Six Decades of Significant Autocorrelation in the U.S. Stock Market
14 Pages Posted: 1 Feb 2008
Date Written: January 20, 2008
Abstract
This paper assesses the autocorrelation patterns in U.S. stock market indices: the S&P 500 (1962-2007), the Dow Jones Industrial Average (1929-2007), and the NASDAQ Composite (1972-2007). Statistically significant lag 1 autocorrelation was observed for all three time series as tested by Monte Carlo simulation. The autocorrelation exhibited by the S&P 500 agreed with the findings of Lo and MacKinlay. The Dow Jones Industrial Average displayed strong autocorrelation at lag 1 from 1940-1986 (163/182 quarters, or 87%, had a positive autocorrelation). For all but three quarters from 1972 to 1997, the NASDAQ Composite had a positive autocorrelation at lag 1 averaging 0.25. 61% of the calendar quarters during that period had statistically significant autocorrelations at lag 1.
Keywords: autocorrelation, ARMA, ARIMA, random walk, time series, Box-Jenkins, autoregressive, moving average, stock market, NASDAQ Composite, Dow Jones Industrial Average, S&P 500, variance ratio test
Suggested Citation: Suggested Citation