Where do women earn more than men? Explaining regional differences in the gender pay gap

Fuchs, Michaela; Rossen, Anja; Weyh, Antje; Wydra‐Somaggio, Gabriele Article — Published Version Where do women earn more than men? Explaining regional differences in the gender pay gap Journal of Regional Science Provided in Cooperation with: John Wiley & Sons Suggested Citation: Fuchs, Michaela; Rossen, Anja; Weyh, Antje; Wydra‐Somaggio, Gabriele (2021) : Where do women earn more than men? Explaining regional differences in the gender pay gap, Journal of Regional Science, ISSN 1467-9787, Wiley, Hoboken, NJ, Vol. 61, Iss. 5, pp. 1065-1086, https://doi.org/10.1111/jors.12532 This Version is available at: https://hdl.handle.net/10419/284838 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/4.0/ J Regional Sci. 2021;61:1065–1086. wileyonlinelibrary.com/journal/jors | 1065 Received: 4 March 2021 | Accepted: 26 March 2021 DOI: 10.1111/jors.12532 ORIGINAL ARTICLE Wheredowomenearnmorethanmen?Explaining regional differences in the gender pay gap Michaela Fuchs 1 |Anja Rossen 2 |Antje Weyh 3 | Gabriele Wydra‐Somaggio 4 1 Regional Research Network, Institute for Employment Research (IAB) Saxony‐Anhalt‐ Thuringia, Halle (Saale), Germany 2 Regional Research Network, Institute for Employment Research (IAB) Bavaria, Nürnberg, Germany 3 Regional Research Network, Institute for Employment Research (IAB) Saxony, Chemnitz, Germany 4 Regional Research Network, Institute for Employment Research (IAB) Rhineland‐ Palatinate‐Saarland, Saarbrücken, Germany Correspondence Anja Rossen, Regional Research Network, Institute for Employment Research (IAB) Bavaria, Thomas‐Mann‐Straße 50, D‐90471 Nürnberg, Germany. Email: [email protected] Abstract This paper provides new evidence on the magnitude and determinants of regional differences in the gender pay gap. On the basis of a comprehensive data set of all full‐time employees in Germany, we explain the profound variation of the gender pay gap at a small‐scale level with theory‐based individual and job‐related characteristics. Using the Oaxaca–Blinder decomposition, we find pronounced spatial differences in the impact of the considered determinants. Whereas gender differences in job‐related characteristics are important drivers in regions with a high gender pay gap, individual characteristics come into effect in regions with a low and negative gap. The results underscore the role played by the establishment composition in a region and the kind of jobs provided for gendered earnings. KEYWORDS decomposition, gender pay gap, Germany, regional labor markets, wages 1|INTRODUCTION Regional inequalities in wages constitute a widespread phenomenon that is of great importance to policymakers and the general public alike. Though their general magnitude and determinants have been subject to intense research (e.g., Combes et al., 2008;López‐Bazo & Motellón, 2012; Pereira & Galego, 2011), much less is known about gender This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made. © 2021 The Authors. Journal of Regional Science published by Wiley Periodicals LLC differences in regional wages. Lower wages of women as illustrated by the gender pay gap (GPG) constitute a highly political and societal issue that persists not only between countries (OECD, 2017) but also between smaller regional units. Germany, for example, being characterized by striking regional disparities on the labor market (OECD, 2005), features a nationwide GPG of 20.1% that hides a range of‐4.3% to 40.3% on the level of the NUTS 3 regions. These profound disparities in the GPG clearly call for an explanation, as they entail unequal chances for men and women in the labor market and social context both across and within regions. However, only a few studies dealing with regional wage disparities take gender issues into account (e.g., Ammermüller et al., 2010; Duranton & Monastiriotis, 2002; Galego & Pereira, 2014), whereas studies that explicitly deal with gender differences in regional wages mainly focus on the contrast between rural and urban regions (Bacolod, 2017; Glaeser & Maré, 2001;Hirschetal.,2013;Nisic,2017). To the best of our knowledge, so far only Murillo Huertas et al. (2017) explicitly investigate regional differences in the GPG in a comprehensive way. They document significant spatial heterogeneity at the NUTS 2 level in Spain, which they trace back to spatial variation in economic, institutional, and demographic factors. As to the GPG itself, a vast body of literature has uncovered a variety of individual and establishment‐related determinants ranging from gender differences in human capital (Becker, 1964) and labor market experience (Combet & Oetsch, 2019; Manning & Swaffield, 2008) as well as in occupational choice (England, 1992; Marini, 1989; Polachek, 1981) to the impact of gender roles and noncognitive skills (Blau & Kahn, 2017) and the sorting into firms (Card et al., 2016; Barth et al., 2017). As these factors are implicitly assumed to be relevant for explaining the GPG at the national level, a fundamental question that arises in regional respect is if and to what degree they also hold at the subnational level. In close relation is the question to what extent gender‐specific wage differences depend on the structure of human capital and firm‐specific factors varying between regions. There are several reasons why this should be the case. Foremost, individual skills and labor market participation that go along with different social norms across regions (Yeandle, 2009) are not evenly distributed across locations (Combes & Gobillon, 2015). Furthermore, regions differ substantially in their sectoral and establishment composition, thereby providing different regional employment opportunity structures for men and women (Hanson & Pratt, 1995; Nisic, 2017; Perales & Vidal, 2015; Petrongolo & Ronchi, 2020). One explanation for these spatial disparities is agglomeration effects that increase productivity and hence wages by improving the quantity and quality of matches between workers and firms in dense labor markets (Glaeser & Maré, 2001). They affect men and women in different ways. For example, due to more and better job opportunities in dense agglomerations, women may experience a lower wage depreciation resulting from time out of employment than women in rural areas (Phimister, 2005). Related, agglomerative forces should lead to higher productivity of specific cognitive and social skills that women display relatively more often than men (Bacolod, 2017). A further factor pertains to the lower spatial mobility of women that restricts their job search to a smaller area and thus hinders them from getting better jobs and higher wages in other regions (Crane, 2007; Petrongolo & Ronchi, 2020). Consequently, the determinants that influence the GPG in general terms should also be distributed unevenly at the regional level. This paper provides new answers to the questions posed above by explaining regional differences in the GPG within Germany, extending the findings of Murillo Huertas et al. (2017) for Spain. To this, we apply the decomposition method of Oaxaca (1973) and Blinder (1973) for all 401 German NUTS 3 regions, which enable a more disaggregated analysis than the NUTS 2 regions used by Murillo Huertas et al. (2017). In addition, our analysis rests on microlevel data for the year 2017 with detailed information on the whole population of employees covered by the German social security system as well as on all establishments employing these persons. The paper contributes to the existing literature in several ways. First, we identify at the national level the role of decidedly regional factors that we consider in addition to the commonly used individual and establishment determinants of the GPG. The second and main contribution consists of examining for each region the impact of the factors identified in the literature as driving the GPG and how this impact differs across the regions. For this purpose, we compare the regional impacts of the factors to assess whether they are valid across all regions or only for specific groups of regions. The paper is structured as follows. Section 2introduces the data, and Section 3presents the decomposition method used in our analysis. Section 4derives the explanatory variables of the GPG against the background of 1066 | FUCHS ET AL. theoretical considerations and empirical evidence. In Section 5, we present the results. While Section 6contains robustness checks, Section 7concludes. 2|DATA Our main data set is the Employee History (BeH) of the Institute of Employment Research (IAB). It covers all employment notifications in Germany, providing information on all employees subject to social security and the establishments they work in. For each employee, the BeH contains a detailed set of sociodemographic and employment characteristics. For each establishment, information on its location and branch of industry is included. As they are not covered by social security, civil servants, persons in military service, and self‐employed are excluded. Due to legal sanctions for misreporting, the data is very reliable. The BeH brings along the decisive advantage of providing the population of employees and establishments in a small‐scale regional perspective, going beyond survey data used by the few related studies of Murillo Huertas et al. (2017), Bacolod (2017), or Nisic (2017). However, there are also some constraints that arise from the specific collection purpose. First, wages are right‐censored, because they are only reported up to the social security contribution limit. Toavoidbiasedestimates,wefollowGartner(2005) and impute censored wages with estimated wages. 1 Second, the German social security data does not contain information on the exact number of hours worked. This implies that we have to explicitly exclude one important source for gender differences in earnings that rests on female part‐time employment (Boll et al., 2016; Manning & Petrongolo, 2008). 2 As an advantage, however, the restriction on full‐time employees allows us to investigate the GPG for a more homogenous group. In addition, we control for the selection into employment by capturing biases induced by non‐employment (Olivetti & Petrongolo, 2008). To this, we estimate for each employee the probability of being full‐time employed by means of a probit model and include the resulting inverse mills ratio as an additional determinant of the GPG in the decomposition. 3 Third, due to the specific data collection purpose, we cannot directly control for marital status and the number of children that restrict women's labor supply (Blau & Kahn, 2017). To indirectly account for related employment interruptions, we additionally consider individual labor market biographies provided by the Integrated Employment Biographies (IEB) of the IAB. This data set contains information on employment spells, benefit receipts, participation in measures of active labor market policy, and job‐search status for every person on a daily basis. Additional information on the establishments comes from the Establishment History Panel (BHP) of the IAB that contains detailed information about all establishments in Germany with at least one employee liable to social security or with at least one marginal part‐time employee. For the construction of further regional variables, we use data from the Federal Employment Agency on the unemployment rates, the German Statistical Office on population and fertility, and the German Microcensus on persons with Catholic denomination. The regional dimension of our analysis is delineated along the 401 German NUTS 3 regions covering kreisfreie Städte and Landkreise. The BeH for 2017 encompasses about 35.4 million employees that we confine to full‐time employees between 15 and 64 years of age. We further disregard apprentices and persons with missing information on wages. To properly calculate variables at the establishment level, for example, wage dispersion, we exclude establishments with one employee. Our final data set encompasses 17,933,369 observations, of which are roughly 12.3 million men and 5.7 million women. 1 Nine percent of the employees in our data set have wages above these thresholds. For the imputation, we estimate the wages with a Tobit regression, using basically the same explanatory variables as in the wage regressions (Gartner, 2005). 2 We take up this issue in the robustness checks and conduct decompositions for full‐and part‐time workers using approximated information on working hours. 3 The probit model contains a set of variables which significantly explain the likelihood to be in employment, but do not explain the wages (e.g., care facilities for children younger than 3 years on the NUTS 3 level). FUCHS ET AL. | 1067 3|METHOD For explaining regional differences in the unadjusted GPG, we follow the seminal work of Oaxaca (1973) and Blinder (1973) that has been widely used in subsequent wage gap studies (e.g., Fortin et al., 2011) and in assessing regional wage disparities (e.g., Murillo Huertas et al., 2017; Pereira & Galego, 2011). The classical Oaxaca–Blinder (OB) decomposition focuses on the gap in mean earnings between male and female workers. Formally, it consists of two estimation steps. First, estimations of the determinants of wages that are based on the Mincerian human capital earnings function (Mincer, 1974) are carried out separately for male (m) and female (f) workers. In a log‐ linear model, log wages (W) are regressed on a set of explanatory factors that comprise individual, establishment, and regional characteristics (X). They are henceforth referred to as endowments and are viewed as observable indicators of productivity differences partly explaining the pay gap. Formally, the regression equations look as follows, with β j representing the estimated coefficient of the characteristic indexed by j, the region under consideration characterized by r, and εdenoting a residual term: 4 ββ ε ∑ =+ + l nW X , rr j r jr jrm, m, 0 m, m, m, (1) ββε ∑ =+ + l nW X . rfr j r j r jrf, , 0 f, f, f, (2) Second, the resulting coefficient estimates, in combination with the gendered endowments, are used to decompose the gender difference in the average wage levels (  W ). This is achieved by replacing gender‐specific log mean wages with the right‐hand side regression results of Equations (1) and (2). Following Blinder (1973), rearranging terms yields the following expression: βββββ ∑∑ −= − + − +− ()()() l nW lnW X X X . rr jr Jr J mr j jr j r jr J rr m, f, m, f, , explained part m, f, f, m, 0 f, 0 unexplained part      (3) The overall or unadjusted GPG is thus split into two components. The first component represents the part of the GPG attributable to gender differences in observed endowments, whereas (  X ) denotes the average characteristics by gender and region. It is, therefore, termed the explained part. The second component is called the unexplained part or adjusted GPG and shows which part of the wage gap is due to the fact that the same endowment generates different market returns for male and female workers. This component also includes the constant. It captures the influence of all unobserved wage determinants on the GPG that we cannot control for in our model due to data restrictions. Such determinants may be personal ability, negotiating skills, or the institutional setting. 4|REGIONAL GENDER PAY GAPS AND THEIR DETERMINANTS Our central variable is the GPG per region. It is based on the daily nominal wages of all female and male full‐time employees between 15 and 64 years of age with their place of work in a specific NUTS 3 region in Germany as of June 30, 2017. Measuring wages at the place of work is preferable to the place of residence (Glaeser & Maré, 2001), as this is where employers are located and wages are determined. We calculate the regional GPG as the difference between the log average wages of men and women (Blau & Kahn, 2017). 4 In containing variables for individuals, establishments, and regions, Equations (1) and (2) combine information on different levels of observation. This multilevel structure can result in inefficient estimates of the coefficients and in biased estimates of the standard errors especially of the variables for the higher level (Moulton, 1990). To deal with this problem, clustering‐robust linear regression techniques are used to estimate standard errors that recognize this clustering of the data. 1068 | FUCHS ET AL. Among the vast range of factors that are responsible for women's lower wages, we focus on the theory‐based individual, establishment, and regional determinants that have been found to be relevant in the explanation of both international and interregional differences in the GPG (Blau & Kahn, 2017; Murillo Huertas et al., 2017). Detailed definitions, as well as descriptive statistics by gender, are provided in Tables A1 and A2 in the appendix. Besides general sociodemographic features like age, qualification, 5 or nationality, individual determinants cover the lower labor market experience of women that leads to a devaluation of their human capital and, thus, to lower wages (Becker, 1964; Blau & Kahn, 2017). We measure the returns of human capital along two different dimensions, namely firm‐specific labor market experience or tenure and career interruptions. 6 This way, we indirectly consider parental leave and childcare periods, which strongly depend on the local provision of childcare facilities. Related, the kind of employment contract matters, specifically holding a temporary contract (Booth et al., 2002). Because temporary positions are more frequently held by women than by men, they contribute not only to lower wages in general but press down the wages of women in particular (Boll et al., 2016). Further individual determinants concern self‐selecting processes in the choice of occupation that go along with labor market segregation. The occupational decision is generally taken at a young age, with interests, personal preferences, and social norms probably being more influential than the consideration of future wages. This behavior might unconsciously be channeled into gender‐specific (entry) jobs (Polachek, 1981), resulting in the observation that women tend to work in low‐paid occupations and men in high‐paid occupations (England, 1992). For example, women often acquire professions in the traditionally female‐dominated fields of caring and nursing, which are generally characterized by lower wages than typical “male”jobs (Marini, 1989). Besides controlling for occupation 7 and, thus, horizontal segregation, we further take into account vertical segregation, which is related to the existence of a “glass ceiling”for women in leading positions (Arulampalam et al., 2007; Fortin et al., 2017). We approximate the under‐representation of women in top jobs by having a supervisory position. As regional labor markets are closely connected via dense interregional commuting patterns between the place of work and the place of residence, we control for each employee's interregional mobility. 8 Because women have lower spatial mobility than men that restricts their job search to a smaller area, they might not benefit from better jobs and higher wages in other regions (Crane, 2007; Petrongolo & Ronchi, 2020). This especially concerns women in rural areas, where their wages are additionally under pressure by less competitive labor markets and consequently higher possibilities of employers to discriminate against them (Hirsch et al., 2013). Again, family responsibilities are a major driving force for women to find a job near their place of living (Nisic, 2017). To consider the divergent mobility patterns between genders, we include a dummy for commuters who live outside the region of work under consideration. Moreover, we control for the selection into the region related to the current job with a dummy, that is, whether a person changed the place of work specifically for the job obtained in 2017. Last, we specify how long an individual has been working in the region under consideration by calculating his or her regional labor market experience. Establishment characteristics cover gender differences in the sorting into firms that provide further reasons for women's lower wages. First of all, women are more likely to be employed in low‐wage firms than men, which can be explained by a combination of sorting and bargaining effects (Barth et al., 2017;Cardetal.,2016). In addition and related to the underrepresentation of women in top jobs, pay‐attractive jobs might be offered on gender‐specific internal markets only that typically exist in large firms (Oi & Idson, 1999). To account for firm‐specific labor markets, we include establishment size as well as the wage level and the wage dispersion within the establishment. The two latter 5 The qualification variable in the BeH is based on reports by employers, which comes along with missing information for some spells in the data set as well as with inconsistencies for a person over time. To improve the information in this variable, we follow the procedure of Fitzenberger et al. (2006) and impute the likely qualification from past or future values. Initially, qualification information was missing for 32% of the spells. After using information from prior and subsequent spells, we end up with 9% missing, which we exclude from the analysis. Conducting the decomposition without imputed qualification variable does not change the results. 6 We do not include general labor market experience because of its high correlation with age and career interruptions. 7 Due to the high degree of collinearity between occupations and sectors, we only include the occupation in our models. 8 We check for any biases due to individual mobility in the robustness checks. FUCHS ET AL. | 1069 variables also serve as proxies for a high‐wage or a low‐wage establishment and the provision of pay‐attractive jobs and career possibilities. A further dimension of gender segregation across firms manifests itself in the vast differences in the sharesofmenandwomeninthefirms’workforce (Card et al., 2016). Therefore, we consider the share of women on all employees within the establishment. Last, we include information on the workforce's qualification structure to account for human capital intensity at the establishment level. Besides individual characteristics of the employees as well as special features of their workplaces, we consider the impact of gender differences in idiosyncratic regional variables on the GPG at the national level. First of all, global agglomeration effects are captured by labor market density, which additionally serves as a proxy for labor market thickness. Regional disparities in the GPG might further result from differences in the local demand for occupations and tasks (Combes & Gobillon, 2015). As women usually predominate in public and hospitality services and men in manufacturing and construction industries (Olivetti & Petrongolo, 2016), the regional sectoral structure should provide divergent employment opportunities for both genders (Hanson & Pratt, 1995; Perales & Vidal, 2015). They are controlled by the share of employees in the service sector, which benefits women. In addition, the female employment rate per region provides information on the respective female labor supply (Olivetti & Petrongolo, 2008). The unemployment rate characterizes the situation on the local labor market and any potential pressure on local wages (Ammermüller et al., 2010). As a further demographic component that simultaneously affects women's lifetime labor force participation (Polachek & Xiang, 2014) we include the fertility rate per region. Religious attachment captures traditional local behavior and social norms. Any specific characteristics in East Germany are taken into account by a dummy variable for employees working in East Germany. 5|RESULTS 5.1 |Descriptive evidence In Germany in the year 2017, full‐time employed women earned on average 97.00 Euro per day and full‐time employed men 118.54 Euro. On the basis of the difference of the log values, this corresponds to an unadjusted GPG of 20.1%. At the regional level, profound disparities emerge (see Figure 1). Most notably, the unadjusted GPG is very low in East Germany, amounting to 5.5% as compared to 21.9% in the Western part. In six East German regions women even earn more than men: Cottbus (−4.3%), Dessau‐Rosslau (−4.0%), Frankfurt/Oder (−3.6%), Stendal (−0.9%), Schwerin (−0.3%), and Märkisch‐Oderland (−0.1%). In contrast, many regions in the South of Germany as well as in the Northwest exhibit a relatively high GPG. The highest GPG (40.3%) can be observed in the Bodenseekreis adjoining Austria and Switzerland. In the Southern districts of Freudenstadt and Dingolfing‐Landau as well as in the city of Ingolstadt the GPG reaches values of around 38%. The two regions with the highest and the lowest GPG feature marked differences between men and women (see Table A3 in the appendix). Most strikingly, men in the Bodenseekreis earn 1.7 times as much as men in Cottbus, whereas women's daily wages are of a quite similar range. One explanation might be the high share of high‐ qualified men in the Bodenseekreis, compared to Cottbus. What is more, the economic setting of the two regions provides different job opportunities for both men and women that seem to reinforce gender‐specific occupational choices. In addition, the Bodenseekreis features a high share of large enterprises paying high wages and showing a high degree of wage dispersion. Looking at the relationship between gendered wages and the GPG across all regions, it becomes evident that women's wages remain relatively stable, whereas men's wages are clearly higher in regions with a high GPG (see Figure S1 in supporting information). Only regions with a negative and very low GPG deviate slightly in that wages are relatively low for both genders. This pattern is confirmed by the correlation coefficient between male wages and the GPG that is much higher than for female wages (0.71 vs. 0.28). Obviously, men's wages drive the regional pattern of the GPG more than women's wages, further suggesting that factors which drive gendered wages are 1070 | FUCHS ET AL. distributed unequally between the regions. In the following, we will assess the reasons for the regional differences in the GPG with the help of the OB decomposition technique. 5.2 |Decomposition of the national gender pay gap For Germany, results of the decomposition show that 6.9 percentage points or about 34.3% of the unadjusted gap of 20.1% can be traced back to gender differences in the explanatory factors included in our analysis, whereas 13.1 percentage points remain as the unexplained part (see Table A4 in the appendix). 9 From the detailed FIGURE 1 The unadjusted gender pay gap in German regions, 2017 FUCHS ET AL. | 1071 decomposition results it becomes clear that establishment characteristics play a prominent role in the explained part, contributing 7.1 percentage points to the overall GPG. Among these, gender differences working in favor of men's wages such as working in an establishment with a high share of female employees and working in a high‐ wage establishment are the prime factors (4.1 and 2.3 percentage points). In contrast, gender differences in individual characteristics that work in favor of women's wages contribute only a total of −0.7 percentage points to the overall gap. Our result that firm characteristics play a decisive role in explaining the GPG is in line with previous findings by Card et al. (2016). Observed gender differences in specific regional characteristics contribute 0.4 percentage points to the GPG. However, this rather low overall contribution hides larger impacts of the single regional characteristics that partly offset each other. First of all, the descriptive evidence in Table A2 in the appendix documents that on average women work in denser labor markets than men. According to the decomposition results, this gender difference reduces the GPG, pointing towards larger benefits from agglomeration economies for women that are in line with previous findings (Bacolod, 2017; Hirsch et al., 2013; Nisic, 2017; Phimister, 2005). This cannot all be put down to the role of the service sector, as the decomposition results on its employment share indicate. On average, women rather work in regions with a slightly higher employment share in services, which increases the GPG by 0.2 percentage points. Similarly, the observation that women work in regions with a higher unemployment rate than men transmits into a slight increase of the GPG, hinting towards a higher bargaining power of establishments in regions with high unemployment and entailing lower wages, particularly for women. The largest regional effect comes from gender differences in working in East and West Germany, leading to an increase in the GPG of 0.3 percentage points. As evidenced in Figure 1, working in East Germany should reduce the GPG. However, it has to be kept in mind that the decomposition is conducted for the whole of Germany. As a consequence, we have to consider differences betweenmenandwomenwhoworkinEast and West Germany, respectively. 10 The positive impact of the East Dummy can then be traced back to the combination of the higher share of women working in East Germany and the significantly higher wages of men working in West Germany. Summarizing the role of the regional determinants, the decomposition shows a significant but small impact of the regional variables on the GPG. 11 This implies that other factors, particularly, individual and firm‐specific characteristics that are unevenly distributed in space are responsible for the profound regional differences in the GPG. We, therefore, proceed with separate decompositions for each region to uncover any regional particularities in the individual and establishment‐specific dimensions. 5.3 |Decomposition of the regional gender pay gaps Figure 2shows the variation of the explained and unexplained parts after the decompositions for the NUTS 3 regions in Germany, which are sorted by the magnitude of their unadjusted GPG. The linear trend lines sum up the coefficients of the two parts across the sorted regions, answering the question of how large the regional differences are with regard to the respective part. The steeper the slope, the more the regions differ. The trend line of the explained part has a positive slope (0.05), hinting towards regional disparities in the relevance of 9 Given the data limitations that are unavoidable when analyzing the GPG, it is a general result for many countries that the explained part makes up the smaller share of the GPG (Boll et al., 2016; Murillo Huertas et al., 2017; OECD, 2017). 10 14% of all men work in East Germany as compared to 16% of all women (see Table A2 in the appendix). 11 An OLS regression of the regional characteristics on the regional unadjusted GPGs yields a significant and negative correlation between the share of the service sector, the female employment rate, the regional unemployment rate, and the workplace in East Germany. Hence, low values in these determinants go along with a high GPG in a region. 1072 | FUCHS ET AL. CONFLICT OF INTERESTS The authors declare that there are no conflict of interests. DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from the Institute for Employment Research. Restrictions apply to the availability of these data, which were used under license for this study. Data are available at https://www.fdz.iab.de/ with the permission of the Institute for Employment Research. ORCID Anja Rossen http://orcid.org/0000-0001-8166-7023 Gabriele Wydra‐Somaggio https://orcid.org/0000-0002-9984-6297 REFERENCES Ammermüller, A., Lucifora, C., Origo, F., & Zwick, T. (2010). Wage flexibility in regional labour markets: Evidence from Italy and Germany. Regional Studies,44(4), 401–421. Arulampalam, W., Booth, A. L., & Bryan, M. L. (2007). Is there a glass ceiling over Europe? Exploring the gender pay gap across the wage distribution. Industrial and Labor Relations Review,60, 163–186. Bacolod, M. (2017). Skills, the gender wage gap, and cities. Journal of Regional Science,57(2), 290–318. Barth, E., Kerr, S. P., & Olivetti, C. (2017). The dynamics of gender earnings differentials: Evidence from establishment data, NBER Working Paper No. 23381. Cambridge, MA. Becker, G. S. (1964). Human capital. A theoretical and empirical analysis with special reference to education. The University of Chicago Press. Blau, F. D., & Kahn, L. M. (2017). The gender wage gap: Extent, trends, and explanations. Journal of Economic Literature, 55(3), 789–865. Blinder, A. S. (1973). Wage discrimination: Reduced form and structural estimates. Journal of Human Resources,8(9), 436–455. Boll, C., & Lagemann, A. (2018). The gender pay gap in the public and private sector in Germany —magnitude, evolution 2010‐2014 and main drivers. HWWI Research Paper,183. Boll, C., Leppin, J., Rossen, A., & Wolf, A.‐R. (2016). In European Commission, Directorate‐General for Justice and Consumers (Ed.), Magnitude and impact factors of the gender pay gap in EU countries. EU Publications. Booth, A., Francesconi, M., & Frank, J. (2002). Temporary jobs: Stepping stones or dead ends? The Economic Journal, 112(480), F189–F213. Card, D., Cardoso, A. R., & Kline, P. (2016). Bargaining, sorting, and the gender wage gap: Quantifying the impact of firms on the relative pay of women. The Quarterly Journal of Economics,131(2), 633–686. Combes, P.‐P., Duranton, G., & Gobillon, L. (2008). Spatial wage disparities: Sorting matters! Journal of Urban Economics, 63(2), 723–742. Combes, P.‐P., & Gobillon, L. (2015). The empirics of agglomeration economies. In G. Duranton, J. V. Henderson, & W. C. Strange (Eds.), Handbook of Regional and Urban Economics (Vol. 5, pp. 247–348). Elsevier. Combet, B., & Oetsch, D. (2019). The gender wage gap opens long before motherhood. Panel evidence on early careers in Switzerland. European Sociological Review,35(3), 332–345. Crane, R. (2007). Is there a quiet revolution in women's travel? Revisiting the gender gap in commuting. Journal of the American Planning Association,73(3), 298–316. Duranton, G., & Monastiriotis, V. (2002). Mind the gaps: the evolution of regional earnings inequalities in the U.K., 1982–1997. Journal of Regional Science,42(2), 219–256. England, P. (1992). Comparable Worth: Theories and Evidence. Aldine de Gruyter. Fitzenberger, B., Osikominu, A., & Völter, R. (2006). Imputation rules to improve the education variable in the IAB employment subsample. Journal of the Applied Social Sciences,126(3), 405–436. Fortin, N. M., Bell, B., & Böhm, M. (2017). Top earnings inequality and the gender pay gap: Canada, Sweden, and the United Kingdom. Labour Economics,47, 107–123. Fortin, N. M., Lemieux, T., & Firpo, S. (2011). Decomposition methods in economics. In O. Ashenfelter, & D. Card (Eds.), Handbook of Labor Economics (Vol. 4, pp. 1–102). North‐Holland. Galego, A., & Pereira, J. (2014). Decomposition of regional wage differences along the wage distribution in Portugal: The importance of covariates. Environment and Planning A,46(10), 2514–2532. Gartner, H. (2005). The imputation of wages above the contribution limit with the German IAB employment sample. FDZ‐Methodenreport 02/2005 (en). FUCHS ET AL. | 1079 Glaeser, E. L., & Maré, D. C. (2001). Cities and skills. Journal of Labor Economics,19(2), 316–342. Goldin, C., Kerr, S. P., Olivetti, C., & Barth, E. (2017). The expanding gender earnings gap: Evidence from the LEHD‐2000 Census. American Economic Review, Papers and Proceedings,107(5), 110–114. Hanson, S., & Pratt, G. (1995). Gender, work, and space. Routledge. Hirsch, B., König, M., & Möller, J. (2013). Is there a gap in the gap? Regional differences in the gender pay gap. Scottish Journal of Political Economy,60(4), 412–439. Kosfeld, R., & Werner, A. (2012). Deutsche Arbeitsmarktregionen–Neuabgrenzung nach den Kreisgebietsreformen 2007–2011. Spatial Research and Planning,70(1), 49–64. López‐Bazo, E., & Motellón, E. (2012). Human capital and regional wage gaps. Regional Studies,46(10), 1347–1365. Manning, A., & Petrongolo, B. (2008). The part‐time pay penalty for women in Britain. The Economic Journal,118(526), F28–F51. Manning, A., & Swaffield, J. (2008). The gender gap in early‐career wage growth. The Economic Journal,118(530), 983–1024. Marini, M. M. (1989). Sex differences in earnings in the United States. Annual Review of Sociology,15(1989), 343–380. Mincer, J. (1974). Schooling, experience and earnings. Columbia University. Minkus, L., & Busch‐Heinzmann, A. (2018). Gender wage inequalities between historical heritage and structural adjustments: A German‐German comparison over time. Social Politics: International Studies in Gender, State & Society,27(1), 156–186. Moulton, B. (1990). An illustration of a pitfall in estimating the effects of aggregate variables on micro units. Review of Economics and Statistics,72(2), 334–338. Murillo Huertas, I. P., Ramos, R., & Simón, H. (2017). Regional differences in the gender wage gap in Spain. Social Indicators Research,134(3), 981–1008. Nisic, N. (2017). Smaller differences in bigger cities? Assessing the regional dimension of the gender wage gap. European Sociological Review,33(2), 292–304. Oaxaca, R. (1973). Male–female wage differentials in urban labor markets. International Economic Review,14(3), 693–709. OECD (2005). OECD employment outlook 2005. OECD. OECD (2017). The pursuit of gender equality. An uphill battle. OECD. Oi, W. Y., & Idson, T. L. (1999). Firm size and wages. In O. C. Ashenfelter, & D. Card (Eds.), Handbook of Labor Economics, 3B (pp. 2165–2214). North Holland. Olivetti, C., & Petrongolo, B. (2008). Unequal pay or unequal employment? A cross‐country analysis of gender gaps. Journal of Labor Economics,26(4), 621–654. Olivetti, C., & Petrongolo, B. (2016). The evolution of gender gaps in industrialized countries. Annual Review of Economics, 8(2016), 405–434. Perales, F., & Vidal, S. (2015). Looking inwards: Towards a geographically sensitive approach to occupational sex segregation. Regional Studies,49(4), 582–598. Pereira, J., & Galego, A. (2011). Regional wage differentials in Portugal: Static and dynamic approaches. Papers in Regional Science,90(3), 529–549. Petrongolo, B., & Ronchi, M. (2020). Gender gaps and the structure of local labor markets (Vol. 64). IZA institute of Labour Economics. Phimister, E. (2005). Urban effects on participation and wages: Are there gender differences? Journal of Urban Economics,58(2005), 513–536. Polachek, S. W. (1981). Occupational self‐selection: A human capital approach to sex differences in occupational structures. The Review of Economics and Statistics,63(1), 60–69. Polachek, S. W., & Xiang, J. J. (2014). The gender pay gap across countries: A human capital approach. In E. Redmount (Ed.), The economics of the family: How the household affects markets and economic growth. Praeger. Rengers,M.,Bringmann,J.,&Holst,E.(2017).ArbeitszeitenundArbeitswünsche.UnterschiedezwischenMikrozensusundSOEP. WISTA,4,11–43. Yeandle, S. (2009). Policy for a change: Local labour market analysis and gender equality. The Policy Press. SUPPORTING INFORMATION Additional Supporting Information may be found online in the supporting information tab for this article. How to cite this article: Fuchs, M., Rossen, A., Weyh, A., & Wydra‐Somaggio, G. (2021). Where do women earn more than men? Explaining regional differences in the gender pay gap. J Regional Sci, 61, 1065–1086. https://doi.org/10.1111/jors.12532 1080 | FUCHS ET AL. APPENDIX A TABLE A1 Definition of the explanatory variables Variable name Definition Individual characteristics Age Dummy 1: 15–19 years, 2: 20–24 years, 3: 25–29 years, 4: 30–34 years, 5: 35–39 years, 6: 40–44 years, 7: 45–49 years, 8: 50–54 years, 9: 55–59 years, 10: 60–64 years Qualification Dummy 1: Low‐qualified (no completed vocational training) 2: Medium‐qualified (completed vocational training) 3: High‐qualified (university degree) Tenure Number of days in employment in the current establishment Career interruption Share of the number of days neither in employment nor in unemployment on the total number of days in the last 20 years (%) Temporary contract Dummy 1: Yes, 0: No Occupation Occupation at the 3‐digit level of the KldB 2010 (144 dummies) Supervisory position Dummy 1: Yes, 0: No Nationality Dummy 1: Foreign, 0: German Commuter Dummy 1: Place of work unequal to place of living, 0: Otherwise Selection into region Dummy 1: Change of place of work before 2017, 0: No change Regional labor market experience Share of the number of days in employment in the current region on the total number of days in employment in the last 20 years (%) Mills ratio Inverse mills ratio from a Probit model for employment Establishment characteristics Establishment size (employees) Dummy 1: <11, 2: 11–20, 3: 21–50, 4: 51–250, 5: >250 Share medium‐qualified employees Share of employees having completed vocational training on all employees (%) Share high‐qualified employees Share of employees holding a university degree on all employees (%) Wage level (€) Median daily wage per establishment (dummy 1: <50, 2: 50–99, 3: 100–199, 4: 200–499, 5: >499) Wage dispersion Absolute deviation from establishment wage median in € Share women Share of women on all employees (%) Regional characteristics Labor market density Share of employees subject to social security contribution on all inhabitants per region in 2017 (%) Share service sector Share of employees subject to social security contribution in services on all employees per region (%) Female employment rate Share of female employees subject to social security contribution on all female inhabitants aged between 15 and 64 years per region in 2017 (%) (Continues) FUCHS ET AL. | 1081 TABLE A1 (Continued) Variable name Definition Unemployment rate Unemployment rate in 2017 (%) Fertility rate Number of live births per woman aged between 15 and 45 years in 2017 Religion Share of inhabitants with Roman Catholic denomination on all inhabitants per region in 2017 (%) East Germany Dummy 1: Yes, 0: No Source: Own compilation. TABLE A2 Descriptive statistics for the explanatory variables, Germany 2017 Variable Number of observations Mean Minimum MaximumMen Women Men Women Individual characteristics Age (years) 15–19 12,255,108 5,678,261 0.00 0.00 0 1 20–24 12,255,108 5,678,261 0.04 0.06 0 1 25–29 12,255,108 5,678,261 0.09 0.14 0 1 30–34 12,255,108 5,678,261 0.12 0.13 0 1 35–39 12,255,108 5,678,261 0.12 0.10 0 1 40–44 12,255,108 5,678,261 0.11 0.09 0 1 45–49 12,255,108 5,678,261 0.14 0.12 0 1 50–54 12,255,108 5,678,261 0.16 0.15 0 1 55–59 12,255,108 5,678,261 0.13 0.13 0 1 60–64 12,255,108 5,678,261 0.07 0.07 0 1 Qualification Low 12,255,108 5,678,261 0.06 0.07 0 1 Medium 12,255,108 5,678,261 0.73 0.71 0 1 High 12,255,108 5,678,261 0.21 0.22 0 1 Tenure 12,255,108 5,678,261 2,629.8 2,413.5 1 6,575 Career interruption 12,255,108 5,678,261 6.09 7.62 0 95.55 Temporary contract 12,255,108 5,678,261 0.11 0.15 0 1 Supervisory position 12,255,108 5,678,261 0.05 0.03 0 1 Nationality 12,255,108 5,678,261 0.08 0.06 0 1 Commuter 12,255,108 5,678,261 0.45 0.39 0 1 1082 | FUCHS ET AL. TABLE A2 (Continued) Variable Number of observations Mean Minimum MaximumMen Women Men Women Selection into region 12,255,108 5,678,261 0.15 0.14 0 1 Regional labor market experience 12,255,108 5,678,261 51.50 50.85 0 99.98 Establishment characteristics Establishment size (employees) <11 12,255,108 5,678,261 0.09 0.11 0 1 11–20 12,255,108 5,678,261 0.09 0.10 0 1 21–50 12,255,108 5,678,261 0.15 0.15 0 1 51–250 12,255,108 5,678,261 0.31 0.30 0 1 >250 12,255,108 5,678,261 0.37 0.35 0 1 Share medium‐qualified employees 12,255,108 5,678,261 68.57 65.44 0 100 Share high‐qualified employees 12,255,108 5,678,261 18.76 22.18 0 100 Wage level (€) <50 12,255,108 5,678,261 0.01 0.02 0 1 50–99 12,255,108 5,678,261 0.38 0.42 0 1 100–199 12,255,108 5,678,261 0.56 0.52 0 1 200–499 12,255,108 5,678,261 0.05 0.04 0 1 >499 12,255,108 5,678,261 0.00 0.00 0 1 Wage dispersion 12,255,108 5,678,261 25.67 25.04 0 259.8 Share women 12,255,108 5,678,261 28.90 58.60 0 100 Regional characteristics Labor market density 12,255,108 5,678,261 438.92 446.32 157.13 1,522.63 Share service sector 12,255,108 5,678,261 69.89 72.04 42.08 69.26 Female employment rate 12,255,108 5,678,261 55.95 55.99 0 1 Unemployment rate 12,255,108 5,678,261 5.72 5.97 1.5 14 Fertility rate 12,255,108 5,678,261 1.62 1.61 37.96 92.38 Religion 12,255,108 5,678,261 30.51 28.52 1.66 87.76 East Germany 12,255,108 5,678,261 0.14 0.16 0 1 Sources: BeH, IEB, BHP, Federal Employment Agency, Federal Statistical Office; own calculations. FUCHS ET AL. | 1083 TABLE A3 Selected characteristics for Germany and the regions with the lowest and highest unadjusted gender pay gap Germany Cottbus Bodenseekreis Daily wages (€): Men 118.54 89.70 150.06 Women 97.00 93.68 100.24 Average age: Men 44 45 43 Women 43 46 41 Share low‐qualified employees (%): Men 6.2 2.6 5.0 Women 6.7 2.1 7.4 Share high‐qualified employees (%): Men 20.6 21.9 30.6 Women 22.1 27.2 21.2 Tenure (days): Men 2,630 2,436 2,808 Women 2,413 2,970 2,178 Share of commuters (%): Men 45.2 51.8 29.6 Women 38.5 46.5 24.4 Share of establishments with >250 employees (%) 36.1 28.9 45.0 Top 3 occupations: Men (%) 1. Occ. in machine‐ building and –operating (7.4) 2. Occ. in warehousing/ logistics, postal and other delivery services, cargo handling (6.2) 3. Occ. in business organization and strategy (5.5) 1. Driver of vehicles in road traffic (6.8) 2. Occ. in machine‐ building and –operating (5.1) 3. Occ. in business organization and strategy (5.0) 1. Occ. in machine‐ building and –operating (13.7) 2. Occ. in metalworking (6.7) 3. Occ. in business organization and strategy (6.9) Top 3 occupations: Women (%) 1. Office clerks and secretaries (11.8) 2. Occ. in business organization and strategy (7.8) 3. Occ. in education and social work, and pedagogic specialists in social care work (5.9) 1. Occ. in public administration (14.2) 2. Office clerks and secretaries (9.5) 3. Occ. in business organization and strategy (8.6) 1. Office clerks and secretaries (11.0) 2. Occ. in education and social work, and pedagogic specialists in social care work (6.5) 3. Occ. in business organization and strategy (5.0) 1084 | FUCHS ET AL. TABLE A3 (Continued) Germany Cottbus Bodenseekreis Share of employees in establishments with a median wage below 100 €(%) 40.5 53.9 27.8 Wage dispersion (€) 25.5 18.7 31.8 Sources: BeH, IEB, BHP; own calculations. TABLE A4 Detailed decomposition results fort he explained part of the gender pay gap (GPG) for Germany and the regions with the lowest and highest unadjusted GPG Germany Cottbus Bodenseekreis Unadjusted GPG 20.05*** −4.33* 40.35*** Explained part 6.90*** −11.46*** 23.11*** Individual characteristics −0.70 −7.10 6.54 Age 0.88*** −0.10 1.47*** Qualification −0.27*** −0.92** 2.09*** Tenure 0.27*** −0.75*** 1.02*** Career interruption 0.19*** −0.04 0.18*** Temporary contract 0.35*** −0.07 0.84*** Occupation −2.84*** ‐4.97*** −0.29 Supervisory position 0.46*** 0.09 0.48*** Nationality −0.09*** −0.03 −0.08 Commuter 0.25*** 0.15*** 0.06 Selection into region 0.01*** 0.11* 0.04 Regional labor market experience 0.06*** −0.66*** 0.62*** Mills ratio 0.03*** 0.11 0.10 Establishment characteristics 7.11 −4.36 16.57 Establishment size 0.25*** 0.17 1.77** Share medium‐qualified employees 0.17*** −0.62** −0.13 Share high‐qualified employees −0.02 1.14*** 0.63* Wage level 2.26*** −5.84*** 4.35*** Wage dispersion 0.39*** −2.93*** 3.85*** Share women 4.07*** 3.73*** 6.11*** Regional characteristics 0.43 Labor market density −0.02*** (Continues) FUCHS ET AL. | 1085 TABLE A4 (Continued) Germany Cottbus Bodenseekreis Share service sector 0.18*** Female employment rate −0.00 Unemployment rate 0.06*** Fertility rate −0.02* Religion 0.02** East Germany 0.27*** Unexplained part 13.15*** 7.13*** 17.24*** ***Significant at the 1% level, **Significant at the 5% level, ***Significant at the 10% level. Sources: BeH, IEB, BHP, Federal Employment Agency, Federal Statistical Office; own calculations. 1086 | FUCHS ET AL.