Charles P. Jones

Empirical anomalies based on unexpected earnings and the importance of risk adjustments

Abstract The purpose of this paper is to reexamine Reinganums study which indicates that abnormal returns could not be earned unexpected quarterly earnings information, and to document precisely the response of stock prices to earnings announcements. This study, using a very large sample of stocks and daily returns, represents the most complete and detailed analysis of quarterly earnings reports that has been performed to date. Our results are contrary to those of Reinganum and show that abnormal returns could have been earned almost any time during the 1970s. Our analysis also indicates that risk adjustments matter little in this type of work. Finally, we find that roughly 50% of the adjustment of stock returns to unexpected quarterly earnings occurs over a 90-day period after the earnings are announced.

An Analysis of the S&P 500 Index and Cowles's Extensions: Price Indexes and Stock Returns, 1870-1999

This article provides a consistent monthly stock price index from January 1871 through 1999. The broadly defined S&P Weekly Index is reconstructed from 1918 and carried forward as the S&P 500 Composite Index to the present. Cowless monthly index is improved in order to provide month-end estimates from February 1885. Cowless estimates of dividends and earnings for this index from 1871 are reevaluated and are carried forward until spliced to the S&P daily estimates that began in 1957. The result is a monthly index of prices, dividends, and earnings based on consistent definitions over a period of 130 years.

A Comparison of Annual Common Stock Returns: 1871-1925 with 1926-85

Lawrence Fisher and James H. Lorie, and Roger G. Ibbotson and Rex A. Sinquefield have documented annual returns on common stocks since 1926. Prior to 1926, due to the work of the Cowles Commission, annual returns can be extended back to January 1871. This study utilizes Alfred Cowless reconstruction of common stock returns to provide a comparison between the periods 1871-1925 and 1926-85. A comparable series of annual returns over the complete 115-year period is developed in both nominal and inflation-adjusted terms. The comparison of the two periods suggests that the inflation-adjusted return averages 6.6 percent with similar variability between the two periods. Copyright 1987 by the University of Chicago.

The Changing Nature of Stock and Bond Volatility

This article examines the changing nature of U.S. stock and bond risk from 1871 through 2000 and the implications for asset allocation. Using geometric means and standard deviations, we examine nominal and inflation-adjusted monthly returns over nonoverlapping 5-year periods, as well as annual returns over periods of approximately 25 years, and we document how stock and bond volatility changed over the period. Our analysis suggests that the relative change in the volatility of stocks and volatility of bonds over the past 50 years has increased the importance of stocks in asset allocation. The change is even more pronounced when inflation is considered. This article examines the changing nature of stock and bond risk from 1871 through 2000 and the implications for asset allocation. Using geometric means and standard deviations, we examine nominal and inflation-adjusted monthly returns over five-year periods, as well as annual returns over periods of approximately 25 years, and document how stock and bond volatility changed over the sample period. Our analysis suggests that the change in the relative volatility of stocks and bonds over the past 50 years has increased the attractiveness of stocks in asset allocation, and the change is even more pronounced when inflation is considered. Since about 1940, stock volatility has fluctuated in a narrow range, and both low and high mean stock returns have been associated with similar levels of volatility. But bond volatility increased during the last 35 years of the series. The best 5-year nominal mean returns on bonds occurred during a 10-year period when bond volatility was at its highest level in history. The geometric mean nominal returns of stocks exceeded those of bonds in 18 of the 26 nonoverlapping five-year periods. Inflation-adjusted geometric mean stock returns were negative in only 3 of the 26 periods, but for bonds, they were negative in 10 of the 26 periods. The inflation-adjusted geometric standard deviation of bonds was 30 percent higher than the nominal standard geometric deviation for the 1871–2000 period. For stocks in this period, however, there was little difference between inflation-adjusted and nominal geometric standard deviations. The relative riskiness of stocks and bonds has undergone a long-term change. Until roughly 1950, the ratio of the two variances (stocks to bonds) was much greater than it has been subsequently except for a single five-year period. An examination of five-year standard deviations indicates that bond risk has increased since the 1960s whereas stock risk has remained relatively steady. The correlation between bond returns and stock returns, although fluctuating, has been increasing. Combined with the increase in bond volatility relative to stock volatility, this rising correlation has important implications for asset allocation. Our analysis of the nominal risk–return trade-off available to investors shows that the situation changed after World War II. For the later two 25-year periods examined here, a 100 percent bond portfolio, or a portfolio invested primarily in bonds, compared unfavorably on a return–risk basis with several portfolios that had larger stock allocations. This outcome was most pronounced in the last period, 1974–2000, when a 70/30 stock/bond allocation had less risk and a much larger return than did a 100 percent bond portfolio. Clearly, during the last half of the 20th century, the changes in relative stock and bond volatility increased the attractiveness of stocks relative to bonds. On an inflation-adjusted basis, the case for portfolios heavily invested in bonds is even weaker than it is on a nominal basis. Bonds are affected more severely when adjusted for the increased risk caused by the covariance of nominal bond returns and inflation.

Ordinal Predictions and the Selection of Common Stocks

rates of return, variances of rates of return, and covariances of rates of return among individual securities. Unfortunately, a feasible method of accurately generating the massive information requirements of the Markowitz model has not been developed. Historical measures of mean rates of return and covariances of rates of return among individual securities have been shown to be unstable over time and to be ineffectual in generating ex ante efficient portfolios.1 Sharpe (10] has dichotomized a securitys total risk into its systematic and residual components. A securitys systematic risk denotes the portion of its standard deviation explained by the market, and its residual risk denotes the portion of its standard ieviation unexplained by the market. The maximum gains from diversification will asymptotically reduce the contribution each security makes to the portfolios standard deviation to the systematic risk of that individual security. That is, as the investor increases the number of securities in his portfolio of common stocks, the residual component of the portfolios standard deviation asymptotically approaches its lower limit of zero*

Estimating Stock Returns

How do we quantify the level of return that an investor can expect in the future? An examination of the historical distribution of total returns reveals declines in dividend yields and new likely lower boundaries for price appreciation. It is often asserted that low dividend yields brought about by higher earnings retention should be followed by greater price appreciation as a firm invests retained earnings into new projects. The available recent evidence refutes this assertion. Barring some significant reversal of current conditions, short-term and possibly intermediate-term returns from stocks will be lower than what many investors may be anticipating.

Asset Allocation Decisions Making the Choice Between Stocks and Bonds

Given renewed interest in bonds, this article considers asset allocation decisions in the context of realized returns for bonds and stocks. The study is based on new and improved data on bond returns covering long periods of time as well as calculations of the probabilities associated with stock bond returns. An analysis of these probabilities using inflation-adjusted returns suggests that the probability of earning sizable compound rates of return with bonds is not only small, but also declines over time. Therefore, despite recent strong performance by bonds, our results suggest investors should be cautious in making asset allocation decisions.

Earnings announcements: Pre- and -post responses

hy does there appear to be such a welldocumented lag in the adjustment of stock returns to earnings announcements? If the market is efficient, returns should quickly reflect information that is publicly available, including its unexpected component, so that no abnormal returns are achievable after earnings announcements. Does the market adjust returns substantially before the earnings announcement date for anticipated extreme earnings surprises, and, if so, by how much? Or does most of the adjustment follow the announcement? Support for the latter possibility would be particularly troublesome for the concept of market efficiency. The purpose of this article is to document precisely the cumulative movement of stock returns relative to the announcement date of quarterly earnings, separated by both the size and the sign ofl the unexpected earnings. We perform this separation, by using a two-way classification relating the unexpected earnings in any given quarter to the unexpected earnings in the preceding quarter. Such a classification provides more detail about the overall relationship between unexpected earnings and stock returns than would be possible using a one-way classification. Our tests employ a large sample of stocks and cover almost the entire decade of the 1970s, providing extensive documentation for the adjustment of stock

The Equity Risk Premium Controversy

The equity risk premium controversy is a puzzle with implications for investors and money managers. Whatever the arguments about a high equity risk premium in the past, there has been a substantial premium over time on average, although typically not as great a realized premium to investors. There is similarly credible evidence that an equivalent or only slightly lower premium may continue into the future.

Long-Term Returns and Risk for Bonds

CHARLES P. JONES is GLU professor of 6nance at the College of Management of North Carolina State University in Raleigh (NC 27695). D espite the vast amount of information available about bond yields and returns, there are differences in the data series involving this important asset class. A good example is seen in the Treasury bond data for 1994. For long-term Treasuries for 1994, Ibbotson Associates reports the yield increasing from 6.54% to 8.09%, for a total return of -7.77%. In contrast, the Lehman Brothers long-term Treasury index shows yields rising from 6.42% to 7.98%, for a total return of -6.94%, a substantial difference in reported results. Some vendors of bond data complain about the validIty and usefulness of competing sources of data supplied to users. There are also gaps in bond data when compared to common stocks, for which we have reconstructed the SP Schwert [1990] has extended market data even farther back. Meanwhde, the experience of bondholders over the years provides strong evidence that bonds are a risky asset for whch reliable, long-term information is needed. This article provides an independent estimate of monthly yields and returns for both government and corporate bonds over a very long period of time. These data are consistent and reliable, and can easily be updated by users of such data on a real-time basis with the methodology outhned here. This data series provides additional insights into the long-term returns and risks from bonds; furthermore, the data are realistic estimates of the actual