We study risk-taking behavior in tournaments where the optimal strategy is to take no risk. By keeping the optimal strategy constant, while varying the competitiveness in the tournaments, we are able to investigate the relationship between competitiveness and excessive risk-taking. In the most competitive tournament, less than 10% of the subjects played the optimal strategy in the first rounds. The majority playing dominated strategies increased their risk-taking during game of play. When we removed feedback about winner’s decisions each round, and when we reduced the number of contestants in the tournaments, subjects significantly reduced their risk-taking.

Myopic loss aversion (MLA) has been proposed as an explanation for the equity premium puzzle, and experiments indicate that investors exhibit behavior consistent with MLA. But a caveat is that a large bulk of financial assets is managed by investment managers whose objectives may differ substantially from those of private investors. Most importantly they manage their clients' money, not their own. In this paper we test experimentally how individuals take risk with other people's (“clients”) money. We find that subjects behave consistently with MLA over their clients' money and take less risk with their clients' money than with their own.

The article analyzes conditions for implementing incentive schemes based on, respectively, joint, relative, and independent performance in a relational contract between a principal and a team of two interacting agents. A main result is that the optimal incentive regime depends crucially on the productivity of the agents. This occurs because agents’ productivities affect the principal’s temptation to renege on the relational contract. The analysis suggests that we will see a higher frequency of relative performance evaluation—and schemes that lie close to independent performance evaluation—as we move from low‐productive to high‐productive environments.

Reward systems based on balanced scorecards often connect pay to an index, that is, a weighted sum of multiple performance measures. We show that such an index contract may indeed be optimal if performance measures are nonverifiable so that the contracting parties must rely on self‐enforcement. Under commonly invoked assumptions (including normally distributed measurements), we show that the weights in the index reflect a tradeoff between distortion and precision for the measures. The efficiency of the contract improves with higher precision of the index measure, because this strengthens incentives, and correlations between measurements may for this reason be beneficial. There is a caveat, however, because the index contract is not necessarily optimal for very precise measurements, although it is shown to be asymptotically optimal. We also consider hybrid measurements, and show that the principal may want to include verifiable performance measures in the relational index contract in order to improve incentives, and that this has noteworthy implications for the formal contract.

Principal-agent models usually invoke the strong assumption that the parties know for sure ex ante whether a variable is verifiable or not. This paper assumes that only the probability of verification is known, and that this probability is endogenously determined. We analyze a principal-agent relationship where the verifiability of the agent's output is determined by the principal's investment in drafting an explicit contract. The model is well suited for analyzing the relationship between explicit contracting, legal courts, trust, and relational contracting. In particular, we show how trust—established through repeated interaction—and legal courts may induce contractual incompleteness.

Millions of investors place their trust in financial advisors who may have incentives to give them bad advice. This may indicate that advisors behave more fairly than economic theory predicts. In this paper, we present results from a large-scale experiment studying advice-giving under conflicting interests. We use a binary dictator game as a baseline and transform it into a situation where the dictator gives advice that may or may not be followed. Our results show that people are averse to giving bad advice. When subjects are given the role of advisor, they behave less selfishly, even when the economic incentives and considerations remain the same as in the baseline dictator game.