IAB-Colloquium (en) Archive - IAB - Institut für Arbeitsmarkt- und Berufsforschung

The Role of Industries in Rising Inequality

This paper provides new evidence on the evolution of earnings inequality over several decades.

While the sharp rise in earnings inequality since the 1980s is well documented, its underlying structure and causes remain contested.

This paper provides new evidence on the evolution of earnings inequality over several decades, based on exceptionally detailed Italian administrative data. We first offer a comprehensive decomposition of the inequality trend across sectors and firms, showing that the predominant component of the rise is sectoral rather than firm-specific. This shift is closely associated with increased sorting between worker skills and sectoral attributes, as well as heightened segregation of workers across sectors. To interpret these patterns, we develop a parsimonious Roy-style model of sectoral choice with assortative matching. Calibrating the model to the data, we find that the observed increase in segregation and sorting can be accounted for by a rise in the dispersion of skill-biased sectoral productivity.

Our findings highlight the central role of sector-level heterogeneity in shaping long-run trends in earnings inequality.

The Eroding Lifetime Return to Cognitive Skills

This study shows how lifetime labor-market returns to cognitive and social skills have changed across Swedish birth cohorts.

We study how lifetime labor-market returns to cognitive and social skills have changed across Swedish birth cohorts. We combine administrative records for cohorts born 1951–1980 with skills measured at military enlistment and follow earnings, employment, and wages from 1985 to 2023. Because complete working lives are not observed for every cohort, we estimate age profiles and predict lifetime outcomes. Social skills are a stronger predictor of labor-market success than cognitive skills, largely through employment, although both skills predict wages. Across cohorts, lifetime returns to cognitive skills decline markedly, while returns to social skills increase. We also compare men and women, using brothers’ skills as proxies for own skills in both groups. Skill returns increased more among women than among men, with the largest relative increase for women’s social skills. The results highlight substantial variation in skill returns by skill type, age, cohort, and gender.

Joint with Erika Forsberg, Peter Fredriksson, Georg Graetz, and Björn Öckert

The Organisational Context of Ethnic Discrimination. Results From a Cross-National Factorial Survey Experiment

The study addresses this gap through a cross-nationally harmonised factorial survey experiment.

While ethnic discrimination in hiring is well documented, little is known about why some organisations – including private companies and public institutions – discriminate more than others. We address this gap through a cross-nationally harmonised factorial survey experiment (FSE) embedded in a comprehensive questionnaire, combining the rigour of experimental design with detailed organisational data.

The survey was conducted among more than 2,000 recruiters evaluating candidates for medium-skilled jobs in four European countries: Germany, Norway, Poland, and Romania. Results show that hiring discrimination is lower in organisations that make hiring decisions collectively, but higher in positions involving frequent customer interaction. Moreover, diversity policy measures implemented within organisations are strongly associated with lower levels of discrimination. Among these, inclusive hiring practices, mentoring or buddy programmes, and support systems for foreign workers appear particularly effective.

The study further demonstrates that the relationship between discrimination and organisational characteristics, such as size and ownership, is largely mediated by the presence of diversity policy measures.
Co-autors: Dominik Buttler, Vegar Bjørnshagen, Marta Palczyńska, Mateusz Smoter

Gender-inclusive language in labour markets

This study leverages a digital field experiment in a labour market setting to examine whether attitudinal effects translate into real-world behaviour.

This study leverages a digital field experiment in a labour market setting to examine whether attitudinal effects translate into real-world behaviour, while offering broader lessons for the design and interpretation of such experiments.

Gender-inclusive language in job descriptions has been shown to impact perceived fit, belonging, and occupational attractiveness, but its impact on actual behaviour remains uncertain. We used a crowdworking platform as an experimental labour market, where sign-up for our advertised study itself serves as an outcome variable. In this setting, we thus tested whether the use of gender-inclusive language in a job advertisement for a maths task increases both the share of women who sign-up and their subsequent performance.

We find no effect of gender-inclusive language on either the share of women or their performance. We provide a look behind the scenes of running a digital field experiment, reflecting on challenges along the way. We used the opportunity to learn about design elements that may have prevented the predicted effect from arising, and the moderating theoretical variables they allude to. One key conjecture that emerged from our assessment is that in tight labour markets, such as the crowdfunding platform we studied, participation is less gendered, with an attenuating effect on gendered performance expectations.

The paper is co-authored with Klarita Gërxhani and Arnout van de Rijt.

Identifying Long-Run Neighborhood Effects via Race-Based Neighborhood Sorting

The study is about children’s adult outcomes are influenced by both their parents’ socioeconomic status (SES) and the SES of the neighborhood.

Children’s adult outcomes are influenced by both their parents’ socioeconomic status (SES) and the SES of the neighborhood where they grew up. We explicitly distinguish between these factors to identify long-run neighborhood effects. We estimate an ordered probit model of neighborhood choice and use it to infer how population shares of parental SES in each neighbor hood vary along both observed and unobserved dimensions. We then regress the average observed outcomes in neighborhoods on these population shares to identify potential outcomes.

Using race based neighborhood sorting as an instrument, we estimate that about half of recent racial inequality in intergenerational mobility is explained by residential segregation.

Joint with Daniel Hartley and Chris Muris

Beyond the Sidelines: Shaping AI Infrastructure with Labor Market Data

This talk argues that as AI systems are deployed into high-risk labor market domains like matching and profiling.

As social scientists, we focus heavily on how AI is changing the labor market. Historically, our discipline has not viewed itself as central to shaping these computational models. This perspective falls short. Both methodologically and substantively, data-producing institutions like the IAB have an urgent opportunity and responsibility, to actively intervene in AI development.

Generative AI pipelines are currently saturated with unscientific behavioral measurements. AI agents routinely deploy unstandardized, fixed-choice questions generated by the models themselves, while Reinforcement Learning from Human Feedback (RLHF) relies on non-probability annotator pools that introduce systemic selection bias. The fundamental machine learning challenge - determining whose values and life circumstances a model represents - is, at its core, a target-population estimation problem. Social science possesses the exact estimation machinery required to diagnose and repair these data structures.

This talk argues that as AI systems are deployed into high-risk labor market domains like matching and profiling, compliance with the EU AI Act demands rigorous auditability. This regulatory shift creates an immediate mandate for the methodologies we produce. In this presentation, we will map how labor market data can directly shape AI pipelines through representative benchmarks, validated instruments, and advanced error diagnosis - including adapting record-linkage traditions to measure what models get wrong about specific demographic groups. Furthermore, we will explore how sampling theory and classic survey methodology can govern the ways AI agents ask questions.

At the UN level, AI readiness for data collected by official institutions has become a central theme. I hope to discuss this element and its broader institutional implications following the presentation.

Is Remote Work Attractive for Women Living in Peripheral Areas? A Survey-Based Experiment in Six EU Countries

This study is the first to explore the extent of job offers with remote work options.

Promoting remote work among women living in peripheral areas is a potential tool to limit gender and geographical imbalances in the labour market. This study is the first to explore the extent to which women living in peripheral municipalities (compared with those from metropolitan ones) are attracted to job offers with remote work options.

The study is based on a CAWI survey with a conjoint experiment conducted in Spring 2026 among 3,785 prime-aged women living in France, Italy, the Netherlands, Poland, Romania, and Sweden.

The initial results indicate that, on average, fully remote work (compared to fully on-site work) increases the probability of accepting a job offer to the same degree in both metropolises and peripheral municipalities, by around 14 percentage points. What is relevant, remote work remains attractive for peripheral respondents with lower than higher education, as well as for those with poor digital skills.

Breaking Barriers or Reinforcing Inequalities? The Impact of Affordable Childcare on Entrepreneurship

This study exploits exogenous variation in the pricing of childcare caused by a nationwide reform.

We exploit exogenous variation in the pricing of childcare caused by a nationwide reform to estimate the impact of lower childcare costs and the propensity of individuals to start a firm. The reform capped the childcare prices, whereas the exogenous variation in the reduction of childcare prices comes from the pre-reform prices, which depended on the place of residence and family type. Decreasing the cost of childcare lowers both the capital and time constraint, which can lead to individuals becoming entrepreneurs.

We evaluate such potential mechanisms both for males and females separately in a difference-in-difference setting up to 5 years post reform. We find that reducing childcare costs increases entrepreneurial activity for both mothers and fathers.