Abstract: Regressions of investment on Tobin's Q are misspecified in the presence of capital gestation lags because they don't distinguish between the value of existing capital and the value of capital at a future date. Current investment should be determined by the anticipated shadow value of capital at the gestation horizon. Under homogeneity conditions analogous to Hayashi[1982], this value is equal to the forecast of an adjusted version of Q. This misspecification helps to explain many pathologies in the literature: attenuated estimates of the coefficient on Q, low R2, and serially-correlated errors. Regressions using aggregate data suggest that (1) endogeneity problems associated with the standard regression of investment on Q can can be eliminated by reversing the regression, (2) forecastable changes in Q provide additional information about investment not captured in current Q, and (3) specifications that explicitly account for gestation lags yield capital adjustment costs of a more reasonable magnitude.

Full paper (397 KB PDF)