A Correspondence Study of Skill and Context Variation in Parental Hiring Probabilities Before, During, and After the COVID-19 Pandemic

Last registered on October 19, 2023

Pre-Trial

Trial Information

General Information

Title
A Correspondence Study of Skill and Context Variation in Parental Hiring Probabilities Before, During, and After the COVID-19 Pandemic
RCT ID
AEARCTR-0004015
Initial registration date
March 25, 2019

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
March 31, 2019, 11:01 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
October 19, 2023, 3:50 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Region

Primary Investigator

Affiliation
University of Bath

Other Primary Investigator(s)

PI Affiliation
University of Bath

Additional Trial Information

Status
Completed
Start date
2019-10-01
End date
2022-06-30
Secondary IDs
ERC 680958
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This is a harmonized correspondence study on employer hiring in Finland and the UK, to compare gender differences with parental status the treatment effect. The over-arching research question has two dimensions. First, this study evaluates whether there are gender differences in relative positive callback rates for parents versus childless married individuals at an early stage of the hiring process, and if they differ across low-, medium- and high-skill occupations. Second, it uses a comparative design to generalize results across two diverse country contexts in terms of institutional support gender employment equality and maternal employment in particular. We select the same three gender-neutral occupations for analysis in each labor market, harmonizing education and work experience in each across the countries. After the start of fielding in both countries, however, the COVID-19 pandemic and associated lockdowns had dramatic effects on the labor markets. We therefore added a new facet of assessing whether the lower demain exacerbated discrimination against mothers and favoring fathers.
External Link(s)

Registration Citation

Citation
Cooke, Lynn Prince and Soyoung Kweon. 2023. "A Correspondence Study of Skill and Context Variation in Parental Hiring Probabilities Before, During, and After the COVID-19 Pandemic." AEA RCT Registry. October 19. https://doi.org/10.1257/rct.4015-1.4
Former Citation
Cooke, Lynn Prince and Soyoung Kweon. 2023. "A Correspondence Study of Skill and Context Variation in Parental Hiring Probabilities Before, During, and After the COVID-19 Pandemic." AEA RCT Registry. October 19. https://www.socialscienceregistry.org/trials/4015/history/197277
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
Intervention (Hidden)
Intervention Start Date
2019-10-01
Intervention End Date
2021-12-31

Primary Outcomes

Primary Outcomes (end points)
The outcome of interest is whether there are group differences in relative positive callback rates versus rejections, and if these differ across countries. A positive callback is a request for interview. Given that few employers formally reject unsuccessful applicants, a non-response for the duration of the study will be coded a rejection. We analyze group and country differences in requests to interview versus rejection to ascertain whether we can reject the null hypothesis that employers do not discriminate in favor of fathers or against mothers as compared with their childless peers during initial screening of the hiring process. We further assess whether there are skill differences in these effects, given that statistical discrimination is argued to be less likely for more highly skilled applicants because of the richness of productivity information on their CVs. Given fielding continued during the pandemic, we further assess whether a drop in labor demand associated with lockdowns increased discrimination against mothers as argued in existing theory, or possibly also increased it against fathers who took on an unusual and high amount of domestic tasks and childcare.
Primary Outcomes (explanation)
We assess discrimination by comparing group differences in callback rates for applicants with commensurate CVs.

Secondary Outcomes

Secondary Outcomes (end points)
Assess bias in the design per Heckman and Siegelman (1993) critique by fielding additional low-quality, low-skill CVs in the post-pandemic period.
Secondary Outcomes (explanation)
Use Neumark's (2012) method for assessing group differences in unobserved productivity by comparing estimates from probit and heteroskedastic probit equations.

Experimental Design

Experimental Design
This is a comparative correspondence study assessing potential gender differences in hiring discrimination around parenthood in a low-, medium- and high-skilled occupation in two countries. For the comparative element of this study, the countries were selected given their variation in institutional supports for both class and gender equality. Finland has more extensive labor market institutions and policy support for dual-earning and caring. The more liberal UK welfare regime minimizes state intervention in market mechanism affecting either class or gender equality. To compare effects across skill levels, three occupations requiring different levels of education were selected that are also less gender-segregated in both countries: call center workers, restaurant managers, and accountants.

Two sets of comparable application materials in terms of employment trajectories and accomplishments were developed for each occupation in each country based on national statistics, data bases, LinkedIn profiles, and discussions with country HR experts and recruiting professionals. The sets of application materials were harmonized to equalize number of years of post-secondary education plus work experience within occupations across countries.

External validity of instruments for each occupation has been confirmed via interviews with HR experts, whereas commensurability of sets of application materials (internal validity) has been validated with in-person student and online abs in each country students and adult samples, in a lab and online.

In fielding, job advertisements will be drawn from online employment databases in each country to provide widest coverage of available jobs. We use a 2x2 within-subject design for effects of parenthood (i.e., for each job ad, two applications will be sent of the same gender, both of whom are married but one is randomly assigned to be a parent (treatment) and the other childless (control)). Parenthood is signaled in the personal information on the resume (doing so is still more normative in Europe). Differences in positive callback rates between men and women are investigated by means of between-subject variation. Gender is manipulated by applicant names, drawn from the most common first names in each country. To evaluate treatment effects, the study documents positive callbacks from employers, comparing group differences in requests for an interview with rejections or non-response. We further divide the fielding period into pre-, pandemic, and post-pandemic periods based on timing of lockdowns and labor demand flows.
Experimental Design Details
This is a comparative correspondence study assessing potential gender differences in hiring discrimination around parenthood in a low-, medium- and high-skilled occupation in two countries. Two sets of comparable application materials were developed for each occupation in each country based on national statistics, data bases, LinkedIn profiles, and discussions with country HR experts and recruiting professionals.

We use a 2x2 within-subject design for effects of parenthood. Two either female or two male sets of application materials will be sent in reply to employer job advertisements. All applicants are married; the treatment of interest is parental status. Parental status is randomly assigned and indicates that the applicant has one approximately 2-year-old child. In Finland, information on family status, along with the age of the child are frequently included among the personal information; we follow suit and signal parenthood and the age of the child by including it with contact information. It is less common to do this in the UK, but HR managers confirmed including family status in the personal information is still done occasionally, in part because the UK receives EU applicants that still follow this practice. We signal the UK child's *age* on the resume under personal interests indicating a hobby with a toddler.

Differences in callback rates between men and women are investigated by means of between-subject variation. Gender is manipulated through applicants’ names, which are clearly identifiable as male or female and selected from each country’s most common first names. In Finland and Germany, it is also usual to include a photo on the resume. Stock professional photos were purchased that approximated the age and expected professional attire of applicants for each occupation.

The definition of skill level follows the International Standard Classification of Occupations (ISCO), and is commensurate across the country-specific occupational classification schemes. High-skilled occupations usually require a tertiary qualification of 3 to 6 years, and include complex problem solving based on specialized knowledge; medium-skilled occupations involve complex technical and practical tasks and often require a post-secondary degree of 1 to 3 years; low-skilled occupations require only the first or possibly second stage of secondary education. The second criterion was that a substantial share of the workforce needed to be employed in the selected occupation. Larger occupations ensure that the number of vacancies is high and that each employer can be utilized only once to minimize the imposition on them. Third, the tasks in the selected occupations should not vary substantially across organizations or sectors. This ensures that the developed application materials would be suitable for the majority of job advertisements and more generalizable. The final selected occupations were accountants (high skill), restaurant managers (medium skill), and call center workers (low skill), based on sex composition, size of occupation, and nature of tasks across organizations. The design aimed for gender balance in each occupation in each country, to avoid conflating discrimination based on parenthood with that based on gender.

For internal validity, characteristics of applicants in each occupation which affect their hiring probability such as age, education and work experience must be kept constant. External validity requires that the applicants be as typical as possible for their occupation and life stage. Achieving internal and external validity in a cross-nationally comparative correspondence study assessing occupations with different skill levels raises particular challenges. For one, although educational requirements for each occupation are fairly comparable across countries by design, national differences in educational systems result in cross-country variation in the age at which applicants in the same occupation might enter as well as graduate from the educational system.

Another consideration is that the length of parents' employment histories is contingent on the age at which they likely entered parenthood, which varies with education. The cross-occupational average age at first birth ranges from almost 28 to almost 33 years among women, with highly-educated women in all countries becoming parents later than medium- and low-skilled ones.

To balance these aspects, we first harmonize according to mothers’ age at first birth in each country for each occupation, enhancing external validity that might affect employers’ perceptions. Second, to aid cross-country comparisons of occupations, we harmonized what we refer to as “total human capital” for each occupation across countries. We define total human capital as the sum of years in post-compulsory schooling plus work experience. Thus, while the length of education, or years of labor market experience in an occupation might differ slightly across countries, the two added together are harmonized for each occupation. Accountants thus have 12 years of post-secondary schooling and experience; restaurant managers 11, and call center workers 10.

A consequence of this two-step harmonization is that the age of applicants within the same occupation differs slightly across countries (~1 year). We justify the decision to harmonize education and experience rather than age because human capital is more salient to potential employers for young adults than age. HR managers and other recruiting experts confirmed the importance of commensurate education and experience over average age. Thus, any country-variation in positive callback rates for a given skill level likely stems from institutional effects rather than the (minor) cross-country variation in ages for each occupation.

Childless women and mothers of the same age may not have comparable years of work experience, however, as mothers tend to interrupt employment as a result of a birth. This raises the possibility that employers may discriminate against mothers not because of their status, but because they may have less experience at a given age than their childless peers. However, determining a plausible career-related interruption for the fictitious applicants is problematic. Not only do the selected countries differ in the length of parental leave, the duration of family-related employment interruptions varies substantially among women within a country. Furthermore, discussions with HR experts highlighted that parental leave that occurs during an ongoing employment episode is seldom itemized on the resume. For this reason, we follow existing studies and keep work experience between childless applicants and parents constant. Additionally, the employment trajectories on each occupational resume were made to be commensurate in career progression and accomplishments. The number of job moves is held constant, as is the size of the firms. Commensurate accomplishments further minimize the importance of deduced employment interruptions.

A final issue is that Finnish men are affected by mandatory conscription of six months to one year after secondary schooling. Consequently, a gap year was added to all Finnish medium- and high-skilled applicants between secondary and tertiary education. This gap accounts for men’s conscription, but works equally well for women, as it is common for all Finnish students to take a gap year before entry into higher education. Low-skilled men, in turn, are assigned six months of military service. Lower-skilled Finnish women are given six months in a job unrelated to call centers.

The commensurability of the application materials was validated via a series of lab and online validation experiments in the two countries.

The selected template of resume and cover letter, order of applications, and names of applicants are assigned randomly, whereas photos are assigned randomly to the Finnish resumes. If the posted job advertisement is more than 50 kilometers from the applicant’s listed home address, the software inserts an additional sentence in the cover letter, that the applicant is looking to move to the region to be closer to family.

Fielding began in the UK in August 2019, and in Finland in September 2019. The first month constituted a piloting phase to refine the types of jobs replied to and confirm that employers view the application sets as commensurate. We started fielding with software that took a random selection of job ads from several online employment websites, but quickly had to revert to using research assistants and the software proved unreliable. During lockdowns and in Finland, rather than a random sample, all suitable jobs were applied to. Top-level information was gathered relating to size of employer, sector, industry, nature of job (part-time), region, desirable special skills, and distance from applicant. The experimental data of interest are invitations to interview orinformal discussions relative to rejections.
Randomization Method
All randomization of application materials is done via a computer software program.
Randomization Unit
Which gender pair to send randomized for responding to suitable job advertisement. Individual randomization in terms of treatment, template, order sent, and for Finland, the CV photo.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
Not applicable.
Sample size: planned number of observations
Responding to 500 job advertisements per occupation in each country (1000 applications sent x 3 occupations x 3 countries)=9000 applications
Sample size (or number of clusters) by treatment arms
We aimed to send a pair of same-gender applications is sent to each of 500 job advertisements for each occupation, one being the control and the other the treatment based on power calculations from published studies at the time of design. However, our response rates were much higher than published studies, indicating the need for a higher number of applications to be sent. Labor demand in Finland and during the pandemic was quite low. These factors made for uneven numbers fielded. In the UK: 1832 call ceter workers, 1260 assistant restaurant managers, and 898 accountant applications were sent. In Finland, we could field a total of only 834 pairs across the two years (470 call center, 464 assistant restaurant managers, and 734 accountant applications).
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
The goal is to have a sufficient sample size of job ads in each occupation to yield beta=.1 (power .9) and alpha=.05 for population gender differences between affirmative call-backs for the treatment (parents) versus controls (childless) across occupations and countries. Note in existing correspondence studies this is often relaxed to beta=.2 (power .8) and alpha=.20. However, studies testing parental discrimination have not published their power calculations so we used McNemar to estimate these as best we could with the published information. As relative callback rates for parents relative to non-parents of each gender differ substantially across occupations and countries, it is difficult to do power calculations to estimate necessary sample size before piloting. The initial plan is therefore to reply to 500 job ads (1000 applications sent) for each occupation in each country as noted above. Funding is available to keep fielding into 2021 if necessary.
Supporting Documents and Materials

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
IRB

Institutional Review Boards (IRBs)

IRB Name
European Research Council Ethics Committee
IRB Approval Date
2016-05-31
IRB Approval Number
ERCEA/BT/ ercea.b.1(2016)3148812
IRB Name
University of Turku Ethics Committee
IRB Approval Date
2016-06-06
IRB Approval Number
Statement 20/2016
IRB Name
Berlin Social Science Center Research Ethics Committee
IRB Approval Date
2016-11-16
IRB Approval Number
N/A
IRB Name
Department of Social and Policy Sciences Ethics Committee, University of Bath
IRB Approval Date
2018-05-01
IRB Approval Number
N/A

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials