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

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

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

Working Papers

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. The health care sector is one leading example where machine learning is hoped to deliver efficiency gains. Antibiotic resistance constitutes a major challenge to health care systems largely due to human antibiotic overuse. In this paper, we investigate how machine learning provides new opportunities to design decision rules that reduce antibiotic use. We focus on urinary tract infections in primary care which constitute one of the leading causes for antibiotic use. Laboratory testing can diagnose bacterial infections but with considerable delay such that physicians often prescribe prior to attaining diagnostic certainty. Using Danish administrative and laboratory data, we find that machine learning methods are capable of predicting the presence of bacteria out-of-sample and complement physician prescribing. We optimize policy rules which delegate a share of prescription decisions to physicians and base the remaining decisions on machine learning predictions. The policy shows a potential to reduce antibiotic prescribing by 8.4 percent and overprescribing by 20.9 percent without assigning fewer prescriptions to patients with bacterial infections. We find that human-algorithm cooperation is essential to achieve efficiency gains.

Human decision-making varies due to differences in information and incentives. This constitutes a substantial challenge for evaluating how machine learning predictions can improve decision outcomes. We tackle this challenge in the context of the major health policy problem of improving efficiency in antibiotic prescribing, the leading cause of antibiotic resistance. We incorporate machine learning predictions on large-scale administrative data into a treatment choice model featuring heterogeneity in patients’ disease risk, physician payoffs, and diagnostic skill. We find substantial variation in the skill to diagnose bacterial urinary tract infections and in how physicians trade off the antibiotic resistance externality against curative benefits. Counterfactual policy evaluation shows that providing predictions to physicians increases efficiency but does not reduce antibiotic use. To reduce prescribing, by close to 10 percent, physicians must be incentivized. Our results highlight the importance of the potential misalignment of decision-makers’ and social planners’ objectives in considering prediction policies, while accentuating the complementarity of machine learning and human expertise.

Human antibiotic consumption is considered the main driver of antibiotic resistance. Reducing human antibiotic consumption without compromising health care quality poses one of the most important global health policy challenges. A crucial condition for designing effective policies is to identify who drives antibiotic treatment decisions, physicians or patient demand. We measure the causal effect of physician practice style on antibiotic intake and health outcomes exploiting variation in patient-physician relations due to physician exits in general practice in Denmark. We estimate that physician practice style accounts for 53 to 56 percent of between-clinic differences in all antibiotic consumption, and for 74 to 81 percent in the consumption of second-line antibiotic drugs. We find little evidence that low prescribing styles adversely affect health outcomes measured as preventable hospitalizations due to infections. Our findings suggest that policies to curb antibiotic resistance are most effective when aimed at improving physician decision-making, in particular when they target high prescribers. High prescribing practice styles are positively associated with physician age and negatively with staff size and the availability of diagnostic tools, suggesting that improvements in the quality of diagnostic information is an important path to improved decisions.

Column (in German): DIW Wochenbericht 29-30/2022. Media coverage in Tagesspiegel, Der StandardKronenzeitungHeise, 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.

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

Google Scholar