Gender and the Labour Market: Do Employers Discriminate According to the Job Characteristics?

M. José González, Universitat Pompeu Fabra
Clara Cortina, Universitat Pompeu Fabra
Jorge Rodriguez, Universitat Pompeu Fabra

Using correspondence testing, we investigate how job characteristics affect gender discriminatory practices. We particularly explore whether discrimination in hiring against women is moderated by certain key characteristics of the occupations: the sex composition (male or female-dominated); the nature of the tasks to be implemented (whether or not managerial and supervisory tasks are required); and the expected educational level (low, medium or high). To do so, we carried out a innovative correspondence study in 2016, whereby we sent two pairs of matched male-female applications (4 fake résumés), rather than just two as it is typically the case, to 1,371 job openings (5,615 résumés) from an heterogeneous selection of occupations in two large cities. Differences in response rates by sex were then used as a tool to assess statistical discrimination. The results show that women''s disadvantage related to the education required for the job applies both to low and high profiles (compared to mid-profiles). We prove that women are discriminated against in non-operational jobs and in jobs requiring either low or high educational credentials. Additionally, women''s applications are given significantly more priority by employers than men''s, only if they apply to a female-dominated occupation.

ExtendedAbstract

 

Genderand the labour market: do employers discriminate according to the jobcharacteristics?

Theme: Economics,Human Capital, and Labour Markets

 

Using correspondencetesting, we investigate how job characteristics affect gender discriminatorypractices. We particularly explore whether discrimination in hiring againstwomen is moderated by certain key characteristics of the occupations: the sexcomposition (whether the presence of more women than men favours a preferencefor women); the nature of the tasks to be implemented (more or less related todecision-making and supervising other’s work); and the required level ofeducation for the position (low, medium or high).

The correspondence studyconsisted of sending pairs of applications to large samples of job openingsfrom fictitious candidates. Candidates'' résumés were equivalent in all respectsbut for some traits that might affect the probability of being called back foran interview. The ultimate goal was detecting gender discrimination at theearly stages of the selection process, since we limit the investigation toestablishing differences in the proportion of men and women called back for aninterview, not on the outcome of these interviews, which the candidatesdeclined to attend. We also investigated the extent to which discriminationvaries depending on the level of qualification of the candidate and on whetherthey have children. To do this, we sent two pairs (4 CVs) of fictitious applicationsmatched on nearly everything except the sex of the candidate to 1,371 jobvacancies (5,615 résumés) advertised on a major Internet search engine over aneight-month period between April and November, 2016.

We estimate two types ofmodels according to whether the dependent variable is: a) a binary variable measuringwhether the candidate was or was not called back by the employer to set up aninterview or to offer him or her the job; or b) a limited interval variablereflecting the call order in which the candidates were called back, from 0 (heor she was never called back, to 1 (she or he was called back in sixth place ),to 2, 3, 4, 5, and 6 (she or he was called back in first place). For the firstset of analyses (a), we use multi-level logit models; for the second set (b), multi-levelordered logit models.

We tested three main hypotheses. First, weexpected positive discriminatory practices for women in jobs which werefemale-dominated, and discrimination against women in male-dominated jobs (sex-compositionhypothesis), as this variable might activate gender stereotypes (typicallymasculine or feminine jobs) or anticipate colleague’s preferences in theworking place. Secondly, we expected discriminatory practices against women tobe more extensive in non-operational jobs, as employers might have higheruncertainty about women’s ability to assume decision-making/managerial andsupervisory tasks (decision-making hypothesis). Third and last, weexpected that discriminatory practices vary according to the educational levelrequired in the job opening (educational expectations hypothesis). Highereducation might be associated to higher costs of replacing the employer and,therefore, this might reinforce female penalty linked to the risk of maternityleave and increase female discrimination. Additionally, higher education mightdecrease the uncertainty about the candidate productivity reducingdiscrimination. Both effects might also counterbalance each other.

The preliminary results(see Tables 1 and 2) show that women''s disadvantage related to the educationrequired for the job applies both to low and high profiles (compared tomid-profiles). We prove that women are discriminated against in non-operationaljobs and in jobs requiring either low or high educational credentials.Additionally, women''s applications were given significantly more priority byemployers than men''s, only when they apply to a female-dominated occupation. This workcontributes to the literature on employment discrimination in two fundamentalways. First, we improve the classical experimental approach applied to correspondencestudies by sending two, rather than just one, set of matched-paired,male-female applications to the same job opening, and alternating the factorson which the pairs of candidates were distinguished (parenthood status andqualifications). This design allowed us to increase the power of the experimentand test rarely explored hypotheses about the possible roots of genderdiscrimination. Second, we provide evidence of employers’ gender discriminationaccording to the occupational characteristics.

Table 1: Number of Applications andCall-Back Rates, by Job Characteristics

Table2: Multilevel logit of being called back for an Interview, by characteristicsof the job, and controlling for candidates’ and application’s characteristics

Baseline: men, low skilled, childless, mixed occupation,operational, mid-education job, Madrid, 0 previous applicants, only 2 applications sent

*Significant at the 0.1 level;**Significant at the 0.05 level; ***Signficantat the 0.001 level

 

 

Presented in Session 1150: Economics, Human Capital, and Labour Markets