Table VII

Monthly Cross-Sectional Regressions of Returns on Prior Return and Prior Earnings Surprises

Cross-sectional regressions are estimated each month from January 1977 to January 1993 of individual stock returns on size, compound return over the prior six months (R6), the abnormal return relative to the equally-weighted market index cumulated from two days before to one day after the most recent past announcement date of quarterly earnings (ABR), unexpected earnings (the change in the most recent past quarterly earnings per share from its value four quarters ago) scaled by the standard deviation of unexpected earnings over the past eight quarters (SUE), and a moving average of the past six months revisions in I/B/E/S mean analyst earnings forecasts relative to beginning-of-month stock price (REV6). In the regression each explanatory variable is expressed in terms of its percentile rank and scaled to fall between zero and one. The dependent variable is the stocks buy-and-hold return either over the subsequent six months (Panel A), or over the next year (Panel B). The reported statistics are the means of the time series of coefficients from the month-by-month regressions, and in parentheses the -statistics relative to the autocorrelation-adjusted standard error of the mean. The sample includes all domestic primary firms on New York Stock Exchange (NYSE), American Stock Exchange (AMEX), and Nasdaq with coverage on the Center for Research in Security Prices (CRSP) and COMPUSTAT.

The continuation in stock price movements over the intermediate term includes a component unrelated to the news in near-term earnings. Finally, a comparison of the results in the two panels reinforces the impression from the

earlier sections that price momentum tends to have longer-lasting effects than earnings momentum.

IV. Are Price and Earnings Momentum Subsequently Corrected?

One way to distinguish between some of the competing explanations for continuations in price movements is to examine whether there is a subsequent correction in the stock price.

In the one-way classifications by prior return (Table II), it is hard to find direct evidence of return reversals in the years following portfolio formation. The raw returns in the second and third following years are not very different across portfolios. There is, however, some tendency for the extreme decile portfolios to concentrate on smaller stocks. The stocks in portfolios 1 and 10 have an average size decile ranking of 2.9, while the average size decile ranking of the stocks in the other portfolios lie between 3.7 and 4.4. The size rankings are based on the breakpoints from the distribution of market capitalization for NYSE stocks. The smaller average capitalization of stocks in the winner portfolio pulls their average return in one direction, but at the same time their lower book-to-market ratio pulls the return in the other direction. All in all, the picture with respect to reversals in the return on the winner portfolio is muddy. On the other hand, the raw returns on the loser portfolio in the following years tend to stay low (the more so taking into account the smaller average capitalization and higher book-to-market ratio of portfolio 1). The lack of direct evidence on reversals tends to call into question the hypothesis that the continuation in prices is induced by positive feedback trading.

Similarly, the one-way sorts by prior earnings surprise (Tables III to V) also fail to turn up signs of subsequent return reversals. Future returns to stocks with bad news about earnings tend to stay relatively low. For both price and earnings momentum, therefore, there do not seem to be any price corrections in subsequent years.

The two-way classifications in Table VI give a sharper verdict on whether the movement in prices is permanent or transitory. Although the portfolios that are ranked highest by both prior return and earnings surprise always have the largest return in the first following year, their returns are not much different from average in the second and third following years. For example, portfolio (3,3) in Panel В of the table has a return of 19.9 percent in the second year, compared to the overall mean return that year of 19.6 percent for all stocks in the sample; its return in the third year is 20 percent compared to 19.5 percent for the entire sample.

At the other end of the scale, the persistence in poor performance is striking. In particular, the doubly-afflicted portfolio (1, 1), with poor past price performance and bad earnings news, continues to suffer a drawn-out decline. Even two and three years after portfolio formation, the portfolios returns in Panels A to С continue to fall below the average. In Panel C, for instance, the returns for portfolio (1, 1) are 16.9 percent and 16.4 percent in the second and third years respectively, which are the lowest returns in each year across the nine

portfolios in the panel. The shortfall in returns would be even more dramatic if the small size and high book-to-market ratio of the loser portfolio were to be taken into account. It might be argued that a more rapid adjustment in the prices of these poorly performing stocks runs up against several obstacles: it is more difficult to enter into short positions than long positions, and security analysts, as we have noted above, tend to acknowledge only gradually the negative prospects for these firms.

In the case of stocks that are ranked highest by prior return, an interesting dichotomy emerges when we condition on whether the past returns are confirmed by earnings news. For example, for those cases in Panel A where the stock is ranked highest by prior return and, in addition, the high past returns are validated by high announcement returns (portfolio (3, 3)), the average first-year return is 27 percent. When the earnings news does not confirm the past returns (portfolio (1, 3)), however, the average first-year return is 21.3 percent, which is only slightly higher than the overall mean first-year return of 20 percent across all stocks in the sample. By the second year, the return on portfolio (1, 3) is 18.3 percent, which is below the overall average of 19.6 percent, while the return on the twice-favored portfolio (3, 3) is about 20 percent. In the same fashion, the results in Panels В and С confirm a reversal in returns in the second year for those cases where high past returns are not supported by similarly favorable news about earnings. Much of the performance of stocks with high price momentum thus occurs when high prior returns are accompanied by favorable news about earnings.

V. Other Tests

A. Price and Earnings Momentum for Large Stocks

In this section, we apply our momentum strategies to a sample composed of larger stocks only. Limiting attention to the larger stocks helps to alleviate potential problems of survivor bias in the sample, and problems with low-priced stocks.10 Stocks with higher market capitalization are also of more interest to institutional investors.

In Tables VIII and IX, the sample comprises stocks whose market capitalization as of the portfolio formation date exceeds the median market value of NYSE stocks. In order to minimize repetition, we report results only for returns in the first year following portfolio formation. Even for this set of large firms, which are more widely followed and for which timely information should be more readily available, there is still evidence that the market adjusts only gradually to the information in past returns or past earnings news. Notably, the one-way sorts in Panel A of Table VIII continue to deliver sizable differences in returns. This is particularly true when stocks are ranked by prior return; the spread in future one-year returns is 14 percent, which is almost as large as the spread for the entire sample in Table II (15.4 percent). A large

10 In the last portfolio formation period, there are only two stocks in our large-stock sample (out of about a thousand eligible stocks) that have prices below five dollars.