Investment-based asset pricing models typically predict a close link between a firm’s stock return and its characteristics at any point in time. Yet, previous studies have primarily focused on the weaker prediction that this link holds on average, finding substantial empirical support. We show how to incorporate the time-series predictions in the estimation and testing of investment-based models using the generalized method of moments. We find that standard specifications of investment-based models with one physical capital input fail to match the time series properties of stock returns in the data, and discuss the implications of the findings for future research.

Using a novel measure of industry exposure to government spending, we show predictable variation in cash flows and stock returns over political cycles. During Democratic presidencies, firms with high government exposure experience higher cash flows and stock returns, while the opposite pattern holds true during Republican presidencies. Business cycles, firm characteristics, and standard risk factors do not account for the pattern in returns across presidencies. An investment strategy that exploits the presidential cycle predictability generates abnormal returns as large as 6.9% per annum. Our results suggest market underreaction to predictable variation in the effect of government spending policies.

What are the economic determinants of a firm’s market value? We answer this question through the lens of a generalized neoclassical model of investment with quasi-fixed labor and three heterogeneous capital inputs. We estimate the structural model using firm-level data on US firms and find that, on average and depending on the industry, installed labor force accounts for 14–21% of firms’ market value, physical capital accounts for 30–40%, knowledge capital accounts for 20–43%, and brand capital accounts for 6–25%. Our analysis provides direct empirical evidence for the importance of labor and intangible capital inputs for understanding firm value.