NBER Working Papers by Laura X. L. Liu

Optimal investment of firms implies that expected stock returns are tied with the expected marginal benefit of investment divided by the marginal cost of investment. Winners have higher expected growth and expected marginal productivity (two major components of the marginal benefit of investment), and earn higher expected stock returns than losers. The investment model succeeds in capturing average momentum profits, reversal of momentum in long horizons, as well as the interaction of momentum with market capitalization, firm age, trading volume, and stock return volatility. However, the model fails to reproduce procyclical momentum profits.

The neoclassical q-theory is a good start to understand the cross section of returns. Under constant return to scale, stock returns equal levered investment returns that are tied directly with characteristics. This equation generates the relations of average returns with book-to-market, investment, and earnings surprises. We estimate the model by minimizing the differences between average stock returns and average levered investment returns via GMM. Our model captures well the average returns of portfolios sorted on capital investment and on size and book-to-market, including the small-stock value premium. Our model is also partially successful in capturing the post-earnings-announcement drift and its higher magnitude in small firms.

Previous work shows that the growth rate of industrial production is a common macroeconomic risk factor in the cross-section of expected returns. We demonstrate the connection between momentum profits and shifts in factor loadings on this macroeconomic variable. Winners have temporarily higher loadings on the growth rate of industrial production than losers. The loading dispersion derives mostly from the high, positive loadings of winners. Depending on model specification, this loading dispersion can explain up to 40% of momentum profits.