The small-cap effect in the predictability of individual stock returns

Highlights

• We investigate the predictability of the daily open-to-open and close-to-close returns.

• The predictability of both types of returns increases from bigger to smaller stocks.

• The increase is stronger for the open-to-open returns.

• We call this the small-cap effect in the predictability of individual stock returns.

• The small-cap effect is more pronounced for the high-momentum stocks.

Abstract

The paper investigates whether the choice of opening prices yields the same predictability of daily returns on individual stocks as the choice of closing prices. Based on the sample of NYSE, Nasdaq, and AMEX listed stocks for the period from January 2009 to November 2013, it is concluded that (a) the degree of predictability and implied forecasting accuracy of both types of returns substantially increase as smaller stocks are examined and (b) this increase is stronger for the open-to-open returns than for the close-to-close returns. This small-cap effect in the predictability of individual stock returns is stronger for the high-momentum stocks.

Introduction

It has become a standard practice in finance research to compute stock returns as the changes in closing prices. Several papers have, however, found empirical evidence suggesting that the time-series behavior of the open-to-open returns on stocks differs substantially from the behavior of the returns computed on the close-to-close basis. Harris (1986), Amihud and Mendelson, 1987, Amihud and Mendelson, 1991, Lockwood and Linn (1990), Stoll and Whaley (1990), Chan, Fong, Kho, and Stulz (1996), Cao, Choe, and Hatheway (1997), Ronen (1997), Hong and Wang (2000), and Datar, So, and Tse (2008), among others, observe that the open-to-open returns on individual stocks are typically more volatile than the close-to-close returns. They argue that this may be due to the presence of private information revealed in trading and the non-synchronization of liquidity trading (Stoll & Whaley, 1990), different trading mechanisms at the open and the close (Amihud & Mendelson, 1987), greater uncertainty resulting from the long non-trading period before the market opening (Amihud & Mendelson, 1991), or methodological issues of the measurement of the opening and closing returns (Ronen, 1997), for instance.

When comparing the time-series properties of the open-to-open and close-to-close returns on individual stocks, it is common to examine exclusively the mean and variation of stock returns over time. It may, however, be of interest to investigate whether there are other (than the first two moments of the distribution) return time-series characteristics that are also different for these two types of stock returns. One of such characteristics can be the extent to which the returns on individual securities are serially correlated and, hence, predictable from past price changes.

The efficient-market theory claims that, in an informationally efficient market, prices of traded assets immediately and fully reflect all relevant information and, therefore, any future price changes are determined entirely by future news. Under this hypothesis, future prices are unpredictable based on the price history and, hence, asset returns of any type are serially uncorrelated at all leads and lags, i.e., follow a random walk.

The existing empirical literature on the predictability of stock returns examines solely the behavior of the close-to-close returns.¹ The aim of this study is to investigate whether the choice of opening prices yields the same degree of predictability of the daily returns on individual stocks as the choice of closing prices and, if the open-to-open and close-to-close stock returns exhibit different degrees of predictability, what firm characteristics account for this difference. To the best of our knowledge, this is the first study that compares (both theoretically and empirically) the extents to which the open-to-open and close-to-close returns on individual stocks are predictable given the stock's price history.

As a measure of the degree of predictability of a particular type of stock returns, we use the proportion of stocks, for which the random walk model is rejected statistically (i.e., the returns of that type are forecastable based on historical price information), in the total number of stocks in a sample. The (statistically) greater this proportion, the greater the extent to which the considered type of stock returns can be predicted by past price changes alone.

Using data on all NYSE, Nasdaq, and AMEX listed individual stocks with complete price histories during the sample period from January 2, 2009 to November 15, 2013, we find that (a) the degree of predictability of both the daily open-to-open and close-to-close returns is statistically significantly greater for the stocks with small capitalization and (b) while there is no statistically significant difference in the degree of predictability of the daily open-to-open and close-to-close returns on the mid- and large-cap stocks, for the small-cap securities the proportion of stocks with forecastable returns is statistically significantly greater for the open-to-open returns than for the returns computed from closing prices. We refer to this dependence of both the absolute and relative degrees of predictability of the daily open-to-open and close-to-close returns on the market value of stocks as the small-cap effect in the predictability of individual stock returns.

The rest of the article proceeds as follows. Section 2 reports the description of the data set used in the tests. Based on the autocorrelation decomposition, Section 3 investigates the factors that may influence the predictability of the daily open-to-open and close-to-close returns. Section 4 describes the testing methodology and reports the test results. Section 5 assesses the size of the improvement in the forecasting accuracy resulting from the more accurate specification of the forecasting model. Section 6 concludes.