【主讲】教沛然 (牛津大学)
【主题】Payoff-based Belief Distortion
【时间】2016年9月14日 (周三) 15：30-17：00
【地点】上海财经大学经济学院楼701室
【语言】英文
【摘要】This paper offers a simple model where experienced payoffs influence subsequent choices through two channels: reinforcement learning and a payoff-based belief distortion. In the first channel, past rewards from an action increases the probability of that action being retaken subsequently. The second channel involves the agent updating beliefs by underweighting evidence that contradicts experienced gain/loss, and maximizing expected utility given distorted beliefs. Three experiments tested the model predictions. A common feature of all experiments is that experienced gain/loss was information-free, i.e. bearing no informational value for Bayesian updating. The first two experiments exogenously manipulated the experience of gain/loss, and found participants' beliefs as well as their updating process to be biased. The third experiment allowed participants to obtain experience endogenously in an asset buying task with clear tension between Bayesian and reinforcement strategies. Participants' beliefs were distorted, and their buying decisions exhibited overweighting of (especially recent) experienced payoffs relative to the Bayesian benchmark.
