## Final Report Summary - BESTDECISION (Behavioural Economics and Strategic Decision Making: Theory, Empirics, and Experiments)

One major line of study reconsiders one of the most frequently used models in all of economics, consumer theory, in which people balance prices and budgets to satisfy preferences for consumption that reflect trade-offs between different goods. An important application is the theory of labor supply, in which workers balance the benefits of leisure against those of earnings. This line seeks to increase the realism of consumer theory’s predictions by generalizing preferences in a direction suggested by Kahneman and Tversky’s ‘prospect theory’. Previous tests of the standard theory of labor supply using cabdrivers—who choose their own hours, as the theory assumes—found it seriously wanting: Drivers who have an unusually profitable morning, signalling a higher ‘wage’, tend to quit earlier, the opposite of what the theory predicts. Either drivers are irrational, or their preferences deviate from traditional assumptions. previous authors have suggested explanations in the spirit of prospect theory, in which drivers care about changes in income as well as levels, and are loss-averse, leading to choices that cluster around daily income targets. More recent empirical work has proposed and tested more formal models, confirming that a generalization of labor supply theory based on prospect theory can give a coherent account of drivers’ choices. But such work has relied on strong, untested ‘structural’ assumptions about the shapes of the functions used to represent drivers’ preferences. The work in this line, with team members Ian Crawford and Juanjuan Meng, establishes generalizations of classic results from ‘nonparametric’ (without functional-form assumptions) consumer theory that encompass 'prospect theory' preferences. The work yields elegant and practical methods to determine whether the models’ power to explain drivers’ anomalous choices comes from the general notions of prospect theory and loss aversion, or from the functional form assumptions, and whether the generalized models explain the data sufficiently better to justify their extra flexibility.

A second line of study concerns the design of optimal bargaining institutions. A classic 1983 paper proposed a novel solution to this question, in which one trader owns an indivisible object and would be willing to sell it for enough money, and another would be willing to buy it if the price is right, but neither knows the other’s value, hence neither knows whether it would be mutually beneficial to trade. The ideal outcome would be for them to trade if and only if the buyer's value is higher than the seller's, at a price that fairly shares the gains. No familiar bargaining institution always achieves this ideal outcome. The classic analysis stepped outside the box by asking whether there is any feasible institution that achieves it, given the incentives that it creates, thus going beyond considering institutions as given and immutable, instead thinking of them as something that can be designed. They used traditional game-theoretic methods, assuming that traders are rational and will make decisions that are in Nash equilibrium, in the sense that my decision is best for me given yours, and vice versa. Assuming that traders would respond to any chosen institution by playing their Nash equilibrium strategies, the classic analysis asked whether there is any institution that ensures the ideal outcome, given the incentives it creates. The answer is ‘No’. In leading cases, familiar institutions like the 'double auction' are as good or better than any feasible institution. But in general, optimal institutions are complex and sensitive to features of the setting that real institutions and real traders ignore. The second line reconsiders the classic analysis, replacing Nash equilibrium by an alternative “class of level-k” models of strategic decision-making, which make weaker behavioral assumptions with more experimental support than Nash equilibrium. Surprisingly, the more realistic models allow an analysis with most of the precision and power of the classic analysis. The unpredictability of traders’ strategic thinking forces the optimal institution to take the much simpler form of a posted-price mechanism, where an optimal price is set, and trade takes place if and only if the buyer's bid is above the seller's ask. Thus using models that are behaviorally more realistic yields results that are less brittle and (perhaps unsurprisingly) more like real-world institutions than in the classic equilibrium-based analysis of optimal bargaining institutions.

A second line of study concerns the design of optimal bargaining institutions. A classic 1983 paper proposed a novel solution to this question, in which one trader owns an indivisible object and would be willing to sell it for enough money, and another would be willing to buy it if the price is right, but neither knows the other’s value, hence neither knows whether it would be mutually beneficial to trade. The ideal outcome would be for them to trade if and only if the buyer's value is higher than the seller's, at a price that fairly shares the gains. No familiar bargaining institution always achieves this ideal outcome. The classic analysis stepped outside the box by asking whether there is any feasible institution that achieves it, given the incentives that it creates, thus going beyond considering institutions as given and immutable, instead thinking of them as something that can be designed. They used traditional game-theoretic methods, assuming that traders are rational and will make decisions that are in Nash equilibrium, in the sense that my decision is best for me given yours, and vice versa. Assuming that traders would respond to any chosen institution by playing their Nash equilibrium strategies, the classic analysis asked whether there is any institution that ensures the ideal outcome, given the incentives it creates. The answer is ‘No’. In leading cases, familiar institutions like the 'double auction' are as good or better than any feasible institution. But in general, optimal institutions are complex and sensitive to features of the setting that real institutions and real traders ignore. The second line reconsiders the classic analysis, replacing Nash equilibrium by an alternative “class of level-k” models of strategic decision-making, which make weaker behavioral assumptions with more experimental support than Nash equilibrium. Surprisingly, the more realistic models allow an analysis with most of the precision and power of the classic analysis. The unpredictability of traders’ strategic thinking forces the optimal institution to take the much simpler form of a posted-price mechanism, where an optimal price is set, and trade takes place if and only if the buyer's bid is above the seller's ask. Thus using models that are behaviorally more realistic yields results that are less brittle and (perhaps unsurprisingly) more like real-world institutions than in the classic equilibrium-based analysis of optimal bargaining institutions.