THE LONG-TERM PRICE EFFECT OF S&P 500 INDEX ADDITION AND EARNINGS QUALITY

Transcription

1 THE LONG-TERM PRICE EFFECT OF S&P 500 INDEX ADDITION AND EARNINGS QUALITY Abstract. This study suggests that inclusion of a firm to the S&P 500 index strengthens managerial incentives for high-quality financial quantitative disclosure, which could explain the observed temporary and long-term price effects of a firm being listed. Discretionary accruals are significantly larger in magnitude before index listing, which signals higher information risk. Five years after S&P 500 addition, discretionary current accruals are about half the magnitude at the time of inclusion. Previous literature posits that investors fixate on earnings itself and overweight the accrual component of current earnings when forecasting future earnings. Empirical studies suggest a profitable arbitrage strategy based on accruals. Our results, however, are consistent with the latest findings that information risk, measured with discretionary accruals, is a priced factor. 1

2 2 The S&P 500 index is the most widely used benchmark for measuring the performance of largecapitalization U.S.-based stocks. It includes U.S. stocks that constitute about 75 percent of the market capitalization of all regularly traded stocks on the U.S. equity market 1. From its inception in 1957 through 2005, the total number of firms added to the index is 957. Standard and Poor s (S&P) introduces changes to the index based on public criteria: the firm should be profitable and a leader in an important U.S. industry, with sufficient liquidity and a relatively diverse ownership structure. In the last decade, about three-quarters of all changes to the index were after involuntary deletions, due to mergers, bankruptcies or other restructuring events. Other involuntary changes occurred because the firms ceased to represent the U.S. economy, either because the industry was no longer representative of the U.S. economy or because the firm was no longer representative of its industry. Both voluntary and involuntary deletions from the index demand additions to the S&P 500, which empirical findings suggest generates significant abnormal returns: between 2.7% (Shleifer, 1986; Harris and Gurel, 1986) and 5.48% (Beneish and Whaley, 1996). In contrast, deletions from the S&P 500 are associated with negative abnormal returns that are larger in magnitude compared to additions: between -10.8% (Beneish and Whaley, 1996) and -12.6% (Lynch and Mendenhall, 1997; Chen et al., 2004). The significant abnormal returns that follow the firm s addition to the S&P 500 (henceforth, referred to simply as the index effect ) have received different interpretations. Information asymmetry and trading costs could explain the abnormal returns; empirical findings, however, are not always consistent with this trading explanation. If the change to the index conveys new information about a firm s future returns, this would also justify the positive price response to the S&P 500 addition. Standard and Poor s rejects this possibility and assures that investment merits of the firm do not influence the selection procedure; nevertheless, empirical studies suggest that there might be new information disclosed with an S&P 500 listing. Dhillon and Johnson (1991) confirm that the information is relevant regarding a security s fundamental value; the option price should not be affected by temporary price movement, but it changes with the S&P addition, and this would occur if new information about the expected distribution of a security s future returns has been released. Jain (1987) argues that it is the S&P reputation that causes the index effect; S&P is known to closely monitor the firms in its indices and its preference towards

3 3 more stable firms could signal to investors that new additions may provide better future returns at a lower risk compared to their market rivals. Denis et al. (2003) focus on the managerial incentives and argue that it is the increase in investors scrutiny that indirectly affects the abnormal returns; because of it, managers of newly indexed S&P firms may devote more time and effort to their companies and, hence, improve the financial performance of their firms to which the stock market responds with a price revision. This study contributes to the debate about the information content of the S&P 500 addition. It recognizes that media attention and investors scrutiny may affect not only managerial performance but also its incentive for high-quality financial disclosure. We argue that after the index listing the information risk, measured by earnings quality, is lower and this explains, along with other factors, the price response to the S&P 500 addition. Denis et al. (2003) argue that investors earnings expectations for newly added firms capture the index effect and, hence, investigate earnings per share forecasts and forecast errors. They explain the upward revision of earnings expectations with earnings improvements because of greater managerial efforts after the S&P 500 addition. We argue that earnings quality may also affect analysts forecasts and their accuracy. The quality of financial reporting can significantly improve the accuracy of analysts predictions. If the index listing reduces earnings management, information risk would lower, demanded returns would decrease and analysts would have more precise financial reports on which they base their estimates. Altogether, this chain of events should contribute to the positive price reaction. If the S&P inclusion itself strengthens managerial incentives for high-quality financial disclosure, this may contribute to the price effect of the S&P 500 additions. We analyze earnings of 202 additions to S&P 500 during the period and draw conclusions about their quality before and after their listing. There are two possible ways in which earnings may be related to the price response that is observed when a stock enters the S&P 500: (1) the quality of earnings and its components may itself be different for these firms, and (2) the market may respond differently to reported earnings for S&P 500 firms. We use current accruals, as a measure of earnings quality, which we decomposed to its expected and discretionary component. Our results confirm that the index listing has a positive effect on managerial

4 4 discretion over accruals, which is significantly lower for S&P 500 firms. After, we examine the long-term effect of the index addition and associate the listing event with the stock returns. We investigate the association between market-adjusted returns and the S&P 500 listing and find results consistent with the latest findings of possible pricing of accruals quality. Literature Survey and Motivation The S&P effect and abnormal returns. Numerous empirical studies examine the effect of index changes on returns and trading volume. Because authors cumulate abnormal returns (CAR) over different window periods (for instance, one to 60 days after the effective inclusion date), the results are not directly compatible. Current findings confirm that new additions to the S&P 500 from 1976 to 1983 had 2.79% abnormal returns (at the inclusion date; Shleifer, 1986), inclusions from 1984 to 1988 had 3.55% CAR (at the announcement date and the day after; Dhillon and Johnson, 1991), additions from 1989 to 1993 had 5.90% CAR (cumulative from the announcement date to the day after the addition; Beneish and Whaley, 1996) and from 1989 to 2000 had 5.44% abnormal returns (at the announcement day; Chen et al., 2004). Current research suggests five competing explanations of the index effect: price pressure, imperfect substitution, liquidity, investor awareness and the certification (information) hypothesis. The long-held assumption that stocks have perfect substitutes, and that perfect elasticity of demand follows, is central to modern finance theory. If securities have a close to perfectly elastic demand curve, then any shock that is devoid of information should not have any (long-term) impact on prevailing prices. The increased demand resulting from index funds and others adding the stock to their portfolios explains how the prices of newly added firms can be momentarily affected after the inclusion. Once the excess demand is satisfied, the index effect should dissipate. This price-pressure explanation argues that stocks have a short-term downward sloping demand curve and that abnormal returns of additions are primarily due to indexing. Shleifer (1986) claims that the price movement is an artifact that demand curves of stocks, likewise other (capital) goods, is downward sloping; index managers replicate the S&P 500 index and this

5 5 generates demand that increases the subsequent stock price. Harris and Gurel (1986) empirically support the findings of price and volume reversals for S&P 500 stocks. Other authors, however, claim that the index effect has a long-term impact on price and volume. In theory, stocks, bonds, puts and calls could be close substitutes for one another; but in practice, different tax treatments and transaction costs can affect investors preferences and render these four options imperfect substitutions. The imperfect-substitute (or long-term downward sloping demand curve) explanation proposes that S&P 500 stocks are not easily replaceable, because they may differ in investment risk and, thereby, trading costs. If S&P 500 firms do not have a substitute, the index effect on price should be at least partially permanent. Wurgler and Zhuravskaya (2002) suggest that the lack of a substitute explains the limited arbitrage with S&P 500 stocks and that risk prevents arbitrage from completely flattening the demand curve of these stocks. Not only the slope of a demand curve, but also liquidity changes accompanying additions to the S&P 500 index are examined in empirical studies. Liquidity can improve without information production if trading volume increases. If market makers face lower inventory costs, overall trading costs decrease and the bid-ask spread reflects this liquidity effect. The liquidity explanation suggests that abnormal returns are due to the decrease in trading costs that follows index inclusion. Erwin and Miller (1998) confirm that the addition of non-optioned stocks, but not of optioned stocks, has a positive effect on liquidity and bid-ask spreads, which is consistent with the liquidity hypothesis. Information asymmetry and trading costs also enter the investor-awareness explanation. With segmented markets, investors hold portfolios that are not completely diversified. Following index inclusion, more investors become informed about a security and hold it for its diversification potential. This explanation proposes that, with the increase in investors awareness, demanded returns decrease due to more informed asset pricing, which causes the price to rise. Chen et al. (2004) emphasize that the generally held belief that S&P 500 firms are well-known is not necessarily true. Their study confirms that increased awareness following S&P additions enhances monitoring by financial analysts, reduces the information asymmetry component of the bid-ask spread, improves access to capital markets, and reduces the cost of under-diversification.

6 6 The certification (information) explanation also posits that index inclusion conveys positive information, but not that it is related to trading costs. Rather, the index addition itself leads to a revision in analysts forecasts about the financial prospects of a firm due to its information content. As mentioned before, Jain (1987) finds excess returns for firms in S&P supplementary indices, even though hedge funds do not try to match them (i.e., the price pressure cannot explain the observed market response). Therefore, S&P s preference towards more stable firms might signal a reduction in the riskiness of firm s securities, which would explain abnormal returns. Denis et al. (2003) find that intensified monitoring makes managers of S&P 500 additions more efficient, which explains excess returns; whereas Dhillon and Johnson (1991) confirm that corporate bonds respond positively to the listing announcement and that the listing suggests new information about a security s fundamental value. Accounting earnings, earnings quality and abnormal returns. The academic literature offers at least two empirical pieces of evidence that reported earnings may not be an objective measure of firm performance due to managerial discretion over its components: (1) earnings manipulation may involve not only accruals with no direct cash flow consequences (for instance, asset write-offs), but also management of real activities with direct cash flow consequences, such as overproduction (Roychowdhury, 2006); and (2) managers may often use their discretion to mislead investors and, therefore, (discretionary) accruals might be regarded as opportunistic (Guay, Kothari and Watts, 1996). However, does this mean that financial analysts can accurately forecast stock returns with financial reports that simply record cash transactions, without accounting manipulations? A series of empirical papers confirms that the properties of earnings and other accounting variables - more notably, accruals - make earnings useful in valuation (e.g., Dechow, 1994) and contracting (Christensen et al., 2005). These scholars have examined time-series properties of accounting information that are relative to its ability to forecast future value-relevant variables and argue that various properties of accounting earnings might yield a more accurate forecast of earnings. These desirable properties are often referred to in the existing literature as earnings (accounting or accruals) quality. Although the concept is still not precisely defined, there are numerous dimensions of earnings quality that empirical studies suggest might enhance investment efficiency. Among earnings quality measures are theoretical

7 7 constructs, sometimes measured jointly, such as (1) accounting conservatism, which captures how timely economic losses are incorporated into the accounting income (Ball et al., 2000); (2) loss avoidance, which can explain how managers report less negative earnings, increase quarterly earnings and beat analysts expectations around the zero with positive financial results (Burgstahler and Dichev, 1997; DeGeorge et al., 1999); (3) earnings smoothing, which reduces earnings volatility over reporting periods (Leuz et al., 2003); and (4) price sensitivity to earnings (i.e., the earnings response coefficient in the earningsprice relation), which suggests how prices adjust to reflect changing expectations about a firm s earnings generating ability (Beaver, Clarke, and Wright, 1979). Although earnings quality is a multi-dimensional concept, empirical research on the topic in the last decade has primarily focused on the role of accruals. Studies link the properties of earnings and its quality with the financial market microstructure (e.g., insider trading; Aboody et al., 2005), with managers decisions (e.g., investment activities and stock split decisions; Louis and Robinson, 2005), and with firm-specific characteristics (e.g., corporate governance structure; Wang, 2006). Accruals distinguish accounting from mere counting of cash and, without them, we would not distinguish cash flow movements associated with a change in the firm s fundamental from cash flows arising from managers ordinary operating investment decisions. However, accruals allow managers discretion; in particular, because the timing of cash transactions does not necessarily match the timing of economic transactions. This dual role of accruals motivates research on earnings decomposition, which tries to disentangle the factors that drive changes in accruals. Probably the most common implementation of earnings management is based on Jones (1991) discretionary accruals model. With a time-series model, Jones estimates an expected or normal level of accruals, and uses the residual as a measure of the discretionary accruals. Recent studies have suggested that there are factors (e.g., firm characteristics), that the Jones model does not consider, although these affect normal accruals (e.g., Ashbaugh, LaFond and Mayhew, 2003). The original or modified Jones discretionary model serves as a proxy for earnings management in numerous studies. Examples include the findings of greater earnings management in firms where executives have a greater sensitivity to equity price movements (Bergstresser and Philippon, 2006), and in firms where the executives are former audit partners (Menon and Williams, 2004).

8 8 Numerous scholars have examined whether financial markets price earnings quality, but empirical findings are still not conclusive. Sloan (1996) posits that investors naively fixate on earnings and overweight the accrual component of current earnings when forecasting future earnings. His results show that a hedge strategy based on accruals earns size-adjusted abnormal returns of 10.4% in the year following portfolio formation, on average, for the time period Since Sloan (1996), a large body of empirical studies has examined which factors are behind the accruals anomaly and how it is possible to find a profitable trading strategy with public information if financial markets are efficient (e.g., Xie, 2001; Hirshleifer et al., 2004; Mashruwala et al., 2006). Scholars have examined various components of accruals to detect which of them contributes significantly to the accruals anomaly (e.g., Xie, 2001). Several empirical studies suggest that the accruals anomaly exists, at least, because (1) there are high transaction costs and arbitrage risk, which explains why the accruals anomaly is not arbitraged away (Mashruwala et al., 2006), and (2) investors either do not have the time to process the large amount of information or do not find it cost-efficient to treat earnings components differently ( limited attention theory, Hirshleifer et al., 2004). Recent studies have questioned previous findings about the accruals anomaly and have related arbitrage strategies suggesting that investors price accruals quality. Francis et al. (2005) find that poorer accruals quality is associated with larger costs of debt and equity. Kraft et al. (2005) replicate with robustness checks three studies claiming the mispricing of accruals and conclude that results are biased towards the investor s fixation hypothesis. Method and Data The original discretionary model suggested by Jones (1991) estimates expected accruals as the fitted value from a regression of total accruals on lagged firm size, the change in firm sales, and gross property plant and equipment scaled by total firm assets. Empirical research over the last decade suggests that current accruals should be preferred and their normal level should be estimated with performance-adjusted accruals models. We use the performance-adjusted accruals method as suggested by Ashbaugh, LaFond and Mayhew (2003). Although we could not apply their methodology with the portfolio technique, our estimation benefits from advantages common to the design of event studies; when we examine the level

9 9 of managerial discretion over accruals before and after the S&P 500 addition, we can use the firm as a control of itself. Our model estimates expected accruals as the fitted value from a regression of current accruals on lagged firm size, the change in firm sales, and the change in accounts receivable scaled by total firm assets. The residual, which we use as a measure of managerial discretion, contains the current accruals variance that is not explained by the model. See Appendix A for more details on the earnings decomposition. During the period , the S&P Corporation announced 419 additions and deletions from the S&P 500. There are two sources for the S&P 500 index changes: (1) the S&P corporate website, with data for the period , and (2) the database accessible at the website of Journal of Finance and disclosed by Chen et al. (2004), with data for the period The main requirement for our sample was to have financial information, available in Datastream Thompson, for at least three consecutive years around the index listings announced in the period The final sample comprises 202 additions. Financial information is obtained from Datastream Thompson. Results Earnings quality around S&P 500 additions. The quality of reported earnings could change with the listing event. To capture the effect of the index addition, this section first examines how current discretionary current accruals (DisAcc) evolved over time. The following section describes information content of DisAcc, which the association between current earnings (and its components) and future earnings reveals. Earnings components and S&P 500 additions. Existing literature still has not resolved the debate around how informative earnings are in comparison to realized cash flows. The success of a firm ultimately depends on its ability to generate cash receipts, but realized cash flows are not usually used as a summary measure of firm performance in executive compensation plans, in debt covenants, nor by investors and creditors. The information asymmetry problem between the firm s management and outside parties contracting with the firm creates the demand for a measure of firm performance which, in contrast to realized cash flows, is not strongly influenced by the timing of cash receipts and disbursements. The

10 10 accrual process provides rules on the timing of cash flow recognition in earnings so that earnings will more closely reflect firm performance 3. Current findings suggest that: (1) for short-term performance (i.e., interim and annual evaluation), earnings are more strongly associated with stock returns than are realized cash flows; and (2) when accruals are large in magnitude (either positive or negative), earnings will more closely reflect firm performance than realized cash flows (Dechow, 1994). Table 1 presents average earnings and its components before and after S&P 500 additions (see Appendix 1 for details about the current earnings decomposition procedure). In contrast with the findings reported by Denis et al. (2003), these results suggest that on average newly added firms do not have higher earnings after the listing. The S&P 500 inclusion, however, could be associated with an increase in the cash flow generating potential of added firms. Although before the addition cash flows represent on average 9.4% of total assets, after the listing year they increase significantly to 11.2%. Along with the increase in cash flows, total accruals decrease further, representing 3.6% of total assets before and 5.4% after index listing. Both components of current accruals are significantly different before and after the index listing. Managerial discretion over accruals is lower after a stock enters the S&P 500, when the financial performance of a firm (namely changes in revenues and accounts receivable) explains the accruals component of reported earnings. [Table 1 about here.] Graph 1 exhibits how the earnings structure changes after the index listing. Cash flows increase gradually after being added to the S&P 500. Negative accruals that predominantly comprise total accruals, however, offset this change in the cash flow component and explain how the increase in reported earnings is not as large as in realized cash flows. Managerial discretion over accruals is lower after the firm enters the S&P 500 index. Graph 2 exhibits the trend in expected and discretionary accruals in a five-year period around the listing event. Discretionary current accruals (DisAcc) decrease consistently over the period; the only exception is the third year after the listing, when they rise slightly. Five years after S&P 500 addition, DisAcc are about half the magnitude at the time of inclusion. [Graph 1 and 2 about here.]

11 11 Information content of the earnings components. The earnings and its components change with the addition of a firm in the S&P 500, but how does this affect its information content? Table 2 presents the accuracy of current earnings and its components in predicting future earnings before and after an S&P 500 addition. We examine the informative role of current earnings (in Model 1), cash flows and accruals (in Model 2), and the three components of earnings (in Model 3) in explaining the variance in future earnings. The firm performance in the current period is affected by a number of factors, among them most likely being its past performance. The association between current and future earnings (the regression coefficient of earnings in Model 1 in Table 2) is positive, suggesting that the financial performance of a firm is determined, among other things, by its past financial results. It is not only the past performance that affects the profitability of a firm; there are a number of other firm characteristics, for instance a strong brand image and corporate values, which may generate revenues for the firm in a longer time horizon. Hence, we expect that the (current) cash flow generating potential of a firm would also positively influence (future) reported earnings and we test this with Models 2 and 3. Our findings confirm that the relation between future earnings and cash flows (i.e., the regression coefficient of cash flows) is positive and statistically significant. Table 2 also provides empirical results that demonstrate how the S&P 500 listing changes the relation between current earnings, or cash flows, and future earnings. The addition of a firm in the S&P 500 may increase investors and media awareness and, hence, may change managerial incentives for high quality financial disclosure. If this occurs, we expect the informative role of earnings to increase with the inclusion (i.e., the relation between current and future earnings to be stronger, with a larger regression coefficient). Our regression analysis demonstrates that with the S&P 500 listing, the marginal effect of current earnings does increase: a one dollar increase in current earnings explains an increase of a 0.68 dollar in future earnings before and 0.75 dollar after an index addition. The relation between future earnings and cash flows is also slightly strengthened with the stock s addition; the marginal effect of cash flows increases by about 0.04 when cash flows and accruals explain the variance in future earnings (in Model 2) and by about 0.01 when discretionary and expected accruals, along with cash flows, are included in Model 3.

12 12 The information content of reported earnings changes with the S&P listing, but how can we conclude which measure of firm performance is preferable in contracting? Realized cash flows or total earnings. Dechow (1994) argues that when accruals are large in magnitude (either positive or negative), earnings will more closely reflect firm performance than realized cash flows. Our results suggest that the S&P 500 listing strengthens the informative role of earnings in comparison to realized cash flows. It is the size of the explained variance in future earnings across the inclusion event that demonstrates how the information content of our accounting variables changes for S&P 500 firms. When cash flows and accruals explain the variance in future earnings (Model 2) instead of current earnings (Model 1) before an index addition, the adjusted R 2 increases (from 49.2% to 53.4%). In contrast, after an index listing, the explanatory role of cash flows and accruals decreases in comparison to current earnings; the explained variance in future earnings (i.e., the adjusted R 2 ) decreases from 52.5% to 47.9%. Prior findings suggest that the size of accruals can significantly affect how informative earnings are compared to realized cash flows (Dechow, 1994). With the index listing, total accruals of S&P 500 firms not only remain negative but also significantly increase in magnitude (See Table 1 for the size of absolute expected and discretionary accruals). Our findings suggest that most likely because of them, earnings increase their informative role and closely reflect firm performance, measured by future earnings, when compared to cash flows (i.e., the explained variance decreases from Model 1 to Model 2 in the year after the inclusion). Table 2 also presents the informative role of discretionary accruals. The decomposition of current accruals does not improve the overall predictive properties of the estimation model (i.e., the explained variance decreases from Model 2 to Model 3), but the discretionary accruals are significant before and after the index listing. Our findings suggest that the association between future earnings and discretionary accruals (DisAcc) changes when a firm enters the S&P 500. While there is a positive relation between DisAcc and future earnings one year before the inclusion, it becomes negative when a firm is listed; the sign also reverses for expected current accruals, but the marginal effect is not significant for this variable. The negative relation between discretionary accruals and future earnings after the index listing could be due to the earnings structure of S&P 500 stocks. With the index addition, current accruals decrease significantly and become negative; on average, they are about 0.5% of total assets before and (-)2.6%

13 13 after inclusion. Additionally, both components of current accruals change significantly. As reported before, discretionary accruals decrease from 6.3% to 5.1% (of total assets), whereas expected accruals are about 15 times larger after the listing event. Although reported earnings decrease along with total and discretionary accruals, cash flows are significantly larger after a stock enters the S&P 500 (these findings are included in Table 1). Therefore, the negative association between future earnings and discretionary accruals confirms that when a firm is listed, and after controlling for other factors (among them being the level of cash flows), future earnings increase not because of managerial discretion over accruals but rather due to the increased realized cash flows. Long-term stock returns and S&P 500 listing. The main challenge of empirical research on S&P 500 additions is to disentangle the effect of various factors that determine the price response. This section examines cumulative market-adjusted returns (MAR) over a period of three years after the listing event. Later, the information content of discretionary accruals and how it explains MAR is detailed. Long-term price effect. Previous research about S&P 500 additions exclusively examines the short-term price effect: up to 60 days after the listing event. If long-term demand curves of newly added stocks are perfectly elastic, short-term providers of liquidity would require a temporary price premium to accommodate the excess demand by index funds during the event week. Nevertheless, new information can also explain the short-term price effect. If investors anticipate an increase in earnings quality based on their previous experience with S&P 500 additions, they would reduce demanded returns due to lower information risk after corporate disclosure of high-quality earnings, thus causing the price after an S&P 500 addition to rise. In contrast to the liquidity and price-pressure explanation, which predicts a price reversal once abnormal demand has subsided, the price effect associated with earnings quality (information risk) may not only be temporary. If the expected increase in earnings quality, which produces abnormal returns at the listing date, really allows for more informative asset pricing after an S&P 500 indexing, then the price revision could be permanent. Its magnitude would be a function of how informative reported earnings are for daily asset-pricing decisions. Therefore, besides the short-term price effect, the index inclusion may also convey positive information about a security to which investors respond gradually over the months following the S&P inclusion. Empirical research proposes that the returns to an

14 14 accruals anomaly are realized over twelve months and other pricing anomalies over even longer horizons, from three to five years. An immediate consensus around expected earnings is difficult to reach when a correct judgement on the truthfulness of financial disclosure and information content of earnings components requires more time. If abnormal returns of newly added firms are partially due to the expected improvement in earnings quality, the positive index effect might strongly influence prevailing prices in the first month(s) but also continuously over the following year(s). The empirical research is not conclusive on which factors determine abnormal returns for added firms, because different explanations may share the same prediction about the price and volume effect of the index listing (e.g., imperfect substitute and certification explanations). To distinguish the competing explanations, scholars usually examine not only how large the price revision is but also if it reverses. Madhavan (2003) suggests a simple approach that examines the relative importance of the temporary and permanent price effect: the decomposition of price movements around index changes (in his study, around the Russell reconstitution event). The long-term price change reflects the permanent effect of index inclusion, which is consistent with the new information and imperfect substitute explanations. The short-term price effect predominantly captures the impact of indexing (i.e., price pressure) and improved liquidity on prevailing prices, but it can also be related to new information. The reversal in stock price in the first day(s) after the inclusion distinguishes a short-term from long-term price effect, as follows: R temp = ln(p 1 ) ln(p 2 ) R perm = ln(p t ) ln(p 0 ), where p 0 is the price in the inclusion month and p t for t=2,..., 11 is the price for the second to eleventh month after addition to the S&P 500. Table 3 summarizes the temporary and permanent price effect in the year following index inclusion. The results confirm that both effects are present after S&P 500 additions. Madhavan (2003) estimates that the temporary price impact for the Russell 2000 addition (5.79%) is much larger than the permanent price impact (1.41%). In contrast, the price effect of addition to the S&P 500 tends to spread over time; the temporary effect (the month after addition) is only 0.01%, whereas the long-term (permanent) effect is larger than 2.40%, independent of the month which serves as a benchmark.

15 15 [Table 3 about here.] Table 4 shows market-adjusted returns before and after S&P 500 addition. Monthly returns are measured relative to the actual S&P 500 index returns 4. Firms that enter the S&P 500 during the period have on average higher mean market-adjusted returns (MAR) after the additions (0.43%) compared to their MAR before the listing event (-0.34%). The difference before-after is statistically significant. [Table 4 about here.] Consistent with the liquidity and price pressure explanations of abnormal returns following S&P 500 addition, some authors argue that the price effect reverses in the days following the announcement (e.g., Beneish and Whaley find a 2.2% price reversal). Other empirical studies support the permanent price effect of S&P 500 inclusion: Chen et al. (2004) document a permanent change in the price of added stocks (positive abnormal returns of 6.4% accumulated over a 60-day window), which they explain with the increase in investor s awareness after the index listing date. Graph 3 exhibits a permanent price effect for S&P additions, not only after the first two months but also in the three years following S&P 500 inclusion. Cumulative (market-adjusted) abnormal returns (CMAR) are calculated by summing the stock s market-adjusted returns over a window of up to three years. CMAR after S&P 500 addition are economically significant: around 6.5% (8.5%) in the first year after the addition, 12.8% (26.1%) up to two years after and 15.9% (42.6%) up to three years after the listing event, during the period ( ). [Graph 3 about here.] Market response to discretionary accruals: information content before and after S&P 500 addition. Our findings suggest that the earnings quality changes with the inclusion of a stock in the S&P 500 and that this event is associated with significant abnormal long-term returns. If the price effect of being listed is related to the more informative role of earnings, we would observe a significant market reaction to the

16 16 discretionary accruals. Table 5 presents our results on this and Appendix B contains the details of the estimation procedure. Prior literature is not conclusive on whether and exactly how much investors use accruals in their valuation and whether this affects the security s price. The anomaly research (e.g., Sloan, 1996) explains the abnormal returns, associated with observable firm attributes, as arising from slow or biased investor responses to information; singed measures of managerial discretion are used in this research. In contrast with these findings, Francis et al. (2005) show that firms with poor accruals quality have higher costs of capital than do firms with good accruals quality. They argue that larger magnitudes of accruals quality are associated with larger required returns because a larger magnitude of accruals quality indicates greater information risk, for which investors require compensation in the form of larger expected returns. In line with the accruals pricing literature, this study argues that investors price the lower information risk associated with the index listing and measured by discretionary accruals. It is the association between past discretionary accruals and current market-adjusted returns (MAR) that reveals the possible pricing of earnings quality. The regression coefficient, ˆβDisAcct 1, is the market response to the managerial discretion over earnings in the last reporting year; it is negative and statistically significant before and after the S&P listing. If the quality of reported earnings is a priced factor, investors would lower demanded returns when high-quality earnings reduce the uncertainly around the role of management in reporting good financial performance. Our findings suggest that this occurs; other things constant, current MAR increase when investors find the information risk, measured by past discretionary accruals, lower than their expectation. The level of cash flows or total earning may also influence abnormal returns of S&P 500 firm. We disentangle their effect with Models 2 and 3 and confirm that discretionary accruals remain a significant factor in explaining the variance in current MAR after including the other accounting variables (i.e., the discretionary component has an effect on abnormal returns that is different from that of earnings itself or the other component, cash flows). If managerial discretion over accruals is a priced factor, how does the S&P 500 listing affect this? Our findings suggest that the information content of discretionary accruals increases along with that of reported earnings; after the listing this component better explains the variance in market-adjusted

17 17 returns (i.e., in Model 1 the R 2 increases from 5,41% before to 6,71% after the inclusion; the same trend is observable in the results obtained with Models 2 and 3). Additionally, the S&P 500 addition may change the role of future discretionary accruals, DisAcc t+1, in price formation. If before the addition future discretion over accruals does not affect MAR, it does after the S&P 500 listing. Moreover, the relation between MAR and (expected) future discretion of management differs after the firm s inclusion. Our findings suggest that one year after the inclusion higher discretion over accruals has a positive effect on abnormal returns. It might be that the lower level of discretionary accruals for S&P 500 firms explains the reversal of this relation, which is negative before the listing. However, it could be that the inclusion event changes the investor s view about the quality of management and its discretion over reported earnings. Prior literature argues that management may attempt to mislead investors with discretionary accruals but may also use its reporting discretion to signal private information to the market (the signalling explanation for earnings management). Louis and Robinson (2005) examine the market reaction to discretionary accruals around stock splits, which prior studies suggest is a tool managers use to signal their optimism about future earnings. They confirm that discretionary accruals are positively correlated with abnormal returns around the split event, and argue that investors perceive the positive discretionary accruals prior to the stock splits as signals of management s optimism rather than the result of opportunistic behavior. The positive association between MAR and future DisAcc might be consistent with this signalling explanation, if the listing event changes either the purpose of discretionary accruals for the management or investors perception of this discretion. The inclusion of a firm in S&P 500 is an important event, and it can easily make the management of newly included firms more optimistic about future earnings. If the management uses discretionary accruals as a signal to the market, it could go along with stock split decisions, as suggested in previous studies. Moreover, the addition of a security in the S&P 500 is recognition that a firm is a leader in its industry; hence, the inclusion may affect the perceived quality of management. Future research can confirm if there is a relation between the index listing, the level of managerial optimism and the investor s perception of managerial discretion over accruals.

18 18 Conclusions This study investigates the relation between earnings quality and the long-term price effect of being added to the S&P 500. The trading-based explanations of the positive abnormal returns of index additions undermine the importance of information risk and earnings informativeness to price formation. We provide evidence of increased earnings quality for S&P 500 firms and argue that the association between market-adjusted returns and discretionary accruals reveals the possible pricing of accruals quality. The larger magnitudes of accruals before the index listing, which indicate greater information risk, could be associated with larger required returns when investors require compensation in the form of larger expected returns. With the index addition, discretionary accruals lower, thus reducing information risk and increasing the price. A large body of literature after Sloan (1996) posits that investors naively fixate on earnings and overweight the accrual component of current earnings when forecasting future earnings. Our results, however, support the role of information risk and accruals pricing, which is consistent with recent findings on the higher costs of capital for firms with poor accruals quality (Francis et al., 2005). A number of issues around the role of earnings quality on stock returns for S&P 500 additions deserve further attention. The earnings decomposition reveals that accruals are consistently negative over the period and that an index listing lowers them further. Previous literature argues that this is a sign of financial reporting conservatism, when firms also frequently report losses (Givoly and Hayn, 2002). Our findings suggest that there might be a trend toward more conservative accounting after the index listing, to which investors could also react positively with a price increase. Additional empirical tests, however, are needed to confirm if financial reporting is significantly more conservative after an index listing and how this affects stock returns before and after an S&P 500 addition. To further comment on the accruals pricing, the association between the index listing and stock split events could also be examined. Louis and Robinson (2005) confirm that investors may perceive discretionary accruals as a sign of managerial optimism. It might be that investors respond differently to stock split events before and after the listing if they believe that discretionary accruals reveal private information about future prospects of the firm only after its listing. If the signalling role of discretionary accruals differs around the index listing, this would confirm that not only information risk but also managerial signals are priced factors. With this study, we

19 19 argue that S&P 500 listing affects earnings quality and that investors price this earnings quality. Further research could confirm how this is linked to accounting conservatism and managerial optimism. Appendix A. Estimating Discretionary Current Accruals Discretionary accruals are estimated with the methodology suggested by Ashbaugh, LaFond and Mayhew (2003). Their estimation procedure provides the discretionary component of current accruals after controlling for firm performance with two alternative measures: namely, accounts receivable and returns on assets. Although they perform the estimation with a portfolio technique, the small number of S&P 500 additions in a single year prohibits this possibility in this study (i.e., usually there are less than 20 additions). The estimation procedure is as follows: CAcc = α 1 (1/laglasset) + α 2 ( Rev AR), where current accruals (CAcc) is net income before extraordinary items, plus depreciation and amortization, minus operating cash flows, scaled by total assets laglasset is total assets at the beginning of the year; Rev is equal to the net sales in year t, less net sales in year t-1, scaled by total assets, and; AR is equal to accounts receivable in year t, less accounts receivable in year t-1, scaled by total assets. The expected current accruals (ExpAcc) are obtained using the estimated parameters. The difference between actual current accruals and ExpAcc is the managerial discretion over accruals: DisAcc. Ashbaugh et al. (2003) suggest that returns on assets can also be used as a measure of financial performance when estimating discretionary accruals. The findings obtained with this alternative measure are not reported, since they do not differ from the estimates obtained with the procedure explained above. The variables are winsored at the 1% level to guarantee that extreme observations do not drive reported results. Appendix B. Estimating the Market Response to Discretionary Accruals The market response to discretionary accruals is measured with the three following models: Model1 : MAR t = α + β 1 DisAcc t 1 + β 2 DisAcc t + β 3 DisAcc t+1 + e Model2 : MAR t = α + β 1 DisAcc t 1 + β 2 DisAcc t + β 3 DisAcc t+1 + β 4 CashF low t 1 + β 5 CashF low t + β 6 CashF low t+1 + e

20 20 Model3 : MAR t = α + β 1 DisAcc t 1 + β 2 DisAcc t + β 3 DisAcc t+1 + β 4 Earnings t 1 + β 5 Earnings t + β 6 Earnings t+1 + e where MAR stands for the average market-adjusted returns in year t, DisAcc - for discretionary current accruals, measured as explained in Appendix A, CashFlow - for operating cash flows, and Earnings - for net income before extraordinary items (all variables are scaled by total assets). The regression coefficients β 1,2,3 capture the effect of past, current and future discretionary components of accruals (DisAcc) on stock returns. The level of cash flows is also expected to affect market-adjusted returns. We disentangle the effect of managerial discretion over accruals from that of the level of cash flows on MAR with Model 2; β 4,5,6 is the effect of past, current and future cash flows on stock returns. Market-adjusted returns depend on reported earnings which affect MAR. We draw conclusions on the individual effect of discretionary accruals with Model 3, where β 4,5,6 represent the effect of past, current and future earnings on stock returns. The explained variance in MAR is captured with the coefficient of determination (R 2 ). Model 1 reveals the information role of discretionary accruals in explaining the variance in marketadjusted returns. Models 2 and 3 include other factors which additionally affect MAR. Expected current accruals are not included in the estimated model, as they do not contribute either individually or jointly to the explained variance. Endnotes 1 While S&P 500 firms are generally large, this is not always the case. There are large companies not in the S&P 500, such as USA Networks and Liberty Media. Also, the widely held belief that S&P 500 firms are well-known is not necessarily true. S&P 500 firms specializing in non-consumer oriented products such as the Fifth Third Corporation and Automatic Data Processing are hardly ever recognized (Source: Chen et al., 2004). 2 There is a significant difference between the two explanations. Jain (1987) argues that S&P s decisions offer information content, even though S&P denies that index inclusion implies any judgment about the future prospects of a company. Denis et al. (2003) state that the inclusion event itself is information-free, but that it improves future performance as it leads to greater scrutiny (or monitoring) of management by investors, and management, in turn, responds with greater effort. 3 Two important accounting principles that guide the production of earnings are the revenue recognition and the matching principle. Management may recognize revenues after providing the service when the cash receipt is reasonably certain (the revenue recognition principle). Cash outlays associated directly with revenues should be accounted for in the period in which the firm recognizes revenues (the matching principle). Additionally, there are accounting conventions, such as objectivity, verifiability and the use of historical cost valuation models, that limit the flexibility of management to manipulate revenue and expense recognition. 4 Market-adjusted returns (MAR) measure the abnormal price movements on a particular day. They are the deviation of a stock s normal returns from that of the market. In many studies about the S&P 500 effect, scholars refer to MAR as abnormal or excess returns.