The gender gap in pay progression: job mobility and job ladders

Background

The median gender pay gap has declined dramatically in the UK from 36.4% in the 1970 (O’Reilly, Smith et al. 2015) to around 18% in the most recent data (ONS 2018). Still, by international standards the pay gap is high: the UK has the fourth largest gender pay gap in the EU and the eighth largest of OECD countries (OECD 2019).

Researchers and policy makers have focused on gender differences in education and labour market experience as the likely drivers of the pay gap. However, today these explanations no longer stand up to scrutiny. Women are on average better educated than men and they are much less likely to withdraw from the labour market for long periods of time. Nevertheless, women earn on average about 10% less than men even when they work full-time and have similar education and labour market experience.

While explanations focusing on women’s potential lower productivity as the cause of the gender pay gap have been thoroughly investigated and found inadequate, there is less evidence on the role played by employers. This research will contribute to addressing this gap.

The standard economic model of the labour market assumes that wages are determined by the market and that individual employers cannot choose the wages they offer to their employees. A different model assumes that for a variety of reasons competition is not perfect and employers have some discretion over the wages they offer. This wage setting power is likely to be weaker when workers are mobile. Mobile workers will leave an employer offering wages below the market rate. However, if workers are relatively immobile, employers can exploit this ‘immobility’ by offering them lower wages. If women are more constrained by family responsibilities in the types of jobs that they will take-up or in the amount of time and effort they can devote to job search, they will generally be more immobile and thus at a disadvantage. Women’s family responsibilities might be ultimately responsible for the gender pay gap but not because they limit their productivity but rather because they reduce their bargaining power with firms.

Project aims

This research project examines the role of employer wage-setting power in driving the gender pay gap in two ways. First, using data from the UK’s largest longitudinal study, it investigates the extent to which job-to-job mobility patterns differ between men and women, and whether any differences can explain the observed gender gap in pay progression. Second, it develops an index of employer wage-setting power based on geographical location, industry and cost of travel and test whether the index can explain gender differences in pay progression.

Tackling the gender pay gap is a widely shared goal among policy makers, political parties, women’s groups, trade-unions and employer organizations. A better understanding of the factors driving the gap is essential to design effective policies. For example, in April 2017, the UK government has mandated large employers report annually on the pay gap in their organization. If women’s lower productivity is to blame for the gender pay gap, such legislation is likely to be ineffective and even counterproductive. On the other hand, mandatory reporting is likely to be more effective if employers’ stronger wage-setting power is a significant factor behind the pay gap. More generally, if employers enjoy significant wage setting power relative to some of their employees, this has implications for legislation on anti-discrimination, the minimum wage, trade-unions and family policy.

Methods

To carry out our analyses, we use the UK Household Longitudinal Study (UKHLS), a large panel survey that follows approximately 40,000 households and interviews yearly all their adult members. In addition to a wide range of demographic and labour market information, the survey collects detailed information about the employment history between interviews (Poster-Vinay and Sepahsalari 2019) . Information is available on job to job changes, job to non-employment and non-employment to job transitions. Following the literature, we focus on gender differences in hourly pay. We estimate, separately for men and women, a series of hourly wage growth equations using fixed effects and random effects regressions. Our focus is on the effects of different types of job mobility on real wage growth while controlling for a wide range of individual characteristics, including family composition, employment and education history, and personality traits. We also examine the extent to which the effect of job mobility changes when parenthood is included among controls and whether the returns to mobility are different for parents and childless persons, and how these effects differ for men and women. The UKHLS collects information about weekly pay as well as the usual number of hours per week allowing for an indirect measure of hourly pay to be constructed. However, it is well known that this measure suffers from a significant amount of measurement error. We use results from previous work (Avram and Harkness 2018), where we have applied the methodology proposed by Skinner, Stuttard et al. (2002) to UKHLS data to obtain more accurate measures of hourly pay.

The labour market concentration literature has used the Herfindahl-Hirschman index (HHI) as a measure of employer wage-setting power (Azar, Marinescu et al. 2019). The HHI is calculated by squaring each firm’s employment share in a given labour market and summing these terms. It is high when there are relatively few firms in the market or when most of employment can be found in a few firms. Conversely, the index is low if there are many firms in the market and employment is relatively equally spread among them. By definition, indices based on smaller geographical areas are higher than indices based on larger areas.

Traditionally, labour markets have been defined by geography and/or occupation and industry. In this case, gender differences in concentration can arise only as a result of occupational/ industry segregation. To induce further variation in our concentration index along gender lines, we exploit the fact that women (and especially mothers) commute shorter distances (Roberts, Hodgson et al. 2011; Roberts and Taylor 2017; Jacob, Munford et al. 2019). Rather than calculating concentration indexes at a pre-specified geographical unit level, we propose a methodology similar to that described by Manning and Petrongolo (2017). Given a worker’s home census output area (COA), we define a series of concentric areas with a progressively larger radius centred on it. These areas include all the COAs within a specified distance (for ex: 5, 10, 15 km etc.) of our anchor COA. We then calculate the HHI corresponding to each area using only firms in our worker’s broad industry group. Thus, for each worker we obtain a series of HH indices that depend on the worker’s home location and on her industry. These indices are strictly ordered in the sense that indices corresponding to areas with a smaller radius are larger than indices corresponding to areas with a larger radius. We select the size of the radius based on the most frequent observed travelled distances in the region the worker lives and test the sensitivity of our results to alternative choices. To construct the HH indices, we use the Business Structure Database (BSD). BSD is derived from the Inter-Departmental Business Register and contains employment information along with geographical location and detailed industry for all businesses that either pay VAT or have at least one PAYE employee. It is estimated that it captures over 2 million enterprises accounting for 99% of the economic activity in Great Britain.

In the next step, we estimate a cost of distance function that depends on location and individual characteristics. We base this estimation on models developed in the transportation literature (Roberts, Hodgson et al. 2011; Jacob, Munford et al. 2019). Our dataset contains yearly information about the commuting time of all employees and biennial information about commuting distances. Following Jacob, Munford et al. (2019), we estimate the impact of commuting time on subjective well-being using individuals who have experienced exogenous shocks in commuting time, separately for men and women. In our specification, we control for characteristics known to influence both commuting choices and 6 subjective well-being, including mode of transport, region, income, and household composition. We allow the effects of commuting time to differ by parenthood status and by the age of the youngest child. We then use information about location and mode of transport to derive the relationship between commuting time and commuting distance. Using these equations, we predict for each worker the impact on subjective well-being of commuting the distances corresponding to our area radiuses (e.g. 5km, 10km, etc.). These predictions represent our individual cost of distance measures. Finally, we then use the estimated cost of distance function to weigh the HHI measures previously derived and construct a single individual level measure of labour market concentration.

We use our individual estimate of labour market concentration as a measure of employer wage-setting power and test whether it can explain wage growth differences between men and women. We first document gender differences in employer wage-setting power, as well as the extent to which these differences are associated with parenthood status, age, educational qualifications and occupation. We then augment our previously estimated hourly wage growth equations using fixed and random effects to include our index of individual labour market concentration. We test whether gender differences in bargaining power can explain observed differences in wage growth. We also examine whether higher market concentration is associated with less wage related job mobility and whether the effects on wage progression are different for leavers and stayers.

Read our explainer: Are women with families paying the price for a lack of mobility? Quantitative and qualitative gender differences in job mobility and how these influence pay

Read our working paper: Gender differences in job mobility and pay progression in the UK