Abstract:

In the first chapter, we elicit beliefs from observers about strategic choices in a matrix game under three incentive structures: a quadratic scoring rule, a linear scoring rule, and a constant payoff. Our experimental evidence suggests that both the lack of proper incentives for stating accurate or truthful beliefs and elicitation from players who have a stake in the game may lead to unreliable belief reports. We find evidence of forecast revision and convergence towards the consensus but little to no evidence of improvements in belief accuracy on the individual or group level following feedback on other group members' forecasts. We also examine heterogeneity in revision heuristics and correlates of revision strategy.

In the second chapter, we investigate theoretically and experimentally a dynamic game of two-sided incomplete information with player types drawn from a commonly known distribution. We elicit beliefs from players about their opponent's type using a quadratic scoring rule. Our two treatments vary a terminal node payoff to generate sharp comparative static predictions in both beliefs and strategies. We find that beliefs track departures from perfect Bayesian equilibrium predictions about strategy choice. We highlight evidence for two deviations from Bayesian beliefs: conservative updating and motivated beliefs. We also examine the extent of rational expectations, common knowledge of rationality, and best response to beliefs.

In the final chapter, we build and experimentally test a model of speculation in binary prediction markets with subjective information flow. The speculation arises from the traders? heterogeneous posteriors as they make different inferences from the subjective pieces of information. We find consistent overpricing of the assets in both incomplete and complete markets where the information flow is a gradually revealed sequence of guesses from an information cascade experiment about the state of the world. There is significantly less overpricing when the information flow contains objective signals from the information cascade environment and traders should theoretically have homogeneous beliefs. We also test other model predictions about asset holdings, asymmetric response to good and bad news, and price dynamics as more information arrives and less time remains in the market.