Research

The Effect of Potential Collusion on Equilibrium Prices (in collaboration with Tom-Reiel Heggedal and Espen R. Moen, PDF)

We investigate how possible collusion, interpreted as a positive prior probability that collusion may occur, influences equilibrium prices in states of the world in which collusion, in fact, does not occur. We explore the mechanism in a theoretical model of consumer search, based on the Stahl (1989)framework, and show that with possible collusion, equilibrium prices are higher even in the absence of collusion. We test the model's predictions in a series of laboratory experiments. The results are qualitatively consistent with our theoretical predictions.

Reputation and Coordination: Experimental Evidence (in collaboration with Tom-Reiel Heggedal, link to paper).

We experimentally study whether reputation facilitates coordination in infinitely repeated games between long-lived and short-lived players. We compare a Control Treatment (CT ) with no history, a History Treatment (HT ) with observable past actions, and a Reputation Treatment (RT ) adding induced commitment types. Theory suggests that reputation facilitates coordination by reducing strategic uncertainty. Results show that efficient coordination rates rise from 6% in CT to 50% in HT and 80% in RT in late supergames. The evidence is consistent with commitment types shifting beliefs and play beyond observable history and has implications for designing online rating systems.

Reputational Cheap Talk with Limited Records (link to paper)

I study a dynamic cheap-talk game where short-lived receivers observe a limited record of a long-lived sender’s honesty. Limiting the record can sustain communication that a full record destroys: when the sender’s payoff is concave in the receiver’s belief, full records reveal her type and communication ends, while limited records sustain it indefinitely, in cycles of rising honesty punctuated by abuse. But above a threshold on the prior of the good state, every stationary equilibrium babbles, whatever the record length. A binary signal of a perfect record supports communication whenever limited records do, and full honesty whenever full records do.

The Role of Beliefs in a Trust Game with Punishment (in collaboration with Michalis Drouvelis, PDF)

We study whether beliefs about others’ willingness to punish increase trust and trustworthiness. Using a novel experimental design, we create environments where people expect punishment to be either likely or unlikely. Our belief manipulation works: participants in the high punishment environment think punishment is twice as likely. Yet trust and trustworthiness remain largely unaffected. We show that this surprising result arises because people react differently to punishment beliefs. For some, punishment strengthens cooperation; for others, it reduces their intrinsic willingness to act trustworthily. These opposing responses cancel out, leaving only small overall effects.

   
Age and Social Trust: Within-Person Evidence from Germany and Switzerland (link to paper)

In cross-sectional surveys generalized trust rises to a peak near midlife. However, age and birth cohort are collinear in a cross-section, so this profile may reflect differences between generations rather than ageing. I separate the two with a within-person first-difference design that removes additively separable cohort effects by construction, using the German Socio-Economic Panel (2003--2023) and the Swiss Household Panel (2002--2024). In both panels the detrended within-person change in trust falls significantly with age: the life-cycle profile is genuinely concave, decelerating as people get older, and the concavity is not a generational artifact. 

Optimal Tax Depreciation Rules under Limited Classes (in collaboration with Jonas Hamang, link to paper)

Tax depreciation systems apply a limited number of statutory rules to a heterogeneous set of capital assets. A common approach is to match economic depreciation as closely as possible for each asset. We show that this principle is generally not aligned with efficiency maximization under the given constraint. By assigning statutory depreciation rates, the tax depreciation system affects relative user costs and thereby the economy-wide allocation of capital across assets. Efficiency therefore depends not only on asset-specific deviations from true user costs, but also on interactions among tax-induced wedges and the economic importance of the affected assets. As a result, optimal policy departs from asset-by-asset accuracy. Instead, it prioritizes coordinating user-cost wedges among interacting assets and allocating more precise treatment to economically important inputs. This perspective yields several implications. Assets with identical economic depreciation may optimally receive different tax treatment, and even when a first-best depreciation rate is available for an asset, it need not be optimal to assign it to that class. Conversely, assets with different depreciation rates may be optimally grouped together. Moreover, improving the accuracy of a given asset in isolation can increase overall inefficiency if it worsens the pattern of wedges across interacting inputs. In fact, there exist cases in which a policy that is strictly less accurate for every asset generates lower efficiency loss by better coordinating wedges across the production system.

Ratings, Externalities and Competition (in collaboration with Thomas de Haan at the University of Bergen).

Information Bundling and Product Differentiation (in collaboration with Thomas de Haan at the University of Bergen).

Product Information and Competition Softening (in collaboration with Tom-Reiel Heggedal).