Unraveling the gender wage gap: Exploring early career patterns among university graduates

Sandner, Malte; Yükselen, Ipek Article — Published Version Unraveling the gender wage gap: Exploring early career patterns among university graduates Scottish Journal of Political Economy Provided in Cooperation with: John Wiley & Sons Suggested Citation: Sandner, Malte; Yükselen, Ipek (2024) : Unraveling the gender wage gap: Exploring early career patterns among university graduates, Scottish Journal of Political Economy, ISSN 1467-9485, Wiley, Hoboken, NJ, Vol. 72, Iss. 2, https://doi.org/10.1111/sjpe.12405 This Version is available at: https://hdl.handle.net/10419/323718 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/4.0/ Scott J Polit Econ. 2025;72:e12405. wileyonlinelibrary.com/journal/sjpe | 1 of 32 https://doi.org/10.1111/sjpe.12405 Received: 29 December 2023 | Accepted: 25 September 2024 DOI: 10.1111/sjpe.12405 SPECIAL ISSUE ARTICLE Unraveling the gender wage gap: Exploring early career patterns among university graduates Malte Sandner1 | Ipek Yükselen2 1Nuremberg Institute for Technology and Institute for Employment Research (IAB), Nuremberg, Germany 2Institute for Employment Research (IAB) and University of Bamberg, Nuremberg, Germany Correspondence Malte Sandner, Nuremberg Institute for Technology and Institut für ArbeitsmarktUnd Berufsforschung (IAB), Nuremberg, Germany. Email: [email protected] Ipek Yükselen, Institute for Employment Research (IAB) and University of Bamberg, Nuremberg, Germany. Email: ipek[email protected]om Abstract A large body of literature has shown that the gender wage gap is small in the first years after graduation and increases gradually with age, largely because of family decisions, often a penalty caused by childbirth. However, the gender wage gap immediately after graduation has received less attention. Using a unique dataset that links 5000 university graduates with master's degrees or equivalent from a large German university to detailed employment records from the German social security register, we specifically analyze the gender wage gap at the first job and its dynamics during the initial years of their careers after graduation. We find that a significant gender wage gap already exists in the first job after graduation, even before most young individuals make family decisions. However, this gender wage gap decreases in the first year after entering the labor market and then increases slowly over time. We attribute this initial decrease in the gender wage gap to female university graduates experiencing greater returns from firm and occupational changes than their male counterparts. This suggests that women may use these changes to address skill mismatches, which are more common among women than men in their first job. KEYWORDS early career, gender wage gap, university graduates JEL CLASSIFICATION I23, J16, J31, J71 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2025 The Author(s). Scottish Journal of Political Economy published by John Wiley & Sons Ltd on behalf of Scottish Economic Society. 2 of 32 | SANDNER and YÜKSELEN 1 | INTRODUCTION Despite advances in women's education and career opportunities in recent decades, a persistent gender wage gap remains prevalent in economically advanced nations (Goldin, 2014; Olivetti & Petrongolo, 2016), this gap is even larger among individuals with higher levels of education (Blau & Kahn, 2017; OECD, 2022). Many studies have examined the gender wage gap among highly educated individuals and found that women's lower labor supply and more frequent career interruptions (mainly due to child care) compared to those of men are the main reasons for this gender wage gap.1 Less is known about the existence and development of a gender wage gap at the beginning of a career. This lack is surprising given the large relevance of starting wages and early career wage growth on university graduates' future labor market outcomes and potentially on the gender wage gap. For example, Oyer (2006), Kahn (2010), and Oreopoulos et al. (2012) show that labor market conditions at the beginning of the career, such as recessions, can have an impact on entry wages and, consequently, on wages in the long run. Moreover, prior wages usually determine wage increases due to promotions within the same firm (Graham et al., 2000); even wage increases as a result of a job change are usually based on previous wages (Hansen & McNichols, 2020). These findings indicate that entry wages play a crucial role in determining future wages over the long run, significantly contributing to the origin of the gender wage gap. Knowledge about the early gender wage gap is also important for developing new or adapting existing counseling policies to provide effective job search strategies and to challenge gender norms in career choices for graduates. Whether a gender wage gap exists in the first years after graduation is theoretically ambiguous. Particularly in the first job, some common reasons for pay differences between men and women, such as familyrelated decisions (e.g., childbirth or marriage), careerrelated developments (e.g., promotions), work experience, and firmspecific networks, may not yet be relevant or less relevant than later in the career.2 Therefore, we expect no or only a small gender wage gap in the initial job, especially when we account for gender differences in the field of study and the characteristics of the employer in the first job. However, particularly in the first job, both the applicants and the firms face considerable uncertainty. Firms can assess only the labor market productivity of candidates without prior work experience based on their university grades and interview performance. Given that women currently tend to have higher GPAs than men (Becker et al., 2010; Francesconi & Parey, 2018), we may even expect a gender wage gap that is conditional on differences in the field of study choice and employer characteristics to favor women in the first job. On the other hand, existing studies show that female applicants negotiate less in job interviews than male applicants do (Babcock & Laschever, 2009; Bertrand, 2011) and may face statistical discrimination at labor market entry (Altonji & Pierret, 2001; Pinkston, 2006). Furthermore, differences in preferences or personality traits, such as risk aversion or overconfidence, can be particularly important at the beginning of a career. Studies show that women are more riskaverse (e.g., Cortés et al., 2023) and less selfconfident than men (e.g., AdameczVölgyi & Shure, 2022), which may lead them to accept lowerpaying job offers. As a result, the gender wage gap could be substantially in favor of men, given differences in field of study and employer characteristics at the first job. This ambiguity about the gender wage gap may be even greater in the years after labor market entry when firms have observed the productivity of their employees or when graduates change jobs to increase their wages. If women earn less than men in their first job as a result of discrimination, this gap may narrow over time as women move to less discriminatory firms or as employers learn about employees' true productivity over time (Altonji & Pierret, 2001; Farber & Gibbons, 1996). Additionally, women may correct initial job choices 1For example, see Adda et al. (2017), Kuziemko et al. (2018), Kleven et al. (2019a), and Cortes and Pan (2020). 2The mean age of German mothers at first birth was 30.5 in 2021 (Federal Statistical Office, 2022), while the average age of labor market entry for women in our sample is 27. Moreover, the average age of women at birth is expected to increase with the level of education. Therefore, this issue is not expected to be of high magnitude in the case of women with a master's degree at labor market entry. | 3 of 32 SANDNER and YÜKSELEN based on gender norms rather than preferences by changing jobs. However, the gender wage gap may also increase over time due to job changes in the early stages of a career, as the literature shows that women generally realize lower returns to job mobility than men (Albrecht et al., 2018; Topel & Ward, 1992). In addition, the gap may widen over time as familyrelated decisions become more important over time. Overall, the gender wage gap at labor market entry and the dynamics of the gender wage gap in the early years of a career remain unclear and thus require examination. This study examines the gender wage gap immediately after entering the labor market and its evolution during the initial years of a career for more than 5000 university graduates with a master's degree or equivalent. We use unique administrative data on graduates of a large German university linked with detailed social security data from the Integrated Employment Biographies (IEB). This linked administrative dataset provides a wide range of information from these two data sources, including sociodemographic characteristics of the graduates, the attained university degree, field of study, the final high school and university grades, the date of enrollment, and the exact timing of graduation, labor market entry, and any occupation or firm changes. Using these data, we first estimate the gender wage gap at labor market entry among university graduates. Our findings show that males have significantly higher wages than females in their first fulltime job immediately after graduation, despite our homogeneous and highly educated sample with high labor market attachment. The estimated unadjusted gender wage gap of approximately 12.5 log points corresponds to approximately 10 euros per day or 300 euros per month. The adjusted gender wage gap is conditional on a comprehensive set of personal and pregraduation controls, including graduation year, age, non(German) citizenship status, field of study, the final university grades, having an apprenticeship degree, worked during study, and the place of the final high school examination, is equal to 6.2 log points. Including occupation fixed effects reduces the gender wage gap to 4.7 log points. Other postgraduation characteristics, such as the timing of the first job, firm fixed effects, the share of women in the firm, and the location of the firm, do not substantially alter the gender wage gap. Second, since both career paths and wages vary widely across fields of study (Altonji et al., 2016), we conduct a subgroup analysis of four broad groups of fields of study: economics and business, mathematics and natural sciences, humanities and social sciences, and medical studies. The results show that the unadjusted (raw) gender wage gap in the first job is prominent in almost all field groups except medical studies. The raw gender wage gaps in each field group are 8.6 log points, 14.1 log points, 10.2 log points and 1.5 log points, respectively. For mathematics and natural sciences, the gender wage gap disappears after controlling for the major subject within the field of study. The adjusted gender wage gap is the highest in the humanities and social sciences, at 9.7 log points. This field group also has the lowest average daily wage in the first job, the highest variation in wages, and the highest share of females. Third, as dynamics are very important, particularly in the early years of a career, and have an impact on future wage growth, we examine the dynamics of the gender wage gap over the first years after labor market entry. Our findings reveal a decrease in the estimated gender wage gap in the first 3 years after labor market entry, followed by an increase in subsequent years. The largest reduction in the wage gap occurs 1 year after labor market entry. Moreover, we demonstrate that this decrease is observed only among economics and business graduates and humanities and social sciences graduates who change both firms and occupations within 1 year of entering the labor market. However, this decline does not occur for graduates from other fields of study or for those who remain in the same firm and occupation. Finally, our analysis focuses on two field groups: economics and business and humanities and social sciences. This analysis shows that women who change firms and occupations after their first job drive the decline in the gender pay gap, as women benefit more from these changes than men. Our data reveal that women are more likely than men to work in a mismatched occupation at the first job. By changing both firms and occupations, women move out of the lowestpaid occupations and are able to correct this mismatch, leading to a greater increase in wages than men. After comparing these empirical findings with several theories in the gender wage gap literature, one explanation for our finding may be that women immediately after graduation 4 of 32 | SANDNER and YÜKSELEN have strong preferences for certain job and firm amenities, such as job meaning and relevance, or they follow certain gender norms about firms and occupations leading them to initially accept mismatched jobs. One year after labor market entry, individuals' preferences or willingness to follow gender norms may change, and they may correct this mismatch by changing occupation and firm. However, our data do not allow for a definitive test of this hypothesis. Our study contributes to the literature in important ways. First, several studies examine the dynamics of the gender wage gap over the life cycle and find evidence that the gender wage gap is smaller at younger ages but increases over time, mainly due to familyrelated decisions (Albrecht et al., 2018; Bertrand et al., 2010; Manning & Swaffield, 2008).3 Although these studies provide valuable insights into the dynamics of the gender wage gap in general, they do not focus on the first job after graduation. The few papers that examine the gender wage gap at the beginning of a career rely primarily on survey data. For example, Cortés et al. (2023) find in a survey of US graduates that women earn 10% less than men in their first job. In a related German study, Francesconi and Parey (2018) found an adjusted gap of 5–10 log points among German college graduates 12–18 months after graduation. In contrast to this literature, we are the only study to investigate the gender wage gap among university graduates using administrative data, with a focus on the first job after graduation.4 Administrative data help to avoid reporting bias that can occur in surveybased studies at the beginning of a career due to frequent job changes. Furthermore, the administrative nature of the data overcomes concerns associated with missing data, response rates, or measurement error due to retrospective questions. Most of the other data used to study the gender wage gap either lack comprehensive information on graduates' pregraduation characteristics (field of study, GPA) or are unable to track graduates as they transition into the labor market and lack information on graduates' occupation, industry, and other important employment characteristics. In contrast, our linked data provide access to accurate and comprehensive measures of human capital determinants of productivity, including academic grades and field of study, as well as detailed information on employment, wages, and occupations. Second, our study provides unique insights into early career job dynamics and their impact on the gender wage gap. At the beginning of careers, a high level of information friction can lead to poor job matches in the labor market for recent graduates (Vesterlund, 1997). Fredriksson et al. (2018) highlight high separation rates and job changes among inexperienced employees due to limited information about the labor market. The literature shows that job changes are in general associated with wage growth but also that men are more likely to change jobs and tend to benefit more from job mobility than women, thereby exacerbating the gender wage gap over time (Albrecht et al., 2018; Del Bono & Vuri, 2011; Manning & Swaffield, 2008; Topel & Ward, 1992). However, these studies do not focus on the first years after labor market entry because observing this crucial early period where returns to job changes may be different is difficult without detailed administrative data. In contrast, we are able to follow all graduates without attrition over time, which allows us to observe the exact timing of any job changes or job search periods within the first few years after labor market entry. This information allows us to observe the share of female and male graduates from each field of study who change jobs and to observe the returns to their mobility, which may have longlasting effects on their future labor market careers. The remainder of this paper is structured as follows. Section 2 describes the dataset and its advantages and shortcomings, characterizes the sample of university graduates used in the analysis, and presents some descriptive statistics. The results of the estimated gender wage gap at labor market entry and the dynamics of the gender wage gap over the first few years of a career are presented in Sections 3 and 4, respectively. We examine gender differences in firm and occupational mobility in Section 5 and the underlying reasons for this mobility in Section 6 before concluding the paper in Section 7. 3The effect of the child penalty on females' labor market outcomes is explored in several studies, for example, Kleven et al. (2019b) and Dustmann et al. (2009). 4The studies by Kunze (2003, 2005) used administrative data but focused on younger graduates who had completed an apprenticeship. | 5 of 32 SANDNER and YÜKSELEN 2 | THE LINKED ADMINISTRATIVE DATASET AND UNIVERSITY GRADUATES 2.1 | Data This study is based on a unique administrative dataset of graduates from a large university linked with registry data from the German Integrated Employment Biographies (IEB) of the Institute for Employment Research (IAB). The linked dataset combines detailed study information on each graduate from the university registry with information on individual employment records covering the whole employment biography of jobs subject to social security contributions from the IEB dataset. The available dataset from the university covers all graduates of this university from 1995 until 2016. During the observation period, almost all the fields of study are considered except for engineering degrees. The data are highly reliable, as they are based on administrative records from the university registry. The dataset provides information on the personal characteristics of each graduate, such as year of birth, gender, nationality, district, and grade of the certificate of general qualification for university admission (Abitur), hereafter referred to as the final high school grade point average (GPA). The dataset also includes studyrelated characteristics at the university, such as the field of study, the type of university degree attained, the final GPA, and the dates of enrollment and graduation. The IEB is a large administrative dataset of individuals' employment biographies provided by the IAB for the period 1975–2019. The information provided by the dataset is highly reliable, as it is a legal requirement in Germany for all employers to provide information on their employees to the German Social Security Administration. The IEB dataset includes individuals in employment covered by social security, excluding selfemployed individuals and civil servants. Thus, the IEB dataset covers approximately 80% of the total labor force in Germany. In addition to the precise timing of employment and outofemployment spells, the dataset provides information on gross daily wages, industry, occupation (threedigit), fulltime status, and other employment characteristics (Dorner et al., 2010). The data from the university are merged with the IEB dataset using a linkage procedure established at the IAB based on an individual's full name, sex, and date of birth, with a 90% match rate (Möller & Rust, 2017).5 2.2 | Sample choice The focus of this study is on individuals with a master's degree or equivalent with available university GPA data. We further focus on graduates who are working fulltime in their first regular job with a daily wage of at least 10 euros6 and we omit graduates who are not fulltime employed in their first job after graduation, that is, parttime, minijobs, internships, working students, etc., even if they subsequently switch to fulltime employment. If an individual has more than one wage spell at a given time, we choose the “main” employment spell as defined by the IAB.7 5Please see the study by Möller and Rust (2017) for a more detailed explanation of the matching procedure. 6Wages are deflated to the year 2010 using the consumer price index. 7Since information on working hours is not available in the linked dataset, we focus on fulltime jobs in order to eliminate a potential bias in the gender wage gap induced by differences in working hours. Since we focus on fulltime employees in our main analysis, the working hours of men and women should be reasonably comparable. However, even if employees are fairly homogeneous in terms of fulltime employment, males may still work more hours than females, allowing them to earn higher wages (e.g., Goldin & Katz, 2016). The study by Francesconi and Parey (2018) documents that differences in hours worked among fulltime employees do not significantly explain the gender wage gap among German graduates approximately 12 to 18 months after graduation. Therefore, we expect that our results are not driven by differences in working hours between fulltime employed female and male graduates. 6 of 32 | SANDNER and YÜKSELEN We focus on master's or equivalent graduates to consider the most policyrelevant group with greater labor market attachment, and the results are easier to interpret for a more homogeneous group.8 Furthermore, returns to master's graduates expected to be higher compared to vocational training, high school education, or only a bachelor's degree (Altonji et al., 2016). In addition, master's graduates have a higher degree of attachment to the labor market and form a relatively homogeneous group, making it easier to identify factors that impact entrylevel wages. They are also an ideal group to study early career gender wage gaps, as child care, a key factor in the wage gap for highly educated individuals, has less of an impact at this stage. Furthermore, we exclude individuals who are older than 35 years (1.5% or 104 individuals). We also omit graduates with a gap of more than 15 months between graduation and their first employment spell (14% or 999 individuals), as these individuals may have spent time abroad or already worked on a selfemployed basis (which is not captured by the data). Since our main analysis focuses on the first fulltime job after graduation and subsequent years, we keep graduates with wage spells at the beginning of their first job and 1 year after their first job (8.6% or 478 individuals were dropped). Finally, after dropping observations with missing values, the final sample for the main wage estimations consists of 5212 individuals. 2.3 | Descriptive statistics Table 1 presents descriptive statistics on labor market entry for our preferred sample of university graduates in fulltime employment, which we use for the wage analysis in the following sections. While Panel A of Table 1 documents pregraduation characteristics, such as university and high school GPA, duration of study, nonGerman citizenship, and others, Panel B presents postgraduation characteristics, including characteristics of the first job. Graduates' university and high school GPAs range from 1 (the best possible grade) to 4 (the lowest passing grade). Panel A of Table 1 shows that both men and women complete their degrees in approximately five and a half years on average, although women graduate at a younger age. The majority of graduates at the university (around 70%) acquire some form of work experience before graduation, with females being more likely to work during their studies than males. In addition, consistent with the literature, the share of female graduates increases with graduation cohort, with the male–female ratio reversing even in the most recent cohort group (2007–2010). This finding is also in line with the overall population of German graduates. Consistent with the literature (see, e.g., Becker et al., 2010), female graduates enroll at the university with better final high school grades and leave the university with slightly better university grades (by around 11% of the sample standard deviation) than males.9 Females are also more than twice as likely to graduate in the humanities and social sciences. Graduates in mathematics and natural sciences account for nearly a quarter of all male graduates, compared to only 8% of all female graduates. Nevertheless, economics and business remain the dominant fields of choice for both genders. Finally, the share of women studying medicine is approximately 10 percentage points higher than that of men. Table 1 also shows that female graduates are more likely than male graduates to earn a magister or state examination degree. The vast majority of graduates have a diploma degree, with a greater proportion of men than women. Table C2 in the Appendix 3 compares our estimation sample with official German register data and other representative studies. For 2010, our sample shows a slightly higher proportion of women (50%) than the data from the Federal Statistical Office (46%). The share of females by field of study is also comparable. The largest difference is observed in mathematics and natural sciences. According to our data, the proportion of females in this field is 18%, 8Another reason for focusing on this group is that the Bologna Process reform was implemented in Germany between 2005 and 2010, and only a small proportion of graduates in the sample have a master's degree. Before the Bologna Process, bachelor's and master's degrees were combined into diploma or magister degrees. 9In our data, final high school grades are available only for the data beginning with the 2001 graduation cohort. | 7 of 32 SANDNER and YÜKSELEN TABLE 1 Descriptive statistics. Male Female Diff.Mean (Std. dev.) Mean (Std. dev.) Panel A: Pregraduation and personal characteristics Final highschool grade (Abitur)2.245 (0.616) 2.077 (0.594) 0.168*** Individuals 1476 1200 Final university grade 2.058 (0.604) 1.998 (0.568) 0.060*** NonGerman citizenship 0.015 (0.121) 0.033 (0.179) −0.019*** Age at graduation 27.238 (1.864) 26.579 (1.906) 0.660*** Duration of study 5.592 (1.308) 5.631 (1.372) −0.039 Apprenticeship 0.065 (0.247) 0.061 (0.239) 0.004 Worked during studies 0.673 (0.469) 0.742 (0.438) −0.069*** Origin in the same federal state as the university 0.876 (0.329) 0.861 (0.346) 0.016 Graduation year 1995–1998 0.283 (0.451) 0.156 (0.363) 0.127*** 1999–2002 0.227 (0.419) 0.172 (0.377) 0.055*** 2003–2006 0.227 (0.419) 0.269 (0.443) −0.042*** 2007–2010 0.263 (0.440) 0.403 (0.491) −0.141*** Field of study Economics and business 0.469 (0.499) 0.328 (0.469) 0.141*** Mathematics and natural sciences 0.224 (0.417) 0.077 (0.267) 0.147*** Humanities and social sciences 0.111 (0.314) 0.300 (0.458) −0.189*** Medical studies 0.196 (0.397) 0.295 (0.456) −0.099*** Type of degree Diploma 0.747 (0.435) 0.576 (0.494) 0.172*** Magister 0.046 (0.210) 0.114 (0.317) −0.068*** Master 0.010 (0.102) 0.015 (0.123) −0.005 State examination (Staatsexamen) 0.196 (0.397) 0.295 (0.456) −0.099*** Individuals 3258 1954 Panel B: Postgraduation characteristics Left the state 0.196 (0.397) 0.179 (0.383) 0.118 Left the city 0.683 (0.465) 0.655 (0.476) 0.028* Mean job search duration 3.747 (3.128) 3.817 (3.039) −0.070 Duration of job search Less than 1 month 0.190 (0.392) 0.161 (0.368) 0.028*** 1–3 months 0.326 (0.469) 0.319 (0.466) 0.007 3–5 months 0.214 (0.410) 0.247 (0.431) −0.033*** More than 5 months 0.270 (0.444) 0.272 (0.445) −0.002 (Continues) 8 of 32 | SANDNER and YÜKSELEN while according to the register data, it is 35%. In our sample, students have a slightly better final high school GPA than in the survey data taken from the study Simeaner et al. (2014) and university grades that are similar to those of the representative sample taken from the survey data used by Francesconi and Parey (2018). The graduates in our sample are, on average, about 5 months younger because we use examination rather than exmatriculation dates. Notably, 11% of our students are nonGerman, compared to 22% in the survey data used by Francesconi and Parey (2018), as we cannot observe individuals in our data if they move to another country after graduation. Panel B of Table 1 presents postgraduation and employment characteristics, such as mobility, the time between graduation and the first fulltime job,10 establishment size, and the share of women in the establishment of the first job.11 The table shows that approximately 70% of the graduates find their first fulltime job outside the city of the university location, with males being slightly more mobile than females. On average, female graduates take longer to find their first job than male graduates, that is, approximately 3.7 and 3.8 months, respectively. A breakdown of the duration of job searches into different categories shows that the share of male graduates with a job search duration of less than 1 month is greater. Finally, female and male graduates tend to work in establishments of similar size. However, in line with the literature, women are more likely to work in establishments with a higher proportion of female employees. A potential explanation for this situation might be the sorting of university graduates into specific industries by gender, resulting in femaledominated industries (Hellerstein et al., 2011). 3 | THE GENDER WAGE GAP AT LABOR MARKET ENTRY We begin our analysis by examining gender wage differences at labor market entry.12 To identify gender differences, we estimate the following regression equation: 10Hereafter referred to as “job search duration”, even though this time is not necessarily spent searching for a job. 11 The data include information only on establishments, not firms. However, in this paper, we use the terms “establishment” and “firm” interchangeably. 12 The term labor market entry refers to the first job after university graduation; we use these terms interchangeably. (1) Yi = 𝛼 + 𝛾Femalei + 𝛽Xi+ϵ i Male Female Diff.Mean (Std. dev.) Mean (Std. dev.) Firm size Less than 25 employees 0.238 (0.426) 0.247 (0.432) −0.009 25–250 employees 0.273 (0.445) 0.266 (0.442) 0.007 250–2000 employees 0.254 (0.436) 0.273 (0.446) −0.018 More than 2000 employees 0.235 (0.424) 0.214 (0.411) 0.020* Share of women in the firm Less than 40% 0.356 (0.479) 0.201 (0.401) 0.155*** 40%–70% 0.405 (0.491) 0.383 (0.486) 0.022 More than 70% 0.239 (0.427) 0.417 (0.493) −0.177*** Individuals 3258 1954 Note: This table shows summary statistics of graduates' pregraduation and postgraduation characteristics. The sample consists of graduates with a master's degree or the equivalent, who worked fulltime at their first job after graduation and who have a wage spell 1 year after their first job. ***, ** and * indicate significance at the 1%, 5%, and 10% levels. TABLE 1 (Continued) | 15 of 32 SANDNER and YÜKSELEN Previous research shows that firm and/or occupational mobility affects wages and contributes to wage growth (Bartel & Borjas, 1981; Topel & Ward, 1992); mobility is especially important in the early stages of a career (Albrecht et al., 2018) and that men tend to benefit more from job mobility than women (Del Bono & Vuri, 2011; Manning & Swaffield, 2008). To analyze whether and to which extend job changes in the earliest career stage explain the drop in the gender wage gap, we continue our investigation by separating the sample into graduates who stay in the same firm and/or occupation (Column 2 of Table 4), those who change occupations but remain in the same firms (Column 3), those who change firms but remain in the same occupation TABLE 4 The gender wage gap by job change status. Dependent variable: Log daily wage Pooled Stayers Only firm changers Only occupation changers Firm and occupation changers (1) (2) (3) (4) (5) Panel A: Economics, business, humanities, and social sciences 1 year after × female 0.036 *** −0.000 0.038 0.061 0.193 *** (0.012) (0.008) (0.063) (0.083) (0.070) 1 year after 0.130 *** 0.098 *** 0.252 *** 0.248 *** 0.271 *** (0.007) (0.006) (0.042) (0.063) (0.039) Female −0.098 *** −0.056 *** −0.156 ** −0.139 * −0.340 *** (0.015) (0.014) (0.068) (0.074) (0.069) Share of females 10.737 0.115 0.101 0.131 Share of males 10.796 0.091 0.058 0.105 Rsquared 0.240 0.274 0.306 0.323 0.328 Individuals 3114 2407 267 194 312 Panel B: Mathematics, natural sciences, and medical studies 1 year after × female −0.001 0.006 −0.010 −0.140 −0.088 (0.011) (0.009) (0.042) (0.132) (0.095) 1 year after 0.150 *** 0.127 *** 0.182 *** 0.341 *** 0.386 *** (0.008) (0.007) (0.031) (0.098) (0.054) Female −0.019 −0.024 0.008 0.074 0.049 (0.016) (0.016) (0.034) (0.171) (0.096) Share of females 10.783 0.144 0.032 0.081 Share of males 10.838 0.086 0.037 0.070 Rsquared 0.393 0.399 0.624 0.499 0.382 Individuals 2098 1718 204 60 137 Note: This table shows the gender wage gap at labor market entry by job change status based on the OLS model specified in Equation (2). The sample consists of graduates with a master's degree or equivalent who work in a fulltime job as their first job after graduation and who have a wage spell 1 year after their first job. The dependent variable is the log gross daily wage at the first job. The estimations include personal and pregraduation characteristics as controls. The personal characteristics include age and having German citizenship. The pregraduation characteristics include duration of study, place of high school final exam, and working during studying. All estimations include the beginning month of the first job as a control. Robust standard errors are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels. 16 of 32 | SANDNER and YÜKSELEN (Column 4), and those individuals who change firms and take up a new occupation (Column 5).21 Our job move categories follow Fitzenberger et al. (2015) who analyze the returns to occupation and firm switcher in a sample of apprentices. Panel A of Table 4, which focuses on graduates in economics, business, humanities, and social sciences, demonstrates that the gender wage gap does not decrease significantly for those individuals who either stay in the same firm and/or occupation 1 year after starting their first job (Columns 2–4). In contrast, female firms and occupation changers increase their wages on average by approximately 19 log points more than their male counterparts (Column 5). This increase must be considered in the context that males who change either firm, occupation, or both benefit from these changes by approximately 25–27 log points, while the stayers increase their wages by only 10 log points. In addition, the initial gender wage gap is larger for individuals who change firms, occupations, or both than for individuals who remain in their occupation in the same firm. The group that changes firms and occupations has the largest initial gap (almost 34 log points). Interestingly, the allocation of men and women to the four groups is relatively similar; thus, differences in shares do not seem to explain the different evolutions of the gender wage gap after 1 year. For mathematics, natural sciences, and medical studies, we find no reduction in the gender wage gap in the first 12 months after starting the first job, and we observe no initial gap for any of the changer groups (Panel B of Table 4). However, even for these fields, the results show that movers have the highest wage growth, between 18 and 39 log points. Overall, the table shows that women in economics, business, humanities, and social science benefit more than men from a completely new start after their first job, which includes a change in firm and occupation. This new start drives the observed reduction in the gender wage gap within 1 year after the first job. As a next step in our analysis, Figure 3 shows the dynamics of wages over 5 years after the first job for stayers and those graduates who change both occupation and firm. Confirming the results shown in Table 4, female and male movers initially earn lower wages on average than stayers. However, the wage difference between stayers and movers is greater for females than for males, due to the very low entry wages of those females who later change both occupation and firm. In summary, women with low entry wages appear to correct their low wages more than men by changing their firm and occupation within 1 year of labor market entry. 6 | WHY IS SWITCHING FIRM AND OCCUPATION IN THE BEGINNING OF THE CAREER MORE BENEFICIAL FOR FEMALES THAN FOR MALES? In this section, we use our rich administrative data to investigate why women benefit more than men from changing firms and occupations after their first job after graduation. Although we are not aiming to identify causal effects for this higher female benefit, we are confident to relief some interesting patterns. We use two estimation approaches to conduct our analysis. First, we estimate whether women who change firms and occupations differ from men who change firms and occupations in terms of demographic characteristics, university outcomes, and characteristics of their first job and whether these gender differences differ from those of stayers. Second, we estimate whether firm and occupation characteristics change differently for males and females after job transitions. At the end of the section, the results of the two estimation approaches are discussed with respect to common theoretical explanations for the gender wage gap. With regard to binary outcomes, we also applied logit estimations, which yielded similar results. Table 5 compares gender differences in personal and pregraduation characteristics (both of which are constant over time) between firms and occupation changers and stayers. In Table 6, Panel A compares gender differences in the characteristics of the first job for individuals who change firms and occupations with those who 21We define an occupation change as when the threedigit occupation code changes. | 17 of 32 SANDNER and YÜKSELEN stay in the same position. Panel B of Table 6 examines gender differences in the characteristics of the first and subsequent jobs 1 year later for occupation and firm changers. Table C5 in the Appendix 3 shows the mean values of all variables in Tables 5 and 6 by gender and the corresponding mean gender differences for stayers and for firm and occupation changers. Row (1) of Table 5 reports the interaction coefficients between the female variable and a dummy for firm and occupation change. The coefficients in Row (1) indicate that none of the personal characteristics, pregraduation characteristics, or job search characteristics exhibit greater differences between males and females who change firms and occupations than between males and females who stay in the same firm and occupation. However, Table 6 demonstrates that out of several characteristics, jobeducation mismatch and occupational rank (Columns 6–11) are two characteristics that differ between males and females who switch occupations and firms in their first job and develop differently after the job change. We separate jobeducation mismatches into two types: horizontal and vertical mismatches. A horizontal mismatch is a fieldoccupation mismatch in which the employee's field of study does not match the field required for the job. A vertical mismatch is a skill mismatch where the skill level of the employee's qualification does not match the requirements of the job. Since our sample includes highly skilled university graduates, only jobs for which university graduates are overqualified are defined as vertical mismatches.22 In addition, occupation rank is a measure that ranks occupations by their average wage (Column 9 of Table 6).23 22A large body of literature has reported that both vertical and horizontal mismatches have a negative effect on wages (Boudarbat & Montmarquette, 2009; Heijke et al., 2003; Robst, 2007; Wolbers, 2003). 23Average wages within threedigit occupation codes are calculated using the SIAB data, which represent 2% of the IEB data. FIGURE 3 The Dynamics of Wages for Job Stayers and Occupation and Firm Changers by Gender. These figures plot the dynamics of wages over 5 years after the first job for stayers (left panel) and for firm and occupation changers (right panel), and the sample sizes are 312, and 137 respectively. The dependent variable is the log gross daily wage. The graduation year, and personal and pregraduation characteristics are added as controls. The personal characteristics include age and having German citizenship. The pregraduation characteristics include duration of study, place of high school final exam, and working during study. All estimations include the beginning month of the first job as a control. Additionally, we control for having a child between the years. 18 of 32 | SANDNER and YÜKSELEN TABLE 5 Personal and pregraduation characteristics of firm and occupation changers. Dependent variables Personal characteristics Pregraduation characteristics Finding first job characteristic Age at the first job Nongerman Duration of study Working during studying Apprent Origin same state as the Uni. Final Uni. grade Duration of job search (1) (2) (3) (4) (5) (6) (7) (8) Female × firm and occupation changers 0.058 −0.021 0.226 −0.012 −0.019 −0.031 0.071 −0.342 (0.269) (0.017) (0.171) (0.053) (0.037) (0.041) (0.073) (0.366) Firm and occupation changers 0.444 *** 0.003 0.063 0.002 0.054 ** 0.021 0.026 −0.035 (0.156) (0.010) (0.096) (0.037) (0.026) (0.027) (0.045) (0.246) Female −0.803 *** 0.026 *** −0.053 0.126 *** −0.010 0.034 ** −0.206 *** 0.141 (0.082) (0.007) (0.053) (0.018) (0.010) (0.014) (0.023) (0.131) Means of dependent variable 27.309 0.023 5.515 0.719 0.074 0.858 2.115 3.859 Rsquared 0.042 0.006 0.003 0.018 0.004 0.002 0.030 0.001 Individuals 2719 2719 2719 2719 2719 2719 2719 2719 Note: The table documents gender differences in personal, pregraduation, and firstjob characteristics, between firms and occupation changers and stayers. The sample size is 2719, including stayers (Column 2, Table 4) and firm and occupation changers (Column 5, Table 4). Each Column is a different estimation that is time invariant. We use the following estimation equation, where the dependent variable changes in each column. The estimations include a female dummy, a dummy variable for firm and occupation changer dummy ( JobChangersi ) (= 1 if an individual changes her or his firm and occupation within 1 year after labor market entry, = 0 if an individual stays at the same firm and occupation), and an interaction of these dummies. Robust standard errors are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels. (4) Yi =𝛽 0 +𝛽 1Femalei +𝛽 2JobChangersi +𝛽 3Femalei ⋅ JobChangersi +𝛾 gradyeari +𝜀 i, | 19 of 32 SANDNER and YÜKSELEN TABLE 6 Job characteristics of firm and occupation changers. Dependent variables Median daily Share of Share of Share of Log Horizontal Vertical Horizontal or Occupation Occupation Occupation Log wage of fulltime employees Parttime employees High qualified employees Women in a firm Firm size Mismatch Mismatch Vertical mismatch Rank Rank < quantile 10 Rank > quantile 90 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Panel A: First job characteristics of firm and occupation changers Female × firm and occupation changers −0.086 * −0.059 −0.013 0.044 −0.204 0.008 0.117 ** 0.070 −22.371 * 0.098 ** −0.025 (0.049) (0.041) (0.032) (0.027) (0.252) (0.056) (0.058) (0.053) (11.815) (0.044) (0.034) Firm and occupation changers −0.120 *** −0.007 −0.063 *** −0.006 −0.341 ** 0.119 *** 0.122 *** 0.150 *** −7.680 0.017 −0.002 (0.028) (0.032) (0.021) (0.017) (0.166) (0.035) (0.039) (0.037) (7.488) (0.024) (0.026) Female −0.041 *** −0.034 ** −0.016 0.071 *** −0.102 0.122 *** 0.005 0.051 ** −15.238 *** 0.013 −0.036 *** (0.014) (0.016) (0.011) (0.009) (0.090) (0.017) (0.021) (0.021) (3.602) (0.012) (0.013) Mean of dependent variables 4.617 0.220 0.372 0.482 5.099 0.219 0.506 0.594 226.031 0.104 0.104 Rsquared 0.029 0.003 0.009 0.035 0.005 0.032 0.014 0.018 0.016 0.009 0.004 Individuals 2719 2719 2719 2719 2719 2719 2719 2719 2719 2719 2719 Panel B: Jobs characteristics before and after the job change within firm and occupation changers Female × year after 0.046 0.103 * 0.011 −0.007 −20.260 −0.060 −0.131 * −0.153 ** 21.816 −0.115 ** −0.035 (0.053) (0.054) (0.036) (0.032) (483.871) (0.048) (0.079) (0.071) (14.263) (0.049) (0.049) 1 year after 0.123 *** −0.030 0.033 −0.022 377.790 −0.028 0.006 0.028 13.199 −0.040 0.057 (0.031) (0.039) (0.024) (0.020) (364.812) (0.034) (0.055) (0.049) (9.661) (0.030) (0.040) Female −0.161 *** −0.048 −0.051 * 0.119 *** −488.739 ** 0.127 ** 0.139 ** 0.122 ** −38.079 *** 0.102 ** −0.061 * (0.047) (0.040) (0.030) (0.027) (235.755) (0.055) (0.055) (0.050) (11.476) (0.044) (0.033) (Continues) 20 of 32 | SANDNER and YÜKSELEN Dependent variables Median daily Share of Share of Share of Log Horizontal Vertical Horizontal or Occupation Occupation Occupation Log wage of fulltime employees Parttime employees High qualified employees Women in a firm Firm size Mismatch Mismatch Vertical mismatch Rank Rank < quantile 10 Rank > quantile 90 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Means of dependent 4.558 0.203 0.338 0.484 4.829 0.296 0.619 0.721 223.341 0.128 0.114 Rsquared 0.113 0.039 0.059 0.085 0.035 0.065 0.060 0.044 0.054 0.047 0.053 Individuals 312 312 312 312 312 312 312 312 312 312 312 Note: Panel A documents the gender differences in the first job characteristics between firm and occupation changers and stayers based on the OLS model specified in Equation (4) The sample size is 2719, including stayers (Column 2, Table 4) and firm and occupation changers (Column 5, Table 4). Panel B uses firm and occupation characteristics as dependent variables, which are timevariant; that is, they may vary before and after the job change. This is similar to the OLS model specified in Equation (2), however, here the focus is only on firm and occupation changers, and the timevariant dependent variables are different in each column. In addition, we control only for graduation year. The sample size is 312 and includes only firm and occupation changers (Column 5, Table 4). The estimations include a female dummy, a dummy variable for 1 year after the job change (=0 for the first job, =1 for the new job 1 year after the first job) and an interaction of these dummies. Robust standard errors in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels. TABLE 6 (Continued) | 21 of 32 SANDNER and YÜKSELEN Specifically, Columns (6) and (7), Panel B of Table 6 show that female job changers are more likely than male job changers to work in horizontally (by 12.7 percentage points) and vertically (by 13.9 percentage points) mismatched first jobs after graduation. However, compared with males, female job changers reduce the frequency of vertical mismatch 1 year after the first job by 13.1 percentage points. After correcting the vertical mismatch, one might expect women to receive a higher wage as soon as they correct their mismatch. Although Column (9), Panel B of Table 6 shows that female job changers work in lowerranked occupations in their first job after graduation, they do not move to (significantly) higherpaid occupations on average compared to men (Row 1, Column 9, Panel B). However, Figure B2 in the Appendix 2 shows that the occupational rank distributions of male and female job changers are quite different at the first job, as females are less likely to work in higherpaid occupations and more likely to work in lowerpaid occupations than males are. After the job change, the distributions of males and females converge, especially in the lower tail, as females predominantly move from lowerpaid occupations to higherpaid occupations. In line with the convergence in the lower tail, Panel B, Column (10) of Table 6 shows that after a firm and occupation change, women reduce the probability of being in the bottom decile of ranked occupations relative to men by 11.5 percentage points. Overall, the results of Tables 5 and 6 demonstrate that females are more likely than males to start in an occupation that is in the bottom tail of the occupation rank distribution and change to higherpaid occupations if they switch both occupation and firm. Moreover, they start in occupations in which they are overqualified and correct this vertical mismatch by changing both occupation and firm. As our data show that correcting the vertical mismatch at the first job explains the decrease in the gender wage gap, the question arises as to why women need to change both firms and occupations to correct the mismatch. We now test common hypotheses in the gender wage gap literature that may explain our findings. Different types of discrimination: A first potential explanation for the decline in vertical mismatch and the gender wage gap within 1 year after the first job is that firms discriminate against women at the hiring stage, when employers do not have sufficient information about the productivity of new hires (Altonji & Blank, 1999). As a first type of discrimination, namely, “screening discrimination”, Pinkston (2003) documents that the productivity signals that employers receive from females are noisier than from males. Therefore, productivity signals at the hiring stage have a smaller or no effect on women's wages, while they have a larger effect on men's wages. In our case of university graduates, since the employer is able to observe the curricula vitae of the applicants, the final university grades may comprise the only signal for the employer. We would expect men with higher grades to not change jobs because they already have a good match in their first job. In contrast, women with higher grades may experience a mismatch compared to their male counterparts at the beginning of their careers and thus change jobs to correct the mismatch. However, our results show that female movers and stayers have better grades on average than their male counterparts do, and both female and male movers have worse grades than stayers. Nevertheless, the difference in the gender gap between movers and stayers is insignificant, as the interaction term is insignificant (Column 7, Table 5). The literature on the gender wage gap suggests that females may face statistical discrimination in the labor market. Accordingly, employers may expect lower productivity from females and hire them for less suitable jobs. Consequently, conditional on being hired, females work in more mismatched jobs and receive lower initial wages at the beginning of their careers. However, over time, as employers learn about the actual productivity of new hires (“employer learning”), such mismatches could be corrected, resulting in higher wages for women (Altonji & Blank, 1999; Altonji & Pierret, 2001; Pinkston, 2003). If females face statistical discrimination, we would expect the gender wage gap to likely narrow not only for those who change firms and occupations but also for stayers. Since we do not find a significant reduction in the gender wage gap for stayers, statistical discrimination is unlikely to explain the differential returns to changing occupations and firms. 22 of 32 | SANDNER and YÜKSELEN Another form of discrimination suggested by the literature is tastebased discrimination, where employers pay women lower wages to compensate for their (or their coworkers') disutility.24 The greater mismatch and lower initial wage of female movers relative to male movers (Table 6) may indicate some form of tastebased discrimination (Becker, 1971). However, if firms discriminate against women, switching to nondiscriminatory firms should be sufficient for women to improve their wages relative to those of men, while an additional change in occupation should not be necessary. As Table 4 shows, this is not the case, as the gender wage gap does not narrow significantly for those who change only firms. Furthermore, if tastebased discrimination explains the gender wage gap, we should observe that women who change firms will move to firms with more women, as these firms typically discriminate less. Contrary to this hypothesis, women who change their firm and occupation are more likely to work in firms with a greater share of women in their first job (Panel A of Table 6). Moreover, our estimation results show that women do not switch to firms with a greater female share than men (Column 4, Panel B of Table 6). Risk aversion, confidence and job searching time: The literature shows that risk aversion and (over)confidence may be an important component of early career job search, with women typically having higher levels of risk aversion (Cortés et al., 2023; Niederle & Vesterlund, 2007) and lower levels of (over)confidence compared to men (AdameczVölgyi & Shure, 2022). More riskaverse and less selfconfident women may have lower reservation wages at the beginning of their careers (Acemoglu & Shimer, 1999; Cox & Oaxaca, 1992; Feinberg, 1977; Pannenberg, 2010; Pissarides, 1974) and thus accept job offers earlier, even if the job pays less and is not a good match (Cortés et al., 2023). However, these women may not be satisfied with lower wages and mismatches and change jobs when they find a higherpaid and better match job. In this case, we would expect women to spend less time searching for a job after graduation than men would, and women who find a job more quickly would be more likely to change jobs. Column (8) of Table 5 shows that although job changers find their first job slightly earlier than stayers, there is no significant gender difference in job search duration for stayers and job changers in the fields of economics, business, humanities and social sciences. In addition, wages do not decrease significantly with the duration of the job search. Job amenities and gender norms: As an alternative explanation, a growing body of literature has reported that women prefer nonwage job amenities such as flexibility or meaning, relevance, or responsibility for the occupation more than men, who have a greater preference for wages (Brenøe & Zölitz, 2020; Flabbi & Moro, 2012; Goldin, 2014; Goldin & Katz, 2011) Additionally, women may follow gender norms when they make their initial career decisions. Changing preferences for certain job attributes or shifting away from gender norms may also be a mechanism for job change. On the one hand, females may change to more flexible jobs in anticipation of having children in the future. However, these changes may not lead to higher wage gains. On the other hand, at the beginning of their careers, women may prefer lowerpaying jobs with a vertical mismatch to compensate for some job amenities and to respond to certain gender norms; however, over time, they change their preferences and switch to higherpaying and less flexible jobs. Based on the assumption that larger firms offer more flexible work arrangements (Albrecht et al., 2018), we test whether women switch to smaller firms. We do not find that females are more likely than men to sort into larger or smaller firms as a result of a job change (Column 5, Table 6). However, our data do not cover other proxies for job amenities, such as the meaning of jobs, schedule adaptability, or telecommuting (De Schouwer & Kesternich, 2022) or whether a job fulfills gender norms. Therefore, we believe that changing preferences for job amenities or shifts away from gender norms may still be important explanations for why women reduce their vertical mismatches and increase their wages when they switch firms and occupations. Overall, this section reveals that women tend to make less optimal decisions than men when they make choices regarding their first job after graduation, leading to mismatches. Women attempt to correct such mismatches by switching firms and occupations. However, since they cannot fully close the initial wage gap through these changes, the results suggest that women should aim to make more informed choices for their first job to avoid the 24 Becker (1964) shows in the model that firms practicing tastebased discrimination cannot survive in the competitive market in the long run. | 23 of 32 SANDNER and YÜKSELEN need for later corrections. Enhanced counseling programs could provide the information necessary to help women make better initial job choices. 7 | CONCLUSION Although many studies have investigated the gender wage gap, the existence and potential explanations for early career gender wage differences remain unclear. This paper analyses the gender wage gap among graduates of a German university with a master's degree or equivalent at the beginning of their careers and over the first years after their labor market entry. We rely on a unique dataset that links administrative data on graduates of a German university with employment registers of the German social security system. This dataset includes extensive information on students' sociodemographic characteristics, educational and labor market outcomes, as well as the exact timing of graduation, labor market entry, and any job changes. We find a significant gender wage gap among university graduates in their first job, which persists even after we include an extensive set of controls. The largest gender wage gap is observed among humanities and social sciences graduates, where the share of females is highest and the average daily wage is lowest. We find no significant gender differences in the wages of mathematics, natural sciences, or medical graduates in their first job after graduation. Moreover, in contrast to previous studies, we find an immediate decrease in the gender wage gap 1 year after labor market entry, which remains relatively stable thereafter. Further analysis shows that the decline in the gender wage gap is concentrated among individuals who change firms and occupations after their first job with a degree in economics, business, humanities, or social sciences. To explain this decrease in the gender wage gap, we also show that female graduates are more likely to start their careers in jobs for which they are overqualified and subsequently correct this skill mismatch, leading to an increase in wages. Correcting this mismatch is costly for females, which may be an additional explanation for the wage gap that emerges later in their careers. Universities have an important opportunity to mitigate the risk of future skill mismatches by implementing counseling interventions. These interventions can provide valuable information on effective job search strategies that can overcome gender norms in career choice, and potential wage losses resulting from skill mismatches, particularly for female students. Our study also highlights significant differences in labor market entry and early career paths depending on the chosen field of study. For this reason, counseling programs that help students understand their career prospects should be tailored specifically to each field of study. By implementing such counseling, universities can provide graduates with the insight they need to navigate the dynamic labor market successfully. ACKNOWLEDGEMENTS We are thankful for valuable comments from Silke Anger, Alex Bryson, Bernd Fitzenberger, Astrid Kunze, Markus Nagler, Uta Schönberg, Nikki Shure, David Wilkinson, and all participants at the Annual LERN Conference 2019, ESPE 2019, AASLE 2019, Workshop on Labour Economics 2021, Research Seminar at the TH Nürnberg, International Conference on The German Labor Market in a Globalized World (ZEW) 2022. Additionally, we are grateful to Joachim Möller and Christopher Rust for support. ORCID Malte Sandner https://orcid.org/0000-0001-8579-4775 Ipek Yükselen https://orcid.org/0009-0008-8885-740X REFERENCES Acemoglu, D. & Shimer, R. (1999) Efficient unemployment insurance. Journal of Political Economy, 107(5), 893–928. AdameczVölgyi, A. & Shure, N. (2022) The gender gap in top jobs – the role of overconfidence. Labour Economics, 79, 102283. 24 of 32 | SANDNER and YÜKSELEN Adda, J., Dustmann, C. & Stevens, K. (2017) The career costs of children. Journal of Political Economy, 125(2), 293–337. Albrecht, J., Bronson, M.A., Thoursie, P.S. & Vroman, S. (2018) The career dynamics of highskilled women and men: evidence from Sweden. European Economic Review, 105, 83–102. Altonji, J.G., Arcidiacono, P. & Maurel, A. (2016) The analysis of field choice in college and graduate school: determinants and wage effects. In: Hanushek, E., Machin, S. & Woessmann, L. (Eds.) Handbook of the economics of education, Vol. 5. Amsterdam, The Netherlands: Elsevier, pp. 305–396. Altonji, J.G. & Blank, R.M. (1999) Chapter 48: race and gender in the labor market. In: Ashenfelter, O., Layard, R. & Card, D. (Eds.) Handbook of labor economics, Vol. 3, Part C. Amsterdam, The Netherlands: Elsevier, pp. 3143–3259. Altonji, J.G., Blom, E. & Meghir, C. (2012) Heterogeneity in human capital investments: high school curriculum, college major, and careers. Annual Review of Economics, 4(1), 185–223. Altonji, J.G. & Pierret, C.R. (2001) Employer learning and statistical discrimination. The Quarterly Journal of Economics, 116(1), 313–350. Angrist, J.D. & Pischke, J.- S. (2009) Mostly harmless econometrics: an empiricist's companion. Princeton, NJ: Princeton University Press. Antoni, M., Ganzer, A., Vom Berge, P. et al. (2016) Sample of integrated labour market biographies (SIAB) 19752014. FDZDatenreport, 04:2016. Institute for Employment Research (IAB), Nuremberg, Germany. Babcock, L. & Laschever, S. (2009) Women don't ask: negotiation and the gender divide. Princeton, NJ: Princeton University Press. Bartel, A.P. & Borjas, G.J. (1981) Wage growth and job turnover: an empirical analysis. In: Rosen, S. (Ed.) Studies in labor markets. Cambridge, MA: National Bureau of Economic Research, Inc, pp. 65–90. Becker, G. (1971) The economics of discrimination. Chicago, IL: University of Chicago Press. Becker, G.S. (1964) Human capital: a theoretical and empirical analysis, with special reference to education. New York, NY: Columbia University Press. Becker, G.S., Hubbard, W.H. & Murphy, K.M. (2010) Explaining the worldwide boom in higher education of women. Journal of Human Capital, 4(3), 203–241. Bertrand, M. (2011) New perspectives on gender. In: Handbook of labor economics, Vol. 4. Amsterdam, The Netherlands: Elsevier, pp. 1543–1590. Bertrand, M., Goldin, C. & Katz, L.F. (2010) Dynamics of the gender gap for young professionals in the financial and corporate sectors. American Economic Journal: Applied Economics, 2(3), 228–255. Black, D.A., Haviland, A.M., Sanders, S.G. & Taylor, L.J. (2008) Gender wage disparities among the highly educated. Journal of Human Resources, 43(3), 630–659. Blau, F.D. & Kahn, L.M. (2017) The gender wage gap: extent, trends, and explanations. Journal of Economic Literature, 55(3), 789–865. Boudarbat, B. & Montmarquette, C. (2009) Choice of fields of study of university Canadian graduates: the role of gender and their parents' education. Education Economics, 17(2), 185–213. Brenøe, A.A. & Zölitz, U. (2020) Exposure to more female peers widens the gender gap in STEM participation. Journal of Labor Economics, 38(4), 1009–1054. Bruns, B. (2019) Changes in workplace heterogeneity and how they widen the gender wage gap. American Economic Journal: Applied Economics, 11(2), 74–113. Cortes, P. & Pan, J. (2020) Children and the remaining gender gaps in the labor market. Technical report, National Bureau of Economic Research. Cortés, P., Pan, J., Pilossoph, L., Reuben, E. & Zafar, B. (2023) Gender differences in job search and the earnings gap: evidence from the field and lab. The Quarterly Journal of Economics, 138(4), 2069–2126. Cox, J.C. & Oaxaca, R.L. (1992) Direct tests of the reservation wage property. The Economic Journal, 102(415), 1423–1432. De Schouwer, T. & Kesternich, I. (2022) It's all about the meaning: work meaning, job flexibility, and the gender wage gap. Discussion Paper Series, DPS 22.04. Antwerp: Department of Economics, KU Leuven. Del Bono, E. & Vuri, D. (2011) Job mobility and the gender wage gap in Italy. Labour Economics, 18(1), 130–142. Deming, D.J. & Noray, K. (2020) Earnings dynamics, changing job skills, and STEM careers. The Quarterly Journal of Economics, 135(4), 1965–2005. Dorner, M., Heining, J., Jacobebbinghaus, P. & Seth, S. (2010) Stichprobe der Integrierten Arbeitsmarktbiografien (SIAB) 19752008. FDZ-Methodenreport, 01:2010. Institute for Employment Research (IAB), Nuremberg, Germany. Dustmann, C., Ludsteck, J. & Schönberg, U. (2009) Revisiting the German wage structure. The Quarterly Journal of Economics, 124(2), 843–881. Farber, H.S. & Gibbons, R. (1996) Learning and wage dynamics. The Quarterly Journal of Economics, 111(4), 1007–1047. Federal Statistical Office. (2022) Mothers at 1st birth and Länder. Wiesbaden, Germany: Federal Statistical Office. Feinberg, R.M. (1977) Risk aversion, risk, and the duration of unemployment. The Review of Economics and Statistics, 59(3), 264–271. | 31 of 32 SANDNER and YÜKSELEN TABLE C5 Descriptive statistics – stayers and firm and occupation changers. Dependent variables Stayers Firm and occupation changers Male Female Male–female Male Female Male–female Panel A: Economics, business, humanities and social sciences Age at the first job 27.467 26.674 0.793*** 27.899 26.990 0.908*** NonGerman 0.014 0.038 −0.024*** 0.017 0.015 0.003 Duration of study 5.527 5.463 0.064 5.576 5.716 −0.140 Working during studying 0.659 0.784 −0.124*** 0.676 0.787 −0.110** Apprenticeship 0.074 0.063 0.011 0.133 0.096 0.037 Origin in the same federal state as the university 0.843 0.876 −0.033** 0.867 0.875 −0.008 Final Uni. grade 2.187 1.976 0.210*** 2.223 2.082 0.140** Duration of job search 3.388 3.518 −0.129 3.110 3.010 0.100 Median daily log wage of fulltime employees in a firm 4.662 4.622 0.041*** 4.543 4.416 0.126*** Share of parttime employees in a firm 0.791 0.757 0.035*** 0.809 0.767 0.042* Share of high qualified employees in a firm 0.395 0.379 0.016 0.332 0.302 0.030 Share of women in a firm 0.452 0.524 −0.071*** 0.446 0.561 −0.115*** Log firm size 5.222 5.119 0.102 4.881 4.575 0.306 Horizontal mismatch occupation 0.148 0.270 −0.122*** 0.267 0.397 −0.130** Vertical mismatch 0.469 0.475 −0.005 0.591 0.713 −0.122** Horizontal or vertical mismatch 0.538 0.588 −0.051** 0.688 0.809 −0.121** Occupation rank 236.061 220.823 15.238*** 228.381 190.772 37.609*** Occupation rank < quantile 10 0.090 0.118 −0.029** 0.102 0.213 −0.111*** Occupation Rank > Quantile 90 0.121 0.085 0.036*** 0.119 0.059 0.060* Observations 1503 904 176 136 Panel B: Mathematics, natural sciences and medical studies Age at the first job 27.476 26.993 0.482*** 28.120 27.027 1.093*** NonGerman 0.014 0.018 −0.004 0.024 0.060 −0.036 Duration of study 5.692 5.870 −0.178** 5.938 6.008 −0.070 Working during studying 0.660 0.661 −0.002 0.747 0.720 0.027 (Continues) 32 of 32 | SANDNER and YÜKSELEN Dependent variables Stayers Firm and occupation changers Male Female Male–female Male Female Male–female Apprenticeship 0.043 0.041 0.001 0.133 0.120 0.013 Origin in the same federal state as the university 0.914 0.855 0.059*** 0.880 0.820 0.060 Final Uni. grade 1.858 2.007 −0.149*** 1.835 1.845 −0.010 Duration of job search 3.330 3.671 −0.341** 2.748 3.462 −0.714 Median daily log wage of fulltime employees in a firm 4.576 4.499 0.077*** 4.306 4.236 0.070 Share of parttime employees in a firm 0.689 0.609 0.080*** 0.677 0.625 0.053 Share of high qualified employees in a firm 0.317 0.242 0.075*** 0.219 0.198 0.021 Share of women in a firm 0.552 0.714 −0.162*** 0.522 0.755 −0.233*** Log firm size 5.594 5.690 −0.095 3.934 4.046 −0.111 Horizontal mismatch occupation 0.196 0.084 0.112*** 0.425 0.260 0.165* Vertical mismatch 0.121 0.100 0.021 0.402 0.460 −0.058 Horizontal or vertical mismatch 0.263 0.139 0.124*** 0.575 0.500 0.075 Occupation rank 285.685 290.981 −5.296 227.080 199.800 27.280 Occupation rank < quantile 10 0.116 0.098 0.018 0.172 0.180 −0.008 Occupation rank > quantile 90 0.372 0.584 −0.212*** 0.172 0.160 0.012 Observations 1148 570 87 50 Note: This table shows summary statistics of graduates' personal, pregraduation, postgraduation and first job characteristics of stayers and of firm and occupation changers. The sample consists of graduates with a master's degree or equivalent who work in a fulltime job as their first job after graduation and who have a wage spell 1 year after their first job. ***, **, and * denote significance at the 1%, 5%, and 10% levels. TABLE C5 (Continued)