The Predictability of Stock Returns: A Cross-Sectional Simulation

Abstract

This paper investigates whether predictable patterns that previous empirical work in finance have isolated appear to be persistent and exploitable by portfolio managers. On a sample that is free from survivorship bias we construct a test wherein we simulate the purchases and sales an investor would undertake to exploit the predictable patterns, charging the appropriate transaction costs for buying and selling and using only publicly available information at the time of decision making. We restrict investment to large companies only to assure that the full cost of transactions is properly accounted for. We confirmed on our sample that contrarian strategies yield sizable excess returns after adjusting for risk, as measured by beta. Using analysts' estimates of long-term growth we construct a test of the Lakonishok, Shleifer, and Vishny (1994) hypothesis. We cannot reject the hypothesis that neither the low-expected-growth portfolio nor the high-expected-growth portfolio yielded any risk-adjusted excess return over the 1980s. Our finding suggests that the superior performance of contrarian strategies cannot adequately be explained by the superior performance of stocks with low expected growth.