Short term market reaction to earnings restatements: value stocks vis-a-vis glamour stocks.

Xu, Tan ; Li, Diane ; Jin, John Jongdae 等

INTRODUCTION

It has long been reported that value (or contrarian) strategies outperform the market since Graham and Dodd (1934), which is counter evidence to the efficient market hypothesis. This strategy calls for buying value stocks and selling glamour stocks. The value stocks are under-priced stocks relative to their intrinsic value indicators such as book value, earnings, cash flows, growth rate, while glamour stocks are over-price stocks relative to their intrinsic value indicators (E.g., Lakonishok et al., 1994).

Various hypotheses have been proposed to explain why the return differential between value stocks and glamour stocks persists so long. For examples, Fama and French (1992, 1993, 1995, 1996) argue that value stocks are judged by the market to have poor earnings prospect and higher risks and, thus, are selling at lower prices relative to their book value (i.e., high BM ratio), while the opposite applies to glamour stocks. In one word, value stocks have higher expected returns because they are riskier than glamour stocks. But they fail to provide sufficient explanations for the return differential.

Lakonishok et al. (1994) argue that the return differential is caused by investors' naive extrapolation of the past sales or earnings growth of a firm into the future: some investors tend to get overly excited about stocks doing very well in the past, usually glamour stocks, and buy them up, while they oversell stocks doing very bad in the past, usually value stocks. Consequently, glamour stocks are overpriced while value stocks are under priced. Thus, when stock prices finally return to the fundamentals in the long horizon, glamour stocks will have lower return than the value stocks. They also argue that the return differential between glamour stocks and value stocks may be due to the career concerns of portfolio managers: i.e., glamour stocks have lower career risk to portfolio managers than value stocks do. Since glamour stocks appear to be prudent investments and hence easy to justify to sponsors, many portfolio managers tilt toward glamour stocks. Conversely, most value stocks have warts and, if they blow up, the portfolio managers will look foolish since they should have known that the value stocks had problems. Thus, portfolio managers stay away from value stocks. (See Haugen (1999)) To test this naive extrapolation hypothesis, Lakonishok et al. (1997) form portfolios of glamour stocks and value stocks each year during 1971 through 1993 and examine the portfolio returns around the earnings announcement days in the post-formation period. They find that the return differentials around the earnings announcement account for approximately 25 to 30 percent of the annual return differentials between value stocks and glamour stocks in the first two to three years following portfolio formation. This may indicate that positive returns of value stocks are from their positive earnings surprises, while negative returns of glamour stocks are from their negative earnings surprises. They also find that the return differential between value stocks and glamour stocks are smaller in large stocks than in small stocks, consistent with the notion that large stocks about which more information is available are less susceptible to mis-pricing than small stocks.

Lo and MacKinly (1990) and Kothari et al. (1995) suggest that the return differential may be due to research design induced biases. Amihud and Mendelson (1986) suggest the return differential may be due to market frictions.

Dechow and Sloan (1997) finds no systematic evidence that stock prices reflect investors' naive extrapolation of past growth in earnings and sales. Instead, they find half of the returns to contrarian strategies can be explained by investors' naive reliance on analysts' growth forecasts. In sum, there is not a consensus or the best theory on explanations for the return differentials between value stocks and glamour stocks, yet.

As Xu et. al. (2006) suggest, earnings restatements due to accounting irregularities may increase the uncertainty of the reporting entity because they usually cause class action lawsuits, management shuffle, restructuring, and even bankruptcy. And earnings restatements may impair the information quality of the reporting entity because restating firm's information may not be as reliable to investment public as it used to be prior to the earnings restatement. These higher uncertainty and lower information quality due to earnings restatements increase the risk premium and stock return volatility of the restating firms (See Aboody (2005), Francis (2005), and Li (2005)). Since the magnitude of changes in uncertainty and information quality due to earnings restatements may vary across firms with different firm characteristics, especially between value firms and glamour firms, the market may react to earnings restatements of value firms differently than those of glamour firms.

Thus, it is a valuable research to examine short-term stock price responses to earnings restatements due to accounting irregularities of value firms vis a vis glamour firms1 as a way of addressing the issue of the return differentials between value firms and glamour firms, which is the purpose of this study.

It is hypothesized that prices of value stocks drop more than those of glamour stocks at the announcement of earnings restatements, if other things being equal. Empirical results of this study show that there are significantly negative Cumulative Abnormal Returns (CAR) over (-1, +1) window and (-5, +5) window surrounding the announcement of earnings restatements. And the short-term impact of earnings restatement announcements on stock prices seems to fade away by the day 1 after the announcement. The results also suggest that CAR do not vary with value/glamour identifiers such as BM, CP, and GS. In order words, CAR of value firms are not significantly different from those of glamour firms around the announcement of earnings restatements.

The remainder of this paper is organized as follows. Development of a testable hypothesis is discussed in the next section that is followed by selection of sample firms and their data. Methodology and measurement of variables are discussed in the third section. And discussions on the empirical tests and their results are followed. Conclusions are addressed in the final section.

HYPOTHESIS DEVELOPMENT

Value firms usually report higher actual earnings than their expected earnings at the initial announcement of actual earnings immediately prior to the earnings restatement, while glamour firms report lower actual earnings than their expected earnings possibly due to extrapolation of the past trend by rational but naive investment public. However, both value firms and glamour firms report lower restated earnings than their initial actual earnings at the announcement date. Thus, investment public would realize that upward price adjustments of value stocks at the initial actual earnings announcement was unwarranted and hence should make downward price adjustments with earnings restatements. They would also realize that downward price adjustments to glamour stocks at the initial actual earnings announcements was not enough and hence should make further downward price adjustments with earnings restatements. If investment public perceives a change from a negative belief on a firm value to a more negative belief (Glamour stocks) less strongly than it does a change from a positive belief to a negative belief (Value stocks), market would drop the price of value stocks more than that of glamour stocks at the announcement of earnings restatements. Another way to explain the differential market reactions to earnings restatements of value firms versus glamour firms would be as follows. For a risk averse investor whose utility function increasingly decreases with a loss and decreasingly increase with a gain, a unit of value loss due to stock price decline from lower investment value point would be more painful than that from a higher investment value point. And hence the market with full of risk averse investors would react to a unit of value loss due to stock price decline from lower investment value point more significantly than that from a higher investment value point. Since investment in value stocks, under-priced stocks, do usually have lower value than investment in glamour stocks, over-priced stocks, do, negative earnings surprises due to accounting irregularities would decrease value stock prices more than glamour stock prices. Another plausible reasoning for the differential market reactions to earnings restatements of value stocks versus glamour stocks would be as follows. Since value stocks which have poor operational performances tend to have much slimmer margin of viability and hence closer to falling into bankruptcy than glamour stocks which have good operational performances do, the market would penalize the bad news about value stocks more severely than the same bad news about glamour stocks. Thus, value stock prices would drop more than glamour stock prices at the announcement of earnings restatements due to accounting irregularities. If other things being equal, a testable hypothesis herefrom would be

Hypothesis: Ceteris Paribus, prices of value stocks drop more than those of glamour stocks at the announcement of earnings restatements.

This hypothesis may appear to be inconsistent with previous findings on contrarian investment theory, suggesting that value stocks fall less after negative earnings surprises relative to glamour stock, while value stocks rise more after positive earnings surprises relative to glamour stocks. These are rational reactions of the capital market to earnings surprises for the following reasons. First, negative earnings surprises of glamour stocks are more surprising to the market than those of value stocks are, while positive earnings surprises of value stocks are more surprising than those of glamour stocks are. Second, the market reacts to the more surprising information in larger magnitude than to the less surprising information. Extending this reasoning to earnings restatements due to accounting irregularities, negative information (downward earnings restatements due to accounting irregularities) about value stocks which performed better than expected with initial actual earnings announcements is more alarming to the market than the same information about glamour stocks which performed worse than expected. Thus, value stocks fall more than glamour stocks do with earnings restatements due to accounting irregularities.

SAMPLE DESCRIPTIONS

A list of earnings restatements due to accounting irregularities announced during January 1997 through December 2002 is obtained from General Accounting Office (GAO). According to GAO's (2002) report, it is the most comprehensive sample during that period and contains 919 earnings restatements announced by 845 public companies. The accounting and stock returns data are drawn from COMPUSTAT and CRSP, respectively. The sample period almost covers the stock market run-up during the late 1990s and its collapse after March 2000. It is the period when the number and magnitude of earnings restatement surge to historic high, providing us a large number of observations. In this period, the public concern on corporate governance grew, leading to the passage of Sarbanes-Oxley Act in July 2002. We exclude earnings restatements announced by American Depository Receipts (ADRs) firms because they are subject to different supervisory requirements.

Comparisons between characteristics of the restating firms and those of all COMPUSTAT firms are presented in Table 1. To measure the statistical significance of the difference between restating firms and all firms, a nonparametric test called Wilcoxon test was conducted because the test avoids the problems caused by skewness and outliers. Since earnings restatements are unevenly distributed across industries (Beasley et al., 2000) and the average size, Book to Market (BM) ratio, and leverage vary from industry to industry, it might be more meaningful to use the industry adjusted indicators. Industry-adjusted variables are calculated by subtracting the industry median value from the raw value of the variables. We identify companies in the same industry by matching their 4-digit historical SIC codes in the fiscal year when earnings restatement was announced. The reason to use the historical SIC code rather than the current SIC code is that some firms might change their industry after the sample period, making current SIC code an imprecise proxy for industry sector in the sample period. The earlier the event day, the more severe the problem is.

Table 1 show that the raw BM ratios of restating firms are lower than those of all firms in 5 out of 6 sample years (1997, 1998, 1999, 2001, & 2002) and the entire sample period. But the differences are not statistically significant in any year. The industry adjusted-BM ratios of restating firms, however, are higher than the industry means in all 6 testing years. And the differences are statistically significant in 3 out of 6 sample years (1999, 2000, & 2002) and the whole sample period. This discrepancy may suggest that restating firms concentrate in industries with more growth opportunities (the lower BM ratio than the overall) but they have less growth opportunities or are considered riskier than their peers (higher BM ratios than the industry mean).

Restating firms are larger in size: the mean market value of the restating firms is significantly larger than that of all COMPUSTAT firms in 5 out of 6 sample years (1998, 1999, 2000, 2001, & 2002) and the whole sample period. Our result is different from the previous results suggesting that restating firms concentrate in small firms (e.g., Beasley et al., 2000). This discrepancy might be due to a significant increase in the number of large restating firms during the sample period. The industry-adjusted market value of the restating firms are significantly higher than zero in 5 out of 6 sample years (1998, 1999, 2000, 2001, & 2002) and the whole sample period, indicating that restating firms are larger than their peers in the same industry.

Restating firms also have a lower leverage in terms of the ratio of total debt to total assets but the difference is significant in year 1997 and for the whole sample period, only, indicating restating firms have lower leverage ratios than all COMPUSTAT firms. And the industry-adjusted leverage is significantly higher than zero in 5 out of 6 years (1997, 1998, 1999, 2000, & 2001) and the whole sample period, indicating that restating firms do have higher leverage ratios than the industry average.

Some companies restated the same financial statement more than once, making the second announcement less informative. To reduce this noise, only the first announcement in the sample is kept if a company announces restatement more than once within the same fiscal year. To isolate the effect of earnings restatement from other factors, companies that announce earnings figure or guidance, or bankruptcy over the (-5, 5) event-date window are excluded. The information on earnings or earnings guidance announcement and bankruptcy announcement is collected from the U.S. news in the Factiva database around the event day of each firm. Stocks selling below one dollar (so-called penny stocks) before earnings restatement are excluded because they have wide bid-ask spreads, high commissions, low liquidity (Conrad and Kaul, 1993) and higher delisting risks. After these procedures, the final sample includes 542 restating firms with 919 earnings restatements due to accounting irregularities.

To examine the effect of mergers and acquisitions of sample firms on their earnings restatements and hence stock returns, reasons for restatements are classified and presented in Table 2. There are only 55 earnings restatements due to mergers and acquisitions out of total 919 earnings restatements, which account for 6 %. And hence mergers and acquisitions are not a major cause of earnings restatements.

METHODOLOGY AND MEASUREMENT OF VARIABLES

Abnormal returns or cumulative abnormal returns (CAR) around the announcement of earnings restatement are estimated to measure the stock price reaction to restatement announcements. Corhay and Rad (1996) show that since stock returns series generally exhibit time-varying volatility, a market model accounting for generalized autoregressive conditional heteroskedastic (GARCH) effects produces more efficient estimators of abnormal returns than a market model estimated using the ordinary least squares (OLS) method. Thus, we used the market model accounting for generalized autoregressive conditional heteroskedastic (GARCH (1, 1)) technique to estimate the abnormal returns during the (-7, +7) event day windows. The model parameters are estimated using the data from 250 to 50 days before the earnings restatement announcement. CAR is the sum of the abnormal returns during the event window.

Like Lakonishok et al.'s (1994), we use the book-to-market (BM) ratio, the cash flow-to-market value (CP) ratio, and the past sales growth (GS) as proxies for the glamour/value characteristics. Glamour stocks are usually stocks with low BM, low CP, and/or high GS, whereas value stocks are usually stocks with high BM, high CP, and/or low GS.

In the univariate test, we divided the sample stocks into five categories by their glamour/value characteristics and compare their average abnormal returns upon restatement announcements. To divide the sample stocks into five categories, a universe of stocks are sorted in ascending order into five deciles at the end of each year by a proxy for the glamour/value characteristics mentioned earlier; we then fit the sample stocks into the five deciles by the proxy. Our universe of stocks consists of all the stocks listed on the NYSE, ASE, and NASDAQ, except for real estate investment trusts (REITs), ADRs, closed-end funds, unit investment, and trusts.

To rank the stocks by the past sales growth, we first calculate the sales growth of all the stocks over the five years prior to the year when the universe stocks were formed; we then calculate the weighted average sales growth of each stock, giving the weight of 5, 4, 3, 2, 1 to its growth in year -1, -2, -3, -4, -5, respectively. We then divide the weighted average sales growth by the number of years that the stock has continuous sale growth. The average past sales growth of new stocks might be inaccurate because they have very few past sales figures. To rule out the influence of new stocks, we require that all the stocks should have sales during the two years prior to the formation year. Stocks falling in decile 1 (5) in the year when earnings restatement is announced have high (low) past sales growth and hence are glamour (value) stocks. If the hypothesis holds, then the stocks in decile 1 should have lower negative CAR than those in declie 5 around the announcement of earnings restatements.

Lakonishok et al. (1994) suggest a two-way classification is a better method to identify value stocks and glamour stocks than a one-way classification. Thus, we also use the combination of BM ratio and past earnings/sales growth to identify the restating firm's glamour/value stock characteristics. At the end of each year, the universe stocks are independently sorted in ascending order into three groups--(1) bottom 30 percent, (2) middle 40 percent, and (3) top 30 percent--by the raw BM ratio and GS, and then take intersections resulting from the two classifications. Restating firms in the low (high) BM high (low) GS group at the end of the year prior to the earnings restatement are glamour (value) stocks. We also use the combination of CP ratio and GS to identify value stocks and glamour stocks.

It has been widely evidenced that there is a positive relationship between earnings surprises and stock price changes and hence abnormal stock returns; i.e., a negative earnings surprise (an excess of the expected earnings over the actual earnings) decreases the stock prices and hence stock returns, while a positive earnings surprise (an excess of the actual earnings over the expected earnings) increases the stock price. And the larger the earnings surprises, the larger stock price changes and abnormal stock returns. (See Beaver et al. (1979))

The following multivariate regression model is estimated to control for the effect of magnitude of earnings restatement on CAR.

[CAR.sub.it] = [alpha] + [[beta].sub.1] [MAG.sub.it] + [[beta].sub.2] [BM.sub.it], + [[epsilon].sub.t] (1)

where [CAR.sub.it] denotes the cumulative abnormal return on firm i over the (-1,1) window; [MAG.sub.it], the restatement magnitude of firm 'i' is the cumulative net change in the firm's net income due to earnings restatement scaled by the shareholders' equity at the end of the quarter prior to the restatement. [BM.sub.it] is the book-to-market ratio. We do not add interactive term (the product of the restatement magnitude and BM) because the correlation between these two independent variables is insignificant. The next, we replaced the BM ratio with the CP ratio and GS, respectively, as a proxy for the glamour/value characteristics and redo the multivariate tests. Although the inclusion of restatement magnitude can improve the explanatory power, it decreases the degree of freedom because there are only 202 observations for restatement magnitude data. Because firms restate financial results in different categories and tax data is not available for some companies, only 202 observations have enough data to compute the restatement magnitude.

In order to control for the influence of MAG and other unknown variables on CAR, we run a two step regression where error terms from the following regression model are regressed over the value/glamour identifiers such as BM and CP.

[CAR.sub.it] = [alpha] + [beta] [MAG.sub.it] + [[epsilon].sub.t] (2)

where [CAR.sub.it] and [MAG.sub.it] are the same as those in regression (1).

The regression model for the error terms from the equation (2) over glamour/value stock identifiers would be as follow:

[RES.sub.it] = [alpha] + [[beta].sub.1] [BM.sub.it] + [[epsilon].sub.t] (3)

[RES.sub.it] = [alpha] + [[beta].sub.1] [CP.sub.it] + [[epsilon].sub.t] (3')

Where BM = the book to market value ratio,

CP = the cash flow to market value ratio.

EMPIRICAL RESULTS

To control for time-varying volatility of stock returns, a market model adjusted for generalized autoregressive conditional heteroskedastic (GARCH (1, 1)) technique is used to estimate abnormal returns. Results from the model are presented in Table 3. The GARCH-adjusted average CAR are -7.42 percent and -8.96 percent in (-1, 1) and (-5, 5) windows, respectively, both of which are statistically significant. None of daily abnormal returns are statistically significant from the day 2 after the announcement of restatements in terms of the standardized cross-sectional (SCS) test (t-statistic) and the generalized sign test (Z-statistic). In sum, the results in Table 2 suggest that there are significantly negative CAR around the announcement of earnings restatements, consistent with previous studies on association between earnings surprises and abnormal stock returns. (See Beaver, Clarke, and Wright (1979)) And the short-term impact of earnings restatement announcements on stock prices seems to fade away by the day 1 after the announcement. Thus, it is reasonable to use CAR for (-1, +1) window to examine differential market reactions to earnings restatements of value stocks vs. glamour stocks.

To see if there is any difference in behavior of CAR between bull market years (1997 to 1999) and bear market years (2000-2002), the GARCH adjusted average CAR for testing periods of the bull market years are computed and presented on Table 4, while the CAR of the bear market years on Table 5. The GARCH-adjusted average CAR for the bull market years are -6.92 percent and -7.78 percent in (-1, 1) and (-5, 5) windows, respectively, both of which are statistically significant at 1 % level. The GARCH-adjusted average CAR for the bear market years are -7.79 percent and -9.85 percent in (-1, 1) and (-5, 5) windows, respectively, both of which are also statistically significant at 1 % level. None of daily abnormal returns are statistically significant from the day 2 after the announcement of restatements for both bull and bear market years. In sum, the results presented in Table 4 for the bull market years and Table 5 for the bear market years are consistent with results in Table 3 for the entire sample years: i.e., there are significantly negative CAR on the date of restatement announcements and one day after.

Average CAR and restatement magnitude of five subgroups sorted by BM, CP, and GS are presented in Panels A, B, C of Table 6, respectively. Restating firms in all sub groups show negative MAG, indicating that restated earnings are lowed than the initial actual earnings preceding the restatement. They also show negative CAR, again. However, there are no significantly different CAR between any two subgroups except the difference between subgroups 1 & 3 sorted by BM and the difference between subgroups 3 and 5 sorted by GS. This indicates that CAR may not vary with any of value/glamour identifiers. These findings imply that the capital market does not react to earnings restatements of value firms differently than to those of glamour firms, inconsistent with the hypothesis.

Average CAR and restatement magnitude of 3 subgroups sorted by the two-way classification method are presented in Table 7. Those statistics of subgroups sorted by BM and GS are presented in Panel A, while those by CP and GS in Panel B of Table 7. The results in table 7 are very similar to those in Table 6 and hence do not support the hypothesis that the market reacts to earnings restatements of value firms more strongly than to those of glamour firms. The results show negative MAG and CAR in all subgroups, while CAR does not vary with any of value/glamour characteristics. Except that there is a significantly different CAR between subgroups 1 & 3 sorted by BM and GS at 10% level.

Results from multivariate regression model (1) using value/glamour identifiers (Such as BM and CP) and MAG as independent variables are presented in Table 8. The regression coefficients of BM and CP are -0.005 and -0.01, respectively. But both of them are not statistically significant and hence do not support the hypothesis.

Results from the two step regression are presented in Table 9. The regression coefficients of BM and CP are -0.0071 (p-value = -1.62) and -0.0875 (p-value = -0.86), respectively, both of which are not statistically significant. Thus, these results also do not support the hypothesis.

Again, to see if there is any structural difference between the bull market years and the bear market years that cause different market reactions to earnings restatements of value stocks vis a vis glamour stocks, the multiple regression model (1), the two step regression models (2), and (3) are estimated for the bull market years and the bear market years, separately. Results from the multiple regression models are presented in Table 10. The regression coefficients of glamour/value proxies such as BM and CP for the bull market years presented in Panel A & C of Table 10 are negative but not statistically significant, while the regression coefficients of glamour/value proxies such as BM and CP for the bear market years presented in Panel B & D of Table 10 are positive but not statistically significant, not supporting the hypothesis.

Results from the two step regression model for the bull market years are in Panel A of Table 11, while those for the bear market years are in Panel B of Table 11. None of the regression coefficients of BM and CP for both bull market years and bear market years are statistically significant, not supporting the hypothesis that the market reacts to earnings restatements of value firms more strongly than to those of glamour firms.

In sum, results shown in Tables 3, 4, 5, 6, 7, 8, 9, 10, and 11 suggest that, in general, the restating firms experience stock price declines around the announcement of earnings restatements but there is no significant difference in this phenomena between value firms and glamour firms, inconsistent with the hypothesis. These results are robust across different measurement of variables. testing methods, and markets (bull market and bear market).

CONCLUSIONS

This paper examines short-term stock price responses to earnings restatements due to accounting irregularities of value firms vis a vis glamour firms as a way of addressing the issue of the return differentials between value firms and glamour firms. It was hypothesized that, Ceteris Paribus, prices of value stocks drop more than those of glamour stocks at the announcement of earnings restatements.

Empirical results of this study show that there are significantly negative CAR over (-1, +1) window and (-5, +5) window surrounding the announcement of earnings restatements, suggesting that market perceives earning restatements due to accounting irregularities, negatively. These results are consistent with deep discounts on stocks of companies infringed by so-called stock option backdating scandals in a sense that bad (good) news decrease (increase) the stock prices. Recently hundreds of companies, many of which are high tech growth companies, are under investigations for their misconducts on the date of granting stock options. Those companies may intentionally change the day that the stock options were granted to an earlier date when their stocks were trading at lower prices, potentially allowing company executives to lock in higher profits when they exercise their options. Thus, the stock prices of those companies drop significantly with the release of the information. The short-term impact of earnings restatement announcements on stock prices seems to fade away by the day 2 after the announcement. However, the results do not show that CAR (-1, +1) vary with any value/glamour identifiers such as BM, CP, and GS, which do not support the hypothesis that value stocks have higher negative CAR than glamour stocks do with earnings restatements. This suggests that the market does not perceive earnings restatements of value firms any differently than those of glamour firms. These results are robust across different measurement of variables, testing methods, and markets (bull market and bear market).

REFERENCES

Aboody, D., J. Hughes, and J. Liu, 2005, Earnings quality, insider trading, and cost of capital, Journal of Accounting Research vol. 43 No. 5, 651-673.

Amihud, D. and H. Mendelson, 1986, Asset pricing and the bid-ask spread. Journal of Financial Economics 20, 223-249.

Ball, R., S. Kothari, and J. Shanken, 1995, Problems in measuring portfolio performance: Application to contrarian investment strategies, Journal of Financial Economics 38, 79-107.

Beasley, M., J. Carcello, D. Hermanson, and P. Lapides, 2000, Fraudulent financial reporting: consideration of industry traits and corporate governance mechanisms, Accounting Horizons 14 (4), 441-454.

Beaver, W., R. Clarke, and W. Wright, 1979, The association between unsystematic security returns and the magnitude of earnings forecast errors, Journal of Accounting Research 17, 316-340.

Conrad, Jennifer and Gautam Kaul, 1993, Long-term market overreaction or biases in computed returns? Journal of Finance 48, 39-63.

Corhay, A. and A. Tourani Rad, Conditional heteroskedasticity adjusted market model and an event study, The Quarterly Review of Economics and Finance 36 (4), 520-538.

Dechow, P., and R. Sloan, 1997, Returns to contrarian investments: test of the naive extrapolations hypothesis, Journal of Financial Economics, 3-27.

Dreman, D., 1998. Contrarian Investment Strategies in the Next Generation, Simon & Schuster, Inc..

Fama, Eugene and Kenneth French, 1992, The cross-section of expected stock returns, Journal of Finance 47, 427-465.

Fama, Eugene and Kenneth French, 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial Economics 33, 3-56.

Fama, Eugene and Kenneth French, 1995, Size and book-to-market factors in earnings and stock returns, Journal of Financial Economics 33, 3-56.

Fama, Eugene and Kenneth French, 1996, Multifactor explanation of asset pricing anomalies, Journal of Finance 51, 55-84.

Francis, J. and M. Smith, 2005, A reexamination of persistence of accruals and cashflows, Journal of Accounting Research, Vol 43 No. 3, 413-452.

General Accounting Office (GAO), 2002, Financial statement restatements: trends, market impacts, regulatory responses, and remaining challenges, GAO-03-138.

Graham, B., and D. Dodd, 1934, Security Analysis, McGraw-Hill.

Kothari, S. P. and Jerold B. Warner, 1997, Measuring long-horizon security price performance. Journal of Financial Economics 43, 301-339.

Hougen, R., 1999, The New Finance: The Case Against Efficient Markets, 2nd edition. Prentice Hall.

Lakonishok, J., A. Shleifer, and R. Vishny, 1994, Contrarian investment, extrapolation, and risk, Journal of Finance 49 (5), 1541-1578.

Lakonishok, J., R. La Porta, A. Shleifer, and R. Vishny, 1997, Good news for value stocks: further evidence on market efficiency, Journal of Finance, Vol. 52 No. 2, 859-874.

Li, G., 2005, Information quality, learning, and stock market returns, Journal Of Financial And Quantitative Analysis, Vol. 40 No.3, 595-620.

Lo, A., and C. Mackinly, 1990, When are contrarian profits due to stock market overreactions? Review of Financial Studies 3, 175-206.

Xu, t., D. Li, and J. Jin, 2006, Long-term market reactions to earnings restatements, working paper.

Tan Xu, Old Dominion University

Diane Li, University of Maryland-Eastern Shore

John Jongdae Jin, California State University-San Bernardino

ENDNOTES

(1) This study adopts the definition made by GAO (2002), i.e., it is "an instance in which a company restates its financial statements because they were not fairly presented in accordance with generally accepted accounting principles (GAAP). This would include material errors and fraud."

Table 1: Descriptive Statistics of Restating Firms and All COMPUSTAT
Firms

Panel A Book-to-market ratio

Year N1 Restating firms N2 All firms Diff

1997 61 0.572 33032 0.648 -0.076
1998 70 0.502 32633 0.772 -0.27
1999 122 0.694 31348 1.009 -0.315
2000 155 0.983 31629 0.854 0.129
2001 173 1.514 30032 1.837 -0.323
2002 91 1.382 27145 1.774 -0.392
Total 672 1.004 185819 1.145 -0.141

Year z-stat Industry t-stat
 Adjusted

1997 -0.71 0.153 1.95
1998 -1.79 0.023 0.54
1999 -1.78 0.153 2.43 *
2000 1.15 0.271 4.05 **
2001 -1.64 0.371 1.75
2002 -0.15 0.353 2.06 *
Total -0.35 0.249 3.89 **

Panel B Market value (Million dollars)

Year N1 Restating firms N2 All firms Diff

1997 64 550.8 36827 1034.13 -483.33
1998 70 2450.27 36857 1292.68 1157.59
1999 128 2234.65 36404 1566.89 667.76
2000 166 1935.09 37222 1902.6 32.43
2001 183 2796.24 35970 1580.11 1216.13
2002 92 2695.2 33886 1496.68 1198.52
Total 703 2230.89 217166 1484.35 746.54

Year z-stat Industry t-stat
 Adjusted

1997 -0.94 379.02 1.81
1998 0.7 2366.71 2.26 *
1999 2.99 ** 1737.54 2.06 *
2000 2.15 * 1815.02 2.07 *
2001 7.98 ** 2601.81 4.09 **
2002 5 74 ** 1716.04 2.16 *
Total 8.44 ** 1921.11 5.66 **

Panel C Total debt / Total asset

Year N1 Restating firms N2 All firms Diff

1997 64 0.33 40002 0.513 -0.184
1998 71 0.234 39410 0.386 -0.152
1999 132 0.309 40310 0.486 -0.177
2000 167 0.281 40616 0.601 -0.32
2001 192 0.269 37973 0.9 -0.63
2002 96 0.132 34591 0.281 -0.149
Total 722 0.283 232902 0.685 -0.402

Year z-stat Industry t-stat
 Adjusted

1997 3.11 ** 0.107 4.10 **
1998 -0.18 0.661 3.17 **
1999 0.47 0.116 3.66 **
2000 0.46 0.947 4 11 **
2001 0.65 0.52 3.52 **
2002 0.84 0.342 1.55
Total 2.10 * 0.779 7.87 **

N1 = the number of restating firms with non-negative value,

N2 = the number of all COMPUSTAT firms with non-negative value,

Diff = the difference between the median (mean) of the restating
firms and those of all the COMPUSTAT firms.

*, and ** = statistical significance at the 5% and 1% levels,
respectively, using a 2-tail test.

Table 2: Reasons for Restatement (by Incidents)

Reasons for Restatement Incidents Percentage
 (%)

Revenue recognition 349 38
Cost or expense related 144 15.7
Others 131 14.3
Restructuring, assets, or inventory 82 8.9
Mergers and Acquisitions 55 6
Securities related 50 5.4
Reclassifications 47 5.1
In-process research and development related 33 3.6
Related-party transactions 28 3
Total 919 100

Table 3: GARCH-Adjusted Abnormal Returns around Earnings Restatement
Announcements

 Day Obs. Mean Abnormal Median Abnormal Positive:
 Return (%) Return (%) Negative

 -7 517 -0.58 -0.25 230:287
 -6 517 0.09 -0.12 246:271
 -5 517 -0.42 -0.36 224:293
 -4 517 -0.16 -0.41 221:296
 -3 517 -0.93 -0.46 204:313
 -2 516 -0.2 -0.12 241:275
 -1 515 -0.25 -0.19 239:276
 0 510 -2.9 -0.94 204:306
 1 505 -4.38 -1.48 187:318
 2 507 0.07 -0.18 239:268
 3 507 0.15 -0.13 241:266
 4 506 -0.08 -0.33 233:273
 5 507 0.01 -0.17 240:267
 6 508 0.04 -0.09 243:265
 7 508 -0.24 -0.22 239:269
 CAR
(-1, +1) 515 -7.42 -3.54 167:348
(-5, +5) 517 -8.96 -3.99 190:327

 Day t-stat Generalized
 Sign Z

 -7 -2.640 ** -1.207
 -6 0.388 0.203
 -5 -1.878 -1.735
 -4 -0.715 -2.000 *
 -3 -4 187 *** -3 497 ***
 -2 -0.913 -0.196
 -1 -1.118 -0.331
 0 -13.128 *** -3.229 **
 1 -19 798 *** -4 550 ***
 2 0.294 0.002
 3 0.665 0.18
 4 -0.366 -0.491
 5 0.032 0.091
 6 0.16 0.315
 7 -1.065 -0.04

(-1, +1) -9 750 *** -6.687 ***
(-5, +5) -8.723 *** -4 731 ***

Abnormal returns = the difference between the actual return and the
predicted returns calculated by the GARCH-adjusted market model.

Obs = the number of sample firms.

* and ** = statistical significance at 5% & 1%, respectively using a
2-tail test.

Table 4: GARCH-Adjusted Abnormal Returns around Earnings Restatement
Announcements for the sub-sample period:
Jan. 1997-Mar. 2000

 Day Obs. Mean Abnormal Median Abnormal Positive:
 Return (%) Return (%) Negative

 -7 214 -0.56 -0.23 95:119
 -6 214 0.11 -0.1 102:112
 -5 214 -0.41 -0.35 93:121
 -4 214 -0.14 -0.39 91:123
 -3 214 -0.9 -0.43 84:130
 -2 213 -0.18 -0.1 99:114
 -1 213 -0.19 -0.13 98:114
 0 210 -2.92 -0.96 83:124
 1 203 -4.35 -1.45 75:127
 2 204 0.09 -0.16 96:108
 3 204 0.18 -0.1 97:107
 4 203 -0.06 -0.31 93:110
 5 204 0.02 -0.16 97:107
 6 205 0.06 -0.07 98:107
 7 205 -0.23 -0.21 96:109
 CAR
(-1, +1) 213 -6.92 -3.21 93:119
(-5, +5) 214 -7.78 -3.37 95:117

 Day t-stat Generalized Sign Z

 -7 -1.134 -1.296
 -6 0.269 0.637
 -5 -1.691 -1.542
 -4 -0.842 -1.066
 -3 -1.153 -3 497 ***
 -2 -3.133 *** -0.196
 -1 -0.978 -0.331
 0 -9 947 *** -2. 975 **
 1 -13.674 *** -5.046 ***
 2 0.638 0.047
 3 0.863 0.894
 4 -0.954 -0.847
 5 0.019 0.185
 6 0.635 0.853
 7 -0.932 -0.053

(-1, +1) -8.532 *** -5 753 ***
(-5, +5) -7 753 *** -4.136 ***

Abnormal returns = the difference between the actual return and the
predicted returns calculated by the GARCH-adjusted market model.

Obs = the number of sample firms.

*, ** and *** = statistical significance at 10%, 5% & 1%, respectively
using a 2-tail test.

Table 5: GARCH-Adjusted Abnormal Returns around Earnings Restatement
Announcements for the sub-sample period:
Apr. 2000-Jun. 2002

 Day Obs. Mean Abnormal Median Abnormal Positive:
 Return (%) Return (%) Negative

 -7 303 -0.59 -0.26 135:168
 -6 303 0.08 -0.13 144:159
 -5 303 -0.43 -0.37 131:172
 -4 303 -0.17 -0.42 130:173
 -3 303 -0.95 -0.48 120:183
 -2 303 -0.21 -0.13 142:161
 -1 302 -0.29 -0.23 140:162
 0 300 -2.89 -0.93 120:180
 1 302 -4.4 -1.5 111:190
 2 303 0.06 -0.19 143:160
 3 303 0.13 -0.15 144:159
 4 303 -0.09 -0.34 140:163
 5 303 0 -0.18 143:160
 6 303 0.03 -0.1 145:158
 7 303 -0.25 -0.23 143:160
 CAR
(-1, +1) 302 -7.79 -3.63 167:348
(-5, +5) 303 -9.85 -4.74 190:327

 Day t-stat Generalized Sign Z

 -7 -2.832 ** -1.436
 -6 0.388 0.424
 -5 -1.578 -1.386
 -4 -1.046 -2.304 *
 -3 -5 327 *** -4.064 ***
 -2 -1.058 -0.296
 -1 -0.903 -0.572
 0 -14.735 *** -3.858 **
 1 -18.858 *** -4 958 ***
 2 0.578 0.043
 3 0.365 0.898
 4 -0.735 -0.459
 5 0.174 0.171
 6 0.016 0.585
 7 -0.965 -0.073

(-1, +1) -9.960 *** -7.001 ***
(-5, +5) -8.683 *** -4.938. ***

Abnormal returns = the difference between the actual return and the
predicted returns calculated by the GARCH-adjusted market model.

Obs = the number of sample firms.

*, ** and *** = statistical significance at 10%, 5% & 1%, respectively
using a 2-tail test.

Table 6: Mean Comparisons between Groups Sorted by One-Way
Classification Method

Panel A. Sample stocks grouped by BM ratio

 Group 1 Group 2 Group 3 Group 4 Group 5
 (Growth) (Value)

Obs. 90 79 101 87 123
MAG -7.21 -2.32 -4.69 -5.91 -7.03
CAR -7.13 -4.83 -9.13 -6.71 -8.07

 t-stat t-stat t-stat
 (1 & 3) (3 & 5) (1 & 5)

Obs.
MAG
CAR 1.69 * 0.70 0.18

Panel B. Sample stocks grouped by CP ratio

 Group 1 Group 2 Group 3 Group 4 Group 5
 (Growth) (Value)

Obs. 86 97 93 79 87
MAG -6.85 -4.62 -3.44 -5.09 -7.95
CAR -7.31 -4.48 -9.53 -6.91 -8.12

 t-stat t-stat t-stat
 (1 & 3) (3 & 5) (1 & 5)

Obs.
MAG
CAR 0.39 0.85 0.18

Panel C. Sample stocks grouped by GS

 Group 1 Group 2 Group 3 Group 4 Group 5
 (Growth) (Value)

Obs. 82 89 84 67 91
MAG -7.47 -3.95 -4.79 -4.03 -7.12
CAR -5.91 -5.61 -8.93 -7.52 -7.4

 t-stat t-stat t-stat
 (1 & 3) (3 & 5) (1 & 5)

Obs.
MAG
CAR 0.63 1.82 * 0.95

Obs = the number of firms in each subgroup.

MAG = the average magnitude of earnings restatements in each
subgroup.

CAR = the average CAR of each subgroup.

*, ** = Statistical significance at the 10% and 5%, respectively.

Table 7: Mean Comparisons between Groups Sorted by Two-Way
Classification Method

Panel A. Sample stocks grouped by both BM and GS

 Group 1 Group 2 Group 3
 (Low BM & (Middle BM (High BM
 High GS) & Middle GS) & Low GS)

Obs. 30 38 81
MAG -5.35 -4.15 -7.22
CAR -5.75 -5.92 -8.21

 t-stat t-stat t-stat
 (1 & 2) (2 & 3) (1 & 3)

Obs.
MAG
CAR 0.56 1.36 1.69 *

Panel B. Sample stocks grouped by both CP and GS

 Group 1 Group 2 Group 3
 (Low BM & (Middle BM (High BM
 High GS) & Middle GS) & Low GS)

Obs. 7 41 67
MAG -6.35 -5.45 -4.70
CAR -6.20 -5.61 -7.09

 t-stat t-stat t-stat
 (1 & 2) (2 & 3) (1 & 3)

Obs.
MAG
CAR 0.92 1.57 0.91

Obs = the number of firms in each subgroup.

MAG = the average magnitude of earnings restatements in each
subgroup.

CAR = the average CAR of each subgroup.

*, ** = Statistical significance at the 10% and 5%, respectively

Table 8: Regressions of CAR on the glamour/value proxies & the
Restating Magnitude

[CAR.sub.it] = [alpha] + [[beta].sub.1] [MAG.sub.it] +
[[beta].sub.2] [BM.sub.it] + [[epsilon].sub.t] (1)

[CAR.sub.it] = [alpha] + [[beta].sub.1] [MAG.sub.it] +
[[beta].sub.2] [CP.sub.it] + [[epsilon].sub.t] (1')

Panel A. Equation (1)

Intercept MAG BM Adj. [R.sup.2]

-6.67 0.03 -0.005 0.07
(-4.22 **) (0.72) (-0.33)

Panel B. Equation (1')

Intercept MAG CP Adj. [R.sup.2]

-5.58 0.03 -0.01 0.08
(-5.13 **) (0.61) (-0.86)

[CAR.sub.it] = CAR of firm 'i' in year 't'.

MAG = the magnitude of earnings restatements.

BM = the book to market value ratio.

CP = the cash flow to market value ratio.

*, ** = Statistical significance at the 10% and 5%, respectively

() = the t-value.

Table 9: Regressions of the residual on the glamour/value proxies

[RES.sub.it] = [alpha] + [[beta].sub.1] [BM.sub.it] +
[[epsilon].sub.t] (3)

[RES.sub.it] = [alpha] + [[beta].sub.1] [CP.sub.it] +
[[epsilon].sub.t] (3')

Intercept BM CP Adj. [R.sup.2]

 0.0023 -0.0071 0.03
 (-1.62)

 0.0017 -0.0875 0.07
 (-0.86)

RES = the residual of the regression of CAR on restatement magnitude.

BM = the book to market value ratio.

CP = the cash flow to market value ratio.

*, ** = Statistical significance at the 10% and 5%, respectively

() = the t-value.

Table 10: Regressions of CAR on the glamour/value proxies & the
Restating Magnitude for two sub-sample periods

[CAR.sub.it] = [alpha] + [[beta].sub.1] [MAG.sub.it] + [[beta].sub.2]
[BM.sub.it] + [[epsilon].sub.t] (1)

[CAR.sub.it] = [alpha] + [[beta].sub.1] [MAG.sub.it] + [[beta].sub.2]
[CP.sub.it] + [[epsilon].sub.t] (1)

Panel A. Equation (1) for the sub-sample period form January 1997 to
March 2000

Intercept MAG BM Adj. [R.sup.2]

 -6.98 0.17 -0.014 0.03
(-2.12 **) (0.59) (-0.04)

Panel B. Equation (1) for the sub-sample period from April 2000 to
June 2002

Intercept MAG BM Adj. [R.sup.2]

 -6.16 0.08 0.01 0.17
(-4.84 **) (0.29) (0.49)

Panel C. Equation (1') for the sub-sample period form January 1997 to
March 2000

Intercept MAG CP Adj. [R.sup.2]

 -6.03 0.01 -0.04 0.1
(-3.75 **) (0.93) (-0.57)

Panel D. Equation (1') for the sub-sample period from April 2000 to
June 2002

Intercept MAG CP Adj. [R.sup.2]

 -5.26 0.03 0.00 0.08
(-2.37 **) (0.61) (0.63)

BM = the book to market value ratio.

CP = the cash flow to market value ratio.

*, ** = Statistical significance at the 10% and 5%, respectively

() = the t-value

Table 11: Regressions of the residual on the glamour/value proxies for
two sub-sample periods

[RES.sub.it] = [alpha] + [[beta].sub.1] [BM.sub.it] +
[[epsilon].sub.t] (3)

[RES.sub.it] = [alpha] + [[beta].sub.1] [CP.sub.it] +
[[epsilon].sub.t] (3')

Panel A. Regression for the sub-sample from January 1997 to March 2000

Intercept BM CP Adj. R2

 0.007 0.011 0.02
 (1.62)

 0.002 -0.105 0.09
 (-1.02)

Panel B. Regression for the sub-sample from April 2000 to June 2002

Intercept BM CP Adj. R2

 0.001 -0.009 0.03
 (-0.82)

 0.002 -0.02 0.03
 (-0.53)

BM = the book to market value ratio.

CP = the cash flow to market value ratio.

*, ** = Statistical significance at the 10% and 5%, respectively

() = the t-value.