We design and conduct a real-effort experiment to jointly estimate present bias and sophistication across effort and monetary domains. Unlike prior work (e.g.,Augenblick and Rabin, 2019; Fedyk, 2024), we do not assume that these parameters are identical across domains. We explain and empirically demonstrate how assuming identical sophistication across money and effort domains can bias the estimates of key parameters. In our online experiment, participants chose to (predicted to) complete 14% (10%) fewer tasks on the same day than on a future day, leading to an estimated present bias ($\beta_e$) over effort of 0.70–0.79, and an estimated sophistication ($\widehat{\beta}_e$) of 0.80–0.88. For money, aggregate present bias ($\beta_m$) is near zero, but there is substantial heterogeneity, with roughly equal numbers of participants exhibiting present bias and future bias. At the individual level, roughly three quarters of all participants correctly anticipate the direction of their bias in both domains, even if not its full magnitude.

We build an experiment to uncover the bandit-like nature of consumer behavior usually masked by the product and time aggregation of consumption data. Subjects make repeated choices between musical styles (either all familiar or unfamiliar), and post-choice satisfaction is observed. We estimate Bayesian bandit models of learning taste by consuming with satiation. Our best model features decreasing random exploration, with openness being associated with higher exploration. Early exploration is more intensive in the unfamiliar treatment and persists throughout the experiment in both treatments. Overall, subjects make choices that deviate from their best prediction 61.5% of the time in the unfamiliar treatment versus 44.7% in the familiar treatment. Our model offers a rational interpretation of random utility in discrete consumer choices which does not rest on perception and/or decision errors.

We experimentally investigate preferences for clumping-versus-separating information in the gain and the loss domains, and also preferences for timing. Our design is motivated by the idea that information preferences may depend on reference points. Subjects participate in two monetary lotteries and choose how to receive the outcome information. For half of the subjects, the lotteries are framed as two gain lotteries; for the other half, as two loss lotteries. Based on Thaler (1985) one can expect that people want to learn the outcomes of the gain lotteries separately and the outcomes of the loss lotteries clumped together (cf. hedonic editing hypothesis). On the other hand, a different reference dependent model by Koszegi and Rabin (2009) relies on expectations-based reference points, and predicts that subjects should prefer clumped information irrespective of the frame.

The results of our experiment show a preference for separating information about gains, and no preference for clumping or separating information about losses. Regarding timing, we find a weak overall preference for receiving information sooner. These findings provide new insights into information preferences. We conclude by discussing policy implications, as well as our additional contributions to related literature.

This study investigates how in-game virtual currencies influence consumer spending behaviour. In an incentivized randomized controlled trial with 753 UK participants, we estimate the willingness-to-pay (WTP) for loot boxes represented as risky and ambiguous lotteries. We test the effects of virtual currencies, money illusion, and cognitive-demanding (non-intuitive) exchange rates, three practices commonly used to sell loot boxes. Our results show that WTP is significantly higher when transactions are conducted in virtual currencies with a 1:1 exchange rate compared to British pounds, contributing to the debate on the neutrality of virtual currencies and experimental currency units. Moreover, participants in our study are affected by money illusion distortions, regardless of the exchange rate’s intuitiveness. However, we find no evidence that non-intuitive exchange rates affect participants’ WTP compared to intuitive exchange rates of comparable size. These findings highlight the behavioural distortions induced by common monetization practices in games and provide empirical support for the European Commission’s call for stricter regulation of virtual currencies in digital environments.

We experimentally investigate whether lying is more likely when the addressees are individuals or groups of individuals, and in the latter case how the probability of lying depends on the group size and the magnitude of the negative externality inflicted by the lie. We employ an observed cheating game, where an individual can reveal or misreport a privately observed number. A misreport can be monetarily beneficial for the liar but imposes a monetary loss on addressees. The privately observed number is also known to the experimenter, who, therefore, can study both whether there is truth or misreporting and, in the latter case, the extent of misreporting. Treatments vary the loss for group members compared to the loss for an individual addressee. We find that groups are never lied to more than individual addressees. When considering how much to deviate from the truth, liars are sensitive to a decreasing loss at the individual level but do not care for an increasing loss at the group level. Groups of different sizes are treated similarly. Social image concerns may explain the results.

The Multiple Price List (MPL) and Switching Multiple Price List (sMPL) provide a useful framework for estimating preference parameters, most usually risk aversion, from a sample of experimental subjects or survey respondents. In this paper, we consider designs in which more than one sMPL is presented to each subject, allowing more than one preference parameter to be estimated simultanously, and we propose a consistent estimator in this setting – the Multivariate Heterogeneous Preference (MHP) estimator. Focusing on the bivariate case of two sMPLs and two preference parameters, we demonstrate that non-standard econometric techniques, namely Monte Carlo integration with importance sampling, are required to implement the MHP estimator. Via a Monte Carlo exercise, we show that our estimator has good finite-sample properties. Finally, we apply the MHP estimator to a real data set and compare the estimates to those obtained using an inconsistent estimator applied in previous studies.

In various organizational settings, a team member is given the authority to make an investment decision that influences the value of the jointly produced surplus. We experimentally investigate the effect of asymmetric status, investment decisions, and the outcome of these decisions on bargaining behavior and outcomes. Agents’ initial contributions to the surplus are determined by their relative performances in a real-effort task. Three treatments vary in how the final surplus value is determined. We observe that when low-contributors take a risk, they are punished (rewarded) for failure (success), whereas high-contributors receive a fixed share independent of the outcome. Analysis of bargaining process variables, subjects’ communication during bargaining, and third parties’ normative judgments provides further insights into the possible mechanism behind this observation.

We investigate whether time pressure exacerbates or mitigates bubbles in laboratory experiments. We find that under high time pressure price volatility is lower and market prices are closer to their fundamental value. This is due to participants using simpler adaptive forecasting strategies, instead of the self-reinforcing extrapolative expectations that they use under low time pressure, and which are conducive to the emergence of bubbles. In addition, by substantially increasing the number of decision periods in our experiment, we find that in the long run prices tend to converge to their fundamental value, also in the absence of time pressure.

Large Language Models (LLMs) have the potential to profoundly transform and enrich experimental economic research. We propose a new software framework, “alter_ego”, which makes it easy to design experiments between LLMs and to integrate LLMs into oTree-based experiments with human subjects. Our toolkit is freely available at github.com/mrpg/ego. To illustrate, we run differently framed prisoner’s dilemmas with interacting machines as well as with human-machine interaction. Framing effects in machine-only treatments are strong and similar to those expected from previous human-only experiments, yet less pronounced and qualitatively different if machines interact with human participants.

Theoretical insights dominate the literature examining the incentive compatibility of payment mechanisms. Despite their elegance, theoretical insights are rarely empirically validated. We fill this gap by empirically exploring the effects of frequently used payment mechanisms using a collective sample of over 3000 participants across two experiments. In Experiment 1, we obtained offer prices to sell a card, systematically varying between-subjects the way subjects received payments over repeated rounds, by either paying for all decisions (and various modifications) or just one, as well as making the payments certain, probabilistic, or purely hypothetical. While we find that the magnitude of the induced value and the range of the prices used to draw a random price significantly affect misbidding behavior, neither the payment mechanism nor the certainty of payment affected misbidding. In Experiment 2, we replaced the BDM mechanism with a Second Price-Auction and found similar results, albeit lower rates of misbidding behavior. Overall, our empirical exercise shows that theoretically relevant elements do not produce empirical differences, while design choices that are theoretically irrelevant produce empirical differences. As such, payment mechanism design considerations should carefully consider the choice architecture in addition to incentive compatibility.

Agents frequently engage with multiple principals simultaneously – for example, when borrowing from several banks or peers. In such settings, principals typically possess less information about the agent’s ability or intentions (e.g., to repay a loan) and must rely on trust. This paper presents experimental evidence from trust games framed in a credit market context to examine the role of reciprocity in interactions involving multiple principals (lenders) and a single agent (borrower). Agents were asked to decide whether to act trustworthily and repay, or to default and act selfishly, after receiving the same credit amount from either one or multiple principals. The results show that reciprocity declines when the number of trusting principals increases. A key mechanism appears to be the reduced marginal harm that an agent’s default imposes on each individual principal. Additionally, agents seem less sensitive to the negative consequences of their actions when multiple principals are affected. These findings suggest that interactions involving multiple principals are behaviorally riskier than bilateral ones. The results have implications for the design of incentive structures in multi-principal-agent environments, such as crowdlending platforms.

Across three studies involving more than 5,000 participants, we provide a comprehensive analysis of the effects of incentivizing responses in the Krupka-Weber norm elicitation task. We consider both the potential benefits of incentivization (higher response quality and mitigation of response biases) and its possible drawbacks (distortion of responses in the direction of norm-unrelated focal points and materialistic values). We find no evidence of undesirable effects of incentives. While we report only modest improvements in response quality, we also show that incentives effectively mitigate response biases that arise when participants’ self-serving motivations conflict with accurate responding.

There is existing evidence that many individuals have preferences regarding selection of numbers in lottery games. Lottery data indicate that the percentage of players who choose their numbers, instead of having numbers randomly assigned, varies widely by lottery game. Differences in number selection mechanisms between games and an expected return maximization motive only present for parimutuel games are both reasons that can explain the variation. Differences in the payoff distributions between lottery games could also be contributing to the observed variation, a novel proposition. An experiment is designed to control for differences in number selection mechanisms and remove the expected return maximization motive, to test for the presence of distribution-dependent number preferences. Results indicate that 40% to 50% of subjects may display such preferences. It is therefore possible that distribution-dependent number preferences contribute to the empirical variation in number selection percentages in lottery games.

We experimentally study how individuals strategically disclose multidimensional information to a Naive Bayes algorithm trained to guess their characteristics. Subjects’ objective is to minimize the algorithm’s accuracy in guessing a target characteristic. We vary what participants know about the algorithm’s functioning and how obvious are the correlations between the target and other characteristics. Optimal disclosure strategies rely on subjects identifying whether the combination of their characteristics is common or not. Information about the algorithm functioning makes subjects identify correlations they otherwise do not see but also overthink. Overall, this information decreases the frequency of optimal disclosure strategies.

We design an experiment to study the effect of asymmetry in the context of group lending with joint liability. The performance of group loan contracts crucially hinges on borrowers engaging in peer monitoring and the common practice is to offer participants of a group loan symmetric contract terms. Our experiment shows that asymmetric contracts, in which monitoring is a dominant strategy for one borrower, increase the monitoring rate, and thus the repayment rate, without leaving borrowers substantially worse off. In addition, asymmetric contracting also raises expected profits of the lending institution. Overall, our experiment reveals that asymmetric group loan contracts are worth considering as part of a policy to maintain both financial stability and higher lender profits.

This paper considers a stationary model of inventory management in a rich setting in which unsold units carry over, in contrast with the full depreciation of unsold units that is implemented in laboratory studies of the news vendor problem. The model permits an array of costs associated with restocking, understocking, depreciation, financing, and holding inventories. The extra dimensions make it possible to hold the optimal inventory constant, while adjusting parameters that change the frequency of stockouts and the risks associated with storage and depreciation. This framework facilitates an investigation of factors that influence the nature and severity of behavioral biases observed in simpler news vendor settings. Optimal inventory decisions are derived and tested with a laboratory experiment. We consider four main questions in the inventory literature: the “pull-to-center” effect, the “recency” effect, the effect of increased up-front costs, and the effect of risk aversion.

This paper analyzes individual behavior in multi-armed bandit problems. We use a between-subjects experiment to implement four bandit problems that vary based on the horizon (indefinite or finite) and the number of bandit arms (two or three). We analyze commonly suggested strategies and find that an overwhelming majority of subjects are best fit by either a probabilistic “win-stay lose-shift” strategy or reinforcement learning. However, we show that subjects violate the assumptions of the probabilistic win-stay lose-shift strategy as switching depends on more than the previous outcome. We design two new “biased” strategies that adapt either reinforcement learning or myopic quantal response by incorporating a bias toward choosing the previous arm. We find that a majority of subjects are best fit by one of these two strategies but also find heterogeneity in subjects’ best-fitting strategies. We show that the performance of our biased strategies is robust to adapting popular strategies from other literatures (e.g., EWA and I-SAW) and using different selection criteria. Additionally, we find that our biased strategies best fit a majority of subjects when analyzing a new treatment with a new set of subjects.

We design an experiment to study the implications of introducing position uncertainty in a social dilemma where eight players decide to contribute to a public good sequentially. Contributions are significantly higher when players make sequential decisions to contribute or not, are uncertain about their position in the sequence, and observe a sample of their predecessors’ choices compared to the simultaneous-move game. Yet, contribution rates remain invariant to the number of agents sampled. Consequently, contributions don’t unravel even with position certainty, and there is no incremental benefit of introducing position uncertainty, contrary to the theoretical prediction. Furthermore, controlling for the sum of contributions observed, individuals contribute less the later in the sequence they are.

In recent years, there has been an increasing interest in motivated memory as a psychological determinant of economic outcomes. According to motivated memory, people tend to better recall pleasant information because it serves their psychological needs. Another phenomenon, however, predicts the same pattern: According to mood congruence, pleasant information is easier to recall for individuals in nonnegative mood, regardless of any psychological needs. Since most people tend to have some need for self-esteem and to experience more positive than negative feelings during the day, the two phenomena predict similar outcomes in most ordinary situations, but not all. To test the predictive power of these two phenomena, we collect data from a laboratory experiment and from a nationally representative survey. We study how individuals in a temporarily induced negative mood (via a video clip) or those who report a low baseline mood (relative to the population) recall negative feedback. Our results meet the predictions of motivated memory: Individuals better recall positive than negative feedback, even when they are in negative mood. Motivated memory is not just a matter of mood.

Although compulsory insurance mitigates the negative externalities caused by uninsured individuals, it raises the issue of insurance crowding out prevention. However, at the theoretical level, compulsory insurance and self-insurance (preventive investments dedicated to loss reduction) are know to be substitutes for risk averters but complements for risk lovers. This paper aims to empirically test these opposite predictions through a laboratory experiment using a model-based design. Our experimental results confirm the theoretical predictions: compulsory insurance and self-insurance are complements for risk lovers and substitutes for risk averters. This study strongly supports public policies advocating mandatory insurance implementation as they enhance risk lovers’ self-insurance investments. Therefore, a risk management scheme combining voluntary top-up and compulsory partial insurance guarantees an optimal risk allocation for risk-averters and increases the investments in self-insurance for risk-lovers.