Soft Floor Auctions: Harnessing Regret to Improve Efficiency and Revenue

with Dirk Bergemann, Kevin Breuer, Peter Cramton, Jack Hirsch and Axel Ockenfels; conditionally accepted at American Economic Review
Working Paper

A soft-floor auction asks bidders to accept an opening price to participate in an ascending auction. If no bidder accepts, lower bids are considered using first-price rules. Soft floors are common despite being irrelevant with standard assumptions. When bidders regret losing, soft-floor auctions are more efficient and profitable than standard optimal auctions. Revenue increases as bidders are inclined to accept the opening price to compete in a regret-free ascending auction. Efficiency is improved since having a soft floor allows for a lower hard reserve price, reducing the frequency of no sale. Theory and experiment confirm these motivations from practice.

How the Design of Ranking Systems and Ability Affect Physician Effort

with Katharina Huesmann, Christian Waibel, and Daniel Wiesen; Management Science vol. 72, no. 2 (2025): 1198-1213.
Paper Policy Paper

While relative performance feedback in the form of rankings appears to be effective in improving healthcare outcomes, it may have either motivating or demotivating effects for individual physicians. Potential factors influencing such effects include a physician’s level of ability and the design of the ranking system itself; however, there is limited understanding of these factors. Using a controlled lab-in-the-field experiment with practicing and future physicians as subjects (N = 352), we systematically analyze effort within small teams under different ranking systems. Exogenously varying the number and position of the thresholds defining the ranking system, we observe that the addition of a threshold to create a new rank is motivating-i.e., increases effort-only among individuals capable of exceeding that threshold; the effort of other individuals may remain unchanged or even decrease. In particular, a highly granular ranking system with ranks spanning the entire range of possible outcomes maximizes overall physician effort: high thresholds serve to motivate high-ability individuals, while moderate and low thresholds provide opportunities for improvement to lower-ability individuals who cannot reach the high thresholds. Our results suggest that, to motivate their teams effectively, clinical leaders should provide rank feedback using a system under which physicians of all ability types can improve their rank through increased effort.

Who Should Know What and Why? Curated Supply of Political Information

Job Market Paper (Experiment 3 in the field)

The information environment can be distorted not only by the spread of falsehoods but also by the selective supply of truths. I study how citizens curate which accurate information others receive based on political alignment. In a pre-registered experiment with 1,200 U.S. adults, senders decide whether to share verified immigration statistics with receivers who hold inaccurate beliefs. Senders provide accurate information at high rates (above 80%), but are approximately 8 percentage points more likely to share facts aligned with their political position than facts that contradict it, even when receivers do not take a downstream political action in the experiment. This selectivity approximately doubles when receivers make a political decision, with the amplification operating primarily through withholding of unaligned information. The theoretical framework organizes the interpretation: baseline selectivity reflects a preference for belief congruence, while its amplification under political action is consistent with an instrumental motive. This curated truth—the selective provision of accurate information that favors the sender's political position—distorts the information environment through a channel not targeted by current policy interventions focused on false content.

Can Second Opinions Improve Health Outcomes? Evidence from a Structured Program for Elective Surgery

with Felix Mindl and Daniel Wiesen
Draft available upon request

We analyze a structured second-opinion program for elective surgeries in Germany using administrative claims data (2013–2019). The program institutionally separates diagnosis from treatment: physicians providing second opinions give independent assessments but are banned from performing the treatment themselves. In orthopedics and neurosurgery, first and second-opinion physicians disagree in 56% of cases. The disagreement is strongly asymmetric: second physicians contradict 63% of surgery recommendations but only 14% of recommendations against surgery—consistent with physician-induced demand. Seventy percent of patients comply with second-opinion recommendations. Using propensity score matching and patient fixed-effects event-study designs, we estimate that compliance reduces monthly health expenditures by about €431–478 (approximately baseline mean), the number of outpatient visits by 0.61, and hospital days by 0.06 over 18 months, with no adverse effects on sick leave. The effects concentrate among patients advised against surgery, who save €710–760 monthly without health deterioration. For these patients, surgeries remain avoided throughout the 18 months following consultation, with symptoms managed instead through conservative therapies. Taken together, these findings provide evidence that structured second-opinion programs effectively curb overtreatment by separating diagnosis from treatment, translating theoretical predictions into realized health and spending outcomes.

Narrative Interventions and Stem Cell Donor Availability

with Michael Haylock, Patrick Kampkötter, Axel Ockenfels, Till Stange and Daniel Wiesen

Abstract

Many people register as stem-cell donors intending to help, yet 30–50% later become unavailable at confirmatory typing (CT). We ask whether behavioral-economics informed narratives can close the gap between intention and follow-through. In a pre-registered field experiment with a globally-operating registry, over 1.8 million registered donors are randomly assigned to one of three narrative videos, emphasizing reciprocity, community, or personal relevance and low costs or a control. We measure stated willingness to donate and link donors to administrative CT availability as matches arrive. Compared to a 94% baseline, the community framing lowers willingness relative to the neutral video by 0.8–1.0 pp, while personal relevance and low costs raise it by 0.3 pp. Early CT outcomes are consistent with these findings. Heterogeneity by gender and behavioral characteristics suggests persuasion is audience-dependent, highlighting the risks of one-size-fits-all campaigns.

Data collection in progress (AEA RCT Registry: AEARCTR-0017571)

Choice Architecture, Sorting, and Availability of Stem Cell Donors

with Patrick Kampkötter, Axel Ockenfels, Till Stange and Daniel Wiesen

Abstract

Stem-cell donor registries seek to recruit large numbers of donors, but also donors who remain available when matched with a patient, creating a potential quantity–quality trade-off. Non-availability at the confirmatory-typing (CT) stage reaches up to 50% in Western countries, undermining search efficiency and potentially delaying life-saving transplants. We report preliminary results from a pre-registered cluster-randomized field experiment with DKMS across 206 school drives in Germany, covering about 48,000 students. At drives in treatment schools, students were informed at sign-up about an explicit post-registration easy-exit option; drives in control schools followed the standard protocol. Contrary to concerns that an easy exit option may weaken commitment, the intervention increased registration rates from 31% to 34%. Only about 1% of treated registrants subsequently exercised the easy-exit. These patterns suggest that an easy-exit can reduce sign-up friction without inducing substantial short-run dropout. The experiment is ongoing, and administrative records will be linked to future CT requests to assess long-run donor availability. Early data show no evidence of reduced availability in the treatment group. The findings suggest that optimizing choice architecture may increase prosocial participation without compromising downstream commitment quality.

Data collection in progress (AEA RCT Registry: AEARCTR-0017570)

Information Acquisition on Digital Platforms: The Effect on Beliefs and Behavior

with Lara Marie Berger

Abstract

Online search is shifting fast: Google desktop searches per user fell about 20% YoY in 2025, ChatGPT use has roughly doubled since 2023, and about 49% of U.S. consumers now use TikTok as a search engine. Yet causal evidence on how the chosen platform shapes what users come to believe is missing. We run a pre-registered between-subjects experiment randomly assigning participants to research three topics (inflation, immigration, personal investments) on Google, ChatGPT, TikTok, or no platform; a browser extension records search behavior, incentive-compatible elicitations capture beliefs, expectations, attitudes, and narratives, and an optional one-week follow-up with monetary encouragement to keep using the assigned platform measures durability. Our pilot (N=59) suggests large platform effects on accuracy and narrative dispersion.

Pilot data collected

I study how institutions shape individual behavior, and how behavioral insights can inform the design of better institutions. I use experiments to test behaviorally motivated theories, drawing on classical laboratory experiments, lab-in-the-field experiments with professionals, and large-scale field experiments. I also apply causal inference methods to large-scale administrative data. I focus on three domains, each transformed by digitalization: information markets, health care markets, and mechanism design.