Are US stock index returns predictable? Evidence from automatic autocorrelation-based tests

Abstract

This article re-examines the evidence of return predictability for three major US stock indices using two recently developed data-driven tests, namely the automatic portmanteau Box–Pierce test and the wild bootstrapped automatic variance ratio test. In tracking the time variation of return predictability via rolling estimation window, we find that those periods with significant return autocorrelations can largely be associated with major exogenous events. Theoretically, the documented time varying nature of predictable patterns is consistent with the adaptive markets hypothesis.

Keywords:

• autocorrelation test
• variance ratio test
• adaptive markets hypothesis
• evolving return predictability
• US stock market

JEL Classification::

• C12
• G14

Notes

¹ The second group of studies examines the profitability of trading strategies based on past returns such as technical trading rules (see the survey paper by Park and Irwin, 2007), momentum and contrarian strategies (see references cited in Chou et al., 2007).

² Two decades later, Fama (1991) reclassifies the weak-form EMH as tests for return predictability in which the coverage has been expanded from past returns to include other financial variables such as the dividend-price ratio, earnings-price ratio, book-to-market ratio and various measures of the interest rates (see references cited in Lim and Brooks, 2011).

³ Nevertheless, there are other competing views on the potential sources of stock return autocorrelations (Boudoukh et al., 1994).

⁴ The conventional wisdom is that developed stock markets are expected to be more efficient than their emerging counterparts, in which the stock returns in the former should exhibit less predictability (Antoniou et al., 1997; Lim and Brooks, 2009b; Griffin et al., 2010).

⁵ Gu and Finnerty (2002) analyse the evolution of market efficiency in the US stock market using 103 years of daily data (1896–1998) for the Dow Jones Industrial Average (DJIA). The authors compute the first-order autocorrelation between daily returns for each individual year using variance ratio, serial correlation and runs tests. Their results show that the magnitude of autocorrelation experiences a sharp decline since late 1970s, indicating that the market has evolved and attained efficiency during the final two decades of their sample period. Lo (2004, 2005) computes the rolling first-order autocorrelation for monthly returns of the S&P Composite Index from January 1871 to April 2003. His graphical plot reveals the degree of efficiency varies through time with surprising result that the US market is more efficient in the 1950s than in the early 1990s. Using a state-space model, Ito and Sugiyama (2009) also report that the US stock market exhibits varying degree of efficiency over the sample period from 1955 to 2006. More specifically, their graphical plot of the time-varying autocorrelation coefficients shows that the US market is the most inefficient during the late 1980s and has become the most efficient at around 2000.