Peter J. BUSH University of Michigan-Flint School of Management Adjunct Professor of Finance

Transcription

1 ANALELE ŞTIINŢIFICE ALE UNIVERSITĂŢII ALEXANDRU IOAN CUZA DIN IAŞI Număr special Ştiinţe Economice 2010 A CROSS-INDUSTRY ANALYSIS OF INVESTORS REACTION TO UNEXPECTED MARKET SURPRISES: EVIDENCE FROM NASDAQ SECTOR- INDICES Peter J. BUSH University of Michigan-Flint School of Management Adjunct Professor of Finance pjbush@umflint.edu Seyed M. MEHDIAN University of Michigan-Flint School of Management Professor of Finance seyed@umich.edu Mark J. PERRY University of Michigan-Flint School of Management Professor of Finance mjperry@umich.edu Abstract We use daily stock returns from the NASDAQ composite index and its eight composite indexes to investigate the reaction of investors to the arrival of unexpected information in framework of Efficient Market Hypothesis (EMH), the Overreaction Hypothesis (OH), and the Uncertain Information Hypothesis (UIH). Consistent with the prediction of the UIH regarding investor behavior, we find strong statistical evidence of a corrective process of significantly positive cumulative abnormal returns following the arrival of both unexpected favorable and unfavorable information for the NASDAQ Composite and four of the sub-indexes. For the Computer and Insurance sub-indices, we observe downward trends in the CARs following the arrival of favorable market surprises and upward trends in response to the arrival of unfavorable surprises, a result consistent with the predictions of the OH. For the Industrials and Transportation sectors, the pattern of investors reaction is not strongly consistent with any of the theories of investor reaction presented here, although the trendless pattern of returns following market surprises for the Transportation index could be explained by the EMH.

2 98 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY One main implication of these mixed empirical results is that investor s reaction varies significantly by sector, highlighting the value of analyzing investor reaction in different segments of the security markets in addition to investigating composite indexes. Keywords: Efficient Market Hypothesis, Rational Investors. Overreaction; Uncertain Information; Unfavorable market surprised JEL classification: D40, G14, G15 I. INTRODUCTION The efficient market hypothesis (EMH) has been part of the essential framework of modern financial theory since it emerged as a prominent theory in the early 1960s. The key idea of the EMH is that financial markets are considered to be efficient when there is an equilibrium state for the prices of the securities traded within that market. In essence, prices are considered to be neither too high nor too low for each of the available traded assets, relative to the fundamental value of those assets. While there are several different forms of the EMH that have been identified, the general concept is that the current price of the financial assets traded within an efficient market has incorporated all available and relevant information, and therefore, there will be no opportunities for an investor to trade a security within that market and obtain anything other than a normal, risk-adjusted return. The EMH is based on the assumptions that investors are rational, that a majority of the traders in the market will react properly to new information, and that as new information becomes available the traders in the markets will act efficiently and with relative precision. In an efficient market environment, the assumption is that there is no systematic bias for investors to skew security prices either positively or negatively from their correct prices, based on the injection of any pertinent and new information into that market. The EMH also suggests that even in the event that some market participants do not react in a rational and expected manner, a majority of the market participants reactions will counteract and offset any potential departures from an equilibrium market state that might initially be brought about by some irrational traders. The end result is that prices in an efficient market are expected to accurately and quickly gravitate towards equilibrium following any shock to the market, regardless of the nature of the information. While there have been a multitude of studies conducted on various securities markets that provide empirical support of the EMH, there has also been a fair amount of skepticism raised regarding some of the primary assumptions of the hypothesis. One of the assumptions most frequently questioned by both industry professionals and academics alike, is the EMH s basic assumption of investor rationality. This fundamental criticism of the EMH has been buttressed by market observations that seem to indicate that counter to the claims of the EMH, potential opportunities to earn a greater-than-normal return do arise in many markets and can be exploited by some investors. In support of these observations, many studies have been conducted that provide evidence that markets are frequently inefficient, such that there is indeed a potential for investors to earn abnormally-high, risk-adjusted returns in a manner that would contradict the EMH. On the basis of the observed evidence that investors do not always react in a way that is consistent with the predictions of the EMH, further studies were completed in order to understand the nature of these markets that are not efficient as predicted by EMH. By virtue

3 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 99 of these studies, and as a contrarian view of investors reactions to market information, multiple theories arose as alternative explanations for the reactions of investors to new information. One of these theories, the Overreaction Hypothesis (OH), observes that investors might overreact in the short term to news that significantly affects asset prices. The basis of the hypothesis is that investors tend to react particularly strongly, and possibly even overreact, to both the arrival of significantly good and bad news, which could create a scenario where there would be a subsequent corrective activity in the prices of securities during a post-event period of gradual adjustment. According to the OH, the expectation would be that the corrective pattern following positive information would be a negative movement in security prices to correct for the initial, positive overreaction to the good news, while the corrective activity following a negative event would be a positive trend in security prices, to correct for the initial, negative overreaction to the bad news. In either case, the corrective adjustments would eventually bring security prices back to a state of equilibrium following an initial overreaction to market-induced surprises. Another of the counter-theories that has been postulated as an alternative to the EMH is the Uncertain Information Hypothesis (UIH). The UIH theory predicts that the volatility of asset returns will increase following the release of unexpected information, because of investor uncertainty about how to correctly react to the news and set appropriate security prices. More specifically, the UIH theory postulates that investors will tend to overreact to bad news in the same manner as predicted by the OH, but in contrast to the OH, the UIH suggests that investors will tend to underreact to good news, based primarily on the general conservatism of investors. Regardless of the type of news, the UIH predicts that the resultant market correction will always be of a positive nature following the initial market reaction. Further studies have found that the tendency is for the volatility of returns following negative shocks is greater than the average volatility of returns following positive news. One other potential investor reaction that has been frequently observed is the tendency for investors to underreact to new information, likely as the result of excessive conservatism of investors. Faced with the arrival of new information, especially information that is unexpected in nature, investors have historically shown a tendency to react with caution and conservatism. Good news is often met with only minor, positive stock price corrections that initially fall below the ultimate, higher equilibrium price level, and the security prices gradually experience upward-pressure over time to correct for the initial underreaction to the positive shock. The same pattern of an upward corrective trend in security prices has been documented in some studies following the arrival of bad news, with appropriate upward market corrections in prices following the initial overly conservative reaction that sets the short-term price below its equilibrium level. Regardless of the specific nature of the reaction of investors to the arrival of surprises, whether the reaction is consistent with the UIH, OH, or Underreaction theories, it has become clear that security prices do not always follow the EMH-expected reactions in all markets at all times. With the release of new information, investors do not always display a consistent tendency to adjust prices to the appropriate level quickly and efficiently, but rather appear frequently to be taking their time to feel out the new situation and make gradual adjustments to security prices. Perhaps this reaction is simply a display of human nature, as people frequently tend to gradually analyze a situation over time and make periodic and continuous adjustments, rather than reacting immediately, decisively, discretely and correctly to the information at hand. The consequence of such conservative and risk-averse investor behavior results in security prices being temporarily mispriced in the short run, at

4 100 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY initially either above or below their eventual correct and equilibrium price. It is during this temporary period of mispricing that an asset market could be said to exhibit features of a market that is inefficient. The present study attempts to examine a gap in the research on market efficiency that could potentially shed further light on investor reaction to market-induced surprises. While almost all of the previous studies focus on individual security prices or market-level stock indices, there has been little research conducted on the potential differences between the price movements of a composite stock market index like the NASDAQ Composite and the price movements of the underlying sub-indices within that composite market index. Using daily returns between the mid-1990s and 2008 for both the NASDAQ Composite Index and its eight component sub-indexes, we examine each index individually and also asses the similarities (or differences) between the return patterns of the composite index and its eight sub-indices. This approach will provide further insights on whether investor reaction is broad-based, universal and consistent for the composite index and its component indexes, which represent a wide variety of different industries like banking, telecommunications and computers, or whether investor reaction varies among the industries represented by the subindices. If different stock market sectors attract different classes of investors, this could lead to differences in investor reaction by NASDAQ sub-index, and our study will therefore help determine the degree to which investor reaction is generally consistent and universal or whether it varies by market segment. To study investors reaction to surprises in the NASDAQ Composite Index and its eight component indexes, we first estimate GARCH (1, 1) models for the daily returns for each index and then compute a time series of standardized residuals from each of these models. We identify significantly favorable market surprises as those shocks that result in standardized residuals greater than 2.50 in absolute value. This procedure provides us with approximately 25 favorable and 25 unfavorable surprises for each index during the sample period (both the number of surprises and sample periods vary slightly by subindex). We then follow a procedure proposed by Brown, Harlow and Tinic (1988) and calculate cumulative abnormal returns (CARs) during a 30-day window after each event to empirically test whether investor reaction is consistent with the EMH, OH or UIH for the nine NASDAQ indexes (composite and 8 sub-indexes). Our empirical findings are somewhat mixed, and although we find strong statistical support for the UIH in the cases of the NASDAQ Composite and four sub-indexes, we also find evidence of the OH for two subindexes, and the EMH for one index, with one index remaining unexplained. These mixed empirical results imply that investor reaction is not universal across industry sectors in the NASDAQ, and in fact varies by sector, highlighting the value of analyzing investor reaction in different segments of the security markets. This paper is organized as follows. In the next section we describe out data and also outline the specific methodology used to empirically examine investor reaction. Section III follows to discuss our empirical results and the major findings of our analysis. Finally, in Section IV we present our summary and conclusions.

5 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 101 II. DATA AND METHODOLOGY a. Data We use daily closing values for the NASDAQ composite index and its eight component sector indices to compute daily returns for each of these nine indexes. Formally, daily returns (R it ) are calculated as: R it = Ln ( I it / I it-1 ) * 100 (1) Where: R it = daily return of index i on day t such that i = 1, 2, 3, 4, 5, 7, 8, 9. I it = closing value of index i on day t. I it-1 = closing value of index i on day t-1. Ln = natural log. To test for the stationarity of these series (R it ), we conducted augmented Dickey-Fuller unit root tests on the returns from each index and the results (not presented here) indicate that all series are stationary in their first differences. Table 1 displays the summary statistics of daily returns for the NASDAQ composite and its eight sub-indices. Compared to the % mean daily return for the overall NASDAQ composite index, five sectors averaged lower returns (banks, industrials, insurance, telecommunications, and transportation) and three sectors averaged higher returns (biotechnology, computers, and other financials). As can be seen, the computer sector index recorded the highest daily mean return (0.055%) and standard deviation (2.1172%) while the daily mean return for the telecommunications sector (0.0231%) generated the lowest return among the sector indices with a relatively high standard deviation (2.01%). The bank sector index registered the lowest daily return volatility (1.05%) with a daily mean return of %. b. Methodology The methodology of this paper contains the following steps: i. Identification of Market Surprises In order to identify market surprises, we first estimate a GARCH (1, 1) model for each index using the calculated daily returns. To remove any serial correlations in the residuals, we incorporate an optimal number of autoregressive lags in each equation using standard time series techniques. To determine market surprises, we next compute standardized residuals from the estimated GARCH (1, 1) models for the composite index and each of the eight component indexes. We identify significantly favorable market surprises as shocks that result in standardized residuals above the value of 2.50 on the day of the news, and significantly unfavorable market surprises as shocks that result in standardized residuals below the value of This procedure provide us with 25 favorable and 25 unfavorable surprises for the NASDAQ composite index, and a total of 181 favorable and 182 unfavorable surprises for the NASDAQ component sub-indices (an average of 22.6 positive shocks and negative shocks per sub-index). Table 2 displays the distributions of identified favorable and unfavorable market surprises for the NASDAQ composite index and its eight subindices. As can be seen in Table 2, the number of market surprises ranges from 18 to 25 favorable shocks, and from 19 to 25 unfavorable shocks, depending on the specific sub-index.

6 102 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY ii. The Post- Market Surprise Volatilities of Returns In order to determine whether market surprises result in a higher volatility of market returns, we track daily returns over a 30-day window following each favorable and unfavorable market surprise for each index. We then calculate and compare the variance of all 30- day post-event periods (following both favorable and unfavorable events) to the variance of non-event days that is, the entire sample period excluding the post-event days, where the variance (Var) is computed using the following standard formula: 2 N J Var t = ( 1/ 1) Rit R ij N j (2) t= 1 where: Var t = the variance of daily returns during time period t. N j = number of days in each category j (j = 1 for all post-market surprise days, 2 for favorable market surprise, 3 for unfavorable market surprise, and 4 for non- market surprise days). R it = daily return of stock index i on day t (i = 1..9). R ij = the average return for each category (post-market surprise or non- market surprise days). We next perform difference-of-variance tests and calculate F-statistics to compare the volatility of post-market surprise days to the volatility of non-market surprise days. The null hypothesis of these tests is that the variances of returns during the post-market surprise windows are equal to (and not significant different from) the variance of returns for the nonmarket surprise days. The rejection of the null hypothesis would indicate that there is a statistically significant difference between the volatility of returns during post-market surprise windows and the volatility of returns in periods that do not follow a market surprise. Based on the UIH hypotheses that the arrival of unexpected information typically generates higher uncertainty and increases post- market surprise volatility, one would then expect the variance of returns during post-market surprise windows to be statistically greater than the variance of returns in non-market surprise windows. Furthermore, we employ similar procedures and statistical techniques to test for any differences between the variances of periods following favorable market surprises and the variances following unfavorable market surprises. iii. Statistical Tests for EMH, OH, and UIH To investigate whether the reactions of investors trading securities based on the NASDAQ composite index and its component indexes react to market surprises in ways that are consistent with the predictions of the EMH, OH or UIH, we employ a procedure used in Mehdian, Perry and Nas (2007). Specifically, we calculate the daily post-event abnormal returns for each index and average them cross-sectionally for all days over the 30-day period following each set of favorable or unfavorable market surprise in each index. Then, we add these 30-day abnormal returns consecutively from day 1 through day 30 to construct cumulative abnormal returns (CARs) for both sets of market surprises (favorable and unfavorable) for each index for the 30-day post-event window period. Formally, let AR itd be the abnormal return for index i on day t following an market surprise d such that t = days, so that:

7 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 103 AR itd Ritd Rnon, i = (2) where d = 1 n, and n denotes the number of favorable or unfavorable market surprises for index i. R itd = Return of Index i on day t for market surprise d. K non, i = Mean return for Index i (i = 1 9) for non-market surprise days. The mean abnormal return on day t is computed as: n AK it = ( 1/ n) AK itd, t = (3) d = 1 The CAR for each Index i is calculated by summing the mean abnormal returns over t days so that: CAR it = CARi( t 1 ) + AK it, and t= 1 30 (4) Following Ruback (1982), the statistical significance of the cumulative abnormal returns are tested by conducting a t-test of the null hypothesis that the CARs for each day are equal to zero during the post-surprise windows for day 1 to day 30. This t statistic follows a Student-t distribution and is calculated as: CARit t Value =, such that (5) Var CAR ) ( it Var ( CARit ) = d Var( AK it ) + (2( d 1) Cov( AK it, AK i( t+1) ) Moreover, we display the graphical representations of CARs for 30 days following favorable and unfavorable market surprises for each index to help determine whether investor reactions to market surprises are consistent with predictions of the EMH, OH or UIH. III. EMPIRICAL RESULTS Table 3 displays daily mean returns for non- market surprise days, all post-market surprise days, favorable post-market surprise days, and unfavorable post-market surprise days, along with the number of days for each sample period in parenthesis. As can be seen in Table 3, the post-market surprise daily mean returns are higher than the non-market surprise daily mean returns, not only for the NASDAQ composite index, but also for all sub-indices except two: the Industrial and Transportation sectors. In addition, as the figures in Table 3 suggest, the daily mean returns for favorable post-market surprise days are higher than the daily mean returns for unfavorable post-market surprise days for four sub-indices: Biotech, Industrial, Financial, and Telecommunication. For the rest of sub-indices as well as the NASDAQ composite index, the daily mean returns for favorable post-market surprise days are less than the daily mean returns for unfavorable post-market surprise. Table 4 shows the variances of daily returns for non-market surprise days, all postmarket surprise days, post-favorable market surprise days, and post-unfavorable market surprise days, along with the sample size for the NASDAQ Composite index and its subindices and two columns of F-statistics. In the first column of F-statistics (column 4 of Table 4), the first F-statistic labeled (a) is for the test of the null hypothesis that the variance of returns for all-market surprise days is equal to the variance of returns for non-market surprise days for each index. The second F-statistic labeled (b) is for the test of the null

8 104 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY hypothesis that the variance of returns during post-favorable surprise days is equal to the variance of returns during non-surprise days for each index. The third F-statistic labeled (c) is the test of the null hypothesis that the variance of returns for post- unfavorable surprises is equal to the variance of returns for non-surprise days for each index. The F-value displayed in the last column of Table 4, labeled (d), is for the test of the null hypothesis that the variance of returns for post-favorable market surprises is equal to the variance of returns for post-unfavorable market surprise days. As can been seen in Table 4, the F-statistics indicate that the variance of returns for post-market surprises is statistically significantly higher than the variance of returns for nonmarket surprise days for all of the NASDAQ indices except for Computers. Therefore, these findings provide evidence to indicate that the volatility of market returns increases significantly in the days following post-market surprises. The findings also suggest that the variance of returns for post-favorable surprise days is statistically significantly higher than the variance of returns for non-surprise days for five NASDAQ sub-indices: Biotechnology, Industrial, Insurance, Telecommunication, and Transportation. On the other hand, the null hypothesis that the variance of returns of post-unfavorable surprises is equal to the variance or returns of non-surprise days is rejected for the NASDAQ composite index and all subindices, which is a sign that market volatility following unfavorable market surprises is significantly higher than the volatility of non-market surprise days. All in all, the findings presented in Table 4 support the hypothesis that the arrival of market surprises results in higher post-surprise market volatility and uncertainty of daily returns. This, of course, is consistent with the prediction of the UIH, which claims that market risk increases during a post-market surprise period. We note by reviewing the last column of Table 4 that the F-statistics are statistically significant for all indices except for the Biotechnology and Telecommunication indices, implying that post-market surprise uncertainty is generally significantly higher in the periods following unfavorable market surprises compared to the uncertainty following favorable market surprises. Tables 5a and 5b show the post-market surprise cumulative abnormal returns (CARs) next to their related t-values for each of the 30 days following the favorable market surprises, and Tables 6a and 6b display the same results for unfavorable market surprises. We test the null hypothesis that the CARs are equal to zero using t-tests based on the formula in Equation 5. In addition, the graphical representations of the CARs for 30-day windows are displayed in Figures 1-9 for each of the NASDAQ indexes. The numbers in Tables 5(a), 5(b), 6(a), and 6(b), and the graphs in Figures 1-9 provide a set of identifiable patterns of daily returns for all nine indices that allows us to determine whether the cumulative returns following market surprises are consistent with the predictions of the EMH, OIH or UI. Specifically, for the NASDAQ Composite index, and four of the component sub-indexes (Banks, Biotechnology, Other Financials, and Telecommunications), the patterns of the CARs are consistent with the prediction of the UIH, because in cases of both favorable and unfavorable market surprises we observe an upward trend in the CARs following the arrival of market surprises. This pattern of daily stock returns implies that the arrival of market surprises creates uncertainty, such that market participants initially price stocks below their fundamental values. However, as time passes and the uncertainly generated by market surprises gradually dissipates, stock prices slowly converge to their fundamental values.

9 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 105 A further examination of Tables 5 and 6 reveals that the trends of the CARs are in line with the prediction of the OH for the Computer and Insurance sub-indices, since we observe downward trends in the CARs following the arrival of favorable market surprises and upward trends in response to the arrival of unfavorable surprises. Therefore, in the case of these two sector indices the reactions of investors following the arrival of market surprises can reasonably be characterized by subsequent price-reversals, implying that a contrarian trading rule of buying current losers and selling current winners may result in higher-thannormal, risk-adjusted returns. It should be noted that the predictions of the OH and UIH are exactly the same in the case of unfavorable market surprises, so it is really the pattern of returns following favorable market surprises that allows us to distinguish between the OH and UIH. Finally, the results for both the Industrials and Transportation indexes are somewhat mixed, while one can plausibly interpret the behavior of CARs in these indices as being consistent with the prediction of the EMH, at least for the Transportation index, where we observe a set of trendless CARs. In order to test the robustness of the above empirical results and our conclusions based on those findings, we next perform OLS estimations of the CARs regressed on a time trend for both favorable and unfavorable market surprises, and those results are presented in Table 7. As can be seen, the trend coefficients for the CARs following both unfavorable and favorable market surprises are positive and statistically significant for the NASDAQ Composite index and the component indexes for Banks, Biotechnology, Other Financial, and Telecommunications, findings which are consistent with the predictions of the UIH. Additionally, we observe that the trend coefficient for favorable market surprises is negative and the trend coefficient for unfavorable surprises is positive for both the Computer and Insurance component indices: a result consistent with the predictions of the OH. Finally, the conclusions made above regarding the Industrial and Transportation indices are supported by the empirical OLS results presented in Table 7, which indicate no significant trends for the Transportation sector following either favorable or unfavorable surprises (mildly supporting EMH), and a significant negative trend following unfavorable news for the Industrial sector (not predicted by any of the theories presented here). IV. SUMMARY AND CONCLUSIONS In this study we investigated the reaction of investors to the arrival of unexpected information (both positive and negative) for the NASDAQ Composite Index and its sub-index components that represent a variety of industry sectors. Daily stock returns from the NASDAQ composite index and its eight composite indexes (Banks, Biotechnology, Computers, Industrial, Insurance, Other Financials, Telecommunications and Transportation) over sample periods from the mid-1990s through 2008 are used to test three competing hypotheses of investor reaction to the arrival of unexpected information: the Efficient Market Hypothesis (EMH), the Overreaction Hypothesis (OH), and the Uncertain Information Hypothesis (UIH). Although our empirical findings are somewhat mixed, we find strong statistical evidence of a corrective process of significantly positive cumulative abnormal returns following the arrival of both favorable and unfavorable information for the NASDAQ Composite and four of the sub-indexes (Banks, Biotechnology, Other Financials, and Telecommunications), and these outcomes are consistent with the prediction of the UIH regarding investor behavior. Specifically, the empirical findings for these five indexes

10 106 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY suggest that investors in these markets systematically set security prices below their fundamental values in response to unexpected economic information. Such behavior is rational according to the UIH, since the arrival of unexpected information (whether favorable or unfavorable) makes the equity market a more risky environment. For the Computer and Insurance sub-indices, we observe downward trends in the CARs following the arrival of favorable market surprises and upward trends in response to the arrival of unfavorable surprises, consistent with the predictions of the OH. Therefore, in the case of these two sector indices, market surprises are followed by subsequent pricereversals, implying that a contrarian trading rule of buying current losers and selling current winners may result in higher-than-normal, risk-adjusted returns. For the remaining two sectors (Industrials and Transportation), the pattern of CARs following the arrival of unexpected information is not strongly consistent with any of the theories of investor reaction presented here, although the trendless pattern of returns following market surprises for the Transportation index could be explained by the EMH. One main implication of these mixed empirical results is that investor reaction is not universal across industry sectors, and in fact varies significantly by sector, highlighting the value of analyzing investor reaction in different segments of the security markets in addition to investigating composite indexes. The findings presented here suggest that further research in this area is warranted to shed light on the dynamics of why and how investor reaction varies by market segment. References [1] Ajayi, R. A., S. Mehdian and M. J. Perry (2006), A Test of US Equity Market Reaction to a Surprises in an Era of High Trading Volume, Applied Financial Economics, Vol. 16, [2] Bremer, M. and R. J. Sweeney (1991), The Reversal of Large Stock-Price Decreases, Journal of Finance, Vol. 46, [3] Brown, K. D. and W. V. Harlow (1988), Market Overreaction: Magnitude and Intensity, Journal of Portfolio Management, Winter 1988, [4] Brown, K. D., W. V. Harlow and S. M. Tinic (1988). Risk, Aversion, Uncertain Information, and Market Efficiency, Journal of Financial Economics, Vol. 22, [5] Chopra, N., J. Lakonishok and J. R. Ritter (1992), Measuring Abnormal Performance: Do [6] Stocks Overreact? Journal of Financial Economics, Vol. 31, [7] DeBondt, W. F. and R. H. Thaler (1985), Does the Stock Market Overreact? Journal of Finance, Vol. 40, [8] DeBondt, W. F. and R. H. Thaler (1987), Further Evidence on Investor Overreaction and Stock Market Seasonality, Journal of Finance, Vol. 42, [9] Howe, J. S. (1986), Evidence of Stock Market Overreaction, Financial Analysis Journal, Vol. 42, [10] Kahneman, D. (2003), A Psychological Perspective on Economics, American Economic Review, May 2003, 93(2), [11] Kahneman D. and A. Tversky (1973), On the Psychology of Prediction, Psychological Review 80,

11 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 107 [12] Liang, Y. and D. J. Mullineaux (1994), Overreaction and Reverse Anticipation: Two Related Puzzles? Journal of Financial Research, Vol. 17, [13] Mehdian, S., M. J. Perry and T. Nas (2007), An Examination of Investor Reaction to Unexpected Political and Economic Events in an Emerging Market: The Case of Turkey, Global Finance Journal, Volume 18, [14] Ruback, R. S. (1982), The Effect of Discretionary Price Control Decisions on Equity Values, Journal of Financial Economics 10, No. 1, Table no. 1 Summary Statistics for the NASDAQ Composite Index and Sector Sub-Indices INDEX Days Mean Daily Return Std. Dev. Maximum Return Minimum Return NASDAQ Composite % 1.65% 14.17% -9.67% (9/30/1994) NASDAQ SECTOR SUB-INDICES Banks % 1.05% 9.95% -5.86% (9/30/1994) Biotechnology % 2.08% 10.81% % (5/1/1996) Computers % 2.12% 18.07% % (7/28/1995) Industrial % 1.50% 9.94% % (9/301994) Insurance % 0.97% 5.51% -4.58% (9/30/1994) Other Financials % 1.74% 12.08% % (1/6/1997) Telecommunications % 2.02% 17.52% -9.89% (5/13/1996) Transportation % 1.34% 7.20% % (9/30/1994) The sample periods above are from the inception dates of the sub-indexes (in parentheses above), and from September 30, 1994 for the composite index, through August 29, 2008 for all indexes.

12 108 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY Table no. 2 Number of Favorable and Unfavorable Market Surprises Identified for the NASDAQ Composite Index and Its Eight Component Sub-Indices Index Favorable Unfavorable COMPOSITE BANKS BIOTECHNOLOGY COMPUTERS INDUSTRIALS INSURANCE OTHER FINANCAL TELECOMMUNICATIONS TRANSPORTATION 24 24

13 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 109

14 110 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY

15 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 111

16 112 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY

17 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 113

18 114 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY

19 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 115 Table no. 7 Regression Analysis of CAR Trends for Favorable and Unfavorable News Index Composite Banks Biotechnology Computer Industrials Insurance Other Financial Telecommunications Transportation Type of Market Surprise Coefficient for Trend Adjusted R-Squared Favorable *** 0.56 Unfavorable *** 0.77 Favorable *** 0.55 Unfavorable *** 0.64 Favorable *** 0.77 Unfavorable *** 0.82 Favorable Unfavorable *** 0.38 Favorable * 0.11 Unfavorable *** 0.80 Favorable ** 0.15 Unfavorable *** 0.81 Favorable *** 0.90 Unfavorable ** 0.20 Favorable Unfavorable ** 0.13 Favorable Unfavorable *** indicates statistical significance at the 1% level, ** at the 5% level and * at the 10% level.

20 116 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY Post-Market Surprise Cumulative Abnormal Returns (CARs) for NASDAQ Sector Indices

21 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 117

22 118 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY

23 A Cross-Industry Analysis of Investors Reaction to Unexpected Market Surprises 119

24 120 Peter J. BUSH, Seyed M. MEHDIAN, Mark J. PERRY