STOCK PRICES

Stock Prices



Stock Prices

Introduction

Recent empirical evidence has lent strong support to the hypothesis of mean-reversion in stock prices. In particular, Fama and French (1988) and Poterba and Summers (1988) report impressive findings that stock prices are mean-reverting (i.e., contain a slowly decaying temporary component) and induce returns characterised by large negative autocorrelations at long horizons.

The reliability of the multi-period return tests employed in these studies has recently been questioned (see, for example, Richardson and Stock, 1989; Kim et al., 1991) and more recently, vector autoregressive analysis of stock prices and dividends has also been used to identify the permanent and temporary components of stock prices (Cochrane, 1994; stockLee, 1995).

In this letter we seek to contribute to this literature in a number of ways. First, we employ a vector autoregression (VAR) of real stock prices and nominal interest rates in order to identify the temporary and permanent components of stock price movements. Further, by decomposing real stock price movements into mutually orthogonal temporary and permanent components, we are able to measure the size and statistical significance of the mean-reverting component of US stock prices. Finally, we examine the sensitivity of the mean-reverting component using robust estimation procedures in order to allow for possible non-normality of the innovations to stock returns and interest rates.

VAR decomposition technique

In order to effect a decomposition of real stock prices into permanent and temporary components, we need to consider the multivariate time series representation of stock returns together with another appropriate variable. We prefer not to consider the multivariate representation of stock returns and dividend changes or stock price-dividend yields in this paper primarily because the primary underlying rationale for considering such a representation — namely the present value model of stock prices — has not been well supported empirically (see Campbell et al., 1997 for a survey of the evidence).

Rather, we consider a general, unrestrictive multivariate analysis of real stock returns and changes in nominal interest rates. While the presence of a strong empirical relationship between these variables is uncontroversial and well documented, our analysis is sufficiently general to allow for various anomalies and puzzles which have been recorded. For example, while the standard present value model suggests a close association between stock returns and interest rates (see again Campbell et al., 1997), the real-world relationship between these two families of financial returns appears to be highly complex and subject to anomalies such as the ...