The effects of a stock index: evidence from the annual rebalancing of the MSCI USA index.

ABSTRACT

This paper examines the effects of the annual rebalancing of the MSCI USA Index on the companies that were added to the index. Although we cannot find considerable announcement effects, our results show some weak evidence that an addition to the MSCI USA Index is good news. Overall, we interpret our findings as consistent with the price pressure hypothesis that is discovered in some previous index studies, which focus on the membership changes in the S&P 500 Index.

1. INTRODUCTION AND LITERATURE REVIEW

The purpose of a stock index is to measure the performance of a stock market, a targeted industry/sector, or a set of stocks in a market such as blue chips or mid-cap stocks. By a rough count, there are at least more than 200 U.S. stock-related indexes in existence. In the Wall Street Journal, more than 50 major stock indexes are reported daily. It is fair to say there is competition among various stock indexes. Whether or not a stock index is successful depends on the acceptance and use of this index by derivative contracts, institutional investors, money managers, newspapers, other media and the investing public. One of the most successful stock indexes in the U.S. is probably the Standard and Poor's 500 Composite Stock Price Index (Thereafter, the S&P 500 Index), which consists of 500 component/member stocks and is provided and maintained by Standard and Poor's Corporation. Naturally, a large portion of prior research focuses on this popular stock index.

Most existing papers examine how the additions to the S&P 500 Index affect the value of these added and deleted stocks. One common result is that the price of an added company increases significantly, resulting in the so-called "S&P effect." Also, some papers, which investigate the volume effect, find that the trading volume increases significantly for both additions and deletions. Although most papers find the significant influence of the S&P 500 Index, interpretations of results are not all the same. Shleifer (1986) suggests that his findings are consistent with the notion that the long-run demand curves of stocks are downward sloping. The result of Harris and Gurel (1986) indicates that the gains upon the announcement of additions are expected to reverse subsequently because the price pressures on the stocks affected are temporary. Lynch and Mendenhall (1997) argue that their findings support both the downward sloping demand curve and price pressure hypotheses. Also, Jain (1987) and Dhillon and Johnson (1991) contend that their findings are due to the information revealed by the changes (the information/signaling hypothesis). In addition, Harris and Gurel (1986), Dhillon and Johnson (1991), and Lynch and Mendenhall (1997) all find an increase in trading volume for stocks added, but none of them credits the price effects to the volume changes (the liquidity hypothesis).

In sum, most previous papers find an addition to the index is good news. One intuitive reason is that many assets, contracts and funds are tied to the S&P 500 Index. Therefore, when there are changes in component stocks, gains of added stocks are likely to be realized due to needed portfolio rebalancing. In addition to the existing interpretations of the "S&P effects," we can also say that the S&P 500 Index is a very successful product of Standard and Poor's Corporation. Indeed, this index has been a cash cow and the flagship index for the famous index provider. There is little doubt that an established, widely used stock index, such as the S&P 500 Index, can exert enormous influence on firm value. One question arises: How about other stock indexes such as the MSCI USA Index, provided by Morgan Stanley Capital International (MSCI)?

This paper investigates how component-stock changes in the MSCI USA Index affect added companies. It differs from previous studies that examine the effects of changes in the S&P 500 Index in that the rebalancing of the S&P 500 Index is based upon the need of industry representation, while most component-stock changes of the MSCI USA Index take place in annual rebalancing. By using a new stock index, we test the theories and hypotheses in a different setting.

We find that while some positive abnormal return is observed, the overall result of the MSCI USA Index rebalancing is consistent with the price pressure hypothesis in that the cumulative average abnormal return over a 56-day period is not significantly different from zero. In addition, we observe that the trading volume of an added company, on average, increases after the announcement of index membership changes.

The remainder of this paper is as follows. Section II discusses MSCI, index competition, and hypotheses. In Section III, we describe the data sources, methodology, and summary statistics. Section IV discusses the results of the event study. We conclude in Section V.

2. MSCI, INDEX COMPETITION, AND HYPOTHESES

2.1. Morgan Stanley Capital International (MSCI)

MSCI (Morgan Stanley Capital International), which is majority-owned by Morgan Stanley and minority-owned by the Capital Group Companies, Inc., provides global indices and benchmark-related products to investors worldwide. In particular, the main business of MSCI is to provide benchmark products and services to the investment-managing community, distribute index and company-level data, and license various MSCI indexes to third parties for the purpose of creating index funds, derivatives, proprietary and structured products.

MSCI claims that its indexes are the most widely used benchmarks by global portfolio managers. By August 2002, approximately $3 trillion of capital managed by institutional investors worldwide used MSCI indexes as benchmarks (MSCI News & Information, August 26, 2002). According to a recent survey conducted by Pensions & Investments, over 90% of international institutional equity assets in the U.S. are benchmarked to MSCI indexes. MSCI estimates that it has a similar market share in Asia. According to Merrill Lynch/Gallup surveys, MSCI is also a major index provider of choice for two thirds of Continental European fund managers compared to other cross-border index providers. Overall, more than 1,500 clients worldwide currently use the MSCI benchmarks.

2.2 Index Competition

Although MSCI is a dominant index provider for cross-border portfolio investments, the company lags behind Standard and Poor's Corporation in the U.S. index market because more U.S. investors investing domestically use S&P indexes as benchmarks. In a word, while MSCI is strong globally, it is weak in the U.S. market, in which Standard and Poor's Corporation dominates. Based on a survey conducted by Standard and Poor's Corporation (Annual Survey of S&P Indexes Assets, April 29, 2003), more than $900 billion of invested assets were indexed to major S&P indexes in 2002, within which about $841 billion was tied to the S&P 500 Index.

To expand its U.S. business, MSCI announced in August 2002 that it was developing a new family of U.S. equity indexes for domestic investors. MSCI also entered into a non-exclusive licensing agreement with Vanguard Group to enable the second-largest U.S. fund company to utilize its new index series (Vanguard Press Release, August 26, 2002). In the meantime, Vanguard Group was seeking shareholder approval on several investment policy changes for various funds and election of a board of trustees for all funds.

On April 3, 2003, the board of trustees of seven Vanguard stock index funds approved the fund company's recommendation to adopt new target benchmark indexes developed by MSCI, dealing a blow to Russell (with 1 Vanguard fund ceased to track its index) and Standard and Poor's Corporation (with 6 Vanguard funds ceased to track its indexes).

So, even though Standard and Poor's indexes have a dominant market share in the U.S. market, there are competing indexes marketed by Dow Jones, Inc., MSCI, Russell, and Wilshire Associates Incorporated. Since the power of the S&P 500 Index has been studied frequently, it would be interesting to investigate whether the changes in component stocks of another major stock index have similar effects, as observed in the S&P 500 Index. In this paper, we focus on the effects of the MSCI USA Index, a country index developed by MSCI.

2.3 Hypotheses

From previous studies, at least four hypotheses have been proposed and tested in order to explain the price/volume effects of the component-stock changes in major stock indexes. We discuss five major hypotheses as follows:

First, the price pressure hypothesis (PPH), advanced by Harris and Gurel (1986), asserts that including a stock in a stock index creates price pressure because of increased demand from some index mutual funds that track this stock index. The investment policy of these index mutual funds forces them to buy the added stock, hence pushing up its price. However, since the demand shift does not change the equilibrium value of the stock, the changed price is expected to reverse to its equilibrium level after the event. That is, the price pressure is likely to be transitory.

Second, if securities are not close substitutes for each other, the long-term demand curve for securities is less than perfectly elastic and hence, downward sloping. Under the downward sloping demand curve (DSDC) hypothesis, argued by Shleifer (1986), an increase in demand for an added stock, such as mandatory index-fund buying, would drive up the share price, given the supply for any particular security. The hypothesis does not anticipate a price reversal because the new equilibrium price reflects the excess demand and is the result of the new distribution of security holders. This hypothesis is also called the imperfect substitutes or distribution effect hypothesis in Harris and Gurel (1986).

Third, the information hypothesis (IH) contends that the announcement of an added stock to a key index is considered conveying new information about the future prospects of the added company, which were not known to investors before the announcement. If the stock market is efficient in transmitting new information, positive information about the added company should be incorporated quickly, hence increasing its price permanently. That is, the price effect is expected to be long-term.

Finally, the liquidity hypothesis (LH) maintains that the inclusion of a stock in an index may lead the stock to be traded more easily, reducing transaction costs and hence increasing in price. McInish and Wood (1992) find that as the trading volume of a stock increases, the bid-ask spread of this stock declines. The reason is that more active trading mitigates the asymmetric-information problem and reduces the cost of trading.

It should be noted that although the assumptions of the above five hypotheses are different, these hypotheses are not necessarily mutually exclusive. For example, the information hypothesis is positively related to the liquidity hypothesis because the announcement of an addition sends a positive signal to investors about the quality of the added company, hence increasing the liquidity of this company's shares.

3. RESEARCH METHOD

The market's response to the addition of firms to the MSCI USA Index is assessed by testing the statistical significance of abnormal returns surrounding the announcement date (April 30, 2003) and the effective date (May 31, 2003, Saturday). Historical stock return data, Russell 3000 Index (used as the market index in the market model) data, and volume data are obtained from Bloomberg for the period May 15, 2002 to June 5, 2003. The final sample includes sixty-four firms added to the MSCI USA Index. Eleven firms were omitted due to confounding releases of information during the event period. The following model is used to estimate the abnormal returns:

A[R.sub.it] = [R.sub.it] - ([[alpha].sub.i] + [[beta].sub.i.sup.*] [R.sub.mt]) (1)

where [R.sub.it] is the return of stock i on day t, [[alpha].sub.i] is the intercept, [[beta].sub.i] is the beta, and [R.sub.mt] is the return of the market (using the Russell 3000 Index) on day t. The estimation period is the 200-day period beginning May 15, 2002 and ending February 28, 2003. The statistical significance of the abnormal returns is determined using Mikkelson and Partch's (1988) Z-statistic:

Z = [1/[square root of N]] [N.summation over (i=1)] [[t2.summation over t=t1] [A[R.sub.it]/[square root of Var [t2.summation over (t=t1)]A[R.sub.it]]]] (2)

where [t.sub.1] is the first day of the examination period window, [t.sub.2] is the last day of the examination period window, N is the number of observations, and the denominator is the square root of the variance of the cumulated prediction error of firm i. This variance is defined to be:

VAR([t2.summation over (t=t1)] A[R.sub.it]) = [V.sup.2.sub.i][T + [[T.sup.2]/ED] + [[([t2.summation over (t=t1)] [R.sub.mt] - T[[bar.R].sub.m]).sup.2]/ED [ED.summation over (t=1)] [([R.sub.mt] - [[bar.R].sub.m]).sup.2]]] (3)

where

[V.sup.2.sub.i] = the residual variance of stock i's market model regression,

T = the number of days in the examination period,

ED is the number of days in the estimation period,

[R.sub.mt] = the market return on day t, and

[[bar.R].sub.m] = the mean market return during the estimation period.

The event window (day -3 to day +3) is used around the announcement date (4/30/03) and the effective date (5/31/03, Saturday) to pinpoint the timing of any average abnormal stock return. Positive and statistically significant average abnormal returns will suggest firm value is enhanced by inclusion of the firm in the MSCI USA Index.

Two-sample t-tests are used to test for changes in volume due to inclusion of the firms in the MSCI USA Index. The percent of shares traded is calculated for three periods. The pre-announcement period is March 24, 2003 to April 24, 2003, the post-announcement period is April 29, 2003 to May 30, 2003, and the post-effective date period is June 2, 2003 to July 2, 2003.

4. RESULTS AND DISCUSSIONS

Table 1 discloses the significance of the average abnormal returns surrounding the announcement date (April 30, 2003) and the effective date (May 31, 2003, Saturday) where sixty-four firms were added to the MSCI USA Index. The statistical significance of the abnormal returns is assessed using Mikkelson and Partch's (1988) Z-statistic. Average abnormal returns before the announcement date (April 25-April 29) are shown just to measure any pre-announcement leakage.

The first column inside the table shows the dates for which average abnormal returns are calculated, the second column shows the average abnormal returns, and the third column shows the z-statistics. The top third of the table pertains to average abnormal returns surrounding the announcement date and the middle third of the table pertains to average abnormal returns surrounding the effective date. The last third of the table pertains to cumulative average abnormal returns covering the entire period surrounding the announcement and effective dates.

The seven average abnormal returns surrounding the announcement date have mixed signs and none of them are statistically significant. The average abnormal returns surrounding the effective date also have mixed signs. One thing worth noting is that the average abnormal return on May 30, 2003, one day before the effective date, is 1.68% and is statistically significant at the 1% level. It seems that there was aggressive index-fund buying (or something similar) on May 30, 2003. It is not clear, however, that why index funds did not buy shares of added companies earlier. The average abnormal return on June 2, 2003 is -0.53%, and the average abnormal return on June 3, 2003 is -0.76%, which is statistically significant at the 5% level. These results seem to favor the price pressure hypothesis discussed previously.

In order to assess the price pressure hypothesis further cumulative average abnormal returns are disclosed at the bottom of Table 1. The cumulative average abnormal return for the period April 30, 2003 to May 30, 2003 is 3.25%, which is statistically significant at the 10% level. This result is largely due to the abnormal return of 1.68% on May 30, 2003. This can be considered as the "MSCI effect," which may be attributed to purchases of added stocks by portfolio managers attempting to mimic the performance of the MSCI USA Index. When this multi-day event window is extended (April 25, 2003 to July 15, 2003) the cumulative average abnormal return is 0.61%, which is not statistically significant. These findings suggest the inclusion of stock in the MSCI USA Index does not permanently increase share value on average.

Table 2 shows results pursuant to the volume of shares traded during the announcement date period and the effective date period. For each firm, the percentage of shares traded is calculated for three periods. The pre-announcement period is March 24, 2003 through April 24, 2003, the post-announcement period is April 29, 2003 through May 30, 2003, and the post-effective date period is June 2, 2003 through July 2, 2003.

Panel A of Table 2 pertains to comparing the average percent of shares traded during the post-effective date period to the post-announcement date period. A t-statistic above 1.96 suggests the percent of shares traded during the post-effective date period is higher than the percent of shares traded during the post-announcement date period. The level of statistical significance is 0.05. For the sixty-four firms, only two firms are associated with a t-statistic of 1.96 or above, suggesting these two firms experienced a rise in trading volume after the effective date. Seven of the firms are associated with a t-statistic of -1.96 or below, suggesting these firms experienced a decline in trading volume after the effective date. Fifty-five firms are associated with t-statistics between -1.96 and +1.96. Twenty-five of the t-statistics are above zero, while thirty-nine are below zero. These results do not seem to suggest trading volume changed on average between these two periods.

Panel B of Table 2 pertains to comparing the average percent of shares traded during the post-effective date period to the pre-announcement date period. Sixteen of the sixty-four t-statistics are 1.96 or above, and five are -1.96 or below. For forty-three firms the t-statistic is between -1.96 and +1.96. Forty-five of the t-statistics are above zero while 19 t-statistics are below zero. These results suggest twenty-five percent of the firms experienced an increase in trading volume between these two periods. The remaining seventy-five percent of the firms either experienced a decrease in trading volume or a statistically insignificant change in trading volume between these two periods.

Panel C of Table 2 discloses more results pursuant to trading volume. For each date in the effective date window (May 28, 2003 through June 5, 2003), the percent of shares traded on a specific date for a firm is divided by the average percent of shares traded during the pre-announcement date period for the firm. A factor greater than one suggests trading volume on the specific date is higher than average. It is notable that this factor is 2.96 for May 30, 2003, the date of the positive abnormal return of 1.68%. The percent of shares traded on this date is almost three times higher than the percent of shares traded on average during the pre-announcement period. This evidence supports the price pressure hypothesis elucidated previously.

5. SUMMARY AND CONCLUSIONS

After examining the abnormal price returns and volume changes of 64 firms added to the MSCI USA Index, three major results emerge. First, we find weak evidence on the announcement date that an addition to the index leads to a significant positive effect. This seems consistent with the fact that the S&P 500 Index dominates in the index-fund market while the MSCI USA Index is far behind the S&P 500 Index.

Second, the average abnormal return observed on May 30, 2003, one day before the effective date, is significantly positive, suggesting that some mutual funds or portfolios might actual follow the MSIC USA Index and hence needed to balance their holdings in response to the member-stock changes announced on April 30, 2003.

Finally, trading volume, on average, does increase significantly for a few days after the announcement date, supporting positive cumulative abnormal returns observed over two time periods, as shown at the bottom of Table 1.

Overall, our paper complements previous studies on indexes in that we investigate the effects of a non-S&P 500 Index. Our results suggest that the MSCI USA Index may be followed and/or tracked by some institutional investors who need to adjust their portfolios whenever there are changes in component stocks of this index.

REFERENCES

Dhillon, Upinder and Johnson, Herb, "Changes in the Standard and Poor's 500 List", Journal of Business_, Vol. 64 (1), 1991, 75-85.

Harris, Lawrence and Gurel, Eitan, "Price and Volume Effects Associated with Changes in the S&P 500 List: New Evidence for the Existence of Price Pressure", Journal of Finance, Vol. 41 (4), 1986, 815-829.

Jain, Prem C., "The Effect on Stock Price from Inclusion in or Exclusion from the S&P 500", Financial Analysts Journal, Vol. 43 (Jan.-Feb.), 1987, 58-65.

Lynch, Anthony W. and Mendenhall, Richard R., "New Evidence on Stock Price Effects Associated with Changes in the S&P 500 Index", Journal of Business, Vol. 70 (3), 1997, 351-383.

McInish, Thomas H. and Wood, Robert A., "An Analysis of Intraday Patterns in Bid/Ask Spreads for NYSE Stocks", Journal of Finance, Vol. 47 (2), 1992, 753-764.

Mikkelson, Wayne H., and Megan Partch M., "Withdrawn Security Offerings", Journal of Financial and Quantitative Analysis, Vol. 23 (2), 1988, 119-133.

Shleifer, Andrei, "Do Demand Curves for Stocks Slope Down"? Journal of Finance, Vol. 41 (3), 1986, 579-590.

Ingyu Chiou, Eastern Illinois University, Charleston, Illinois, USA

Stephen J. Larson, Eastern Illinois University, Charleston, Illinois, USA

Dr. Ingyu Chiou earned his Ph.D. at New York University in 1998. He is an Assistant Professor of Finance at Eastern Illinois University.

Dr. Stephen J. Larson earned his Ph.D. at Florida Atlantic University in 1998. He publishes mainly in the area of stock market overreaction. He is an Assistant Professor of Finance at Eastern Illinois University.

TABLE 1. THE STATISTICAL SIGNIFICANCE OF THE AVERAGE ABNORMAL RETURNS

The statistical significance of the average abnormal returns for
sixty-four firms added to the MSCI USA

Index is disclosed

Date              Average Abnormal Return   Z-Statistic

Announcement Date Period
4/25/2003         0.03%                     0.42
4/28/2003         -0.16%                    (0.58)
4/29/2003         -0.08%                    (0.63)
4/30/2003         0.21%                     0.69
5/1/2003          -0.10%                    (0.57)
5/2/2003          0.20%                     0.54
5/5/2003          0.06%                     0.07

Effective Date Period
5/28/2003         -0.06%                    (0.17)
5/29/2003         0.43%                     1.11
5/30/2003         1.68% ***                 8.49
6/2/2003          -0.53%                    (1.56)
6/3/2003          -0.76% **                 (2.84)
6/4/2003          0.27%                     0.30
6/5/2003          0.14%                     0.28

Cumulative Average Abnormal Returns
4/30 to 5/30/03   3.25% *    1.73
4/25 to 7/15/03   0.61%      0.24

Statistically significant at the 0.1 (*), 0.05 (**),
or 0.01 (***) level

TABLE 2. THE STATISTICAL SIGNIFICANCE OF CHANGES IN TRADING VOLUME

A summary of t-statistics is given for three periods pursuant to
sixty-four firms added to the MSCI USA Index. The pre-announcement
date period is March 24, 2003 to April 24, 2003, the
post-announcement date period is April 29, 2003 to May 30, 2003, and
the post-effective date period is June 2, 2003 to July 2,
2003. The number of trading days in each period is 23 so the cut-off
t-statistic for the 5% level is 2.0154.

Panel A. Changes in percent of shares traded: Post-effective date
period minus post-announcement date period.

Number of t-statistics 2.0154 or above: 2
Number of t-statistics between -2.0154 and +2.0154: 57
Number of t-statistics -2.0154 or below: 5

Number of positive t-statistics: 25
Number of negative t-statistics: 39

Panel B. Changes in percent of shares traded: Post-effective
date period minus pre-announcement date period.

Number of t-statistics 2.0154 or above: 16
Number of t-statistics between -2.0154 and +2.0154: 43
Number of t-statistics -2.0154 or below: 5

Number of positive t-statistics: 45
Number of negative t-statistics: 19

Panel C. Percent of shares traded on a specific date divided by the
average percentage of shares traded during the pre-announcement
period (March 24, 2003 through April 24, 2003).

Date        % Shares Traded/Average % shares
          traded during pre-announcement period

5/28/03                   1.26
5/29/03                   1.53
5/30/03                   2.96
6/2/03                    1.53
6/3/03                    1.21
6/4/03                    1.20
6/5/03                    1.34