Learning and the improving relationship between investment and q∗

We show that the relationship between aggregate investment and Tobin’s q has become remarkably tight in recent years, contrasting with earlier times. We attribute this change to the growing empirical dispersion in Tobin’s q, which we document in both in the cross-section and the time-series. To study the source of this dispersion, we augment a standard investment model with learning. Information acquisition endogenously amplifies volatility in the firm’s value function. Perhaps counterintuitively, the investment-q regression works better for research-intensive industries, a growing segment of the economy, despite their greater stock of intangible assets. We confirm the model’s predictions in the data, and we disentangle our learning mechanism from measurement error in q. ∗Andrei: [email protected]; Mann: [email protected]; Moyen: [email protected]. We thank Andrew Abel for helpful comments on an early version of this paper.