Social Learning, Misperception, and Imperfect Information

Description

Social learning refers to situations in which individuals draw inference from other individuals' actions to learn about an underlying state of the world. The classic literature on social learning assumes that individuals have a correct understanding of their environment. However, growing empirical evidence shows that individuals are subject to systematic misperceptions of the external environment. For example, consumers may misinterpret the interests or tastes of other consumers, and investors may over- or under-estimate the returns of financial securities. A burgeoning theoretical literature demonstrated that certain types of misperceptions can cause large inefficiencies in social learning. This recent literature focuses on a stylized model in which individuals have perfect information about the actions of all other individuals. In practice, however, individuals have complex and imperfect information about others' actions. For instance, consumers consult bestseller lists to learn about other consumers' purchase decisions, and investors use aggregate financial statistics to infer other investors' investment choices. In order to understand and regulate these markets, it is important to study the joint impact of misperception and imperfect information on social learning. This requires us to move beyond the literature’s simple stylized model with perfect information.In the proposed project, I will develop a general framework to understand how imperfect information and misperception jointly impact social learning. I will provide a unified theory on the dynamics and long-run outcome of social learning, putting little restriction on the specific forms of the individuals' misperception and imperfect information. Three earlier studies, including one by the PI, have examined the role of imperfect information in classic social learning models, which involves individuals accurately understanding their environment. However, no study has analyzed the role of imperfect information in social learning when individuals also have misperceptions.More specificly, I will first compare the model of the proposed project with existing models in the literature which analyses misperception under perfect information. The preliminary results show that learning dynamics exhibit fundamentally different properties in our framework. In particular, a new class of learning dynamics referred to as "incomplete learning" may emerge in our model, but it is impossible to occur in the existing models with perfect information. Then, I will examine different forms of imperfect information and misperception to identify general patterns of social learning in our framework. Ultimately, I seek a unified characterization to classify social learning environments in terms of the induced learning outcomes.