard and mis- little is known about whether these types of matched employment in promoting workers’ employment histories shape employers’ evalu- opportunities. ations of candidates during the job applicant Alternatively, histories of nonstandard and screening process. Understanding the conse- mismatched employment may result in penal- quences of nonstandard and mismatched ties for workers similar to those stemming employment for workers’ hiring outcomes is from long-term unemployment. For instance, therefore necessary for a more comprehensive employers may screen out workers with non- account of labor market stratification in the standard or mismatched employment histories current economic landscape. in favor of workers with standard, full-time To address this gap in existing research, employment histories that match their skill this article examines three questions. First, level. This would support the notion that the what are the consequences of having a history labor market is becoming segmented into jobs of part-time work, temporary agency employ- that provide mobility opportunities and those ment, or skills underutilization for workers’ that are “dead ends” (Kalleberg, Reskin, and subsequent labor market opportunities? Sec- Hudson 2000). In this case, access to full- ond, given the gendered history of these time, standard employment at one’s level of employment arrangements—particularly part- skill and experience may serve as an impor- time and temporary agency employment—in tant axis of stratification in the labor market. the United States, do these consequences vary The consequences of different employment for male and female workers? Finally, what histories may also depend in important ways mechanisms account for the effects of non- on a worker’s gender. Unlike skills underutili- standard or mismatched employment histories zation, part-time work and temporary agency on workers’ labor market outcomes? employment arose in the U.S. economy as Workers experience nonstandard employ- heavily feminized types of work (Hatton 2011; ment arrangements (e.g., part-time work and Kalleberg 2000), potentially indicating that a temporary agency employment) as well as female worker is on the “mommy track” or mismatched employment positions (e.g., skills does not live up to the “ideal worker” standard underutilization) for a variety of reasons. (Acker 1990; Williams 2001). Thus, part-time Sometimes these types of positions are sought and temporary employment may serve as a as an alternative to unemployment. In other proxy for a worker’s parental status. In the cases, these positions are used to balance com- case of women, this may result in a “mother- peting demands outside of work. Little hood penalty” in hiring opportunities (Correll, research, however, examines the effect of Benard, and Paik 2007). On the other hand, these positions on workers’ future employ- employers might perceive nonstandard ment outcomes. On the one hand, this type of employment histories as a common experi- employment may protect workers from the ence for women. Thus, a nonstandard work negative effects of long-term unemployment, history may provide employers with limited signaling to prospective employers that work- information about a female worker’s quality, ers have maintained their skills and are moti- therefore reducing any negative influence of vated to work (Becker 1964; Marler, Barringer, nonstandard work histories on women’s hiring and Milkovich 2002; Ruhm 1991). This sug- outcomes. For men, by contrast, employers gests that any job—even a nonstandard or may take such a history as a signal that the Pedulla 3 male worker was unable to find a full-time, which workers are assumed to be employed standard job, raising concerns about worker full-time, in a permanent position, and in a quality. Thus, the protective or penalizing job commensurate with their skills and expe- nature of nonstandard employment is likely to rience (Kalleberg 2007, 2009). Given that depend on a worker’s gender, but the direction these conditions go unmet for a large propor- of the difference is uncertain. tion of workers in the “new economy,” it is Existing scholarship that addresses ques- important to understand the consequences of tions about the consequences of nonstandard these forms of employment on workers’ labor and mismatched employment relies largely on market outcomes. This article focuses on standard labor force survey data (see Addison, three employment situations—part-time Cotti, and Surfield 2009; Addison and Sur- work, temporary agency employment, and field 2009; Mavromaras, Sloane, and Wei skills underutilization—falling into two over- 2015). This has limited the ability of extant arching employment categories: nonstandard research to examine employers’ demand-side work and mismatched work. Nonstandard preferences, and has left open concerns about employment is generally characterized as biased estimates due to unobserved selection work that is not full-time, not expected to processes. This article analyzes original data continue indefinitely, or not performed at the from two experiments: (1) a field experiment legal employer’s place of business, nor under examining actual hiring decisions in five the legal employer’s direction (Kalleberg major U.S. labor markets; and (2) a survey 2000). Part-time work and employment experiment conducted with hiring decision- through a temporary help agency clearly meet makers at U.S.-based firms. These experi- these criteria. Mismatched employment ments allow me to focus directly on how occurs when individuals’ skills or preferences employers’ demand-side preferences shape do not fit the characteristics of their job (Kal- the consequences of nonstandard and mis- leberg 2007). Working in a job that is beneath matched employment histories, examine how a worker’s level of skill, education, or experi- those consequences vary by the gender of the ence—often referred to as skills underutiliza- worker, and alleviate concerns about unob- tion (McKee-Ryan and Harvey 2011) or served selection (see Pager 2007). The pri- overqualification (Vaisey 2006)—is a classic mary manipulations in both experiments were example of mismatched employment. These workers’ gender (signaled with male and two classifications may overlap; under certain female names) and the most recent work his- conditions, part-time work and temporary tories presented on applicants’ résumés. Each agency employment can also be mismatched application was randomly assigned 12 months employment. For example, when workers of recent employment experience, either a want a full-time, permanent job but are able full-time, standard job, a part-time job, to obtain only a part-time position or work employment through a temporary help agency, through a temporary help agency, they would a job below the applicant’s skill level, or a be in a mismatched employment relationship spell of unemployment. The results provide (Kalleberg 2007; McKee-Ryan and Harvey gender-specific causal estimates of the effects 2011). of nonstandard and mismatched employment Researchers have identified multiple fac- histories for workers’ future labor market tors implicated in the rise of nonstandard and opportunities. mismatched employment (Kalleberg 2000; Vaisey 2006). Global economic integration has increased competition for U.S. firms, cre- Nonstandard and ating incentives for companies to outsource Mismatched Employment work to lower-wage countries and implement Nonstandard and mismatched work deviate “flexible,” nonstandard employment relations from common conceptions of employment in for their U.S. employees (Kalleberg 2009). American Sociological Review 4

Legal changes have also paved the way for of nonfarm employment in the United States employers to alter the employment contract (Autor 2003), and then remained relatively and increase the use of nonstandard employ- stable at these higher levels (Bureau of Labor ment relations (Autor 2003; Gonos 1997). Statistics 2005). A majority of THA workers Additionally, changes in key labor market (nearly 60 percent) are in those positions invol- institutions, such as the decline in the power untarily (Bureau of Labor Statistics 2005). of organized labor (Clawson and Clawson Women historically dominated THA employ- 1999), have likely enabled the emergence of ment as the sector developed after World War nonstandard positions in the U.S. labor mar- II (Hatton 2011), but THA workers are now ket, placing workers in less desirable working roughly half male and half female (Bureau of arrangements, such as skills underutilization. Labor Statistics 2005). Technological changes that improved com- Finally, skills underutilization describes munication and information systems also workers employed in jobs for which they have likely played a role in the increase of non- excessive skills, education, or experience standard employment relations by, for exam- (Erdogan and Bauer 2011). There is less ple, enabling employers to more easily research on skills underutilization than part- coordinate their labor needs with temporary time or temporary work, in part due to the chal- help agencies (Kalleberg 2000; Schilling and lenges with operationalizing the construct using Steensma 2001). Moreover, the changing edu- survey data. However, Vaisey (2006) finds that cation landscape and shifts in the occupational skills underutilization (defined as excessive structure of the U.S. economy are likely education for one’s job) increased significantly implicated in the rise of skills underutiliza- between 1972 and the early 2000s. One key tion (Vaisey 2006). Regardless of the cause, difference between skills underutilization and part-time work, temporary employment, and part-time or temporary agency employment is skills underutilization have become common that women have not been historically overrep- experiences for U.S. workers. resented in this type of position (Vaisey 2006). Part-time employment is defined as work- ing less than 35 hours per week and is the most prevalent form of nonstandard work. Roughly Penalized or Protected? 20 percent of the U.S. workforce is employed Social scientists have examined the forces in part-time positions, and approximately a behind the rise of nonstandard and mis- quarter of part-time workers are in those posi- matched employment in the United States and tions involuntarily, preferring a full-time job the consequences of those employment expe- (Bureau of Labor Statistics 2013; Kalleberg riences for workers currently in those posi- 2000). Although the gender gap in part-time tions (Autor 2003; Epstein et al. 1999; work has declined over time, significant gen- Kalleberg et al. 2000). Limited scholarship, der differences remain. Currently, over 70 per- however, addresses how employers evaluate cent of part-time workers in the United States candidates with these types of employment between the ages of 25 and 54 are women histories during the job application process (Bureau of Labor Statistics 2015). and how those evaluations vary by gender. We Temporary help agency employment cap- therefore need additional research aimed at tures workers who are on the payroll of one understanding how experience with nonstan- firm (the “temp agency”) but who perform dard or mismatched employment may shape their tasks on a temporary basis at a separate male and female workers’ ability to get a job firm. Employment through temporary help in the future. At its core, this is an issue about agencies (THAs) has risen dramatically over the demand side of the job matching process, the past 30 years. The THA sector grew at an focusing on employers’ perceptions of male annual rate of 11 percent between 1979 and and female workers with nonstandard and 1995, more than five times the rate of growth mismatched employment histories and, Pedulla 5 ultimately, which applicants employers choose critical role that perceived competence plays to hire. Extending the voluminous literature in affecting the labor market opportunities of on how employers make hiring decisions (see unemployed workers, I argue that histories of Moss and Tilly 2001; Oyer and Schaefer 2011; nonstandard and mismatched employment are Rivera 2012), I build on insights offered by likely also read as signals of competence that theories of labor market signaling (Bills 2003; shape workers’ hiring outcomes. Spence 1973) as well as scholarship on unem- The literature on the signaling effects of ployment scarring (Gangl 2006; Ruhm 1991) unemployment generally limits its focus to and “ideal worker” norms (Acker 1990; Wil- signals of competence, but other types of sig- liams 2001). I draw on this work to conceptu- nals are possible. Sociological scholarship on alize how nonstandard and mismatched the cultural construction of the “ideal worker” employment influences hiring outcomes, the norm (Acker 1990; Williams 2001) highlights mechanisms that drive those effects, and how the penalties that accrue to workers who occupy those consequences may vary by gender. social positions that are perceived to be incom- patible with workplace performance expecta- tions (Turco 2010). This work points toward a Signaling and the Job Applicant second dimension that can be signaled by a Screening Process worker’s employment history: commitment. During the job applicant screening process, Perceived commitment—the expected effort employers are often faced with dozens, or and dedication that a worker is perceived to put even hundreds, of applications for a single forth—is central to the evaluation of workers vacancy. Because obtaining information about (Bielby and Bielby 2002; Correll and Benard the quality of a worker from a job application 2006; Turco 2010). And, perceived commit- can be difficult, employers likely utilize sig- ment may be of particular relevance in under- nals from application materials to make infer- standing how employers evaluate workers’ ences about the quality of potential employees employment histories (Leung 2014), including (Bills 2003; Ma and Weiss 1993; Spence histories of nonstandard and mismatched 1 1973). What signals may be sent by a history employment. I argue that, together, these two of nonstandard or mismatched employment? dimensions—competence and commitment— Two bodies of scholarship are particularly play a central role in affecting employers’ eval- useful in conceptualizing the content of the uations of workers with nonstandard and signals sent by a history of nonstandard or mismatched work histories during the job mismatched work. First, there is the literature applicant screening process. on the scarring effects of unemployment, which examines if and how histories of unem- ployment affect workers’ future earnings and The Potential Penalties of employment opportunities (Gangl 2006; Nonstandard and Mismatched Ruhm 1991). Building on theories of labor Employment Histories market signaling, this work argues that a spell The competence and commitment signals sent of unemployment may serve as a signal that by histories of nonstandard and mismatched the worker possesses some unobservable neg- employment may lead to penalties for workers, ative quality, such as lower levels of compe- compared to workers who have remained in tence or productivity (Eriksson and Rooth full-time jobs at the appropriate level of skill 2014; Kroft, Lange, and Notowidigdo 2013).2 and experience. Negative competence signals Existing research shows that competence is a are likely to be strongest if potential employers key dimension of social perception (Fiske perceive job applicants’ recent employment et al. 2002), with important consequences for history as resulting from involuntary job loss the evaluation of workers (Correll et al. 2007; followed by the inability to find full-time, stan- Cuddy, Fiske, and Glick 2004). Given the dard employment at their skill level.3 An American Sociological Review 6 inability to find a full-time, standard job that negatively shape workers’ labor market out- matches a worker’s skill and experience— comes. Using survey data, Ferber and Waldfo- when that is the type of position desired—may gel (1998) find negative associations between be read by employers as signaling a deficiency histories of part-time and temporary employ- in ability, skill, or competence (Karren and ment and workers’ future earnings, compared Sherman 2012). In this way, the involuntary to workers who remained in full-time employ- nature of most skills underutilization arrange- ment. Additionally, Mavromaras and col- ments likely sends strong negative signals leagues (2015) analyze panel data from about a worker’s competence. Similarly, invol- Australia and find that workers with histories untary part-time work and temporary employ- of skills underutilization are more likely to be ment may suggest to an employer that a job unemployed in the future, suggesting a “scar- applicant is not competent or productive ring” effect for histories of skills underutiliza- enough to find and keep a full-time job. In fact, tion. However, the relative importance of when part-time work and temporary employ- supply- and demand-side forces in these stud- ment are involuntary, they represent a form of ies is unknown, and researchers have been mismatched employment—similar to skills unable to tease apart the mechanisms underly- underutilization—because there is a lack of fit ing these results. between workers’ desires and their jobs (Kal- leberg 2007). Insofar as employers perceive a The Potential Protective Force of worker’s part-time or temporary agency Nonstandard and Mismatched employment history as involuntary, those employment histories will likely send negative Employment signals about an applicant’s competence. It is also possible that nonstandard and mis- A history of nonstandard or mismatched matched employment positions might provide employment may also send negative signals workers with an advantage over long-term about a worker’s commitment. Even though unemployment. Insofar as employers priori- survey data suggest there are few systematic tize an applicant’s competence during the hir- differences between the reported commitment ing process, workers who have any job—even levels of full-time, standard workers and other if it is not a full-time, standard position— types of workers, particularly part-time work- should fare better than unemployed workers ers (Kalleberg 1995), employers may perceive who, by definition, have not been working workers with nonstandard and mismatched (Becker 1964; Korpi and Levin 2001). Com- employment histories as less committed. pared to workers with histories of long-term These negative perceptions will likely be par- unemployment, employers may perceive ticularly strong if a future employer perceives applicants with histories of nonstandard and a worker’s move into nonstandard or mis- mismatched employment as keeping their matched employment as voluntary, resulting, skills more updated (Nollen 1996; see also Yu for example, from the worker attempting to 2012), improving the competence signal they balance paid employment and family respon- send to future employers. This benefit over sibilities (Epstein et al. 1999; Williams 2001). long-term unemployment should be stronger Because decisions about “voluntarily” moving for part-time or temporary workers, compared away from full-time, standard employment to workers in positions of skills underutiliza- are heavily gendered, the consequences of tion, particularly if they remain employed in these decisions may be different for men and their desired occupation. Nonstandard or mis- women. This issue will be discussed in detail matched employment may also send positive below. commitment signals to future employers, Some empirical evidence supports the compared to workers who remain unem- aforementioned notion that histories of non- ployed. Taking a part-time position, tempo- standard and mismatched employment may rary job, or job below one’s skill level may Pedulla 7 signal to future employers that a worker is form (Williams 2001). For many women in willing to do “whatever it takes” to remain the United States, these competing demands employed, even if that means taking an unde- intersect with unsupportive workplace prac- sirable job (i.e., nonstandard or mismatched). tices and social policies that make it more Empirically, there is some suggestive evi- challenging for them to live up to the “ideal dence that nonstandard employment may worker” norm (Gornick and Meyers 2003; buffer workers against the negative effects of Kelly et al. 2010). Even if women are able to unemployment. Addison and Surfield (2009) balance work and family demands, employers find, using labor force survey data, that jobless are likely more concerned about this set of individuals who obtain nonstandard employ- issues for women than they are for men. ment of multiple kinds, including temporary In contrast to the masculine construction agency employment, are more likely to be of the “ideal worker,” part-time and tempo- employed one year later, compared to the job- rary employment (although not skills under­ less who continue to search for work. Workers utilization) arose in the United States as who move into these nonstandard positions highly feminized positions in the labor mar- also have employment continuity that is simi- ket (Hatton­ 2011; Williams 2001). Part-time lar to full-time, standard employees. It is jobs have historically been viewed as part unclear, however, whether these findings of the “mommy track” (Williams 2001)—an derive from employer preferences, job seeker employment option for women attempting to behavior, or some combination thereof. Addi- balance the “competing devotions” of work tionally, the causal nature of the protective and family life (Blair-Loy 2003; Feldman force of nonstandard work remains uncertain, 1990). Similarly, temporary agency employ- given the challenges of dealing with various ment developed as a form of women’s work selection processes when utilizing standard (Vosko 2000). Hatton (2011) argues that dur- survey data. Ultimately, while there are rea- ing the emergence of the THA industry after sons to believe that having a job—any job— World War II, industry leaders intentionally will provide workers with an advantage over defined temporary jobs as “women’s jobs” to joblessness during the job application process, avoid confrontations with organized labor. research to date has not been able to provide Today, however, although female workers strong causal evidence to support or reject this make up nearly three-quarters of all part-time claim. workers between the ages of 25 and 54, there is approximate gender parity in the THA sec- tor (Bureau of Labor Statistics 2005, 2014). The Gendered How might the masculine construction of Consequences of the “ideal worker” intersect with the gendered Employment Histories histories of part-time and temporary work dur- The notion of the “ideal worker” norm has ing the job application process? On the one been used to understand persistent gender hand, employers may make gendered assump- inequality in the labor market (Acker 1990; tions about the selection processes that lead Williams 2001). The argument is that employ- workers into nonstandard employment posi- ers expect their workers, especially white- tions. For female applicants, histories of part- collar workers, to limit the aspects of their time or temporary employment may lead personal lives that may conflict with their paid employers to perceive them as being on the work responsibilities, to work long hours, and “mommy track,” having significant demands to be willing to travel frequently (Kelly et al. outside the workplace that will conflict with 2010). Given that demands outside the work- their ability to perform on the job. Indeed, place—such as childcare and household experimental research finds that women are work—fall disproportionately on women, the assumed to be in part-time positions to deal “ideal worker” generally takes a masculine with domestic and family duties (Eagly and American Sociological Review 8

Steffen 1986). For women, therefore, non- are stronger for men (Ferber and Waldfogel standard employment may serve as a proxy 1998). In Canada, there is some evidence that for motherhood, raising concerns about their women are more likely than men to exit tem- competence and commitment as well as their porary jobs for full-time employment, sug- compliance with the “ideal worker” norm. In gesting that temporary work is less scarring turn, women may face greater penalties than for female workers (Fuller 2011). Although men for nonstandard employment histories, not exactly the same as having a nonstandard similar to the “motherhood penalty” faced at employment history, some experimental the hiring interface (Correll et al. 2007). research finds that men are penalized more On the other hand, consistent with some heavily than women for taking a leave of models of labor market signaling (see Spence absence or needing to leave work for family 1973), signals will likely be most consequen- reasons (Allen and Russell 1999; Butler and tial when they are either costly or scarce. Skattebo 2004). Given the disproportionate concentration of Together, existing theoretical perspectives women in part-time positions (Kalleberg indicate that the effects of part-time and tem- 2000) and, historically, temporary agency porary agency employment histories will vary employment (Hatton 2011), nonstandard by the gender of the worker. Yet, whether men employment histories for women are not or women will face more severe penalties scarce and therefore may have limited signal- remains unclear. Given that there are limited ing power. Given the perceived commonality gender differences in experiencing skills of nonstandard employment histories among underutilization in the United States, and this women, employers may obtain limited addi- type of employment does not have a femin- tional information about female job applicants ized history, there is little reason to think that with such histories, reducing any penalties employers will treat men and women differ- women may face as a result. For men, how- ently based on this factor. ever, part-time or temporary work histories are more rare and, therefore, may serve as stronger signals to employers. Such an Methodological employment history may trigger employers’ Considerations concerns about whether there is something The aforementioned studies about the conse- deficient about a male worker—potentially his quences of nonstandard and mismatched competence or commitment—making him employment rely on observational data, leav- unable to find a full-time, standard job (Eagly ing open the possibility that workers’ selec- and Steffen 1986). Additionally, men with tion into nonstandard or mismatched part-time and temporary employment histories employment, employers’ demand-side prefer- may be seen as violating standard breadwin- ences, or some unobservable worker or ning models of masculinity (see Cha 2010). employer characteristics drive these associa- Significant research documents the ways that tions. To my knowledge, only one U.S.-based violating gender stereotypes can result in study has attempted to deal with these social and economic sanctions (Rudman and ­endogeneity concerns by using a quasi-­ Phelan 2008). This suggests that men will be experimental research design. Autor and more heavily penalized than women for histo- Houseman (2010) address the problem of ries of part-time and temporary agency selection bias by exploiting the random employment. assignment of people in Detroit’s welfare-to- There is some preliminary empirical sup- work program to different types of job place- port for this line of thinking. For example, in ments (i.e., a temporary help agency the United States, histories of part-time work placement versus no job placement). They are associated with lower future earnings for find quite different results depending on both men and women, but the negative effects whether they correct for unobserved selection Pedulla 9 processes. Specifically, after correcting for nonstandard and mismatched employment selection, they find that temporary agency histories. This approach allows me to ascer- employment was actually no better than going tain the ways various employment histories without a job placement. Yet, temporary intersect with gender to shape workers’ expe- agency employment appears significantly riences at the hiring interface while removing beneficial when no corrections for selection concerns about unobserved selection pro- processes are made. While the generalizabil- cesses and omitted variables bias. ity of the Autor and Housemen (2010) study is unknown, their findings clearly suggest that selection bias makes identifying the causal The Field Experiment effects of temporary employment difficult What are the consequences of nonstandard using observational data. Given this chal- and mismatched employment histories for lenge, experimental research designs that workers as they apply for jobs? And, how do alleviate concerns about selection bias and these effects differ by the gender of the bias due to omitted variables—and that spe- worker? To examine these questions, I ana- cifically enable a direct investigation of lyze original data from a field experiment. employers’ demand-side preferences—are This involved submitting 2,420 applications vital to furthering our understanding of the to 1,210 job openings between November consequences of nonstandard and mismatched 2012 and June 2013.4 After sending each employment histories. application, I tracked the callbacks (i.e., posi- To address the methodological issues in tive employer responses) received by each existing research and identify employers’ application. The overall callback rate for the responses to workers with different employ- field experiment was 7.5 percent, which is ment histories, I implemented complemen- consistent with previous studies that use simi- tary field and survey experiments examining lar methods (Bertrand and Mullainathan the effects of nonstandard and mismatched 2004; Correll et al. 2007). employment for male and female workers’ The field experiment had two axes of vari- hiring outcomes. In the analysis, I first utilize ation. First, the experiment varied the most data from the field experiment, where ficti- recent employment experience on the appli- tious job applications were sent to apply for cant’s résumé. Each résumé was randomly real job openings, to examine how nonstand- assigned 12 months5 of recent work experi- ard and mismatched employment histories ence: either a full-time, standard job, a part- intersect with gender to affect hiring out- time job, a job through a temporary comes in the actual labor market. The field employment agency, a job below an individu- experiment, however, provides information al’s skill level, or a spell of unemployment. only about whether an employer responds The second axis of variation in the experiment positively to a job application. It does not was the applicant’s gender, which was sig- provide any details about the mechanisms naled using gendered names.6 The male names linking histories of nonstandard or mis- were Jon Murphy and Matthew Stevens; the matched employment to future employment female names were Katherine Murphy and outcomes. To identify these intervening pro- Emily Stevens. A résumé and a cover letter cesses, I analyze data from the survey experi- were included with each job application. Each ment, which used the same experimental cover letter was crafted with similar language, manipulations as the field experiment and while also accurately reflecting the work his- collected information on hiring decision- tory presented on the résumé. The cover letter makers’ perceptions of job applicants’ compe- for each experimental condition remained tence and commitment. The survey experiment consistent across employers, except that each thus enables an analysis of the mechanisms letter was personalized with the employer’s that may account for the consequences of name and the job title for the open position. American Sociological Review 10

Because two résumés were submitted for each skill and experience. Finally, a spell of unem- job opening, I constructed two résumé tem- ployment was presented on workers’ résumés plates that were similar in content but aestheti- by indicating that their most recent job ended cally distinct.7 Each résumé indicated that the one year before the application date.11 To applicant graduated from one of two large, ensure résumés in the unemployment condi- public universities in the Midwest with similar tion had the same number of employment rankings by U.S. News and World Report. experiences as résumés in the other condi- Thus, an important scope condition of the tions, a summer internship in college was findings is that they are limited to workers added to résumés for the “unemployed” work- with a college education.8 After graduating ers. Part A of the Appendix presents examples from college, each résumé indicated that the of the experimental treatments used in the applicant had a first job that lasted for just field experiment. under two years. Each applicant then had Applications were submitted to four differ- a second job that lasted for nearly four and a ent job types that varied in the level and type half years. All applicants then transitioned to a of skill they required as well as their gender new job, which is where the experimental composition: sales, accounting/bookkeep-­ manipulations were implemented. The stand- ing, project management/management, and ard, full-time résumés were pretested before administrative/clerical job types. The résumés using them in the experiment, and they submitted for each job type had an employ- received similar ratings on key dimensions of ment history with relevant experience for that perceived skill and experience.9 occupation. The applications were submitted Histories of nonstandard and mismatched to job openings in five major U.S. labor mar- employment were carefully signaled on work- kets—New York City, Atlanta, Chicago, Los ers’ résumés.10 Part-time work was presented Angeles, and Boston—to add geographic on a worker’s résumé by including “part-time”­ diversity to the analysis. Employment histo- in parentheses after the occupational title for ries for each applicant were geographically the most recent job on the full-time, standard specific to the labor market in which the résumé. This method of signaling part-time applicant was applying. Each résumé also work experience is consistent with how work- included a local phone number and a local ers often present this information in online address. Each phone number had its own résumé banks. Temporary agency employ­ voice mailbox and a unique gender-specific ment was presented on the applicant’s résumé voice recording where employers could leave as working through one of two leading tempo- messages for the applicant. The applicants’ rary help agencies and in the worker’s chosen street addresses were located a few blocks occupation (e.g., accounting, management). away from each other in each city. The The description of the tasks and responsibili- addresses were real, but the apartment num- ties completed as a temporary worker were bers were fictitious.12 very similar in content to those presented on The sample of job openings for the experi- the full-time, standard résumé. While the part- ment was drawn from a leading national online time and temporary agency employment job-posting website and therefore represents a ­histories were in the worker’s occupation of broad cross-section of job openings. Using a choice, skills underutilization—for all ­workers national job-posting website ensured some —was denoted as working as a sales associate level of consistency in the jobs being posted at a large retail store, working with customers across labor markets. To collect the job open- in the retail space. The year of employment in ings that met the search criteria for the experi- this position followed approximately six years ment, I worked with a computer programmer of work experience in professional, skilled to design a script that executed the needed jobs, and thus clearly indicates that these searches. Each search was for a particular job workers were employed at a level below their type (e.g., administrative assistant), within a Pedulla 11

20-mile radius of each city, posted over the standard errors clustered at the level of the job previous 30 days, and that could be applied for opening (results presented in Part C of the directly through the job-posting website.13 Appendix).14 I begin by examining the conse- After collecting the job openings that matched quences of nonstandard and mismatched these requirements, duplicate postings from employment histories for men and then turn the same employer were removed to reduce the to the consequences for women. I then com- likelihood that employers would perceive the pare the callback rates for men and women résumés as fictitious. within each employment history category. After the final set of job openings was Male job applicants received a 10.4 percent selected for a given job type in a given city, I callback rate in the full-time condition. In the randomly assigned each job opening to a other conditions, male job applicants received a demographic category (male or female) and statistically significant lower callback rate to receive applications with two different (part-time: 10.4 percent versus 4.8 percent, employment histories. However, the rand- |z| = 2.18, p < .05; skills underutilization: 10.4 omization ensured that each employer percent versus 4.7 percent, |z| = 2.07, p < .05; received at least one application with either unemployment: 10.4 percent versus 4.2 per- the full-time or unemployment treatments. cent, |z| = 3.11, p < .01). The one exception, Two applications were sent to each employer, however, was temporary agency employment separated by one day. The names at the top of (10.4 percent versus 7.1 percent, |z| = 1.21, the résumés, the formats of the résumés, and p = .23), where no statistically significant effect the order of the résumés were randomized was detected. The results also indicate that, for and counterbalanced to ensure that these men, none of the nonstandard or mismatched aspects of the job application would not be employment history categories received statis- correlated with the treatment. Part B of the tically significantly higher callback rates than Appendix presents the distribution of the the unemployment condition. Together, these characteristics of the applications submitted results indicate that part-time work and skills in the field experiment. underutilization are as scarring for male work- The primary outcome variable for the field ers as a year of unemployment. However, for experiment was whether the applicant men, temporary agency employment histories received a positive response or callback from are statistically indistinguishable from histories the employer via phone or e-mail. Responses of full-time, standard employment at a work- were coded as callbacks if the employer er’s skill level and histories of unemployment. requested an interview with the applicant, or The callback rate for women in the full- if the employer asked the applicant to contact time, standard employment condition was on them to discuss the position in more depth. par with the callback rate for men (10.4 per- Auto-generated responses and simple requests cent), but the consequences of the various for more information were not coded as posi- work histories appear quite different for tive responses. female job applicants. Skills underutilization is the only employment category where female job applicants received a callback rate Field Experiment Results that was statistically significantly lower than Figure 1 presents the main field experiment the full-time condition (10.4 percent versus results as a bar graph with the callback rate 5.2 percent, |z| = 2.05, p < .05). In terms of for each employment history category, broken temporary agency employment and unem- down by the worker’s gender. For statistical ployment, the callback rates are slightly lower tests, I use z-tests for differences in propor- than in the full-time, standard employment tions and present results for two-tailed tests condition, but these differences are not statis- throughout. The results are nearly identical tically significantly different from having a when I use logistic regression models with full-time work history. I found no reduction at American Sociological Review 12

Figure 1. Callback Rates, by Employment History and Gender Source: Original experimental audit study data. Note: All statistical tests are z-tests for differences in proportions. Statistical significance comparing given employment history to full-time employment history: *p < .05; **p < .01 (two-tailed tests). Statistical significance comparing male and female workers in the same employment history category: +p < .05 (two-tailed tests). all in the callback rate for women with histo- (|z| = .42, p = .68). Male and female applicants ries of part-time work (10.4 versus 10.9 per- with a skills underutilization history also cent). Women with histories of part-time received callbacks from employers at similar employment—the most heavily gendered rates (4.7 percent for men and 5.2 percent for labor market position under investigation— women, |z| = .23, p = .82). However, a margin- do not face any penalties compared to women ally significant gender difference emerges for with histories of full-time employment. Over- histories of unemployment. For applicants with all, for female job applicants, there seem to be histories of unemployment, men received posi- limited negative consequences of part-time tive responses 4.2 percent of the time, com- work, temporary agency employment, and pared with 7.5 percent for women (|z| = 1.89, unemployment, but strong penalties for histo- p = .059). ries of skills underutilization. What might explain the different callback Finally, I compare the callback rates for rates for male and female applicants with part- male and female job applicants in each employ- time work histories? One possibility is that the ment history category. In the full-time, standard differential effect of part-time work for men condition, male and female job applicants and women may vary by the gender composi- received the same callback rate from employ- tion of the different occupations in the field ers, 10.4 percent for men and 10.4 percent for experiment. Indeed, the gender composition women. The next cluster of columns in Figure­ 1 of these occupations ranges from 38 percent examines positive responses for résumés with women in management occupations to 73 per- part-time employment histories. Here, we see a cent women in administrative and clerical statistically significant gender difference. Men occupations.16 In more heavily feminized with a part-time history received positive occupations, such as administrative and cleri- responses 4.8 percent of the time, compared cal jobs, part-time work may be more preva- with a 10.9 percent positive response rate for lent for women than in male-dominated women with part-time histories (|z| = 2.14, occupations, thus limiting the strength of the p < .05).15 Men and women with temporary signal sent by a part-time history for female agency employment histories had similar call- applicants in these occupations. In this case, back rates of 7.1 and 8.3 percent, respectively part-time employment histories may have Pedulla 13

Figure 2. Callback Rates for Part-Time Employment Histories, by Occupation and Gender Source: Original experimental audit study data. little effect on women’s hiring outcomes in or female applicants compared to having a female-dominated occupations, but a strong history of full-time, standard employment. effect in male-dominated occupations. It is The field experiment findings, however, are also possible, however, that stereotypes about not able to provide insights into the demand- gender and employment histories transcend a side mechanisms underlying employers’ worker’s occupation, resulting in similar gen- responses to job applicants with different dered patterns across occupations. To examine employment histories. For that task, I turn to this possibility, I compared the callback rates results from the survey experiment. for men and women with a history of part- time employment, broken down by occupa- tional category. Figure 2 presents these results. The Survey Experiment The descriptive evidence indicates that, in The field experiment results provide compel- each occupation, female applicants with part- ling evidence about the effects of nonstandard time employment histories receive a higher and mismatched employment in the actual callback rate than male applicants.17 Thus, the labor market and how they differ by gender, gendered nature of particular occupations but those data cannot examine the reasons does not seem to be driving the gender-­ why nonstandard and mismatched employ- differentiated consequences of part-time ment histories shape employers’ evaluations employment histories. of job applicants. Thus, to complement the The field experiment demonstrates impor- field experiment, I conducted an Internet- tant heterogeneity in the consequences of based survey experiment with individuals in different types of employment histories for U.S. firms who make hiring decisions for men and women. Male applicants with histo- their companies. The survey experiment was ries of part-time employment, skills underuti- conducted between December 6, 2012, and lization, and unemployment are heavily January 4, 2013. Most hiring studies that use penalized, whereas female applicants are experimental methods are conducted on negatively affected only if they have histories undergraduate or graduate students (e.g., Cor- of skills underutilization. The results also rell et al. 2007). The survey experiment pre- indicate that female job applicants with histo- sented here therefore advances research ries of part-time employment and, to some methodology in this area by surveying indi- extent, unemployment fare better than men viduals who make actual hiring decisions and with a similar background. Temporary agency who work in five broad occupational ­groups: employment does not appear to penalize male human resources managers, human resources American Sociological Review 14 assistants/associates, business executives, other axis. Each respondent was randomly mid-level managers, and business owners. assigned to review either two résumés belong- The respondents in the sample are part of an ing to men or two résumés belonging to online, opt-in panel and thus are not a random women, but having different employment his- probability sample. Any potential limits on tories (at least one of which was either a full- generalizability, however, do not affect the time or an unemployed résumé). Thus, the ability to generate internally valid, causal gender manipulation was between subjects, estimates of the effects of interest from the which reduces concerns about social desira- survey-experimental research design. bility bias by making it less likely a respond- Table 1 presents descriptive statistics about ent would identify gender as a key issue of the 903 respondents in the survey experiment. interest, and the employment history manipu- Roughly half (52.9 percent) of the respondents lation was within subjects. The format and are men, 71.8 percent are white, the vast order of the résumés, as well as the names at majority have at least a college degree, and 26 the top of the résumés, were randomized and percent work in firms with 500 or more counter-balanced. employees. To provide a sense of how respondents in the survey experiment compare to national estimates of similar individuals, Variable Construction Table 1 also includes a “National Estimates” After reviewing each résumé, respondents column. The national estimates, however, are were asked to evaluate the applicant. To par- not exactly comparable to the survey sample’s allel the outcome variable in the field experi- characteristics and thus are provided mainly ment, respondents were asked on a five-point as a guide for the reader. Additional details scale: “How likely would you be to recom- about characteristics of the sample of respond- mend that your company interview this appli- ents, how the national estimates were gener- cant?” Responses to this item were then ated and differ from the survey experiment converted into a dichotomized variable with sample, and the sample selection process are the “very likely” category equal to 1 and the discussed in the online supplement (http://asr other categories equal to 0; 27.3 percent of .sagepub.com/supplemental). job applicants were “very likely” to be rec- Once respondents were qualified to par- ommended for interviews. Coding the inter- ticipate in the survey, they were asked to view likelihood variable in this way makes review and evaluate two experimentally sense theoretically, because only applicants manipulated résumés for an open accounting who attained the “very likely” category on the clerk position at their company. The account- “interview likelihood” measure in the survey ing position was selected because nonstand- context would have been likely to receive ard work is common in the accounting callbacks in the field experiment context. profession, and most companies have some- Thus, the dichotomous measure in the survey body who performs an accounting or book- experiment most closely parallels the out- keeping role. It also parallels the accounting/ come measure in the field experiment.18 bookkeeping category of jobs applied to in Respondents were also asked to evaluate the field experiment. The two axes of varia- the applicant along a series of dimensions. tion on the résumés in the survey experiment This was designed to isolate the mechanisms were the same as in the field experiment. The leading to the consequences of nonstandard most recent employment history of the appli- and mismatched employment: perceived cant (full-time, part-time, temporary agency, competence and perceived commitment. skills underutilization, or unemployment) I combined the following items to generate was varied along one axis and the gender of a scale of perceived competence. Respondents the applicant, using the same names as in the were asked: “On a scale from one to seven, field experiment, was manipulated along the how strongly do you agree or disagree with the Pedulla 15

Table 1. Survey Experiment Respondent and Firm Characteristics

Sample National Percent/Median Estimates

Respondent Characteristics Male 52.9% 50.7% Race/Ethnicity White 71.8% 83.8% Black 10.3% 8.6% Hispanic 6.6% 10.1% Asian 6.3% 4.6% Other/Multiple Race 5.0% Education High School or Less 7.6% 16.3% Some College 17.6% 32.6% College 53.6% 35.3% Graduate School 21.2% 15.8% Income (Median) $67,500 $88,007 Age (Median) 40.5 44.3 Job Tenure in Years (Median) 5 Firm Characteristics Number of Employees Fewer than 10 17.4% 78.6% Between 10 and 99 37.8% 19.6% Between 100 and 499 18.8% 1.5% 500 or more 26.0% .3% Industry Agriculture, Mining, Construction 5.4% 13.6% Education and Health 16.6% 11.8% Financial and Information 16.3% 10.3% Leisure and Hospitality 7.5% 10.0% Manufacturing 14.4% 4.7% Professional and Business Services 16.8% 24.7% Public Administration 4.9% 5.6% Retail 9.3% 11.7% Transportation, Utilities, Wholesale 8.3% 8.5% Other .7% .2% Sample Size 903

Note: While 903 respondents qualified for the survey, the sample sizes for the respondent and firm characteristics range from 893 to 899 in Table 1 due to item nonresponse. National estimates of respondent characteristics are derived from various sources at the Bureau of Labor Statistics and the Census Bureau. National job tenure estimates, by detailed occupation, were not available and therefore are not included. Additional details are discussed in the online supplement. National firm characteristic estimates are from the U.S. Census Bureau (2008) Statistics of U.S. Businesses. Collapsing some categories from the census data was necessary to align the data with the firm size and industry categories in the survey. following statements about this applicant?” applicant is skilled,” “the applicant has rele- Responses ranged from “strongly disagree” to vant work experience,” and “the applicant has “strongly agree.” The statements used to create adequate accounting experience.” Addition- the competence measure were “the applicant is ally, the competence scale included the follow- competent,” “the applicant is productive,” “the ing items with five response categories: (1) American Sociological Review 16

“Compared to similar employees who already employers’ responses about whether they work at your company, how quickly do you would be “very likely” to recommend that think this applicant would learn how to per- their company interview the applicant. In form new tasks?” and (2) “Compared to simi- essence, this analysis seeks to determine lar employees who already work at your whether the main findings from the field company, how much relevant experience in experiment replicate in the survey experiment accounting and bookkeeping does this appli- context. Because the interview likelihood cant have?” These items combined with an variable is binary, I used logistic regression alpha of .89, and the standardized scale is used models in the analyses. I first examine the throughout. consequences of nonstandard and mismatched To generate a scale of perceived commit- employment histories separately for male and ment, I combined four survey items. Using a female job applicants. Then, I test whether seven-point scale, respondents were asked to gender differences within each employment respond to the statement “the applicant is com- history category are statistically significant. mitted.” Then, on a five-point scale, respond- Model 1 in Table 2 examines the conse- ents were asked, “Compared to similar quences of nonstandard and mismatched employees who already work at your com- employment histories for male applicants. pany, how committed do you think this appli- The results demonstrate that men with histo- cant would be to their job if they were hired?” ries of part-time employment, skills underuti- Also on a five-point scale, respondents were lization, and unemployment are heavily asked, “If your company needed to ask this penalized in terms of their interview likeli- applicant to work extra hours, how likely is it hood. For example, male applicants with his- that this applicant would meet that request?” tories of part-time work have approximately Finally, respondents were asked, “If this appli- 40 percent lower odds (exp[–.513 = .599]) of cant were to be hired at your company, how being “very likely” to be recommended for an long do you think that they would stay?” The interview, compared to male workers with five response categories ranged from “less full-time, standard employment histories. than 1 year” to “more than 4 years.” These four There is also a marginally significant nega- items combined with an alpha of .72, and the tive effect of temporary agency employment scale used in the analyses is standardized. for men ( p < .10). Additional tests indicate The key predictor variables for the analysis that none of the nonstandard or mismatched are the different employment histories on the employment history categories are statisti- résumé that the respondent reviewed—full- cally significantly different from unemploy- time, part-time, temporary agency, skills under- ment. Thus, the consequences of nonstandard utilization, or unemployment—and the gender and mismatched employment for male appli- of the applicant. All models include controls for cants in the survey experiment are very simi- the order the résumés were reviewed, the name lar to those found in the field experiment. at the top of the résumé, and the format of the Next, Model 2 in Table 2 examines the con- résumé.19 Listwise deletion is used to deal with sequences of nonstandard and mismatched missing data and only respondents who pro- employment histories for female applicants. I vided interview recommendations for both find that the only employment history category applicants that they reviewed are kept in the where female applicants are penalized is the analytic sample.20 All analyses adjust for the skills underutilization category. The scarring fact that respondents evaluated two résumés by consequence of unemployment for women is clustering the standard errors by respondent. marginally statistically significant ( p < .10). There are no discernable differences in the interview recommendations for women with Interview Likelihood histories of part-time or temporary employment The first analyses examine whether nonstan- versus a history of full-time employment. There dard and mismatched work histories affect are also no statistically significant differences Pedulla 17

Table 2. Logistic Regression Models of the Consequences of Nonstandard and Mismatched Employment Histories on Being “Very Likely” to Be Recommended for an Interview

Male Applicants Female Applicants (1) (2)

Employment History Full-Time (omitted) Part-Time –.513* –.255 (.236) (.220) Temporary Agency –.445 –.127 (.235) (.222) Skills Underutilization –.747** –.610** (.270) (.233) Unemployment –.447* –.320 (.195) (.182) Constant –.734*** –.513** (.178) (.163) n (clusters) 439 452 n (observations) 878 904

Note: Clustered standard errors are in parentheses. Log-odds presented. All models include controls for the order in which résumés were presented, the format of the résumé, and the name on the résumé. *p < .05; **p < .01; ***p < .001 (two-tailed tests).

between having a history of nonstandard or applicant’s gender in their evaluations (see mismatched employment and having a history Heerwig and McCabe 2009). There is also of unemployment. Again, the findings for evidence that social categories, such as race female applicants in the survey experiment and gender, are more likely to be used as align closely with results presented in the field heuristic devices when time is scarce (Fiske experiment. 1998). In the field experiment, hiring man- The results presented in Table 2, however, agers are likely screening hundreds of appli- do not test for differences in the interview cants in a short time period, making the recommendation likelihood for male and activation of gender stereotypes more likely female applicants within each employment than in the survey context, where respond- category. This is where the results diverge ents had as much time as they wanted to between the field and survey experiments. review two résumés. Finally, the survey The field experiment found that female experiment asked respondents to review applicants received higher callback rates résumés for one job type—an accounting than male applicants in the part-time work clerk position—whereas the field experi- and unemployment history categories. How- ment examined four occupational groups, ever, this is not the case in the survey experi- which included but were not limited to ment. There are no statistically significant accounting and bookkeeping positions. Dif- differences in being “very likely” to be rec- ferences in the job types under investigation ommended for an interview between male might contribute to the different gender find- and female applicants within employment ings in the field and survey experiments.21 history categories in the survey experiment. Why might this discrepancy exist? It is dif- ficult to address this issue empirically, but Mediating Effects of Perceived the difference may be related to social desir- Competence and Commitment ability biases that arise in the survey context, Finally, I explore whether employers’ percep- which may limit respondents’ use of an tions of job applicants’ competence and American Sociological Review 18 commitment can account for the reduced means the 95 percent confidence interval interview likelihood faced by applicants with around the ACME does not include zero. For certain histories of nonstandard and mis- clarity, Table 3 only presents the mediation matched employment. Because the gender results for the types of nonstandard and mis- differences within each employment history matched employment histories that penalized category are not statistically significant in the workers in a statistically significant manner survey experiment, it is not possible to test for in the analysis presented in Table 2. the mechanisms underlying gender differ- As Panel A in Table 3 shows, perceived ences in the consequences of nonstandard commitment mediates 40.4 percent of the employment. negative effect of part-time employment his- To examine whether hiring decision-­ tories for men. Perceived competence, in con- makers’ perceptions of applicants’ competence trast, does not mediate the effects of part-time and commitment can account for the conse- work for male applicants. Skills underutiliza- quences of nonstandard and mismatched tion, in turn, is scarring for male applicants employment histories, I utilize the average due to both perceived competence and per- causal mediation analysis framework proposed ceived commitment. For female applicants, by Imai, Keele, and Tingley (2010)—rather skills underutilization appears penalizing due than linear structural equation modeling.22 to perceived competence—which explains Imai and colleagues (2010) argue that linear 94.1 percent of the effect—but not perceived structural equation modeling approaches to commitment. Together, these empirical results mediation are limited because they do not offer provide compelling evidence that perceived a general definition of causal mediation beyond competence and perceived commitment are a specific statistical model and they do not important mechanisms linking histories of generalize to nonlinear models. To overcome nonstandard and mismatched employment these limitations, the method of Imai and col- with workers’ hiring outcomes. However, leagues (2010) derives a formal definition of these findings also reveal the heterogeneous causal mediation from the potential outcomes reasons why different employment histories framework for causal inference and presents a are scarring for workers and that those rea- clear set of assumptions under which causal sons vary with the gender of the worker. mediation effects are non-parametrically iden- tified. These aspects of the Imai and colleagues (2010) approach remove the need for model- Discussion and specific assumptions to interpret a mediation Conclusions effect as causal. Additionally, the nonparamet- The increased prevalence of nonstandard and ric identification result of the Imai and col- mismatched employment over the past four leagues (2010) approach results in an decades has occurred at the same time as estimation procedure for mediation analysis employers have become increasingly reliant that easily generalizes to nonlinear models, on the external labor market to fill vacancies. which are necessary for dependent variables In this respect, workers’ employment histo- such as the binary interview likelihood out- ries have arguably taken on a more important come examined in this manuscript.23 role in the hiring process. Yet, limited research Table 3 presents results from the mediation has examined how job applicants’ histories of analysis (Panel A is for male applicants; Panel nonstandard and mismatched employment B is for female applicants).24 For each type of shape employers’ decision-making during the employment history, the table presents the job applicant screening process. Employers average causal mediation effect (ACME) as may perceive workers with nonstandard and well as the proportion of the total effect mismatched employment histories as being explained by either competence or commit- less competent or committed, penalizing them ment perceptions. If the ACME is bold, it compared to workers with full-time, standard Pedulla 19

Table 3. Mediation Analysis of the Role of Perceived Competence and Commitment in Explaining the Effects of Nonstandard and Mismatched Employment Histories on Interview Recommendations

Perceived Competence Perceived Commitment

Proportion of Proportion of Total Effect Total Effect ACME Mediated ACME Mediated

A. Male Applicants Part-Time –.008 .098 –.037 .404 [–.042, .028] [–.072, –.000] Skills Underutilization –.090 .685 –.040 .318 [–.128, –.053] [–.077, –.003] Unemployment –.032 .366 –.008 .098 [–.063, –.001] [–.038, .023] B. Female Applicants Skills Underutilization –.109 .941 –.034 .301 [–.150, –.066] [–.073, .004]

Note: 95% confidence intervals are in brackets. Estimates were derived from 1,500 simulations; standard errors are clustered by respondent. Bold ACME estimates indicate the confidence interval does not include zero. employment histories at their level of skill economic climate improves (Jaimovich and and experience. At the same time, however, Siu 2012), these negative effects of skills any job may be better than no job in the eyes underutilization likely have broad implications of future employers. Therefore, it is also pos- for workers’ employment opportunities as the sible that nonstandard and mismatched work recovery from the Great Recession continues. may buffer workers against the scarring con- The results also indicate that the conse- sequences of long-term unemployment. quences of part-time work histories are con- Given existing data sources, it has been tingent on the job applicant’s gender. Men challenging for previous research to examine face severe penalties for part-time work histo- the set of theoretical issues explored in this ries, but women experience no penalties. article. To address these limitations, my analy- Overall, female job applicants with part-time ses drew on original data from a field experi- work histories receive a nearly identical call- ment and a complementary survey experiment. back rate to women with full-time, standard The findings demonstrate that skills underutili- work histories. Yet, men with part-time histo- zation is deeply scarring for both male and ries have a similar callback rate to men with a female job applicants. Indeed, a history of spell of unemployment. The results also sug- skills underutilization is as scarring for work- gest that women with part-time histories fare ers as a year of unemployment. As hypothe- better than men with part-time histories across sized, skills underutilization sends strong all four occupations in the field experiment, negative signals about both male and female which vary in their gender composition. applicants’ competence. For male applicants, Additionally, results from the survey experi- histories of skills underutilization are also ment provide compelling evidence that part- penalizing because they send a negative com- time work leads to penalties for male workers mitment signal. Given recent evidence that because it raises concerns about their level of jobs in the middle of the skills distribution are commitment. lost at high rates during economic recessions, Finally, there are no discernable penalizing and that these jobs tend not to rebound as the or protective effects of a history of temporary American Sociological Review 20 agency employment for either male or female should be explored in more depth in future job applicants. This lack of consequence of research. Additionally, all applications were temporary agency employment is intriguing. submitted in response to online job postings. Perhaps temporary agency work positively Consequently, it is not possible to assess signals to future employers that a worker has whether the results would look different if obtained broad and varied knowledge and applications were submitted through informal skills from working in different companies referral networks. The experimental manipu- (Marler et al. 2002). It is also possible that lations are also limited to workers’ most recent well-known temporary agencies, such as the work experiences. Future research should ones used for the résumés in this study, pro- examine whether the consequences of non- vide a positive signal to future employers standard and mismatched employment vary about the quality of the worker—potentially with the timing of those spells in a worker’s marking them as a “good” employee. These employment trajectory. Finally, the empirical potential explanations could be fruitful ave- results are limited to college-educated work- nues for future scholarship. ers of a particular age in specific labor markets Together, these findings indicate that dif- and occupations. While there are theoretical ferent types of employment histories send reasons to believe that the findings should distinct signals and have highly varied conse- generalize beyond the empirical scope of this quences, which also vary by the gender of the project, empirically testing the generalizabil- worker. Yet, this article is not without limita- ity of these findings—particularly, whether tions. First, the empirical findings leave open similar patterns emerge in the low-skilled important questions about the mechanisms labor market—opens up an important avenue underlying the gender differences in the con- for future inquiry. sequences of part-time work histories. Addi- Notwithstanding these limitations, this tionally, the discrepancy between the gender article makes important contributions to findings in the field and survey experiments research on the changing nature of employ- raises important methodological questions ment in the United States. First, the article about how, when, and why demographic char- extends theoretical insights from the literature acteristics (e.g., gender) produce evidence of on unemployment scarring by demonstrating bias and discrimination in survey experiments. that other types of labor market positions— Previous survey and lab studies that have con- specifically, nonstandard and mismatched ducted employment experiments using stu- employment—send consequential signals to dent samples have found moderating effects future employers. At the same time, the arti- of workers’ gender (Castilla and Benard 2010; cle combines insights from the literature on Correll et al. 2007), but the sample for the unemployment scarring with gender scholar- survey experiment in this article consisted of ship on the “ideal worker” norm to suggest actual hiring decision-makers and the gender that commitment, in addition to competence, differences that emerged are relatively muted. is an important signal sent by nonstandard Research disentangling why these discrepan- and mismatched employment histories. Addi- cies exist would assist in moving forward tionally, the findings contribute to sociologi- survey-experimental methodologies. cal scholarship on inequality by demonstrating The analyses presented here are limited to that access to full-time, standard employment employers’ decision-making at the initial that utilizes one’s skills appears to have applicant screening stage. Thus, they do not become a key stratifying force in the “new provide information on final hiring decisions, economy.” The effects of nonstandard and wage setting, promotions, or terminations. mismatched work are varied, but part-time While previous research suggests there are employment for men, and skills underutiliza- penalties for part-time employment histories tion for both men and women, severely limit in terms of future compensation (Ferber and workers’ labor market opportunities. Workers Waldfogel 1998), this topic and related issues with these histories may find themselves Pedulla 21 stuck in undesirable labor market positions, here raise questions about whether all types of facing challenges as they attempt to transition jobs actually open up new labor market oppor- into more stable employment. These findings tunities for workers. Indeed, certain types of encourage a shift from research to date that employment positions appear to send negative has focused primarily on the consequences of signals to future employers about workers’ nonstandard and mismatched employment for competence and commitment, penalizing them workers’ earnings, benefits, autonomy, and in similar ways to remaining unemployed. control while they are working in a nonstand- The theoretical integration and empirical ard or mismatched position. More research is findings presented in this article advance con- needed to understand how these types of temporary scholarship on the consequences of employment arrangements may have conse- the changing economic landscape. The increase quences for workers’ labor market mobility of nonstandard and mismatched employment and, ultimately, their long-term economic and relations in the United States affects workers social trajectories. not just while they are in those positions, but in Gender differences in the effects of part- certain cases also limits their opportunities as time work (and, to some extent, unemploy- they attempt to transition into their next job. ment) also contribute to sociological theories Additionally, there is important gender varia- of the “ideal worker.” The findings suggest tion in the consequences of part-time employ- that employers have already incorporated cer- ment, shedding light on the complex ways that tain types of employment experiences into gender infuses the contemporary labor market. their understandings of female labor force Together, these findings deepen current knowl- participation; therefore occupying those labor edge about how changing economic structures market positions does not violate the “ideal shape workers’ employment opportunities and worker” norm for women. Men, however, begin to identify the mechanisms through appear to be expected to maintain full-time, which those consequences operate. standard—in other words, primary breadwinner —employment trajectories to comply with “ideal worker” expectations. As theories of Appendix labor market signaling might predict, this Part A. Field Experiment Treatments empirical pattern may emerge due to gender This appendix provides examples of the dif- differences in rates of part-time employment, ferent employment histories used in the field whereby the relative prevalence of part-time experiment. The following examples are work among women likely limits the signal it drawn from the résumés for administrative sends to future employers. Thus, changes in assistant openings in Boston, Massachusetts. the underlying gender distribution of part- The employers’ names have been altered. time workers could potentially shift the ways Each treatment was the applicant’s work his- part-time employment histories differentially tory for the 12 months prior to submitting the affect men and women. Future research would job application. be well served to examine this set of issues. The findings from this research also com- plicate “work first” public policy prescriptions 1. Full-Time, Standard: that argue that any job is better than no job. Technology Company – Boston, MA Many workforce development programs are June 2012 – Present based on the premise that assisting a worker to obtain employment, any employment, will Office Manager & Executive Assistant serve as a “stepping stone” to better jobs in the •• Coordinate all office management tasks, future. While there are certainly good reasons which includes working with computer that people take any job they can find—spe- and phone system vendors, maintaining cifically in cases where economic hardship is necessary levels of office supplies, and imminent—the experimental data presented managing all office filing systems. American Sociological Review 22

•• Answer and screen incoming phone letters and other correspondence for calls, coordinate travel arrangements, executive staff. and draft memos and letters for execu- •• Coordinating conferences, meetings, and tive staff. retreats for staff, managers, and clients. •• Plan and coordinate all aspects of meet- •• Developing and improving office coordi- ings for executive staff and key nation systems, such as ordering supplies stakeholders. and updating administrative technology.

2. Part-Time: 4. Skills Underutilization: Technology Company – Boston, MA Large Retailer – Boston, MA June 2012 – Present June 2012 – Present Office Manager & Executive Assistant (Part- Sales Representative Time) •• Provide high-quality customer assis- •• Coordinate all office management tasks, tance in merchandise selection and other which includes working with computer service areas. and phone system vendors, maintaining •• Maintain high level of cleanliness and a necessary levels of office supplies, and welcoming environment on the retail managing all office filing systems. floor. •• Answer and screen incoming phone •• Build and strengthen relationships with calls, coordinate travel arrangements, repeat customers. and draft memos and letters for execu- tive staff. 5. Unemployment: The most recent job was omit- •• Plan and coordinate all aspects of meet- ted in the unemployment condition. To ensure this ings for executive staff and key résumé was of a similar length to and had the same stakeholders. number of work experiences as résumés in the oth- er conditions, a college internship was added to the 3. Temporary Employment Agency: applicant’s work history:

Temp Agency – Boston, MA Anonymous Bank – Boston, MA June 2012 – Present Summer 2004 Temporary Administrative Assistant Summer Intern Serve as a temporary Administrative Assistant •• Assisted with meeting and conference through [Name of Temp Agency]. Assignments at planning, scheduling, and answering different companies have included: phones. •• Drafted memos and correspondence and •• Answering incoming phone calls, sched- participated in special projects on an as- uling travel arrangements, and writing needed basis. Pedulla 23

Part B. Distribution of Field Experiment Applications

Table B1. Distribution of Applications Submitted in Experimental Audit Study

Frequency Percent

Employment History Full-Time 665 27.5 Part-Time 361 14.9 Temporary Agency 327 13.5 Skills Underutilization 342 14.1 Unemployment 725 30.0 Total 2,420 100.0 Applicant Gender Male 1,198 49.5 Female 1,222 50.5 Total 2,420 100.0 Labor Market Atlanta 318 13.1 Boston 484 20.0 Chicago 404 16.7 Los Angeles 484 20.0 New York City 730 30.2 Total 2,420 100.0 Occupation Accounting/Bookkeeping 372 15.4 Administrative/Clerical 416 17.2 Project Management/Management 828 34.2 Sales 804 33.2 Total 2,420 100.0

Source: Original experimental audit study data.

Part C. Field Experiment Regression Results

Table C1. Logistic Regression Models of the Consequences of Employment Histories for Receiving a Callback from an Employer, by Applicant Gender

Male Female All Applicants Applicants Applicants (1) (2) (3)

Employment History Full-Time (omitted) Part-Time –.821* .0484 –.821* (.341) (.272) (.341) Temporary Agency –.419 –.251 –.419 (.285) (.305) (.285) Skills Underutilization –.859* –.748* –.859* (.397) (.348) (.397) Unemployment –.960** –.352 –.960** (.294) (.255) (.294)

(continued) American Sociological Review 24

Table C1. (continued)

Male Female All Applicants Applicants Applicants (1) (2) (3)

Interactions Part-Time × Female .869* (.436) Temporary × Female .168 (.417) Underutilization × Female .111 (.528) Unemployment × Female .608 (.389) Female Applicant .00457 (.254) Constant –2.158*** –2.154*** –2.158*** (.179) (.181) (.179) n (clusters) 599 611 1,210 n (observations) 1,198 1,222 2,420

Note: Clustered standard errors are in parentheses. Log-odds presented. *p < .05; **p < .01; ***p < .001 (two-tailed tests).

Acknowledgments and Gregory 2001), in addition to a signaling mecha- nism. However, the human capital mechanism and I gratefully acknowledge the support and guidance offered the signaling mechanism are empirically indistin- by Devah Pager, Viviana Zelizer, Paul DiMaggio, and guishable during the initial point of job applicant Rourke O’Brien throughout the entirety of this project. I screening, because employers are not able to directly also thank Maria Abascal, Sarah Brayne, Liz Derickson, observe whether a worker has lost human capital due Luke Elliott, Crystal Feimster, Jennifer Glass, Angelina to an unemployment spell. Thus, insofar as employ- Grigoryeva, Ryan Hagen, Anna Haley-Lock, Trevor ers are concerned about human capital deterioration Hoppe, Patrick Ishizuka, Carly Knight, Karen Levy, Ken- for unemployed workers when making hiring deci- Hou Lin, Sara McLanahan, Kristin Perkins, Becky Pettit, sions, it likely manifests in the content of the nega- Vanessa Rumble, Meredith Sadin, Matt Salganik, Mat- tive signal sent by a history of unemployment. thew Salisbury, Kate Weisshaar, and the ASR editors and 3. By involuntary job loss, I am referring to being anonymous reviewers. Any mistakes are my own. fired or laid off, rather than being displaced due to a plant closing (see Gibbons and Katz 1991). Funding 4. The findings may be influenced by the particular economic climate in which the study was con- Funding for this project was provided by the National ducted: the recovery from the Great Recession. In Science Foundation (SES 1203135); the Horowitz Foun- times of economic distress, employers may be more dation for Social Policy; the Employment Instability, likely to perceive nonstandard and mismatched Family Well-Being, and Social Policy Network at the employment histories as being outside the worker’s University of Chicago; NICHD (5 R24 HD042849); control and therefore penalize them less. It is also Princeton University’s Department of Sociology; and the possible, however, that in times of economic dis- Fellowship of Woodrow Wilson Scholars. tress, more individuals are unemployed and looking for work, enabling employers to be more selective during the hiring process. Future research should Notes examine this set of issues. 1. The insights offered by the literature on signaling 5. I selected a treatment period of 12 months because and the hiring process are not entirely distinct from of the need to keep the duration of the treatment notions of statistical discrimination (see Phelps 1972). equal across conditions, and the need to use an 2. The unemployment scarring literature also points to amount of time that would be appropriate for each a human capital mechanism (Arulampalam, Gregg, employment history. Pedulla 25

6. A separate set of field experiment conditions 15. There is also a positive and statistically significant included a set of African American racialized interaction between having a part-time history and names to examine how race shapes the effects of being a female applicant in a logistic regression nonstandard and mismatched work. Those findings model predicting callbacks (see Model 3 in Table are presented separately. C1 in the Appendix). 7. There is no statistically significant difference in the 16. Additionally, 48.6 percent of workers in sales occu- callback rate for the distinct résumé templates. pations and 72.9 percent of workers in accounting 8. This study examines college-educated workers, in and bookkeeping occupations are female (Bureau part, for methodological reasons. Many low-skilled of Labor Statistics 2014). jobs for workers without any college education are 17. Given the large reduction in sample size when ana- not posted on national job websites. lyzing the data by occupation, these differences 9. The pretesting was conducted through Amazon generally do not reach statistical significance. I also .com’s Mechanical Turk platform. examined interactions between each occupation and 10. To determine how to signal each employment his- each employment history. The results indicate that tory, I examined résumés in online résumé banks. the consequences of nonstandard and mismatched 11. Unemployment was signaled through dates that employment do not differ by the occupation to the applicant did not have a job. The formal defi- which the application was submitted. nition of unemployment is that an individual does 18. When the full interview likelihood scale is used, the not have work and is looking for work. The second results diverge slightly. Female applicants are more component of the definition is not formally signaled, likely than male applicants to remain in the highest although the jobless individual is clearly looking for interview likelihood category (“very likely”), but work at the time the application is submitted. To the if they do not receive the highest interview likeli- employer, unemployment and joblessness are indis- hood rating, they are more severely penalized than tinguishable; I thus refer to the condition with a spell men. When women do not receive a “very likely” of joblessness as the “unemployment” condition. interview recommendation rating, they generally 12. The anonymized field and survey experiment data receive a rating of “somewhat likely,” two points necessary to reproduce the analyses in this article down the scale. Men, however, are more likely than will be made available through the Dataverse Net- women to face some penalty for nonstandard or work (http://www.thedata.org) no later than July mismatched employment. When men are penalized, 2016. though, they generally receive the “likely” rating, 13. In a few cases, I limited the search to jobs posted one point down the scale. within fewer than 30 days. In these cases, the com- 19. Results are consistent when these controls are puter script would not run for the full 30-day search excluded. period, but worked for these shorter amounts of time. 20. Twelve respondents did not provide an interview The level of education included in the search criteria recommendation for at least one of the applicants was also different across occupations. For accounting they reviewed. Therefore, 891 respondents are and sales jobs, the education level was limited to jobs included in the analyses in Table 2. requiring an associate’s or bachelor’s degree. For the 21. However, I replicated the survey experiment with project manager/manager openings, the search was a nonrandom sample of respondents from Amazon limited to jobs requiring a bachelor’s degree. Finally, .com’s Mechanical Turk and asked them to evaluate I did not limit the administrative assistant searches résumés for a managerial position. Again, I found no by education, because many employers did not spec- moderating effects of gender on the interview like- ify any education level requirement for this job type. lihood of workers with a part-time history, which Additionally, some job openings required complet- provides some evidence against the argument that ing intensive applications on the employer’s website, the discrepancy in the gender findings between the which the IRB protocol did not cover and which field and survey experiments is due primarily to the often required essay questions that would have made occupation under investigation. it more difficult to ensure that differences in answers 22. For additional details about the exact procedure were not responsible for driving the differences in used to estimate the average causal mediation callbacks. Thus, applications were not submitted for effect, see Imai and colleagues (2011:773–74). Imai these job openings. and colleagues (2010) also provide techniques for 14. The findings are consistent when controls are examining the sensitivity of the mediation analysis included for occupation, labor market, and their findings to the key identifying assumptions of the interaction, as well as when linear probability approach. The sensitivity parameters for the media- regression models are used. Additionally, the tion analyses presented here were in line with pre- findings are consistent when job-posting-specific vious studies (see Imai et al. 2011), suggesting the random effects are included in the full regression findings are generally robust. model with controls for occupation, labor market, 23. Each mediation analysis presented in this article was and their interaction. conducted using 1,500 simulations. The mediation­ American Sociological Review 26

results are robust to the inclusion of controls for the Bureau of Labor Statistics. 2005. “Contingent and Alter- hiring decision-makers’ sex, race, age, marital sta- native Employment Arrangements, February 2005.” tus, income (logged), and education. Retrieved February 9, 2012 (http://www.bls.gov/ 24. The empirical results are similar when a regression- news.release/History/conemp.txt). based framework is used to examine the role of Bureau of Labor Statistics. 2013. “Persons at Work by perceived competence and perceived commitment Occupation, Sex, and Usual Full- or Part-time Sta- in explaining the effects of nonstandard and mis- tus.” Retrieved August 24, 2013 (http://www.bls.gov/ matched employment. cps/cpsaat23.htm). Bureau of Labor Statistics. 2014. “Employed Persons by Detailed Occupation, Sex, Race, and Hispanic or References Latino Ethnicity.” Retrieved March 6, 2015 (http:// Acker, Joan. 1990. “Hierarchies, Jobs and Bodies: A The- www.bls.gov/cps/cpsaat11.htm). ory of Gendered Organizations.” Gender & Society Bureau of Labor Statistics. 2015. “Employed and Unem- 4(2):139–58. ployed Full- and Part-Time Workers by Age, Sex, Addison, John T., Chad Cotti, and Christopher J. Surfield. Race, and Hispanic or Latino Ethnicity.” Retrieved 2009. “Atypical Work: Who Gets It, and Where Does May 15, 2015 (http://www.bls.gov/cps/cpsaat08.htm). It Lead? Some U.S. Evidence Using the NLSY79.” Butler, Adam B., and Amie Skattebo. 2004. “What IZA Discussion Paper No. 4444, Bonn, Germany. Is Acceptable for Women May Not Be for Men: Addison, John T., and Christopher J. Surfield. 2009. The Effect of Family Conflicts with Work on Job-­ “Does Atypical Work Help the Jobless? Evidence Performance Ratings.” Journal of Occupational and from a CAEAS/CPS Cohort Analysis.” Applied Eco- Organizational Psychology 77(4):553–64. nomics 41(9):1077–87. Cappelli, Peter. 2001. “Assessing the Decline of Internal Allen, Tammy D., and Joyce E. A. Russell. 1999. “Paren- Labor Markets.” Pp. 207–245 in Sourcebook of Labor tal Leave of Absence: Some Not So Family-Friendly Markets: Evolving Structures and Processes, edited Implications.” Journal of Applied Social Psychology by I. Berg and A. L. Kalleberg. New York: Kluwer 29(1):166–91. Academic / Plenum Publishers. Arulampalam, Wiji, Paul Gregg, and Mary Gregory. Castilla, Emilio J., and Stephen Benard. 2010. “The Para- 2001. “Unemployment Scarring.” Economic Journal dox of Meritocracy in Organizations.” Administrative 111(475):F577–F584. Science Quarterly 55(4):543–676. Autor, David H. 2003. “Outsourcing at Will: The Con- Cha, Youngjoo. 2010. “Reinforcing Separate Spheres: tribution of Unjust Dismissal Doctrine to the Growth The Effect of Spousal Overwork on the Employment of Employment Outsourcing.” Journal of Labor Eco- of Men and Women in Dual-Earner Households.” nomics 21(1):1–42. American Sociological Review 75(2):303–329. Autor, David H., and Susan N. Houseman. 2010. “Do Clawson, Dan, and Mary Ann Clawson. 1999. “What Temporary-Help Jobs Improve Labor Market Out- Has Happened to the U.S. Labor Movement? Union comes for Low-Skilled Workers? Evidence from Decline and Renewal.” Annual Review of Sociology ‘Work First.’” Applied Economics 2(3):96–128. 25:95–119. Becker, Gary S. 1964. Human Capital: A Theoretical and Correll, Shelley J., and Stephen Benard. 2006. “Biased Empirical Analysis, with Special Reference to Educa- Estimators? Comparing Status and Statistical ­Theories tion. New York: Columbia University Press. of Gender Discrimination.” Social Psychology of the Bertrand, Marianne, and Sendhil Mullainathan. 2004. Workplace 23:89–116. “Are Emily and Greg More Employable than Laki- Correll, Shelley J., Stephen Benard, and In Paik. 2007. sha and Jamal? A Field Experiment on Labor Mar- “Getting a Job: Is There a Motherhood Penalty?” ket Discrimination.” American Economic Review American Journal of Sociology 112(5):1297–1338. 94(4):991–1013. Cuddy, Amy J. C., Susan T. Fiske, and Peter Glick. 2004. Bielby, William T., and Denise D. Bielby. 2002. “Tell- “When Professionals Become Mothers, Warmth ing Stories about Gender and Effort: Social Science Doesn't Cut the Ice.” Journal of Social Issues Narratives About Who Works Hard for the Money.” 60(4):701–718. Pp. 193–217 in The New Economic Sociology: Devel- Eagly, Alice H., and Valerie J. Steffen. 1986. “Gender opments in an Emerging Field, edited by M. F. Guil- Stereotypes, Occupational Roles, and Beliefs about len, R. Collins, P. England, and M. Meyer. New York: Part-Time Employees.” Psychology of Women Quar- Russell Sage Foundation. terly 10(3):252–62. Bills, David B. 2003. “Credentials, Signals, and Screens: Epstein, Cynthia Fuchs, Carroll Seron, Bonnie Oglensky, Explaining the Relationship between Schooling and and Robert Sauté. 1999. The Part-Time Paradox: Job Assignment.” Review of Educational Research Time Norms, Professional Life, Family and Gender. 73(4):441–69. New York: Routledge. Blair-Loy, Mary. 2003. Competing Devotions: Career Erdogan, Berrin, and Talya N. Bauer. 2011. “The Impact and Family among Women Executives. Cambridge, of Underemployment and Turnover and Career MA: Harvard University Press. Trajectories.” Pp. 215–32 in Underemployment: Pedulla 27

­Psychological, Economic, and Social Challenges, NBER Working Paper Series #18334, Cambridge, edited by D. C. Maynard and D. C. Feldman. New MA. York: Springer. Kalleberg, Arne L. 1995. “Part-Time Work and Workers Eriksson, Stefan, and Dan-Olof Rooth. 2014. “Do in the United States: Correlates and Policy Issues.” Employers Use Unemployment as a Sorting Criterion Washington & Lee Law Review 52(3):771–98. When Hiring? Evidence from a Field Experiment.” Kalleberg, Arne L. 2000. “Nonstandard Employment American Economic Review 104(3):1014–39. Relations: Part-Time, Temporary and Contract Feldman, Daniel C. 1990. “Reconceptualizing the Nature Work.” Annual Review of Sociology 26:341–65. and Consequences of Part-Time Work.” Academy of Kalleberg, Arne L. 2007. The Mismatched Worker. New Management Review 15(1):103–112. York: W.W. Norton & Company. Ferber, Marianne A., and Jane Waldfogel. 1998. “The Kalleberg, Arne L. 2009. “Precarious Work, Insecure Long-Term Consequences of Nontraditional Employ- Workers: Employment Relations in Transition.” ment.” Monthly Labor Review 121(May):3–12. American Sociological Review 74(1):1–22. Fiske, Susan T. 1998. “Stereotyping, Prejudice, and Dis- Kalleberg, Arne L., Barbara F. Reskin, and Kenneth crimination.” Pp. 357–411 in Handbook of Social Hudson. 2000. “Bad Jobs in America: Standard and Psychology, edited by D. Gilbert, S. T. Fiske, and Nonstandard Employment Relations and Job Quality G. Lindzey. Boston: McGraw-Hill. in the United States.” American Sociological Review Fiske, Susan T., Amy J. C. Cuddy, Peter Glick, and Jun 65(2):256–78. Xu. 2002. “A Model of (Often Mixed) Stereotype Karren, Ronald, and Kim Sherman. 2012. “Layoffs and Content: Competence and Warmth Respectively Fol- Unemployment Discrimination: A New Stigma.” low from Perceived Status and Competition.” Journal Journal of Managerial Psychology 27(8):848–63. of Personality and Social Psychology 82(6):878–902. Kelly, Erin L., Samantha K. Ammons, Kelly Chermack, Fuller, Sylvia. 2011. “Up and On or Down and Out? Gen- and Phyllis Moen. 2010. “Gendered Challenge, Gen- der, Immigration and the Consequences of Temporary dered Response: Confronting the Ideal Worker Norm Employment in Canada.” Research in Social Stratifi- in a White-Collar Organization.” Gender & Society cation and Mobility 29(2):155–80. 24(3):281–303. Gangl, Markus. 2006. “Scar Effects of Unemployment: Korpi, Tomas, and Henrik Levin. 2001. “Precarious Foot- An Assessment of Institutional Complementarities.” ing: Temporary Employment as a Stepping Stone Out American Sociological Review 71(6):986–1013. of Unemployment in Sweden.” Work, Employment, & Gibbons, Robert, and Lawrence F. Katz. 1991. “Layoffs Society 15(1):127–48. and Lemons.” Journal of Labor Economics 9(4): Kroft, Kory, Fabian Lange, and Matthew J. Noto- 351–80. widigdo. 2013. “Duration Dependence and Labor Gonos, George. 1997. “The Contest Over ‘Employer’ Market Conditions: Theory and Evidence from a Status in the Post-War United States: The Case of Field Experiment.” Quarterly Journal of Economics Temporary Help Firms.” Law & Society Review 128(3):1123–67. 31(1):81–110. Leung, Ming D. 2014. “Dilettante or Renaissance Person? Gornick, Janet C., and Marcia K. Meyers. 2003. Families How the Order of Job Experiences Affects Hiring in That Work: Policies for Reconciling Parenthood and an External Labor Market.” American Sociological Employment. New York: Russell Sage Foundation. Review 79(1):136–58. Hatton, Erin. 2011. The Temp Economy: From Kelly Girls Ma, Ching-to Albert, and Andrew M. Weiss. 1993. “A to Permatemps in Postwar America. Philadelphia: Signaling Theory of Unemployment.” European Eco- Temple University Press. nomic Review 37(1):135–57. Heerwig, Jennifer A., and Brian J. McCabe. 2009. “Edu- Marler, Janet H., Melissa Woodard Barringer, and George cation and Social Desirability Bias: The Case of a T. Milkovich. 2002. “Boundaryless and Traditional Black Presidential Candidate.” Social Science Quar- Contingent Employees: Worlds Apart.” Journal of terly 90(3):674–86. Organizational Behavior 23(4):425–53. Hollister, Matissa. 2011. “Employment Stability in the Mavromaras, Kostas, Peter Sloane, and Zhang Wei. 2015. U.S. Labor Market: Rhetoric versus Reality.” Annual “The Scarring Effects of Unemployment, Low Pay, Review of Sociology 37:305–324. and Skills Under-utilization in Australia Compared.” Imai, Kosuke, Luke Keele, and Dustin Tingley. 2010. “A Applied Economics 47(23):2413–29. General Approach to Causal Mediation Analysis.” McKee-Ryan, Frances M., and Jaron Harvey. 2011. “‘I Psychological Methods 15(4):309–334. Have a Job But…’: A Review of Underemployment.” Imai, Kosuke, Luke Keele, Dustin Tingley, and Tep- Journal of Management 37(4):962–96. pei Yamamoto. 2011. “Unpacking the Black Box of Moss, Philip I., and Chris Tilly. 2001. Stories Employers Causality: Learning about Causal Mechanisms from Tell: Race, Skill, and Hiring in America. New York: Experimental and Observational Studies.” American Russell Sage Foundation. Political Science Review 105(4):765–89. Nollen, Stanley D. 1996. “Negative Aspects of Tem- Jaimovich, Nir, and Henry E. Siu. 2012. “The Trend is porary Employment.” Journal of Labor Research the Cycle: Job Polarization and Jobless Recoveries.” 17(4):567–82. American Sociological Review 28

Oyer, Paul, and Scott Schaefer. 2011. “Personnel Eco- Turco, Catherine J. 2010. “Cultural Foundations of Token- nomics: Hiring and Incentives.” Handbook of Labor ism: Evidence from the Leveraged Buyout Industry.” Economics 4:1769–1823. American Sociological Review 75(6):894–913. Pager, Devah. 2007. “The Use of Field Experiments for U.S. Census Bureau. 2008. “Statistics of U.S. Businesses: Studies of Employment Discrimination: Contribu- 2008.” Retrieved May 25, 2014 (https://www.census tions, Critiques, and Directions for Future Research.” .gov/epcd/susb/latest/us/US--.HTM). Annals of the American Academy of Political and Vaisey, Stephen. 2006. “Education and its Discontents: Social Science 609(1):104–133. Overqualification in America, 1972–2002.” Social Phelps, Edmund S. 1972. “The Statistical Theory of Forces 85(2):834–64. Racism and Sexism.” American Economic Review Vosko, Leah F. 2000. Temporary Work: The Gendered 62(4):659–61. Rise of a Precarious Employment Relationship. Rivera, Lauren A. 2012. “Hiring as Cultural Matching: Toronto: University of Toronto Press. The Case of Elite Professional Service Firms.” Amer- Williams, Joan C. 2001. Unbending Gender: Why Fam- ican Sociological Review 77(6):999–1022. ily and Work Conflict and What to Do about It. New Rudman, Laurie A., and Julie E. Phelan. 2008. “Back- York: Oxford University Press. lash Effects for Disconfirming Gender Stereotypes in Yu, Wei-hsin. 2012. “Better Off Jobless? Scarring Effects Organizations.” Research in Organizational Behavior of Contingent Employment in Japan.” Social Forces 28:61–79. 90(3):735–68. Ruhm, Christopher J. 1991. “Are Workers Permanently Scarred by Job Displacements?” American Economic David S. Pedulla is Assistant Professor of Sociology Review 81(1):319–24. and Faculty Research Associate in the Population Schilling, Melissa A., and H. Kevin Steensma. 2001. “The Research Center at the University of Texas at Austin. Use of Modular Organizational Forms: An Industry- Using primarily experimental and quantitative Level Analysis.” Academy of Management Journal approaches, his research examines the processes under- 44(6):1149–68. lying racial and gender stratification in the labor market Smith, Vicki. 1997. “New Forms of Work Organization.” as well as the consequences of nonstandard, contingent, Annual Review of Sociology 23:315–39. and precarious employment on workers’ social and eco- Spence, Michael. 1973. “Job Market Signaling.” Quar- nomic outcomes. terly Journal of Economics 87(3):355–74.