Research
Publications
One thing leads to another: Evidence on the scope and persistence of behavioral spillovers, with Alexander Götz and Renate Schubert, Journal of Public Economics, 236, 2024.
Evaluations of economic interventions usually focus on one target behavior. This study extends the evaluation scope to multiple untargeted behaviors. We evaluate a hot water saving intervention in a natural field experiment. Despite an exclusive focus on hot water, the intervention changes multiple behaviors. Notably, we find a 5.6 percent reduction in room heating energy consumption that persists one year after the intervention. We show that the room heating spillover has important welfare implications.
Do mobile applications foster sustainable mobility? Evidence from a field experiment, with Alexander Götz, Ioana Marinica, Luca Mosetti, and Renate Schubert, Swiss Journal of Economics and Statistics, 160:12, 2024.
Mobile applications hold promise to foster sustainable mobility behavior, but evaluations of their effectiveness are subject to a number of empirical challenges. We conduct a randomized controlled trial with three distinctive features: unobtrusive tracking of the control group, limited sample attrition, and a representative sample. In our study, 410 participants track their mobility behavior over a five week period. After one week, the treatment group engages with the user interface of the "Swiss Climate Challenge App". The user interface combines information on individual CO2 emissions with gamification features. We find a treatment effect that implies a 9.8% reduction in emissions caused by access to the mobile application. While we lack the statistical power to exclude a zero average effect, we find statistically significant emission reductions in the second half of the intervention period, among subjects in medium population density areas, and among men. Our findings suggest that mobile applications could generate considerable net benefits, but larger studies will be needed for validation.
Cheap search, picky workers? Evidence from a field experiment, Economics Bulletin 42(4), 2022.
Search frictions impede the labor market. Despite this indisputable fact, it is a priori unclear how job search costs affect search duration and unemployment: lower search costs make it easier to find a job, reducing search duration and unemployment, but may also increase the reservation wage, increasing search duration and unemployment. I collaborate with a recruiting company to directly test the effects of lower search costs in a field experiment among approximately 400 IT professionals in Switzerland. I find that workers are more likely to search for detailed job information, but not to file a job application, when search costs are lower. These findings are consistent with an increase in the reservation wage. Lower search costs might lead to picky workers, but fail to ultimately reduce search duration and unemployment.
Investigating survivorship bias: The case of the 1918 flu pandemic, with Joël Floris, Laurent Kaiser, Kaspar Staub, and Ulrich Woitek, Applied Economics Letters 29(21), 2022.
Estimates of the effect of fetal health shocks may suffer from survivorship bias. The fetal origins literature seemingly agrees that survivorship bias is innocuous in the sense that it induces a bias toward zero. Arguably, however, selective mortality can imply a bias away from zero. In the case of the 1918 flu pandemic, a suppressed immune system may have been protective against the most severe consequences of infection. We use historical birth records from the maternity hospital of Bern, Switzerland, to evaluate this possibility. Our results suggest that a careful consideration of survivorship bias is imperative for the evaluation of the 1918 flu pandemic and other fetal health shocks.
Cutting fertility? Effects of cesarean deliveries on subsequent fertility and maternal labor supply, with Martin Halla, Gerald J. Pruckner, and Pilar García-Gómez, Journal of Health Economics 72, 2020.
Despite the growing incidence of cesarean deliveries (CDs), procedure costs and benefits continue to be controversially discussed. In this study, we identify the effects of CDs on subsequent fertility and maternal labor supply by exploiting the fact that obstetricians are less likely to undertake CDs on weekends and public holidays and have a greater incentive to perform them on Fridays and days preceding public holidays. To do so, we adopt high-quality administrative data from Austria. Women giving birth on different days of the week are pre-treatment observationally identical. Our instrumental variable estimates show that a non-planned CD at parity decreases lifecycle fertility by almost 13.6%. This reduction in fertility translates into a temporary increase in maternal employment.
Working papers
The tragedy of the common heating bill, with Mateus Souza (draft available upon request).
We study the distortion that arises when apartment buildings share a common heating bill. This situation resembles the tragedy of the commons. Individual metering solves the problem, but does it improve welfare? We propose a theoretical framework with shared billing and environmental externalities to derive a sufficient statistic for the welfare effect of individual metering. We then analyze the introduction of individual metering in 265 apartment buildings in Switzerland. Our event study estimates show that individual metering reduces annual heating expenses. Machine learning estimates reveal substantial heterogeneity in treatment and welfare effects, suggesting a role for targeted interventions in this context.
Predictive algorithms for apartment rental demand using multimodal data, with Silvia Romanato, Alexander Sternfeld, Marguerite Thery, Sébastien Houde, and Andrew Sonta (draft available upon request).
Apartment listings include tabular data (e.g., number of rooms, construction year, price), text descriptions, and images. We use state-of-the-art methods to extract information from these three data types: Convolutional Neural Networks for images, BERT for text, and Gradient Boosted Trees for tabular data. We then combine the three predictions in a machine learning ensemble to predict demand.