Research

Publications

Column (in German): DIW Wochenbericht 13-14/2021.

Column (in German): ZEW News 9/2009.

Column (in German): DIW Wochenbericht 14/2014.

Working Papers

We analyze how machine learning predictions may improve antibiotic prescribing in the context of the global health policy challenge of increasing antibiotic resistance. Estimating a binary antibiotic treatment choice model, we find variation in the skill to diagnose bacterial urinary tract infections and in how general practitioners trade off the expected cost of resistance against antibiotic curative benefits. In counterfactual analyses we find that providing machine learning predictions of bacterial infections to physicians increases prescribing efficiency. However, to achieve the policy objective of reducing antibiotic prescribing, physicians must also be incentivized. Our results highlight the potential misalignment of social and heterogeneous individual objectives in utilizing machine learning for prediction policy problems.

Columns: DIW Weekly Report 19/2019 (in English), DIW Wochenbericht 19/2019 and Oekonomenstimme (in German).

Inefficient human decisions are driven by biases and limited information. Health care is one leading example where machine learning is hoped to deliver efficiency gains. Antibiotic resistance constitutes a major challenge to health care systems due to human antibiotic overuse. We investigate how machine learning provides new opportunities to reduce antibiotic use, with the help of physicians. We focus on urinary tract infections in primary care, a leading cause for antibiotic use, where physicians often prescribe prior to attaining diagnostic certainty. Symptom assessment and rapid testing provide diagnostic information with limited accuracy, while laboratory testing can diagnose bacterial infections with considerable delay. Linking Danish administrative and laboratory data, we optimize policy rules which base initial prescription decisions on machine learning predictions and delegate decisions to physicians where these benefit most from private information at the point-of-care. We find human-algorithm complementarity is essential to achieve efficiency gains with a potential reduction in antibiotic prescribing by 8.1 percent and in overprescribing by 20.3 percent.

In the fight against antibiotic resistance, reducing antibiotic consumption while preserving healthcare quality presents a critical health policy challenge. We investigate the role of practice styles in patients’ antibiotic intake using exogenous variation in patient-physician assignment. Practice style heterogeneity explains 49% of the differences in overall antibiotic use and 83% of the differences in second-line antibiotic use between primary care providers. We find no evidence that high prescribing is linked to better treatment quality or fewer adverse health outcomes. Policies improving physician decision-making, particularly among high-prescribers, may be effective in reducing antibiotic consumption while sustaining healthcare quality.

Column (in German): DIW Wochenbericht 29-30/2022. Media coverage in Tagesspiegel, Der Standard, Kronenzeitung, Heise, PC-Welt, BR, and Computerwoche podcast (in German).

We quantify Facebook’s ability to build shadow profiles by tracking individuals across the web, irrespective of whether they are users of the social network. For a representative sample of US Internet users, we find that Facebook is able to track about 40% of the browsing time of both users and non-users of Facebook, including on privacy-sensitive domains and across user demographics. We show that the collected browsing data can produce accurate predictions of personal information that is valuable for advertisers, such as age or gender. Because Facebook users reveal their demographic information to the platform, and because the browsing behavior of users and non-users of Facebook overlaps, users impose a data externality on non-users by allowing Facebook to infer their personal information.

Column (in German): DIW Wochenbericht 27/2023. Media coverage in taz, Table.Media, finanzen.net, and ntv (in German).

The tracking of online user behavior has been essential for the construction of consumer profiles to help platforms monetize their services by selling targeted advertisements. We analyze web browsing data to show how prediction quality of consumer profiles varies across platforms depending on the size and scope of user data available to them. We find decreasing returns to the number of users observed and the number of websites tracked. Combining web browsing data with demographic information, two heterogeneous sources of user information which are available to some online platforms, provides a sizable increase in prediction quality. For Google, we find more slowly decreasing returns compared to other trackers with an increase in both the number of users and websites tracked. Finally, we document that proposed data combination policies may level the playing field with respect to the returns to data.

We analyze the effects of a hypothetical payment card fee regulation on bank profits, consumer welfare, and merchant welfare. We model consumers’ and merchants’ bank choices for debit card services, cardholders’ demand for card usage (conditional on bank choice), and how banks account for these in setting card fees to their customers. To estimate the model, we use bank-level data and survey data from the Norwegian debit card scheme, BankAxept. We conduct counterfactual exercises to analyze the effects of interchange fee regulations in the debit card scheme.

Former title: Considerations on partially identified regression models, Centre for European Economic Research Discussion Paper 2012-024, BETA (Bureau d’Economie Théorique et Appliquée) Working Paper No. 2012-07, Online Appendix: Download

Work in progress

My research repositories and social media profiles

SSRN
Google Scholar
arXiv
Bluesky
LinkedIn
Mastodon
Twitter