Collective Learning and Firms' Investment

In the standard neoclassical investment theory, decision makers are assumed to have perfect knowledge of the parameters characterizing firms' operation. From the perspectives of firm managers, however, many of underlying parameters are unknown at the beginning, and instead, they are what the managers need to learn over time by observing related variables. Among such parameter uncertainties, this project focuses on the uncertainty regarding the systematic risk of a firm's productivity, that is the covariance between the productivity and the pricing kernel. This project proposes a dynamic model where the agent learns about the systematic risk from the up-to-date history of productivity and accordingly makes capital investment decisions. A key element is that the systematic risk is industry-specific, so firms learn about it by observing not only their own productivity shocks but also shocks to peers in the same industry. This collective learning takes place because firms that belong to the same industry tend to have similar systematic risks. Once calibrated, the model will predict certain quantitative relationships among learning-related variables, capital investment, and stock returns, which can be tested empirically. The preliminary result shows that a firm's capital investment increases with a learning-related variable, the cumulative number of firm-year observations in the industry to which the firm belongs. This evidence supports the presence of the collective-learning mechanism.