Document Sample

July 26, 2007 Why Has the Market Reaction to Corporate Earnings Announcements Increased over Time?* Christo Pirinsky and Qinghai Wang * Christo Pirinsky is from the College of Business and Economics, California State University, Fullerton, 2600 E. Nutwood; Suite 1060, Fullerton, CA 92831 and Qinghai Wang is from the School of Business Administration, University of Wisconsin – Milwaukee, Milwaukee, WI 53201. Pirinsky can be reached at 714-278-2217 or cpirinsky@fullerton.edu; Wang can be reached at 414-229-4775 or wangq@uwm.edu. We thank Shmuel Baruch, Valentin Dimitrov, Simi Kedia, Feng Zhao and seminar participants at Rutgers University for helpful comments. Why Has the Market Reaction to Corporate Earnings Announcements Increased Over Time? Abstract Stock price reactions to earnings announcements have been increasing over the last three decades, while at the same time the usefulness of reported earnings information to investors has been deteriorating. To reconcile these two findings, we propose an explanation that relates the value of new information (and hence its price impact) to both the precision of the information signal as well as the uncertainty of underlying fundamentals. In a world of more uncertain fundamentals, prices will react more at information releases since the new information would resolve more uncertainty. We confirm this hypothesis by showing that measures of the average cash-flow uncertainty in the market can explain the up-ward trend in stock price reactions at earnings announcements. JEL classification: G14; G20 Keywords: Earnings announcement, stock price reaction, asset pricing 2 Stock return volatility around earnings announcements has increased substantially over the last three decades (see Landsman and Maydew (2002)). This trend has been recognized not only by academic researchers but also by corporate executives and financial market regulators. A recent Associated Press report states that “[q]uarterly earnings announcements look more than ever like prize fights: hyped events that often leave the pugilists battered and the fans drained.”1 Many corporate executives consider stock price reactions at earnings announcements excessive and have attempted to distance investors from quarterly earnings information. In the same report, the Washington Post Chairman and CEO Donald Graham comments “We don’t do quarters. Quarterly earnings are not in the top 100 things you should care about if you want to value the company. ... If you care about that sort of thing, you shouldn't own our stock.” The upward trend in earnings-related return volatility seems surprising given that the relative importance of earnings has actually decreased over time. For example, Ely and Waymire (1999), Francis and Schipper (1999) and Lev and Zarowin (1999) find that the usefulness of reported earnings, cash flows, and book (equity) values to investors has been deteriorating over time. They argue that this trend is largely driven by the increase in the amount and quality of alternative investment information in the market.2 The last few decades have seen tremendous growth of the security analysis industry and numerous financial innovations, all of which facilitate the extraction of information from firms prior to earnings announcements. As a result, one would expect stock price volatility at earnings announcement to decrease over time. Thus, the stronger stock price reaction at earnings announcements seems difficult to reconcile with the declining value of earnings information. 1 See “Should We Hear From Companies More Often Than Four Times a Year?” by Ellen Simon , July 22, 2005, Associated Press. 2 Ryan and Zarowin (2003) also show that earnings increasingly reflect news with a lag relative to stock prices, which implies that the relative importance of earnings relative to non-announcement information tends to decrease over time. Kim and Kross (2005), however, show that the ability of earnings in predicting future operating cash flows has been increasing, suggesting that earnings information is not less valuable for predicting future earnings. 1 The underlying reasons for the upward trend in earnings-related stock return volatility are still unclear. Most of the proposed explanations relate stock price reactions at earnings announcements to the patterns of information disclosure by firms. For example, Francis, Schipper, and Vincent (2002a) argue that the increase in earnings announcement return volatility could be explained by the recent increase in concurrent disclosures by companies around earnings announcements. In a related vein, Collins, Li and Xie (2005) argue that the recent increase in return volatility around earnings announcements is caused by stronger price reactions to “Street” earnings (i.e., I/B/E/S reported actual earnings), in comparison to weaker reactions to GAAP earnings. Another view relates the magnitude of earnings-announcement stock returns to the frequency of disclosure. Along this line, Harvey Pitt, the former chairman of the SEC, suggests in a 2002 Congressional testimony that “[p]ublic companies should be required to make affirmative disclosures of unquestionably material information in real time, including providing updates to prior disclosure.” In this paper we propose an alternative explanation for the up-ward trend in earnings- related volatility in the stock market. Within a simple framework, we show that the value of new information (and hence its price impact) increases in both the precision of the information signal as well as the uncertainty of underlying fundamentals. Most previous studies elaborate on the first aspect of earnings announcements, namely the quality of information disclosure. Here we elaborate on the second one – the uncertainty of firm fundamentals. If the uncertainty of firm fundamentals has increased over time, than this trend could potentially explain the increased stock price reaction at earnings announcements. We show that the uncertainty of firm fundamentals, as measured by earnings volatility, has indeed increased substantially over the last three decades. Consistent with our conjecture, we also find that the average earnings volatility in the market is able to explain the time-series trend in the average stock price reactions at earnings announcements. 2 We begin our empirical analysis by providing comprehensive evidence on the increasing price reactions to earnings news over the period of 1978 to 2004. Over the sample period, stock price reactions to earnings announcements have increased significantly. This increase is uniform for both positive and negative earnings surprises and for stocks listed on both NYSE and NASDAQ. While large stocks experience more dramatic increases in return volatility around earnings announcements, the upward trend is also well pronounced among small stocks. We next confirm that both the information production outside of earnings announcements and the quality of earnings information available prior to earnings announcements have indeed increased over time. We use financial analysts earnings forecast to measure information production and quality of earnings information outside of earnings announcements. We show that both the number of firms covered by financial analysts and the number of analysts covering a firm have increased substantially over the sample period. More importantly, we find that the uncertainty surrounding corporate earnings announcements has decreased over time: the magnitude of both earnings surprises (measured by financial analysts’ forecasts errors) and the uncertainty of earnings forecasts (measured by dispersion of analyst forecasts) have decreased dramatically over the last three decades. The down-ward trend in both earnings surprises and analyst forecast uncertainty suggest that the up-ward trend in stock return volatility around earnings announcements does not reflect an increased importance of earnings information relative to market expectations. Consistent with our predictions, we also show that the uncertainty of underlying firm fundamentals has increased substantially over the last three decades. We use volatility of quarterly earnings (volatility of return on equity) to measure uncertainty of firm fundamentals. Results based on several alternative earnings volatility measures provide consistent evidence on the changing uncertainty of firm fundamentals. We also show that earnings yields (earnings relative to market 3 price) decreased significantly over the sample period, suggesting market valuation of corporate earnings has also increased. To examine the link between firm fundamentals and earnings announcement volatility, we first estimate cross-sectional regressions of absolute earning announcement returns on determinants of stock return volatility around earnings announcements. The cross-sectional analysis confirms that, after controlling for earnings surprises and other firm characteristics, firms with more volatile earnings and firms with lower earnings yields exhibit stronger reactions at earnings announcements. We further implement a direct time-series test of the link between earnings volatility and stock price reactions around earnings announcements. We regress average absolute earnings- announcement stock returns on average earnings uncertainty measures over the sample period in a linear time trend model. We show that earnings volatility is the most important factor explaining the time-series trend of average stock price reactions around earnings announcements. We conclude our empirical analysis by examining whether the increased price reaction to earnings news is related to changes in investor behavior, particularly investor reactions to earnings news. Many academic studies have shown that, following earnings announcements, stock prices continue to drift in the direction of the earnings news (i.e., post-earnings announcement drift). One common interpretation of this phenomenon is investor underreaction to earnings information (see Bernard and Thomas (1990)). It is possible that the increased price reaction to earnings news could be a ‘correction’ of the underreaction phenomenon. If this is the case, we would expect that the post-earnings announcement drift would weaken (or even disappear) over time. We show, however, that the magnitude of the post-earnings announcement drift remains unchanged over the sample period. This suggests the up-ward trend in earnings-related volatility in the market is not related to changes in investor trading behavior, or arbitrage-trading activities related to earnings 4 announcements. This result is inconsistent with the claims made by corporate executives that investor reactions to quarterly earnings news are excessive and lends further support for the fundamentals-based explanation of the increased volatility around earnings announcements. In sum, we find that stock price reactions to earnings announcements have increased significantly over the last three decades. Consistent with previous studies, we also confirm that the relative importance of earnings announcements has deteriorated over time. We argue that the up- ward trend of earnings-related price volatility can be explained by the up-ward trend in the uncertainty of firm fundamentals. In a world of more uncertain fundamentals, prices will react more at information releases since information here would resolve more uncertainty. Our paper is closely related to the recent finding of Canpbell, Lettau, Malkiel, and Xu (2001) that the idiosyncratic volatility of stock returns has increased over time. The underlying reasons for such an increase are controversial. For example, Fink, Fink, Grullon, and Weston (2005) argue that the effect is driven by the increased propensity of firms to go public earlier in their life. Wei and Zhang (2005) explain the trend in idiosyncratic volatility with the recent increase in the volatility of corporate earnings. We show that volatility at earnings announcements follows strongly this upward trend. Since earnings announcement stock returns represent a significant fraction of aggregate stock return volatility, our results also shed some light on the increase of idiosyncratic volatility. The paper is organized as follows. Section I documents the time-series trends in earnings reactions. Section II discusses the basic hypotheses. Section III studies the time-series trends of earnings informativeness and firm fundamentals. Section IV presents cross-sectional and time- series analysis on the determinants of price reactions to earnings announcements; and Section V studies the time-series trends in the post-earnings announcement drift. We conclude in Section VI. 5 I. Time-series Trends in Earnings Reactions I.A. Data The sample includes stocks listed on the New York Stock Exchange (NYSE), American Stock Exchange (AMEX), and NASDAQ that appear on the CRSP and COMPUSTAT tapes. The sample period is from 1978 to 2004; it spans 27 years of the most resent stock market history. We exclude real estate investment trusts (REITs), American Depository Receipts (ADRs), and closed- end mutual funds. To mitigate microstructure effects, we also exclude stocks priced below $5 and stocks in the lowest market capitalization decile based on NYSE breakpoints as of the end of the previous calendar year.3 We obtain data on corporate earnings from Compustat quarterly files. Earnings are measured before extraordinary items and discontinued operations. We use the calendar year and quarter variable in COMPUSTAT to align quarterly earnings across firms. Table I presents the summary statistics for the sample. Our sample has 1,556 stocks in 1978, more than 3,000 stocks in the 90s, and 2,314 stocks at the end of 2004. The analyst coverage data is from I/B/E/S. I/B/E/S starts reporting analyst quarterly earnings forecast in 1984.4 The percentage of stocks with analyst coverage in the sample increases steadily over time. While in 1984, the fraction of our sample with analyst coverage was 50 percent, this fraction reaches 97 percent in 2004. The average number of analysts covering a stock also increases from around 4 in 1984 to close to 9 in 2004. The substantial increase in analyst activity over the last 30 years suggests that the overall amount of earnings-related information in the market has increased as well. 3 Another reason for removing the lowest decile firms from the sample is that Compustat earnings information is largely unavailable for these firms. 4 I/B/E/S begins reporting annual earning forecast in 1978, but analyst coverage in the late 70’s and early 80’s is generally much lower. Because we study price reactions to quarterly earnings announcements, we use analyst coverage for quarterly earnings throughout the paper. 6 Institutional holdings data originates from the CDA Spectrum 13F Filings database and it starts in 1980. Under the 1978 amendment to the Securities and Exchange Act of 1934, all institutional investors managing a portfolio with an investment value of $100 million or more are required to file quarterly 13F reports to the SEC, listing their equity positions greater than 10,000 shares or $200,000 in market value as of the last date of each quarter. For each stock in the sample, we calculate the level of institutional ownership using the percentage of shares owned by institutions relative to total shares outstanding at the end of each quarter. The institutional presence in the market has also increased substantially over time. While in 1980 the average institutional ownership in a stock was 21 percent, in 1980 the average institutional ownership in a stock was 60 percent. I.B. Time Trend of Stock Price Reactions to Earnings News We start our analysis by estimating the time-series properties of stock price reactions at earnings announcements. Our results in this section are consistent with results from previous studies. Our sample, however, significantly exceeds the samples in most of these studies. For instance, Landsman and Maydew (2002) conduct their study on 1000 randomly selected stocks, while Francis, Schipper, and Vincent (2002a)) examine only 400 large stocks. In addition, we also provide a more detailed look at the trend in earnings-announcement volatility across small and large stocks, different exchanges, and positive and negative reactions. We calculate earnings-announcement stock returns over a three day window (-1, 0, +1) centered at the earnings announcement date (date 0). Afterwards we subtract from the earnings- announcement return the corresponding three day return of a portfolio of stocks with similar market capitalization. The benchmark portfolios are based on NYSE market capitalization deciles 7 at the end of the previous year. Throughout the paper, we use this excess return to measure stock price reaction to earnings news. For each quarter we calculate equal- and value-weighted averages of the absolute earnings announcement excess returns. Panel A of Figure 1 presents the time series of the quarterly averages. We observe a strong upward trend in the average stock price reactions at earnings announcement dates. The average absolute stock return at earnings announcements has increased from 2.3 percent at the beginning of 1978 to almost 5 percent at the end of 2004. The equal- weighted results appear uniformly stronger than the value-weighted results, indicating a size effect in the reactions. Panel B of Figure 1 plots separately the average positive and negative earnings reactions. We observe that the magnitudes of the positive and negative returns around earnings announcements are remarkably close. Interestingly, stocks become much more sensitive to earnings announcements in periods of high market uncertainty, such as the market crash in 1987 and the burst of the internet bubble in 2000 and 2001. For example the average stock price reaction at the fourth quarter earnings of 1987 was 6.1 percent and at the fourth quarter earnings of 2000 as high as 8.3 percent. Note that these responses are not mechanically related to the crash. All price reactions are already adjusted for the market. The responses are symmetric – present for both positive and negative returns. Figure 2 reports the time series of average stock price reactions for five price reaction quintiles. Each quarter, we sort stocks into five quintiles based on the excess stock returns over the three day period (in this case, we do not use absolute values) and calculate the average stock price reaction for each quintile in each quarter. The chart provides information about the dynamics of the overall distribution of earnings reactions over time. The trend in the extreme reactions is well pronounced. For example the average return of the top (bottom) twenty percent of earnings 8 reactions was 4.4 (-4.4) percent in 1978 and it has increased (decreased) steadily over the sample period reaching 9.5 (-9.1) percent at the end of 2004. Figure 3 reports the time series of average stock price reactions for five size quintiles. On average, small stocks are associated with stronger earnings reactions than large stocks – the average earnings release returns for large stocks are 3.2 percent while for small stocks they are 6.1 percent. Interestingly, the discrepancy across size disappears in crisis periods (years 1987 and 2000) in which the earnings responses are equally strong across all stocks. The up-trend in earnings responses is well pronounced across all five quintiles. For example, the average absolute returns of the large-stock quintile increase from 1.7 percent in 1978 to 3.5 percent in 2004, while the returns of the small-stock quintile increase from 4.6 percent in 1978 to 6.4 percent in 2004. We also observe that the average stock price reactions at earnings announcement dates have increased across the two major exchanges NYSE and NASDAQ (Figure 4). While the trend for NASDAQ-traded stocks, it is also well-pronounced for NYSE-traded stocks. In Table II we examine the statistical significance of the time trend in earnings reactions by estimating the following linear time trend model: ERt bT t (1) We estimate the model in (1) for both equal-weighted and value-weighted average absolute price reactions (ERt). In addition, we examine the time trends of price reactions for various portfolios across negative and positive price reactions; extreme positive and negative price reactions; the smallest and largest market capitalization quintiles; and NYSE and NASDAQ stocks.5 All specifications detect a significant positive time trend in earnings reactions over time. 5 We have also estimated an alternative specifications of the model in (1), including the lagged value of price reactions to earnings announcements from the previous quarter and dummy variables for the two quarters with the largest price reactions. All results are qualitatively similar. 9 We have also conducted a unit root test for the same time series and we reject the null-hypothesis for the existence of a unit root in the data. Table III presents average earnings announcement returns for three subperiods: 1978-1986, 1987-1995, and 1996-2004 and tests the difference of the price reactions between the early and late subperiods. Consistent with the results from the charts and the time trend model, the upward trend in earnings-announcement returns is also well pronounced across subperiods. The last two rows of the table indicate that the differences across subperiods are statistically significant. Here, the trend is also symmetric and equally strong for both negative and positive reactions. Price reactions to earnings news are stronger for small stocks than for large stocks over the full sample period. While both small and large stocks experience significant increase in their price reactions around earnings announcements, the increase for large stocks is greater. II. Basic Hypotheses In this section we outline a parsimonious model about the stock price volatility at information releases. The objective of this model is to illustrate the importance of both the quality of the information signal and the uncertainty of underlying fundamentals for the reaction at the announcement. Earnings are a signal about future cash-flows. Consider a setup in which the current stock price, P , of a stock is equal to the discounted expected future cash-flows (dividends) x x2 P 1 ... (2) 1 r (1 r ) 2 Assume that all future dividends are independently and normally distributed random variables x ~ ( x , x ) . Under these assumptions, the price of the stock would be given by 2 10 E ( x) P (3) r Now consider an information signal (earnings announcement) about the distribution of future dividends in the form s x ~ ~ ~ , (4) ~ ~ where the noise of the signal is normally distributed as ~ (0, 2 ) . The volatility of the signal, 2 , reflects the noise level of the signal. Before the earnings announcement, the price of the stock is equal to the present value of x . After the release of the signal, ~ , the new price of the x 1 ~ expected future cash-flows P s r r stock is equal to the discounted expected future cash flows conditional on the new information, P ~ ~ , or: 1 xs r 1 x 2 P( s ) x 2 s x (5) r x 2 The variance of the stock price change at the information release is given by x 4 VarP( s) (6) r 2 ( x 2 ) 2 PROPOSITION 1: The volatility of the stock price at earnings announcements is given by (6) and it (i) Increases with the precision of the signal (decreases with the noise of the signal) (ii) Increases with the volatility of future cash flows (iii) Decreases with the level of discount rates 11 The proposition follows directly from (5). It is intuitively clear that the magnitude of stock price reaction at the earnings announcement would be proportional to the precision of the signal (or inversely related to the noise of the signal 2 ). More precise signals would be associated with stronger reactions and less precise signals with smaller reactions. When the signal is extremely noisy ( 2 is very large), the variance at the announcement would approach zero, indicating that in this case the information content of the announcement is extremely low. The stock price reaction at the announcement increases with the precision of the signal and takes the highest value when the volatility of the noise is equal to zero. In this case, the volatility of the price change will be equal to the volatility of the future cash flow x . 2 The second part of the proposition indicates that the volatility at the earnings announcement is proportional to the volatility of underlying cash-flows. The intuition for this result is that information about more uncertain fundamentals is more valuable. In the extreme case of deterministic cash-flows, information would have no value since it doesn’t add anything to what we already know. The third part of the proposition reflects changes in the discount rates. Overall, changes in discount rates are determined by changes in market risk premium, which in turn is determined by both the risk level of stock returns and the aggregate risk preference of investors. We do not elaborate on this variable since Vuolteenaho (2002) shows that changes in future cash flows rather than discount rates explain the majority of the variation in individual stock returns. We thus summarize the above results in the following two hypotheses: Hypotheses 1: Stock price reaction at earnings announcements is proportional to the precision of the information signal about future cash flows. Hypotheses 2: Stock price reaction at earnings announcements is proportional to the volatility of future cash flows. 12 III. Time Series Trends in Earnings-related Variables Guided by the predictions of the model from the previous section, here we explore why the magnitude of earnings announcement returns has increased over time. Is it because the information quality in earnings as a signal has increased or because the patterns of corporate cash flows have changed? We start our analysis by identifying a set of variables that could potentially explain the increase in stock return volatility around earnings announcements and study the time trends in these variables. A. Information precision In this subsection, we concentrate on variables measuring the noise in earnings information. The value of earnings information can be summarized by the level of ‘surprise’ of the earnings news relative to market expectations. We use the following surprise-variables: Standardized Unexpected Earnings (SUE). SUE is a commonly used earnings surprise measure based on the assumption that earnings follow a seasonal random walk process. To estimate SUE, we follow Foster, Olsen and Shevlin (1984) and assume that firm earnings follow a seasonal random walk. Under this assumption SUE is defined as the difference of current earnings and earnings from the same quarter of the previous year normalized with book value of equity. Analyst Forecast Error (AFE). Financial analyst earnings forecasts serve as a proxy for market expectations about future corporate earnings. AFE is defined as the difference between the actual quarterly earnings per share (EPS) and the most recent mean consensus analyst forecast for the quarter. The difference is then scaled by both the book value and the market value of equity per share at the end of the previous quarter.6 We use the unadjusted forecast data from I/B/E/S to 6 We report results based on the mean analyst forecast, results using median analyst forecast are not tabulated but exhibit similar patterns. 13 measure AFE. As discussed in Dieter, Malloy and Scherbina (2002), using the standard I/B/E/S dataset (i.e., after adjusted for stock splits) will lead to much smaller or the disappearance of analyst forecast surprises for stocks that have experienced many stock splits. The analyst forecast error has several advantages as an earnings surprise measure relative to alternative measures based on time series forecast of earnings, such as SUE. For example, AFE does not rely on a forecasting model and is based on a larger information set. This measure is also widely used in both the accounting and finance literature. In a recent study, Livnat and Mendenhall (2006) show that earnings surprises calculated from analyst and time series forecast (i.e., SUE) differ systematically and AFE provides a less noisy measure of earnings surprise. One shortcoming, though, is that I/B/E/S starts covering quarterly earnings forecasts in 1984. Also, the number of firms under coverage, particularly in the early period is small. The analyst forecast error measures the overall level of informativeness of the earnings announcement. Larger errors are associated with greater informational content of the earnings announcement, since in this case the announcement resolves higher uncertainty. Analyst Forecast Uncertainty (AFU). This is a measure of the differences of opinion among financial analysts about future earnings. We again obtain the standard deviations from the unadjusted earnings forecasts. We scale the standard deviations of earnings forecasts using market value of equity per share and book value of equity per share. Unlike AFE, AFU captures the uncertainty of the earnings forecast itself. According to Hypothesis 1, stock price volatility at earnings announcements is expected to be higher when the divergence of opinions about fundamentals is higher. Figure 5 plots the time-series of equal- and value-weighted standardized unexpected earnings (SUE). Here, we do not identify a strong trend in the earnings surprises over time. We note, however, that these surprises are based on a simple statistical model for forming expectations 14 about future earnings. In reality, investors might use a much wider set of information to predict earnings. Figure 6 presents the time-series of equal- and value-weighted analyst forecast errors and standard deviations. Panel A plots the average analyst forecast error, defined as the difference between the actual quarterly earnings per share and the most recent mean consensus analyst forecast for that quarter scaled by the stock price at the end of previous quarter. We report the time trends of analyst forecast error and analyst forecast dispersion scaled by book value of equity in Table IV. We observe that analyst forecast errors have decreased steadily over time – from 1.29 in 1978 to 0.28 in 2004. Such trend could reflect the recent advances in financial market towards greater information production and better corporate disclosure. Panel B of Figure 6 presents the time-series of equal- and value-weighted analyst forecast uncertainty, defined as the standard deviation of analyst earnings forecasts divided by the stock price. According to Hypothesis 1, the higher the divergence of investor opinion about fundamentals, the higher the sensitivity of prices to new information would be. Surprisingly, we find that the divergence of analyst expectations has decreased over time – from 0.56 percent in 1978 to 0.11 percent in 2004. Not only are analysts now able to predict earnings announcement outcomes more accurately than thirty years ago, but the level of disagreement among them has also decreased steadily over time. One could argue that the downward trend in AFE and AFU (Figure 6) variables is driven by the fact that stock prices have increased over the sample period, especially during the late 90s. As a result, we might be simply rediscovering the trend towards higher valuations. This, however, is not the case. When recalculating the same ratios normalized with book value of equity instead of market value of equity, we find that that the downward trend here is also very well pronounced. For 15 example, analyst forecast errors scaled by book value of equity have also decreased steadily over time – from 1.47 in 1978 to 0.76 in 2004. Dispersion of analyst forecast shows similar trend (Table IV). Results based on analyst earnings forecasts show clearly that the quality of market predictions about future earnings has improved considerably over the sample period and this seems inconsistent with the upward trend in earnings announcement stock return volatility. The increased ability of the market to predict earnings implies that the usefulness of the earnings as a signal has deteriorated over time. Consider, for instance, the case of perfect foresight in which analysts are able to predict the exact outcome of future earnings announcements. In this case investors’ attention would shift entirely from corporate announcements to analyst announcements. Earnings announcements would become a non-event and there will be no reaction at the announcement at all. This does not imply that earnings information per se has become less important over time. It implies that the information has advanced over time and investors learn about it earlier7. As a consequence, changing quality of earnings announcement information seems an unlikely explanation of the upward trend in earnings announcement volatility. B. Firm Fundamentals In this subsection, we identify a set of variables related to the uncertainty of firm fundamentals. Return on Equity (ROE). Return on equity is the ratio of current earnings relative to book value of equity. It is a measure of firm profitability. Companies with high return on equity are more likely to be more mature firms with low growth rates of future cash flows. As a result, we expect 7 Francis, Schipper, and Vincent (2002b) directly examine whether investors’ use of analyst forecasts substitutes for their use of earnings announcements information. They find that market reacts to information in analyst reports but the informativeness of earnings announcements as measured by stock price reactions to the announcements is not eroded by competing information in the form of analyst reports. 16 these companies to be associated with lower return volatility around earnings announcements. More important, we use ROE to construct a direct measure of earnings uncertainty as we discuss below. Volatility of Return on Equity (VROE). This is the standard deviation of quarterly ROE calculated from the previous five years of data. We delete estimates based on fewer than 12 observations. We use this variable as the measure of the uncertainty of firm fundamentals. According to Hypothesis 2, the magnitude of stock returns at earnings announcements is positively related to the volatility of fundamentals. Earnings Yield (EY). Earnings yield is the ratio of current earnings relative to market value of equity and is the inverse of the price-earnings ratio. Unlike ROE, which measures corporate profitability, earnings yield serves as a measure of market valuations of corporate earnings. To certain extent, EY is related to the discount rate specified in Equation (6). Panel A of Figure 7 plots the time-series of equal- and value-weighted averages of return on equity. We observe that ROE decreases over time but only for equal-weighted averages which indicates that the decrease in profitability was concentrated in smaller stocks. The equal-weighted average reaches the lowest level at the end of 2000 when it becomes negative. Value-weighted ROE does not show significant time trend, suggesting that level of overall corporate profitability, especially for large stocks, remains unchanged over the sample period. Panel B of Figure 7 plots the time-series of equal- and value-weighted earnings yields in the market. Earnings yields have declined significantly over time – from 3.2 in 1978 to 1.1 in 2004. Since earnings yields are inversely related to price/earnings ratios, this trend also represents a steady increase in the price-earnings ratios over the last 27 years. The time trend of earnings yields, combined with the relative stability of ROE, indicates that the increase in the valuation ratios over 17 the sample period is largely driven by increased market valuations rather than decreased firm profitability. Figure 8 presents the time-series of equal- and value-weighted averages of the standard deviation of ROE calculated based on the previous five years of data. We observe that the volatility of corporate earnings increases over time. In 1978, it was 1.86 percent and in 2004 it was 4.7 percent. The value-weighted volatility line lies below the equal-weighted line, indicating that the fundamentals of large stock are less volatile then the fundamentals of small stocks. Despite the lower magnitude, the up-ward trend here is even stronger – the value-weighted volatility increased from 1.15 percent in 1978 to 3.57 percent in 2004. Valuation ratios and earnings volatility span two different dimensions of uncertainty of future cash-flows. Higher valuation ratios, on one hand, indicate that today the market capitalizes long-term growth opportunities more heavily than it used to three decades ago. Earnings volatility, on the other hand, indicates that not only long-term fundamentals but also short-term cash-flows have become more uncertain. These results are consistent with the observed up-ward trend in stock price reactions around earnings announcements. As summarized in Hypothesis 2 of the previous section, more uncertain fundamentals would make prices more responsive to information releases. Table IV represents a different way of looking at the data. It summarizes the analyst forecast error, analyst forecast uncertainty, the standard deviation of ROE, and earnings yield for three subperiods: 1978-1986, 1987-1995, and 1996-2004. Each one of the above four variables is normalized with both book value of equity and market value of equity. The last two rows test the statistical significance of the difference in means between the first and the last subperiods. The subperiod results are consistent with the results from the charts. They confirm that analyst forecast error, analyst forecast uncertainty, and earnings yields have decreased over time, while earnings 18 volatility has increased. Note that ratios based on market prices are usually more volatile than ratios based on book values since market values are more volatile than book values. In sum, the preliminary evidence on the increased volatility around earnings announcements is inconsistent with the information precision stories since the relative importance of earnings information is deteriorating over time. The upward trend in earnings announcement stock return volatility is more consistent with the changing fundamentals stories. In the next section, we directly test the relation between firm fundamentals and the price reactions to earnings announcements. IV. Firm Fundamentals and Price Reaction to Earnings News In this section we study the importance of firm fundamentals in explaining the time trend of stock return volatility around earnings announcements. IV.A. Cross sectional analysis We first implement cross-sectional analysis of the relation between price reactions at earnings announcements and various earnings and firm characteristics. Given the high correlation between earnings yield and ROE, we include only ROE in the analysis. In addition to the earnings variables introduced in the previous section, we include the following control variables: Size. We measure stock size with the market value of firm equity. Firm size, in logarithm, is sampled at the end of the previous year. Firm size is related to the level of information asymmetry and is positively correlated with both analyst coverage and institutional investor holdings. As we show earlier, firm size is related to both earnings surprises and return volatility around earnings announcements. 19 Book-to-market Ratio (BE/ME). Stock’s book-to-market ratio is defined as the ratio of book value of equity over market value of equity at the end of the previous year. It is a relative valuation ratio for firm’s equity and it reflects the present value of growth opportunities. We use book-to-market as a valuation metric in the cross-sectional regressions since it is less noisy relative to alternative metrics, such as price-earnings ratios. The sensitivity of stock prices to new information is stronger when the growth rate of future cash-flows is higher. Since long-term growth rates are difficult to measure directly, book to market ratios serve as a proxy for growth opportunities – low book-to-market ratios (high valuations) indicate high growth rates; high book- to-market ratios (low valuations) indicate low growth rates. Leverage. We measure leverage using the ratio of total debt to total assets. Firm leverage is related to both the volatility of earnings and overall firm risk. Age. We measure the age of firms in our sample using the number of years the firm has been in the CRSP database. Firm age is related to maturity and earnings volatility of the firm. Firm age could also be related to the level of information asymmetry of the firm. We use logarithm of age in the regression. Institutional Holdings. The level of institutional ownership is calculated as the percentage of shares owned by institutions relative to total number of shares outstanding at the end of each quarter. There are two general ways through which institutional investors could have an impact on the way prices react to earnings information. On the one hand, if institutional investors are better informed, then they should be less surprised by the announcement. As a result, the presence of institutional investors in a stock would be associated with smaller stock price reactions on earnings announcement days8. On the other hand, if institutions are more likely to respond to information 8 Consistent with the information story, a variety of empirical studies provide evidence that institutions posses better information and information-processing skills than individual investors (see Grinblatt and Titman (1989, 1993), Daniel, Grinblatt, Titman, and Wermers (1997), Wermers (1999, 2000), and Bennett, Sias, and Starks (2003)). 20 than individual investors, stocks with higher institutional ownership could react more strongly to earnings information. In Table V we estimate the cross-sectional regressions for each quarter. The table reports time series averages of the coefficient estimates and t-statistics for the three subperiods: 1978- 1986, 1987-1995, and 1996-2004. In Model 1 we use SUE as a measure of earnings surprise. In Model 2, we substitute this variable with analyst-related variables and add institutional ownership. We note that the results here are not simply a multivariate extension of the univariate results form the previous section. In the previous section, we looked at the time-series trends in the explanatory variables – here we study the cross-sectional sensitivity of absolute earnings returns to these variables, particularly the earnings information variables and measures of uncertainty of firm fundamentals. This cross-sectional analysis provides additional insight on the trend in earnings announcement returns, since a variable might contribute to the trend in two separate ways – through changes in its levels or through changes in its relative importance over time. The cross-sectional results confirm that stocks with more volatile earnings (higher volatility of return on equity) have stronger reactions at earnings announcements. The sensitivity of earnings announcement stock returns to earnings volatility has also increased over time in both models. The increasing significance of VROE in explaining the cross-sectional variation of earnings-announcement volatility and the fact that VROE has indeed increased strongly over time indicate that the uncertainty of firm fundamentals plays an important role in explaining the up-ward trend in stock price reactions around earnings announcements. Price reactions to earnings announcements are also positively related to earnings surprises - both SUE and AFE are significant in the regressions. Interestingly, the importance of earnings surprises (AFE) in the cross-sectional regressions increases over time, wile at the same time, the level of earnings surprises decreases. In other words, missing an analyst expectation would result in 21 a larger stock price reaction today than thirty years ago. At the same time, however, the chance of missing an expectation today is much smaller than it used to be thirty years ago. The dispersion of financial analyst forecasts (AFU) does not significantly relate to the magnitude of the price reactions at earnings announcements. The other control variables are also significantly related to price reactions at earnings announcements. Size is negatively related to earnings announcement returns. This is consistent with Figure 4, indicating that small stocks react stronger to earnings announcements. Firms with higher book-to-market ratios have weaker earnings announcement reactions. This is consistent with the interpretation that high book-to-market firms have lower valuation ratios and as a result lower future growth opportunities. Both leverages and firm age are negatively related to the price volatility at earnings announcements. The result for firm age is not surprising, since more mature firms are generally perceived as being less risky. The causality for leverage could go both ways. On one hand, leverage could increase stock price volatility since it increase the overall risk level of the firm. On the other hand, the leverage ratio could be endogenously determined by earnings volatility and firm risks. As a result, less risky firms would have higher leverage ratios. Less profitable firms experience stronger price reactions at earnings announcements. The economic and statistical significance of ROE in the cross-sectional regression also increases over the sample period. The profitability of the average firm, however, has decreases over time. As a result, the net effect of ROE on the return volatility around earnings announcements is unclear. Lastly, the institutional ownership variable is positively related to the level of price reactions. IV. B. The time series evidence 22 In this subsection we examine directly the link between fundamentals uncertainty and the upward trend in earnings announcement volatility by including earnings volatility as an independent variable in the linear time trend model of equation (1): ERt b1T b2VROE t t (7) We estimate the model in (7) for both equal-weighted and value-weighted average price reactions at earnings announcements along with equal-weighted and value-weighted earnings volatility. Table VI shows that average earnings volatility in the market explains the time trend of average price reactions at earnings announcements. In both the equal-weighted and value-weighted regressions, the time trend variable becomes insignificant, while the earnings volatility variables remain highly significant. The above results are robust to various specifications that include lagged values of price reactions and earning volatility. When substituting the volatility of return on equity with other variables, such as profitability (ROE), earnings yields, and various measures of earnings surprises and forecast uncertainty, we show that these variables are unable to explain the time trend of return volatility around earnings announcements. V. Changes of Price Reactions and Post-earnings Announcement Drift The magnitude of the stock price reactions at earnings announcements has increased over time. We argue that this up-ward trend in return volatility around earnings announcements is due to the fact that the fundamentals of the typical firm in the market have become more uncertain over time. Many corporate executives and regulators, however, take a different stand on the issue. They argue that stock price reactions at earnings announcements are excessive (see the Introduction) and can not be justified by fundamentals. 23 In this subsection we shed further light on this question by looking at the stock price behavior following the announcement. Many academic studies have documented robust evidence of post-earnings announcement drift and one common interpretation of this phenomenon is underreaction to earnings information (see Bernard and Thomas (1989)). If investors now overreact to earnings information, we would expect that the post-earnings announcement drift would weaken (or even disappear) over time. In Table VII we sort stocks into quintiles according to their earnings surprises and earnings announcement returns and calculate abnormal returns for the most positive (first panel) and most negative (last panel) quintiles over 5-, 10-, 20-, and 60-day periods after earnings announcements. We calculate the abnormal return based on the difference between the stock’s cumulative return over the period and the cumulative return of its corresponding size decile portfolio for the same period. We calculate average abnormal stock returns for the three subperiods: 1978-1986, 1987- 1995, and 1996-2004. Panel A reports results for earnings surprise portfolios sorted on AFE. Consistent with the earlier results, the price reactions at the earnings announcement for both the “Most Negative” and “Most Positive” earnings surprise portfolios are increasing over time. The magnitude of the post- earnings announcement drift for these portfolios, however, remains substantially unchanged over the sample period. Panel B reports results for earnings surprise portfolios based on the actual price reactions at the announcement. Compared with the stock return results in Panel A, there are some noticeable differences. Stock price drifts following the most negative stock price reactions here are markedly stronger for the latest period. Abnormal 60 day-returns for the latest period are -2.48%, while for the first period they are -0.97%. More interestingly, there is a consistent price continuation 24 following negative price reactions in the latest period over short holding periods, while stock prices experience even short term reversals following negative price reactions in the earlier period. In sum, even though price reactions around earnings announcements have increased over time, post-earnings announcement price drifts remain significant. In fact, the magnitude of the drift has even increased following extreme negative earnings news. This suggests that the market does not overreact to earnings information and lends further support to the fundamentals-explanation of the increasing earnings-related volatility. VI. Conclusion Stock price reactions at earnings announcements have increased substantially over the last three decades. At the same time the relative importance of earnings has actually decreased over time. The last few decades have seen tremendous growth of the security analysis industry and numerous financial innovations, all of which facilitate the extraction of information from firms prior to earnings announcements. As a result, one would expect stock price volatility at earnings announcement to decrease over time. To reconcile these two findings we propose an explanation based on the fact the uncertainty of firm fundamentals has increased over time. In a world of more uncertain fundamentals, prices will react more at information releases since information here would resolve more uncertainty. We confirm this hypothesis by showing that measures of the average cash-flow uncertainty in the market can explain the up-ward trend in stock price reactions at earnings announcements. Our analysis suggests that the claim often made by corporate executives that investors react too aggressively to earning news is unfounded. Our findings also imply that, given the change of firm fundamentals, increasing the frequency of information disclosure may not be helpful in 25 reducing the magnitude of stock price reactions to information. If the increased volatility in markets reflects changes in fundamental variables, then this volatility, while undesirable, may be unavoidable. 26 References Bennett, J., R. Sias, and L. Starks, 2003, Greener pastures and the impact of dynamic institutional preferences, Review of Financial Studies 16, 1203-1238. Bernard, V.L., Thomas, J.K., 1989. Post-earnings-announcement drift, delayed price response or risk premium? Journal of Accounting Research 27, 1–36. Bernard, V.L., Thomas, J.K., 1990. Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics 13, 305–340. Collins, D., O. Li, and H. Xie, 2005, What drives the increased informativeness of earnings announcements over time? Working paper, University of Iowa. Daniel, Kent, Mark Grinblatt, Sheridan Titman, and Russ Wermers, 1997, Measuring mutual fund performance with characteristic based benchmarks, Journal of Finance 52, 1035-1058. Dieter, K., C. Malloy, and A. Scherbina, 2002, Differences of opinion and the cross-section of stock returns, Journal of Finance 57, 2113-2141. John Y. Campbell, Martin Lettau, Burton G. Malkiel, and Yexiao Xu, 2001, Have individual stocks become more volatile? An empirical exploration of idiosyncratic risk, Journal of Finance 56, 1-43. Ely, K., and G. Waymire, 1999, Accounting standard setting organizations and earnings relevance: longitudinal evidence from NYSE common stocks, 1927-93, Journal of Accounting Research 37, 293-317. Fink, Jason, Kristin Fink, Gustavo Grullon, and James P. Weston, 2005, IPO vintage and the rise of idiosyncratic risk, Working paper, Rice University. Foster, G., C. Olsen, and T. Shevlin, 1984, Earnings releases, anomalies, and the behavior of securities returns, The Accounting Review 59, 574-603. Francis, Jennifer, and Katherine Schipper, 1999, Have financial statements lost their relevance? Journal of Accounting Research 37, 319-352. Francis, Jennifer, Katherine Schipper, and Linda Vincent, 2002a, Expanded disclosures and the increased usefulness of earnings announcements, The Accounting Review 77, 515-546 Francis, Jennifer, Katherine Schipper, and Linda Vincent, 2002b, Earnings announcements and competing information, Journal of Accounting and Economics 33, 313-342 Grinblatt, M. and S. Titman, 1989, Portfolio performance evaluation: Old issues and new insights, Review of Financial Studies 2, 393-422. Grinblatt, M. and S. Titman, 1993, Performance measurement without benchmarks: An examination of mutual fund returns, Journal of Business 66, 47-68. 27 Kim, M. and W. Kross, 2005, The ability of earnings to predict future operating cash flows has been increasing-not decreasing, Journal of Accounting Research 43, 751-780. Landsman, A., and Maydew, 2001, Beaver (1968) revisited: Has the information content of quarterly earnings announcements declined in the past three decades? Working paper, University of North Carolina Chapel Hill Lev, Baruch, and P. Zarowin, 1999, The boundaries of financial reporting and how to extend them, Journal of Accounting Research 37, 353-385. Vuolteenaho, Tuomo, 2002, What drives firm-level stock returns? Journal of Finance 57, 233-264. Fama, E.F., MacBeth, J.D., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political Economy 81, 607–636. Rendleman, R.J., Jones, C.P., Latane, H.A., 1982. Empirical anomalies based on unexpected earnings and the importance of risk adjustment. Journal of Financial Economics 10, 269–287. Ryan, S.G., and P.A. Zarowin, 2003, Why has the contemporaneous linear returns-earnings relation declined?, Accounting Review 78, 523-553. Wei, Steven X., and Chu Zhang, 2005, Why did individual stocks become more volatile?, Journal of Business, forthcoming. Wermers, R., 1999, Mutual fund trading and the impact on stock prices, Journal of Finance 54, 581-622. Wermers, R., 2000, Mutual fund performance: An empirical decomposition into stock-picking talent, style, transaction costs, and expenses, Journal of Finance 55, 1655-1695. Wurgler, J., 2000, Financial markets and the allocation of capital, Journal of Financial Economics 58, 187-214. 28 Table I Summary Statistics Presented are the total number of stocks, the percentage of stocks with analyst coverage, the average number of analysts per stock, the percentage of stocks with positive institutional ownership, and the average level of institutional ownership relative to total number of shares outstanding. The sample includes domestic common stocks traded on NYSE, AMEX, and NASDAQ from 1978 to 2004. We exclude REITs, closed-end funds, ADRs, stocks priced below $5, and stocks in the lowest market capitalization decile based on NYSE break points as of the end of the previous year. Percentage of Stocks with Average Average Level of Number of Analyst Number of Institutional Year Stocks Coverage Analysts Ownership 1978 1556 --- --- --- 1980 1544 --- --- 0.21 1983 1900 --- --- 0.28 1986 2003 0.60 4.4 0.31 1989 2019 0.83 5.6 0.36 1992 2371 0.92 5.5 0.41 1995 3069 0.91 5.4 0.44 1998 3112 0.94 6.3 0.48 2001 2768 0.88 7.3 0.53 2004 2314 0.97 8.6 0.60 29 Table II Linear Trend in Earnings Announcement Stock Returns We calculate earnings-announcement stock returns as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. The benchmark portfolios are based on NYSE market capitalization deciles as of the end of the previous year. Afterwards, we calculate equal- and value-weighted averages of the absolute value of the earnings announcement stock returns for the whole market and for the portfolios of positive and negative price reactions; extreme positive and extreme negative price reactions; the quintiles of the smallest and largest stocks; and NYSE and Nasdaq stocks. The table estimates time series regressions of average portfolio earnings reactions on a linear time-trend variable. Intercept T-stat. Time T-stat. Equally-weighted 2.55 16.58 0.031 12.65 Value-weighted 1.93 12.37 0.023 9.22 Negative Reactions 2.43 15.35 0.0324 12.87 Positive reactions 2.68 15.76 0.0297 10.95 Most negative 4.58 15.39 0.0629 13.29 Most positive 4.94 15.44 0.0618 12.12 Smallest quintile 4.47 26.76 0.0309 11.6 Largest quintile 1.77 9.07 0.0264 8.51 NYSE stocks 2.57 19.54 0.0187 8.93 Nasdaq stocks 2.35 13.19 0.0459 16.22 30 Table III Average Stock Returns at Earnings Announcements We calculate earnings-announcement stock returns as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. The benchmark portfolios are based on NYSE market capitalization deciles as of the end of the previous year. Then, for each quarter, we construct equal- and value-weighted averages of absolute earnings announcement returns (first two columns), average positive and negative announcement returns (next two columns), and average announcement returns for quintile portfolios of small and large stocks (last two columns). The table reports means and medians of the above time series for three subperiods: 1978-1986, 1987-1995, and 1996-2004. The last two rows test the statistical significance of the difference in means between the first and the last subperiods. EW VW Negative Positive Small Stocks Large Stocks Period 1978-1986 Mean 0.032 0.025 -0.060 0.062 0.039 0.024 Median 0.022 0.018 -0.049 0.051 0.027 0.018 Period 1987-1995 Mean 0.040 0.029 -0.076 0.078 0.049 0.028 Median 0.026 0.020 -0.059 0.064 0.034 0.019 Period 1996-2004 Mean 0.056 0.044 -0.106 0.111 0.059 0.044 Median 0.035 0.029 -0.082 0.089 0.037 0.029 Difference b/w first and last period Mean 0.024 0.019 -0.047 0.049 0.020 0.020 T-stat. 84.28 49.95 -73.20 82.41 33.96 41.72 31 Table IV Average Earnings Variables across Sub-periods For each quarter, we construct averages of the following four variables: analyst forecast error, defined as the difference between the actual quarterly earnings per share (EPS) and the most recent mean consensus analyst forecast for that quarter; analyst forecast uncertainty, defined as the standard deviations of unadjusted earnings forecasts; earnings volatility, defined as the standard deviation of ROE calculated from the previous five years of data; and earnings yield, defined as the scaled current earnings. We normalize each one of the above four variables with both book value of equity and market value of equity. The table reports means and medians of the above time series for three subperiods: 1978-1986, 1987-1995, and 1996-2004. The last two rows test the statistical significance of the difference in means between the first and the last subperiods. Analyst Forecast Error AFU Earnings Volatility Earnings Yield BE ME BE ME BE ME BE ME Period 1978-1986 Mean 1.397 1.003 0.538 0.380 2.353 2.243 3.142 2.289 Median 0.500 0.300 0.300 0.200 1.435 1.339 3.550 2.318 Period 1987-1995 Mean 1.016 0.576 0.500 0.257 3.986 2.394 2.306 1.064 Median 0.400 0.200 0.300 0.100 2.227 1.236 3.170 1.487 Period 1996-2004 Mean 0.843 0.313 0.423 0.140 4.813 2.229 1.402 0.458 Median 0.300 0.100 0.200 0.100 2.594 1.143 2.870 1.134 Difference b/w first and last period Mean -0.550 -0.690 -0.120 -0.240 2.500 -0.010 -1.740 -1.830 T-stat. -18.06 -27.30 -9.27 -23.20 113.19 -0.88 -49.45 -88.51 32 Table V Cross-sectional Determinant of Earnings Announcement Returns First, for each quarter, we regress the absolute value of the three-day earnings announcement stock return on the following set of independent variables: the logarithm of the market capitalization of the stock at the end of the previous year, the book-to-market equity ratio at the end of the previous year (BE / ME); leverage, defined as ratio of total debt to assets; Age, defined as the logarithm of the number of years the stock has been in the CRSP database; the ratio of current earnings and book value of equity (ROE); the standard deviation of ROE calculated based on the previous five years of data (VROE); standardized unexpected earnings (SUE), defined as the difference between current earnings and earnings form the same quarter of the previous year normalized with book value of equity; analyst forecast error (AFE), defined as the difference between the actual quarterly earnings per share and the most recent mean consensus analyst forecast for that quarter scaled by the stock price at the end of previous quarter; analyst forecast uncertainty (AFU), defined as the standard deviation of analyst earnings forecast divided by the stock price; and the fraction of shares held by institutions at the end of the previous quarter. Afterwards, we estimate time-series averages of the estimated coefficients for three subperiods: 1978-1986, 1987-1995, and 1996-2004. 33 (1) (2) 78-86 87-95 96-04 78-86 87-95 96-04 Intercept 0.0529 0.0751 0.0879 0.0706 0.0760 0.0880 44.58 39.66 36.63 8.58 62.16 45.14 Log Size -0.0025 -0.0025 -0.0012 -0.0033 -0.0032 -0.0027 -11.73 -17.40 -3.55 -2.35 -27.91 -13.92 BE / ME -0.0049 -0.0068 -0.0106 -0.0057 -0.0082 -0.0116 -10.29 -8.29 -7.69 -3.65 -12.12 -11.60 Leverage 0.0004 -0.0085 -0.0165 -0.0028 -0.0101 -0.0225 0.32 -6.52 -9.56 -1.53 -8.77 -15.84 Age -0.0020 -0.0069 -0.0093 -0.0053 -0.0069 -0.0089 -4.45 -19.60 -16.81 -8.57 -24.90 -23.18 ROE -0.0178 -0.0111 -0.0285 -0.1028 -0.0246 -0.0302 -3.28 -2.73 -5.43 -2.11 -6.45 -5.36 VROE 0.0796 0.0520 0.1033 0.0220 0.0595 0.1077 8.72 9.94 11.91 0.49 16.30 14.44 SUE 0.1132 0.0924 0.1304 11.10 9.24 4.80 AFE 0.1702 0.2999 0.8752 3.35 9.40 8.34 AFU 0.0440 0.1165 0.0249 0.31 1.83 0.15 Inst. Holding 0.0051 0.0125 0.0212 1.57 14.42 17.18 Adj. R-square 0.059 0.072 0.077 0.078 0.095 0.092 34 Table VI Linear Trends in Earnings Announcement Stock Returns We calculate earnings-announcement stock returns as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. The benchmark portfolios are based on NYSE market capitalization deciles as of the end of the previous year. Afterwards, we calculate equal- and value-weighted averages of the absolute value of the earnings announcement stock returns for each quarter. The table estimates time series regressions of absolute earnings returns on a linear time-trend variable (first two columns) and on a linear time-trend variable and lagged average earnings announcement returns (last two columns). Equally-weighted Value-weighted Coefficient T-stat. Coefficient T-stat. Alpha 0.0137 3.08 0.0082 1.88 Time 0.0001 1.15 -0.0001 -0.47 VROE 0.6433 2.83 0.9949 2.72 R-square 0.6296 0.4815 35 Table VII Post Earnings Announcements Drifts across Sub-periods For each quarter, we sort earnings surprises based on analyst forecast error and price reaction to earnings announcements into five equal sized portfolios. Analyst forecast error is defined as the difference between the actual quarterly earnings per share (EPS) and the most recent mean consensus analyst forecast for that quarter normalized by book value of equity; price reaction to earnings-announcement stock returns is defined as the cumulative return over a three day window centered at the earnings announcement date net of the cumulative stock return of a benchmark portfolio of stocks with similar market capitalization. We calculate size adjusted cumulative stock returns for the most negative and most positive earnings surprise portfolios over 5, 10, 20 and 60 trading days after the three-day announcement period for three subperiods: 1978-1986, 1987-1995, and 1996-2004. Panel A reports results based on earnings surprise ranking and Panel B reports results based on price reactions ranking. Panel A: Based on earnings surprise ranking period reaction 5 day 10 day 20 day 60 day Most Negative 1978-1986 -1.16** 0.21 0.08 0.02 -1.32** 1987-1995 -1.90** 0.00 0.01 -0.22** -2.27** 1996-2004 -2.61** -0.31** -0.32** -0.30** -1.86** Most Positive 1978-1986 1.59** 0.20** 0.72** 1.09** 1.91** 1987-1995 2.34** -0.03 0.45** 1.13** 2.55** 1996-2004 2.85** 0.11** 0.40** 0.97** 1.28** Panel B: Based on price reaction ranking period reaction 5 day 10 day 20 day 60 day Most Negative 1978-1986 -5.99** 0.24** 0.42** 0.31** -0.97** 1987-1995 -7.57** 0.11 0.18 0.06 -1.41** 1996-2004 -10.64** -0.43** -0.58** -0.57** -2.48** Most Positive 1978-1986 6.20** 0.12 0.31** 0.50** 0.75** 1987-1995 7.85** -0.12 0.09 0.61** 1.33** 1996-2004 11.08** -0.03 0.23** 0.71** 0.73** ** denotes significance at 5 percent level. 36 EW VW Negative Positive 10.0% 9.0% 9.0% 8.0% 8.0% 7.0% 7.0% 6.0% 6.0% 5.0% 5.0% 4.0% 4.0% 3.0% 3.0% 2.0% 2.0% 1.0% 1.0% 0.0% 0.0% 78 79 80 81 83 84 85 86 88 89 90 91 93 94 95 96 98 99 00 01 03 04 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 Panel A. Absolute value Panel B. Positive vs. Negative Figure 1. Average Earnings Announcement Stock Returns. Earnings-announcement stock returns are calculated as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. Panel A plots the time-series of equal- and value-weighted averages of absolute earnings announcement returns. Panel B plots the time series of the average positive and negative announcement returns. Q1 (Negative) Q2 Q3 Q4 Q5 (Positive) 20.0% 15.0% 10.0% 5.0% 0.0% 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 -5.0% -10.0% -15.0% -20.0% Figure 2. Average Earnings Announcement Stock Returns Across Return Quintiles. Earnings- announcement stock returns are calculated as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. The figure plots the time series of average earnings announcement returns across quintile portfolios stratified on the magnitude of the return. 37 12.0% Q1 (Small) Q2 Q3 Q4 Q5 (Large) 10.0% 8.0% 6.0% 1 4.0% 2.0% 0.0% 1978 1979 1980 1981 1983 1984 1985 1986 1988 1989 1990 1991 1993 1994 1995 1996 1998 1999 2000 2001 2003 2004 Figure 3. Average Earnings Announcement Stock Returns Across Size Quintiles. Earnings- announcement stock returns are calculated as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. The figure plots the time series of average earnings announcement returns across size-quintile portfolios. NYSE NASDAQ 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% 1978 1979 1981 1982 1984 1985 1987 1988 1990 1991 1993 1994 1996 1997 1999 2000 2002 2003 Figure 4. Average Earnings Announcement Stock Returns: NYSE vs. NASDAQ. Earnings- announcement stock returns are calculated as the cumulative return of the stock over a three day window centered at the earnings announcement date net of the cumulative return of a benchmark portfolio of stocks with similar market capitalization. The figure plots the time-series of average earnings announcement returns across NYSE and NASDAQ. 38 SUE ew SUE vw 2.50 2.00 1.50 1.00 0.50 0.00 1978 1979 1981 1982 1984 1985 1987 1988 1990 1991 1993 1994 1996 1997 1999 2000 2002 2003 Figure 5. Standardized Unexpected Earnings. Presented are the time-series of equal- and value-weighted standardized unexpected earnings (SUE), defined as the difference between current earnings and earnings form the same quarter of the previous year normalized with book value of equity. AFE ew AFE vw AFU ew AFU vw 1.40 0.60 1.20 0.50 1.00 0.40 0.80 0.30 0.60 0.40 0.20 0.20 0.10 0.00 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 0.00 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Panel A. Analyst Forecast Error Panel B. Analyst Forecast Uncertainty Figure 6. Analyst Forecasts. Panel A presents the time-series of equal- and value-weighted analyst forecast errors, defined as the difference between the actual quarterly earnings per share and the most recent mean consensus analyst forecast for that quarter scaled by the stock price at the end of previous quarter. Panel B presents the time-series of equal- and value-weighted analyst forecast uncertainty, defined as the standard deviation of analyst earnings forecasts divided by the stock price. 39 EY ew EY vw ROE ew ROE vw 0.05 7.00% 6.00% 0.04 5.00% 0.03 4.00% 0.02 3.00% 2.00% 0.01 1.00% 0 1978 1979 1981 1982 1984 1985 1987 1988 1990 1991 1993 1994 1996 1997 1999 2000 2002 2003 0.00% 1978 1979 1981 1982 1984 1985 1987 1988 1990 1991 1993 1994 1996 1997 1999 2000 2002 2003 -0.01 -1.00% -2.00% -0.02 Panel A. ROE Panel B. Earnings Yield Figure 7. Time Trend of ROE and Earnings Yield. Panel A plots the time-series of equal- and value-weighted averages of the ratio of current earnings over book value of equity. Panel B Presented are the time-series of equal- and value-weighted yearnings yield averages in the market, defined as the ratio of current earnings and market value of equity. VROE ew VROE vw 7.00% 6.00% 5.00% 4.00% 3.00% 2.00% 1.00% 0.00% 1978 1979 1980 1981 1983 1984 1985 1986 1988 1989 1990 1991 1993 1994 1995 1996 1998 1999 2000 2001 2003 2004 Figure 8. Volatility of ROE. plots the time-series of equal- and value-weighted averages of the standard deviation of ROE calculated based on the previous five years of data. 40

DOCUMENT INFO

Shared By:

Categories:

Tags:

Stats:

views: | 3 |

posted: | 7/13/2012 |

language: | |

pages: | 42 |

How are you planning on using Docstoc?
BUSINESS
PERSONAL

By registering with docstoc.com you agree to our
privacy policy and
terms of service, and to receive content and offer notifications.

Docstoc is the premier online destination to start and grow small businesses. It hosts the best quality and widest selection of professional documents (over 20 million) and resources including expert videos, articles and productivity tools to make every small business better.

Search or Browse for any specific document or resource you need for your business. Or explore our curated resources for Starting a Business, Growing a Business or for Professional Development.

Feel free to Contact Us with any questions you might have.