Testing the Q theory of investment in the frequency domain

We revisit the empirical performance of the Q theory of investment, explicitly taking into account the frequency dependence of investment, Tobin's Q, and cash flow. The time series are decomposed into orthogonal components of different frequencies using wavelet multiresolution analysis. We find that the Q theory fits the data much better than might be expected (both in-sample and out-of-sample) when the frequency relationship between the variables is taken into account. Merging the wavelet approach and proxies for Q recently suggested in the investment literature also significantly improves the quality of short-term forecasts.