Additions to and deletions from an open-ended market index: evidence from the Australian all ordinaries.

Abstract:

The shares of companies added to (deleted from) closed-end indices such as the S&P 500 experience significant positive (negative) abnormal returns. We examine companies added to or deleted from an open-ended market index where changes occur regularly and can be predicted in advance. Our results are broadly consistent with those for closed-end indices and are robust to benchmarks used to control for market, size and industry effects. Analysis of daily returns and volume shows significant effects around the change date. By construction, companies entering the index outperform index benchmarks prior to inclusion. After controlling for selection bias, additions outperformed over the 4 months prior to inclusion.

Keywords:

INDEX; ABNORMAL RETURN; TRADING VOLUME.

1. Introduction

Share market indices such as the Australian All Ordinaries Share Price Index (AOI) are intended to provide a broad measure of share price movements, but there is evidence that inclusion in (exclusion from) an index can directly affect the price of a company's shares. Several US studies report that stocks added to (deleted from) the S&P 500 Index experience abnormal returns that average about 3% (minus 1.5%) when the change is announced. Beneish and Whaley (1996) and Lynch and Mendenhall (1997) found that the total price effects associated with addition to (deletion from) the S&P500 Index since 1989 are as large as 7.2% (minus 14.1%).

In Australia, inclusion in (exclusion from) the AOI, which is an open-ended index, are determined by criteria based on market capitalisation and liquidity. (1) Any company that satisfies these criteria is included in the AOI. In contrast to the S&P 500 index, which is a closed-end index, the total number of companies in the AOI can vary over time and companies likely to be included in (removed from) the AOI can be identified in advance with considerable confidence. (2) Therefore, astute investors and possibly `risk arbitrageurs' may be able to come into the Australian market at a much earlier date than is possible in the US. (3) Consequently, the market impact of changes in the composition of the AOI may differ from that associated with changes in the S&P 500.

During 1998, the possible effects of removal from the AOI on the share prices and capital-raising ability of small companies became prominent issues in the context of changes made by the Australian Stock Exchange (ASX) to the rules that determine the composition of the AOI. The changes infuriated funds managers and company directors, some of whom attributed share price falls to actual or expected removal from the AOI. There were suggestions of legal action against the ASX by directors of some excluded companies, and a threat to establish a rival index. The controversy was even linked to the possible establishment of a rival stock exchange. (4)

This paper has two purposes. First, we analyse the effects of additions to and deletions from the AOI around the actual change date, using daily data. We examine whether the abnormal returns and abnormal trading activity documented in the US studies, and attributed largely to trading by index funds, are also found in Australia which has a smaller index fund sector. Second, we examine returns on the shares of companies added to the index over a 25-month period centred on the change date, with a particular focus on the 6 months immediately preceding the actual change month. The objective of this analysis is to ascertain whether astute investors may be able to earn abnormal returns, and the possible timing of these returns, after removing the selection bias inherent in the criteria used by the ASX to select new entrants to the AOI.

Section 2 of this paper contains the literature review. It reviews the US and UK evidence on the effects of additions to and deletions from share market indices as well as possible explanations for the observed price and volume effects. Section 3 outlines the construction of the AOI prior to 1 July 1998 and the controversial changes introduced by the ASX with effect from 1 July 1998. Section 4 outlines the nature and sources of the data and the methodology used to examine the price and trading activity effects associated with addition to and deletion from the AOI. Section 5 sets out the results and section 6 presents conclusions.

2. Literature Review

2.1 US Evidence on Index Additions and Deletions

Several US studies have found that changes in the composition of the S&P 500 are associated with significant changes in both share prices and trading volume. (5) Goetzmann and Garry (1986) reported that six of the seven stocks deleted from the index on 30 November 1983 experienced a temporary increase in volume and a permanent price decline that averaged almost 2% on the first trading day after the change was announced and implemented. (6)

Harris and Gurel (1986) studied both additions to and deletions from the S&P 500 over the period 1973 to 1983. Additions to the S&P 500 were associated with permanent volume increases and temporary price increases that were significantly larger during the second half of the sample period. For 1978 to 1983, the 84 additions recorded a mean excess return of 3.13% on the announcement day, (7) whereas for the 110 additions from 1973 to 1977 the mean excess return was only 0.21%. For 1978 to 1983 the 13 deletions that were not caused by a merger, bankruptcy or tender offer, experienced a temporary increase in volume on the announcement day and a mean price decrease of 1.4% which was not reversed over the next several days.

Shleifer (1986) studied additions to the S&P 500 over the period 1966 to 1983, with market model parameters estimated separately using daily returns before and after inclusion in the index. (8) Prior to September 1976, no significant announcement-day price effects were detected for a sample of 144 additions. (9) For 102 additions in the period September 1976 to 1980 the mean abnormal return on the announcement day was 2.27%, and for a 1981 to 1983 sub-sample it was 3.19%.

Jain (1987) studied both additions to and deletions from the S&P 500 from November 1977 to December 1983. For 87 additions the mean excess return on the announcement day, estimated using the market model, was 3.07%, while for 22 deletions the mean announcement day excess return was minus 1.16%.

Dhillon and Johnson (1991) examined 187 additions to the S&P 500 from 1978 to 1988 again using daily data and the market model. (10) The announcement effect documented in earlier studies continued unabated in the 1984 to 1988 period. Specifically, the mean announcement-day excess return for this period was 3.334%. Dhillon and Johnson (1991) tracked returns for 60 trading days after the announcement and, in contrast to Harris and Gurel (1986), found no evidence of any price reversal. (11)

Lynch and Mendenhall (1997) (LM) examined S&P 500 additions and deletions from March 1990 to April 1995. This later period is of particular interest because since October 1989 Standard and Poor's has, whenever possible, announced changes in the composition of the index one week prior to their implementation. (12) Additions (deletions) were associated with significant positive (negative) abnormal returns of 3.158% (minus 6.263%) on the announcement day. (13) Further abnormal returns were then recorded between the announcement day and the change day, followed by a partial reversal on and immediately after the change day, which suggests that there were temporary price effects caused by index fund trading. The total price effect of being added to or deleted from the index was estimated using the Cumulative Abnormal Return (CAR) from the announcement day through 10 days after the change day. The mean CAR over this window was 4.8% for additions and minus 14.1% for deletions.

LM found that additions and deletions were also associated with similar abnormal volume. It was significant on six consecutive days starting with the announcement day and largest on the day prior to the change. This finding is consistent with index funds doing most of their trading on this day to minimise tracking error.

All the studies reviewed so far used closing price data. The study by Beneish and Whaley (1996) of additions to the S&P500 differs from the previous studies in that by using opening and closing stock price data they were able to separate the overnight return from the intra-day return on the day immediately following the announcement day. The sample period from January 1986 to June 1994 was split into two periods delineated by October 1989, the date that S&P began pre-announcing changes to the S&P 500 index. Their return results in the later sub-period are similar to those found by Lynch and Mendenhall (1997). Beneish and Whaley (1996) use the S&P 500 futures as the adjustment for market return and include the overnight return in their reported results. They report that the overnight return is 3.1% while the total abnormal return from the announcement day close to the change day close is 7.2%.

Further, Beneish and Whaley (1996) analyse trading activity using three measures; namely, trading volume, trade size and quoted bid-ask spreads. They find that trading volume increases from the announcement day and peaks on the change day. Average trade sizes are also reported to be larger. They found that 'trading activity' is abnormal and statistically significant when compared to a pre-announcement period. They attribute the return and trading activity results to index funds, which became increasingly important over their sample period, and the actions of risk arbitrageurs. Their results for the bid-ask spread are not as strong when compared to the other two measures of trading activity.

Price adjustments associated with addition to or removal from an index have also been examined in the UK. Brealey (2000) found that stocks added to both the FTSE-All Share and the FTSE 100 indexes experienced, on average, a positive abnormal return over an 11-day period that included both the announcement and the effective day. However, this abnormal return was not significant. Deletions were associated with significant negative abnormal returns. The CAR over the eleven days was -4.5% for deletions from the All-Share index and -2.0% in the case of the FTSE 100 index. (14)

In summary, several studies have reported significant price effects associated with changes in the composition of sharemarket indices, particularly the S&P 500. Over the period 1976 to 1988 when Standard and Poor's announced and implemented changes in the index sample simultaneously, additions (deletions) were associated with an average abnormal return of approximately 3% (minus 1.5%) on the first trading day after the change. The majority of studies found that the price changes were sustained over subsequent trading days. Since October 1989 when Standard and Poor's has generally announced index changes a week in advance, a price response which is both larger (particularly for deletions) and more complex has been found. The results of these US studies appear to be robust to variations in the methodology used. In addition, various researchers have documented abnormal trading activity after the announcement of index changes. They note that this effect has increased over time which may be related to the growth of index funds and the activities of risk arbitrageurs.

2.2 Explanations for Price and Volume Effects Associated with Index Changes

Four possible explanations have been proposed for the price effects associated with changes in the composition of sharemarket indices. One is that decisions to include or exclude a stock may convey information to investors (the information hypothesis). Standard and Poor's is known to prefer stable firms for its indices, so inclusion of a firm in the S&P 500 may be perceived as a reduction in the risk of its securities. (15) Another possibility is that heavy trading by index funds around the time of the change temporarily moves security prices away from their equilibrium values (the price pressure hypothesis).

As first recognised by Shleifer (1986) and Harris and Gurel (1986) a permanent price change associated with addition to or deletion from an index could also be explained by stocks having long run demand curves which are downward sloping (downward-sloping demand hypothesis). When a firm enters a major index such as the S&P 500, index funds will buy its shares which removes a substantial proportion of them from circulation. If demand curves for stocks slope downward, the reduced supply will cause an increase in the equilibrium price. Conversely, removal from an index should be accompanied by a price decrease.

Finally, inclusion in an index may have valuation consequences because it increases a stock's liquidity (the liquidity hypothesis). According to this view, inclusion may be followed by closer scrutiny of the company by investors, greater institutional interest in the stock and therefore an increase in public information about it. As a result, the stock will be held more widely, will become more liquid and the bid-ask spread will fall which lowers the required rate of return on the stock and leads to a price increase. (16)

Evidence on the validity of these competing hypotheses is mixed. The price pressure (PP) hypothesis is supported by Harris and Gurel (1986) while Shleifer (1986) supports the downward-sloping demand curve (DS) hypothesis but is unable to rule out other possible explanations. Both Jain (1987) and Dhillon and Johnson (1991) support the information hypothesis and reject the PP hypothesis.

Lynch and Mendenhall (1997) were able to carry out more refined tests because they examined a period when index changes were announced prior to implementation. They conclude that the data are consistent with the DS and PP hypotheses and acknowledge that information and liquidity effects may contribute to the observed price changes, but argue that these effects can be ruled out as complete explanations. Beneish and Whaley (1996) rule out the liquidity hypothesis but appear to support the DS and PP hypotheses.

3. Sharemarket Indices and the Market Impact of Index Portfolio Changes

Sharemarket indices can differ in terms of features such as format (open or closed end), the criteria used to determine eligibility for inclusion, the frequency with which the index portfolio is reviewed and the time period, if any, between notification and implementation of changes in the portfolio. As noted earlier, there are differences between the AOI and the S&P 500, which means that the market impact of portfolio changes may differ between the two indices.

3.1 The All Ordinaries Index (17)

The main ASX index is the All Ordinaries Index (AOI) which is based on a sample or portfolio of companies designed to approximate the behaviour of the `investible universe' of listed domestic companies. The investible universe is defined by the ASX as `all companies above a minimum (small) size with consistent and adequate daily turnover'. (18) The AOI is an open-ended index where the index portfolio contains all companies that meet specified criteria related to size (market capitalisation) and liquidity. (19)

The number of companies that meet the criteria can vary considerably over time and in the period January 1990 to July 1998 the number of companies in the index portfolio ranged from 229 to 330 with an average of 285. (20) Table 1 shows the number of companies added to and removed from the index in the January 1992--July 1998 period which covers the period of our study.

3.2 The Market Impact of Index Portfolio Changes

Several institutional differences suggest that the market impact of changes to the AOI index portfolio may differ from those found for the S&P 500. First, the AOI is an open-ended index whose portfolio is reviewed at defined intervals, using publicly disclosed criteria that are sufficient conditions for determination of changes. In contrast, the S&P 500 as a closed end index is reviewed infrequently, and decisions on changes are guided by a set of general criteria. Importantly, the AOI portfolio is determined by a transparent process that ensures that changes are predictable in advance, whereas changes to the S&P 500 are not known until they are announced. These differences should permit astute investors and possibly `risk arbitrageurs' to enter the Australian market at a much earlier date than is possible in the US. Consequently, it may be that the price effects, if any, of changes in the composition of the AOI are different from, and spread over a longer time period than, those that accompany changes in the S&P 500.

A second difference is that index funds are relatively smaller and have a shorter history in the Australian market than in the US. For example, the Vanguard Index Trust--500 Portfolio fund had an asset value of $14 million in 197621 and `at the end of 1997, the value of the US market indexed to the S&P500 was about $600 billion.' (22) In Australia, a survey of indexed funds managers in June 1999 found that assets under indexed management totalled $60.3 billion, or 10.6% of total assets. The proportion of Australian equities that were indexed increased from 12% in June 1997 to 15% in June 1999. (23) While published statistics focus on the recent rapid growth in indexing it is not a new phenomenon in the Australian market. For example, retail index funds have been available at least since 1988 and Australia's biggest investment manager, AMP Asset Management started running indexed investments in 1993. (24) Also, the limited statistics that are available on index funds in Australia are likely to understate the demand for the shares of AOI companies because many active funds typically have part of their portfolio structured to approximate the index. (25)

Third, the two indices differ in coverage. The S&P500 index covers approximately 70% of the market capitalisation of the US equity market while the AOI provides about 90% market capitalisation coverage of the Australian market. (26) Therefore, apart from large newly-listed companies that may be added to the AOI within six weeks of listing, companies entering and leaving the index portfolio are relatively smaller than those entering and leaving the S&P500. Consequently, changes in demand could give rise to larger price changes than those observed in the US. On the other hand, the shares of small and/or illiquid companies entering and leaving the AOI may not be held by index funds except for those that follow a policy of strict index replication. (27) If this is so, inclusion in (exclusion from) the AOI may stimulate a smaller abnormal price change or trading activity or none at all. Consistent with this suggestion, the ASX, after reviewing the evidence available at the time, concluded, `It is clear that the market response to portfolio changes for the AOI compared to the S&P500 is muted.' (28) However, the belief that inclusion in (exclusion from) the AOI has significant price effects is widespread and strongly held, particularly by managers and directors of companies that no longer qualify for inclusion. For example, the chairman of Hills Industries told shareholders that removal from the AOI had lowered the company's share price by at least 10%. (29) Companies facing exclusion from the AOI are also reported to have suffered `an average underperformance of around 30% for the four months leading up to exclusion.' (30) Under these circumstances, the performance, of companies added to and deleted from the AOI is worthy of serious examination.

4. Data and Methodology

4.1 The Sample

All companies added to and deleted from the All Ordinaries Index from January 1992 to July 1998 were identified using the ASX publication Monthly Index Analysis (MIA) and the actual date of each addition (deletion) was recorded from this publication. The MIA, which contains the first official notification of forthcoming changes in the composition of the various ASX indices `comes out around the 10th of the month providing over 2 weeks notice of forthcoming changes'. (31)

In this study, the analysis was performed at the daily and monthly levels. For the daily analysis, share prices adjusted for capitalisation changes together with details of dividends were obtained from IRESS for the period January 1994 to 30 September 1998. (32) These data were used to calculate daily share accumulation returns. The All Ordinaries Accumulation Index, the Small Ordinaries Accumulation Index (33) and Industry Accumulation Indices used as benchmarks were collected from IRESS. Daily trading volume for each company in the sample was also collected from IRESS.

Companies are included in the final sample only if they have a matching industry index on IRESS and a price history for one year prior to their addition (deletion) date. The latter requirement precludes any confounding problem associated with initial public offerings that were also added to the index. Companies deleted from the index were also excluded if the deletion was due to merger, takeover or liquidation. Thus our deletions sample consists only of companies deleted from the index due to insufficient size or lack of liquidity. Furthermore, for the daily analysis, companies were excluded from our sample if, during the period from day -10 to day + 5 relative to the change date, there were announcements of profit, takeovers or liquidation.

For the monthly analysis, price relatives for each company and the CRIF value-weighted market index were extracted from the AGSM CRIF Share Price Relative database up to 31 December 1997. (34) From 1 January 1998 to 31 December 1998, share prices adjusted for capitalisation changes together with details of dividends were obtained from the IRESS database. These data were used to calculate monthly share accumulation returns for calendar year 1998. The market index used for the period prior to 1998 was the CRIF value-weighted market index. For 1998, we used the All Ordinaries Accumulation Index taken from IRESS while the Small Ordinaries and Industry Accumulation Indexes were taken from the IRESS database for the entire sample period. The exclusion filters used for the daily analysis were also used at the monthly level.

4.2 Methodology

4.2.1 Analysis of Returns All returns calculated in this study are continuously compounded. The raw returns may be influenced by general market movements, the company's size or its industry sector. These factors are controlled by adjusting the raw returns using various benchmarks. Our main approach is analogous to the zero-one version of the familiar market model, so that

(1) [R'.sub.i] = [R.sub.i] - [R.sub.B]

where: [R.sub.i] = the return on company i; [R.sub.B] = the return on the benchmark; and [R'.sub.i] = the adjusted return. (35)

The main aim of the daily analysis is to ascertain as precisely as possible the timing and magnitude of any price effects that may be attributable to trading by investors who are influenced by changes in the index portfolio. A priori, we expect trading by index funds that seek to replicate the AOI to occur close to the change date because the performance of such funds is assessed by measures of tracking error. Other investors, including those who seek to approximate the AOI may buy (sell) the shares of companies to be added to (deleted from) the index once they can be confident that the change will occur. Thus a price response might occur when a forthcoming change is published in the MIA or when a stock is added to the Watch List, included in the MIA since July 1997, which foreshadows possible future changes in the composition of indices. However, publication dates for the MIA cannot be identified precisely and, in many cases, information about AOI changes almost certainly reached the market at an earlier time. As noted earlier, astute investors should be able to confidently identify the companies that satisfy the criteria outlined in section 3. Therefore, in contrast to the S&P 500 there is no single announcement date for each change to the AOI; rather there is a long event window and important price effects may occur outside the period normally examined in event studies.

We use a standard event study methodology with our window of analysis extending from day -60 to + day 60 relative to the addition (deletion) date as the change date is the only date that we know with certainty. We divided our sample into three groups. They are additions to the index, deletions from the index due to the failure to meet the minimum market capitalisation requirements (`size deletions'), and deletions due to the failure to meet the liquidity requirements (`liquidity deletions'). Size and liquidity deletions are examined separately because the returns associated with them are likely to differ. In the case of size deletions, negative abnormal returns prior to deletion may be the cause of the change in the index rather than caused by the change. This is not the case for liquidity deletions, which might therefore, provide a cleaner setting for the isolation of any index effect. (36) Our final samples comprise 31 additions, 49 size deletions and 17 liquidity deletions where the change date occurs between January 1995 and July 1998. (37) Abnormal returns from the zero-one model were calculated for each day and each company over the 121-day event window that we examined and averaged on a daily basis for each of the addition, size deletion and liquidity deletion groups. (38) Returns on day t are aggregated and the average abnormal return calculated for each of the days t, where t ranges from -60 to +60 relative to the change date. The analysis was repeated using the Small Ordinaries Accumulation Index to test whether any effect was related to company size. Finally, it was repeated with industry accumulation indices to control for industry specific effects.

While analysis of daily returns should reveal any abnormal returns close to the change date, analysis of monthly returns allows us to examine a larger sample that is not affected by any survivorship bias. (39) Monthly returns were analysed for a sample of 93 companies added to the index between January 1992 and December 1997. Our window of analysis is from month -12 to month +12 relative to the change month. Initially, we used the same benchmarks employed in the daily analysis to control for general market movements, size and industry effects. However, it is unlikely that the use of these, or any other similar, benchmarks will enable us to determine the magnitude of any `index effect'--that is, price effects caused by a forthcoming change in the composition of the index. The problem is that the process used to select companies for addition to the index introduces a selection bias that is likely to dominate any index effect. The index additions that we study are small companies that have performed well relative to the rest of the market, such that their market capitalisation increased to exceed the index threshold over the 6-month period prior to their inclusion. In effect at month -1, the ASX has looked back at those companies that at month -7 were too small for inclusion, but then for a period of six successive months, have a market capitalisation greater than the threshold for inclusion. Companies that meet the above criteria are added to the index. Clearly, these companies are, by construction, out-performers over the 6 months prior to inclusion, and are expected to exhibit returns that are `abnormal' relative to any market index.

Essentially, companies added to the index have three features in common: small market capitalisation; above-average realised returns over a 6-month period and addition to the AOI portfolio. To isolate any `index' effect we form a control portfolio that has all of these features except addition to the AOI. (40) Specifically, we replicate the procedure used by the ASX to select companies for inclusion in the index by tracking their market capitalisation over a 6-month `qualifying period', except that we use a lower inclusion threshold of 0.01% of total domestic market capitalisation. In addition, the market capitalisation of these companies cannot exceed 0.022% of total domestic market capitalisation in the 6-month qualifying period. (41) Companies that meet these criteria are included in our control portfolio. We use the returns from this control portfolio to calculate adjusted returns as defined in equation 1.

4.2.2 Trading Activity Tests of trading activity were performed using a similar methodology to Beneish and Whaley (1996). In accordance with their approach we use daily trading volume as a measure of trading activity. (42) Because changes in index composition depend on company size and liquidity for a period of up to 6 months prior to the change, our pre-test period is from -12 months to -7 months (days -260 to -131) relative to the change date. For each company in our sample, mean daily trading volume was calculated for the 6-month pre-test period. These averages were used as benchmarks for the subsequent test period. (43) An abnormal trading volume ratio was calculated for each company by dividing the daily trading volume in the test period by mean volume in the pre-test period. For the addition and deletion portfolios, the ratio is averaged across companies for each day in the test period. If daily trading volume on any day in the test period is greater than normal, the ratio will be greater than one.

5. Results

5.1 Analysis of Daily Returns

The main objective of the daily analysis is to identify, as precisely as possible, the timing and magnitude of any effects associated with trading by investors who adjust their portfolios in response to changes in the AOI. While we present results for the period from day -60 to day +60, our primary focus is on the change date and the few days that surround it, particularly the period from day -10 to day +10. (44) As stated earlier, we believe that because of concerns about tracking error, `pure' index funds will only buy (sell) close to the change date. Also, we seek to document whether the buying (selling) pressure of `pure' index funds and its effects on share prices are reversed in subsequent days.

Table 2 shows the adjusted daily returns for the additions portfolio. Using various benchmark indices to control for general market movements, size and industry effects, all the results show that the adjusted return on day -1 is positive and statistically significant at the 1% level. Moreover, the adjusted return on day -1 is typically more than twice that on any of the other 120 days in the test period. The return on day 0 is negative but not statistically significant. There also appear to be isolated cases, mostly after the addition date, when there are both negative and positive adjusted returns that are statistically significant at the 5% level. The table shows that increased demand for shares is associated with average returns, after adjusting for market movements, of + 2.6% (+4.0%) in the day (week) before the addition date. This is consistent with the findings of US studies.

Table 3 shows the adjusted daily returns for the size deletions portfolio. The table shows that there are negative returns which are statistically significant at the 1% level for two of the three benchmarks on day -5 and for all benchmarks on days -2, and -1. Further, irrespective of the benchmark used, all the adjusted returns on day 0 are positive and statistically significant at the 1% level. This may be indicative of a price reversal effect when selling pressure by index funds has stopped. The average market-adjusted price change associated with removal from the index appears to be in the order of-3.3% (-10.7%) in the day (week) before the deletion date. Again, this is broadly consistent with US evidence, particularly that of Lynch and Mendenhall (1997).

Table 4 shows the adjusted daily returns for the portfolio of companies deleted due to inadequate liquidity. The results are not as strong as those shown in table 3. In the immediate period before removal from the All Ordinaries Index, only day -1 has a negative return that is statistically significant at the 1% level when the analysis controls for general market movements and industry effects. The average adjusted returns on day +1 are negative in all three cases and statistically significant when controlling for size. The effect of investor selling, after controlling for industry effects, is of the order of-2% (-3.2%) in the day (week) before the deletion date. Controlling for general market movements, the adjusted returns are -2.4% (-4.7%) in the day (week) before the deletion date.

Figures 1 to 3 show daily CARs for the addition, size deletion and liquidity deletion portfolios. For additions, figure 1 reveals two patterns of returns. First, there is an increase in returns from -30 days (-6 weeks) to the addition date where the returns peak. Then, all these returns are reversed over the next six weeks. Second, and within this broader pattern, is a substantial price jump at day -1 that is partially reversed on day 0. These patterns are robust to the benchmark used and could indicate some support for the price pressure hypothesis.

[FIGURES 1-3 OMITTED]

Figure 2 shows the daily CAR for the size deletion portfolio. The return pattern is negative sloping from day -60. However, it appears that there is a more rapid decline from about day -5 that bottoms at day -1. This decline is partially reversed at day 0.

Figure 3 shows the daily CAR for the liquidity deletion portfolio. For all three benchmarks, the patterns are similar until about 25 days before the deletion date. For the general market and size benchmarks, there is a continued decline in the returns from days -25 to + 60. From days -5 to -1, there is a much more rapid decline that is partially reversed on day 0. When the industry benchmark is used, the returns appear to be random, except for a decline from days -5 to +1, followed by an apparent reversal over the next 15 to 20 days.

5.2 Analysis of Monthly Returns

Table 5 shows the adjusted monthly returns for the portfolio of companies that were added to the index.

Table 5 shows the returns adjusted for general market movements, size and industry effects over the period from -12 to + 12 months relative to the addition month. For all three benchmarks, months -11, -3 and -1 have positive adjusted returns that are statistically different from zero at the 5% level. (45) For additions, month -4 is statistically significant when the analysis controls for industry effects. An F-test showed that the standard deviation in the test period was higher than, and statistically different from, the standard deviation in the pre-test period at the 5% level when the general market movements and size benchmarks were used. (46) As we use the standard deviation in the test period for our t-tests, we should not overstate the number of observations that are statistically significant. (47)

As discussed in section 4.2, selection of companies for addition to the AOI is biased toward selection of companies that have outperformed the market in the months leading up to their inclusion. Not surprisingly, table 5 shows that for all benchmarks, the adjusted returns are positive in every month prior to the change month. The results for the additions portfolio adjusted for the selection bias are reported in table 6. They suggest that there is a sizeable `index' effect after removing the out-performance inherent in the selection process. The first statistically significant positive adjusted return occurs at month -4 and apart from month -6 all adjusted returns are positive. The CAR for months -4 to month 0 is 25.1% table 7 shows the summary statistics for our control portfolio.

5.3 Trading Activity

We now re-examine the daily data with our focus on trading volume. The samples are the same as those in the returns analysis except that the size and liquidity deletions are combined in one portfolio. Table 8a shows the abnormal trading volume ratio for companies added to the index from days -10 to +10 relative to the addition date. (48)

The mean volume ratios are statistically significant at least at the 5% level for days -10, -8, -5, -1, +3, +4 and +7. On day -1, average trading volume is 5.294 times the average daily volume in the pre-test period. This result is highly significant with a t-statistic of 3.35. On day 0 the average trading volume is 3.345 times the average in the pre-test period, but it is not statistically significant at the 5% level. (49) The mean volume ratio and the t-statistics may be affected by outliers. Therefore, we also report the median and employ the non-parametric Mann-Whitney test. This test compares the distribution of volume ratios for each day of the test period with the distribution during the pre-test period from days -260 to -131. The median volume ratio is greatest on day -1 and the Mann-Whitney test probabilities indicate that the distributions differ significantly at the 5% level on all days in the test period except for days -3, -2 and +8. Taken together, the results suggest that trading volume is significantly elevated around the time of inclusion and is greatest on the last day before inclusion in the index.

In addition to the increased trading volume on day -1 and several of the surrounding days, there may be other longer term changes in trading activity associated with index changes. For example, greater investor `interest' in index stocks might be reflected in higher volume after a company's shares are added to the AOI. To determine whether there are any long term patterns we average the daily abnormal trading volume ratio over successive 20-day intervals that cover the whole period from day -260 to day +60 relative to the addition date. The 20-day averages are shown in table 8b which also shows the number of days in each interval for which the results were statistically significant at the 5% level. We present these results for the whole period for two reasons. First, the stability or otherwise of the 20-day averages over the first part of the period provides an indication of the adequacy of the pre-test period as a benchmark. Second, there may be abnormal trading prior to the day -60 to day +60 test period used in the returns analysis.

The results in table 8b show that apart from indications of an increase in trading activity between days -160 and -141, trading volume is reasonably stable over the pre-test period. There are unmistakable signs that increased trading activity begins well before addition to the index and persists for at least 60 days after addition. During the whole period from day -100 to day +60 average trading volume is at least twice the normal level during the pre-test period. The pattern observed in table 8 may reflect purchases by active investors who are able to identify, with reasonable accuracy, companies likely to be added to the index. These investors may include `risk arbitrageurs' that buy, once a company can be identified as a likely addition to the AOI, with the intention of selling at a profit close to the change date. Consistent with this suggestion, the time at which volume begins to increase corresponds with the time at which the positive adjusted returns reported in table 6 become significant.

Table 9a shows trading volume for the deletions portfolio from days -10 to + 10 relative to the deletion date. The mean trading volume ratio is statistically significant on all days from -4 to -1. Mean volume peaks at 3.611 times normal with a t-statistic of 5.03 on day -1 but is not significant on the deletion day. Consistent with the results for additions, the median volume ratio is greatest on day -1. In this case, the Mann-Whitney test indicates that the distributions differ significantly at the 5% level on 10 of the 21 days in the test period. These results provide strong evidence of increased trading volume around the deletion date, which may be indicative of selling by index funds.

Table 9b shows the average trading volume over 20-day intervals from days -260 to +60 relative to the deletion date. As for additions, there is evidence of greater trading volume around the deletion date. However, the effect is less pronounced and, in contrast to the results for additions, any increase in trading activity is minor until the period from days -20 to -1, but the increase that occurs appears to persist over the 60 days following the deletion date

6. Conclusions

This paper provides evidence that the effects of changes to an open-ended market index are broadly consistent with the empirical evidence based on closed-end indices. Daily analysis of the AOI reveals significant positive (negative) abnormal returns immediately prior to addition (deletion). The abnormal performance peaks (bottoms) on the day prior to the change and there is evidence of price reversals on the day of the change. Our results based on daily returns are consistent with significant trading by index funds keen to minimise tracking error. The trading volume results may also support the view that index funds are active close to the change date.

In addition, changes to the index are associated with extended periods of elevated trading activity which may be due to the fact that, in contrast to the S&P 500 index, changes to the AOI can be predicted in advance. For companies added to the AOI, this activity starts well before addition and may be consistent with active investors buying on the basis of a momentum strategy or in anticipation of a company's addition to the index. In comparison, trading activity for the deletions portfolio begins to increase at a much later date. For additions and to a lesser extent deletions, there is evidence that increased trading activity persists for at least 60 days after the change.

The process used to select companies for addition to (deletion from) the AOI introduces a selection bias which ensures that index additions (deletions) will perform well (poorly), relative to usual benchmarks, in the months leading up to the change. After controlling for this bias, we find evidence of significant positive returns over the 4-month period prior to inclusion in the AOI.

(Date of receipt of final transcript: December, 2001. Accepted by Stephen Gray, Area Editor.)

Appendix A

The All Ordinaries Index (50)

Inclusion in the AOI

Any listed, domestic company is eligible for inclusion in the AOI if its market capitalisation is at least 0.022% of total domestic market capitalisation and it meets a liquidity requirement. The liquidity requirements were revised with effect from 1 July 1998. Prior to I July 1998, a company needed to have an average normal turnover on the ASX in excess of 0.5% of its quoted shares per month over the previous six months. After 1 July 1998, the liquidity criterion for inclusion was that the company's monthly median liquidity, relative to the ASX median market liquidity must be at least 50% over the previous six months.

Exclusion from the AOI

The index portfolio is reviewed monthly and, apart from merger, takeover or liquidation, a company may be removed for reasons related to changes in its size and liquidity. A company is removed if its market capitalisation falls below 0.015% (0.01%) of total market capitalisation for a period of six (three) months. A company is also removed if the monthly turnover in its shares falls below a specified level for an extended period. Prior to 1 July 1998, this minimum level was 0.2% of its quoted shares per month. Larger index companies with turnover approaching this lower level could be down-weighted to reflect the low availability of their shares.

After 1 July 1998, stocks that failed to meet the liquidity target for inclusion in the AOI would be down-weighted or removed from the AOI on a quarterly basis as follows:

(a) companies with relative liquidity less than 33% of the market would be subject to a down-weighting factor of 75%;

(b) companies with relative liquidity less than 25% of the market would be down-weighted to 50%;

(c) companies with relative liquidity less than 17% would be down-weighted to 25%; and

(d) companies with relative liquidity less than 12.5% would be removed from the AOI.

Each company's weight in the AOI depends on its index capitalisation which is equal to its market capitalisation multiplied by any down-weighting factor. Consequently, from 1 July 1998 onwards, a company can be removed from the index for any of three reasons. First, failure to meet the minimum market capitalisation requirements. Second, if the liquidity of the company relative to market liquidity falls below 12.5% for six consecutive months. Third, if its index capitalisation after down-weighting falls below the minimum percentages of total market capitalisation as shown above.

Appendix B

General Guidelines for Adding to and Removing Stocks from the S&P 50051

Adding Stocks to the S&P Indices

1. Market Value: S&P indices are market-value weighted.

2. Industry Group Classification: Companies selected for the S&P indices represent a broad range of industry segments within the US economy.

3. Capitalisation: Ownership of a company's outstanding common shares is carefully analysed to screen out closely held companies.

4. Trading Activity: The trading volume of a company's stock is analysed on a daily, monthly and annual basis to ensure ample liquidity and efficient share pricing.

5. Fundamental analysis: Both the financial and operating condition of a company are rigorously analysed. The goal is to add companies to the indices that are relatively stable and will keep turnover low.

6. Emerging Industries: Companies in emerging industries and/or new industry groups-industry groups currently not represented in the indices-are candidates as long as they meet the guidelines listed above.

Removing Stocks from the S&P Indices

1. Merger, Acquisition, LBO: A company is removed from the indices as close as possible to the actual transaction date.

2. Bankruptcy: A company is removed from the indices immediately after chapter 11 filing or as soon as an alternative recapitalisation plan that changes the company's debt/equity mix is approved by shareholders.

3. Restructuring: Each company's restructuring plan is analysed in depth. The restructured company as well as any spin-offs are reviewed for index inclusion or exclusion.

4. Lack of representation: A company can be removed from the indices because it no longer meets current criteria for inclusion and/or is no longer representative of its industry group.

Table 1
Summary Statistics for Additions to and Deletions from the All
Ordinaries Index (#)

Calendar Year              1992   1993   1994   1995

Additions                    39     38     58     23

Deletions
  Size                        7      4      6      6
  Liquidity                   5     11      3      5
  Others (*)                  4      3     12      9
No. of Companies on AOI:    260    280    317    320
31 December

Calendar Year              1996   1997   July   Total
                                         1998

Additions                    46     18     17    239

Deletions
  Size                       11     28     33     95
  Liquidity                   4      2     10     36
  Others *                   23     18      8     77
No. of Companies on AOI:    328    298      -      -
31 December

Note:

* Others refer to mergers/takeovers and liquidations; and

(#) Source Monthly Index Analysis from 1992 to July 1998 and
ASX Fact Book 1999.
Table 2
Daily Adjusted Returns for the Period from Day -60 to Day +60
Relative to the Addition Date (#a)

        General Market
Day         Control            Size Control        Industry Control

-43    0.011 *     (2.10)    0.009       (1.73)    0.009       (1.87)
  :    :                     :                     :
  :    :                     :                     :
  :    :                     :                     :
-10   -0.000      (-0.05)    0.003       (0.53)    0.001       (0.16)
 -9    0.002       (0.31)    0.003       (0.59)    0.001       (0.11)
 -8   -0.002      (-0.41)   -0.001      (-0.17)   -0.003      (-0.67)
 -7   -0.003      (-0.66)   -0.007      (-1.45)   -0.004      (-0.77)
 -6    0.003       (0.61)    0.005       (0.99)    0.001       (0.26)
 -5    0.007       (1.38)    0.008       (1.63)    0.007       (1.42)
 -4    0.000       (0.05)    0.003       (0.67)    0.001       (0.29)
 -3    0.005       (0.90)    0.003       (0.62)    0.006       (1.21)
 -2    0.002       (0.30)    0.000      (-0.04)    0.003       (0.54)
 -1    0.026 **    (5.17)    0.022 **    (4.45)    0.025 **    (5.09)
  0   -0.008      (-1.50)   -0.006      (-1.25)   -0.006      (-1.17)
 +1    0.000       (0.09)   -0.003      (-0.62)    0.001       (0.10)
 +2    0.003       (0.57)    0.001       (0.30)    0.005       (1.10)
 +3    0.003       (0.51)    0.007       (1.39)    0.001       (0.17)
 +4    0.003       (0.62)    0.002       (0.50)    0.002       (0.50)
 +5    0.002       (0.35)    0.002       (0.47)    0.001       (0.24)
 +6   -0.003      (-0.56)   -0.002      (-0.42)   -0.001      (-0.19)
 +7    0.004       (0.81)    0.004       (0.91)    0.005       (1.16)
 +8   -0.002      (-0.42)   -0.003      (-0.64)    0.001       (0.23)
 +9   -0.007      (-1.48)   -0.006      (-1.28)   -0.007      (-1.34)
+10    0.003       (0.52)    0.000       (0.02)    0.002       (0.42)
  :    :                     :                     :
+30   -0.010      (-2.03)   -0.011 *    (-2.21)   -0.010      (-1.90)
  :    :                     :                     :
+32   -0.012 *    (-2.30)   -0.100      (-2.01)   -0.010 *    (-2.07)
  :    :                     :                     :
+34    0.011 *     (2.10)    0.011 *     (2.31)    0.011 *     (2.15)
  :    :                     :                     :
       :                                           :
  :    :                     :                     :
       :                                           :
  :    :                     :                     :
       :                     :
  :    :                     :                     :
  :    :                     :                     :
  :    :                     :                     :
+60   -0.013 **   (-2.59)   -0.010      (-1.95)   -0.011 *    (-2.30)

Note:

* Statistically significant at the 5% level for a two-tailed test;

** Statistically significant at the 1% level for a two-tailed test;

(#) Standard deviations used for the t-test are from the testing
period; and

(a) t-statistics are in brackets.
Table 3
Daily Adjusted Returns for the Period from Day -60 to Day +60
Relative to the Size Deletion Date (#a)

        General Market
Day         Control            Size Control        Industry Control

-49   -0.016      (-1.85)   -0.015      (-1.83)   -0.018 *    (-2.20)
  :    :                     :                     :
  :    :                     :                     :
       :                     :                     :
  :    :                     :                     :
  :    :                     :                     :
-21   -0.017 *    (-2.01)   -0.017 *    (-2.14)   -0.015      (-1.89)
  :    :                     :                     :
  :    :                     :                     :
-10   -0.016      (-1.88)   -0.011      (-1.33)   -0.013      (-1.65)
 -9   -0.011      (-1.30)   -0.012      (-1.43)   -0.012      (-1.53)
 -8   -0.002      (-0.28)   -0.003      (-0.43)    0.001       (0.18)
 -7   -0.012      (-1.35)   -0.009      (-1.10)   -0.012      (-1.49)
 -6    0.009       (1.03)    0.008       (1.04)    0.007       (0.85)
 -5   -0.027 **   (-3.17)   -0.024 **   (-3.01)   -0.017 *    (-2.09)
 -4   -0.017      (-1.97)   -0.013      (-1.64)   -0.015      (-1.92)
 -3    0.006       (0.67)    0.007       (0.86)   -0.000      (-0.03)
 -2   -0.036 **   (-4.25)   -0.033 **   (-4.15)   -0.031 **   (-3.89)
 -1   -0.033 **   (-3.84)   -0.033 **   (-4.12)   -0.033 **   (-4.19)
  0    0.026 **    (3.00)    0.028 **    (3.42)    0.025 **    (3.07)
 +1    0.004       (0.46)    0.005       (0.67)    0.005       (0.64)
 +2   -0.011      (-1.34)   -0.004      (-0.53)   -0.004      (-0.44)
 +3   -0.002      (-0.19)    0.003      (-0.35)   -0.000      (-0.05)
 +4    0.001       (0.17)    0.001       (0.18)   -0.001      (-0.14)
 +5    0.000      (-0.02)    0.000      (-0.05)   -0.002      (-0.22)
 +6   -0.005      (-0.54)   -0.004      (-0.46)   -0.000      (-0.06)
 +7    0.002       (0.24)   -0.001      (-0.16)    0.004       (0.45)
 +8    0.004       (0.43)    0.002       (0.22)   -0.002      (-0.19)
 +9   -0.006      (-0.67)   -0.005      (-0.62)   -0.007      (-0.84)
+10   -0.013      (-1.55)   -0.010      (-1.22)   -0.006      (-0.80)
  :    :                     :                     :
  :    :                     :                     :
  :    :                     :                     :
  :    :                     :                     :
  :    :                     :                     :
       :                     :                     :
  :    :                     :                     :
  :    :                     :                     :

Note:

* Statistically significant at the 5% level for a two-tailed test;

** Statistically significant at the 1% level for a two-tailed test;

(#) Standard deviations used for the t-test are from the testing
period; and

(a) t-statistics are in brackets.
Table 4
Daily Adjusted Returns for the Period from Day -60 to Day +60
Relative to the Liquidity Deletion Date (#a)

        General Market
Day         Control            Size Control        Industry Control

-48    0.018 *     (2.66)    0.014 *     (2.50)    0.017 *     (2.56)
  :    :                     :                     :
-46    0.014       (2.02)    0.012 *     (2.12)    0.013       (2.07)

  :    :                     :                     :
  :    :                     :                     :
-10   -0.013      (-2.00)   -0.009      (-1.58)   -0.010      (-1.58)
 -9   -0.004      (-0.55)   -0.003      (-0.49)   -0.001      (-0.08)
 -8    0.003       (0.44)    0.002       (0.35)    0.006       (0.85)
 -7    0.000       (0.04)    0.007       (1.15)    0.001       (0.19)
 -6   -0.013      (-2.00)   -0.015 *    (-2.62)   -0.014 *    (-2.12)
 -5   -0.003      (-0.44)   -0.002      (-0.42)   -0.001      (-0.10)
 -4    0.000       (0.04)    0.000      (-0.02)    0.003       (0.49)
 -3   -0.012      (-1.81)   -0.008      (-1.38)   -0.007      (-1.09)
 -2   -0.008      (-1.23)   -0.005      (-0.78)   -0.007      (-1.04)
 -1   -0.024 **   (-3.61)   -0.019 **   (-3.21)   -0.020 **   (-3.09)
  0    0.003       (0.44)    0.003       (0.61)   -0.003      (-0.45)
 +1   -0.014      (-2.09)   -0.015 *    (-2.57)   -0.013      (-2.03)
 +2    0.002       (0.26)    0.001       (0.25)    0.002       (0.31)
 +3    0.004       (0.56)    0.006       (1.03)    0.008       (1.21)
 +4   -0.003      (-0.45)   -0.001      (-0.24)   -0.002      (-0.44)
 +5    0.005       (0.79)    0.006       (1.05)    0.005       (0.77)
 +6   -0.010      (-1.42)   -0.007      (-1.22)   -0.007      (-1.12)
 +7    0.010       (1.48)    0.008       (1.38)    0.009       (1.34)
 +8    0.002       (0.32)    0.005       (0.81)    0.004       (0.60)
 +9   -0.007      (-1.05)   -0.006      (-1.06)   -0.006      (-0.99)
+10   -0.001      (-0.18)    0.002       (0.31)    0.002       (0.32)
  :    :                     :                     :

  :    :                     :                     :

+48   -0.015 *    (-2.21)   -0.011      (-1.91)   -0.012      (-1.86)
  :    :                     :                     :

+60   -0.012      (-1.82)   -0.012 *    (-2.11)   -0.013      (-1.99)

Note:

* Statistically significant at the 5% level for a two-tailed test;

** Statistically significant at the 1% level for a two-tailed test;

(#) Standard deviations used for the t-test are from the testing
period; and

(a) t-statistics are in brackets.
Table 5
Monthly Adjusted Returns for the Period from Month -12 to
Month +12 Relative to the Addition Month (#a)

          General Market
Month        Control            Size Control       Industry Control

-12      0.034      (0.80)    0.034      (0.83)    0.031      (0.82)
-11      0.089 *    (2.07)    0.086 *    (2.08)    0.075 *    (1.96)
-10      0.063      (1.46)    0.057      (1.38)    0.048      (1.25)
 -9      0.058      (1.35)    0.054      (1.31)    0.048      (1.26)
 -8      0.044      (1.02)    0.041      (0.98)    0.040      (1.04)
 -7      0.050      (1.18)    0.048      (1.17)    0.052      (1.36)
 -6      0.059      (1.37)    0.056      (1.36)    0.053      (1.39)
 -5      0.058      (1.36)    0.055      (1.34)    0.072      (1.89)
 -4      0.076      (1.78)    0.074      (1.78)    0.082 *    (2.15)
 -3      0.090 *    (2.10)    0.084 *    (2.04)    0.078 *    (2.03)
 -2      0.080      (1.86)    0.073      (1.76)    0.075      (1.95)
 -1      0.095 *    (2.21)    0.090 *    (2.17)    0.084 *    (2.20)
  0      0.044      (1.02)    0.037      (0.89)    0.041      (1.08)
  1     -0.003     (-0.06)   -0.006     (-0.14)   -0.009     (-0.23)
  2      0.018      (0.42)    0.020      (0.48)    0.021      (0.54)
  3     -0.039     (-0.91)   -0.039     (-0.95)   -0.032     (-0.85)
  4     -0.015     (-0.34)   -0.017     (-0.42)   -0.009     (-0.23)
  5     -0.009     (-0.21)   -0.012     (-0.28)   -0.010     (-0.26)
  6      0.001      (0.01)   -0.003     (-0.08)    0.003      (0.08)
  7     -0.002     (-0.05)   -0.004     (-0.11)   -0.004     (-0.09)
  8      0.004      (0.09)    0.004      (0.10)    0.019      (0.50)
  9     -0.022     (-0.50)   -0.021     (-0.50)   -0.008     (-0.20)
 10     -0.037     (-0.85)   -0.036     (-0.87)   -0.020     (-0.53)
 11     -0.033     (-0.77)   -0.034     (-0.81)   -0.034     (-0.89)
 12     -0.001     (-0.02)   -0.001     (-0.02)   -0.002     (-0.06)

Note:

* Statistically significant at the 5% level for a two-tailed test;

(#) Standard deviations used for the t-test are from the testing
period; and

(a) t-statistics are in brackets.
Table 6
Monthly Returns Adjusted Against an Out-Performance Control
Group From -6 Months Relative to the Addition Month (#a)

        Outperformance Control

 Month

  -6     -0.108 ***     (-4.54)
  -5      0.005          (0.22)
  -4      0.047 *        (1.97)
  -3      0.064 **       (2.71)
  -2      0.046 *        (1.93)
  -1      0.084 **       (3.55)
   0      0.010          (0.43)

Note:

* Statistically significant at 10% level for two-tailed test;

** Statistically significant at 5% level for two-tailed test;

*** Statistically significant at 1% level for two-tailed test;

(#) standard deviations are calculated in the pre-test period from
month -12 to -24 relative to addition month; and

(a) t-statistics are in brackets.
Table 7
Summary Statistics for Control Portfolio

Number of Companies in Control Portfolio          132
Number of Companies in the Portfolio Per Month:
  Mean                                            3
  Median                                          2
  Minimum                                         1
  Maximum                                         7

                  Monthly Returns

 Month       Average     Maximum    Minimum

  -6          0.201       0.933     -0.345
  -5          0.067       0.597     -0.243
  -4          0.057       0.405     -0.275
  -3          0.017       0.523     -0.223
  -2          0.027       0.593     -0.288
  -1          0.013       0.353     -0.321
   0          0.010       0.326     -0.279

        Market Capitalisation at Month 0 ($ million)
Average           Maximum             Minimum

 39.2               80.7               18.9
Table 8a
Abnormal Trading Volume Ratio for the Period from Days -10 to
            +10 Relative to the Addition Date (#a)

                          Abnormal Trading
                            Volume Ratio

   Day     Mean 4.76 **  t-Statistic for  Median        Mann
              (2.93)          Mean                 Whitney Test
              1.941                                Probability (b)
              (0.73)

  -10        4.761 **         2.93         0.789       <0.001
   -9        1.941            0.73         0.908        0.002
   -8        4.329 **         2.60         0.674        0.010
   -7        2.278            1.00         0.796       <0.001
   -6        2.805            1.41         1.522       <0.001
   -5        3.848 *          2.22         0.521        0.002
   -4        2.367            1.07         0.692        0.002
   -3        1.965            0.75         0.478        0.169
   -2        0.811           -0.15         0.320        0.465
   -1        5.294 **         3.35         1.903       <0.001
    0        3.345            1.83         0.868       <0.001
   +1        2.264            0.99         0.844        0.001
   +2        1.850            0.66         1.024        0.026
   +3        6.119 **         3.99         0.986        0.001
   +4        4.472 **         2.71         1.102       <0.001
   +5        2.503            1.17         0.735        0.015
   +6        2.556            1.21         0.411        0.023
   +7        3.709 *          2.11         0.976        0.003
   +8        0.958           -0.03         0.408        0.109
   +9        1.835            0.65         0.910       <0.001
   +10       2.008            0.79         0.926       <0.001

Note:

* Statistically significant at 5% level for a two tailed test;

** Statistically significant at 1% level for a two tailed test;

(#) The Abnormal Trading Volume Ratio is calculated by dividing the
daily volume for each stock by the average daily trading volume for
that stock in the pre-test period from day -260 to day -131;

(a) Means and t-statistics are for trimmed data set. Medians and
Mann-Whitney test are for the untrimmed data; and

(b) The Mann-Whitney Test Probability is an estimate of the probability
that the distribution on the test day is the same as the distribution
in the pre-test period.
Table 8b
Average Trading Volume Over 20-day Intervals for Additions

      Interval (a)

   From          Until   Mean Volume Ratio (b)  Number Significant (c)

 -260         -241               0.84                      0
 -240         -221               0.93                      1
 -220         -201               0.81                      1
 -200         -181               1.01                      2
 -180         -161               0.89                      0
 -160         -141               1.44                      3
 -140         -121               1.23                      1
 -120         -101               1.79                      1
 -100          -81               2.15                      3
  -80          -61               2.62                      2
  -60          -41               2.56                      3
  -40          -21               3.14                      7
  -20           -1               2.51                      4
   +1          +20               2.85                      6
  +21          +40               2.24                      3
  +41          +60               2.20                      2

Note: (a) Intervals are measured in trading days relative to the
addition date;

(b) Arithmetic mean of the individual daily observations. The ratio
is the stock's daily trading volume divided by the average trading
volume during the period from 260 to 131 days prior to addition; and

(c) Number of days for which Abnormal Trading Volume ratio is
statistically significant at the 5% level.
Table 9a
Abnormal Trading Volume Ratio for the Period from Days -10 to
+10 Relative to the Deletion Date (#a)

                       Abnormal Trading Volume Ratio

Day    Mean 4.76 **  t-Statistic for   Median   Mann Whitney
          (2.93)           Mean                 Test Probability (b)
           1.941
          (0.73)

 -10      1.254            0.49         0.407          0.343
  -9      1.317            0.61         0.484          0.193
  -8      1.411            0.79         0.543          0.289
  -7      0.609            0.75         0.382          0.480
  -6      1.589            1.14         0.499          0.011
  -5      1.692            1.33         0.429          0.387
  -4      2.857 **         3.58         0.746         <0.001
  -3      2.884 **         3.63         0.496          0.051
  -2      2.394 **         2.69         0.884          0.002
  -1      3.611 **         5.03         2.241         <0.001
   0      1.593            1.14         0.729          0.001
  +1      1.725            1.40         0.684          0.007
  +2      1.787            1.52         0.881          0.006
  +3      1.500            0.96         0.581          0.006
  +4      1.006            0.01         0.582          0.079
  +5      0.851           -0.29         0.257          0.481
  +6      1.756            1.46         0.376          0.636
  +7      1.734            1.41         0.191          0.042
  +8      1.604            1.16         0.394          0.942
  +9      1.293            0.56         0.381          0.951
 +10      1.481            0.93         0.625          0.016

Note: * Statistically significant at 5% level for a two tailed test;

** Statistically significant at 1% level for a two tailed test;

(#) The Abnormal Trading Volume Ratio is calculated by dividing the
daily volume for each stock by the average daily trading volume for
that stock in the pre-test period from day -260 to day -131;

(a) Means and t-statistics are for trimmed data set. Medians and
Mann-Whitney test are for the untrimmed data; and

(b) The Mann-Whitney Test Probability is an estimate of the probability
that the distribution on the test day is the same as the distribution
in the pre-test period.
Table 9b
Average Trading Volume over 20-Day Intervals for Deletions

 Interval (a)

 From    Until   Mean Volume Ratio (b)   Number Significant (c)

 -260    -241           1.17                    1
 -240    -221           1.05                    3
 -220    -201           1.08                    3
 -200    -181           0.99                    0
 -180    -161           0.81                    0
 -160    -141           0.92                    0
 -140    -121           0.94                    1
 -120    -101           0.84                    0
 -100     -81           0.96                    0
  -80     -61           1.25                    2
  -60     -41           1.04                    0
  -40     -21           1.17                    0
  -20      -1           1.63                    4
   +1     +20           1.57                    3
  +21     +40           1.38                    1
  +41     +60           1.41                    2

Note: (a) Intervals are measured in trading days relative to the
deletion date;

(b) Arithmetic mean of the individual daily observations. The ratio
is the stock's daily trading volume divided by the average trading
volume during the period from 260 to 131 days prior to deletion; and

(c) Number of days for which Abnormal Trading Volume ratio is
statistically significant at the 5% level.

(1.) After completion of this study, the AOI was restructured and since 3 April 2000 it has been a closed-end index comprising 500 companies. The criteria for inclusion are based only on company size (market capitalisation) and there is no liquidity requirement. In this paper, all references to the AOI relate to the conditions that applied at the time of the study.

(2.) If a company meets the capitalisation and liquidity criteria for six consecutive months, it is automatically added to the AOI. However, decisions regarding inclusion in (exclusion from) the S&P 500 index are guided by a set of six (four) general criteria. Refer to section 3.1 and appendix A for details of the AOI prior to 3 April 2000 and to appendix B for the guidelines on inclusion in (exclusion from) the S&P 500 index.

(3.) Beneish and Whaley (1996) argue that because many index funds wait until the day of the change in the index to rebalance their portfolios, `risk arbitrageurs' can buy ahead of the index funds in the expectation of selling at a higher price a few days later. Technically, this does not constitute arbitrage and in this case `risk arbitrage' refers to speculation that impending purchases by index funds will result in price increases. A momentum effect is another possible explanation for abnormal returns observed in the months prior to an addition to the index (e.g. Chan, Jegadeesh & Lakonishok 1996 and Hong & Stein 1999). We thank the editor Stephen Gray for this comment.

(4.) See, for example, Burke and Anderson, `Major investors threaten to set up their own index', Australian Financial Review, 11 April 1997, Mullane; `ASX index plans infuriate investors', Australian Financial Review, 28 October 1998; Mellish, `Class action mooted', The Age, 7 December 1998 and Mullane, `ASX caught in frank exchange of views', Australian Financial Review, 9 December 1998.

(5.) In addition to the studies discussed below, related evidence is provided by Woolridge and Ghosh (1986), Pruitt and Wei (1989), Edmister, Graham and Pirie (1994) and Collins, Wansley and Robinson (1995).

(6.) One stock was excluded from the analysis because it was also subject to a takeover bid.

(7.) The initial price increase appeared to be cumulatively reversed over the succeeding period. For example, the mean excess return of 3.13% on day 1 was followed by a mean cumulative excess return of minus 2.49 % from day 2 to day 21.

(8.) Shleifer (1986, fn. 5) reported that combining the pre- and post-event estimation periods made little difference to the results and that the results were not materially different when returns were not corrected for market movements.

(10.) Shleifer (1986) notes that prior to September 1976 it was not possible to determine the dates on which S&P500 portfolio changes were made. Jain (1987) states that precise identification of announcement dates became possible only from November 1977 and that prior to that time announcements of changes were made only periodically (every three to six months).

(10.) Market model parameters were estimated using pre-event data. Dhillon and Johnson (1991) reported that `in general our results are robust to changes in methodology. The results are virtually identical whether the value-weighted or the equally weighted index is used or whether or not log price relatives are used' (fn. 1, p. 77).

(11.) Dhillon and Johnson (1991) reproduced Harris and Gurel's (1986) results for additions from 1978 to 1983 but extended the time period to 121 days centred on the announcement. Apart from the significant return on the announcement day there was a steady decline in the Cumulative Abnormal Returns throughout the 121-day period which Dhillon and Johnson (1991)suggested could be due to a faulty risk adjustment.

(12.) In contrast to most of the prior studies, LM used the market-adjusted or `zero-one' model but noted that results obtained using the market model were very similar. LM also note that since index additions (deletions) are likely to have performed well (poorly) just prior to the change, use of return data from this period to estimate market model parameters may produce biased alpha estimates. Accordingly, the market model was estimated using returns from 872 to 673 days prior to the announcement day.

(13.) All results quoted here are for `clean' samples consisting only of firms not affected by merger or spin-off activity around the time of the index change.

(14.) The author noted that as the principal criterion for inclusion in these indexes is the stock's market capitalisation, the changes he examines may be partly anticipated. Therefore, abnormal returns measured over the 11-day period may underestimate the impact on prices.

(15.) This suggestion is advanced by Jain (1987).

(16.) This explanation of the link between liquidity and valuation is provided by Shleifer (1986).

(17.) This section relies heavily on `Review of the All Ordinaries Index', ASX Consultation Paper, January 1999.

(18.) Op cit, p. 7.

(19.) See appendix A for details of the criteria for inclusion in and exclusion from the AOI. While most market indices are based on a fixed number of stocks, the AOI is rare, but not unique, in being an open-ended index at the time of this study. A similar approach is used in constructing the MSCI series. See ASX Consultation paper, p. 31.

(20.) ASX Consultation Paper, p. 13.

(21.) See Beneish and Whaley(1996), p. 1911.

(22.) Mace, `Indexing sets portfolio pace', Superfunds No. 220, November 1998b, p. 78.

(23.) `State of play in the indexed funds management market', Rainmaker Roundup, June Quarter 1999, p. 6.

(24.) The MLC MasterKey Share Index Fund formerly managed by the Australia Bank opened on February 22, 1988. The reference to AMP is by S. Hely, `Active-passive debate goes on', Superfunds, no. 227, July 1999, p. 3.

(25.) According to statistics attributed to the ASX manager of index services Mr. John Elfverson, most actively managed funds in Australia `had a tracking error of 150 to 250 basis points, indicating that many managers paid fairly close attention to the index.' See J. Mace, `Indexing--label error', Superfunds no. 217, August 1998a, p. 34. A similar point is made by Brealey (2000) who notes that the common practice of measuring fund performance against that of a market index may make fund managers reluctant to invest in non-index stocks. See R.A. Brealey, `Stock prices, stock indexes and index funds', Bank of England Quarterly Bulletin, February 2000, pp. 61-8.

(26.) ASX Consultation paper, pp. 30-1.

(27.) Discussions with fund managers indicate that some managers of Australian index funds seek to approximate the AOI by holding large companies in exact index weights, combined with a stratified sampling approach for small companies in the AOI.

(28.) ASX Consultation paper, p. 43.

(29.) R. Hill-Ling, Chairman's Address, Hills Industries Annual General Meeting, 1998.

(30.) M. Mullane, `ASX gives ground on All Ords', Australian Financial Review, December 12-13, 1998, p. 37. In other reports in the financial press this finding has been attributed to research by Macquarie Equities.

(31.) ASX Consultation paper, p. 44. We also examined date-stamped copies of the MIA held in a stockbroker's library. The results of this examination support the statement that it provides about two weeks' notice of changes in the AOI portfolio.

(32.) IRESS is a real time database provided by Bridge/DFS Pty Ltd. It retains share price data only for companies that are currently listed on the ASX and this may introduce a survivorship bias in the daily analysis. However, a comparison at the monthly level for the same period of the sample of companies drawn from IRESS and the AGSM CRIF price relative database which does not have a survivorship bias, gives similar results.

(33.) The Small Ordinaries Index sample consists of the companies in the AOI less the companies in the ASX 100 index. Thus the companies in our samples are at the lower end of the size range covered by the Small Ordinaries Index.

(34.) The AGSM CRIF share price relative database is a historical database with monthly data. It retains the price history of companies that are delisted, liquidated, merged or taken over by other companies.

(35.) The zero-one model is equivalent to the market model where alpha is zero and beta is equal to one. This analysis was also performed using weekly data. We repeated the weekly analysis using the market model adjusted for thin trading using the Dimson method. A comparison of these results against those using the zero-one model found that the pattern and direction of the abnormal returns were similar but the size of these abnormal returns varied according to the risk adjustment used. The weekly results are not reported in this paper but are available upon request from the authors.

(36.) While liquidity deletions have the potential to provide such a setting the situation is complicated by two factors. First, our sample exhibits event date clustering in that many of the liquidity deletions occurred in July 1998. Second, it is difficult to determine which companies will be subject to deletion because, to remain in the AOI it is necessary to exceed the required minimum level in only one month out of six. For example, after the new liquidity requirements to apply from 1 July 1998 were announced, more than 20 companies were identified as likely to be deleted on that date due to low liquidity. However, several of these companies experienced increased liquidity such that less than half of them were actually removed from the AOI in July 1998.

(37.) Nine of the seventeen liquidity deletions occurred on 1 July 1998.

(38.) We use the zero-one model because the US evidence shows that the results are robust to the methodology used. Lynch and Mendenhall (1997) report results from the zero-one model but note that very similar results were obtained using the market model Beneish and Whaley (1996) also adjust only for market returns. Further, as noted in footnote 34, when the market model and the zero-one model were used to examine weekly returns associated with index additions and deletions, both models yielded similar results.

(39.) As noted earlier, IRESS retains data only for companies currently listed on the ASX but this limitation does not apply to the AGSM CRIF monthly price relative database.

(40.) We are grateful to Rob Brown for suggesting this approach.

(41.) Therefore, our control group excludes any companies that were added to the AOI during or shortly after the 6-month qualifying period. We also examined deletions using monthly data and used an analogous approach to select a portfolio that would serve as a control for size deletions. The number of companies that met the criteria for inclusion in the control group was too small to yield meaningful results. Monthly returns for deletions are not reported because the determination of size deletions involves a selection bias that will induce underperformance relative to index benchmarks.

(42.) Like Beneish and Whaley(1996) we also used average daily trade size as a measure of trading activity. The results, which are not reported, were weaker than those for trading volume but revealed some evidence of an increase in average trade size prior to changes in the AOI portfolio.

(43.) Trading volume was adjusted for capitalisation changes such as rights and bonus issues. This information was taken from IRESS and cross-checked against the ASX Annual Stockmarket Summaries and annual reports.

(44.) As stated at footnote 29, the ASX provides about 2 weeks' notice of forthcoming changes to the AOI. Therefore the examination of returns from day -10 to the inclusion date takes place in a setting that is essentially identical to the analysis by Beneish and Whaley (1996). Any abnormal returns observed prior to day -10 may overstate the returns that could be earned by investors who buy shares in the expectation that particular stocks will be added to the index. While investors who are aware of the criteria used by the ASX should be able to identify likely additions to the AOI with a high degree of accuracy, some uncertainty remains until an announcement has been made by the ASX. We thank the editor, Stephen Gray for this comment.

(45.) The insignificance of month 0 may appear to be inconsistent with the daily analysis. However, it must be remembered that most changes occur at the beginning of each month. Prices increase before the addition date and if there is a subsequent reversal, the overall price change for the month may be insignificant.

(46.) The test period consists of the months -12 to +12 relative to the change month. The pre-testing period consists of the 12 months immediately preceding the testing period.

(47.) Binomial tests were also performed for all benchmarks. For each month of the test period, the hypothesis that the frequency of positive abnormal returns in the test period is equal to their frequency in the pre-test period, was tested for statistical significance. The pattern and statistical significance are generally similar to those reported in table 5. These results are not reported but are available from the authors.

(48.) We perform part of this analysis on both trimmed and untrimmed data sets. In the trimmed data set, we have removed the most extreme individual observations representing 0.1% of the data. Trimming had little effect on the results except in removing some outliers beyond day -60. The reported means and t-statistics are based on the trimmed data set while the medians and Mann-Whitney test results are based on the untrimmed data.

(49.) In the calculation of the t-statistic we use the standard deviation of the volume ratio from -60 to +60 days relative to the change date. This standard deviation is larger than for the pre-test period and will be biased against finding a statistically significant ratio.

(50.) This appendix is based on the criteria that applied at the time of the study. As noted earlier, the AOI has been a closed-end index since 3 April 2000.

(51.) This is taken directly from the S&P web page at http://www.spglobal.com/lg6faqlnkfaq.html

References

ASX consultation paper, January 1999, `Review of the All Ordinaries Index'.

ASX working document, April 1999, `Rules for the construction of the All Ordinaries Index'.

ASX `Monthly Index Analysis' 1992 to 1998, various issues.

ASX (1995-98), `Annual Stockmarket Summaries for Taxation'.

Beneish, M.D. & Whaley, R.E. 1996, `An anatomy of the `S&P Game': The effects of changing the rules', Journal of Finance, vol. 51, December, pp. 1909-30.

Brealey, R.A. 2000, `Stock prices, stock indexes and index funds', Bank of England Quarterly Bulletin, February, pp. 61-8.

Burke, F. & Anderson, S. 1997, `Major investors threaten to set up their own index', Australian Financial Review, April 11, p. 35.

Chan, L.K.C., Jegadeesh, J. & Lakonishok, J. 1996, `Momentum strategies', Journal of Finance, vol. 51, pp. 1681-713.

Collins, M.C. Wansley, J.W. & Robinson, B. 1995, `Price and volume effects associated with the creation of Standard & Poor's Midcap Index', Journal of Financial Research, vol. 18, Fall, pp. 329-50.

Dhillon, U. & Johnson, H. 1991, `Changes in the Standard and Poor's 500 list', Journal of Business, vol. 64, pp. 75-85.

Edmister, R., Graham, A.S. & Pirie, W. 1994, `Excess returns of index replacement stocks: Evidence of liquidity and substitutability', Journal of Financial Research, vol. 17, Fall, pp. 333-46.

Goetzmann, W. & Garry, M. 1986, `Does delisting from the S&P 500 affect stock price?', Financial Analysts Journal, vol. 42, pp. 64-9.

Harris, L. & Gurel, E. 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, vol. 41, pp. 815-29.

Hely, S. 1999, `Active-passive debate goes on', Superfunds, no. 227, July, p. 33.

Hill-Ling, R. 1998, Chairman's Address, Hills Industries Annual General Meeting

Hong, H. & Stein, J.C. 1999, `A unified theory of underreaction, momentum trading and overreaction in asset markets, Journal of Finance, vol. 54, pp. 2143-83.

Jain, P.C. 1987, `The effect on stock price of inclusion in and exclusion from the S&P 500', Financial Analysts Journal, vol. 43, pp. 58-65.

Lynch, A.W. & Mendenhall, R.R. 1997, `New evidence on stock price effects associated with changes in the S&P 500 index', Journal of Business, vol. 70, pp. 351-83.

Mace, J. 1998a `Indexing--label error?', Superfunds, no. 217, August, p. 34.

Mace, J. 1998b, `Indexing sets portfolio pace', Superfunds, no. 220, November, p. 78.

Mellish, M. 1998, `Class action mooted', The Age, 7 December, p. B1.

Mullane, M. 1998, `ASX index plans infuriate investors', Australian Financial Review, 28 October, p. 23.

Mullane, M. 1998, `ASX caught in frank exchange of views', Australian Financial Review, 9 December, p. 1-14.

Mullane, M. `ASX gives ground on All Ords', Australian Financial Review, December 12-13, 1998, p. 37.

Pruit, S.W. & Wei, K.C.J. 1989, `Institutional ownership and changes in the S&P 500', Journal of Finance, vol. 44, pp. 509-13.

Rainmaker Information, 1999, `State of play in the indexed funds management market', Rainmaker Roundup, June Quarter, p. 6.

Shleifer, A. 1986, `Do demand curves for stocks slope down?', Journal of Finance, vol. 41, pp. 579-90.

Woolridge, J.R. & Ghosh, C. 1986, `Institutional trading and security prices: The case of changes in the composition of the S&P 500 index', Journal of Financial Research, vol. 9, pp. 13-24.

Howard W.H. Chan ([dagger]) Peter F. Howard ([dagger])

[dagger] Department of Accounting and Finance, Monash University, Clayton VIC 3168. Email: [e-mail]

The authors acknowledge the financial support provided by a Monash University Faculty of Business and Economics research grant, the helpful support of the IRESS data service and the research assistance of Alicia Low and Melissa Graham. The authors are grateful for comments from seminar participants at RMIT and Monash universities, conference participants at the 1999 Asia-Pacific Finance conference, Alan Farley, Barry Fairley, Douglas Lob and especially Rob Brown.