Mitigating manipulation in committees: Just let them talk!
[Working Paper on SSRN]
[Preregistration]
[Replication Materials]
Keywords: Group Decision Making, Committees, Manipulation, Face-to-face Communication, Delphi Technique
Many decisions rest on the collective judgment of small groups like committees or teams. However,
some group members may have hidden agendas and manipulate this judgment to influence decisions in
their favor. Utilizing an incentivized experiment, I compare how objective accuracy and perceived
trustworthiness of judgments from groups using two out-of-the-shelf interaction formats - the Delphi technique and face-to-face meetings -
are affected by manipulation.
Without manipulation, Delphi is more accurate. With manipulation, face-to-face is more accurate.
Only in situations with manipulation the more accurate interaction format is perceived as more
trustworthy. With manipulation, sharing of (truthful) information decreases in Delphi but not in
face-to-face interaction.
Debt Aversion: Theory and Measurement
[Working Paper]
[Replication Materials]
Keywords: Debt Aversion, Intertemporal Choice, Risk and Time Preferences
We propose a model of debt aversion and conduct an experiment involving real debt and saving contracts to jointly elicit debt aversion with preferences over time, risk, and intertemporal losses. We find that 89% of participants are debt averse, which strongly influences their choices. We estimate the “borrowing premium” – the compensation an verage person would require to accept getting into debt – to be around 16% of the prin- cipal. Building on these findings, we validate a survey module to measure debt aversion in real-world contexts where complex laboratory experiments are infeasible. Applying the survey module to a large, representative sample of German households, we find that debt aversion is both statistically and economically significantly associated with actual debt-taking and financial health.
#ManyDaughters: Multi-Analyst Project Exploring the Effect of Daughters on Preferences and Behaviors
[Project Website]
Pre-analysis plans (PAPs) are challenging to write for complex observational data because researchers cannot anticipate the data structure before committing to a specification. Granting limited data access might improve specifications, but it might also invite tailoring them to the observed sample. We provide the first large-scale experimental evaluation of this tradeoff. In a many-analyst study, 127 research teams wrote a PAP and analysis code to test four hypotheses on the effect of having daughters on attitudes, preferences, and behavior, a much-studied but inconclusive topic. The underlying data come from the German Socio-Economic Panel (SOEP). Teams were randomized to write their PAP with or without access to a 5% subsample, and all code was run on the held-out 95%. Access lowered self-reported effort but neither measurably improved PAP quality (by peer ratings) nor reduced the dispersion of estimates across teams; nor did it increase data-driven specification. Thus, the feared downside and the hoped-for upside of partial access both failed to materialize. None of the four daughters hypotheses is supported in pooled tests, yet for each hypothesis individual teams obtain significant results in both directions, a direct illustration of analytical heterogeneity.
ManyLabsDACH: A Multi-Lab Experimental Economics Study across Germany, Austria, and Switzerland
[Project Website]
This multi-lab study will unite economics labs across Germany, Austria, and Switzerlands. We aim to (i) test a mechanism for selecting experimental designs that can be included in many-labs studies. (ii) To conduct high-powered meta-analytic tests of the hypotheses (treatment effects) evaluated in the selected experimental design. (iii) To estimate the population heterogeneity in these treatment effects across the participating laboratories (Holzmeister et al., 2024). (iv) To examine whether treatment effects estimated in the many-lab data collection differ from treatment effects estimated in an online data collection. (v) To compare the student population in the DACH region (Germany, Austria, and Switzerland) with a representative sample of the German population, in collaboration with SOEP, one of the largest and most established household panel studies in the world.
Expanding Moral Wiggle Room: The Role of Moral Context
We test the generalizability of the classic moral wiggle room paper in a large-scale preregistered online data collection (n=12,000). Replicating the original design online, leads to remarkably similar results with an effect size that is 84% of the original. Following the original authors, the moral wiggle room effect can be due to a “self-deceptive” and/or “other deceptive” motive. Holding the other-deceptive motive constant across conditions, decreases the effect size to 66% of the original effect size. Increasing the vulnerability of the recipients reduces the effect by about a third to 56% of the original in line with our hypothesis of the importance of moral context; also controlling for the other-deceptive motive further decreases the effect to 40% of the original. Increased vulnerability also modestly increases the revealing rate. We conclude that the effect as such is highly generalizable, primarily motivated by self-deception, and importantly moderated by the vulnerability of the recipients.
Examining the generalizability of research findings from archival data
[Proceedings of the National Academy of Sciences, 2022]
Keywords: Generalizability, Archival Data, Reproducibility, Strategic Management, Forecasting