Peer-reviewed Publications:

Earnings Inequality and Working Hours Mismatch (with Carsten Schröder). Labour Economics 76 (2022): 102184

Based on data from the German Socio-Economic Panel, we document a significant rise in monthly earnings inequality between 1993 and 2018. The main contributors are inter-temporal increases in working hours inequality and increases in the covariance between working hours and hourly wages, while changes in the distribution of hourly wages play a minor role. Applying a novel double decomposition technique reveals that these results are particularly pronounced in the growing groups of female employees and service sector employees. If employees had been able to realize their desired optimal working hours, the increase in inequality would have been more moderate. This is mainly due to the fact that employees with low hourly wages work less than desired, a finding that is reinforced over time—even after controlling for various covariates. 


Inspections and Compliance: Enforcement of the Minimum-Wage Law (with Alexandra Fedorets) FinanzArchiv: Public Finance Analysis 77.1 (2021): 1-58

This paper is the first to empirically study the relationship between spatial distributions of labor market inspections and noncompliance with Germany's minimum-wage law. Combining novel administrative data with large-scale longitudinal survey data, we document that the inspection probability is higher in regions with higher noncompliance. This implies risk-based allocation of the inspection efforts and, hence, its endogeneity. Using fixed effects and an instrumental-variable approach, we show that higher inspection efforts have a limited effect on compliance. Based on a theoretical framework and international evidence, we discuss challenges for law enforcement, the political importance of compliance, and possible improvement measures. 

Recent Working Papers

Out For Good: Labor Market Effects of Transitory and Persistent Health Shocks (with Johannes König)

Health shocks limit individuals’ participation in the labor market and pose a major risk to household welfare. In this paper, we derive two novel health shock indicators using machine learning based on sick days and hospitalizations: one for transitory and one for persistent shocks. In an event study framework, we show their respective effects on employment, yearly working hours, and labor earnings, but also partner earnings and household net income. Persistent shocks induce large negative employment effects that end up impacting household net incomes. In contrast, transitory shocks induce only minor employment effects that leave household net incomes unaffected.

Is There a Desired Added Worker Effect? Evidence from Involuntary Job Losses (with Rick Glaubitz)

Existing research has found little to no evidence for an added worker effect. However, studies to date have only analysed individuals' actual labor supply responses to their partners' job loss, neglecting to consider a potential mismatch between desired and actual labor supply adjustments. Using data from the German Socio-Economic Panel (SOEP), we study individuals' changes in actual and desired working hours after their partners' involuntary job loss in an event study design. Our results show that neither desired nor actual working hours change significantly. Thus, we provide first evidence that the absence of the added worker effect is in line with individuals' stated labor supply preferences and is not the result of an inability to realise desired working hours.

Further Work in Progress

Levels and Drivers of Lifetime Earnings (with Carsten Schröder and Yogam Tchokni)

We use administrative earnings records linked with rich panel survey data to provide empirical facts about the levels and drivers of lifetime earnings in Germany. Our novel data base allows us to provide a much more comprehensive picture than previous studies, as it also covers self-employed, civil servants and individuals with low labor market attachment. We complement the earnings information in the administrative records with rich background information from the survey, so that we can evaluate drivers of lifetime earnings. We show that exogenous determinants like sex or parental background as well as educational or occupational choices and household formation are important drivers of lifetime earnings. Labor market participation operates as a mediator that explains some but not all of the associations between lifetime earnings and its drivers.

The beyondpareto command for optimal extreme value index estimation (with Johannes König, Isabella Retter, Christian Schluter, and Carsten Schröder)

This paper introduces the command beyondpareto which estimates the extreme value index for distributions that are Pareto-like, i.e. whose upper tails are regularly varying and eventually become Pareto. The estimation is based on rank-size regressions, and the threshold value for the upper order statistics included in the final regression is determined optimally by minimizing the asymptotic mean-squared error (AMSE). An essential diagnostic tool for evaluating the fit of the estimated extreme value index is a Pareto quantile-quantile (QQ) plot, provided in the accompanying command pqqplot. The usefulness of our estimation approach is illustrated in several real-world examples focusing on the upper tail of the German wealth and city size distribution.

Dealing with income censoring in register data (with Johannes König, Isabella Retter, Christian Schluter, Carsten Schröder, and Yogam Tchokni)

Register data on income and wealth are often censored. For example, in Germany, earnings in social security insurance registers are censored at the assessment or contribution ceiling. Here we provide a guide for the imputation of censored data if the top tail of the distribution is Pareto-like. The Pareto distribution comes with two parameters: a minimum threshold value, from which the existence of a power function applies, and a shape parameter (also known as the tail or Pareto index). It stipulates that, when plotted in a log-log plot, the distribution is represented by a straight line. First, drawing on non-censored earnings data, we show that the right tail of the German earnings distribution is Pareto like. Second, we show how plausible tail parameters can be estimated from censored data. In particular, this involves, as we will show, the appropriate choice of the Pareto threshold and the handling of censored observations in the construction of the distribution function.

Job Tasks and Workers' Health

Technological progress and automation are rapidly changing the task composition of jobs. In this paper, I analyze the impact of occupational routine task intensity on workers' mental and physical health. By combining individual-level health information of German workers with data on occupational task profiles and applying an instrumental variable strategy, I find that male and female workers are oppositely affected by occupational routine task intensity. For women routine tasks are more likely cognitive routine tasks that negatively affect mental health. For men, routine tasks are more likely manual routine tasks, that negatively affect physical health but have a positive effect on mental health. When considering the overall workforce, the effects on mental health balance out, but a significant negative effect of routine task intensity on physical health remains.