Stocks added to or deleted from the S&P 500 Index: A comprehensive long term analysis. Abstract

Transcription

1 Stocks added to or deleted from the S&P 500 Index: A comprehensive long term analysis Abstract The long-run price performance of stocks added to or deleted from the S&P 500 Index from 1962 to 2003 was analyzed. Both addition and deletion stocks are shown to have outperformed the index, and deletion stocks outperformed addition stocks. The operating performance of addition (deletion) firms is shown to have been consistently better (poorer) than the relevant industry average both before and after the changes. Institutional ownership of deletion stocks decreased prior to their deletion but increased again in the post-deletion period. Insiders buy deletion stocks more than addition stocks. JEL Classification: G11, G12, G14 Keywords: S&P 500 Index Revision, Long-run Performance, Operating Performance, Institutional Ownership, Long-term Overreaction 1

2 1. Introduction Changes in index constituent stocks are valuable information for investors. While studies on the short-term performance of the stocks concerned are well documented, very few have investigated the long-run performance of stocks added to or deleted from an index 1. One reason is most scholars focus on the short run and their hypotheses can be tested by simply examining short-term stock performance. However, we believe that it is also important to investigate the long-run performance of such stocks to validate whether or not those hypotheses are equally applicable to long-run stock price movements or changes in the operating efficiency, about which knowledge is rather limited. This study involved a comprehensive analysis of the long-run performance of stocks added to or deleted from the S&P 500 Index. It asked, what is their long-run performance and what forces drive that performance if it is different from certain benchmarks? We try to understand the results from the point of view of both operating performance and the stock trading environment, including changes in institutional ownership, liquidity, analysts forecasts, insider trading as well as investor awareness. A period five years before and after the addition or deletion from the index was considered. Relating asset pricing to dividend growth, the current stock price should equal the present value of future dividends or cash flows. Both the future dividends/cash flows and the required return will be affected if a stock is added to or deleted from an index. The operating performance of added stocks will perhaps improve if investors and analysts monitor the company more closely (Denis, McConnell, Ovtchinnikov, and Yu, 2003). Moreover, being added to an index such as the S&P 500 Index could also signal that the company is in a leading industry and its prospects are bright (Cai, 2008). Operating cash flow and future dividend payments may increase as a result. Addition to an index could enhance the liquidity of the stock and the liquidity premium could 1 Except that Cai (2008) examined the long-term impact of Russell 2000 Index rebalancing on portfolio evaluation. 2

3 decrease as a result. Increased analyst coverage helps reduce information asymmetry. In addition, more investors become aware the company and the shadow cost will probably decrease in the future (Chen, Noronha, and Singal, 2004). All these factors help to reduce the required return. Therefore, the increasing future dividends coupled with decreased required return suggests that the expected returns will be higher for added stocks in the long run. For deleted stocks, the overall required return will increase because liquidity will decrease (Hedge & McDermott, 2003), though investor awareness perhaps will not (Chen et al., 2004). Meanwhile, there is no evidence either in the scholarly literature or from Standard & Poor s that deletion from the index is a reliable signal of poor company prospects. 2 But no published study has investigated the long-run operating performance of deleted stocks and the future dividends/cash flows are unknown. The long-run performance of deleted stocks remains unclear. Investors and analysts will monitor a management s investments and operations more closely once a company has been added to an index. Management therefore has to create more value for the shareholders and operate the company more efficiently (Denis et al., 2003). However, the same argument can also apply to deleted stocks. It is bad news for the company if the stock is deleted from the index. The management will face great pressure from shareholders and will be forced to work harder to improve the company s operations and create more value for the shareholders. 3 Such changes usually take several years to bring results. If the deleted stock s performance improves, it will eventually be reflected in the share price. Whether added stocks or deleted stocks will outperform the market is an empirical question that also needs to be addressed. 2 The criteria for inclusion in the S&P 500 Index include market capitalization, liquidity, domicile, public float, sector classification, financial viability, and treatment of IPOs. A stock can be deleted from the index if the company is involved in a merger, acquisition or significant restructuring, or if it substantially violates one or more of the addition criteria (Standard and Poor s, 2011). 3 For example, Chen, Harford, and Li (2007) found that concentrated long-term holdings by independent institutions will lead them to specialize in monitoring the company. Brav, Jiang, Partnoy, and Thomas (2008) investigated hedge fund activism and found that target firms experience increases in payout, operating performance, and CEO turnover after activism takes hold. Admati and Pfleiderer (2009) created a model predicting that a credible threat of exit from a large shareholder often reduces agency costs between managers and shareholders. 3

4 Denis et al. (2003) shown that once a stock is added to the S&P 500 Index, analysts earnings forecasts rise, as do realized earnings in the coming year. But Elliott, Van Ness, Walker, and Warr (2006) do not confirm those results. This study takes a step further by looking at changes in operating performance, such as in financial ratios adjusted by industry averages. It compared the previous five years operating performance with that of the following five years in an attempt to provide a clear answer to the debate. It is also important to investigate any long-run changes in institutional ownership, liquidity, analysts forecasts, insider trading as well as investor awareness. Any might provide insights and help to explain the changes in the trading environment. The underlying assets of S&P Index component stocks are now worth aboutus$1.1trillion, and many investment products such as index funds, exchange traded funds, index futures and options are linked to the S&P 500 Index. Index additions and deletions will thus affect the trading environment and stock prices in general. It would be interesting from an investor s point of view to know whether a long-term buy and hold strategy based on index revisions could be reliably profitable or even beat the market. If addition to an index indeed certifies a company s long-term prospects, investors might buy and hold the stock for the long run, but no empirical evidence supports this proposal. 2. Literature Review Most previous studies of index composition changes have focused on the addition of stocks to an index. In the literature, six hypotheses are suggested to explain the observed increases in stock prices associated with addition to an index. The price pressure hypothesis suggests that the change stimulates excess demand which generates upward price pressure that will persuade current investors to sell the stock prematurely, but that pressure will abate once the momentary demand is satisfied. This hypothesis assumes that the long-term demand curve of the stock is perfectly elastic, but in the short term it is downward-sloping (Scholes, 1972). Studies supporting the hypothesis in relation to the S&P 500 Index include those of Harris and Gurel 4

5 (1986) and Wurgler and Zhuravskaya (2002). The hypothesis is also supported by evidence from other U.S. indexes and other markets, such as the Russell 2000 Index (Biktimirov, Cowan, and Jordan, 2004), the S&P Small Cap 600 Index (Shankar and Miller, 2006), the TSE 300 Index (Chung and Kryzanowski, 1998), the FTSE 100 Index (Mase, 2007), and the ISE-100 and ISE-30 indices (Bildik and Gulay, 2008). The downward-sloping demand curve hypothesis is also called the imperfect substitutes hypothesis. It proposes that there is a permanent stock price increase (decrease) if a stock added to (deleted from) an index. Studies that support this hypothesis for the S&P 500 Index include those of Shleifer (1986) and Lynch and Mendenhall (1997). The hypothesis has also been supported using data from other market indices such as the TSE 300 Index (Kaul, Mehrotra, and Morck, 2000) and the Nikkei 500 Index (Liu, 2000). There is also a liquidity effect hypothesis which states that liquidity will change if a stock is added to or deleted from an index, and liquidity change can produce information (Chen et al., 2004). For example, if a stock is added to an index the amount of information will increase due to investors greater interest and information produced by analysts, the media and other financial intermediaries as a result. The adverse selection component of liquidity will decrease, resulting in a positive price response. Even without an increase in information, increased trading volume will reduce the inventory cost component of liquidity and also result in a positive price response (Chen et al., 2004). Other studies of the S&P 500 Index such as those of Hedge and McDermott (2003) and Becker-Blease and Paul (2006) support the liquidity effect hypothesis. Studies of the DJIA (Beneish and Gardner, 1995) and the TSE 300 Index (Chung and Kryzanowski, 1998) also provide supporting evidence. The improved operating performance hypothesis states that as investors become aware of a firm because of its addition to an index, they are more likely to own shares in the company, monitor the firm more closely, and exert pressure on the firm to improve performance (Denis et al., 2003). Tests of this hypothesis have looked at the revisions of analysts earnings forecasts and 5

6 realized earnings once a stock is added to the S&P 500 Index, but they have shown contradictory results (Denis et al., 2003; Elliott et al., 2006). The increased investor awareness hypothesis, also called the shadow cost hypothesis (Merton, 1987), states that in segmented markets investors hold incompletely diversified portfolios. The equilibrium return required by less than fully diversified investors will be higher than that required by the full-information capital asset pricing model. The difference between the two returns is the shadow cost. If a stock is added to an index, more investors will be aware of it and hold it for its diversification potential. As a result, the stock s shadow cost will fall, resulting in an increase in its price (Chen et al., 2004). Elliott et al. (2006) reported that increased investor awareness is the primary factor behind a cross-section of abnormal announcement returns invovling the S&P 500 Index. The information hypothesis or certification hypothesis states that adding a firm to a major index in some sense certifies it as a leading firm in a leading industry. Cai (2007) found that being added to the S&P 500 Index seems to convey favorable information about the firm and its industry. Carter (2011) found similar evidence, also with the S&P 500. Liu (2008) examined the Nikkei 225 index and found the return series become more (less) random and less (more) predictable for stocks added to (deleted from) the index, and these changes are related to changes in the information environment for the stocks involved. Platikanova (2008) found that discretionary accruals decrease significantly once a company has been added to the S&P 500 index, which greatly improves earnings quality. 3. Data Changes in the S&P 500 Index constituent stocks from July 1962 to December 2003 were analyzed. The data from July 1962 to December 2000, also used by Chen et al. (2004), can be 6

7 downloaded from the Journal of Finance website 4 ; data from 2001 to 2003 were collected by following the methods developed by Chen et al. instructions. The stock prices, returns, trading volume, number of shares outstanding and market capitalizations were collected from the CRSP database. Related accounting data is provided by the Compustat database. Fama-French threefactor data is available on French s website 5. Institutional ownership and insider trading data are published in the Thomson Financial Institutional database and Insider database. Analysts forecasts were obtained from IBES. Like Chen et al. (2004), we exclude added stocks if the additions were due to a merger, takeover, spinoff, or changes in share classes. We also exclude deleted stocks if the deletions were due to a merger, takeover, spin-off, divestiture, bankruptcy or liquidation, a buyout, a suspension or delisting from the NYSE, or a change in share classes. Figure 1 plots the number of changes in the S&P 500 Index from 1962 to On average 22 stocks were added to the index and 22 deleted each year (range 8 to 60). Figure 1 shows a bell shaped pattern to the number of changes. For example, the number of changes was eight in 1977, increasing to 30 in 1984, and later decreasing to nine in In general, the index increased from 1962 to 2003, from 63 in 1962 to 1,469 in 1999, but it then index dropped significantly to 879. [ Figure 1 ] Table 1 reports the characteristics of the sample stocks. SIC is the first two digits of the standard industry code reported by the CRSP. If the number of observations is less than five, we combine them into the Other category. Firm size is defined as the closing price at the end of month t-1 times the number of shares outstanding in millions of dollars. RET_P1Y is the cumulative raw return from month t-12 to t-1. RET_P5Y is the cumulative raw return from month

8 t-60 to t-1. M/B is the market-to-book ratio at month t-1. The added stocks are from 42 industries and the deleted stocks are from only 17. Table 1 shows that the firms added were much larger than those deleted. The average firm size for the addition stocks is $1,143million, while for the deletion stocks it is only $175million. Second, the stocks added performed better in both the short and long term, while for the deletion stocks the returns are negative on average. Note that overall the average past oneyear and five-year return of addition stocks are 22% and 194%, respectively while the average past one-year and five-year return of deletion stocks are -2% and -14%. In general the addition stocks can be described as growth stocks with higher M/B ratios and the deletion stocks as value stocks with lower M/B ratios. The average M/B ratio of addition stocks is 2.39 and for deletion stocks is [ Table 1 here ] 4. Long-run Stock Performance 4.1 Buy and hold raw returns and stock market index adjusted returns From investors point of view, they care about both raw cumulative return and the market adjusted return, especially fund managers. Table 2 reports the three-year and five-year raw returns year by year as well as market adjusted cumulative returns. We consider three benchmarks: the Dow Jones Industrial Average Index, the S&P 500 Index, and the CRSP Value Weighed Index, all adjusted for dividends. Table 2 shows that both of addition stocks and deletion stocks record positive raw returns in three-year and five-year post event periods. For addition stocks the cumulative raw return averages 40% over three years and 75% over five. For deletion stocks, the cumulated raw returns are 68% and 107% in three- and five-year periods respectively. The short-term price pressure on addition stocks is well documented, but we also examined this issue, though it is not the focus of the study. Table 2 reports the returns over the [1, 36] and [0, 36] month periods. The difference in 8

9 returns with the two event windows roughly shows the effect of price pressure. The event month return is 4% for addition stocks and -7% for deletion stocks. This is consistent with what has previously been reported about short-term price pressure. 6 The table shows that both addition stocks and deletion stocks beat most indices. The addition stocks outperformed the Dow Jones Industrial Average Index by 19% over three years and by 38% over five. They also outperformed the S&P 500, even though those stocks were themselves index components. The addition stocks also outperformed the CRSP value weighted index over a five year period. And the long-run performance of the deletion stocks was even better. They outperformed the Dow Jones Industrial Average Index by 57% over three years and by 87% over five years. The deletion stocks also outperformed the S&P 500 Index and the CRSP value weighted index over both periods. The difference in returns between the addition and deletion stocks was 28% over three years (statistically significant with a t-value of 3.22) and the difference was 32% over five years (also significant with a t-value of 2.62). These results are consistent with those of Cai (2008), who investigated the long-term impact of Russell 2000 index rebalancing. [Table 2 here] Survival bias may explain why the deletion stocks outperformed the addition stocks over the long run. To address this issue we checked for any additional delistings of the addition and deletion stocks over the subsequent five years. In our sample period, 99 addition stocks and 55 deletion stocks were later delisted. Including their returns in the cumulated returns does not appreciably alter the results shown in Table Calendar-time abnormal returns 6 Short-term price pressure over shorter windows was also consistent with prior studies. The results are available on request. 7 Table of delisting time for addition and deletion stocks, and market adjusted returns with delisted returns are available on request. 9

10 Another popular method to measure abnormal return is the calendar time approach. For each calendar month in our sample period we formed equal-weighted and value-weighted portfolios of the firms added or deleted in the previous five years. We then regressed the portfolio s excess return against the Fama-French three factors and Carhart s momentum factor. The excess return was computed as follows: R Aor D pt R ft m R R SMB HML MOM (3) mt ft s t h t m t t where R pt is the portfolio s return for month t, A or D in the superscript indicates added or deleted stocks, R ft is the risk-free interest rate, (R mt -R ft ) is the entire market s excess return, SMB t is the difference in returns between a portfolio of small and large cap stocks, HML t is the difference in returns between a portfolio of high and low book-to-market ratio stocks, and MOM t is the highest prior monthly portfolio return minus the lowest over the prior 2-12 month period. The expected value of the intercept, α, which measures the monthly abnormal return, is zero under the null hypothesis of no abnormal performance. The regression results are reported in Table 3. Panel A reports the value weighted results. The intercept of addition stocks is 0.9 percent and significant at the 1 percent level of confidence. The intercept of deletion stocks is 0.6 percent and also significant at the 1 percent level. So for the value weighted portfolio, the higher returns of addition stocks and deletion stocks are not fully captured by the three factors. The coefficient on the SML factor for the addition stocks is not significant, but for the deletion stocks it is with a t-value of Size apparently contributes to explaining the better long-run performance of the deleted stocks. This is consistent with Table 1 which shows that deleted firms are much smaller than added ones. The HML factor is significant for both addition and deletion stocks, but the signs of the coefficients are opposite. The negative HML coefficient of the addition stocks indicates that they were financially healthy firms with little distress risk. The positive coefficient for the deletion stocks indicates that the 10

11 higher returns of deletion stocks may have been contributed by their greater distress risk. The momentum factor does not help explain the abnormal returns of either addition or deletion stocks. Panel B of Table 3 reports the results with the equal weighted portfolios. The lack of significance for the intercepts means the returns are well explained by the factors in the models. The factor loading is similar to Panel A, though the signs of the addition stocks are positive. The deletion stocks still have higher SMB and HML factor loadings. In addition, the MOM factor plays a minor but significant role in explaining the returns. [ Table 3 here ] These results have strong implications for real world investments. They are consistent with the conventional wisdom that investors should buy and hold the stocks which have just been added to the S&P 500 index. They are inconsistent, however, with the conventional wisdom that investors should sell stocks which is have been deleted. Our empirical results show that deletion stocks outperform the benchmarks as well as stocks just added to the S&P Changes in Operating Performance Following the lead of Loughran and Ritter (1997), we used profit margin, ROA, OIBD/Assets, (CE+RD)/Assets, and M/B to proxy for a firm s overall operating efficiency. 9 Those five ratios were adjusted by the respective industry averages and compared over the fiscal years t- 5 to year t+5. Panels A and B of Table 4 report the results for addition stocks and deletion stocks respectively. Panel A shows that once adjusted by the industry average, all numbers are positive. 8 We do not use the matching method to compute the abnormal returns because the stocks in the S&P 500 Index are leading companies in America s leading industries. It is impossible to find a set of matching firms similar to S&P 500 index constituent stocks. 9 Profit Margin is defined as net income divided by sales (COMPUSTAT item #172)/(COMPUSTAT item #121). ROA is defined as net income over total assets (COMPUSTAT item #172)/(COMPUSTAT item #6). OIBD/Assets is operating income divided by total assets, where operating income is defined as operating income before depreciation, amortization, and taxes, plus interest income (COMPUSTAT item #13 + COMPUSTAT item #62)/(COMPUSTAT item #6). (CE+RD)/Assets is capital and R&D expenditures as a proportion of total assets (COMPUSTAT item #128 + COMPUSTAT item #46) /(COMPUSTAT item #6). M/B is the firm s market-to-book ratio (COMPUSTAT item #54 * COMPUSTAT item #199)/(COMPUSTAT item #60). 11

12 It means that the performance of firms newly added to S&P 500 Index was usually above their industry average. Panel B shows that once adjusted by the industry average, all numbers are negative. Stocks just deleted from the S&P 500 underperformed their industries. [ Table 4 here ] Panel C of Table 4 reports the Z-statistics of Wilcoxon matched-pairs signed-rank tests for the differences between deletion stocks and addition stocks reported in panels A and B. The Z-statistics clearly show that there is a significant difference in operating performance between addition stocks and deletion stocks in both the pre- and post-event periods. Panel D reports the Wilcoxon test results for the difference between deletion stocks and addition stocks reported in panels A and B from [-5, -1] to [1, 5]. For addition stocks, the difference between the pre- and post-event periods is significant. The results show that though the post-event operating performance of addition stocks was above their industry averages, the industry adjusted performance was reversed. For the deletion stocks, however, we found no consistent evidence of changes in industry-adjusted performance. These results suggest that the outperforming stock prices of addition firms are in the long-run supported by their strong operating performance. However, we found no significant evidence of improved operating performance among deleted stocks in the post-event period. Any stock price outperformance among the deletion stocks is not related to the underlying operating performance of the firms. 6. Changes in the Trading Environment 6.1 Institutional ownership Certain institutions such as pension funds only invest in S&P 500 Index constituent stocks. Index funds and exchange traded funds passively buy and sell stocks when Standard and Poor s announce changes. We would expect institutional ownership to increase if a stock added to the S&P 500 Index and to decrease if it is deleted. We created two institutional ownership proxies: 12

13 IH_NO and IH_RATIO. IH_NO is the number of institutions holding the stock; IH_RATIO is the proportion of stock held by institutions defined as the total number of shares held by institutions divided the number of shares outstanding. Because ownership data is only available from 1980 and we need to investigate ownership five years before and after changes, the analysis was limited to additions and deletions from 1985 to Table 5 reports the results. The proxies of additions increased significantly from year t-5 to t. The main reason is probably that some addition stocks were quite young and were growth stocks as Table 2 suggested. Panel A of Table 5 shows that both the number and proportion of institutional owners increased significantly from year t-5 to year t-1 for addition stocks. The median IH_RATIO increased from 0.51 to 0.6 while the median IH_NO increased from 101 to 180. For the deletion stocks, both IH_RATIO and IH_NO were stable from year t-5 to year t-1. The median IH_RATIO is 0.55 at year t-5 and 0.58 at year t-1, and the corresponding IH_NO values are 116 and 122. [ Table 5 here ] Around the years when a stock is added to and deleted from the S&P 500 (from year t-1 to t+1), our results confirm the convention wisdom that institutions buy stocks added to the S&P 500 Index and sell stocks deleted from the index. For example, the IH_NO of addition stocks increased from 180 at year t-1 to 225 at year t, and further increased to 229 at year t+1. The results are stronger for the deletion stocks. The IH_NO values for deletion stocks are 122 at year t-1, 89 at t, and 69 at t+1. The IH_RATIOs of deletion stocks are 0.58 at t-1, 0.49 at t, and 0.47 at t+1. In the long-run (t+1 to t+5), there is not much further increase in institutional holdings of addition stocks. IH_NO increases from 229 to 246 from year t+1 to t+5, but IH_RATIO only increases from 0.59 to In contrast with conventional wisdom, there is a significant increase in institutional holding of deletion stocks as well. IH_NO increased from 69 to 115 from year t+1 to t+5; corresponding to an IH_RATIO increase from 0.47 to Taking together IH_NO and 13

14 IH_RATIO at year t+5, the values for addition stocks are 246 and 0.62; for deletion stocks they are 115 and This means that the institutional ownership is more concentrated within deletion stocks than addition stocks. Though not reported, we classified the mutual funds as either index-related or nonindexrelated and found a significant increase in IH_RATIO for addition stocks and a significant decrease for deletion stocks among index-related funds Analysts forecasts We focus on two properties of analysts forecasts: coverage (COVERAGE) and the forecasts dispersion (DISPERSION). COVERAGE is defined as the number of analysts following a stock; DISPERSION is the standard deviation of analysts annual earnings per share forecasts for a fiscal year divided by the average of their earnings per share (EPS) forecasts. We expect that COVERAGE will increase significantly after addition and decrease significantly after deletion. The DISPERSION of deletion stocks should be higher than that of addition stocks because their future is less clear. We also expect DISPERSION to be greater around the deletion period. Figure 2 reports analyst coverage and the dispersion of analysts earnings forecasts five years before and after stocks were added to or deleted from the S&P 500 from 1979 to Consistent with our predictions, COVERAGE of addition stocks increased from year t-5 to year t and further increased to year t+2. COVERAGE of deletion stocks decreased from year t-2 to year t and further decreased to year t+1. COVERAGE is quite stable from year t+2 to t+5 for both additions and deletions. Second, the DISPERSION of deletion stocks is consistently higher than that of addition stocks. The DISPERSION of addition stocks is quite stable from year t-5 to t+5, which means that in general the analysts had a consensus view about their earnings. There is an uptrend in the DISPERSION of deletion stocks from year t-5 to year t, remaining high at year t In our mutual funds subset, we also find that for nonindex-related funds for both addition and deletion stocks, IH_RATIO decreased from quarter t-1 to t. Those results are available on request. 14

15 DISPERSION decreased significantly, however, at year t+2 and remained at a lower level in years t+3 to t+5. [ Figure 2 here ] 6.3 Liquidity We constructed three proxies for liquidity. TURNOVER is the trading volume divided by the number of shares outstanding. ADJTURNOVER is the market adjusted turnover, which is defined as the TURNOVER divided by an adjustment factor (ADJFACTOR). ADJFACTOR is the monthly CRSP turnover, and we set January 1950 as the base month with value 1. ILLIQUIDITY (Amihud, 2002) was defined as D 1 im Rimd ILLIQUIDIT Yit, (8) D VOLD im t 1 imd where D im is the number of days for which data is available for stock i in month m, R imd is the return on stock i on day d of month t and VOLD imd is the respective daily volume in dollars. We compared the differences in liquidity, defined as the post-event liquidity minus the pre-event liquidity, using various event windows from 10 months to 60 months. The null hypothesis is that there is no difference in liquidity before and after a revision. Table 6 reports the results. TURNOVER increased for both addition stocks and deletion stocks over all event windows. Table 6 also shows that TURNOVER increased monotonically as the window widened from 10 months to 60 months. For example, the difference in TURNOVER for addition stocks is , , , , , and for 10 months up to 60 months. Surprisingly, though, the turnover of deletion stocks does not decrease. One reason may be that investors remain aware of those stocks and keep trading them. And of course market turnover has generally increased over the past four decades. ADJTURNOVER was constructed to remove the increasing turnover of the general market. [ Table 6 here ] 15

16 ADJTURNOVER of the addition stocks decreased significantly, so their turnover decreased relative to the overall market. It also indicates that those new added stocks were liquid before being added to the S&P 500. The difference in ADJTURNOVER of the deletion stocks is not significant. Investors apparently remain aware of those stocks in their trading. ILLIQUIDITY for the addition stocks is unchanged, and it is not much changed for the deletion stocks. In general all are relatively liquid. So there is no clear evidence of changes in liquidity. For addition stocks, TURNOVER is positive, but once adjusted for the market it becomes negative. TURNOVER for deletion stocks is positive, but once adjusted for the market it becomes not significant. ILLIQUIDITY for both addition and deletion stocks is unchanged. 6.4 Shadow Costs Chen et al. (2004) has proposed an investor awareness hypothesis to explain the shortterm asymmetric price responses of stocks added to and deleted from the S&P 500 Index. The idea is that investor awareness of addition stocks increases and as a consequence shadow costs will be reduced; but investors do not become unaware of it after it has been removed from the index. We investigated shadow cost over a long event period from year t-5 to t+5. Following Chen et al. (2004) s method, shadow cost was computed as ShadowCost t Re siduals tan darddev FirmSize t 0. (9) S&P500MarketCap 0 NumberofShareholders t Here FirmSize 0 (the market value of equity) and S&P500MarketCap 0 are measured on the announcement date of an index change. ResidualStandardDev t is the standard deviation of the difference between the firm s return and the S&P 500 total return in a 252-day from year t-5 to year t before the index change announcement (or after the effective day for the post period). NumberofShareholders t is the most recent figure available prior to the announcement date. Shadow costs were multiplied by

17 Table 7 reports the results. First column reports the mean and second column reports the median values. The third and fourth columns report the mean shadow costs after and before October The shadow cost of addition stocks decreased significantly. For example, the median of shadow costs are at year t-1, but at t and at t+5. The pattern is similar for the mean before and after October [ Table 7 here ] The shadow cost of deletion stocks increased slightly from year t-5 to t and maintained that level afterwards. The median shadow costs are 0.233, and at year t-1, t and t+5. There is, however, a significant difference before and after October Before October 1989 the shadow cost increased from year t-5 to t+5; after that, we do not observe such a pattern. The shadow cost was generally higher before October The difference is clearest among addition stocks. Their shadow cost is higher than that of deletion stocks. This is consistent with number of institutional investors in Panel A of Table 5. Those newly added stocks are often young growth stocks with little listing history on exchanges while the deletion stocks have been listed on exchanges much longer. 6.5 Insider trading Previous scholarly work has shown that insider trading conveys new information about companies. 11 Insider trading in addition and deletion stocks might provide insight about the 11 Lakonishok and Lee (2001) found that insiders in the aggregate are contrarian investors and they predict market movements better than simple contrarian strategies. They found that the informativeness of insiders activities comes from their purchases, while insider selling appears to have no predictive ability. Piotroski and Roulstone (2005) document that insiders are both contrarians and the possessors of superior information about misevaluation. Their trades are positively related to the firm s future earnings and bookto-market ratio, and inversely related to recent stock returns. Hsieh et al. (2005) studied the informativeness of insider trades and analyst recommendations jointly. They found that insider trading is informative when signaling positive information, and analyst recommendations are informative when conveying negative information. Fidrmuc et al. (2006) found that directors trades convey new information about a firm s prospects in U.K. They found that for poorly performing firms and those close to financial distress there are stronger market reactions to director s purchases (sales) significantly positive (negative). 17

18 companies prospects. For example, if the management of a deleted firm is confident about the future of the company or believe the stock is oversold they will take the opportunity to buy. Insider trading data are available from We investigated insider trading five years before and after stocks were added to or deleted from the S&P 500 from 1991 to Follow Hsith et al. (2005), we defined PVA as the value of insider buys as a percentage of the sum of insider buys and sells, and PNA as the number of insider buys as a percentage of all insider transactions. Insiders include CEOs, CFOs, directors, officers, presidents and vice presidents. Table 8 reports the results. We interpret the results based on the smoother mean values reported in Panel B, but those results are consistent with median values reported in Panel A. Insider buying among deletion stocks is consistently higher than among addition stocks. The fraction of insider buying among deletion stocks increased significantly from year t-5 to t, but there was no such pattern among addition stocks. The PVA of deletion stocks is 0.41 at year t-5 and 0.70 at t; for addition stocks the corresponding values are 0.21 and Those results provide strong evidence that management of the deleted firms often consider their firms undervalued. [ Table 8 here] The PVA of deletion stocks decreased from year t to t+5 and the prices had stabilized. The PVA and PNA of addition stocks remain quite stable from year t-5 to t+5. The operating performance of addition firms is consistently better than the industry average, as Table 4 has reported. 7. An Alternative View of Long-term Overreaction DeBondt and Thaler (1985) have documented a long-run overreaction of stock prices over a three to five year period. We also tested for overreaction among addition and deletion stock prices. The following univariate regression was evaluated. 18

19 Future Re turn Past Return, (10) i,t i, t where FutureReturn refers to the buy and hold returns (BHR) and CRSP value-weighted index adjusted buy and hold returns (BHAR) over three and five years after addition or deletion. PastReturn refers to the BHR and BHAR over the three and five years prior to addition or deletion. For example, BHR_3Y is the buy and hold raw return in the [1, 36] month interval, and BHAR_P5Y is the buy and hold abnormal return over the [-1, -60] month window. Table 9 shows that the positive BHR and BHAR of addition stocks cannot be explained by the long-run overreaction of past BHR and BHAR. Table 1 clearly shows that added stocks are past return winners and deletions have been losers. Table 2 shows that the BHR and BHAR are significantly positive, as would be expected given excellent past performance. The results suggest that addition stocks have a momentum effect supporting further increases in the long run. However, Table 9 shows that there is a negative relationship between past BHR and future BHR. This means future BHR tends to be smaller than past BHR. The corresponding results with BHAR show no significant relationship. [ Table 9 here] Panel B of Table 9 shows a stronger negative relationship between past and future BHR. The results are significant at the 5% level for a five-year period. Compared with the coefficients for addition stocks, the magnitudes for deletion stocks are large, suggesting that deletion stocks have a stronger long-run overreaction, i.e., if the stock has had poor past performance, the future performance will be good. The R-squared of deletion stocks is also larger than for addition stocks, especially for five-year BHR. However, the results with BHAR are again not significant. The results in Table 2 remain unexplained. 8. Conclusion 19

20 Using historical changes of the S&P 500 index constituent stocks from 1962 to 2003, we investigated stocks added to or deleted from the S&P 500 Index. Both addition and deletion stocks subsequently outperform the index. Deletion stocks outperform the addition stocks over the subsequent three or five years. The operating performance of addition (deletion) firms is consistently better (worse) than the relevant industry average. No evidence shows that deletion stocks operating performance improves after deletion. Institutional ownership of addition stocks increases prior to the addition, but on average there are no further increases in the post-addition period. The institutional ownership of deletion stocks decreases prior to the deletion and then increases in the post-deletion period. There is no evidence of any change in a share s liquidity before or after an index change. The shadow cost associated with addition stocks decreases after addition, but there is no reduction for deletion stocks. Insiders buying of deletion stocks is higher than of addition stocks, showing that management insiders are often confident of a deleted company s prospects and feel the shares are undervalued. These results have strong implications for real world investing, as we have found that deletion stocks outperform addition stocks and also the market indices. This has been the first published study to provide a comprehensive analysis of the long-run performance of stocks added to or deleted from the S&P 500. It has uncovered fundamental evidence that deletion stocks outperform addition stocks in the long run. These results have strong implications for long-term buy-and-hold investors. 20

21 References Admati, Anat R., and Paul Pfleiderer, 2009, The Wall Street walk and shareholder activism: Exit as a form of voice, Review of Financial Studies 22, Becker-Blease, John R., and Donna L. Paul, 2006, Stock liquidity and investment opportunities: Evidence from index additions, Financial Management 35, Beneish, Messod D., and J. Gardner, 1995, Information costs and liquidity effects from changes in the Dow Jones Industrial Average list, Journal of Financial and Quantitative Analysis 30, Biktimirov, Ernest N., Arnold R. Cowan and Bradford D. Jordan, 2004, Do demand curves for small stocks slope down? Journal of Financial Research 27, Bildik, Recep, and Guzhan Gulay, 2008, The effects of changes in index composition on stock prices and volume: Evidence from the Istanbul stock exchange, International Review of Financial Analysis 17, Brav, Alon, Wei Jiang, Frank Partnoy, and Randall Thomas, 2008, Hedge fund activism, corporate governance, and firm performance, Journal of Finance 63, Cai, Jie, 2007, What s in the news? Information content of S&P 500 additions, Financial Management 36, Cai, Jie, 2008, Long-term impact of Russell 2000 Index rebalancing, Financial Analysts Journal 64, Carter, Kelly E., 2011, Downstream comovement: Evidence from S&P 500 Index addition announcements, Working paper, University of South Florida. Chen, Honghui, Gregory Noronha, and Vijay Singal, 2004, The price response to S&P 500 Index additions and deletions: Evidence of asymmetry and a new explanation, Journal of Finance 59, Chen Xia, Jarrad Harford, and Kai Li, 2007, Monitoring: Which institutions matter?, Journal of Financial Economics 86, Chung, Richard, and Lawrence Kryzanowski, 1998, Are the market effects associated with revisions to the TSE300 index robust?, Multinational Finance Journal 2, Cremers, K.J. Martijn, and Antti Petajisto, 2009, How active is your fund manager? A new measure that predicts performance, Review of Financial Studies 22, DeBondt, Werner and Richard Thaler, 1985, Does the stock market overreact?, Journal of Finance 40, Denis, Diane K., John J. McConnell, Alexei V. Ovtchinnikov, and Yun Yu, 2003, S&P 500 Index additions and earnings expectations, Journal of Finance 58,

22 Elliott, William B., Bonnie F. Van Ness, Mark D. Walker, and Richard S. Warr, 2006, What drives the S&P 500 inclusion effect? An analytical survey, Financial Management 35, Fidrmuc, Jana P., Marc Georgen, and Luc Renneboog, 2006, Insider trading, news releases, and ownership concentration, Journal of Finance 61, Harris, Lawrence, and Eitan Gurel, 1986, Price and volume effects associated with changes in the S&P 500 list: New evidence for the existence of price pressures, Journal of Finance 41, Hegde, Shantaram P., and John B. McDermott, 2003, The liquidity effects of revisions to the S&P 500 Index: An empirical analysis, Journal of Financial Markets 6, Hertzel, Michael, Michael Lemmon, James S. Linck, and Lynn Rees, 2002, Long-run performance following private placements of equity, Journal of Finance 56, Hsieh, Jim, Lilian Ng, and Qinghai Wang, 2005, How informative are analyst recommendations and insider trades?, Working paper, George Mason University. Kaul, Aditya, Vikas Mehrotra, and Randall Morck, 2000, Demand curves for stocks do slope down: New evidence from an index weights adjustment, Journal of Finance 55, Lakonishok, Josef, and Inmoo Lee, 2001, Are insider trades informative?, Review of Financial Studies 14, Liu, Shinhua, 2008, Changes in the Nikkei 500: New evidence for downward sloping demand curves for stocks, International Review of Finance 1, Liu, Shinhua, 2009, Index membership and predictability of stock returns: The case of the Nikkei 225, Pacific-Basin Finance Journal 17, Loughran, Tim, and Jay R. Ritter, 1997, The operating performance of firms conducting seasoned equity offerings, Journal of Finance 52, Lynch, Anthony W., and Richard R. Mendenhall, 1997, New evidence on stock price effects associated with changes in the S&P 500 Index, Journal of Business 70, Mase, Bryan, 2007, The impact of changes in the FTSE 100 Index, Financial Review 41, Merton, Robert C., 1987, Presidential address: A simple model of capital market equilibrium with incomplete information, Journal of Finance 42, Piotroski, Joseph D., and Darren R. Roulstone, 2005, Do insider trades reflect both contrarian beliefs and superior knowledge about future cash flow realizations?, Journal of Accounting and Economics 39, Platikanova, Petya, 2008, Long-term price effect of S&P 500 addition and earnings quality, Financial Analysts Journal 64, Scholes, Myron S., 1972, The market for securities: Substitution vs. price pressure and the effect of information on share prices, Journal of Business 45,

23 Shankar, S. Gowri, and James M. Miller, 2006, Market reaction to changes in the S&P SmallCap 600 Index, Financial Review 41, Shleifer, Andrei., 1986, Do demand curves for stocks slope downward? Journal of Finance 41, Shumway, Tyler, 1997, The delisting bias in CRSP data, Journal of Finance 52, Standard and Poor s, 2011, S&P U.S. indices methodology. Wurgler, Jeffrey., and Ekaterina V. Zhuravskaya, 2002, Does arbitrage flatten demand curves for stocks?, Journal of Business 75, Yan, Sterling Xueming, and Zhe Zhang, 2009, Institutional investors and equity returns: Are short-term institutions better informed?, Review of Financial Studies 22,

24 Table 1: Summary Statistics of Sample Stocks This table reports the median characteristics of sample firms added to or deleted from the S&P 500 Index between July 1962 and December SIC is the first two digits of the firm s standard industry code. Firm size is the closing price at the end of month t-1 times the number of shares outstanding in millions of dollars. RET_P1Y is the cumulative raw return from month t-12 to t-1. RET_P5Y is the cumulative raw return from month t-60 to t-1. M/B is the market-to-book ratio at month t-1. The sample comprises 788 addition stocks and 244 deletion stocks. SIC Firm Size Additions RET_ P1Y RET_ P5Y M/B Firm Size Deletions RET_ P1Y RET_ P5Y M/B

25 Other Average

26 Table 2: Market Adjusted Buy-and-hold Long-run Returns This table reports the mean market adjusted returns of stocks added to or deleted from the S&P 500 index from July 1962 to December The sample comprises 788 addition stocks and 244 deletion stocks. DJIA denotes the Dow Jones Industrial Average, SP500 is the Standard and Poor s 500 and CRSP denotes the Centre for Research in Securities Prices. *, **, *** indicate significance at the 10%, 5%, and 1% level of confidence, respectively. Post Event Months [0, 36] [1, 36] [1, 60] Additions Raw returns DJIA Return S&P 500 Return CRSP Value-weighed Index Return Raw-DJIA 0.22*** 0.19*** 0.38*** T-value Raw-SP *** 0.18*** 0.36*** T-value t Raw-CRSP 0.09** ** T-value Deletions Raw returns DJIA Return S&P 500 Return CRSP Value-weighed Index Return Raw-DJIA 0.49*** 0.57*** 0.87*** T-value Raw-SP *** 0.54*** 0.80*** T-value Raw-CRSP 0.32*** 0.41*** 0.54*** T-value Raw Ret (Deletions) Raw Ret (Additions) 0.17* 0.28*** 0.32*** T-value