thesis for gender pay gap

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

The persistence of pay inequality: The gender pay gap in an anonymous online labor market

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing

* E-mail: [email protected] (LL); [email protected] (LB)

Affiliation Department of Psychology, Lander College, Flushing, New York, United States of America

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing

Affiliation Department of Computer Science, Lander College, Flushing, New York, United States of America

Roles Formal analysis, Writing – original draft, Writing – review & editing

Affiliation Department of Health Policy & Management, Mailman School of Public Health, Columbia University, New York, New York, United States of America

Roles Conceptualization, Writing – review & editing

Affiliation Department of Clinical Psychology, Columbia University, New York, New York, United States of America

Roles Formal analysis

Affiliation Department of Computer Science, Stern College for Women, New York, New York, United States of America

Roles Conceptualization, Methodology, Writing – original draft, Writing – review & editing

Affiliation Department of Epidemiology, Mailman School of Public Health, Columbia University New York, New York, United States of America

• Leib Litman, 
• Jonathan Robinson, 
• Zohn Rosen, 
• Cheskie Rosenzweig, 
• Joshua Waxman, 
• Lisa M. Bates

• Published: February 21, 2020
• https://doi.org/10.1371/journal.pone.0229383
• Reader Comments

Studies of the gender pay gap are seldom able to simultaneously account for the range of alternative putative mechanisms underlying it. Using CloudResearch, an online microtask platform connecting employers to workers who perform research-related tasks, we examine whether gender pay discrepancies are still evident in a labor market characterized by anonymity, relatively homogeneous work, and flexibility. For 22,271 Mechanical Turk workers who participated in nearly 5 million tasks, we analyze hourly earnings by gender, controlling for key covariates which have been shown previously to lead to differential pay for men and women. On average, women’s hourly earnings were 10.5% lower than men’s. Several factors contributed to the gender pay gap, including the tendency for women to select tasks that have a lower advertised hourly pay. This study provides evidence that gender pay gaps can arise despite the absence of overt discrimination, labor segregation, and inflexible work arrangements, even after experience, education, and other human capital factors are controlled for. Findings highlight the need to examine other possible causes of the gender pay gap. Potential strategies for reducing the pay gap on online labor markets are also discussed.

Citation: Litman L, Robinson J, Rosen Z, Rosenzweig C, Waxman J, Bates LM (2020) The persistence of pay inequality: The gender pay gap in an anonymous online labor market. PLoS ONE 15(2): e0229383. https://doi.org/10.1371/journal.pone.0229383

Editor: Luís A. Nunes Amaral, Northwestern University, UNITED STATES

Received: March 5, 2019; Accepted: February 5, 2020; Published: February 21, 2020

Copyright: © 2020 Litman et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: Due to the sensitive nature of some of the data, and the terms of service of the websites used during data collection (including CloudResearch and MTurk), CloudResearch cannot release the full data set to make it publically available. The data are on CloudResearch's Sequel servers located at Queens College in the city of New York. CloudResearch makes data available to be accessed by researchers for replication purposes, on the CloudResearch premises, in the same way the data were accessed and analysed by the authors of this manuscript. The contact person at CloudResearch who can help researchers access the data set is Tzvi Abberbock, who can be reached at [email protected] .

Funding: The authors received no specific funding for this work.

Competing interests: We have read the journal's policy and the authors of this manuscript have the following potential competing interest: Several of the authors are employed at Cloud Research (previously TurkPrime), the database from which the data were queried. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Introduction

The gender pay gap, the disparity in earnings between male and female workers, has been the focus of empirical research in the US for decades, as well as legislative and executive action under the Obama administration [ 1 , 2 ]. Trends dating back to the 1960s show a long period in which women’s earnings were approximately 60% of their male counterparts, followed by increases in women’s earnings starting in the 1980s, which began to narrow, but not close, the gap which persists today [ 3 ]. More recent data from 2014 show that overall, the median weekly earnings of women working full time were 79–83% of what men earned [ 4 – 9 ].

The extensive literature seeking to explain the gender pay gap and its trajectory over time in traditional labor markets suggests it is a function of multiple structural and individual-level processes that reflect both the near-term and cumulative effects of gender relations and roles over the life course. Broadly speaking, the drivers of the gender pay gap can be categorized as: 1) human capital or productivity factors such as education, skills, and workforce experience; 2) industry or occupational segregation, which some estimates suggest accounts for approximately half of the pay gap; 3) gender-specific temporal flexibility constraints which can affect promotions and remuneration; and finally, 4) gender discrimination operating in hiring, promotion, task assignment, and/or compensation. The latter mechanism is often estimated by inference as a function of unexplained residual effects of gender on payment after accounting for other factors, an approach which is most persuasive in studies of narrowly restricted populations of workers such as lawyers [ 10 ] and academics of specific disciplines [ 11 ]. A recent estimate suggests this unexplained gender difference in earnings can account for approximately 40% of the pay gap [ 3 ]. However, more direct estimations of discriminatory processes are also available from experimental evidence, including field audit and lab-based studies [ 12 – 14 ]. Finally, gender pay gaps have also been attributed to differential discrimination encountered by men and women on the basis of parental status, often known as the ‘motherhood penalty’ [ 15 ].

Non-traditional ‘gig economy’ labor markets and the gender pay gap

In recent years there has been a dramatic rise in nontraditional ‘gig economy’ labor markets, which entail independent workers hired for single projects or tasks often on a short-term basis with minimal contractual engagement. “Microtask” platforms such as Amazon Mechanical Turk (MTurk) and Crowdflower have become a major sector of the gig economy, offering a source of easily accessible supplementary income through performance of small tasks online at a time and place convenient to the worker. Available tasks can range from categorizing receipts to transcription and proofreading services, and are posted online by the prospective employer. Workers registered with the platform then elect to perform the advertised tasks and receive compensation upon completion of satisfactory work [ 16 ]. An estimated 0.4% of US adults are currently receiving income from such platforms each month [ 17 ], and microtask work is a growing sector of the service economy in the United States [ 18 ]. Although still relatively small, these emerging labor market environments provide a unique opportunity to investigate the gender pay gap in ways not possible within traditional labor markets, due to features (described below) that allow researchers to simultaneously account for multiple putative mechanisms thought to underlie the pay gap.

The present study utilizes the Amazon Mechanical Turk (MTurk) platform as a case study to examine whether a gender pay gap remains evident when the main causes of the pay gap identified in the literature do not apply or can be accounted for in a single investigation. MTurk is an online microtask platform that connects employers (‘requesters’) to employees (‘workers’) who perform jobs called “Human Intelligence Tasks” (HITs). The platform allows requesters to post tasks on a dashboard with a short description of the HIT, the compensation being offered, and the time the HIT is expected to take. When complete, the requester either approves or rejects the work based on quality. If approved, payment is quickly accessible to workers. The gender of workers who complete these HITs is not known to the requesters, but was accessible to researchers for the present study (along with other sociodemographic information and pay rates) based on metadata collected through CloudResearch (formerly TurkPrime), a platform commonly used to conduct social and behavioral research on MTurk [ 19 ].

Evaluating pay rates of workers on MTurk requires estimating the pay per hour of each task that a worker accepts which can then be averaged together. All HITs posted on MTurk through CloudResearch display how much a HIT pays and an estimated time that it takes for that HIT to be completed. Workers use this information to determine what the corresponding hourly pay rate of a task is likely to be, and much of our analysis of the gender pay gap is based on this advertised pay rate of all completed surveys. We also calculate an estimate of the gender pay gap based on actual completion times to examine potential differences in task completion speed, which we refer to as estimated actual wages (see Methods section for details).

Previous studies have found that both task completion time and the selection of tasks influences the gender pay gap in at least some gig economy markets. For example, a gender pay gap was observed among Uber drivers, with men consistently earning higher pay than women [ 20 ]. Some of the contributing factors to this pay gap include that male Uber drivers selected different tasks than female drivers, including being more willing to work at night and to work in neighborhoods that were perceived to be more dangerous. Male drivers were also likely to drive faster than their female counterparts. These findings show that person-level factors like task selection, and speed can influence the gender pay gap within gig economy markets.

MTurk is uniquely suited to examine the gender pay gap because it is possible to account simultaneously for multiple structural and individual-level factors that have been shown to produce pay gaps. These include discrimination, work heterogeneity (leading to occupational segregation), and job flexibility, as well as human capital factors such as experience and education.

Discrimination.

When employers post their HITs on MTurk they have no way of knowing the demographic characteristics of the workers who accept those tasks, including their gender. While MTurk allows for selective recruitment of specific demographic groups, the MTurk tasks examined in this study are exclusively open to all workers, independent of their gender or other demographic characteristics. Therefore, features of the worker’s identity that might be the basis for discrimination cannot factor into an employer’s decision-making regarding hiring or pay.

Task heterogeneity.

Another factor making MTurk uniquely suited for the examination of the gender pay gap is the relative homogeneity of tasks performed by the workers, minimizing the potential influence of gender differences in the type of work pursued on earnings and the pay gap. Work on the MTurk platform consists mostly of short tasks such as 10–15 minute surveys and categorization tasks. In addition, the only information that workers have available to them to choose tasks, other than pay, is the tasks’ titles and descriptions. We additionally classified tasks based on similarity and accounted for possible task heterogeneity effects in our analyses.

Job flexibility.

MTurk is not characterized by the same inflexibilities as are often encountered in traditional labor markets. Workers can work at any time of the day or day of the week. This increased flexibility may be expected to provide more opportunities for participation in this labor market for those who are otherwise constrained by family or other obligations.

Human capital factors.

It is possible that the more experienced workers could learn over time how to identify higher paying tasks by virtue of, for example, identifying qualities of tasks that can be completed more quickly than the advertised required time estimate. Further, if experience is correlated with gender, it could contribute to a gender pay gap and thus needs to be controlled for. Using CloudResearch metadata, we are able to account for experience on the platform. Additionally, we account for multiple sociodemographic variables, including age, marital status, parental status, education, income (from all sources), and race using the sociodemographic data available through CloudResearch.

Expected gender pay gap findings on MTurk

Due to the aforementioned factors that are unique to the MTurk marketplace–e.g., anonymity, self-selection into tasks, relative homogeneity of the tasks performed, and flexible work scheduling–we did not expect a gender pay gap to be evident on the platform to the same extent as in traditional labor markets. However, potential gender differences in task selection and completion speed, which have implications for earnings, merit further consideration. For example, though we expect the relative homogeneity of the MTurk tasks to minimize gender differences in task selection that could mimic occupational segregation, we do account for potential subtle residual differences in tasks that could differentially attract male and female workers and indirectly lead to pay differentials if those tasks that are preferentially selected by men pay a higher rate. To do this we categorize all tasks based on their descriptions using K-clustering and add the clusters as covariates to our models. In addition, we separately examine the gender pay gap within each topic-cluster.

In addition, if workers who are experienced on the platform are better able to find higher paying HITs, and if experience is correlated with gender, it may lead to gender differences in earnings. Theoretically, other factors that may vary with gender could also influence task selection. Previous studies of the pay gap in traditional markets indicate that reservation wages, defined as the pay threshold at which a person is willing to accept work, may be lower among women with children compared to women without, and to that of men as well [ 21 ]. Thus, if women on MTurk are more likely to have young children than men, they may be more willing to accept available work even if it pays relatively poorly. Other factors such as income, education level, and age may similarly influence reservation wages if they are associated with opportunities to find work outside of microtask platforms. To the extent that these demographics correlate with gender they may give rise to a gender pay gap. Therefore we consider age, experience on MTurk, education, income, marital status, and parental status as covariates in our models.

Task completion speed may vary by gender for several reasons, including potential gender differences in past experience on the platform. We examine the estimated actual pay gap per hour based on HIT payment and estimated actual completion time to examine the effects of completion speed on the wage gap. We also examine the gender pay gap based on advertised pay rates, which are not dependent on completion speed and more directly measure how gender differences in task selection can lead to a pay gap. Below, we explain how these were calculated based on meta-data from CloudResearch.

To summarize, the overall goal of the present study was to explore whether gender pay differentials arise within a unique, non-traditional and anonymous online labor market, where known drivers of the gender pay gap either do not apply or can be accounted for statistically.

Materials and methods

Amazon mechanical turk and cloudresearch..

Started in 2005, the original purpose of the Amazon Mechanical Turk (MTurk) platform was to allow requesters to crowdsource tasks that could not easily be handled by existing technological solutions such as receipt copying, image categorization, and website testing. As of 2010, researchers increasingly began using MTurk for a wide variety of research tasks in the social, behavioral, and medical sciences, and it is currently used by thousands of academic researchers across hundreds of academic departments [ 22 ]. These research-related HITs are typically listed on the platform in generic terms such as, “Ten-minute social science study,” or “A study about public opinion attitudes.”

Because MTurk was not originally designed solely for research purposes, its interface is not optimized for some scientific applications. For this reason, third party add-on toolkits have been created that offer critical research tools for scientific use. One such platform, CloudResearch (formerly TurkPrime), allows requesters to manage multiple research functions, such as applying sampling criteria and facilitating longitudinal studies, through a link to their MTurk account. CloudResearch’s functionality has been described extensively elsewhere [ 19 ]. While the demographic characteristics of workers are not available to MTurk requesters, we were able to retroactively identify the gender and other demographic characteristics of workers through the CloudResearch platform. CloudResearch also facilitates access to data for each HIT, including pay, estimated length, and title.

The study was an analysis of previously collected metadata, which were analyzed anonymously. We complied with the terms of service for all data collected from CloudResearch, and MTurk. The approving institutional review board for this study was IntegReview.

Analytic sample.

We analyzed the nearly 5 million tasks completed during an 18-month period between January 2016 and June 2017 by 12,312 female and 9,959 male workers who had complete data on key demographic characteristics. To be included in the analysis a HIT had to be fully completed, not just accepted, by the worker, and had to be accepted (paid for) by the requester. Although the vast majority of HITs were open to both males and females, a small percentage of HITs are intended for a specific gender. Because our goal was to exclusively analyze HITs for which the requesters did not know the gender of workers, we excluded any HITs using gender-specific inclusion or exclusion criteria from the analyses. In addition, we removed from the analysis any HITs that were part of follow-up studies in which it would be possible for the requester to know the gender of the worker from the prior data collection. Finally, where possible, CloudResearch tracks demographic information on workers across multiple HITs over time. To minimize misclassification of gender, we excluded the 0.3% of assignments for which gender was unknown with at least 95% consistency across HITs.

The main exposure variable is worker gender and the outcome variables are estimated actual hourly pay accrued through completing HITs, and advertised hourly pay for completed HITs. Estimated actual hourly wages are based on the estimated length in minutes and compensation in dollars per HIT as posted on the dashboard by the requester. We refer to actual pay as estimated because sometimes people work multiple assignments at the same time (which is allowed on the platform), or may simultaneously perform other unrelated activities and therefore not work on the HIT the entire time the task is open. We also considered several covariates to approximate human capital factors that could potentially influence earnings on this platform, including marital status, education, household income, number of children, race/ethnicity, age, and experience (number of HITs previously completed). Additional covariates included task length, task cluster (see below), and the serial order with which workers accepted the HIT in order to account for potential differences in HIT acceptance speed that may relate to the pay gap.

Database and analytic approach.

Data were exported from CloudResearch’s database into Stata in long-form format to represent each task on a single row. For the purposes of this paper, we use “HIT” and “study” interchangeably to refer to a study put up on the MTurk dashboard which aims to collect data from multiple participants. A HIT or study consist of multiple “assignments” which is a single task completed by a single participant. Columns represented variables such as demographic information, payment, and estimated HIT length. Column variables also included unique IDs for workers, HITs (a single study posted by a requester), and requesters, allowing for a multi-level modeling analytic approach with assignments nested within workers. Individual assignments (a single task completed by a single worker) were the unit of analysis for all models.

Linear regression models were used to calculate the gender pay gap using two dependent variables 1) women’s estimated actual earnings relative to men’s and 2) women’s selection of tasks based on advertised earnings relative to men’s. We first examined the actual pay model, to see the gender pay gap when including an estimate of task completion speed, and then adjusted this model for advertised hourly pay to determine if and to what extent a propensity for men to select more remunerative tasks was evident and driving any observed gender pay gap. We additionally ran separate models using women’s advertised earnings relative to men’s as the dependent variable to examine task selection effects more directly. The fully adjusted models controlled for the human capital-related covariates, excluding household income and education which were balanced across genders. These models also tested for interactions between gender and each of the covariates by adding individual interaction terms to the adjusted model. To control for within-worker clustering, Huber-White standard error corrections were used in all models.

Cluster analysis.

To explore the potential influence of any residual task heterogeneity and gender preference for specific task type as the cause of the gender pay gap, we use K-means clustering analysis (seed = 0) to categorize the types of tasks into clusters based on the descriptions that workers use to choose the tasks they perform. We excluded from this clustering any tasks which contained certain gendered words (such as “male”, “female”, etc.) and any tasks which had fewer than 30 respondents. We stripped out all punctuation, symbols and digits from the titles, so as to remove any reference to estimated compensation or duration. The features we clustered on were the presence or absence of 5,140 distinct words that appeared across all titles. We then present the distribution of tasks across these clusters as well as average pay by gender and the gender pay gap within each cluster.

The demographics of the analytic sample are presented in Table 1 . Men and women completed comparable numbers of tasks during the study period; 2,396,978 (48.6%) for men and 2,539,229 (51.4%) for women.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pone.0229383.t001

In Table 2 we measure the differences in remuneration between genders, and then decompose any observed pay gap into task completion speed, task selection, and then demographic and structural factors. Model 1 shows the unadjusted regression model of gender differences in estimated actual pay, and indicates that, on average, tasks completed by women paid 60 (10.5%) cents less per hour compared to tasks completed by men (t = 17.4, p < .0001), with the mean estimated actual pay across genders being $5.70 per hour.

https://doi.org/10.1371/journal.pone.0229383.t002

In Model 2, adjusting for advertised hourly pay, the gender pay gap dropped to 46 cents indicating that 14 cents of the pay gap is attributable to gender differences in the selection of tasks (t = 8.6, p < .0001). Finally, after the inclusion of covariates and their interactions in Model 3, the gender pay differential was further attenuated to 32 cents (t = 6.7, p < .0001). The remaining 32 cent difference (56.6%) in earnings is inferred to be attributable to gender differences in HIT completion speed.

Task selection analyses

Although completion speed appears to account for a significant portion of the pay gap, of particular interest are gender differences in task selection. Beyond structural factors such as education, household composition and completion speed, task selection accounts for a meaningful portion of the gender pay gap. As a reminder, the pay rate and expected completion time are posted for every HIT, so why women would select less remunerative tasks on average than men do is an important question to explore. In the next section of the paper we perform a set of analyses to examine factors that could account for this observed gender difference in task selection.

Advertised hourly pay.

To examine gender differences in task selection, we used linear regression to directly examine whether the advertised hourly pay differed for tasks accepted by male and female workers. We first ran a simple model ( Table 3 ; Model 3A) on the full dataset of 4.93 million HITs, with gender as the predictor and advertised hourly pay as the outcome including no other covariates. The unadjusted regression results (Model 4) shown in Table 3 , indicates that, summed across all clusters and demographic groups, tasks completed by women were advertised as paying 28 cents (95% CI: $0.25-$0.31) less per hour (5.8%) compared to tasks completed by men (t = 21.8, p < .0001).

https://doi.org/10.1371/journal.pone.0229383.t003

Model 5 examines whether the remuneration differences for tasks selected by men and women remains significant in the presence of multiple covariates included in the previous model and their interactions. The advertised pay differential for tasks selected by women compared to men was attenuated to 21 cents (4.3%), and remained statistically significant (t = 9.9, p < .0001). This estimate closely corresponded to the inferred influence of task selection reported in Table 2 . Tests of gender by covariate interactions were significant only in the cases of age and marital status; the pay differential in tasks selected by men and women decreased with age and was more pronounced among single versus currently or previously married women.

To further examine what factors may account for the observed gender differences in task selection we plotted the observed pay gap within demographic and other covariate groups. Table 4 shows the distribution of tasks completed by men and women, as well as mean earnings and the pay gap across all demographic groups, based on the advertised (not actual) hourly pay for HITs selected (hereafter referred to as “advertised hourly pay” and the “advertised pay gap”). The average task was advertised to pay $4.88 per hour (95% CI $4.69, $5.10).

https://doi.org/10.1371/journal.pone.0229383.t004

The pattern across demographic characteristics shows that the advertised hourly pay gap between genders is pervasive. Notably, a significant advertised gender pay gap is evident in every level of each covariate considered in Table 4 , but more pronounced among some subgroups of workers. For example, the advertised pay gap was highest among the youngest workers ($0.31 per hour for workers age 18–29), and decreased linearly with age, declining to $0.13 per hour among workers age 60+. Advertised houry gender pay gaps were evident across all levels of education and income considered.

To further examine the potential influence of human capital factors on the advertised hourly pay gap, Table 5 presents the average advertised pay for selected tasks by level of experience on the CloudResearch platform. Workers were grouped into 4 experience levels, based on the number of prior HITs completed: Those who completed fewer than 100 HITs, between 100 and 500 HITs, between 500 and 1,000 HITs, and more than 1,000 HITs. A significant gender difference in advertised hourly pay was observed within each of these four experience groups. The advertised hourly pay for tasks selected by both male and female workers increased with experience, while the gender pay gap decreases. There was some evidence that male workers have more cumulative experience with the platform: 43% of male workers had the highest level of experience (previously completing 1,001–10,000 HITs) compared to only 33% of women.

https://doi.org/10.1371/journal.pone.0229383.t005

Table 5 also explores the influence of task heterogeneity upon HIT selection and the gender gap in advertised hourly pay. K-means clustering was used to group HITs into 20 clusters initially based on the presence or absence of 5,140 distinct words appearing in HIT titles. Clusters with fewer than 50,000 completed tasks were then excluded from analysis. This resulted in 13 clusters which accounted for 94.3% of submitted work assignments (HITs).

The themes of all clusters as well as the average hourly advertised pay for men and women within each cluster are presented in the second panel of Table 5 . The clusters included categories such as Games, Decision making, Product evaluation, Psychology studies, and Short Surveys. We did not observe a gender preference for any of the clusters. Specifically, for every cluster, the proportion of males was no smaller than 46.6% (consistent with the slightly lower proportion of males on the platform, see Table 1 ) and no larger than 50.2%. As shown in Table 5 , the gender pay gap was observed within each of the clusters. These results suggest that residual task heterogeneity, a proxy for occupational segregation, is not likely to contribute to a gender pay gap in this market.

Task length was defined as the advertised estimated duration of a HIT. Table 6 presents the advertised hourly gender pay gaps for five categories of HIT length, which ranged from a few minutes to over 1 hour. Again, a significant advertised hourly gender pay gap was observed in each category.

https://doi.org/10.1371/journal.pone.0229383.t006

Finally, we conducted additional supplementary analyses to determine if other plausible factors such as HIT timing could account for the gender pay gap. We explored temporal factors including hour of the day and day of the week. Each completed task was grouped based on the hour and day in which it was completed. A significant advertised gender pay gap was observed within each of the 24 hours of the day and for every day of the week demonstrating that HIT timing could not account for the observed gender gap (results available in Supplementary Materials).

In this study we examined the gender pay gap on an anonymous online platform across an 18-month period, during which close to five million tasks were completed by over 20,000 unique workers. Due to factors that are unique to the Mechanical Turk online marketplace–such as anonymity, self-selection into tasks, relative homogeneity of the tasks performed, and flexible work scheduling–we did not expect earnings to differ by gender on this platform. However, contrary to our expectations, a robust and persistent gender pay gap was observed.

The average estimated actual pay on MTurk over the course of the examined time period was $5.70 per hour, with the gender pay differential being 10.5%. Importantly, gig economy platforms differ from more traditional labor markets in that hourly pay largely depends on the speed with which tasks are completed. For this reason, an analysis of gender differences in actual earned pay will be affected by gender differences in task completion speed. Unfortunately, we were not able to directly measure the speed with which workers complete tasks and account for this factor in our analysis. This is because workers have the ability to accept multiple HITs at the same time and multiple HITs can sit dormant in a queue, waiting for workers to begin to work on them. Therefore, the actual time that many workers spend working on tasks is likely less than what is indicated in the metadata available. For this reason, the estimated average actual hourly rate of $5.70 is likely an underestimate and the gender gap in actual pay cannot be precisely measured. We infer however, by the residual gender pay gap after accounting for other factors, that as much as 57% (or $.32) of the pay differential may be attributable to task completion speed. There are multiple plausible explanations for gender differences in task completion speed. For example, women may be more meticulous at performing tasks and, thus, may take longer at completing them. There may also be a skill factor related to men’s greater experience on the platform (see Table 5 ), such that men may be faster on average at completing tasks than women.

However, our findings also revealed another component of a gender pay gap on this platform–gender differences in the selection of tasks based on their advertised pay. Because the speed with which workers complete tasks does not impact these estimates, we conducted extensive analyses to try to explain this gender gap and the reasons why women appear on average to be selecting tasks that pay less compared to men. These results pertaining to the advertised gender pay gap constitute the main focus of this study and the discussion that follows.

The overall advertised hourly pay was $4.88. The gender pay gap in the advertised hourly pay was $0.28, or 5.8% of the advertised pay. Once a gender earnings differential was observed based on advertised pay, we expected to fully explain it by controlling for key structural and individual-level covariates. The covariates that we examined included experience, age, income, education, family composition, race, number of children, task length, the speed of accepting a task, and thirteen types of subtasks. We additionally examined the time of day and day of the week as potential explanatory factors. Again, contrary to our expectations, we observed that the pay gap persisted even after these potential confounders were controlled for. Indeed, separate analyses that examined the advertised pay gap within each subcategory of the covariates showed that the pay gap is ubiquitous, and persisted within each of the ninety sub-groups examined. These findings allows us to rule out multiple mechanisms that are known drivers of the pay gap in traditional labor markets and other gig economy marketplaces. To our knowledge this is the only study that has observed a pay gap across such diverse categories of workers and conditions, in an anonymous marketplace, while simultaneously controlling for virtually all variables that are traditionally implicated as causes of the gender pay gap.

Individual-level factors

Individual-level factors such as parental status and family composition are a common source of the gender pay gap in traditional labor markets [ 15 ] . Single mothers have previously been shown to have lower reservation wages compared to other men and women [ 21 ]. In traditional labor markets lower reservation wages lead single mothers to be willing to accept lower-paying work, contributing to a larger gender pay gap in this group. This pattern may extend to gig economy markets, in which single mothers may look to online labor markets as a source of supplementary income to help take care of their children, potentially leading them to become less discriminating in their choice of tasks and more willing to work for lower pay. Since female MTurk workers are 20% more likely than men to have children (see Table 1 ), it was critical to examine whether the gender pay gap may be driven by factors associated with family composition.

An examination of the advertised gender pay gap among individuals who differed in their marital and parental status showed that while married workers and those with children are indeed willing to work for lower pay (suggesting that family circumstances do affect reservation wages and may thus affect the willingness of online workers to accept lower-paying online tasks), women’s hourly pay is consistently lower than men’s within both single and married subgroups of workers, and among workers who do and do not have children. Indeed, contrary to expectations, the advertised gender pay gap was highest among those workers who are single, and among those who do not have any children. This observation shows that it is not possible for parental and family status to account for the observed pay gap in the present study, since it is precisely among unmarried individuals and those without children that the largest pay gap is observed.

Age was another factor that we considered to potentially explain the gender pay gap. In the present sample, the hourly pay of older individuals is substantially lower than that of younger workers; and women on the platform are five years older on average compared to men (see Table 1 ). However, having examined the gender pay gap separately within five different age cohorts we found that the largest pay gap occurs in the two youngest cohort groups: those between 18 and 29, and between 30 and 39 years of age. These are also the largest cohorts, responsible for 64% of completed work in total.

Younger workers are also most likely to have never been married or to not have any children. Thus, taken together, the results of the subgroup analyses are consistent in showing that the largest pay gap does not emerge from factors relating to parental, family, or age-related person-level factors. Similar patterns were found for race, education, and income. Specifically, a significant gender pay gap was observed within each subgroup of every one of these variables, showing that person-level factors relating to demographics are not driving the pay gap on this platform.

Experience is a factor that has an influence on the pay gap in both traditional and gig economy labor markets [ 20 ] . As noted above, experienced workers may be faster and more efficient at completing tasks in this platform, but also potentially more savvy at selecting more remunerative tasks compared to less experienced workers if, for example, they are better at selecting tasks that will take less time to complete than estimated on the dashboard [ 20 ]. On MTurk, men are overall more experienced than women. However, experience does not account for the gender gap in advertised pay in the present study. Inexperienced workers comprise the vast majority of the Mechanical Turk workforce, accounting for 67% of all completed tasks (see Table 5 ). Yet within this inexperienced group, there is a consistent male earning advantage based on the advertised pay for tasks performed. Further, controlling for the effect of experience in our models has a minimal effect on attenuating the gender pay gap.

Task heterogeneity

Another important source of the gender pay gap in both traditional and gig economy labor markets is task heterogeneity. In traditional labor markets men are disproportionately represented in lucrative fields, such as those in the tech sector [ 23 ]. While the workspace within MTurk is relatively homogeneous compared to the traditional labor market, there is still some variety in the kinds of tasks that are available, and men and women may have been expected to have preferences that influence choices among these.

To examine whether there is a gender preference for specific tasks, we systematically analyzed the textual descriptions of all tasks included in this study. These textual descriptions were available for all workers to examine on their dashboards, along with information about pay. The clustering algorithm revealed thirteen categories of tasks such as games, decision making, several different kinds of survey tasks, and psychology studies.We did not observe any evidence of gender preference for any of the task types. Within each of the thirteen clusters the distribution of tasks was approximately equally split between men and women. Thus, there is no evidence that women as a group have an overall preference for specific tasks compared to men. Critically, the gender pay gap was also observed within each one of these thirteen clusters.

Another potential source of heterogeneity is task length. Based on traditional labor markets, one plausible hypothesis about what may drive women’s preferences for specific tasks is that women may select tasks that differ in their duration. For example, women may be more likely to use the platform for supplemental income, while men may be more likely to work on HITs as their primary income source. Women may thus select shorter tasks relative to their male counterparts. If the shorter tasks pay less money, this would result in what appears to be a gender pay gap.

However, we did not observe gender differences in task selection based on task duration. For example, having divided tasks into their advertised length, the tasks are preferred equally by men and women. Furthermore, the shorter tasks’ hourly pay is substantially higher on average compared to longer tasks.

Additional evidence that scheduling factors do not drive the gender pay gap is that it was observed within all hourly and daily intervals (See S1 and S2 Tables in Appendix). These data are consistent with the results presented above regarding personal level factors, showing that the majority of male and female Mechanical Turk workers are single, young, and have no children. Thus, while in traditional labor markets task heterogeneity and labor segmentation is often driven by family and other life circumstances, the cohort examined in this study does not appear to be affected by these factors.

Practical implications of a gender pay gap on online platforms for social and behavioral science research

The present findings have important implications for online participant recruitment in the social and behavioral sciences, and also have theoretical implications for understanding the mechanisms that give rise to the gender pay gap. The last ten years have seen a revolution in data collection practices in the social and behavioral sciences, as laboratory-based data collection has slowly and steadily been moving online [ 16 , 24 ]. Mechanical Turk is by far the most widely used source of human participants online, with thousands of published peer-reviewed papers utilizing Mechanical Turk to recruit at least some of their human participants [ 25 ]. The present findings suggest both a challenge and an opportunity for researchers utilizing online platforms for participant recruitment. Our findings clearly reveal for the first time that sampling research participants on anonymous online platforms tends to produce gender pay inequities, and that this happens independent of demographics or type of task. While it is not clear from our findings what the exact cause of this inequity is, what is clear is that the online sampling environment produces similar gender pay inequities as those observed in other more traditional labor markets, after controlling for relevant covariates.

This finding is inherently surprising since many mechanisms that are known to produce the gender pay gap in traditional labor markets are not at play in online microtasks environments. Regardless of what the generative mechanisms of the gender pay gap on online microtask platforms might be, researchers may wish to consider whether changes in their sampling practices may produce more equitable pay outcomes. Unlike traditional labor markets, online data collection platforms have built-in tools that can allow researchers to easily fix gender pay inequities. Researchers can simply utilize gender quotas, for example, to fix the ratio of male and female participants that they recruit. These simple fixes in sampling practices will not only produce more equitable pay outcomes but are also most likely advantageous for reducing sampling bias due to gender being correlated with pay. Thus, while our results point to a ubiquitous discrepancy in pay between men and women on online microtask platforms, such inequities have relatively easy fixes on online gig economy marketplaces such as MTurk, compared to traditional labor markets where gender-based pay inequities have often remained intractable.

Other gig economy markets

As discussed in the introduction, a gender wage gap has been demonstrated on Uber, a gig economy transportation marketplace [ 20 ], where men earn approximately 7% more than women. However, unlike in the present study, the gender wage gap on Uber was fully explained by three factors; a) driving speed predicted higher wages, with men driving faster than women, b) men were more likely than women to drive in congested locations which resulted in better pay, c) experience working for Uber predicted higher wages, with men being more experienced. Thus, contrary to our findings, the gender wage gap in gig economy markets studied thus far are fully explained by task heterogeneity, experience, and task completion speed. To our knowledge, the results presented in the present study are the first to show that the gender wage gap can emerge independent of these factors.

Generalizability

Every labor market is characterized by a unique population of workers that are almost by definition not a representation of the general population outside of that labor market. Likewise, Mechanical Turk is characterized by a unique population of workers that is known to differ from the general population in several ways. Mechanical Turk workers are younger, better educated, less likely to be married or have children, less likely to be religious, and more likely to have a lower income compared to the general United States population [ 24 ]. The goal of the present study was not to uncover universal mechanisms that generate the gender pay gap across all labor markets and demographic groups. Rather, the goal was to examine a highly unique labor environment, characterized by factors that should make this labor market immune to the emergence of a gender pay gap.

Previous theories accounting for the pay gap have identified specific generating mechanisms relating to structural and personal factors, in addition to discrimination, as playing a role in the emergence of the gender pay gap. This study examined the work of over 20,000 individuals completing over 5 million tasks, under conditions where standard mechanisms that generate the gender pay gap have been controlled for. Nevertheless, a gender pay gap emerged in this environment, which cannot be accounted for by structural factors, demographic background, task preferences, or discrimination. Thus, these results reveal that the gender pay gap can emerge—in at least some labor markets—in which discrimination is absent and other key factors are accounted for. These results show that factors which have been identified to date as giving rise to the gender pay gap are not sufficient to explain the pay gap in at least some labor markets.

Potential mechanisms

While we cannot know from the results of this study what the actual mechanism is that generates the gender pay gap on online platforms, we suggest that it may be coming from outside of the platform. The particular characteristics of this labor market—such as anonymity, relative task homogeneity, and flexibility—suggest that, everything else being equal, women working in this platform have a greater propensity to choose less remunerative opportunities relative to men. It may be that these choices are driven by women having a lower reservation wage compared to men [ 21 , 26 ]. Previous research among student populations and in traditional labor markets has shown that women report lower pay or reward expectations than men [ 27 – 29 ]. Lower pay expectations among women are attributed to justifiable anticipation of differential returns to labor due to factors such as gender discrimination and/or a systematic psychological bias toward pessimism relative to an overly optimistic propensity among men [ 30 ].

Our results show that even if the bias of employers is removed by hiding the gender of workers as happens on MTurk, it seems that women may select lower paying opportunities themselves because their lower reservation wage influences the types of tasks they are willing to work on. It may be that women do this because cumulative experiences of pervasive discrimination lead women to undervalue their labor. In turn, women’s experiences with earning lower pay compared to men on traditional labor markets may lower women’s pay expectations on gig economy markets. Thus, consistent with these lowered expectations, women lower their reservation wages and may thus be more likely than men to settle for lower paying tasks.

More broadly, gender norms, psychological attributes, and non-cognitive skills, have recently become the subject of investigation as a potential source for the gender pay gap [ 3 ], and the present findings indicate the importance of such mechanisms being further explored, particularly in the context of task selection. More research will be required to explore the potential psychological and antecedent structural mechanisms underlying differential task selection and expectations of compensation for time spent on microtask platforms, with potential relevance to the gender pay gap in traditional labor markets as well. What these results do show is that pay discrepancies can emerge despite the absence of discrimination in at least some circumstances. These results should be of particular interest for researchers who may wish to see a more equitable online labor market for academic research, and also suggest that novel and heretofore unexplored mechanisms may be at play in generating these pay discrepancies.

A final note about framing: we are aware that explanations of the gender pay gap that invoke elements of women’s agency and, more specifically, “choices” risk both; a) diminishing or distracting from important structural factors, and b) “naturalizing” the status quo of gender inequality [ 30 ] . As Connor and Fiske (2019) argue, causal attributions for the gender pay gap to “unconstrained choices” by women, common as part of human capital explanations, may have the effect, intended or otherwise, of reinforcing system-justifying ideologies that serve to perpetuate inequality. By explicitly locating women’s economic decision making on the MTurk platform in the broader context of inegalitarian gender norms and labor market experiences outside of it (as above), we seek to distance our interpretation of our findings from implicit endorsement of traditional gender roles and economic arrangements and to promote further investigation of how the observed gender pay gap in this niche of the gig economy may reflect both broader gender inequalities and opportunities for structural remedies.

Supporting information

S1 table. distribution of hits, average pays, and gender pay gaps by hour of day..

https://doi.org/10.1371/journal.pone.0229383.s001

S2 Table. Distribution of HITs, average pays, and gender pay gaps by day of the week.

https://doi.org/10.1371/journal.pone.0229383.s002

• 1. United States Equal Employment Opportunity Commission, Lily Ledbetter Fair Pay Act of 2009 (2009), available at https://www.eeoc.gov/eeoc/publications/brochure- equal_pay_and_ledbetter_act.cfm, accessed on 11/12/2018.
• 2. United States Department of Labor (DOL), Office of Federal Contract Compliance Programs (OFCCP), Pay Transparency Nondiscrimination Provision, available at https://www.dol.gov/ofccp/PayTransparencyNondiscrimination.html , accessed on 11/12/2018.
• View Article
• Google Scholar
• 4. United States Department of Labor (DOL), Bureau of Labor Statistics (BLS) (2016) Women’s earning 83 percent of men’s, but vary by occupation. TED Econ Dly , available at https://www.bls.gov/opub/ted/2016/womens-earnings-83-percent-of-mens-but-vary-by-occupation.htm , accessed on 11/12/2018.
• 5. Davis A (2015) Women still earn less than men across the board (Economic Policy Institute, 2015), available at http://www.epi.org/publication/women-still-earn-less-than-men-across-the-board/ , accessed on 11/12/2018.
• 6. “Gender Pay Inequality: Consequences for Women, Families and the Economy” (Joint Economic Committee, 2016). [no author]
• 7. Hartmann H, Hayes J, Clark J (2014) “How Equal Pay for Working Women would Reduce Poverty and Grow the American Economy” (Institute for Women’s Policy Research, 2014).
• 8. OECD (2015) In it together: Why Less Inequality Benefits All (OECD Publishing, Paris) available at http://www.oecd.org/els/soc/OECD2015-In-It-Together-Chapter1-Overview-Inequality.pdf , accessed on 11/12/2018.
• PubMed/NCBI
• 16. Litman L, Robinson J (In Press) Conducting Online Research on Amazon Mechanical Turk and Beyond. Sage Publications.
• 17. Farrell D, Greig F (2016) Paychecks, paydays, and the online platform economy: Big data on income volatility. JP Morgan Chase Institute.
• 18. Kuek SC, Paradi-Guilford C, Fayomi T, Imaizumi S, Ipeirotis P, Pina P, Singh M (2015) The global opportunity in online outsourcing (World Bank Group, 2015) Available at http://documents.worldbank.org/curated/en/138371468000900555/pdf/ACS14228-ESW-white-cover-P149016-Box391478B-PUBLIC-World-Bank-Global-OO-Study-WB-Rpt-FinalS.pdf , accessed on 11/12/2018.
• 23. Bureau of Labor Statistics, U.S. Department of Labor, Labor Force Statistics from the Current Population Survey, Household Data Annual Averages. Employed persons by detailed occupation, sex, race, and Hispanic or Latino ethnicity, on the Internet at https://www.bls.gov/cps/cpsaat11.htm (visited 9/3/18).

Advertisement

The gender pay gap in the USA: a matching study

• Original Paper
• Open access
• Published: 05 September 2019
• Volume 33 , pages 271–305, ( 2020 )

Cite this article

You have full access to this open access article

• Katie Meara 1 ,
• Francesco Pastore 2 &
• Allan Webster 1  

48k Accesses

58 Citations

38 Altmetric

Explore all metrics

This study examines the gender wage gap in the USA using two separate cross-sections from the Current Population Survey (CPS). The extensive literature on this subject includes wage decompositions that divide the gender wage gap into “explained” and “unexplained” components. One of the problems with this approach is the heterogeneity of the sample data. In order to address the difficulties of comparing like with like, this study uses a number of different matching techniques to obtain estimates of the gap. By controlling for a wide range of other influences, in effect, we estimate the direct effect of simply being female on wages. However, a number of other factors, such as parenthood, gender segregation, part-time working, and unionization, contribute to the gender wage gap. This means that it is not just the core “like for like” comparison between male and female wages that matters but also how gender wage differences interact with other influences. The literature has noted the existence of these interactions, but precise or systematic estimates of such effects remain scarce. The most innovative contribution of this study is to do that. Our findings imply that the idea of a single uniform gender pay gap is perhaps less useful than an understanding of how gender wages are shaped by multiple different forces.

Similar content being viewed by others

Gender wage gaps in Ghana: a comparison across different selection models

Gender-Specific Wage Structure and the Gender Wage Gap in the U.S. Labor Market

Analysing the glass ceiling and sticky floor effects in Bangladesh: evidence, extent and elements

Avoid common mistakes on your manuscript.

1 Introduction

This study estimates the gender pay gap in the USA using several different matching estimators. We first justify the use of matching estimators by using an Oaxaca recentered influence function (RIF) model to estimate the gender pay gap. Other authors using a similar approach have found the “unexplained” component of the gender pay gap to be high. Some of these, including Kassenboehmer and Sinning ( 2014 ) and Töpfer ( 2017 ), attribute this to heterogeneity within their sample. A similar analysis in this study also finds a high “unexplained” component, which implies a heterogeneity problem.

Where heterogeneity is an issue, a well-established approach is to use a matching estimator—see, for example, Ñopo ( 2008 ). This study therefore relies on several matching estimators for its core analysis. These are discussed from the methodological perspective later, but matching involves a number of conceptual issues which are central to the approach of this study. A matching approach creates a control group (of males) which, as far as possible, matches the treated group (female) in all relevant characteristics. For the estimator not to be biased, relevant characteristics such as part-time working and union membership must be included as covariates. The result is an estimate of the gap between male and female pay that controls for all relevant observable characteristics, including unionization and part-time work. Estimating a pure “gender” effect on wages is one of the advantages of using a matching estimator, but the process of creating a control group omits other more indirect ways by which women are paid less.

For example, working part-time typically involves a substantially lower hourly rate of pay than working full-time, as this study confirms. A much higher proportion of females work part-time than do males. Likewise, unionized workers exhibit significantly higher hourly pay than non-unionized workers, and females are much less likely to be unionized than males. A matching approach is intended to capture the effect on wages of being female and needs to control for overlapping effects like part-time work or union membership. Methodologically this is sound, but it must be properly understood that there is more to the matter. In terms of hourly pay, females are also disadvantaged by, say, working part-time and being less likely to be unionized. It is proper to ignore such effects in a matching estimate of the pure “gender effect,” but this study emphasizes that such estimates do not capture the full extent of the wage disadvantages faced by females.

The main focus of this study is, within a matching framework, to examine the important interactions between gender and other relevant characteristics. Union membership and part-time work are two of these. The study also considers the effects of parenthood, age, and gender segregation. An important part of the approach taken is the inverse probability weighted regression adjustment (IPWRA) matching estimator. There are important statistical advantages from using an IPWRA estimator (mainly its “double robustness” property), but the key reason for using IPWRA is behavioral more than statistical. The IPWRA estimator can work with two treatment effects and hence estimate the effects of interactions between gender and another variable. For example, consider female and part-time as treatment variables. The IPWRA approach can simultaneously give the following treatment effects on hourly wages: (a) being female, (b) working part time, and (c) both being female and working part time (an interaction effect).

The conceptual relevance of these interactions is not new in the literature, as Blau and Kahn ( 2017 ) make clear, but such interaction effects have not previously been formally estimated in a consistent manner, if at all. The contribution of the paper is to provide clear evidence that a basic matching estimate of the gender pay gap is useful but does not tell the whole story. An analysis which includes not just a “gender only” effect on wages but also interactions between this gender effect and other key covariates (such as part-time work) is a much richer one. This is the main contribution of the study.

Section 2 provides a review of the literature. The data used by the study, which are two samples taken from the US Current Population Survey (CPS) for the period October 2011 to March 2012 and for the period October 2017 to March 2018, are described in Section 3 , and the methodological approach is described in Section 4 . The matching analysis with a single treatment effect is presented in Section 5 and the IPWRA analysis in Section 6 . Section 7 presents the conclusions of the study.

2 Review of literature

Blau and Kahn ( 2017 ) present a comprehensive review of what is now an extensive literature on the gender pay gap in the USA. A number of themes arising in this literature are developed further in this paper. Blau and Kahn ( 2017 ) present detailed empirical evidence to show that some of the core issues have changed since the 1970s. Several of these are of particular relevance for this paper. Firstly, the gender wage gap has fallen dramatically but still remains sizeable. This is perhaps surprising given that the gap in education has been reversed in favor of women. They find that the gender wage gap has fallen from about 36–38% in 1970 to between 18 and 21% in 2010. The analysis presented in this study does not consider long-term changes but does confirm that a substantial wage gap remains.

In their meta-analysis of a total of 263 papers, Weichselbaumer and Winter-Ebmer ( 2005 ) also find evidence of a global reduction of the gender wage gap. At the same time that the gender wage gap was narrowing, the human capital factors used to explain the gap (education and actual work experience) were either moving in favor of women or strongly declining. Beaudry and Lewis ( 2014 ) associate the declining gender wage gap in the USA with changes in the price of skills, related to skill-biased technical change. In another US study, Borghans et al. ( 2014 ) find the decline in the gender wage gap to be associated with a growth in the importance of people skills. In a rare natural experiment, Flory et al. ( 2014 ) link the gap in gender wages to female aversion to competitive work environments.

Blau and Kahn ( 2017 ) report that the gender gap in years of education has reversed from − 0.2 to + 0.2 between 1981 and 2011 for the USA. The gap in years of work experience fell from 7 in 1981 to 1.4 years in 2011. In consequence, the role of these traditional factors in the gender wage gap has shrunk. Together, education and work experience explained about 27% of the gap in 1981 but only around 8% in 2010. A number of other explanatory factors have also reduced in significance, such as the effect of unionization on male wages. Despite this decline, the evidence presented in this study shows that unionization still plays a part in gender wage differences. Blau and Kahn ( 2017 ) show that, in contrast, some other factors have become increasingly important. For example, they find that gender segregation by occupation and industry has become of much greater consequence—accounting for only about 27% of the gap in 1980 but about 49% in 2010. The role of gender segregation is another theme which this study seeks to develop further.

The link between gender segregation and the gender wage gap has long since been made. Polachek ( 1981 ) constructs a model in which female earnings potential depreciates during temporary exits from the labor force while males remaining in the labor force see their earnings potential appreciate from continued skill development. The expectation of interruptions to work experience affects female investment in skills and, hence, occupational choice. Maternity drives women to self-segregate into jobs which are less innovative and less skill driven—occupations that tend to be paid less. Cobb-Clark and Moschion ( 2017 ) provide evidence from Australia that gender differences in educational performance exist at an early stage and vary according to socio-economic status.

A number of studies have tried to assess the extent of occupational segregation in the USA and elsewhere by means of the Duncan and Duncan ( 1955 ) segregation index. Blau and Kahn ( 2013a , b ) find that the segregation index fell from 64.5 in 1970 to 51.0 in 2009. The decline was more rapid in the 1970s than in the 1980s and even more gradual in the following years. As Blau and Kahn ( 2017 ) note, even the diminished value of 51% still represents a high degree of occupational segregation. Unsurprisingly (given the known role that segregation has in explaining the gender wage gap), the high value of the segregation index relative to 2009 confirms that occupational and industry differences by gender still remain sizeable. This study also reports gender segregation indices for the USA with similar findings.

Hegewisch et al. ( 2010 ) find similar evidence of a declining degree of segregation in the USA. Moreover, they link gender segregation to the gender wage gap, finding a negative relationship between the share of women in employment in an occupation and the gender wage gap. Tomaskovic-Devey and Skaggs ( 2002 ) also link gender segregation to the gender wage gap, finding further evidence of the role of industries as a source of wage inequality. Levanon et al. ( 2009 ) consider the view that gender segregation and the gender wage gap are causally related by two sociological processes—devaluation and queuing—using US Census data. Their analysis found some evidence of devaluation (valuing the work of females less) but little evidence of queuing (employers preferring to hire males).

Other studies drew similar conclusions to the USA for other countries. For instance, Barón and Cobb-Clark ( 2010 ) find an important effect of occupational segregation on the gender wage gap in Australia. They find the gender wage gap to be fully explained by productivity characteristics but not fully explained for high-wage workers. Olsen and Walby ( 2004 ) find evidence from the UK that labor market rigidities—including the segregation of women into certain occupations and into smaller, non-unionized firms—were responsible for about 36% of the gender wage gap. Walby and Olsen ( 2002 ) also find both occupational and industrial segregation to have been prevalent in the UK. Livanos and Pouliakas ( 2012 ), in a study of Greece, find that gender segregation with respect to educational subject explained part of the gender wage gap. Pastore and Verashchagina ( 2011 ) find that the gender wage gap more than doubled during the transition from plan to market in Belarus, particularly because women have experienced increasing segregation in low-wage industries.

Polachek ( 1985 ) further extends this link between gender wages and a life cycle view of occupational choice. Polachek ( 2014 ) finds the gender pay gap to be smaller between single men and women and larger between married men and women. This is attributable to his life cycle model of human capital and the resulting different occupational structures between the genders. To the extent that educational choices by women are related to eventual occupational choices, the study of Danish labor markets by Humlum et al. ( 2019 ) suggests that these may also be affected by parental attitudes to labor markets. The role of maternity and aging on female earnings is confirmed by a comparatively recent strand of the literature which focuses on the labor market behavior of young people to try to ascertain at which stage the gender pay gap first arises. Many studies have found little or no gender wage gap among young people. A gap emerges after maternity and widens as workers age. Manning and Swaffield ( 2008 ) provide an early study of this type for the UK. In a study of US MBAs, Bertrand et al. ( 2010 ) attribute a growing gender wage gap that increased with age to career interruptions as well as differences in training and weekly hours of work. More recently, similar findings have been noted for several developing countries—see, for example, Pastore ( 2010 ) and Pastore et al. ( 2016 ). This study provides recent evidence for the USA which confirms the existence of much narrower differences in gender wages for younger than older workers.

Some research has been aimed at locating the gap along the earning distribution to understand whether it is generalized or whether it is attributable to particular groups of individuals with specific skill levels. Blau and Kahn ( 1997 ) find increased demand for highly skilled workers to have widened the gender wage gap. In their study covering 11 countries, Arulampalam et al. ( 2007 ) find evidence of a tendency for the gender pay gap to be concentrated mainly among the low-skill (so-called sticky floor effect) and the high-skill (so-called glass ceiling effect) workers. Examples of the latter include managerial positions, particularly senior management, and many highly paid liberal professions (Goldin, 2014 ). In these types of jobs, not only education and human capital are of importance but also relationships of trust with customers. This makes the role of some individuals hard to substitute and, in consequence, requires flexibility with respect to hours of work—conditions that are often not easily met by women. Olivetti ( 2006 ) provides a new measure of the returns to work experience, using PSID data for the USA. Her analysis shows that there has been a convergence in the rate of returns to work experience by gender, with female returns increasing more rapidly than those of men. This is attributed to the diffusion of new technologies that favor the skills of women more than those of men.

Sulis ( 2012 ), in a study of Italy, found that search frictions, productivity, and discrimination all shaped the gender wage gap. Another issue related to maternity is the prevalence of part-time working by women. Part-time working attracts lower hourly rates of pay and has often been identified as an important contributor to the gender wage gap. Blau et al. ( 2013 ) found that US policies encouraged women to undertake part-time work in lower level jobs. Ermisch and Wright ( 1993 ) provide evidence that women in the UK received lower wages in part-time than in full-time work. Moreover, as noted above, Goldin ( 2014 ) emphasizes the role of flexible working times in highly paid occupations and senior positions. This, in turn, is an argument to support the view that the preference of women for part-time work might tend to exclude women from such types of jobs. The role of part-time working in creating gender wage differences is another focal point of the analysis presented in this study.

Several studies have tried to understand the origins of discrimination and have found evidence that they are related to the persistence of traditional views regarding the gender division of roles in society. Fortin ( 2005 ) finds perceptions of the role of women in the home and in society to have a significant effect on the gender wage gap—that anti-egalitarian views are associated with a higher gender wage inequality. Pastore and Tenaglia ( 2013 ) find evidence of the role that different religious denominations have in favoring or hindering female employment—as a consequence of a different degree of secularization and of views regarding traditional gender roles and the male breadwinner family model.

Gauchat et al. ( 2012 ) examine other potential effects on gender wage inequality in the USA, such as the effects of globalization, finding that it contributes to a reduced gender pay gap. Oostendorp ( 2009 ) finds evidence that the occupational gender wage gap tends to decrease with respect to trade and foreign direct investment in richer countries but found little evidence of any effect in poorer countries. In a study of wages in India, Menon and Van der Meulen Rodgers ( 2009 ) even find the gender wage gap to increase with respect to openness to international trade.

All of the key themes developed by this paper have been previously considered in one way or another by the existing literature. At the heart of the gender pay gap is a sense that women are paid less than men for undertaking essentially the same work. Matching techniques offer the opportunity to better compare like with like, and such comparisons are of considerable importance. But the literature makes clear that female employment is typically not like male employment. For example, gender segregation, part-time working, parenthood, and unionization are all factors which affect differences between male and female wages. The contribution of this paper is to provide systematic and robust evidence on how these factors interact with the core “like for like” gender pay gap. It finds, for example, that being both a female and a part-time worker results in a much greater disadvantage in hourly wages than just being female. In so doing, it implies that the concept of a single gender pay gap is a too simplistic representation of reality.

3.1 Data overview

The study uses two cross-section samples taken from the monthly US Current Population Survey (CPS), the first for October 2011 to March 2012 and the second for October 2017 to March 2018. Since both cross-sections comprise different individuals, it is not possible to formally test for changes between the two periods, but the intention was to check whether key conclusions change between the two periods. The full number of observations for the first sample was 907,775 and for the second 877,776. This sample includes non-responses and individuals who were not in employment at the time. For much of the analysis, the effective sample was necessarily limited to those individuals for whom sufficient information to obtain their usual hourly earnings existed. This amounted to 77,097 individuals for the first sample and 76,308 for the second. It should also be noted that the Stata software automatically removes observations for which there are missing values so the actual number of observations used in any one task may vary from these totals. The first sample (October 2011 to March 2012) comprised 51.6% females and 48.4% males, and the second sample (October 2017 to March 2018) had exactly the same proportions.

3.2 Sample characteristics

Table 1 provides employment rates of males and females for both samples. Participation rates for both males and females increased in the six years between the two samples. In both cases, the proportion of females not in the labor force was about 10% higher than that of males. Lower overall participation rates for females were not the only key difference from males. In both samples, the proportion of females working part time was substantially higher than that of males. In the second later sample, this became more exaggerated with the proportion of females engaged in part-time work being roughly double compared with that of males.

As Blau and Kahn ( 2017 ) note, the existence of gender segregation implies that industry and occupational differences between male and female employment are important contributory factors to gender differences in wages. To assess the extent and evolution of gender segregation, Table 2 reports gender segregation indices for CPS data over a much longer period (March 2005 to March 2018) than those used for the rest of the study. These indices suggest a gradual decline in gender segregation by occupation between March 2005 and March 2018, but the overall degree of segregation by the end still remained substantial. For segregation by industry, there is very little evidence of longer term change. Segregation by industry is lower than that by occupation but still of consequence. It is worth noting carefully that the values of gender segregation indices are necessarily affected by how both “occupation” and “industry” are defined. The narrower the definitions, the more likely one is to observe a greater degree of gender segregation.

These findings are consistent with other studies of gender segregation in US labor markets. Most notably, Blau et al. ( 2013 ) find a value of 51% for occupational segregation in 2009 compared with about 52% in March and September 2009 in this study. The results are also consistent with the findings of Hegewisch et al. ( 2010 ) on occupational segregation. The findings support the view of Blau and Kahn ( 2017 ) that the decline in gender segregation observed in earlier decades has stalled at levels that still represent a high degree of occupational segregation. Available existing evidence on segregation by industry is much more limited so providing such evidence is one of the contributions of this study.

The analysis necessarily used the CPS definitions of both occupation and industry. Detailed definitions of both industry and occupation were used. Due to changes in definitions over the period, the precise number of each varied, but there were at least 600 occupation and 250 industry categories included throughout. It is recognized that such definitions can never be wholly satisfactory and that the results could have been significantly affected by a different alternative set of definitions.

Another relevant feature of the data is that women exhibited lower rates of unionization than men. In the first sample (October 2011 to March 2012), 12.8% of males and 11.4% of females were unionized. In the second sample (October 2017 to March 2018), the comparable proportions were 11.0% for males and 9.9% for females.

3.3 Variables

Much of the analysis was concerned with the effect of gender on wages. For this, the outcome (dependent) variable was the lhwage, the log of usual hourly earnings. For most of the analysis, the key treatment variable was female (0 if male, 1 if female).

The following variables were used mainly as covariates but also served as treatment variables in some instances:

parttime, 0 if full time and 1 if part time

young, 0 if 25 or over and 1 if under 25

parent, 1 if a parent of a child aged under 18 but 0 if not

union, 1 if a union member but 0 if not.

The following variables were used as covariates only:

married, 1 if married but 0 if not

edyears, number of years of education

hours, the usual number of weekly hours worked

exper, expected experience (explained further below)

migrant, 0 if born in the USA but 1 if not

regional dummy variables

dummy variables for race

occupational dummy variables

sector dummy variables.

Both the occupational and sector dummies used the standard CPS definitions. It is recognized that occupations and industries are impossible to define in a wholly satisfactory way and that variations in these definitions could result in quite results for these dummy variables.

To calculate expected experience for each individual in the model, a probit model was used to estimate (separately) the probability of employment at each age starting at 15 and ending at 65. The role of expected experience (and of gender differences in the effect of parenthood) as a determinant of the gender pay gap was first advanced by Polachek ( 1975 ). In this paper, the model of expected experience was of the general form:

where empl is the (0, 1) variable for whether the individual was employed and D is a vector of regional and race dummy variables.

The marginal effects (probabilities) were then used to calculate the probability that each individual would have been in employment at each age from 15 to 65. These were then added together to give the expected experience in years. Given space constraints, the results are not reported here but are available from the authors on request.

4 Methodology

4.1 wage decompositions using recentered influence functions.

Firpo et al. ( 2018 ) offer an extension of the Oaxaca-Blinder wage decomposition using recentered influence functions (RIF). The technique involves two steps, the first of which is to divide the wage distribution into a composition and structure effect using a reweighted procedure (where the weights are estimated). The second step estimates structure and composition effects for each covariate; essentially in a manner similar to that of Oaxaca-Blinder. The key difference is that, using the method developed by Firpo et al. ( 2009 ) and Fortin et al. ( 2011 ), the dependent variable of the regression is replaced by the appropriate RIF. To implement this procedure, we used the oaxaca_rif routine in Stata .

Authors using different data sets than those of this study have used Oaxaca RIF decompositions to estimate the gender pay gap. Some of these, such as Kassenboehmer and Sinning ( 2014 ) and Töpfer ( 2017 ), found a high proportion of unexplained gender differences which they attributed to heterogeneity in their data. Wage decompositions were not a focus of this study. Our main purpose in producing such estimates was to demonstrate that similar problems existed with the two data sets used for this study. The evidence that similar issues exist with the CPS data is intended to support the use of matching estimators in this study. A summary of the results of the Oaxaca RIF analysis is presented in the Appendix . More detailed results are available from the authors on request. The interpretation of the results needs some care. In particular, the “unexplained” component is open to misinterpretation and differing points of view. Further details are not provided here since this study argues that a different methodological approach is more suited to its topic.

4.2 Matching with a single treatment variable

The existing empirical literature emphasizes the need to compare like with like with respect to gender pay differences. Some authors, including Ñopo ( 2008 ) and Frölich ( 2007 ), have advocated the use of matching estimators for this purpose. Both authors propose these techniques as an alternative to the decompositions of the type proposed by Blinder ( 1973 ) and Oaxaca ( 1973 ). For example, Ñopo ( 2008 ) argues that matching addresses the “out of support” problem inherent in Blinder-Oaxaca wage decomposition models. Section 4.1 above argued that a more modern version of wage decompositions using RIF is still subject to heterogeneity issues. Matching approaches are well equipped to deal with heterogeneity issues. In addition, the heart of the matching approach (the selection of a carefully matched control group) has considerable intuitive appeal in any attempt to compare like with like.

A matching approach starts by defining an outcome variable (log of hourly earnings) and a (0, 1) treatment variable (female). It seeks to establish whether a statistically significant difference exists in the log of hourly earnings between the treated (female) group and the untreated (male) group. The procedure selects a control group from the untreated (male) group which is selected to be, as far as possible, identical in all other relevant observable characteristics to the treated (female) group.

A key issue for all matching techniques is the “missing data” problem. For example, the treatment variable (say being female) is observed, but, to compare male and female wages accurately, we would need to know what would have happened if the same individual had been born male. This clearly cannot be observed, and the “missing data” problem is how best to replicate it from an appropriate counterfactual. With a single treatment variable, this means selecting an appropriate control group.

This study uses three different approaches to the selection of the control group. These are propensity score (PS) matching (using kernel density matching), matching by Mahalanobis distance, and coarsened exact matching (CEM). Given the widespread use of the first two matching techniques in the literature, no further explanation is offered here. The CEM technique is a more recent addition to the matching toolbox: see Iacus et al. ( 2012 ). For matching by both propensity score and by Mahalanobis distance, the treated group is not changed and the only “matching” occurs in the creation of a control group. With coarsened exact matching, the process excludes all those observations from the treated group for which a nearly exact match on all covariates cannot be found. CEM sets a maximum difference in the covariates between the treated and untreated groups and removes observations from both groups where no nearly exact match exists. In many respects, this makes it a more rigorous attempt to compare like with like, but, unlike the other approaches, it results in sample size reductions.

Neither PS nor Mahalanobis matching techniques remove those observations from the treated group that are “difficult” to match closely. In consequence, an issue arises of how closely the control group matches the treated group (sometimes referred to as “bias on observables”). For each analysis using both techniques, the match between the two groups was checked using the psmatch2 routine in Stata. The resulting graphs are reported in the separate appendices available from https://www.researchgate.net/publication/331703104_Meara_Pastore_Webster_specification_checks .

A further more intractable problem is the risk of bias on unobservables: an excluded confounding variable may have biased the results. This study uses a large number of covariates in the treatment model in an attempt to reduce this risk (see Section 3 ). However, as King and Nielsen ( 2016 ) have pointed out, doing this can create a risk of a different form of bias: from matching on irrelevant variables. To limit that risk, all covariates included in the probit (treatment) model were first tested for statistical significance in a regression model with the outcome as the dependent variable. These regressions are not reported but details are available from the authors on request.

The approach taken in this study reflects conceptual as well as statistical issues. For matching estimators to be unbiased, they need to include all relevant observables. This means that in estimating the gender pay gap, the technique should control for other covariates that are known to also affect the difference in gender wages. These include the effects of gender segregation, part-time working, unionization, and parenthood. It is, of course, central to the study to estimate the gender wage gap on as close to a “like for like” basis as possible. However, it is also important to recognize that this is an estimate of the direct consequence of gender on wages and that there are other less direct mechanisms that affect gender wages. The approach of this study is to identify how the gender pay gap changes when these “indirect” effects of being female are taken into account.

The CPS data reveal, as expected, that part-time working is more common among females than males and that females are less unionized. The study first uses matching to show that, with the CPS data, there existed a union wage premium and an hourly wage discount for working part time. Next, the study estimated the core (like for like) gender pay gap for both samples. This is estimated firstly with industry and occupation dummies. It was then re-estimated without these dummy variables to identify the effect of gender segregation on the gender pay gap. For the remainder of the matching analysis, the sample was sub-divided into two according to one of the key covariates. These were used to show how the gender pay gap varies between one group and another. For example, the sample was divided into young (under 25) and older workers and the gender pay gap estimated for each. A similar approach was taken for part-time working, union membership, and parenthood. These provided a key insight into how each of these variables influences differences in gender wages.

4.3 Matching with inverse probability weighted regression adjustment (IPWRA)

The IPWRA estimator derived by Cattaneo ( 2010 ) and Cattaneo et al. ( 2013 ) differs from most matching estimators in that it estimates both a treatment model and an outcome model. The treatment model is similar to most matching models. It estimates the probability of the treatment variable (female in this case) being associated with each of a number of characteristics. Many matching models use probit for this purpose. In this study, the IPWRA treatment model used a logit model.

The treatment model gives the probability of, say, observing a female given that one observes a part-time worker. That is, the treatment model is used to assign a sampling probability for each observation. The inverse of this probability is then used to weight each observation in the outcome models. The inverse probabilities are used to address the “missing data” problem. Using these inverse probabilities, in essence, creates a counterfactual to address the missing data issue. The technique next estimates a number of (inverse probability) weighted regression outcome models, one for each treatment level. Each of these produces a series of treatment-specific predicted outcomes, one for each treatment level. The means of these predicted outcomes are then used to estimate the treatment effect.

The IPWRA estimator can be shown to have some important statistical properties. The most important of these is the property of “double robustness”: see Cattaneo ( 2010 ) and Cattaneo et al. ( 2013 ). That is, if either the treatment model or the outcome model is incorrectly specified but the other is correctly specified, then the estimates are still consistent. This means that it is only necessary for one of the two to be correctly specified for the estimator to be consistent. As a corollary, it is necessary to assume that at least one of the treatment or outcome models does not exclude a confounding variable.

Hirano et al. ( 2003 ) have shown that doubly robust estimators (which include IPWRA) exhibit a lower bias than estimators without the double robustness property. Another common problem with matching models is mis-matching on irrelevant variables. King and Nielsen ( 2016 ) point out that IPWRA estimators are less prone to mis-matching on irrelevant observables.

From the perspective of this paper, the reasons for using the IPWRA are not just for the desirable statistical properties of the estimator but also for the questions that it can address. The model is specified to work with a number of discrete treatment levels. This means that it can be adapted to work with more than one treatment variable. For example, suppose that that we have two (0, 1) treatment variables: female and parttime. This can be adapted into four treatment levels:

Treatment level 0: female = 0 and parttime = 0

Treatment level 1: female = 1 and parttime = 0

Treatment level 2: female = 0 and parttime = 1

Treatment level 3: female = 1 and parttime = 1

In this way, it is possible to use the IPWRA to estimate both treatment effects separately and to estimate their joint (interaction) effect when both apply. It is this feature that makes it particularly useful for analyzing the interaction between gender and other related influences such as part-time working, unionization, and parenthood.

In this study, the outcome variable for all IPWRA models was the log of hourly wages. For both the treatment and outcome models, the full set of covariates listed in the preceding section was used. An important assumption of the IPWRA model is known as the overlap assumption. This means that every individual must have a positive probability of receiving each treatment level. For example, it must be possible that union members can be male and can be female. If unions excluded all males or all females, the overlap assumption would be violated. Stata produces graphical checks for the overlap assumption. These are not reported for the IPWRA models in Section 6 but are available in separate appendices available from https://www.researchgate.net/publication/331703104_Meara_Pastore_Webster_specification_checks .

Finally, as with other matching models, the IPWRA analysis assumes that treatments and outcomes are statistically independent (conditional mean independence).

4.4 Interpretation of results

For both the single treatment and the IPWRA matching analysis, the outcome variable is the log of hourly wages. Consequently, the average treatment effect on the treated (ATT) is the difference in the log of wages between, say, females and males. This is often interpreted as the percentage difference in wages. However, the difference in logs is only a linear approximation (by means of a Taylor expansion) of the true percentage difference. This approximation (as can be seen in our results) is only accurate when the difference between the two sets of wages is small. Since the precise percentage difference can readily be derived from the matching output, this is reported together with the relevant ATT throughout this paper, except for the CEM analysis (for which the ATT is estimated differently and correctly reflects the exact percentage difference).

5 Matching analysis with a single treatment variable

5.1 treatment effects of part-time working and union membership.

This section provides a supporting analysis for work to follow on the gender pay gap. Earlier analysis of the CPS data (Section 3 ) has shown that women are less likely than men to be unionized but more likely to be working part time. The purpose of this analysis is to demonstrate that, with the CPS data, both union membership and part-time working have significant effects on wages in their own right.

Table 3 presents matching estimates of the reduction in hourly wages from working part time and the wage premium from being a union member. These are for the full sample and made use of the full set of covariates listed in Section 4 earlier, including industry, occupation, race, and region dummies. Results are for propensity score (kernel density) matching and use a second set of estimates (from matching by Mahalanobis distance) as a robustness check. Since this is a supporting analysis, we do not also provide a set of CEM estimates (as is done with later analysis) in the interests of being concise.

Table 3 shows a statistically significant premium for union membership according to the PS matching estimator. The results (statistically significant at 99% confidence) imply a union wage premium of about 14% for our first sample and about 13% for the second. The Mahalanobis estimates for the first sample are comparable with those of the PS estimator for the first sample (a premium of about 14%) but slightly lower for the second sample (a premium of about 11%). Both estimators support a substantial and statistically significant union wage premium in each sample.

For part-time working, our results consistently show a substantial and statistically significantly lower hourly wage than for full-time working. Propensity score estimates for both our samples are comparable: a part-time discount of about 19% in October 2011 to March 2012 and of about 21% in October 2017 to March 2018. Estimates for matching by Mahalanobis distance are again comparable across the two samples—discounts of about 14% and 16%—but are somewhat lower than those for the propensity score estimator. Nonetheless, both estimators support a conclusion that a substantial disadvantage in hourly wages exists from working on a part-time basis.

This study reported earlier that, for our samples from the US CPS data, women were more likely to work part time and less likely to be unionized. The analysis in this section has shown that, for the same data, both characteristics would contribute to an overall difference between male and female wages that goes beyond the impact of the direct effect of gender alone. This is a key point to be explored further in this study. It implies that a “like for like” comparison of the direct effect of gender on wages is not the only effect that merits consideration.

5.2 Treatment effects of gender

This section focuses on matching estimates for the gender pay gap in the US using both our samples. As discussed earlier, it is important that the matching process makes use of all relevant observed covariates. Not to do so would expose the estimates to an increased risk of bias on unobservables. The resulting estimate is, in consequence, an estimate of the effect on wages of being female with the effects of all other observed covariates controlled by the matching process. Such estimates are unquestionably useful but give rise to two sets of concerns. These are not really statistical but are important for our understanding of gender wage differences. Firstly, we know from the literature that gender wage differences can vary by, for example, age group and that gender segregation affects gender wage differences. It is important to understand these factors. Secondly, the process of matching selects controls (males) which are similar in terms of, say, parenthood, part-time working, or union membership. All of these can affect gender wage differences. In short, there needs to be an estimate of the effect of gender on wages where, as far as possible, like is compared with like. But in so doing, it is important not to neglect other more indirect routes by which gender wage differences occur.

In this section we start by estimating the gender pay gap for both our samples. The main estimate of the gender pay pap quite properly controls for the effect on wages of the concentration of women in lower paid occupations or industries (gender segregation). To identify the effects of gender segregation, we repeat the analysis but without industry or sector dummy variables. Next, we consider the effect of age on the gender wage differences by applying our matching estimates to two sub-samples—young (under 25) and older. Since part-time working results in lower hourly wages (see the preceding section), we then estimate separate gender wage gaps for part-time and full-time workers. Separate gender pay gaps are then estimated for parents and non-parents and for union members and non-members. The purpose of all of these is to provide a much richer analysis and interpretation than just the direct effect of gender on wages.

Table 4 reports the results of this analysis using propensity score (PS) matching (kernel density), Table 5 repeats the analysis for matching by Mahalanobis distance, and Table 6 also repeats the analysis using coarsened exact matching (CEM). The PS matching (Table 4 ) is included since it is the most widely understood matching technique. Matching by Mahalanobis distance (Table 5 ) and matching by the CEM technique (Table 6 ) are both included as robustness checks on the findings of the PS matching analysis.

The PS matching analysis (Table 4 ) produced an estimate of a statistically significant gender pay gap of about 13% for the October 2011 to March 2012 sample and of about 12% for the October 2017 to March 2018 sample. Comparable estimates using (a) Mahalanobis distance (Table 5 ) and (b) CEM (Table 6 ) were (a) 13% and 10.5% and (b) 12% and 14%. In all cases, these estimates were statistically significant at 99% confidence. These estimates represent the gender pay gap resulting from the direct effect of being female. That is, the secondary effects of, for example, part-time working, parenthood, or union membership are included in the controls and not in the estimate.

Table 4 shows the effect of taking into account gender segregation by means of industry and occupation dummy variables. Removing these industry and occupation dummies increased the estimate of the gender pay gap to 15% for the first sample and to 16% for the second. A comparable effect was observed with both the Mahalanobis and CEM estimators (Tables  5 and 6 ). Interpretation of these findings is important. It is not necessary to choose between estimates with industry and occupation dummy variables and those without. Both convey complementary information. To the extent to which the matching was successful in comparing like with like, the estimates for, say, the second sample showed that being female involved hourly wages that were typically 13% less than those for males. Since this estimate controls for differences in industry and occupation, it does not take into account gender segregation. When we allow for the effects of females being more concentrated in lower paid industries and occupations, the comparable estimate is a pay gap of 17%. As with Blau and Kahn ( 2017 ), this supports the conclusion that gender segregation by industry and by occupation is important in understanding gender wage differences.

The next sub-division of the sample was between young (under 25) and older. Previous studies have found the gender pay gap to be smaller or even non-existent for younger workers. With the PS matching (Table 4 ), this study finds a small but statistically significant gender pay gap for young individuals, of about 2% in our first sample and about 3% in the second. Both the Mahalanobis distance matching (Table 5 ) and the CEM (Table 6 ) analysis found no statistically significant gender pay differences (at 95% confidence) for young workers. These findings contrast sharply for the estimates of the gender pay gap for older workers. For each of the three estimators, these were statistically significant and substantially higher than those for young workers. The PS matching estimates (Table 4 ) imply a gender pay gap of about 13% for older workers in the first sample and of about 14% in the second sample. Mahalanobis distance (Table 5 ) and CEM (Table 6 ) yield similar results. The sharp difference in the gender pay gap between young and older workers has some obvious potential implications for the role of marriage and parenthood in gender pay differences. These are discussed further later.

Sub-dividing the sample by part-time and full-time workers produces some further interesting findings. The PS matching analysis (Table 4 ) suggests a statistically significant but small gender pay gap for part-time workers. For this first sample, this was estimated at 3% and, for the second sample, 6%. Both Mahalanobis and CEM techniques (Tables 5 and 6 ) found no statistically significant (at 95%) gender pay difference between male and female part-time workers. The gender pay gap for full-time workers estimated by PS matching (Table 4 ) was statistically significant and substantial for both samples—14% for the first sample and 15% for the second. Both Mahalanobis and CEM techniques produced similar estimates (Tables 5 and 6 ). The finding of no statistically significant gender difference in the hourly wages of part-time workers is of consequence. Evidence presented earlier shows both that a higher proportion of females than males work part time and that part-time working involves its own gap in hourly pay relative to full time. That there is little or no gender pay difference between male and female part-time workers implies that the interaction between gender and part-time effects is of importance. That is, the role of part-time working in the gender pay gap is more through the pay disadvantage of part-time working than any significant gender wage difference between part-time workers. This is further analyzed in the next section.

The division of both samples by parenthood finds a statistically significant gender pay gap for both parents (of children under 18) and for non-parents in both samples, according to all three of the matching estimators used. In every case, the estimated wage gap for parents was substantially greater than that for non-parents. For example, the estimated wage gap for parents using PS matching was about 17% in the first sample and about 18% in the second sample. The comparable estimates for non-parents were 10% and 12%. These findings complement those with respect to age, which imply changes in the gender pay gap at ages consistent with parenthood. They also complement the existing literature which finds a role for parenthood affecting the gender pay gap, not least through its impact on experience and human capital. Again, the role of parenthood is further analyzed in the next section.

The last sub-division of the samples was with respect to union membership. Again all three matching estimators find a statistically significant gender pay gap for both samples and for both union and non-union members. In almost all cases, the estimated gender pay gap for union members is greater than that for non-members. With PS matching, the gender pay gap for union members in the first sample was estimated at about 12% and for non-members at 11%. For the second sample, the comparable estimates were 16% and 13%. These findings imply a contradictory effect of union membership on gender wages. Union membership, as shown earlier, involves a wage premium which, given low female unionization, should widen the gender pay gap. In contrast, the gender pay gap not only exists between male and female union members but also is higher than that for those who are not unionized. This implies that to fully understand the net overall effect of the interaction between unionization and gender on pay, further analysis is needed. This is provided in the next section.

6 IPWRA analysis for the full sample

6.1 with gender and part-time working as treatments.

Table 7 presents the results of the IPWRA analysis with both female and parttime as treatment variables. The two treatment variables were combined to produce the following composite treatment levels:

Treatment level 0—male full time (female = 0 and parttime = 0)

Treatment level 1—female full time (female = 1 and parttime = 0)

Treatment level 2—male part time (female = 0 and parttime = 1)

Treatment level 3—both female and part time (female = 1 and parttime = 1)

The results are divided into two parts—absolute and relative treatment effects. Absolute effects are the treatment effects where the control group is treatment level 0 (comparable male full-time workers). Relative effects compare the other (non-zero) treatment levels with each other. In particular, treatment effects were estimated for:

Treatment level 1 (female full time) relative to treatment level 2 (male part time)

Treatment level 1 (female full time) relative to treatment level 3 (female part time)

Treatment level 2 (male part time) relative to treatment level 3 (female part time).

In a similar manner to the earlier matching analysis, the full set of variables listed in Section 4 was used to construct the relevant treatment and outcome models in each case.

The absolute effects presented in Table 7 produce some interesting findings. Firstly, the gender pay gap between male and female full-time workers was 14% in both the earlier and later of the two samples. These are values consistent with the earlier matching analysis. Secondly, the analysis confirms a substantial gap in hourly pay rates between part-time and full-time workers. The gap in hourly pay between full-time and part-time males was about 24% in both samples. This confirms the earlier findings that part-time working involves a substantial disadvantage in hourly pay rates relative to full-time working. Lastly, the (separate) pay gaps for being female and for working part time re-enforce each other when it comes to the pay gap between part-time women and full-time men. For the earlier sample, this estimated gap in pay was about 27% and for the later sample approximately 28%. This provides clear evidence that the prevalence of part-time working is an important mechanism by which the “like for like” gender pay gap is worsened. That is, it shows that the wage disadvantage of being female is substantially worsened when the prevalence of female part-time working is taken into account.

For the relative effects, female part-time working was found to result in substantially lower hourly wages compared with all female workers. This gap was found to be about 15% in the earlier sample and 16.5% in the later one. This provides evidence that the gap between part-time and full-time rates exists for females as well as for males. Female part-time workers were also found to have statistically significantly lower hourly wages than comparable part-time workers of both genders. However, the gender pay gap among part-time workers was comparatively modest—about 3% in both samples. Finally, part-time males were found to have substantially lower wages than females (both part and full time). This implies that the wage disadvantage of working part time is larger than the disadvantage from being female. This finding emphasizes the importance of including the wage disadvantages of part-time working within the understanding of gender wage differences.

The outcome of the IPWRA analysis of gender and part-time working performs two key functions. Firstly, it shows that the disadvantages of working part time and the prevalence of part-time working among females are both relevant and important for understanding gender wage differences. Secondly, it provides a robustness check on many of the earlier findings of the matching analysis. Since there are also no substantial behavioral differences between the two different time periods, the main findings are not just robust with respect to choice of estimator but also robust with respect to the choice between the two cross-sections.

6.2 With gender and union membership as treatments

Table 8 presents the results of the IPWRA analysis using both gender and unionization as treatments. The following composite treatment levels were used:

Treatment level 0—male non-union (female = 0 and union = 0)

Treatment level 1—female non-union (female = 1 and union = 0)

Treatment level 2—male union (female = 0 and union = 1)

Treatment level 3—both female and union (female = 1 and union = 1)

In this case, the absolute effects are the treatment effects in relation to the control group of non-union males (treatment level 0).

Relative effects compare:

Treatment level 1 (female non-union) with treatment level 2 (male union)

Treatment level 1 (female non-union) with treatment level 3 (female union)

Treatment level 2 (male union) with treatment level 3 (female union).

As before, the full set of variables listed in Section 4 was used to construct the relevant treatment and outcome models. These included industry and occupation dummy variables.

Table 8 finds a gender pay gap between non-unionized females and non-unionized males of about 14% in the earlier sample and around 15% in the later one. Again this is consistent with the preceding estimates of the “like for like” gender pay gap. The results also provide evidence of a substantial union wage premium. Male workers benefited from a union wage premium of approximately 18% in the October 2011 to March 2012 sample and of about 17% in the October 2017 to March 2018 sample. Relative to non-unionized males, the effect of female union membership was to reduce the gender pay gap to about 8% in the earlier sample and about 10% in the later sample. That is, the existence of a union wage premium helps to reduce the overall pay gap for females but does not eliminate it.

The relative treatment effects also produce some interesting and relevant findings. One of these is that there exists a gender pay gap within unionized labor. In the earlier sample, female union members were typically paid about 13% less than comparable males and in the later sample about 16% less. For women, as with men, the results show a union wage premium but this is smaller than that for males. The estimated female wage premium was 8.5% in the earlier sample and about 6% in the later one, both less than one half of the male union wage premium. The estimated gender pay gap between non-unionized females and unionized males is in the order of 40% for both samples.

As with part-time working, the IPWRA analysis shows that a strict “like for like” comparison between male and female wages ignores another indirect mechanism by which female wages are disadvantaged. For both male and female workers, there is a union wage premium, although the premium for women is lower. That females are less likely to be unionized also means that any given union wage premium does less to reduce the overall difference in gender wages. A combination of union premium and gender wage gap leads to very large differences in hourly pay rates between non-unionized females and unionized males.

6.3 With gender and parenthood as treatments

This analysis considers composite treatments derived from the two (0, 1) treatment variables female and parent. The following composite treatment levels were used:

Treatment level 0—male non-parent (female = 0 and parent = 0)

Treatment level 1—female non-parent (female = 1 and parent = 0)

Treatment level 2—male parent (female = 0 and parent = 1)

Treatment level 3—both female and parent (female = 1 and parent = 1)

Absolute treatment effects were in comparison to the control group of treatment level 0 (male non-parents).

Treatment level 1 (female non-parent) with treatment level 3 (female parent)

Treatment level 1 (female non-parent) with treatment level 2 (female parent)

Treatment level 2 (male parent) with treatment level 3 (female parent).

Table 9 presents the results of this analysis. For non-parents, the core (“like for like”) gender pay gap was statistically significant in both the October 2011 to March 2012 and the October 2017 to March 2018 samples (about 10% in the first sample and about 11% in the second). The effect of being a male parent (relative to comparable male non-parents) was estimated to result in a statistically significant wage premium of about 8% in the first sample and about 3% in the second. The (absolute) effect of being both female and a parent implies a wage disadvantage of about 5% compared with male non-parents in the first sample and about 11% in the second.

The relative effects are of particular interest. For females, as with males, the results suggest that a statistically significant wage premium exists for parents in relation to non-parents. This premium was estimated at just under 4% for both samples. Within the sub-sample of all parents, the results show a substantial wage disadvantage from being a female parent (in relation to male parents). This disadvantage was estimated at 14.2% for the first sample and 14.7% for the second. Lastly, the results suggest that the effect of parenthood is to widen the gender pay gap. The estimated treatment effect (in relation to all females) of being a male parent implied a gender wage gap of about 22% in the October 2011 to March 2012 sample and of about 24% in the October 2017 to March 2018 sample.

The finding that parenthood is a further source of wage disadvantage for females is, perhaps, not surprising but important to be supported with evidence. These findings do, however, need careful interpretation. The data include only those females in employment at the time of the relevant surveys. The CPS data identifies parents of children under 18 years at the time of survey. This means that they are not capable of incorporating past adverse effects on human capital for those parents whose offspring are now adults. Despite these limitations, the analysis offers evidence which supports the existing literature which emphasizes the role of female parenthood in understanding the gender pay gap.

6.4 With gender and youth as treatments

Table 10 presents the IPWRA analysis which considers composite treatments derived from the treatment variables female and youth (defined as age under 25). The following composite treatment levels were defined:

Treatment level 0—older male (female = 0 and youth = 0)

Treatment level 1—older female (female = 1 and youth = 0)

Treatment level 2—young male (female = 0 and youth = 1)

Treatment level 3—young female (female = 1 and youth =  1)

Absolute treatment effects were in comparison to the control group of treatment level 0 (older males).

Treatment level 1 (older female) with treatment level 3 (young female)

Treatment level 1 (older female) with treatment level 2 (young male)

Treatment level 2 (young male) with treatment level 3 (young female).

The results presented in Table 10 imply a gender pay gap for those aged 25 or over of about 12% in the October 20011 to March 2012 sample and of 12.5% for the October 2017 to March 2018 sample. For those aged under 25 years, there was also a statistically significant gender pay gap but of much smaller magnitude. For both samples, this was estimated at approximately 3%.

For males, the effect of being young, unsurprisingly, results in statistically significantly lower hourly wages compared with being older. For the earlier sample, the gap was estimated at about 25% and for the later sample at about 22%. For females, the comparable effect was a gap of about 21% for the earlier sample and around 20% for the later one. Given that both being young and being female involve lower hourly wages, it is not wholly surprising that both effects re-enforce each other to create a substantial wage gap between young females and older males. For the earlier sample, this gap was estimated at about 27% and for the later sample at just over 25%.

7 Conclusions

The existing literature on the gender pay gap is extensive and the range of potential causes very numerous. This study has, for example, only touched on a sub-set of the wide range of issues covered by Blau and Kahn ( 2017 ). However, there remains a scope for formal statistical analysis. Not all relevant propositions have been tested. Estimations of the gender pay gap through Oaxaca RIF wage decompositions are still beset with concerns relating to the unexplained component and heterogeneity within the sample. Matching estimators provide a stronger basis for controlling for heterogeneity. In a sense, they provide more reassurance that the “unexplained” gender pay gap is in fact not explained by observable characteristics such as part-time working or parenthood.

Despite the strengths of a matching approach in controlling for covariates other than gender, it is too easy to overlook that some of these are also relevant to understanding gender wage differences. Part of the contribution of this study is that it does not ignore many of the more relevant covariates. It shows that when the concentration of women in lower paid occupations and industries (gender segregation) are taken into account, then the gender pay gap increases. It shows that the gap in hourly wages is much smaller for part-time than for full-time workers and for younger than for older workers and, in some cases, not even statistically significant.

The main contribution of this study is in looking at how these key mechanisms by which females are further disadvantaged interact with the gender effect itself. The IPWRA analysis estimates (for October 2017 to March 2018) a gender pay gap of about 15% and a gap in hourly wages from working part time (compared to full time) of about 27%. For those individuals who are both a female and a part-time worker, the gap compared with that for full-time males was estimated at 31%. This shows that part-time working has as important an effect on gender wage differences as the direct “like for like” gender effect.

The matching analysis also showed the gender pay gap for unionized workers to be higher than that for non-unionized workers. It also showed that unionized workers of both genders benefit from a union wage premium. The IPWRA analysis shows that the net effect of union membership is that female union members face a smaller gender pay gap than other workers. That is, despite the gender pay gap being greater for unionized females than for non-unionized females, the existence of the union wage premium means that they face a lower gender pay gap overall.

This paper used a matching approach to obtain as close as possible a “like for like” estimate of the gender pay gap and then examined how the gender pay gap changes with respect to other influences on gender wage differences such as gender segregation, part-time working, and low female unionization. The extensive literature on gender pay means that these have all been discussed somewhere previously. The contribution of this paper is to provide explicit, soundly based estimates of these interactions. This offers a much richer understanding of the way in which different sources of disadvantage for females interact in the creation of gender pay differences. In some instances, it implies that it might be better not to think of a single gender pay gap but of a series of different pay gaps for different groups.

Arulampalam W, Booth AL, Bryan ML (2007) Is there a glass ceiling over Europe? Exploring the gender pay gap across the wage distribution. ILR Rev 60(2):163–186

Google Scholar  

Barón JD, Cobb-Clark DA (2010) Occupational segregation and the gender wage gap in private-and public-sector employment: a distributional analysis*. Econ Record 86(273):227–246

Beaudry P, Lewis E (2014) Do male-female wage differentials reflect differences in the return to skill? Cross-city evidence from 1980-2000. Am Econ J Appl Econ 6(2):178–194

Bertrand M, Goldin C, Katz LF (2010) Dynamics of the gender gap for young professionals in the financial and corporate sectors. Am Econ J Appl Econ 2(3):228–255

Blau FD, Kahn LM (1997) Swimming upstream: trends in the gender wage differential in the 1980s. J Labor Econ 15:1–42

Blau FD, Kahn LM (2013a) The feasibility and importance of adding measures of actual experience to cross-sectional data collection. J Labor Econ 31(S1):S17–S58

Blau FD, Kahn LM (2013b) Female labor supply: why is the United States falling behind? Am Econ Rev 103(3):251–256

Blau FD, Kahn LM (2017) The gender wage gap: extent, trends, and explanations. J Econ Lit 55(3):789–865

Blau FD, Brummund P, Liu AYH (2013) Trends in occupational segregation by gender 1970–2009: adjusting for the impact of changes in the occupational coding system. Demography 50(2):471–492

Blinder AS (1973) Wage discrimination: reduced form and structural estimates. J Human Resour 436–455

Borghans L, Ter Weel B, Weinberg BA (2014) People skills and the labor-market outcomes of underrepresented groups. ILR Rev 67(2):287–334

Cattaneo MD (2010) Efficient semiparametric estimation of multi-valued treatment effects under ignorability. J Econ 155(2):138–154

Cattaneo MD, Drukker DM, Holland AD (2013) Estimation of multivalued treatment effects under conditional independence. Stata J 13(3):407–450

Cobb-Clark DA, Moschion J (2017) Gender gaps in early educational achievement. J Popul Econ 30:1093–1134

Duncan OD, Duncan B (1955) A methodological analysis of segregation indexes. Am Sociol Rev 20(2):210–217

Ermisch JF, Wright RE (1993) Wage offers and full-time and part-time employment by British women. J Hum Resour 28:111–133

Firpo S, Fortin NM, Lemieux T (2009) Unconditional quantile regressions. Econometrica 77(3):953–973

Firpo S, Fortin N, Lemieux T (2018) Decomposing wage distributions using recentered influence function regressions. Econometrics 6(2):28

Flory JA, Leibbrandt A, List JA (2014) Do competitive workplaces deter female workers? A large-scale natural field experiment on job entry decisions. Rev Econ Stud 82(1):122–155

Fortin NM (2005) Gender role attitudes and the labour-market outcomes of women across OECD countries. Oxf Rev Econ Policy 21(3):416–438

Fortin NM, Lemieux T, Firpo S (2011) Decomposition methods. In O. Ashenfelter and D. Card, eds., Handbook of labor economics, Vol. 4A, chap. 1, pp. 1 102. North-Holland, Amsterdam

Frölich M (2007) Propensity score matching without conditional independence assumption—with an application to the gender wage gap in the United Kingdom. Econ J 10(2):359–407

Gauchat G, Kelly M, Wallace M (2012) Occupational gender segregation, globalization, and gender earnings inequality in US metropolitan areas. Gend Soc 26(5):718–747

Goldin C (2014) A grand gender convergence: its last chapter. Am Econ Rev 104(4):1091–1119

Hegewisch A, Liepmann H, Hayes J, Hartmann H (2010) Separate and not equal? Gender segregation in the labor market and the gender wage gap. IWPR Briefing Paper, 377

Hirano K, Imbens GW, Ridder G (2003) Efficient estimation of average treatment effects using the estimated propensity score. Econometrica 71(4):1161–1189

Humlum MK, Nandrup AB, Smith N (2019) Closing or reproducing the gender gap? Parental transmission, social norms and education choice. J Popul Econ 32(2):455–500

Iacus SM, King G, Porro G (2012) Causal inference without balance checking: coarsened exact matching. Polit Anal 20(1):1–24

Kassenboehmer SC, Sinning MG (2014) Distributional changes in the gender wage gap. ILR Rev 67(2):335–361

King G, Nielsen R (2016) Why propensity scores should not be used for matching. Copy at http://j. mp/1sexgVw Download Citation BibTex Tagged XML Download Paper, 378

Levanon A, England P, Allison P (2009) Occupational feminization and pay: assessing causal dynamics using 1950–2000 US census data. Soc Forces 88(2):865–891

Livanos I, Pouliakas K (2012) Educational segregation and the gender wage gap in Greece. J Econ Stud 39(5):554–575

Manning A, Swaffield J (2008) The gender gap in early-career wage growth. Econ J 118(530):983–1024

Menon N, Van der Meulen Rodgers Y (2009) International trade and the gender wage gap: new evidence from India’s manufacturing sector. World Dev 37(5):965–981

Ñopo H (2008) Matching as a tool to decompose wage gaps. Rev Econ Stat 90(2):290–299

Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. Int Econ Rev 693–709

Olivetti C (2006) Changes in women’s hours of market work: the role of returns to experience. Rev Econ Dyn 9(4):557–587

Olsen W, Walby S (2004) Modelling gender pay gaps. Equal Opportunities Commission, Manchester

Oostendorp RH (2009) Globalization and the gender wage gap. World Bank Econ Rev 23(1):141–161

Pastore F (2010) The gender gap in early career in Mongolia. Int J Manpow 31(2):188–207

Pastore F, Tenaglia S (2013) Ora et non Labora? A test of the impact of religion on female labor supply, IZA DP No 7356

Pastore F, Verashchagina A (2011) When does transition increase the gender wage gap? Econ Transit 19(2):333–369

Pastore F, Sattar S, Sinha N, Tiongson ER (2016) When do gender wage differences emerge? A study of Azerbaijan’s labor market. World Bank policy research working paper 7613

Polachek SW (1975) Potential biases in measuring male-female discrimination. J Hum Resour 10:205–229

Polachek SW (1981) Occupational self-selection: a human capital approach to sex differences in occupational structure. Rev Econ Stat 63:60–69

Polachek SW (1985) Occupation segregation: a defense of human capital predictions. J Hum Resour 20(3):437–440

Polachek SW (2014) Equal pay legislation and the gender wage gap. IZA World of Labor

Sulis G (2012) Gender wage differentials in Italy: a structural estimation approach. J Popul Econ 25(1):53–87

Tomaskovic-Devey D, Skaggs S (2002) Sex segregation, labor process organization, and gender earnings inequality. Am J Sociol 108(1):102–128

Töpfer M (2017) Detailed RIF decomposition with selection: the gender pay gap in Italy (No. 26-2017). Hohenheim Discussion Papers in Business, Economics and Social Sciences

Walby S, Olsen W (2002) The impact of women’s position in the labour market on pay and implications for UK productivity. Women and Equality Unit, London

Weichselbaumer D, Winter-Ebmer R (2005) A meta-analysis of the international gender wage gap. J Econ Surv 19(3):479–511

Download references

Acknowledgement

The authors would like to thank the anonymous referees for helpful comments and suggestions.

Author information

Authors and affiliations.

Department of Accounting, Finance and Economics, Bournemouth University, Holdenhurst Road, Bournemouth, BH8 8EB, UK

Katie Meara & Allan Webster

Department of Economics, University of Campania Luigi Vanvitelli, Corso Gran Priorato di Malta, I-81043, Capua (Caserta), Italy

Francesco Pastore

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Allan Webster .

Additional information

Responsible Editor: Klaus F. Zimmermann

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix. Oaxaca RIF decomposition of the gender pay gap

• Robust standard errors are reported for the basic model and bootstrapped standard errors for the reweighted model
• Q10 = 10th percentile, Q50 = median, and Q90 = 90th percentile
• Dependent variable = log of hourly wages
• Covariates:
• • marital status (0, 1)
• • expected experience
• • number of years of education
• • migrant (0, 1)
• • parenthood (0, 1)
• • usual hours of work
• • part-time (0, 1)
• • union membership (0, 1)
• • race dummy variables
• • region dummy variables
• • industry and occupation dummy variables
• Variables used for reweighting:

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Reprints and permissions

About this article

Meara, K., Pastore, F. & Webster, A. The gender pay gap in the USA: a matching study. J Popul Econ 33 , 271–305 (2020). https://doi.org/10.1007/s00148-019-00743-8

Download citation

Received : 17 March 2017

Accepted : 22 July 2019

Published : 05 September 2019

Issue Date : January 2020

DOI : https://doi.org/10.1007/s00148-019-00743-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

JEL classification

• Gender wage gap
• Part-time work
• Gender segregation
• Unionization
• Sample selection bias
• Inverse probability weighted regression adjustment estimator
• Find a journal
• Publish with us
• Track your research

Downloadable Content

Mind the gap: exploring the impact of the gender wage gap towards women's academic success and career aspirations

• Masters Thesis
• Gervacio, Alyzza Colleen A.
• Valiquette L'Heureux, Anais
• David, Ariane
• Political Science
• California State University, Northridge
• Dissertations, Academic -- CSUN -- Political Science.
• Gender Wage Gap
• Higher Education
• Glass Ceiling
• Educational Attainment
• Devaluation
• Occupational Segregation
• Human Capital
• http://hdl.handle.net/20.500.12680/sb397g94d
• by Alyzza Colleen Gervacio

Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.

Report | Wages, Incomes, and Wealth

“Women’s work” and the gender pay gap : How discrimination, societal norms, and other forces affect women’s occupational choices—and their pay

Report • By Jessica Schieder and Elise Gould • July 20, 2016

Download PDF

Press release

Share this page:

What this report finds: Women are paid 79 cents for every dollar paid to men—despite the fact that over the last several decades millions more women have joined the workforce and made huge gains in their educational attainment. Too often it is assumed that this pay gap is not evidence of discrimination, but is instead a statistical artifact of failing to adjust for factors that could drive earnings differences between men and women. However, these factors—particularly occupational differences between women and men—are themselves often affected by gender bias. For example, by the time a woman earns her first dollar, her occupational choice is the culmination of years of education, guidance by mentors, expectations set by those who raised her, hiring practices of firms, and widespread norms and expectations about work–family balance held by employers, co-workers, and society. In other words, even though women disproportionately enter lower-paid, female-dominated occupations, this decision is shaped by discrimination, societal norms, and other forces beyond women’s control.

Why it matters, and how to fix it: The gender wage gap is real—and hurts women across the board by suppressing their earnings and making it harder to balance work and family. Serious attempts to understand the gender wage gap should not include shifting the blame to women for not earning more. Rather, these attempts should examine where our economy provides unequal opportunities for women at every point of their education, training, and career choices.

Introduction and key findings

Women are paid 79 cents for every dollar paid to men (Hegewisch and DuMonthier 2016). This is despite the fact that over the last several decades millions more women have joined the workforce and made huge gains in their educational attainment.

Critics of this widely cited statistic claim it is not solid evidence of economic discrimination against women because it is unadjusted for characteristics other than gender that can affect earnings, such as years of education, work experience, and location. Many of these skeptics contend that the gender wage gap is driven not by discrimination, but instead by voluntary choices made by men and women—particularly the choice of occupation in which they work. And occupational differences certainly do matter—occupation and industry account for about half of the overall gender wage gap (Blau and Kahn 2016).

To isolate the impact of overt gender discrimination—such as a woman being paid less than her male coworker for doing the exact same job—it is typical to adjust for such characteristics. But these adjusted statistics can radically understate the potential for gender discrimination to suppress women’s earnings. This is because gender discrimination does not occur only in employers’ pay-setting practices. It can happen at every stage leading to women’s labor market outcomes.

Take one key example: occupation of employment. While controlling for occupation does indeed reduce the measured gender wage gap, the sorting of genders into different occupations can itself be driven (at least in part) by discrimination. By the time a woman earns her first dollar, her occupational choice is the culmination of years of education, guidance by mentors, expectations set by those who raised her, hiring practices of firms, and widespread norms and expectations about work–family balance held by employers, co-workers, and society. In other words, even though women disproportionately enter lower-paid, female-dominated occupations, this decision is shaped by discrimination, societal norms, and other forces beyond women’s control.

This paper explains why gender occupational sorting is itself part of the discrimination women face, examines how this sorting is shaped by societal and economic forces, and explains that gender pay gaps are present even  within  occupations.

Key points include:

• Gender pay gaps within occupations persist, even after accounting for years of experience, hours worked, and education.
• Decisions women make about their occupation and career do not happen in a vacuum—they are also shaped by society.
• The long hours required by the highest-paid occupations can make it difficult for women to succeed, since women tend to shoulder the majority of family caretaking duties.
• Many professions dominated by women are low paid, and professions that have become female-dominated have become lower paid.

This report examines wages on an hourly basis. Technically, this is an adjusted gender wage gap measure. As opposed to weekly or annual earnings, hourly earnings ignore the fact that men work more hours on average throughout a week or year. Thus, the hourly gender wage gap is a bit smaller than the 79 percent figure cited earlier. This minor adjustment allows for a comparison of women’s and men’s wages without assuming that women, who still shoulder a disproportionate amount of responsibilities at home, would be able or willing to work as many hours as their male counterparts. Examining the hourly gender wage gap allows for a more thorough conversation about how many factors create the wage gap women experience when they cash their paychecks.

Within-occupation gender wage gaps are large—and persist after controlling for education and other factors

Those keen on downplaying the gender wage gap often claim women voluntarily choose lower pay by disproportionately going into stereotypically female professions or by seeking out lower-paid positions. But even when men and women work in the same occupation—whether as hairdressers, cosmetologists, nurses, teachers, computer engineers, mechanical engineers, or construction workers—men make more, on average, than women (CPS microdata 2011–2015).

As a thought experiment, imagine if women’s occupational distribution mirrored men’s. For example, if 2 percent of men are carpenters, suppose 2 percent of women become carpenters. What would this do to the wage gap? After controlling for differences in education and preferences for full-time work, Goldin (2014) finds that 32 percent of the gender pay gap would be closed.

However, leaving women in their current occupations and just closing the gaps between women and their male counterparts within occupations (e.g., if male and female civil engineers made the same per hour) would close 68 percent of the gap. This means examining why waiters and waitresses, for example, with the same education and work experience do not make the same amount per hour. To quote Goldin:

Another way to measure the effect of occupation is to ask what would happen to the aggregate gender gap if one equalized earnings by gender within each occupation or, instead, evened their proportions for each occupation. The answer is that equalizing earnings within each occupation matters far more than equalizing the proportions by each occupation. (Goldin 2014)

This phenomenon is not limited to low-skilled occupations, and women cannot educate themselves out of the gender wage gap (at least in terms of broad formal credentials). Indeed, women’s educational attainment outpaces men’s; 37.0 percent of women have a college or advanced degree, as compared with 32.5 percent of men (CPS ORG 2015). Furthermore, women earn less per hour at every education level, on average. As shown in Figure A , men with a college degree make more per hour than women with an advanced degree. Likewise, men with a high school degree make more per hour than women who attended college but did not graduate. Even straight out of college, women make $4 less per hour than men—a gap that has grown since 2000 (Kroeger, Cooke, and Gould 2016).

Women earn less than men at every education level : Average hourly wages, by gender and education, 2015

The data below can be saved or copied directly into Excel.

The data underlying the figure.

Source :  EPI analysis of Current Population Survey Outgoing Rotation Group microdata

Copy the code below to embed this chart on your website.

Steering women to certain educational and professional career paths—as well as outright discrimination—can lead to different occupational outcomes

The gender pay gap is driven at least in part by the cumulative impact of many instances over the course of women’s lives when they are treated differently than their male peers. Girls can be steered toward gender-normative careers from a very early age. At a time when parental influence is key, parents are often more likely to expect their sons, rather than their daughters, to work in science, technology, engineering, or mathematics (STEM) fields, even when their daughters perform at the same level in mathematics (OECD 2015).

Expectations can become a self-fulfilling prophecy. A 2005 study found third-grade girls rated their math competency scores much lower than boys’, even when these girls’ performance did not lag behind that of their male counterparts (Herbert and Stipek 2005). Similarly, in states where people were more likely to say that “women [are] better suited for home” and “math is for boys,” girls were more likely to have lower math scores and higher reading scores (Pope and Sydnor 2010). While this only establishes a correlation, there is no reason to believe gender aptitude in reading and math would otherwise be related to geography. Parental expectations can impact performance by influencing their children’s self-confidence because self-confidence is associated with higher test scores (OECD 2015).

By the time young women graduate from high school and enter college, they already evaluate their career opportunities differently than young men do. Figure B shows college freshmen’s intended majors by gender. While women have increasingly gone into medical school and continue to dominate the nursing field, women are significantly less likely to arrive at college interested in engineering, computer science, or physics, as compared with their male counterparts.

Women arrive at college less interested in STEM fields as compared with their male counterparts : Intent of first-year college students to major in select STEM fields, by gender, 2014

Source:  EPI adaptation of Corbett and Hill (2015) analysis of Eagan et al. (2014)

These decisions to allow doors to lucrative job opportunities to close do not take place in a vacuum. Many factors might make it difficult for a young woman to see herself working in computer science or a similarly remunerative field. A particularly depressing example is the well-publicized evidence of sexism in the tech industry (Hewlett et al. 2008). Unfortunately, tech isn’t the only STEM field with this problem.

Young women may be discouraged from certain career paths because of industry culture. Even for women who go against the grain and pursue STEM careers, if employers in the industry foster an environment hostile to women’s participation, the share of women in these occupations will be limited. One 2008 study found that “52 percent of highly qualified females working for SET [science, technology, and engineering] companies quit their jobs, driven out by hostile work environments and extreme job pressures” (Hewlett et al. 2008). Extreme job pressures are defined as working more than 100 hours per week, needing to be available 24/7, working with or managing colleagues in multiple time zones, and feeling pressure to put in extensive face time (Hewlett et al. 2008). As compared with men, more than twice as many women engage in housework on a daily basis, and women spend twice as much time caring for other household members (BLS 2015). Because of these cultural norms, women are less likely to be able to handle these extreme work pressures. In addition, 63 percent of women in SET workplaces experience sexual harassment (Hewlett et al. 2008). To make matters worse, 51 percent abandon their SET training when they quit their job. All of these factors play a role in steering women away from highly paid occupations, particularly in STEM fields.

The long hours required for some of the highest-paid occupations are incompatible with historically gendered family responsibilities

Those seeking to downplay the gender wage gap often suggest that women who work hard enough and reach the apex of their field will see the full fruits of their labor. In reality, however, the gender wage gap is wider for those with higher earnings. Women in the top 95th percentile of the wage distribution experience a much larger gender pay gap than lower-paid women.

Again, this large gender pay gap between the highest earners is partially driven by gender bias. Harvard economist Claudia Goldin (2014) posits that high-wage firms have adopted pay-setting practices that disproportionately reward individuals who work very long and very particular hours. This means that even if men and women are equally productive per hour, individuals—disproportionately men—who are more likely to work excessive hours and be available at particular off-hours are paid more highly (Hersch and Stratton 2002; Goldin 2014; Landers, Rebitzer, and Taylor 1996).

It is clear why this disadvantages women. Social norms and expectations exert pressure on women to bear a disproportionate share of domestic work—particularly caring for children and elderly parents. This can make it particularly difficult for them (relative to their male peers) to be available at the drop of a hat on a Sunday evening after working a 60-hour week. To the extent that availability to work long and particular hours makes the difference between getting a promotion or seeing one’s career stagnate, women are disadvantaged.

And this disadvantage is reinforced in a vicious circle. Imagine a household where both members of a male–female couple have similarly demanding jobs. One partner’s career is likely to be prioritized if a grandparent is hospitalized or a child’s babysitter is sick. If the past history of employer pay-setting practices that disadvantage women has led to an already-existing gender wage gap for this couple, it can be seen as “rational” for this couple to prioritize the male’s career. This perpetuates the expectation that it always makes sense for women to shoulder the majority of domestic work, and further exacerbates the gender wage gap.

Female-dominated professions pay less, but it’s a chicken-and-egg phenomenon

Many women do go into low-paying female-dominated industries. Home health aides, for example, are much more likely to be women. But research suggests that women are making a logical choice, given existing constraints . This is because they will likely not see a significant pay boost if they try to buck convention and enter male-dominated occupations. Exceptions certainly exist, particularly in the civil service or in unionized workplaces (Anderson, Hegewisch, and Hayes 2015). However, if women in female-dominated occupations were to go into male-dominated occupations, they would often have similar or lower expected wages as compared with their female counterparts in female-dominated occupations (Pitts 2002). Thus, many women going into female-dominated occupations are actually situating themselves to earn higher wages. These choices thereby maximize their wages (Pitts 2002). This holds true for all categories of women except for the most educated, who are more likely to earn more in a male profession than a female profession. There is also evidence that if it becomes more lucrative for women to move into male-dominated professions, women will do exactly this (Pitts 2002). In short, occupational choice is heavily influenced by existing constraints based on gender and pay-setting across occupations.

To make matters worse, when women increasingly enter a field, the average pay in that field tends to decline, relative to other fields. Levanon, England, and Allison (2009) found that when more women entered an industry, the relative pay of that industry 10 years later was lower. Specifically, they found evidence of devaluation—meaning the proportion of women in an occupation impacts the pay for that industry because work done by women is devalued.

Computer programming is an example of a field that has shifted from being a very mixed profession, often associated with secretarial work in the past, to being a lucrative, male-dominated profession (Miller 2016; Oldenziel 1999). While computer programming has evolved into a more technically demanding occupation in recent decades, there is no skills-based reason why the field needed to become such a male-dominated profession. When men flooded the field, pay went up. In contrast, when women became park rangers, pay in that field went down (Miller 2016).

Further compounding this problem is that many professions where pay is set too low by market forces, but which clearly provide enormous social benefits when done well, are female-dominated. Key examples range from home health workers who care for seniors, to teachers and child care workers who educate today’s children. If closing gender pay differences can help boost pay and professionalism in these key sectors, it would be a huge win for the economy and society.

The gender wage gap is real—and hurts women across the board. Too often it is assumed that this gap is not evidence of discrimination, but is instead a statistical artifact of failing to adjust for factors that could drive earnings differences between men and women. However, these factors—particularly occupational differences between women and men—are themselves affected by gender bias. Serious attempts to understand the gender wage gap should not include shifting the blame to women for not earning more. Rather, these attempts should examine where our economy provides unequal opportunities for women at every point of their education, training, and career choices.

— This paper was made possible by a grant from the Peter G. Peterson Foundation. The statements made and views expressed are solely the responsibility of the authors.

— The authors wish to thank Josh Bivens, Barbara Gault, and Heidi Hartman for their helpful comments.

About the authors

Jessica Schieder joined EPI in 2015. As a research assistant, she supports the research of EPI’s economists on topics such as the labor market, wage trends, executive compensation, and inequality. Prior to joining EPI, Jessica worked at the Center for Effective Government (formerly OMB Watch) as a revenue and spending policies analyst, where she examined how budget and tax policy decisions impact working families. She holds a bachelor’s degree in international political economy from Georgetown University.

Elise Gould , senior economist, joined EPI in 2003. Her research areas include wages, poverty, economic mobility, and health care. She is a co-author of The State of Working America, 12th Edition . In the past, she has authored a chapter on health in The State of Working America 2008/09; co-authored a book on health insurance coverage in retirement; published in venues such as The Chronicle of Higher Education ,  Challenge Magazine , and Tax Notes; and written for academic journals including Health Economics , Health Affairs, Journal of Aging and Social Policy, Risk Management & Insurance Review, Environmental Health Perspectives , and International Journal of Health Services . She holds a master’s in public affairs from the University of Texas at Austin and a Ph.D. in economics from the University of Wisconsin at Madison.

Anderson, Julie, Ariane Hegewisch, and Jeff Hayes 2015. The Union Advantage for Women . Institute for Women’s Policy Research.

Blau, Francine D., and Lawrence M. Kahn 2016. The Gender Wage Gap: Extent, Trends, and Explanations . National Bureau of Economic Research, Working Paper No. 21913.

Bureau of Labor Statistics (BLS). 2015. American Time Use Survey public data series. U.S. Census Bureau.

Corbett, Christianne, and Catherine Hill. 2015. Solving the Equation: The Variables for Women’s Success in Engineering and Computing . American Association of University Women (AAUW).

Current Population Survey Outgoing Rotation Group microdata (CPS ORG). 2011–2015. Survey conducted by the Bureau of the Census for the Bureau of Labor Statistics [ machine-readable microdata file ]. U.S. Census Bureau.

Goldin, Claudia. 2014. “ A Grand Gender Convergence: Its Last Chapter .” American Economic Review, vol. 104, no. 4, 1091–1119.

Hegewisch, Ariane, and Asha DuMonthier. 2016. The Gender Wage Gap: 2015; Earnings Differences by Race and Ethnicity . Institute for Women’s Policy Research.

Herbert, Jennifer, and Deborah Stipek. 2005. “The Emergence of Gender Difference in Children’s Perceptions of Their Academic Competence.” Journal of Applied Developmental Psychology , vol. 26, no. 3, 276–295.

Hersch, Joni, and Leslie S. Stratton. 2002. “ Housework and Wages .” The Journal of Human Resources , vol. 37, no. 1, 217–229.

Hewlett, Sylvia Ann, Carolyn Buck Luce, Lisa J. Servon, Laura Sherbin, Peggy Shiller, Eytan Sosnovich, and Karen Sumberg. 2008. The Athena Factor: Reversing the Brain Drain in Science, Engineering, and Technology . Harvard Business Review.

Kroeger, Teresa, Tanyell Cooke, and Elise Gould. 2016.  The Class of 2016: The Labor Market Is Still Far from Ideal for Young Graduates . Economic Policy Institute.

Landers, Renee M., James B. Rebitzer, and Lowell J. Taylor. 1996. “ Rat Race Redux: Adverse Selection in the Determination of Work Hours in Law Firms .” American Economic Review , vol. 86, no. 3, 329–348.

Levanon, Asaf, Paula England, and Paul Allison. 2009. “Occupational Feminization and Pay: Assessing Causal Dynamics Using 1950-2000 U.S. Census Data.” Social Forces, vol. 88, no. 2, 865–892.

Miller, Claire Cain. 2016. “As Women Take Over a Male-Dominated Field, the Pay Drops.” New York Times , March 18.

Oldenziel, Ruth. 1999. Making Technology Masculine: Men, Women, and Modern Machines in America, 1870-1945 . Amsterdam: Amsterdam University Press.

Organisation for Economic Co-operation and Development (OECD). 2015. The ABC of Gender Equality in Education: Aptitude, Behavior, Confidence .

Pitts, Melissa M. 2002. Why Choose Women’s Work If It Pays Less? A Structural Model of Occupational Choice. Federal Reserve Bank of Atlanta, Working Paper 2002-30.

Pope, Devin G., and Justin R. Sydnor. 2010. “ Geographic Variation in the Gender Differences in Test Scores .” Journal of Economic Perspectives , vol. 24, no. 2, 95–108.

See related work on Wages, Incomes, and Wealth | Women

See more work by Jessica Schieder and Elise Gould

Numbers, Facts and Trends Shaping Your World

Read our research on:

Full Topic List

Regions & Countries

• Publications
• Our Methods
• Short Reads
• Tools & Resources

Read Our Research On:

The Enduring Grip of the Gender Pay Gap

Table of contents, how the gender pay gap increases with age, mothers with children at home tend to be less engaged with the workplace, while fathers are more active, employed mothers earn about the same as similarly educated women without children at home; both groups earn less than fathers, progress in closing the gender pay gap has slowed despite gains in women’s education, gender pay gap differs widely by race and ethnicity, broader economic forces may impact men’s and women’s earnings in different ways, what’s next for the gender pay gap.

The gender pay gap – the difference between the earnings of men and women – has barely closed in the United States in the past two decades. In 2022, American women typically earned 82 cents for every dollar earned by men. That was about the same as in 2002, when they earned 80 cents to the dollar. The slow pace at which the gender pay gap has narrowed this century contrasts sharply with the progress in the preceding two decades: In 1982, women earned just 65 cents to each dollar earned by men.

There is no single explanation for why progress toward narrowing the pay gap has all but stalled in the 21st century. Women generally begin their careers closer to wage parity with men, but they lose ground as they age and progress through their work lives, a pattern that has remained consistent over time. The pay gap persists even though women today are more likely than men to have graduated from college. In fact, the pay gap between college-educated women and men is not any narrower than the one between women and men who do not have a college degree. This points to the dominant role of other factors that still set women back or give men an advantage.

One of these factors is parenthood. Mothers ages 25 to 44 are less likely to be in the labor force than women of the same age who do not have children at home, and they tend to work fewer hours each week when employed. This can reduce the earnings of some mothers, although evidence suggests the effect is either modest overall or short-lived for many. On the other hand, fathers are more likely to be in the labor force – and to work more hours each week – than men without children at home. This is linked to an increase in the pay of fathers – a phenomenon referred to as the “ fatherhood wage premium ” – and tends to widen the gender pay gap.

Related: Gender pay gap in U.S. hasn’t changed much in two decades

Family needs can also influence the types of jobs women and men pursue , contributing to gender segregation across occupations. Differential treatment of women, including gender stereotypes and discrimination , may also play a role. And the gender wage gap varies widely by race and ethnicity.

Pew Research Center conducted this study to better understand how women’s pay compared with men’s pay in the U.S. in the economic aftermath of the COVID-19 outbreak .

The study is based on the analysis of monthly Current Population Survey (CPS) data from January 1982 to December 2022 monthly files ( IPUMS ). The CPS is the U.S. government’s official source for monthly estimates of unemployment . For a quarter of the sample each month, the CPS also records data on usual hourly earnings for hourly workers and usual weekly earnings and hours worked for other workers. In this report, monthly CPS files were combined to create annual files to boost sample sizes and to analyze the gender pay gap in greater detail.

The comparison between women’s and men’s pay is based on their median hourly earnings. For workers who are not hourly workers, hourly earnings were computed as the ratio of usual weekly earnings to usual weekly hours worked. The samples include employed workers ages 16 and older with positive earnings, working full time or part time, including those for whom earnings were imputed by the Census Bureau . Self-employed workers are excluded because their earnings are not recorded in the CPS.

The COVID-19 outbreak affected data collection efforts by the U.S. government in its surveys, especially in 2020 and 2021, limiting in-person data collection and affecting the response rate. It is possible that some measures of economic outcomes and how they vary across demographic groups are affected by these changes in data collection.

“Mothers” and “fathers” refer to women and men 16 and older who have an own child younger than 18 living in the household.

The U.S. labor force, used interchangeably with the workforce in this analysis, consists of people 16 and older who are either employed or actively looking for work.

White, Black and Asian workers include those who report being only one race and who are not Hispanic. Hispanics are of any race. Asian workers include Pacific Islanders. Other racial and ethnic groups are included in all totals but are not shown separately.

“High school graduate” refers to those who have a high school diploma or its equivalent, such as a General Education Development (GED) certificate, and those who had completed 12th grade, but their diploma status was unclear (those who had finished 12th grade but not received a diploma are excluded). “Some college” include workers with an associate degree and those who attended college but did not obtain a degree.

Younger women – those ages 25 to 34 and early in their work lives – have edged closer to wage parity with men in recent years. Starting in 2007, their earnings have consistently stood at about 90 cents to the dollar or more compared with men of the same age. But even as pay parity might appear in reach for women at the start of their careers, the wage gap tends to increase as they age.

Consider, for example, women who were ages 25 to 34 in 2010. In that year, they earned 92% as much as men their age, compared with 83% for women overall. But by 2022, this group of women, now ages 37 to 46, earned only 84% as much as men of the same age. This pattern repeats itself for groups of women who were ages 25 to 34 in earlier years – say, 2005 or 2000 – and it may well be the future for women entering the workforce now.

A good share of the increase in the gender pay gap takes place when women are between the ages of 35 and 44. In 2022, women ages 25 to 34 earned about 92% as much as men of the same ages, but women ages 35 to 44 and 45 to 54 earned 83% as much. The ratio dropped to 79% among those ages 55 to 64. This general pattern has not changed in at least four decades.

The increase in the pay gap coincides with the age at which women are more likely to have children under 18 at home. In 2022, 40% of employed women ages 25 to 34 had at least one child at home. The same was true for 66% of women ages 35 to 44 but for fewer – 39% – among women ages 45 to 54. Only 6% of employed women ages 55 to 64 had children at home in 2022.

Similarly, the share of employed men with children at home peaks between the ages of 35 to 44, standing at 58% in 2022. This is also when fathers tend to receive higher pay, even as the pay of employed mothers in same age group is unaffected.

Parenthood leads some women to put their careers on hold, whether by choice or necessity, but it has the opposite effect among men. In 2022, 70% of mothers ages 25 to 34 had a job or were looking for one, compared with 84% of women of the same age without children at home. This amounted to the withdrawal of 1.4 million younger mothers from the workforce. Moreover, when they are employed, younger mothers tend to put in a shorter workweek – by two hours per week, on average – than other women their age. Reduced engagement with the workplace among younger mothers is also a long-running phenomenon.

Fathers, however, are more likely to hold a job or be looking for one than men who don’t have children at home, and this is true throughout the prime of their working years , from ages 25 to 54. Among those who do have a job, fathers also work a bit more each week, on average, than men who do not have children at home.

As a result, the gender gap in workplace activity is greater among those who have children at home than among those who do not. For example, among those ages 35 to 44, 94% of fathers are active in the workforce, compared with 75% of mothers – a gap of 19 percentage points. But among those with no children at home in this age group, 84% of men and 78% of women are active in the workforce – a gap of 6 points.

These patterns contribute to the gap in workplace activity between men and women overall. As of 2022, 68% of men ages 16 and older – with or without children at home – are either employed or seeking employment. That compares with 57% of women, a difference of 11 percentage points. This gap was as wide as 24 points in 1982, but it narrowed to 14 points by 2002. Men overall also worked about three hours more per week at a job than women in 2022, on average, down from a gap of about six hours per week in 1982.

Parenthood affects the hourly earnings of employed women and men in unexpected ways. While employed mothers overall appear to earn less than employed women without children at home, the gap is driven mainly by differences in educational attainment between the two groups. Among women with similar levels of education, there is little gap in the earnings of mothers and non-mothers. However, fathers earn more than other workers, including other men without children at home, regardless of education level. This phenomenon – known as the fatherhood wage premium – is one of the main ways that parenthood affects the gender pay gap among employed workers.

Motherhood does have important effects on the potential earnings of women. Women who experience breaks in their careers after becoming mothers sacrifice at least some of their earnings . Some mothers may never work for pay after having children, passing on earnings altogether. But it is difficult to know what the earnings of mothers might have been and, as a result, it is hard to know for certain what the full effect of motherhood is on women’s earnings. Estimates suggest that motherhood may account for much of the current shortfall in the earnings potential of women overall. 1

Among employed men and women, the impact of parenting is felt most among those ages 25 to 54, when they are most likely to have children under 18 at home. In 2022, mothers ages 25 to 34 earned 85% as much as fathers that age, but women without children at home earned 97% as much as fathers. In contrast, employed women ages 35 to 44 – with or without children – both earned about 80% as much as fathers. The table turns for women ages 45 to 54, with mothers earning more than women with no children at home. Among those ages 35 to 44 or 45 to 54, men without children earned only 84% as much as fathers.

But these patterns in the earnings of employed mothers and women with no children at home are influenced greatly by differences in education levels between the two. Among employed women ages 25 to 34, some 61% of women without children at home had a bachelor’s degree or higher level of education in 2022, compared with 37% of mothers. It follows that among women ages 25 to 34, those without children at home (a more highly educated group, on average) earned more than women with at least one child at home. Conversely, employed mothers ages 45 to 54 were more likely than other women to have at least a bachelor’s degree – 58% vs. 42%. For that reason, mothers ages 45 to 54 earned more than women without children. 2

When the earnings of mothers are compared with those of women without children at home who have the same level of education, the differences either narrow or go away. Among employed women ages 25 to 34 with at least a bachelor’s degree, both mothers and women without children at home earned 80% as much as fathers in 2022. Among women ages 25 to 34 with a high school diploma and no further education, mothers earned 79% as much as fathers and women with no children at home earned 84% as much. The narrowing of the gap in earnings of mothers and women without children at home after controlling for education level also extends to other age groups.

Thus, among the employed, the effect of parenthood on the gender pay gap does not seem to be driven by a decrease in mothers’ earnings relative to women without children at home. Instead, the widening of the pay gap with parenthood appears to be driven more by an increase in the earnings of fathers. Fathers ages 25 to 54 not only earn more than mothers the same age, they also earn more than men with no children at home. Nonetheless, men without children at home still earn more than women with or without children at home.

Although there is little gap in the earnings of employed mothers and women with no children at home who have the same level of education, there is a lingering gap in workplace engagement between the two groups. Whether they had at least a bachelor’s degree or were high school graduates, mothers ages 25 to 34 are less likely to hold a job or be looking for one. Similarly, younger mothers on average work fewer hours than women without children at home each week, regardless of their education level. The opposite is true for fathers compared with men without children at home.

The share of women with at least a bachelor’s degree has increased steadily since 1982 – and faster than among men. In 1982, 20% of employed women ages 25 and older had a bachelor’s degree or higher level of education, compared with 26% of employed men. By 2022, 48% of employed women had at least a bachelor’s degree, compared with 41% of men. Still, women did not see the pay gap close to the same extent from 2002 to 2022 as they did from 1982 to 2002.

In part, this may be linked to how the gains from going to college have changed in recent decades, for women and men alike. The college wage premium – the boost in earnings workers get from a college degree – increased rapidly during the 1980s. But the rise in the premium slowed down over time and came to a halt around 2010. This likely reduced the relative growth in the earnings of women.

Although gains in education have raised the average earnings of women and have narrowed the gender pay gap overall, college-educated women are no closer to wage parity with their male counterparts than other women. In 2022, women with at least a bachelor’s degree earned 79% as much as men who were college graduates, and women who were high school graduates earned 81% as much as men with the same level of education. This underscores the challenges faced by women of all education levels in closing the pay gap.

Notably, the gender wage gap has closed more among workers without a four-year college degree than among those who do have a bachelor’s degree or more education. For example, the wage gap for women without a high school diploma narrowed from 62% in 1982 to 83% in 2022 relative to men at the same education level. But it closed only from 69% to 79% among bachelor’s degree holders over the same period. This is because only men with at least a bachelor’s degree experienced positive wage growth from 1982 to 2022; all other men saw their real wages decrease. Meanwhile, the real earnings of women increased regardless of their level of education.

As women have improved their level of education in recent decades, they’ve also increased their share of employment in higher-paying occupations, such as managerial, business and finance, legal, and computer, science and engineering (STEM) occupations. In 1982, women accounted for only 26% of employment in managerial occupations. By 2022, their share had risen to 40%. Women also substantially increased their presence in social, arts and media occupations. Over the same period, the shares of women in several lower-paying fields, such as administrative support jobs and food preparation and serving occupations, fell significantly.

Even so, women are still underrepresented in managerial and STEM occupations – along with construction, repair and production, and transportation occupations – when compared with their share of employment overall. And there has been virtually no change in the degree to which women are over represented in education, health care, and personal care and services occupations – the last of which are lower paying than the average across all occupations. The distribution of women and men across occupations remains one of the drivers of the gender pay gap . But the degree to which this distribution is the result of personal choices or gender stereotypes is not entirely clear.

Looking across racial and ethnic groups, a wide gulf separates the earnings of Black and Hispanic women from the earnings of White men. 3 In 2022, Black women earned 70% as much as White men and Hispanic women earned only 65% as much. The ratio for White women stood at 83%, about the same as the earnings gap overall, while Asian women were closer to parity with White men, making 93% as much.

The pay gap narrowed for all groups of women from 1982 to 2022, but more so for White women than for Black and Hispanic women. The earnings gap for Asian women narrowed by about 17 percentage points from 2002 to 2022, but data for this group is not available for 1982.

To some extent, the gender wage gap varies by race and ethnicity because of differences in education, experience, occupation and other factors that drive the gender wage gap for women overall. But researchers have uncovered new evidence of hiring discrimination against various racial and ethnic groups, along with discrimination against other groups, such as LGBTQ and disabled workers. Discrimination in hiring may feed into differences in earnings by shutting out workers from opportunities.

Changes in the gender pay gap are also shaped by economic factors that sometimes drive men’s and women’s earnings in distinctive ways. Because men and women tend to work in different types of jobs and industries, their earnings may respond differently to external pressures.

More specifically, men’s earnings essentially didn’t change from 1982 to 2002. Potential reasons for that include a more rapid decline in union membership among men, a shift away from jobs calling for more physical skills, and global competition that sharply reduced employment in manufacturing in the 1980s. At the same time, women’s earnings increased substantially as they raised their level of education and shifted toward higher-paying occupations.

But in some ways, the economic climate has proved less favorable for women this century. For reasons that are not entirely clear, women’s employment was slower to recover from the Great Recession of 2007-2009. More recently, the COVID-19 recession took on the moniker “ she-cession ” because of the pressure on jobs disproportionately held by women . Amid a broader slowdown in earnings growth from 2000 to 2015, the increase in women’s earnings from 2002 to 2022 was not much greater than the increase in men’s earnings, limiting the closure in the gender pay gap over the period.

Higher education, a shift to higher-paying occupations and more labor market experience have helped women narrow the gender pay gap since 1982. But even as women have continued to outpace men in educational attainment, the pay gap has been stuck in a holding pattern since 2002, ranging from 80 to 85 cents to the dollar.

More sustained progress in closing the pay gap may depend on deeper changes in societal and cultural norms and in workplace flexibility that affect how men and women balance their careers and family lives . Even in countries that have taken the lead in implementing family-friendly policies, such as Denmark, parenthood continues to drive a significant wedge in the earnings of men and women. New research suggests that family-friendly policies in the U.S. may be keeping the pay gap from closing. Gender stereotypes and discrimination, though difficult to quantify, also appear to be among the “last-mile” hurdles impeding further progress.

What is the gender wage gap in your metropolitan area? Find out with our pay gap calculator

901 E St. NW, Suite 300 Washington, DC 20004 USA (+1) 202-419-4300 | Main (+1) 202-857-8562 | Fax (+1) 202-419-4372 |  Media Inquiries

Research Topics

• Email Newsletters

ABOUT PEW RESEARCH CENTER  Pew Research Center is a nonpartisan fact tank that informs the public about the issues, attitudes and trends shaping the world. It conducts public opinion polling, demographic research, media content analysis and other empirical social science research. Pew Research Center does not take policy positions. It is a subsidiary of  The Pew Charitable Trusts .

© 2024 Pew Research Center

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

The persistence of pay inequality: The gender pay gap in an anonymous online labor market

Leib litman.

1 Department of Psychology, Lander College, Flushing, New York, United States of America

Jonathan Robinson

2 Department of Computer Science, Lander College, Flushing, New York, United States of America

3 Department of Health Policy & Management, Mailman School of Public Health, Columbia University, New York, New York, United States of America

Cheskie Rosenzweig

4 Department of Clinical Psychology, Columbia University, New York, New York, United States of America

Joshua Waxman

5 Department of Computer Science, Stern College for Women, New York, New York, United States of America

Lisa M. Bates

6 Department of Epidemiology, Mailman School of Public Health, Columbia University New York, New York, United States of America

Associated Data

Due to the sensitive nature of some of the data, and the terms of service of the websites used during data collection (including CloudResearch and MTurk), CloudResearch cannot release the full data set to make it publically available. The data are on CloudResearch's Sequel servers located at Queens College in the city of New York. CloudResearch makes data available to be accessed by researchers for replication purposes, on the CloudResearch premises, in the same way the data were accessed and analysed by the authors of this manuscript. The contact person at CloudResearch who can help researchers access the data set is Tzvi Abberbock, who can be reached at [email protected] .

Studies of the gender pay gap are seldom able to simultaneously account for the range of alternative putative mechanisms underlying it. Using CloudResearch, an online microtask platform connecting employers to workers who perform research-related tasks, we examine whether gender pay discrepancies are still evident in a labor market characterized by anonymity, relatively homogeneous work, and flexibility. For 22,271 Mechanical Turk workers who participated in nearly 5 million tasks, we analyze hourly earnings by gender, controlling for key covariates which have been shown previously to lead to differential pay for men and women. On average, women’s hourly earnings were 10.5% lower than men’s. Several factors contributed to the gender pay gap, including the tendency for women to select tasks that have a lower advertised hourly pay. This study provides evidence that gender pay gaps can arise despite the absence of overt discrimination, labor segregation, and inflexible work arrangements, even after experience, education, and other human capital factors are controlled for. Findings highlight the need to examine other possible causes of the gender pay gap. Potential strategies for reducing the pay gap on online labor markets are also discussed.

Introduction

The gender pay gap, the disparity in earnings between male and female workers, has been the focus of empirical research in the US for decades, as well as legislative and executive action under the Obama administration [ 1 , 2 ]. Trends dating back to the 1960s show a long period in which women’s earnings were approximately 60% of their male counterparts, followed by increases in women’s earnings starting in the 1980s, which began to narrow, but not close, the gap which persists today [ 3 ]. More recent data from 2014 show that overall, the median weekly earnings of women working full time were 79–83% of what men earned [ 4 – 9 ].

The extensive literature seeking to explain the gender pay gap and its trajectory over time in traditional labor markets suggests it is a function of multiple structural and individual-level processes that reflect both the near-term and cumulative effects of gender relations and roles over the life course. Broadly speaking, the drivers of the gender pay gap can be categorized as: 1) human capital or productivity factors such as education, skills, and workforce experience; 2) industry or occupational segregation, which some estimates suggest accounts for approximately half of the pay gap; 3) gender-specific temporal flexibility constraints which can affect promotions and remuneration; and finally, 4) gender discrimination operating in hiring, promotion, task assignment, and/or compensation. The latter mechanism is often estimated by inference as a function of unexplained residual effects of gender on payment after accounting for other factors, an approach which is most persuasive in studies of narrowly restricted populations of workers such as lawyers [ 10 ] and academics of specific disciplines [ 11 ]. A recent estimate suggests this unexplained gender difference in earnings can account for approximately 40% of the pay gap [ 3 ]. However, more direct estimations of discriminatory processes are also available from experimental evidence, including field audit and lab-based studies [ 12 – 14 ]. Finally, gender pay gaps have also been attributed to differential discrimination encountered by men and women on the basis of parental status, often known as the ‘motherhood penalty’ [ 15 ].

Non-traditional ‘gig economy’ labor markets and the gender pay gap

In recent years there has been a dramatic rise in nontraditional ‘gig economy’ labor markets, which entail independent workers hired for single projects or tasks often on a short-term basis with minimal contractual engagement. “Microtask” platforms such as Amazon Mechanical Turk (MTurk) and Crowdflower have become a major sector of the gig economy, offering a source of easily accessible supplementary income through performance of small tasks online at a time and place convenient to the worker. Available tasks can range from categorizing receipts to transcription and proofreading services, and are posted online by the prospective employer. Workers registered with the platform then elect to perform the advertised tasks and receive compensation upon completion of satisfactory work [ 16 ]. An estimated 0.4% of US adults are currently receiving income from such platforms each month [ 17 ], and microtask work is a growing sector of the service economy in the United States [ 18 ]. Although still relatively small, these emerging labor market environments provide a unique opportunity to investigate the gender pay gap in ways not possible within traditional labor markets, due to features (described below) that allow researchers to simultaneously account for multiple putative mechanisms thought to underlie the pay gap.

The present study utilizes the Amazon Mechanical Turk (MTurk) platform as a case study to examine whether a gender pay gap remains evident when the main causes of the pay gap identified in the literature do not apply or can be accounted for in a single investigation. MTurk is an online microtask platform that connects employers (‘requesters’) to employees (‘workers’) who perform jobs called “Human Intelligence Tasks” (HITs). The platform allows requesters to post tasks on a dashboard with a short description of the HIT, the compensation being offered, and the time the HIT is expected to take. When complete, the requester either approves or rejects the work based on quality. If approved, payment is quickly accessible to workers. The gender of workers who complete these HITs is not known to the requesters, but was accessible to researchers for the present study (along with other sociodemographic information and pay rates) based on metadata collected through CloudResearch (formerly TurkPrime), a platform commonly used to conduct social and behavioral research on MTurk [ 19 ].

Evaluating pay rates of workers on MTurk requires estimating the pay per hour of each task that a worker accepts which can then be averaged together. All HITs posted on MTurk through CloudResearch display how much a HIT pays and an estimated time that it takes for that HIT to be completed. Workers use this information to determine what the corresponding hourly pay rate of a task is likely to be, and much of our analysis of the gender pay gap is based on this advertised pay rate of all completed surveys. We also calculate an estimate of the gender pay gap based on actual completion times to examine potential differences in task completion speed, which we refer to as estimated actual wages (see Methods section for details).

Previous studies have found that both task completion time and the selection of tasks influences the gender pay gap in at least some gig economy markets. For example, a gender pay gap was observed among Uber drivers, with men consistently earning higher pay than women [ 20 ]. Some of the contributing factors to this pay gap include that male Uber drivers selected different tasks than female drivers, including being more willing to work at night and to work in neighborhoods that were perceived to be more dangerous. Male drivers were also likely to drive faster than their female counterparts. These findings show that person-level factors like task selection, and speed can influence the gender pay gap within gig economy markets.

MTurk is uniquely suited to examine the gender pay gap because it is possible to account simultaneously for multiple structural and individual-level factors that have been shown to produce pay gaps. These include discrimination, work heterogeneity (leading to occupational segregation), and job flexibility, as well as human capital factors such as experience and education.

Discrimination

When employers post their HITs on MTurk they have no way of knowing the demographic characteristics of the workers who accept those tasks, including their gender. While MTurk allows for selective recruitment of specific demographic groups, the MTurk tasks examined in this study are exclusively open to all workers, independent of their gender or other demographic characteristics. Therefore, features of the worker’s identity that might be the basis for discrimination cannot factor into an employer’s decision-making regarding hiring or pay.

Task heterogeneity

Another factor making MTurk uniquely suited for the examination of the gender pay gap is the relative homogeneity of tasks performed by the workers, minimizing the potential influence of gender differences in the type of work pursued on earnings and the pay gap. Work on the MTurk platform consists mostly of short tasks such as 10–15 minute surveys and categorization tasks. In addition, the only information that workers have available to them to choose tasks, other than pay, is the tasks’ titles and descriptions. We additionally classified tasks based on similarity and accounted for possible task heterogeneity effects in our analyses.

Job flexibility

MTurk is not characterized by the same inflexibilities as are often encountered in traditional labor markets. Workers can work at any time of the day or day of the week. This increased flexibility may be expected to provide more opportunities for participation in this labor market for those who are otherwise constrained by family or other obligations.

Human capital factors

It is possible that the more experienced workers could learn over time how to identify higher paying tasks by virtue of, for example, identifying qualities of tasks that can be completed more quickly than the advertised required time estimate. Further, if experience is correlated with gender, it could contribute to a gender pay gap and thus needs to be controlled for. Using CloudResearch metadata, we are able to account for experience on the platform. Additionally, we account for multiple sociodemographic variables, including age, marital status, parental status, education, income (from all sources), and race using the sociodemographic data available through CloudResearch.

Expected gender pay gap findings on MTurk

Due to the aforementioned factors that are unique to the MTurk marketplace–e.g., anonymity, self-selection into tasks, relative homogeneity of the tasks performed, and flexible work scheduling–we did not expect a gender pay gap to be evident on the platform to the same extent as in traditional labor markets. However, potential gender differences in task selection and completion speed, which have implications for earnings, merit further consideration. For example, though we expect the relative homogeneity of the MTurk tasks to minimize gender differences in task selection that could mimic occupational segregation, we do account for potential subtle residual differences in tasks that could differentially attract male and female workers and indirectly lead to pay differentials if those tasks that are preferentially selected by men pay a higher rate. To do this we categorize all tasks based on their descriptions using K-clustering and add the clusters as covariates to our models. In addition, we separately examine the gender pay gap within each topic-cluster.

In addition, if workers who are experienced on the platform are better able to find higher paying HITs, and if experience is correlated with gender, it may lead to gender differences in earnings. Theoretically, other factors that may vary with gender could also influence task selection. Previous studies of the pay gap in traditional markets indicate that reservation wages, defined as the pay threshold at which a person is willing to accept work, may be lower among women with children compared to women without, and to that of men as well [ 21 ]. Thus, if women on MTurk are more likely to have young children than men, they may be more willing to accept available work even if it pays relatively poorly. Other factors such as income, education level, and age may similarly influence reservation wages if they are associated with opportunities to find work outside of microtask platforms. To the extent that these demographics correlate with gender they may give rise to a gender pay gap. Therefore we consider age, experience on MTurk, education, income, marital status, and parental status as covariates in our models.

Task completion speed may vary by gender for several reasons, including potential gender differences in past experience on the platform. We examine the estimated actual pay gap per hour based on HIT payment and estimated actual completion time to examine the effects of completion speed on the wage gap. We also examine the gender pay gap based on advertised pay rates, which are not dependent on completion speed and more directly measure how gender differences in task selection can lead to a pay gap. Below, we explain how these were calculated based on meta-data from CloudResearch.

To summarize, the overall goal of the present study was to explore whether gender pay differentials arise within a unique, non-traditional and anonymous online labor market, where known drivers of the gender pay gap either do not apply or can be accounted for statistically.

Materials and methods

Amazon mechanical turk and cloudresearch.

Started in 2005, the original purpose of the Amazon Mechanical Turk (MTurk) platform was to allow requesters to crowdsource tasks that could not easily be handled by existing technological solutions such as receipt copying, image categorization, and website testing. As of 2010, researchers increasingly began using MTurk for a wide variety of research tasks in the social, behavioral, and medical sciences, and it is currently used by thousands of academic researchers across hundreds of academic departments [ 22 ]. These research-related HITs are typically listed on the platform in generic terms such as, “Ten-minute social science study,” or “A study about public opinion attitudes.”

Because MTurk was not originally designed solely for research purposes, its interface is not optimized for some scientific applications. For this reason, third party add-on toolkits have been created that offer critical research tools for scientific use. One such platform, CloudResearch (formerly TurkPrime), allows requesters to manage multiple research functions, such as applying sampling criteria and facilitating longitudinal studies, through a link to their MTurk account. CloudResearch’s functionality has been described extensively elsewhere [ 19 ]. While the demographic characteristics of workers are not available to MTurk requesters, we were able to retroactively identify the gender and other demographic characteristics of workers through the CloudResearch platform. CloudResearch also facilitates access to data for each HIT, including pay, estimated length, and title.

The study was an analysis of previously collected metadata, which were analyzed anonymously. We complied with the terms of service for all data collected from CloudResearch, and MTurk. The approving institutional review board for this study was IntegReview.

Analytic sample

We analyzed the nearly 5 million tasks completed during an 18-month period between January 2016 and June 2017 by 12,312 female and 9,959 male workers who had complete data on key demographic characteristics. To be included in the analysis a HIT had to be fully completed, not just accepted, by the worker, and had to be accepted (paid for) by the requester. Although the vast majority of HITs were open to both males and females, a small percentage of HITs are intended for a specific gender. Because our goal was to exclusively analyze HITs for which the requesters did not know the gender of workers, we excluded any HITs using gender-specific inclusion or exclusion criteria from the analyses. In addition, we removed from the analysis any HITs that were part of follow-up studies in which it would be possible for the requester to know the gender of the worker from the prior data collection. Finally, where possible, CloudResearch tracks demographic information on workers across multiple HITs over time. To minimize misclassification of gender, we excluded the 0.3% of assignments for which gender was unknown with at least 95% consistency across HITs.

The main exposure variable is worker gender and the outcome variables are estimated actual hourly pay accrued through completing HITs, and advertised hourly pay for completed HITs. Estimated actual hourly wages are based on the estimated length in minutes and compensation in dollars per HIT as posted on the dashboard by the requester. We refer to actual pay as estimated because sometimes people work multiple assignments at the same time (which is allowed on the platform), or may simultaneously perform other unrelated activities and therefore not work on the HIT the entire time the task is open. We also considered several covariates to approximate human capital factors that could potentially influence earnings on this platform, including marital status, education, household income, number of children, race/ethnicity, age, and experience (number of HITs previously completed). Additional covariates included task length, task cluster (see below), and the serial order with which workers accepted the HIT in order to account for potential differences in HIT acceptance speed that may relate to the pay gap.

Database and analytic approach

Data were exported from CloudResearch’s database into Stata in long-form format to represent each task on a single row. For the purposes of this paper, we use “HIT” and “study” interchangeably to refer to a study put up on the MTurk dashboard which aims to collect data from multiple participants. A HIT or study consist of multiple “assignments” which is a single task completed by a single participant. Columns represented variables such as demographic information, payment, and estimated HIT length. Column variables also included unique IDs for workers, HITs (a single study posted by a requester), and requesters, allowing for a multi-level modeling analytic approach with assignments nested within workers. Individual assignments (a single task completed by a single worker) were the unit of analysis for all models.

Linear regression models were used to calculate the gender pay gap using two dependent variables 1) women’s estimated actual earnings relative to men’s and 2) women’s selection of tasks based on advertised earnings relative to men’s. We first examined the actual pay model, to see the gender pay gap when including an estimate of task completion speed, and then adjusted this model for advertised hourly pay to determine if and to what extent a propensity for men to select more remunerative tasks was evident and driving any observed gender pay gap. We additionally ran separate models using women’s advertised earnings relative to men’s as the dependent variable to examine task selection effects more directly. The fully adjusted models controlled for the human capital-related covariates, excluding household income and education which were balanced across genders. These models also tested for interactions between gender and each of the covariates by adding individual interaction terms to the adjusted model. To control for within-worker clustering, Huber-White standard error corrections were used in all models.

Cluster analysis

To explore the potential influence of any residual task heterogeneity and gender preference for specific task type as the cause of the gender pay gap, we use K-means clustering analysis (seed = 0) to categorize the types of tasks into clusters based on the descriptions that workers use to choose the tasks they perform. We excluded from this clustering any tasks which contained certain gendered words (such as “male”, “female”, etc.) and any tasks which had fewer than 30 respondents. We stripped out all punctuation, symbols and digits from the titles, so as to remove any reference to estimated compensation or duration. The features we clustered on were the presence or absence of 5,140 distinct words that appeared across all titles. We then present the distribution of tasks across these clusters as well as average pay by gender and the gender pay gap within each cluster.

The demographics of the analytic sample are presented in Table 1 . Men and women completed comparable numbers of tasks during the study period; 2,396,978 (48.6%) for men and 2,539,229 (51.4%) for women.

In Table 2 we measure the differences in remuneration between genders, and then decompose any observed pay gap into task completion speed, task selection, and then demographic and structural factors. Model 1 shows the unadjusted regression model of gender differences in estimated actual pay, and indicates that, on average, tasks completed by women paid 60 (10.5%) cents less per hour compared to tasks completed by men (t = 17.4, p < .0001), with the mean estimated actual pay across genders being $5.70 per hour.

*Model adjusted for race, marital status, number of children and task clusters as categorical covariates, and age, HIT acceptance speed, and number of HITs as continuous covariates.

In Model 2, adjusting for advertised hourly pay, the gender pay gap dropped to 46 cents indicating that 14 cents of the pay gap is attributable to gender differences in the selection of tasks (t = 8.6, p < .0001). Finally, after the inclusion of covariates and their interactions in Model 3, the gender pay differential was further attenuated to 32 cents (t = 6.7, p < .0001). The remaining 32 cent difference (56.6%) in earnings is inferred to be attributable to gender differences in HIT completion speed.

Task selection analyses

Although completion speed appears to account for a significant portion of the pay gap, of particular interest are gender differences in task selection. Beyond structural factors such as education, household composition and completion speed, task selection accounts for a meaningful portion of the gender pay gap. As a reminder, the pay rate and expected completion time are posted for every HIT, so why women would select less remunerative tasks on average than men do is an important question to explore. In the next section of the paper we perform a set of analyses to examine factors that could account for this observed gender difference in task selection.

Advertised hourly pay

To examine gender differences in task selection, we used linear regression to directly examine whether the advertised hourly pay differed for tasks accepted by male and female workers. We first ran a simple model ( Table 3 ; Model 3A) on the full dataset of 4.93 million HITs, with gender as the predictor and advertised hourly pay as the outcome including no other covariates. The unadjusted regression results (Model 4) shown in Table 3 , indicates that, summed across all clusters and demographic groups, tasks completed by women were advertised as paying 28 cents (95% CI: $0.25-$0.31) less per hour (5.8%) compared to tasks completed by men (t = 21.8, p < .0001).

*Models adjusted for race, marital status, number of children, and task clusters as categorical covariates, and age, HIT acceptance speed, and number of HITs as continuous covariates.

Model 5 examines whether the remuneration differences for tasks selected by men and women remains significant in the presence of multiple covariates included in the previous model and their interactions. The advertised pay differential for tasks selected by women compared to men was attenuated to 21 cents (4.3%), and remained statistically significant (t = 9.9, p < .0001). This estimate closely corresponded to the inferred influence of task selection reported in Table 2 . Tests of gender by covariate interactions were significant only in the cases of age and marital status; the pay differential in tasks selected by men and women decreased with age and was more pronounced among single versus currently or previously married women.

To further examine what factors may account for the observed gender differences in task selection we plotted the observed pay gap within demographic and other covariate groups. Table 4 shows the distribution of tasks completed by men and women, as well as mean earnings and the pay gap across all demographic groups, based on the advertised (not actual) hourly pay for HITs selected (hereafter referred to as “advertised hourly pay” and the “advertised pay gap”). The average task was advertised to pay $4.88 per hour (95% CI $4.69, $5.10).

The pattern across demographic characteristics shows that the advertised hourly pay gap between genders is pervasive. Notably, a significant advertised gender pay gap is evident in every level of each covariate considered in Table 4 , but more pronounced among some subgroups of workers. For example, the advertised pay gap was highest among the youngest workers ($0.31 per hour for workers age 18–29), and decreased linearly with age, declining to $0.13 per hour among workers age 60+. Advertised houry gender pay gaps were evident across all levels of education and income considered.

To further examine the potential influence of human capital factors on the advertised hourly pay gap, Table 5 presents the average advertised pay for selected tasks by level of experience on the CloudResearch platform. Workers were grouped into 4 experience levels, based on the number of prior HITs completed: Those who completed fewer than 100 HITs, between 100 and 500 HITs, between 500 and 1,000 HITs, and more than 1,000 HITs. A significant gender difference in advertised hourly pay was observed within each of these four experience groups. The advertised hourly pay for tasks selected by both male and female workers increased with experience, while the gender pay gap decreases. There was some evidence that male workers have more cumulative experience with the platform: 43% of male workers had the highest level of experience (previously completing 1,001–10,000 HITs) compared to only 33% of women.

Table 5 also explores the influence of task heterogeneity upon HIT selection and the gender gap in advertised hourly pay. K-means clustering was used to group HITs into 20 clusters initially based on the presence or absence of 5,140 distinct words appearing in HIT titles. Clusters with fewer than 50,000 completed tasks were then excluded from analysis. This resulted in 13 clusters which accounted for 94.3% of submitted work assignments (HITs).

The themes of all clusters as well as the average hourly advertised pay for men and women within each cluster are presented in the second panel of Table 5 . The clusters included categories such as Games, Decision making, Product evaluation, Psychology studies, and Short Surveys. We did not observe a gender preference for any of the clusters. Specifically, for every cluster, the proportion of males was no smaller than 46.6% (consistent with the slightly lower proportion of males on the platform, see Table 1 ) and no larger than 50.2%. As shown in Table 5 , the gender pay gap was observed within each of the clusters. These results suggest that residual task heterogeneity, a proxy for occupational segregation, is not likely to contribute to a gender pay gap in this market.

Task length was defined as the advertised estimated duration of a HIT. Table 6 presents the advertised hourly gender pay gaps for five categories of HIT length, which ranged from a few minutes to over 1 hour. Again, a significant advertised hourly gender pay gap was observed in each category.

Finally, we conducted additional supplementary analyses to determine if other plausible factors such as HIT timing could account for the gender pay gap. We explored temporal factors including hour of the day and day of the week. Each completed task was grouped based on the hour and day in which it was completed. A significant advertised gender pay gap was observed within each of the 24 hours of the day and for every day of the week demonstrating that HIT timing could not account for the observed gender gap (results available in Supplementary Materials).

In this study we examined the gender pay gap on an anonymous online platform across an 18-month period, during which close to five million tasks were completed by over 20,000 unique workers. Due to factors that are unique to the Mechanical Turk online marketplace–such as anonymity, self-selection into tasks, relative homogeneity of the tasks performed, and flexible work scheduling–we did not expect earnings to differ by gender on this platform. However, contrary to our expectations, a robust and persistent gender pay gap was observed.

The average estimated actual pay on MTurk over the course of the examined time period was $5.70 per hour, with the gender pay differential being 10.5%. Importantly, gig economy platforms differ from more traditional labor markets in that hourly pay largely depends on the speed with which tasks are completed. For this reason, an analysis of gender differences in actual earned pay will be affected by gender differences in task completion speed. Unfortunately, we were not able to directly measure the speed with which workers complete tasks and account for this factor in our analysis. This is because workers have the ability to accept multiple HITs at the same time and multiple HITs can sit dormant in a queue, waiting for workers to begin to work on them. Therefore, the actual time that many workers spend working on tasks is likely less than what is indicated in the metadata available. For this reason, the estimated average actual hourly rate of $5.70 is likely an underestimate and the gender gap in actual pay cannot be precisely measured. We infer however, by the residual gender pay gap after accounting for other factors, that as much as 57% (or $.32) of the pay differential may be attributable to task completion speed. There are multiple plausible explanations for gender differences in task completion speed. For example, women may be more meticulous at performing tasks and, thus, may take longer at completing them. There may also be a skill factor related to men’s greater experience on the platform (see Table 5 ), such that men may be faster on average at completing tasks than women.

However, our findings also revealed another component of a gender pay gap on this platform–gender differences in the selection of tasks based on their advertised pay. Because the speed with which workers complete tasks does not impact these estimates, we conducted extensive analyses to try to explain this gender gap and the reasons why women appear on average to be selecting tasks that pay less compared to men. These results pertaining to the advertised gender pay gap constitute the main focus of this study and the discussion that follows.

The overall advertised hourly pay was $4.88. The gender pay gap in the advertised hourly pay was $0.28, or 5.8% of the advertised pay. Once a gender earnings differential was observed based on advertised pay, we expected to fully explain it by controlling for key structural and individual-level covariates. The covariates that we examined included experience, age, income, education, family composition, race, number of children, task length, the speed of accepting a task, and thirteen types of subtasks. We additionally examined the time of day and day of the week as potential explanatory factors. Again, contrary to our expectations, we observed that the pay gap persisted even after these potential confounders were controlled for. Indeed, separate analyses that examined the advertised pay gap within each subcategory of the covariates showed that the pay gap is ubiquitous, and persisted within each of the ninety sub-groups examined. These findings allows us to rule out multiple mechanisms that are known drivers of the pay gap in traditional labor markets and other gig economy marketplaces. To our knowledge this is the only study that has observed a pay gap across such diverse categories of workers and conditions, in an anonymous marketplace, while simultaneously controlling for virtually all variables that are traditionally implicated as causes of the gender pay gap.

Individual-level factors

Individual-level factors such as parental status and family composition are a common source of the gender pay gap in traditional labor markets [ 15 ] . Single mothers have previously been shown to have lower reservation wages compared to other men and women [ 21 ]. In traditional labor markets lower reservation wages lead single mothers to be willing to accept lower-paying work, contributing to a larger gender pay gap in this group. This pattern may extend to gig economy markets, in which single mothers may look to online labor markets as a source of supplementary income to help take care of their children, potentially leading them to become less discriminating in their choice of tasks and more willing to work for lower pay. Since female MTurk workers are 20% more likely than men to have children (see Table 1 ), it was critical to examine whether the gender pay gap may be driven by factors associated with family composition.

An examination of the advertised gender pay gap among individuals who differed in their marital and parental status showed that while married workers and those with children are indeed willing to work for lower pay (suggesting that family circumstances do affect reservation wages and may thus affect the willingness of online workers to accept lower-paying online tasks), women’s hourly pay is consistently lower than men’s within both single and married subgroups of workers, and among workers who do and do not have children. Indeed, contrary to expectations, the advertised gender pay gap was highest among those workers who are single, and among those who do not have any children. This observation shows that it is not possible for parental and family status to account for the observed pay gap in the present study, since it is precisely among unmarried individuals and those without children that the largest pay gap is observed.

Age was another factor that we considered to potentially explain the gender pay gap. In the present sample, the hourly pay of older individuals is substantially lower than that of younger workers; and women on the platform are five years older on average compared to men (see Table 1 ). However, having examined the gender pay gap separately within five different age cohorts we found that the largest pay gap occurs in the two youngest cohort groups: those between 18 and 29, and between 30 and 39 years of age. These are also the largest cohorts, responsible for 64% of completed work in total.

Younger workers are also most likely to have never been married or to not have any children. Thus, taken together, the results of the subgroup analyses are consistent in showing that the largest pay gap does not emerge from factors relating to parental, family, or age-related person-level factors. Similar patterns were found for race, education, and income. Specifically, a significant gender pay gap was observed within each subgroup of every one of these variables, showing that person-level factors relating to demographics are not driving the pay gap on this platform.

Experience is a factor that has an influence on the pay gap in both traditional and gig economy labor markets [ 20 ] . As noted above, experienced workers may be faster and more efficient at completing tasks in this platform, but also potentially more savvy at selecting more remunerative tasks compared to less experienced workers if, for example, they are better at selecting tasks that will take less time to complete than estimated on the dashboard [ 20 ]. On MTurk, men are overall more experienced than women. However, experience does not account for the gender gap in advertised pay in the present study. Inexperienced workers comprise the vast majority of the Mechanical Turk workforce, accounting for 67% of all completed tasks (see Table 5 ). Yet within this inexperienced group, there is a consistent male earning advantage based on the advertised pay for tasks performed. Further, controlling for the effect of experience in our models has a minimal effect on attenuating the gender pay gap.

Another important source of the gender pay gap in both traditional and gig economy labor markets is task heterogeneity. In traditional labor markets men are disproportionately represented in lucrative fields, such as those in the tech sector [ 23 ]. While the workspace within MTurk is relatively homogeneous compared to the traditional labor market, there is still some variety in the kinds of tasks that are available, and men and women may have been expected to have preferences that influence choices among these.

To examine whether there is a gender preference for specific tasks, we systematically analyzed the textual descriptions of all tasks included in this study. These textual descriptions were available for all workers to examine on their dashboards, along with information about pay. The clustering algorithm revealed thirteen categories of tasks such as games, decision making, several different kinds of survey tasks, and psychology studies.We did not observe any evidence of gender preference for any of the task types. Within each of the thirteen clusters the distribution of tasks was approximately equally split between men and women. Thus, there is no evidence that women as a group have an overall preference for specific tasks compared to men. Critically, the gender pay gap was also observed within each one of these thirteen clusters.

Another potential source of heterogeneity is task length. Based on traditional labor markets, one plausible hypothesis about what may drive women’s preferences for specific tasks is that women may select tasks that differ in their duration. For example, women may be more likely to use the platform for supplemental income, while men may be more likely to work on HITs as their primary income source. Women may thus select shorter tasks relative to their male counterparts. If the shorter tasks pay less money, this would result in what appears to be a gender pay gap.

However, we did not observe gender differences in task selection based on task duration. For example, having divided tasks into their advertised length, the tasks are preferred equally by men and women. Furthermore, the shorter tasks’ hourly pay is substantially higher on average compared to longer tasks.

Additional evidence that scheduling factors do not drive the gender pay gap is that it was observed within all hourly and daily intervals (See S1 and S2 Tables in Appendix). These data are consistent with the results presented above regarding personal level factors, showing that the majority of male and female Mechanical Turk workers are single, young, and have no children. Thus, while in traditional labor markets task heterogeneity and labor segmentation is often driven by family and other life circumstances, the cohort examined in this study does not appear to be affected by these factors.

Practical implications of a gender pay gap on online platforms for social and behavioral science research

The present findings have important implications for online participant recruitment in the social and behavioral sciences, and also have theoretical implications for understanding the mechanisms that give rise to the gender pay gap. The last ten years have seen a revolution in data collection practices in the social and behavioral sciences, as laboratory-based data collection has slowly and steadily been moving online [ 16 , 24 ]. Mechanical Turk is by far the most widely used source of human participants online, with thousands of published peer-reviewed papers utilizing Mechanical Turk to recruit at least some of their human participants [ 25 ]. The present findings suggest both a challenge and an opportunity for researchers utilizing online platforms for participant recruitment. Our findings clearly reveal for the first time that sampling research participants on anonymous online platforms tends to produce gender pay inequities, and that this happens independent of demographics or type of task. While it is not clear from our findings what the exact cause of this inequity is, what is clear is that the online sampling environment produces similar gender pay inequities as those observed in other more traditional labor markets, after controlling for relevant covariates.

This finding is inherently surprising since many mechanisms that are known to produce the gender pay gap in traditional labor markets are not at play in online microtasks environments. Regardless of what the generative mechanisms of the gender pay gap on online microtask platforms might be, researchers may wish to consider whether changes in their sampling practices may produce more equitable pay outcomes. Unlike traditional labor markets, online data collection platforms have built-in tools that can allow researchers to easily fix gender pay inequities. Researchers can simply utilize gender quotas, for example, to fix the ratio of male and female participants that they recruit. These simple fixes in sampling practices will not only produce more equitable pay outcomes but are also most likely advantageous for reducing sampling bias due to gender being correlated with pay. Thus, while our results point to a ubiquitous discrepancy in pay between men and women on online microtask platforms, such inequities have relatively easy fixes on online gig economy marketplaces such as MTurk, compared to traditional labor markets where gender-based pay inequities have often remained intractable.

Other gig economy markets

As discussed in the introduction, a gender wage gap has been demonstrated on Uber, a gig economy transportation marketplace [ 20 ], where men earn approximately 7% more than women. However, unlike in the present study, the gender wage gap on Uber was fully explained by three factors; a) driving speed predicted higher wages, with men driving faster than women, b) men were more likely than women to drive in congested locations which resulted in better pay, c) experience working for Uber predicted higher wages, with men being more experienced. Thus, contrary to our findings, the gender wage gap in gig economy markets studied thus far are fully explained by task heterogeneity, experience, and task completion speed. To our knowledge, the results presented in the present study are the first to show that the gender wage gap can emerge independent of these factors.

Generalizability

Every labor market is characterized by a unique population of workers that are almost by definition not a representation of the general population outside of that labor market. Likewise, Mechanical Turk is characterized by a unique population of workers that is known to differ from the general population in several ways. Mechanical Turk workers are younger, better educated, less likely to be married or have children, less likely to be religious, and more likely to have a lower income compared to the general United States population [ 24 ]. The goal of the present study was not to uncover universal mechanisms that generate the gender pay gap across all labor markets and demographic groups. Rather, the goal was to examine a highly unique labor environment, characterized by factors that should make this labor market immune to the emergence of a gender pay gap.

Previous theories accounting for the pay gap have identified specific generating mechanisms relating to structural and personal factors, in addition to discrimination, as playing a role in the emergence of the gender pay gap. This study examined the work of over 20,000 individuals completing over 5 million tasks, under conditions where standard mechanisms that generate the gender pay gap have been controlled for. Nevertheless, a gender pay gap emerged in this environment, which cannot be accounted for by structural factors, demographic background, task preferences, or discrimination. Thus, these results reveal that the gender pay gap can emerge—in at least some labor markets—in which discrimination is absent and other key factors are accounted for. These results show that factors which have been identified to date as giving rise to the gender pay gap are not sufficient to explain the pay gap in at least some labor markets.

Potential mechanisms

While we cannot know from the results of this study what the actual mechanism is that generates the gender pay gap on online platforms, we suggest that it may be coming from outside of the platform. The particular characteristics of this labor market—such as anonymity, relative task homogeneity, and flexibility—suggest that, everything else being equal, women working in this platform have a greater propensity to choose less remunerative opportunities relative to men. It may be that these choices are driven by women having a lower reservation wage compared to men [ 21 , 26 ]. Previous research among student populations and in traditional labor markets has shown that women report lower pay or reward expectations than men [ 27 – 29 ]. Lower pay expectations among women are attributed to justifiable anticipation of differential returns to labor due to factors such as gender discrimination and/or a systematic psychological bias toward pessimism relative to an overly optimistic propensity among men [ 30 ].

Our results show that even if the bias of employers is removed by hiding the gender of workers as happens on MTurk, it seems that women may select lower paying opportunities themselves because their lower reservation wage influences the types of tasks they are willing to work on. It may be that women do this because cumulative experiences of pervasive discrimination lead women to undervalue their labor. In turn, women’s experiences with earning lower pay compared to men on traditional labor markets may lower women’s pay expectations on gig economy markets. Thus, consistent with these lowered expectations, women lower their reservation wages and may thus be more likely than men to settle for lower paying tasks.

More broadly, gender norms, psychological attributes, and non-cognitive skills, have recently become the subject of investigation as a potential source for the gender pay gap [ 3 ], and the present findings indicate the importance of such mechanisms being further explored, particularly in the context of task selection. More research will be required to explore the potential psychological and antecedent structural mechanisms underlying differential task selection and expectations of compensation for time spent on microtask platforms, with potential relevance to the gender pay gap in traditional labor markets as well. What these results do show is that pay discrepancies can emerge despite the absence of discrimination in at least some circumstances. These results should be of particular interest for researchers who may wish to see a more equitable online labor market for academic research, and also suggest that novel and heretofore unexplored mechanisms may be at play in generating these pay discrepancies.

A final note about framing: we are aware that explanations of the gender pay gap that invoke elements of women’s agency and, more specifically, “choices” risk both; a) diminishing or distracting from important structural factors, and b) “naturalizing” the status quo of gender inequality [ 30 ] . As Connor and Fiske (2019) argue, causal attributions for the gender pay gap to “unconstrained choices” by women, common as part of human capital explanations, may have the effect, intended or otherwise, of reinforcing system-justifying ideologies that serve to perpetuate inequality. By explicitly locating women’s economic decision making on the MTurk platform in the broader context of inegalitarian gender norms and labor market experiences outside of it (as above), we seek to distance our interpretation of our findings from implicit endorsement of traditional gender roles and economic arrangements and to promote further investigation of how the observed gender pay gap in this niche of the gig economy may reflect both broader gender inequalities and opportunities for structural remedies.

Supporting information

Funding statement.

The authors received no specific funding for this work.

Data Availability

Why does the gender wage gap persist? ‘Comparable worth’ bias

Kadie Philp, commissioner and CAO of Ontario’s Pay Equity Commission, says “structural and unconscious bias influences the way we value the contributions made by women and men in the labour market.” Jessica Lee

Is the work performed by a daycare worker as valuable as that of an electrician? Is a restaurant server’s work as valuable as an IT professional’s?

More to the point: Is the work historically or stereotypically done by women as valuable as the work historically or stereotypically done by men?

These are questions Ontario’s pay equity process answers. Seeking to redress systemic discrimination in wages, the process of pay equity is designed to close gender wage gaps by providing equal pay for work of equal value. While the gender wage gap in Canada is slowly narrowing, many women in Canada today continue to be paid less than men. Even with gains made in equity for women in the workplace (the wage gap decreased from 16 per cent in 2007 to 12 per cent in 2022), jobs predominantly held by women, such as librarians or child-care workers, persist in being lower paid than those typically held by men.

This disparity is due in part to biases around the “comparable worth” of different jobs, says Kadie Philp, commissioner and CAO of Ontario’s Pay Equity Commission .

“Women have historically been clustered in specific types of jobs, and those jobs have been undervalued for various reasons, some linked to early structures of the labour market during the industrial revolution and others linked to prescribed gender roles that saw women’s participation in the labour market as an extension of their unpaid domestic work,” says Ms. Philp, who also hosts a podcast called Level the Paying Field .

“Through data analysis, we can see how both structural and unconscious bias influences the way we value the contributions made by men and women in the labour market.”

For example, research has shown that when women move into a formerly male-dominated sector, wages decrease, says Ms. Philp. She cites one study that found when women moved into computer design, wages in that job class fell 34 per cent.

Ms. Philp says the real and significant cost of comparable worth bias is that companies are “losing out on attracting and retaining talent by not recognizing and rewarding their contributions equitably.”

Leveraging data to mitigate bias

Dr. Anthony Greenwald, professor emeritus of the University of Washington and co-author of Blind Spot, Hidden Biases of Good People, has been studying the concept of bias for decades. He says that implicit biases, or unconsciously held negative beliefs against a specific social group, are typically developed in early childhood. “People don’t acquire them voluntarily and aren’t aware they are acquiring them,” he says.

To help people educate themselves on their own implicit biases, Dr. Greenwald and a colleague co-created the Implicit Association Test in 1998. Users are asked to make a series of choices as rapidly as possible. The result they receive is based on the speed of the response. It has since been taken by millions of users and used in multiple research studies.

“Later research showed what it was measuring was importantly related to discriminatory judgement and behaviour,” says Dr. Greenwald, a recent guest on the Level the Playing Field podcast.

While his research also found that implicit bias is impossible to change, he says that one way to mitigate its effects is for organizations to apply an epidemiological-like study to examine their gender wage gap. The process he outlines leads to uncovering disparity based on bias and correcting it.

“You take a large group of people and sort them by age, race or ethnicity to see if there are differences. Almost always, you will find differences,” he says. “The problem is that most employers don’t do it routinely.”

Ms. Philp says she feels many employers have made substantive efforts in trying to address systemic or implicit discrimination in their compensation, promotion and hiring practices. But, like Dr. Greenwald, she says “companies could be doing better in leveraging the data they have.”

If disparities are found related to race, gender, socioeconomic status or age, companies can correct them, says Ms. Philp. “That is exactly what the pay equity process does.”

System-wide efforts to overcome bias in comparable worth have been ongoing, including the establishment of federal and provincial pay equity legislation in a number of provinces including Ontario and Quebec . Ms. Philp notes that over the past 20 years, the gender wage gap has narrowed in every province in Canada, ranging from six to 11 per cent.

“[Gains] haven’t been huge, but it shows there is traction and attention is being paid to this in the labour market,” she says.

One way for companies to address pay equity is for companies to use comparable worth to evaluate jobs and determine their value to a company or business, Ms. Philp says. This involves a gender-neutral ranking of the value of comparable skills across occupations, regardless of sex. If jobs traditionally held by women are determined to have comparable worth to jobs held primarily by men, they should be paid commensurably.

Holding employers’ feet to the fire

Another way to address gender wage gaps is for provinces to introduce pay transparency legislation, which would require salaries be included on any publicly advertised job posting. So far, B.C. has introduced this legislation in 2023 and Ontario is poised to introduce such a requirement in 2024.

Laura Williams, managing partner of Williams HR Law LLP, says while there may be some initial bristling by companies at salary disclosure requirements, it has potential to be an important piece in reducing gender bias in pay.

Companies generally do not have their compensation practices in order, Ms. Williams says, “so they’re flying by the seat of their pants, and this has become a challenge.”

Pay transparency laws will hold employers’ feet to the fire, causing companies to become a “lot more methodical around setting their pay [rates].”

And, she adds, pay transparency has the potential to improve pay equity because it will assist in the pay negotiation process, an area where men have traditionally been stronger negotiators.

“I believe it will make a change. I don’t think it will be the be all and end all,” says Ms. Williams.

Ms. Philp says she has seen results in some jurisdictions that require strict reporting of any pay disparities that are found during pay equity analysis and posting of those results. “Certainly where there is rigorous reporting requirements, the wage gap is closing more effectively,” she says.

But she is hopeful that greater awareness of pay equity will change the landscape, rather than harsher penalties, as more companies voluntarily move in the right direction.

“We have to celebrate those successes and thank the companies and leaders who have committed to this agenda, so we can have a vibrant and healthy and fairly compensated labour market.”

Advertising feature produced by Globe Content Studio with Pay Equity Office. The Globe’s editorial department was not involved.

Interact with The Globe

Women in Hawaiʻi face average 50% gender pay gap

• March 12, 2024

The average male in Hawaiʻi out-earned the average female by 50%, a smaller margin than the 69% gap observed nationally, but according to a new University of Hawaiʻi Economic Research Organization ( UHERO ) blog , the margin is “still very large.” UHERO cites data from the American Community Survey spanning 2015 to 2022.

“These aggregate figures, however, overlook crucial factors such as educational attainment and occupation, which significantly impact earnings,” wrote UHERO Research Economist Rachel Inafuku, the author of the blog. “An accurate measure of the gender pay gap needs to consider whether women with similar skill levels and educational backgrounds earn less than their male counterparts.”

When examining earnings across education levels from 2015 to 2022, lifetime earnings are consistently higher for men than women, even within the same level of education. The smallest discrepancy is found among individuals with professional or doctoral degrees, where men earn 33% to 35% more than women holding the same degree.

The most substantial gap surfaces among those who attended college but didn’t complete their degree: men in this group earn 63% more than their female counterparts. For a woman to match the lifetime earnings of a man with some college experience but no degree, she would need to attain at least a master’s degree. At every educational tier, women in the subsequent degree bracket fail to surpass the earnings of men at the previous education level.

Pay gap across occupations

Among the 130 occupations examined in UHERO ’s sample, men earned more than women in 82% of these careers, while 18% of these professions saw women earning more. Occupations such as financial managers, chief executives, and pharmacists exhibited some of the lowest ratios of female to male earnings, indicating that women working these jobs earn much less than men. Conversely, roles like hosts/hostesses, bartenders, and paralegals show the highest ratios of female to male earnings.

The gender pay gap has been a focal point in economics especially within the past year when Claudia Goldin won the Nobel Prize for uncovering key drivers of gender differences in the labor market. While female labor participation increased substantially over the past century, Goldin found that the earnings gap between men and women in the U.S. hardly closed over a long period of time.

What drives Hawaiʻi ’s gender pay gap?

Several national studies have found that women are more likely to prioritize home activities, such as reducing their work hours to care for their children, which often results in career sacrifices.

“Here in Hawaiʻi , the data aligns with Goldin’s findings,” Inafuku wrote. “Throughout their mid to late 20’s, men earn slightly more than women. After that, this pay differential widens as men’s wages grow at a much faster rate than women’s, and the gap continues over the course of a worker’s career.”

Read the entire blog on UHERO ’s website .

UHERO is housed in UH Mānoa’s College of Social Sciences .

Related Posts:

• Gender pay gap present in 74% of Hawaiʻi jobs, more…
• Future of high stock prices, market discussed by…
• Universal preschool in Hawaiʻi: benefits now and later
• previous post: Navigator Nainoa Thompson to be honored at international meeting of geographers
• next post: New renderings preview UH Cancer Center’s clinical trials research hub

If required, information contained on this website can be made available in an alternative format upon request. Get Adobe Acrobat Reader

About Calendar COVID-19 Updates Directory Emergency Information For Media MyUH Work at UH

Gagana Samoa

Kapasen Chuuk

Kajin Majôl

ʻŌlelo Hawaiʻi

• Administrative

Tochigi Has Largest Gender Pay Gap in Japan

男女の賃金差、最大は栃木　全国順位公表、最小は高知―政府ｐｔ.

Related News

2024.09.03 08:38 Harris Opposes Nippon Steel's Plan to Buy U.S. Steel

2024.09.03 07:22 Japan Companies' Internal Reserves Top 600 T. Yen for 1st Time

2024.09.02 23:32 Japan's Satomi Wins 2nd Straight Paralympic Badminton Gold

水星探査機、到着１年延期＝エンジン不調で―ＪＡＸＡなど

日本が白星発進＝サッカーＵ２０女子Ｗ杯

山口が左膝手術＝Ｊ１神戸

西武の岡田が現役引退＝プロ野球

人気ゲームの世界大会、アジア初開催へ＝札幌市〔地域〕

日本人メダリスト〔パリパラリンピック〕

【近代五種】佐藤大宗〔パリ五輪〕

【アーティスティックスイミング】日本代表〔パリ五輪〕

【ゴルフ】山下美夢有〔パリ五輪〕

レスリング〔パリ五輪〕

【スポーツクライミング】安楽宙斗〔パリ五輪〕

【カヌー】羽根田卓也〔パリ五輪〕

2024.09.03 14:24 ニュース 京都国際が優勝報告会　全国高校野球

2024.09.02 17:49 ニュース 日本人初の大リーガー　村上雅則さんが講演

2024.09.02 16:45 ニュース 井上尚ら計量パス　３日に世界４団体防衛戦　ボクシング

2024.09.02 16:29 芸能・エンタメ パリ五輪スケボー日本代表・小野寺吟雲選手、「ニベアクリーム」の新CMに出演（メイキング インタビュー／小野寺吟雲 アイナ・ジ・エンド）

リアルタイムランキング

• 1 奄美のマングース、根絶を宣言＝６年以上捕獲なし―環境省
• 2 円相場、１４６円６３～６４銭＝３日正午現在
• 3 イスラエル武器輸出、一部停止
• 4 車いすラグビーの日本、初の金＝バド男子単の梶原連覇―パリ・パラリンピック〔パラリンピック〕
• 5 大谷、２安打に３盗塁　大リーグ
• 2 魚大量死で非常事態宣言　ギリシャ
• 3 競泳の木村が金メダル＝ボッチャの杉村、準々決勝敗退―パリ・パラリンピック〔パラリンピック〕
• 4 競泳の窪田が銀＝車いすテニスの上地は２回戦へ―パリ・パラリンピック〔パラリンピック〕
• 5 広瀬前参院議員を在宅起訴＝秘書給与３５０万円詐取の罪―東京地検

ディープフェイク 08.30 15:49

アンモニア 08.30 14:38

百条委員会 08.30 14:20

断熱リフォーム 08.26 15:30

海洋地球研究船『みらい』 08.26 14:29