Information content of S&P 500 index additions: A reexamination using Russell 1000 reconstitutions

Transcription

1 Information content of S&P 500 index additions: A reexamination using Russell 1000 reconstitutions Swaminathan Kalpathy Washington State University swamik@wsu.edu Mukunthan Santhanakrishnan Idaho State University santmuku@isu.edu January 2008 Preliminary draft. Please do not cite or quote without authors permission

2 Information content of S&P 500 index additions: A reexamination using Russell 1000 reconstitutions Abstract Existing literature on the announcement effects of S&P 500 index additions generally supports the hypothesis that demand curves for stocks slope downward. This explanation assumes that index addition announcements are free of any information content. More recent evidence, however, indicates that index additions are not completely free of information. Examining Russell 1000 reconstitutions, which are based on transparent and objective rules, and comparing them with S&P 500 additions, which are made by S&P index committee, enables us to unravel the two explanations. We uncover two interesting findings. First, we show that it is important to control for prior performance before examining financial analysts forecasting behavior. Second, after controlling for prior performance and other characteristics that are known to influence index inclusion, we find that analysts revise their earnings expectations upwards for S&P 500 index additions and not for Russell 1000 reconstitutions. Moreover, we find strong evidence that S&P 500 firms exhibit a permanent upward shift in stock price in the two years following index additions. In contrast, Russell 1000 firms witness complete reversal of stock returns following reconstitutions compared to size and book-to-market matched control firms. These results are consistent with S&P 500 index additions conveying new information to the market.

3 1. Introduction The topic of S&P 500 index additions has invited considerable attention among academics, practitioners, and capital market participants. One of the stylized facts surrounding the announcement of index additions is the positive stock market reaction to the event. 1 This price response to index additions has attracted a rich debate on the relative information content of index addition announcements. The two sides to the debate can be broadly summarized as follows. On one side is the idea that index additions represent an upward shift in the demand curve for stocks. In the absence of perfect substitutes, the increased demand from investors (mainly institutional investors) causes an increase in stock prices following index addition. This almost entirely assumes that index additions are devoid of any additional information content. On the other side of the debate is the idea that the price reaction could, in part, reflect information about future firm prospects. The objective of this paper is to disentangle these two explanations non mutually exclusive explanations. Shleifer (1986) was among the earlier studies documenting the abnormal returns associated with firms that get added to the S&P 500 index. This result is supported by Lynch and Mendenhall (1997) and Wurgler and Zhuravskaya (2002). In addition, Wurgler and Zhuravskaya (2002) show that abnormal returns are related to the difficulty in finding perfect substitutes for the firm getting added to the index supporting the downward sloping demand curve hypothesis. 1 See Harris and Gurel (1986), Shleifer (1986), Jain (1987), Dhillon and Johnson (1991), Beneish and Whaley (1996), Lynch and Mendenhall (1997), Wurgler and Zhuravskaya (2002), Denis, McConnell, Ovtchinnikov, and Yu (2003), Chen, Noronha and Singal (2004), Elliott, Van Ness, Walker and Warr (2006), for evidence on price effects around S&P 500 index additions. 3

4 Academic work providing traction to the information-based story surrounding index additions includes Jain (1987), Dhillon and Johnson (1991), Denis, McConnell, Ovtchinnikov, and Yu (2003), Becker-Blease and Paul (2006), and Cai (2007). Jain (1987), in documenting a positive price reaction around S&P 500 index additions, finds similar price reaction to firms that get added to supplementary S&P indexes. Since index funds do not typically have holdings in firms in the supplementary S&P indexes, Jain (1987) concludes that S&P 500 index additions are not completely information-free. Dhillon and Johnson (1991) provide evidence of a price increase of nonconvertible bonds and options of firms that get added to the index. Denis et al. (2003) examine whether financial analysts revise their earnings expectations around index addition announcements. Their evidence suggests that financial analysts revise their expectations upward for earnings in the period subsequent to index addition announcements. They also find that earnings improve following index additions. Becker-Blease and Paul (2006) find a permanent increase in stock market liquidity following index additions, and increases in capital expenditures that are related to the increased liquidity. 2 Finally, Cai (2007) presents evidence that rival firms in the same industry as the index addition firm witness a positive price reaction. While there seems to be some evidence that index additions are not completely information-free, one interesting question is whether we can unravel the informationbased hypothesis from the downward sloping demand curve hypothesis. While this is a difficult question to address, we employ an approach that attempts to sever the link between the two competing explanations. Specifically, we contrast firms that are added 2 Hegde and McDermott (2003) also show a permanent increase in liquidity for firms that get added to the S&P 500 index. 4

5 to the S&P 500 index with firms that are added to the Russell 1000 index. 3 While the S&P 500 and Russell 1000 are among the most commonly used U.S. equity benchmarks 4, the main difference between these two indices is the selection process. Firms added to the S&P 500 index have to satisfy several conditions and are chosen based on such as size, liquidity, industry representation on the index.. 5 The interesting feature about Russell indices is the transparency involved in the process. The index is constituted every year based strictly on market capitalization. All firms are sorted on market capitalization as of the 31 st of May each year. The top 1000 firms in terms of market cap constitute the Russell 1000 while the next 2000 firms constitute the Russell 2000 index. Together, these two indexes constitute the Russell 3000 firms. We use the fact that identity of firms that are added to the Russell 1000 index is known before the reconstitution to examine the information content of S&P 500 index additions. Improved firm value associated with index addition may occur in three likely scenarios for improved firm performance. First, improved performance can result from better product market outcomes as a result of the spillover of the financial market visibility into product markets. Second, performance improvements may arise from the bonding of managers to the discipline of the capital markets (Denis et al. (2003)) leading to more managerial effort production; Third, improved firm performance may result from the firm taking on more projects in light of a reduction in the overall of cost of capital 3 See Madhavan (2003), Biktimirov, Cowan and Jordan (2004), Chen (2006), and Carino and Pritamani (2007) for evidence on announcement effects for Russell index additions. 4 In 2006, $1.58 trillion tracked firms in the S&P 500 index and $252.2 billion tracked firms in the Russell 1000 index (Smith and George (2006)). 5 See for a discussion on index change methodology 5

6 following index additions. Interestingly, some (or most) of the above explanations likely apply to firms that get added to either the S&P 500 and Russell 1000 indexes. Whether there are any differential effects in how capital market participants respond to S&P 500 index additions and Russell 1000 index reconstitutions is largely an empirical question. Nonetheless, there is an important dimension that provides a clear contrast to the two settings, namely selection criteria. While S&P 500 index additions involve some subjective judgments on the part of the selection committee, there is no reason to believe that there is any subjectivity involved with Russell 1000 reconstitutions. As a result, any possible certification provided by the S&P index selection committee may be value relevant for a couple of reasons. First, as we pointed out earlier, there may be some positive spillover effects by way of favorable product market outcomes for firms subsequent to index additions. 6 Second, S&P may be picking firms that are ex-ante expected to have more favorable prospects in the future. Following the approach in Denis et al. (2003), we look at how financial analysts respond to index addition announcements. If there is information content in index additions, it is likely picked up by financial analysts by way of changes in their earnings expectations. For the sample of firms that get added to the S&P 500 index, our results largely mirror the findings in Denis et al. (2003). We find that financial analysts revise their expectations upward for earnings subsequent to the announcement of index addition relative to a control sample of firms matched on industry, size and liquidity that do not get added to the index. Surprisingly, we find similar results for the sample of firms that 6 For example, suppliers may provide more favorable terms of credit in their dealings with the index addition firms in response to the certification provided by S&P. It is also conceivable that product warranties offered by S&P 500 index addition firms are perceived by customers to be more valuable. 6

7 are added to the Russell 1000 index; namely, an upward revision in earnings relative to a control sample of firms matched on industry and size. Upon further examination, we find that both the S&P 500 and Russell 1000 index addition firms exhibit strong performance, relative to control firms, prior to being added to the index. Givoly and Lakonishok (1979), Brown, Foster and Noreen (1985) and Abarbanell (1991) document a positive relation between prior stock returns and earnings forecast revisions. If prior stock prices contain information about future earnings, then financial analysts impound this information in formulating their expectations of earnings. Therefore, prior returns incrementally reflect changes in earnings expectations. 7 We draw on this motivation and include prior year stock returns along with the other criteria that influence index inclusion to form new control samples. We then compare the change in earnings expectations of sample firms relative to the new set of control firms. The main findings with the sample of S&P 500 index addition firms are statistically significant at conventional levels. Interestingly, after controlling for prior performance, we do not find any evidence that the changes in earnings expectations for the Russell 1000 index addition firms are different from those for control firms. We interpret this as support for the idea that S&P 500 index additions are not completely free of information. Next, we examine the stock price performance in the two years following index addition. The information-based story would suggest that S&P 500 index addition firms exhibit a permanent increase in stock price subsequent from being added to the index. In contrast, the lack of information content associated with Russell 1000 reconstitutions would suggest that Russell 1000 firms are less likely to witness a permanent increase in 7 McNichols and O Brien (1997) show that analysts, in issuing recommendations and forecasts, self-select into firms that they expect to do well. If there is persistence in prior performance, then the recommendations and/or forecasts could reflect this self-selection. 7

8 stock price. Indeed, this is exactly what we find by examining stock returns of of index addition firms relative to stock returns of size and book-to-market matched control firms (Barber and Lyon (1997)). We are not aware of any paper in the literature that documents the long-run price pattern subsequent to index additions. Our results suggest that we need to examine a sufficiently long period following index addition given that Russell 1000 addition firms start exhibiting reversal in stock returns from a period beginning five months following reconstitution. Our contribution to the literature is threefold. First, through our empirical tests we are able to disentangle the information based explanation of the announcement effects of index additions from the broader explanation that demand curves for stock slope downward. Evidence in prior studies is unable to completely rule out that the two explanations are not mutually exclusive. While we acknowledge that the two explanations may presumably co-exist, we are able to somewhat isolate the downward sloping demand curve story in examining the information content of index additions. Second, we show the importance of controlling for prior performance for studies that examine analyst forecasting behavior around corporate events. Because the likelihood of observing some corporate event is not completely independent of prior performance, it is important to control for this before examining how financial analysts revise their expectations for earnings. Third, we examine long-run stock returns following index additions to detect permanent shifts in stock price. While our evidence for S&P 500 firms suggests that abnormal stock returns following index additions are not reversed in the two years following the event, results for Russell 1000 reconstitutions are quite the opposite. In the two years following index addition, Russell 1000 firms experience a 8

9 complete reversal in stock returns; the reversal begins in five months subsequent to index addition. Our evidence suggests that the choice of the time interval is quite critical in making inferences about the permanent shifts in stock price. The remainder of the paper is organized as follows. Section 2 describes the data and methodology for the study, Section 3 discusses our main results, and Section 4 concludes. 2. Background, Data and Methodology In this section, we present the institutional details regarding addition of firms to the S&P 500 index and the Russell 1000 index. Specifically, we contrast the criteria for being included in the S&P 500 and the Russell 1000 indices. We then describe the data and methodology. A. Institutional Details The S&P 500 constituents are determined by the Index committee based on a set of criteria including market capitalization, liquidity, industry representation, and financial viability. Recently, the committee has also required that the company on the index be domiciled within the United States and have public float of at least 50%. The constituents of the index are changed if a company in the index violates one or more of the criteria or is involved in merger, acquisition or substantial restructuring. The changes to the index are announced throughout the year, typically a week to 10 days ahead of the effective date. In contrast to the S&P Index committee s discretionary choice of companies to be added to the index, Russell 1000 index constituents are the largest 1000 firms at the end of May every year. Specifically, firms traded on all major U.S. exchanges are sorted on 9

10 market capitalization at the end of May of each year and the top 1000 firms based on market cap are assigned membership to the Russell 1000 index. A preliminary list of constituents, based on the May end market cap, is announced in the second week of June, the final reconstituted list is announced in the fourth week of June, and the index changes are effective at the end of June. We utilize these differences in the index addition process as the experimental setting to examine the information content of index additions. If index additions convey information, we would expect the firms added to the Russell 1000 index to convey less information relative to S&P 500 firms, due to the transparent and objective rules based selection of Russell constituents. B. Data We expanded the sample of S&P 500 index additions provided in Jeffrey Wurgler s website ( with index additions for years 2001 through 2003 from Standard & Poors Website. Many of the index changes are a result of corporate events such as a merger, acquisition or restructuring. We drop companies if they were added to the index as a result of these corporate transactions. We are left with 309 firms that are eligible for further analysis. We also require that these sample companies have analysts earnings forecasts data in I/B/E/S database and stock return data in CRSP database. We begin our sample period in 1987 because I/B/E/S data on analysts earnings forecasts are available starting in Our sample period ends in We obtain monthly Russell 1000 index constituents for the years 1990 through 2005 from the Russell Investments Group. Since Russell indices are reconstituted once a 10

11 year in June, we can identify additions to the Russell 1000 index. We require that these companies also have analysts earnings forecast data in the I/B/E/S database and stock return data in the CRSP database. We are left with 332 companies for further analysis. C. Methodology We use analysts forecasts of expected earnings as a proxy for market expectations. If the index addition event has any information content regarding the added firms future financial performance, we would expect analysts to impound this information into expectations of future earnings. Adopting the procedure employed in Denis et al. (2003), we examine the change in analysts expectations around the index addition event. We also require that index addition announcements occur at least three months prior to the end of the fiscal year end for the forecast to be considered a current year forecast. If not, we use the forecast for the next year as the current year forecast and use that measure in estimating the change in expectations. Also, to control for the analysts tendency to walk down their forecasts as the fiscal year end approaches, we need to use appropriate benchmarks when interpreting the change in expectations around index additions. In choosing our benchmark portfolios, we identify companies that could have been added to the index, but were not added. In the case of additions to the S&P 500 index, Standard and Poors criteria suggest some important firm characteristics that feature in the selection process, namely industry representation, size and liquidity. Hence the benchmark portfolio consists of firms matched with the index addition firms on industry, market capitalization and liquidity (we refer to these firms as ISL firms hereinafter). Companies that have data in 11

12 both CRSP and I/B/E/S databases, but are not current members of the S&P 500 index, are sorted into one of the 12 Fama-French industry portfolios based on their SIC codes. Each industry portfolio is divided into three portfolios based on their market capitalization. Finally, within each industry and market capitalization portfolio, firms are sorted into three liquidity portfolios. We define liquidity as the three-year average of monthly volume divided by shares outstanding for that company at the end of the month. This sorting procedure results in 108 portfolios. Each newly added stock is compared with its appropriate industry, size and liquidity portfolio. In the case of Russell 1000 index, the main criteria is the market capitalization at the end of May in the year the company is added to the index. Hence we form benchmark portfolios based on industry and market capitalization and require that the companies have market capitalization rank greater For the sample companies and the companies in the benchmark portfolios, we estimate the change in expectations using the following procedure. To calculate preannouncement earnings expectations for a given company, we retain the most recent forecast made by all analysts covering the firm before the announcement date, as long the most recent forecast is no more than 4 months old. We use the average of these analysts forecasts as the consensus forecast before the index addition. To calculate postannouncement earnings expectations for a given company, we retain the first forecast after the index addition, as long as it is not more than 4 months away from the index addition announcement. We use forecasts from analysts who were following the 8 This procedure is a little bit different from our procedure for selecting control firms for S&P 500 index additions. With Russell 1000, the main factor for selection is firm size so we are picking control firms that are closest to the sample firm in terms of market cap. We also control for industry representation because analysts expectations of earnings presumably incorporate industry-wide factors, at least in part. 12

13 company prior to the index addition in order to avoid inducing any bias by including new analysts. We use the average of these analysts forecasts as the consensus forecast after the index addition. We calculate the raw changes in forecasts by subtracting the consensus EPS forecast before the announcement from the consensus EPS forecast after the announcement as FE i = FE i, after FE i, before where FE i is the change in the EPS forecast for company i, FE i,before is the consensus EPS forecast before the announcement for company i, and FE i,after is the consensus EPS forecast before the announcement for company i. Further, to standardize by share price, we divide the change in the EPS forecast by the company s stock price as of the end of the month prior to the announcement month as PFE i = (FE i, after FE i, before )/P i, t-1 where PFE i is the change in the EPS forecast for company i standardized by share price, FE i,after and FE i,before are as defined above, and P i,t-1 is company i stock price as of the end of the month prior to the announcement month. To standardize by earnings per share, for those companies that have a positive consensus EPS forecast before the announcement, we divide the change in the forecast by the consensus EPS forecast before the announcement as EFE i = (FE i, after FE i, before )/ FE i, before where EFE i is the change in the EPS forecast for company i standardized by EPS and FE i,after and FE i,before are as defined above. 13

14 3. Findings A. Change in analyst expectations of earnings In this section, we discuss the findings from studying the changes in analysts expectations, using the methodology discussed in Section 2 and present the results in Table 1. Changes in expectations for the additions to the S&P 500 and Russell 1000 indices are presented in Panel A and Panel B respectively. We find that the firms added to the S&P 500 index experience a statistically insignificant decrease in analysts forecasts of $0.013 (p-value = 0.459). As discussed earlier, we need to compare these results with the change in expectations for the benchmark portfolio, to control for analysts tendency to walk down their earnings forecasts as the fiscal year end approaches. We find that analysts lower earnings forecasts by $0.068 for the ISL matching firms, leading to a difference of $0.055 between the sample and control firms. This difference is statistically significant at the 1% level. In addition, standardizing by share price and earnings prior to index addition, we find a difference of 0.15% and 5.97% also significant at the 1% level 9. We find similar results for the firms added to the Russell 1000 index. The sample firms experience statistically insignificant increases in analysts forecasts of $0.01 (pvalue=0.257). Nevertheless, the difference in earnings forecasts between sample firms and control firms is positive $0.03 and significant at the 5% level. In addition, standardizing by share price, we find a difference of 0.11% also significant at the 5% level. When we standardize by earnings forecasts prior to the index addition, the difference is 2.58% but not statistically significant at conventional levels. 9 Although we report mean values in the paper, we also compared the median values of the sample and control firms and found qualitatively similar results. 14

15 Overall, these results suggest that market participants positively revise their expectations for the financial performance of firms getting added to both the S&P 500 index and the Russell 1000 index. Our results for the S&P 500 addition firms mirror the findings documented in Denis et al (2003) for S&P 500 index additions from 1987 to The interesting question is why the Russell 1000 additions are associated with upward revisions in earnings forecasts, relative to control firms. We examine the effect of prior performance in the subsequent sub-section. B. Performance prior to index addition Previous studies (Givoly and Lakonishok (1979), Brown, Foster and Noreen (1985) and Abarbanell (1991)) document the positive relation between prior stock returns and earnings forecast revisions. If prior stock prices contain information about future earnings, then financial analysts likely impound this information in formulating their expectations of earnings. Therefore we need to examine if there is a difference between the prior performance of sample and benchmark firms, and control for this in our tests if there are significant differences. We focus on the stock return performance in the year prior to the index addition. For the firms added to the S&P 500 index, we use the time period between one year prior to the announcement date and the day before announcement. Since the Russell 1000 index is reconstituted in June, we use the 12 month period (June to May) prior to the reconstitution. In Table 2, we summarize the raw buy-and-hold returns and marketadjusted returns for sample and control firms. Panel A provides comparison of returns for the sample firms that are added to the S&P 500 index and the control firms; Panel B 15

16 compares the returns of sample firms that are added to the Russell 1000 index and the control firms. The results indicate that the sample firms substantially outperform the benchmark portfolio in the year prior to index addition. The difference in market-adjusted stock returns between the sample firms added to the S&P 500 index and median of the benchmark portfolio is 17.6% in the year prior to the index addition. This difference is also statistically significant at the 1% level. The sample firms that are added to the Russell 1000 index exhibit an even greater difference relative to benchmark firms; 73.34%. This is possibly due to the fact that firms are added to the Russell 1000 index based on market capitalization rather than a subjective decision by the Index committee. For a firm to enter the Russell 1000 index, either as a result of an upgrade from the Russell 2000 index or a completely new addition, the firm would have had to perform well to dislodge an existing firm from the Russell 1000 index. These results, in conjunction with previous studies that show the influence of past performance on analysts forecast, warrant a re-examination of the change in analyst forecasts. C. Earnings expectations controlling for prior performance In order to control for past performance, we seek to identify control firms with performance that resembles that of the sample firms and also satisfy the criteria for index addition. Therefore, we start with the benchmark portfolio formed based on industry, size and liquidity (industry and size in case of Russell 1000 firms) and from this subsample we identify five matching firms with the closest raw stock returns in the year prior to date of announcement of index addition. Changes in expectations for the 16

17 additions to the S&P 500 and Russell 1000 indices are presented in Panel A and Panel B respectively of Table 3. We find that the analyst forecasts for the matched firms still decrease after the index addition announcement. However, the magnitude of the difference in change in expectations of sample firms relative to control firms is now lower compared to what we found earlier without controlling for prior performance. In the case of S&P 500 firms the change in expectations for the control firms is $-0.02 (p-values: 0.007) and the difference between the sample and control firms is $ When standardized by share price and earnings expectations prior to the index addition, the difference is 0.15% and 3.77% respectively. These differences are statistically significant at the 10% and 5% level respectively. These results lead us to conclude that firms that are added to the S&P 500 index experience greater earnings expectations after the index addition relative to control firms after even after we control for prior performance. Panel B provides the change in analysts forecasts for firms added to the Russell 1000 index relative to control firms. Results indicate that the change in earnings expectations for sample firms is lower than matched firms change in earnings expectations. The difference is $0.033 and when standardized by share price and consensus earnings prior to index addition, the difference is 0.068% and 1.82% respectively. However these differences are not statistically significant at conventional levels of statistical significance. These results indicate that the change in analysts earnings expectations for firms added to the Russell 1000 index is not different from change in analysts earnings expectations for similar matched firms. The significant results we found earlier were likely driven by prior performance. 17

18 Overall, the results suggest that analysts react positively to firms added to the S&P 500 index when measured by their earnings forecasts. These results persist even after controlling for the influence of the firms past performance on analysts expectations. On the other hand, firms added to the Russell 1000 index experience an increase in earnings expectations. This increased earnings expectations seem to be primarily driven by past performance of these companies. Putting these pieces of results together, we conclude that the market participants treat additions to the S&P 500 index as information events, after we account for the alternative (but not mutually exclusive) explanation that demand curves for stocks slope downward. D. Stock returns following index additions The information-based story suggests that firms added to the S&P 500 index are likely to witness a permanent increase in stock price subsequent to being added to the index. In contrast, the lack of information content associated with Russell 1000 reconstitutions suggests that Russell 1000 firms are less likely to witness a permanent increase in stock price. We examine this hypothesis using abnormal stock returns in the two years subsequent to index additions. Following Barber and Lyon (1997), for each index addition firm, we find a control firm with a market value of equity that lies within 70% to 130% of the sample firm s market value of equity. Within this subset of possible matches, we select a control firm that has the closest book-to-market value of equity. We measure market value of equity at the end of June of the year t, i.e. the year of addition of the firm to the index. We measure book-to-market value of equity at the end of year t-1. We calculate the buy-and- 18

19 hold excess returns (BHER) as the difference in buy-and-hold returns between the sample firm and the matched control firm. In Panel A of Table 4, we report BHER for S&P 500 index addition firms. Year 0 is the fiscal year the sample firm gets added to the index. Results suggest that S&P 500 index addition firms outperform control firms in the year prior to being added to the index. The mean (median) BHER is 17.39% (14.31%). The mean BHER is statistically significant at the 10% level while the median BHER is significant at the 1% level. This is consistent with our findings reported in Table 2 where we define control firms as a portfolio of firms matched on industry, size and liquidity. Starting in the year of the addition to the index and leading up to two years following the index addition, we find no evidence of return reversals. The BHER in Year 0, Year +1 and Years +1 and +2 combined is not statistically distinguishable from 0. For the Russell 1000 reconstitution sample (see Panel B of Table 4), we find a significant run-up in the year prior to the index addition. The mean (median) BHER is 39.05% (12.44%). These numbers are statistically significant at conventional levels. This is consistent with our findings for Russell 1000 firms reported in Table 2 where we define control firms as a portfolio of firms matched on industry and size. The positive abnormal returns continue in the year of the addition. The mean (median) BHER is 26.54% (10.07%) and is significant at the 1% level. Interestingly, we find that stock returns start to reverse in the years following index addition. In the year subsequent to the index addition, the mean (median) BHER is -8.1% (-11.99%) and is significant at conventional levels. For the two years following index addition, i.e. Years +1 and +2 combined, the mean (median) BHER is % (-13.46%). The median BHER is 19

20 statistically significant at the 5% level. Our overall results indicate that firms that get added to the Russell 1000 index witness a reversal in stock returns in the two years subsequent to being added to the index. Results in Table 4 are reported on a fiscal year basis. In order to understand how the sample firms contrast to matched control firms in event time, we compare the buyand-hold returns on a daily basis with the matching firms from day -30 to day +504 relative to index announcement date. Results are presented in Figure 1. Panel A results indicate that S&P 500 additions firms experience significant abnormal return around the effective date and no subsequent reversal in the two year period subsequent to the index addition. In contrast, the Russell 1000 companies experience reversal in returns 100 trading days after index reconstitution. Further, these results point to the importance of examining the long term returns beyond the immediate window examined in previous studies. Overall, our results indicate that the S&P 500 index addition firms witness a permanent increase in stock price for a period of up to two years following index addition. The Russell 1000 index addition firms, on the other hand, witness a run up in the period leading to the event that is more than fully reversed in the two years following index addition. Our findings corroborate the hypothesis that S&P 500 index additions are not completely free of information. They also indicate that it is important for researchers to examine a long time period following index additions considering that for the Russell 1000 firms we start to witness stock return reversals only five months subsequent to index addition. 20

21 4. Conclusion and discussion In this paper, we reexamine the debate surrounding the relative information content of S&P 500 index additions. Some studies have used the S&P 500 index addition events to motivate the investigation of whether the demand curves for stock slope downward, and have found evidence in support of this hypothesis. Nevertheless, this assumes that index additions are completely free of any information content concerning the firm s future prospects. Evidence from a growing body of literature suggests that the index additions are not completely free of information. While separating these two stories could provide us interesting insights, as a practical matter it becomes very hard to isolate this using only a sample of S&P 500 index additions. We embark on an alternative exercise by examining whether S&P 500 index additions have any information content by comparing this to a sample of firms that get added to the Russell 1000 index. We use evidence in prior literature on Russell index reconstitutions that finds support for the idea that demand curves for stocks added to this index slope downward. We also note in our study that controlling for prior performance is quite important. Firms added to the S&P 500 index and the Russell 1000 index exhibit strong performance before additions. After we control for size, industry representation, liquidity and prior performance, we find evidence that financial analysts revise their earnings expectations upward for firms added to the S&P 500 index. Nevertheless, we do not find any evidence that financial analysts revise their earnings expectations for firms added to the Russell 1000 index. We interpret this as being support for the hypothesis that S&P 500 index additions are not completely free of information. 21

22 We provide complementary evidence by examining stock returns in the period following index addition. Our results suggest that the S&P 500 index firms witness an increase in stock price following index addition that is not reversed for a period of up to two years; in other words the positive shift in stock price is permanent. On the other hand, the Russell 1000 firms witness a run up in the period leading to the index addition that is more than fully reversed in the two years following the event; in other words the positive shift in stock price in transitory. Our study complements evidence in existing literature that finds that S&P 500 index additions are not completely free of information. While it is possible that the subjectivity involved in the selection process reveals new information about the firm to the market during the time of index addition announcement, it is also conceivable that the addition of the firm to the index causes firm performance to improve. For example, improved visibility of the firm in the financial market may have positive spillover effects in the product markets thereby leading to better firm performance subsequent to index addition. Financial analysts, cognizant of this effect, impound this information, among other pieces of information, in formulating their revised forecasts of earnings. Examining these interactions could be an interesting topic for future research. 22

23 References Abarbanell, J.S., 1991, Do Analysts Earnings Forecasts Incorporate Information in Prior Stock Price Changes? Journal of Accounting and Economics 14, Becker-Blease, J.R. and D.L. Paul, 2006, Stock Liquidity and Investment Opportunities: Evidence from Index Additions, Financial Management 35, Beneish, M.D. and R.E. Whaley, 1996, An Anatomy of the S&P Game : The Effects of Changing the Rules, Journal of Finance 51, Biktimirov, E.N., A.R. Cowan, and B.D. Jordan, 2004, Do Demand Curve For Small Stocks Slope Down? Journal of Financial Research 27, Brown, P., G. Foster, and E. Noreen, 1985, Security analyst multi-year earnings forecasts and the capital market, Studies in Accounting Research 21, American Accounting Association. Cai, J., 2007, What s in the News? Information Content of S&P 500 Additions, Financial Management, Forthcoming. Carino, D.R., and M. Pritamani, 2007, Price Pressure at Russell Index Reconstitution, Russell Investment Group working paper. Chen, H., G. Noronha, and V. Singal, 2004, The Asymmetric Price Response to S&P 500 Index Additions and Deletions: Evidence and Explanation, Journal of Finance 59, Chen, H., 2006, On Russell Index Reconstitution, Review of Quantitative Finance and Accounting 26, Denis, D.K., J. J. McConnell, A. V. Ovtchinnikov, and Y. Yu, 2003, S&P Additions and Earnings Expectations, Journal of Finance 58, Dhillon, U. and H. Johnson, 1991, Changes in the Standard and Poor's 500 List, Journal of Business 64, Elliott, W.B., B.F. Van Ness, M.D. Walker, and R.S. Warr, 2006, What Drives the S&P 500 Inclusion Effect? An Analytical Survey, Financial Management 35, Givoly, D. and J. Lakonishok, 1979, The Information Content of Financial Analysts Forecasts of Earnings, Journal of Accounting and Economics 1, Harris, L. and E. 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,

24 Hegde, S. and J.B. McDermott, 2003, The Liquidity Effects of Revisions to the S&P 500 Index: An Empirical Analysis, Journal of Financial Markets 6, Jain, P., 1987, The Effect on Stock Price of Inclusion in or Exclusion from the S&P 500, Financial Analysts Journal 43, Lynch, A.W. and R.R. Mendenhall, 1997, New Evidence on Stock Price Effects Associated with Changes in the S&P 500 Index, Journal of Business 70, McNichols, M. and P.C. O Brien, 1997, Self-Selection and Analyst Coverage, Journal of Accounting Research 35, Madhavan, A., 2003, The Russell Reconstitution Effect, Financial Analysts Journal 59, Shleifer, A., 1986, Do Demand Curves Slope Down? Journal of Finance 41, Wurgler, J. and E. Zhuravskaya, 2002, Does Arbitrage Flatten Demand Curves for Stocks? Journal of Business 75,

25 Figure 1: Comparison of sample and match firm mean buy and hold excess return for index additions In this figure, we present the buy and Hold excess return for firms added to the S&P 500 index and Russell 1000 index from -30 to +504 relative to the index addition date. The excess performance measures are calculated as the performance of the sample firm minus the performance of a matched control firm. The buy-and-hold excess stock return is calculated by following the approach in Barber and Lyon (1997). Panel A: Buy and Hold Excess Return for S&P 500 index additions relative to addition date Panel B: Buy and Hold Excess Return for Russell 1000 index additions relative to addition date 25

26 Table 1: Changes in Analysts EPS forecasts for companies added to the indices Forecasts of EPS are obtained from I/B/E/S database for firms that are added to the S&P 500 index. Average EPS forecasts preceding the index addition announcements are compared with average EPS forecasts following the announcements. Changes in the average EPS forecasts for the sample and control firms are reported in this table. The sample in Panel A consists of 309 firms added to the S&P 500 index from 1987 through The control firms are firms in the same Fama-French 12 industry portfolios and the same size and liquidity portfolios as the newly added firms. The sample in Panel B consists for 332 firms added to the Russell 1000 index from 1990 through The control firms are firms in the same Fama-French 12 industry portfolios and the same size portfolios as the newly added firms. Mean Difference is the average of the differences between the change in EPS forecasts for sample and control firms. p-values in parenthesis indicate whether the reported numbers above are significantly different from zero. Panel A: Companies added to the S&P 500 Index Sample Size Sample Firms Control Firms Mean Difference EPS Forecast Change (0.459) (0.001) (0.006) EPS Forecast Change % -0.20% 0.15% standardized by price (0.157) (0.001) (0.001) EPS Forecast Change % -6.41% 5.97% standardized by EPS (0.931) (0.001) (0.001) Panel B: Companies added to the Russell 1000 Index Sample Size Sample Firms Control Firms Mean Difference EPS Forecast Change (0.257) (0.018) (0.038) EPS Forecast Change % -0.07% 0.11% standardized by price (0.296) (0.038) (.034) EPS Forecast Change % -0.58% 2.46% standardized by EPS (0.525) (0.059) (0.410) 26

27 Table 2: Performance in the prior year This table reports the average stock returns of firms in the year prior to the index addition. The corresponding numbers for the control firms are the median stock returns of the portfolio of benchmark firms. Stock returns are defined as the total buy-and-hold returns as well as the market-adjusted stock returns using the CRSP value weighted index. Panel A presents results for the S&P 500 index additions, and Panel B presents results for the Russell 1000 index additions. p-values in parenthesis indicate whether the reported numbers above are significantly different from zero. Panel A: Prior year performance of firms added to the S&P 500 index Sample Size Sample Firms Control Firms Difference Raw Returns % 31.42% 9.46% (0.001) Market Adjusted Return % 13.55% 17.57% (CRSP Value Weighted Index) (0.001) Panel B: Prior year performance of firms added to the Russell 1000 index Sample Size Sample Firms Control Firms Difference Raw Returns % 37.04% 44.30% (0.001) Market Adjusted Return % 24.98% 73.34% (CRSP Value Weighted Index) (0.001) 27

28 Table 3: Changes in Analysts EPS forecasts for companies controlling for Performance Forecasts of EPS are obtained from I/B/E/S database for firms that are added to the S&P 500 index. Average EPS forecasts preceding the index addition announcements are compared with average EPS forecasts following the announcements. Changes in the average EPS forecasts for the sample and control firms are reported in this table. The sample in Panel A consists of 309 firms added to the S&P 500 index from 1987 through The control firms are 5 firms with similar returns and in the same Fama-French 12 industry portfolios and the same size and liquidity portfolios as the newly added firms. The sample in Panel B consists for 213 firms added to the Russell 1000 index from 1990 through The control firms are 5 firms with similar returns and in the same Fama-French 12 industry portfolios and the same size portfolios as the newly added firms. Mean Difference is the average of the differences between the change in EPS forecasts for sample and control firms. p-values in parenthesis indicate whether the reported numbers above are significantly different from zero. Panel A: Companies added to the S&P 500 Index Sample Size Sample Firms Control Firms Difference EPS Forecast Change (0.459) (0.007) (0.64) EPS Forecast Change % -0.19% 0.15% standardized by price (0.157) (0.013) (0.071) EPS Forecast Change % -3.73% 3.77% standardized by EPS (0.931) (0.014) (0.025) Panel B: Companies added to the Russell 1000 Index Sample Size Sample Firms Control Firms Difference EPS Forecast Change (0.333) (0.218) (0.118) EPS Forecast Change % -0.04% 0.068% standardized by price (0.561) (0.470) (0.670) EPS Forecast Change % -1.2% 1.82% standardized by EPS (0.907) (0.306) (0.354) 28

29 Table 4: Performance surrounding addition to the index This table provides details of performance surrounding the addition companies to S&P 500 index during the period between 1987 through 2003 and Russell 1000 index between 1990 and The excess performance measures are calculated as the performance of the sample firm minus the performance of a matched control firm. The buy-and-hold excess stock return is calculated by following the approach in Barber and Lyon (1997). For each sample firm, we find a control firm with a market value of equity that lies within 70% to 130% of the sample firm s market value of equity. Within this subset of possible matches, we select a control firm that has the closest book-to-market value of equity. We measure market value of equity at the end of June of the year t, i.e. the year of addition of the firm to the index. We measure book-to-market value of equity at the end of year t-1. The p-values reported in the table are for testing whether the mean and median excess performance measures are equal to zero. We use the t-test for testing the statistical significance of the mean, and the Wilcoxon sign rank test for testing the statistical significance of the median. Panel A: Companies added to the S&P 500 Index Year 1 Year 0 Year +1 Years +1 and +2 Mean BHER 17.39% 8.44% -3.11% -2.3% p-value Median BHER 14.31% 5.19% -2.95% -3.85% p-value N Panel B: Companies added to the Russell 1000 Index Year 1 Year 0 Year +1 Years +1 and +2 Mean BHER 39.05% 26.54% -8.10% % p-value Median BHER 12.44% 10.07% % % p-value N