Broadening the Scope of Experimental Observational Learning Studies

Observational learning is an important component in many economic decisions. Observing others making similar decisions is a perfectly rational way to gain information before making a choice. The dissemination of technology, sequential voting and retirement saving are among the environments in which observational learning can play a vital part. As the study of observational learning requires careful control of beliefs, motivation and observation, its study has largely occurred in experimental laboratories. Most observational learning studies follow the model of Bikchandani, Hirschleifer and Welch (1992), in which decision-makers receive a binary private signal and a binary choice. They normally assume that the value of observed decisions is valuable only for their informational content, and that decisions are made in a predetermined order. My dissertation involves three papers that relax several assumptions typical of observational learning studies. Chapter 1 allows for payoff externalities, while chapters 2 and 3 study observational learning when the timing of decisions is endogenous. Chapter 1 is a quantitative study of the observational learning process in the presence of payoff externalities, both positive and negative. Using a continuous-signal, binary choice setting, I am able to isolate individual subjects' beliefs, and on aggregate to find the influence of information and payoff externalities on decisions. I find that payoff externalities play a large role in the observational learning process, and that subjects are more sensitive to them than to the information contained in observed decisions. My results suggest that payoff externalities can have a larger effect on the quality of decisions than predicted by theory. Chapters 2 and 3 employ study observational learning in a setting where the timing of decisions is endogenous. Chapter 2 explores observational learning with un-correlated decision problems, so subjects cannot update their beliefs by observing the behavior of others. I find that the observation of one's peers can help decision-making, even when this observation has no informational content. Subjects learn to make decisions earlier when delaying them is costly. Chapter 3 builds on chapter 2 by adding payoff externalities. A 2 x 2 design manipulates the existence of both payoff externalities and correlated decision problems. Payoff externalities tend to improve decision-making. I posit two explanations. The first is that positive payoff externalities mitigate risk aversion, allowing subjects to make the choice with the highest expected value. The second is that subjects who are insensitive to the informational content in actions that they observe still reap their benefit if they are sensitive to payoff externalities. This design more closely resembles technology adoption decisions than the typical setup, as payoff externalities are often present and agents can engage in costly delay in order to gain more information.