Research

Abstract: We study how the threat of entry affects service quantity and quality of general practitioners (GPs). We leverage Germany’s needs-based primary care planning system, in which the likelihood of new GPs reduces by 20 percentage points when primary care coverage exceeds a cut-off. We compile novel data covering all German primary care regions and up to 30,000 GP-level observations from 2014 to 2019. Reduced threat of entry lowers patient satisfaction for incumbent GPs without nearby competitors but not in areas with competitors. We find no effects on working hours or quality measures at the regional level including hospitalizations and mortality.

Abstract: This study quantifies the relationship between workplace digitalization, characterized by the increasing use of frontier technologies, and workers’ health outcomes using novel and representative German linked employer-employee data. Analyzing changes in individual-level technology use between 2011 and 2019, we find that digitalization induces similar shifts toward more complex and service-oriented tasks for all workers but exacerbates health disparities between cognitive and manual workers. Unlike mature, computer-based technologies, recent frontier technologies significantly degrade manual workers’ subjective health and increase sick leave, while leaving cognitive workers unaffected. Our findings suggest that these negative effects are mitigated in firms that provide training and support to help workers adjust to technological changes.

Abstract: New technologies drive productivity growth, yet the distribution of gains may be unequal. We study how labor market institutions – specifically shop-floor worker representation – mediate the impact of automation. Combining German individual-level administrative records with plant-level data on industrial robot adoption, we find that works councils reduce the separation risk for incumbent workers during automation events. When labor markets are tight and replacement costs are high, incumbent workers become more valuable from the firm’s perspective. Consequently, we document that the moderating effects of works councils diminish. Older workers, who face greater challenges reallocating to new employers, benefit the most from organized labor in terms of wages and employment. Finally, we observe that works councils do not hinder robot adoption; rather, they spur the use of higher-quality robots, encourage more worker training during robot adoption, and foster higher productivity growth thereafter.

Abstract: Healthcare systems worldwide face increasing nurse shortages, but the consequences remain poorly understood. This paper studies how nurse scarcity in hospitals affects care provision and patient health. I exploit the 2011 Swiss franc stabilization, which increased the salience to cross-border commute from Germany to Switzerland and led to an outflow of nurses in German hospitals depending on their distance to the border. Using rich universal patient-, hospital-, and county-level German and Swiss administrative data in a matched difference-in-differences design, I show that border hospitals lose around 12 percent of their nursing staff. This leads to lower care intensity and a reallocation of services towards urgent cases (triage) while healthcare demand or supply outside hospitals remains unchanged. Consequently, in-hospital mortality rises by 4.4 percent – concentrated among emergency and older patients – and life expectancy decreases by 0.28 statistical life years, with no evidence of offsetting gains in Switzerland. These results highlight that nurse scarcity shapes hospital production and widens health disparities across patients and regions.

Abstract: Artificial intelligence (AI) is diffusing rapidly in the workplace, yet aggregate productivity gains remain limited. This paper examines the dual diffusion of AI  - through both formal, employer-led and informal, employee-initiated adoption - as potential explanation. Using a representative survey of nearly 10,000 employees in Germany, we document a high extensive but low intensive margin of usage: while 64 percent use AI tools, only 20 percent use them frequently. This diffusion is strongly skill-biased and depends less on establishment and regional characteristics. While formality is associated with more frequent usage, training, AI-based supervision, and higher perceived productivity gains, it does not broaden access. These patterns suggest that widespread informal usage can coexist with limited productivity effects when complementary investments and organizational integration lag behind.

Abstract: This paper capitalizes on a nationwide German healthcare reform that limited general practitioners (GPs) in selecting locations to establish their practices. Due to these entry restrictions, a significant number of regions experienced a decline in primary care coverage, depending on pre-reform characteristics and with minimal regional adjustments. By integrating comprehensive datasets at the patient, practice, and regional level within a multi-period difference-in-differences framework, we estimate how reduced access to primary care impacts service provision and population health. Furthermore, precise geo-location data of practices allow us to assess how the effect varies depending on the horizontal and vertical market structure.

Abstract: In Germany, more than half of employees are already using artificial intelligence (AI) in the workplace, although the majority do so informally. This suggests that employees perceive AI applications as supportive tools, yet it also highlights that firms are falling behind employees’ expectations regarding formal AI implementation. This report presents findings from the DiWa-Be 2.0 survey, a representative cross-sectional survey conducted in 2024, covering approximately 9,800 employees subject to social insurance contributions in Germany. The survey aims to provide a database for evaluating the impact of technological change - particularly AI - on the workplace. It specifically focuses on changes in job tasks and skill requirements, working conditions and organizational structures, participation in further training, and employee health. The results indicate that the use of AI varies substantially by individual and occupational characteristics such as occupational segment, education, age, and gender. While only about one-third of employees without formal educational qualifications use AI, nearly 80% of employees holding university degrees or advanced vocational qualifications (Meister or Techniker) do so. Initial multivariate analyses show that employees who have intensified their AI usage over the past five years report higher complexity in their job tasks, especially in the areas of writing, programming, and mathematics. Furthermore, intensified AI use correlates with increased job autonomy but also with higher work intensity. However, no statistically significant correlation between AI usage and employee health was found. Moreover, there are no differences between AI users and non-users regarding their participation in continuing training activities.