Spot Market Effects Surrounding Compositional Changes to the FTSE 100: Transitory or Permanent?
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Spot Market Effects Surrounding Compositional Changes to the FTSE 100: Transitory or Permanent? KWAKU K. OPONG AND PHILIP A. HAMILL* *The authors are respectively: Kwaku K. Opong, Reader in Accounting and Finance, Department of Accounting and Finance, University of Glasgow, 65-73 Southpark Avenue, Glasgow G12 8LE, UK. Philip A. Hamill, Lecturer in Finance, Queen’s University Belfast. Address for correspondence: Philip A. Hamill, Lecturer in Finance, School of Management & Economics, 22 University Square, Queen’s University, Belfast BT7 1NN, UK. Tel: UK-02890 273278: E-mail: p.a.hamill@qub.ac.uk
ABSTRACT Spot Market Effects Surrounding Compositional Changes to the FTSE-100: Transitory or Permanent? This paper examines the effect on share’s price and volumes traded when it is included in, or deleted from, the FTSE 100 from 1984 to 1999. An Event Study Methodology approach, in conjunction with the ZD test, was adopted. A priori, we developed a number of competing hypotheses from the extant literature. These included The Imperfect-Substitutes-Hypothesis, Price-Pressure-Hypothesis, Liquidity-Hypothesis, Price-Volatility-Hypothesis and explanations based on Agency-Costs. Our empirical analysis reports a significant price increase prior to additions to the FTSE 100, which is followed by a price reversal, which appears to persist over the long-term. The results for deletions are virtually symmetric. The only difference being that the magnitude of the post-event price decrease for additions is much greater than the magnitude of the increase experienced by deletions. The findings reported for additions are consistent with the predictions from the Price-Volatility-Hypothesis. At present there is no adequate model to explain the price and volume patterns experienced by firms deleted from the FTSE 100. Also, these findings appear to be an anomalous violation of semi-strong-form market efficiency. In conclusion, spot market effects surrounding compositional changes to the FTSE 100 appear to be permanent, at least over the period we study. Keywords: FTSE 100, Event study methodology, ZD test, additions, deletions, semi-strong- form market efficiency 2
1. INTRODUCTION This paper examines the impact on prices and the volumes traded for stocks which were either included in, or deleted from, the FTSE 100. The majority of the extant literature has focused on spot market effects for firms entering or exiting the S&P 500. This literature provides empirical findings which have been explained by a number of competing hypotheses. The initial focus of the US literature was to disentangle the effect of alternative listing rules. Initially, from 1976 to 1989, changes in the composition of the S&P 500 were announced after the close of trading. The change became effective the following morning’s open. In October 1989 S&P’s regulations changed. Under the new regime the announcement of a change to the index was made after the market close, however, the change did not become effective until five days later. Papers which examined this issue conclude that under the initial regime there was a temporary effect, whereas, in the latter period it was permanent (Shleifer, 1986; Dhillion and Johnson, 1991; Beneish and Whaley, 1996; Lynch and Mendenhall, 1997). The temporary effect was explained in relation to the Price-Pressure-Hypothesis whereas the Imperfect-Substitutes- Hypothesis accounted for a permanent effect. Recent papers, which extend the post-event analysis, propose the Price-Volatility-Hypothesis to explain a price reversal for longer horizons (Malkiel and Radisich, 2001; Cooper and Woglom, 2003). The extant literature allows us to develop a strategy, based on the average behaviour of cumulative abnormal returns, to help disentangle the competing hypotheses. We adopt an Event Study Methodology in conjunction with the robust ZD-test to analyze mean, and cumulative, abnormal returns and trading volumes. Our sample of firms consists of 162 additions to, and 160 deletions from, the FTSE 100 from 1984 to 1999 for abnormal returns calculations. For 3
trading volume we have 132 additions with 123 deletions. We examine the effects on prices and trading volume from twenty-one days before the composition of the index changes to sixty days after. A number of significant findings are reported for both additions to, and deletions from, the FTSE 100. Firms being included in the FTSE 100 experience a significant price increase during the twenty-one day period prior to inclusion. Theoretically, index-funds, wishing to minimize tracker error, would purchase the stock at the price it enters the index; which would be the day before entry. This finding suggests that risk-arbitrageurs enter the market to purchase the stock in anticipation of selling it to tracker funds at a profit. On the day the stock enters the FTSE 100 its stock prices experiences a reversal. This finding is at odds with all the US literature. In the period immediately following entry into the S&P 500 all US literature reports persistence. The price reversal experienced by UK firms persists, a finding which is consistent with recent US evidence. In terms of the hypotheses developed, a priori, the Price-Volatility-Hypothesis most closely resembles the empirical findings for UK data. Deletions from the FTSE 100 experience similar, but reverse, price and volume patterns. In the twenty-one day prior to exiting the FTSE 100 deletion experience a significant price decline which is followed by a price reversal. In the subsequent period the positive price effect persists and is significant. At present there is no adequate model to explain this reported result. In the period surrounding deletion from the FTSE 100 there is an obvious Price-Pressure effect. The evidence provided in this paper for additions and deletions appear to be an anomalous violation of semi-strong-form market efficiency. While we document this apparent anomaly, we do not test whether investors could have earned significant economic returns from a strategy based on this public information. We feel that the existing theoretical and empirical evidence for the UK is 4
inadequate and our understanding of these findings could benefit from additional contributions to the literature. The remainder of this paper is organised as follows. Section 2 discusses the relevant literature while section 3 develops testable hypothesis. Section 4 describes the dataset while section 5 provides a methodological overview. Sections 6 and 7 discuss the empirical analysis and summarise the main findings reported in the paper. 2. LITERATURE REVIEW The majority of literature examining the effect of stock inclusion (deletion) to (from) an index has centred on the Standard and Poor’s 500 Stock Index (S&P 500). The extant literature provides conflicting evidence for the S&P 500, which to a large extent, was accounted for by changes in the regulatory environment. Initially, from 1976 to 1989, changes in the composition of the S&P 500 were announced after the close of trading with the change becoming effective the following morning’s open. In October 1989 S&P’s regulations changed. Under the new regime the announcement of a change to the index is made after the market close, but, the change does not become effective until five days later. During the initial period the observed price effects appears to be temporary. Harris & Gurel (1986) report that prices increase by more than 3% immediately following an addition announcement but that the price increase is gradually, but completely, reversed nearly two weeks after addition. They attribute this price behaviour to the Price-Pressure-Hypothesis. Alternatively, Shleifer (1986), Dhillion and Johnson (1991), Beneish and Whaley (1996) and Lynch and Mendenhall (1997) provide evidence of a permanent effect; consistent with Imperfect-Substitutes-Hypothesis. Under this hypothesis firm’s long-run demand curves are downward sloping, as opposed to being perfectly 5
elastic, demand curves. The results reported by Beneish and Whaley and Mendenhall and Lynch for the Cumulative Abnormal Returns (CARs) from the announcement day to the effective change day are mathematically similar, but, the explanations they provide diverge considerably. Mendenhall and Lynch conclude that this significant CAR is an anomalous violation of semi-strong-form market efficiency, arguing that investors could have earned economically significant abnormal returns from a strategy based on publicly available information. This similar result reported by Beneish and Whaley is attributed to a market learning mechanism, which they predict should diminish over time, and, in the limit case a price effect will be limited to the announcement day as index funds rebalance earlier. Empirical evidence is provided to this effect. Malkiel and Radisich (2001) question the veracity of the documented permanent effect. Instead, they extend their analysis beyond that of previous studies, providing evidence of a price reversal for longer horizons. While a price reversal is reported, no explanation is given for it. Cooper and Woglom (2003) propose the Price-Volatility-Hypothesis, which they test empirically, to explain this observed price effect. They argue that trading effects should lead to a permanent effect on stock price volatility. The initial price increase prior to inclusion in the S&P 500 is a function of both short run excess demand, which, is associated with a permanent reduction in the supply of the stock as index funds incorporate the stock into their non-traded portfolio. As a consequence the stock experiences increased volatility post-addition from the shock to the Downward Sloping Demand Curve. Also, a significant reduction in the supply of the stock should lead to higher trading volume which could also affect volatility. In these circumstances the value of the stock would be expected to decline due to increased price volatility leading to higher post-addition risk premium, even though a firm’s future cash-flows 6
are unaffected, they are now discounted at a higher rate. Consistent with Malkiel and Radisich (2001), Cooper and Woglom (2003) also report a price reversal for longer horizons. The Liquidity-Hypothesis is also consistent with a permanent price and volume effect. Amihud and Mendelson (1986) argue that the required rate of return on a stock varies with expected trading costs. If the excess demand from index funds for an addition to the index results in decreased trading costs, the present value of all future trading costs will fall and the stock price will experience a permanent rise. The arguments surrounding the Price Pressure Hypothesis, Imperfect-Substitutes-Hypothesis, Price-Volatility-Hypothesis and the Liquidity-Hypothesis try to explain observed price and volume patterns in relation to trading effects. Alternatively, changed fundamentals could possibly explain a permanent price increase. Fundamental effects include the Information- Content-Hypothesis and the Agency-Costs Hypothesis. Under the Information-Content- Hypothesis inclusion signal ‘good news’ about the firm’s long-term future performance, and vice versa. Whereas the Agency-Hypothesis rests on the notion that increased investor scrutiny reduces Agency Costs and raises future profits. For the Information-Content-Hypothesis to hold, changes in the composition of the index must be motivated by a decision criteria which is based on non-public information. In the UK it is unlikely that this hypothesis could explain price and volume effects for changes in the composition of the FTSE 100 as the decision to include (exclude) a stock is mechanical in nature and in the public domain1. Also, Cooper and Woglom (2003) point out that while hypothesis based on changed fundamentals make a permanent prices increase consistent with the Efficient Markets Hypothesis, they don’t explain an initial price increase followed by a subsequent reversal. 7
McIlkenny, Opong and Watson (1996) provide an empirical analysis of price and volume effects for compositional changes to the FTSE 100 from March 1984 to December 1992. McIlkenny et al report significant negative price and volume effects for deleted firms in the three days prior to the Steering Committee Meeting (SCM) meeting. For additions they report significant positive price activity on the day of the SCM. In their subsequent analysis McIlkenny et al failed to offer any convincing arguments for, or against, any of the competing hypothesis. Therefore, the aim of this paper is to contribute to the literature by drawing firmer conclusions with respect to the economics associated with changes in the composition of the FTSE 100. 3. HYPOTHESES DEVELOPMENT Our previous discussion allows us to develop a strategy, based on the average behaviour of CARs, to help disentangle the competing hypotheses. These include: 1. Pre-Change Day CAR: running from twenty-one days before the change day (CD) to the day before the Change Day (CD-1). 2. Release CAR: runs from the CD to the Release Ending (RE) Day. 3. Post Release CAR: runs from RE+1 to CD+60. 4. Combined CAR: runs from running from twenty-one days before the CD to the Release Ending (RE) Day. 5. Permanent CAR: running from the CD-21 to CD+60. Table 1 provides a summary of the prediction for the CARs for each of the hypothesis around the CD and for five CAR windows. The predictions with respect to deletions are the exact opposite. 8
INSERT TABLE 1 HERE These hypothesized effects for the Price-Pressure-Hypothesis and Imperfect-Substitutes- Hypothesis assume that most funds rebalance their portfolios the day before the stock is included (deleted) in (from) the index. This price pattern would be anticipated if there is no speculative trading. A pure price pressure effect does not affect the fundamental determinants of firm value and should dissipate when index funds have satisfied their demand for the added (deleted) stock. Lynch and Mendenhall (1997) refer to the day when index funds have completed their trades as the release ending day. The theoretical model for determining this day is provided by Keim and Madhavan (1996). The release ending day has been reached when trading volume has returned to its normal post-change level2. For additions to the FTSE 100 the release ending day was estimated to be three days after the stock was added to the index while for deletions the release ending day was estimated to be four days after the stock was deleted from the index. Keim and Madhavan also predict that most index funds rebalance their portfolios on the day before stocks are added (deleted) to (from) the tracked index (Change Day, CD-1) in a bid to minimise tracking error3. Consequently, the price pressure implies that the largest temporary price effect occurs the day prior to the composition of the index changing (CD-1), followed by a price reversal on the change day. This analysis assumes no speculative trading. In practice, price effects will also depend on the extent to which speculators attempt to profit on index fund’s unavoidable need to purchase (sell) stock additions (deletions) to (from) the FTSE 100 and the degree to which index funds trade in the run-up to the change day. This possibility is captured in the pre-change day CAR. However, regardless of the extent of pre- change day trading, price release starts on the change day as index-fund trading dissipates. The 9
Imperfect-Substitutes-Hypothesis predicts a zero CAR over the Release CAR, if this holding period is negative, a price reversal, this would be indicative of the Price-Pressure-Hypothesis. If the price reversal is only partial, then this implies a permanent effect associated with inclusion (exclusion) to (from) the FTSE 100. As the firms enters the index, index fund buying leads to a permanent reduction in the supply of the company’s stock, as it is amalgamated into non-traded portfolios causing the market clearing price to increase. For deletions the exact opposite is the case. A price effect could occur in the period prior to the change day. Speculators, anticipating the up-coming price change, and possibly index funds, to trade in stocks prior to the change day. In table 1 we refer to this as the Price-Pressure-Hypothesis with interaction. A partial price reversal, or alternatively a permanent price effect, would be anticipated if long-term demand curves slope downward. In order to investigate potential permanent effects longer CAR holding periods need to be examined4. The predictions for the Liquidity Hypothesis assume that firms added to (deleted from) the index experience increase (decreased) liquidity. This is based on the assumption of an informationally efficient market. In terms of trading volume, if index-fund trading is concentrated on the day before the change day, then, for both additions and deletions, we should observe the largest mean abnormal volume over during the Pre-Change Day CAR on the day before change day. The Liquidity Hypothesis implies that the mean abnormal volume is positive (negative) after the change day for additions (deletions). The distinguishing feature of the Price-Volatility-Hypothesis is the prediction of a price reversal over the Post-Release CAR, whereas the competing hypotheses predict a zero CAR. The prediction of a zero Release CAR implies that any pre-event price hike persists for a period. 10
4. DATA DESCRIPTION The Steering Committee of the FTSE 100 meets every quarter to review those firms that constitute the FTSE 100 Index. Prior to June 1992, the meetings took place on the third Wednesday of the quarter. Thus there were meetings on the third Wednesday in March, June, September and December. Therefore, a firm is either deleted or added on the first working day of the following month. Beginning June 1992, the meetings took place on the second Wednesday of the quarter and a firm is deleted or added after seven working days excluding the day of the meeting. For a firm already in the FTSE 100 to be deleted, it must fall below the 110th in terms of market value of all the U.K firms listed on the London Stock Exchange. Conversely, for a firm to be included in the FTSE 100, it must rise above being the UK's 90th largest company. The Steering Committee selects those companies to be added or deleted based on their closing market values on the day preceding the meeting. There are times when firms are deleted from the FTSE 100 due to events like mergers and takeover. Such firms were deleted from the study. The process of including firms into the and deleting firms from the FTSE 100 by Steering Committee started in 1984. The sample therefore consists of all those firms that have been deleted from or included in the FTSE-100 from March 1984 to September 2000. The initial sample of firms coming in and out of the FTSE-100 since 1984 totalled 439. The final sample of firms chosen for the study had to satisfy the following criteria: of (1) The deletion or inclusion of a firm from (in) the FTSE-100 must be the result of a firm satisfying the market value criteria. (2) There must not be any interim or final earnings announcements during the test period. (3) Daily share price data must be available for both the estimation and test period. The rationale for criteria (1) and (2) is to exclude other variables that could have confounding effects on the study and criteria (3) is to enable the estimation of the parameters for the models 11
used in the study and to standardise data across firms. The final sample comprised 162 additions to the FTSE 100 and 160 deletions. The addition (deletions) dates were obtained from the Primark/Datastream. Share price, dividend and trading volume data were also obtained from Primark/Datastream. 5. METHODOLOGY (i) Shareholder Wealth Effects – ZD test Following the recommendations of Hamill, Opong & McGregor (2002) we conduct an information content analysis using the ZD test. It is capable of accounting for multiple mis- specification of the market, providing robust variance estimates when calculating the significance of Cumulative Abnormal Returns (CARs). Daily logarithmic returns were calculated using: y i ,t = ln [( Pi ,t + Di ,t ) ( Pi ,t −1)] x i ,t = ln( Pm ,t ) − ln( Pm,t −1) where: P i ,t is the price of security i on day t, D i,t is the dividends paid during period t, P i ,t −1 is the price of security i at the end of period t-1, and P m, t is the price of the market index on day t. Figure 1 outlines the notation used for the estimation and test period. We have the single- 12
index market model written in vector form as5: INSERT FIGURE 1 ABOUT HERE y it = xit β i + u it (1) In this study, T=150 trading days, while e= 161 (event day) and the last day in the event period (T+m) is T=60. where xit = (1, xit ) and β′ i = (α i , β ) or, in matrix form as i y i = Xi β i + ui (2) where y′ = (y i i1 ,..........., y iT ), u′ = (u i i1 ,......, u ), iT and 1,.......,1 X′ = (x′ i i1 ,......, x′ ) = x iT i1 x ,....., iT The estimated model is used to forecast m future observations yi *′ =( y i ,T +1,..., y i ,T + m ) using the matrix of future observations; *′ *′ 1,......,1 X *′ = x ,......, x = i i , T +1 i,T + m i , T + 1,......, ix, T + m x and the OLS estimator β̂ i = (X′i Xi ) −1 X′i yi . The vector of prediction errors is then u * ′ = ( û i i ,T +1 ,......, û i ,T + m ), obtained from: 13
ui * = y i*− X i * β̂ i (3) where y* i is the return on the firm over the test period, X i * is a typical m × 2 OLS matrix of market returns over the test period and β̂ i is the vector of OLS estimated parameters. The cumulative sum of forecast errors over the event window (T+m1, T+m2) is: m2 T+ ∑ uˆ iτ = Cui* ( 4) τ =T + m 1 where 'C' is an appropriately designed 1 × n selection vector which has the elements taking the value unity if uˆ τ i is contained in the event window and zero if it is not. The covariance matrix is given by: Di = (X′i Xi T ) Qi (X′i Xi T ) −1 −1 (5) where Qi is an estimate of E (X′i ui u′i Xi T ) which can be approximated by: ∑ (x′it xi ,t − s + xi ,t − s xit )uˆ it uˆ i ,t − s (6) T p T ˆi =T Q −1 ∑ x ′it xit uˆ it2 + T −1 ∑ t =1 s =1 t = s +1 1/ 3 for p chosen to be approximately T ; p should be increased until the truncations become trivial. ˆ is thus an estimate of the average of the variances of x′ u Q it it plus a i term that takes into account the covariances between x′ uit it and x′ u i ,t − s i ,t − s , the 14
number of covariances being truncated at s=p through the assumption of mixing (see, Mills, 1993, Chapter 5). In these circumstances we have: ( ) E Cui * ui *′ C′ = σ i CC′ + CXi * Di Xi *′ C′ / T = V D 2 i (7 ) where: u ′i ui σ i2 = (T − 2) The Cumulative Abnormal Return (CAR) over the event window is: N CARm1,m2 = N ∑Cui (8) −1 * i=1 and using the expression in equations 7 and 8 we have: Z D = N ×CAR m1,m 2×V D ~ N (0,1) (9) −1 / 2 where: V D =∑i=1 V D i N The statistic ZD can be used to test the hypothesis: H0 : CARm1,m2 = 0 versus H1: CARm1,m2 ≠ 0 (ii) Trading Volume Effects 15
The number of firms used in the volume analysis were 123 for firms coming out of FTSE-100 and 132 for firms joining the FTSE 100. The methodology adopted for the analysis of trading volume in the period when firms either come into the FTSE 100 or out of it assumes that the volume of shares traded on a particular day equals to the average of shares traded in the estimation period. The expected volume of shares traded in a particular day is given by: − V i ,t = Vτ (10) where − Vi,,t is the percentage of firm i's shares traded in day t and V τ is the average of firm i’s shares traded based on 129 daily observations in the estimation period from day -150 to -22. Abnormal trading volume is therefore given by − ε i ,t = V i ,t = Vτ (11) Abnormal volume is estimated over test period days -21 to +150. Significance tests are conducted using the following equation: εt tε i , t = (12) Sε τ where Seϑ = [var (ARϑ)]1/2 with var estimated over the 129 days, -150 to -22. In addition, equation (8) is used to compute cumulative excess trading volume for the test period. 16
6. RESULTS The results for mean and cumulative abnormal returns and trading volume, for stocks added to the FTSE 100, are reported in table 2. Mean abnormal returns for five days prior to inclusion and five days after inclusion are reported both for returns and trading volume. In the pre-event period, from day -5 to -1, the mean abnormal returns in four out of the five days running up to the event day are positive. The day immediately before the composition of the FTSE 100 changes experiences the highest mean abnormal, approximately 1.5 per cent with an associated t-statistic of 9.13. Figure 2 depicts this fact graphically. It is obvious the plot of mean abnormal returns that day -1 experienced the greatest price hike in the pre-event period, and all subsequent periods. INSERT TABLE 2 AND FIGURE 2 ABOUT HERE On the day the stock becomes a constituent of the FTSE 100 the mean abnormal return is negative and significant at the 5 per cent level with a t-statistic of -2.5. This is also illustrated graphically in figure 2 which highlights the sharp decline on day 0. This price reversal persists over the next five days, all of which experience negative mean abnormal returns, with days 1 to 4 experiencing a significant decline at the 5 per cent level, or less6. In relation to trading volume, day -1 experiences the highest mean abnormal trading volume of 1.50, approximately, while the event day experiences the second highest level of trading volume at 0.19. Both are significant at the 1 per cent level, or less. INSERT FIGURE 3 ABOUT HERE 17
Figure 3 highlights changes in mean abnormal trading volume for additions. The CARs. in panel B of table 2 provide additional insights. The Pre-change-day CAR is positive at approximately 3.3 per cent and significant at the 1 per cent level. Trading volume for the Pre-change day CAR was 1.97 with an associated t-statistic of 3.86. Over the Release CAR holding period we observe a decline of approximately 2.24 per cent which is significant. This reverses a large proportion of the 3.3 percent increase reported for the Pre-Change-Day CAR. This cancelling out effect over the -21 to 3 day period is evidenced by the Combined CAR which indicates a net positive effect over this period which is insignificant. In order to extend our analysis further we a examine the Post-Release CAR and the Permanent CAR. We observe a significant decline of approximately 4.6 per cent. For the Permanent CAR we observe a significant decline in the region of 18 per cent. The result for the Combined trading volume CAR indicates and the Permanent trading volume CAR suggest that there was significant increase in trading volume in period spanning the pre-event period and the Release period. The Permanent CAR is insignificant. INSERT FIGURE 4 & 5 ABOUT HERE Figure 4 provides a graphical illustration of the preceding discussion. It is obvious from the CAR plot that initially share prices experience an initial price hike which is quickly reversed after the stock is included in the FTSE 100. What is strikingly evident from this graph is the persistent price decline of these stocks. Even though we don’t report statistics for CARs beyond sixty days a casual analysis of this graph suggests that this price pattern doesn’t dissipate. Overall the price and volume results reported in table 2 would suggest that in the run-up to the inclusion of a stock in the FTSE 100 the price and volume activity is concentrated on the day 18
immediately before inclusion, but, that there is also significant price and volume activity over the pre-event period. From the day the stock is included in the FTSE 100 the price experiences a reversal, which appears to be persistent. In terms of our prior hypotheses this result doesn’t fit perfectly with any one hypothesis. The patterns which emerged from day -21 to 3 suggest a temporary price pressure effect, which with intervention of speculators, inflates prices. The apparent persistent price decline in the post-release period is consistent with the predictions from the Price-Volatility-Hypothesis. The UK pattern contrasts with the US literature. The permanent effects reported by Beneish and Whaley (1996) and Lynch and Mendenhall (1997) were driven by price effects from the announcement day to the change day, with abrupt increase in prices beginning the day after announcement. The gradual price increase in the run-up to the day the composition of the FTSE 100 changes can be accounted for by information symmetry between the FTSE Steering committee and market participants. The decision criteria is mechanical and is not based on an assessment of the firms’s future performance by the Steering committee. This full information decision criteria appears to have the effect of leading to a price increase much earlier, in relation to the change day, for firms added to the FTSE 100. In this study we reported a significant pre-change day CAR for a twenty-one day holding period of 3.3 per cent whereas for US literature the ‘price pop’, as it is referred to, is confined to the five days between the announcement day and the change day. Also, where as in the US the positive price effect persists for a short horizon after the change day, in the UK the release period experiences a swift price reversal. In terms of the post-release period the persistent price decline is consistent with the predictions from the Price-Volatility-Hypothesis and the empirical findings of Malkiel and Radisich (2001) and Cooper and Woglom (2003). One striking feature of the data is the extent of the decline for the UK. It must be stressed that even though the price decline reported 19
here is consistent with the Price-Volatility Hypothesis, it may not explain precisely the UK experience. While some of the assumptions which underpin the model are reasonable for the UK we would argue that there are significant institutional differences which may need to be taken account of when developing predictions for the UK, for example the mechanical listing rule which applies in the UK. Figure 4 suggests an almost a mirror image for additions and deletion. The only apparent different between the two events is that the magnitude of the price decline for additions is greater than the magnitude of the price increase for deletions. INSERT TABLE 3 FIGURE 6, 7 & 8 ABOUT HERE The results reported in table 3 for deletions from the FTSE 100 exhibit a similar, but reverse pattern, to additions. Again we experience significant mean and cumulative abnormal returns and trading volume. The mean abnormal returns for days -2, -1, 0, 1 and 5 are significant at the 5 per cent level, or less. Mean abnormal trading volume is significant on days -5, -4, -2, and -1. It is obvious from figure 6 that the harp decline on CD-1 is reversed on CD. The highest mean abnormal trading volume occurred on day -1. It was 0.72, which is approximately half the highest mean abnormal trading volume for additions on the same day. Also, whereas the level of trading volume is highly significant for additions on day 0 it isn’t for deletions on the same day. Figure 7 highlights changes in mean abnormal trading volume. The pre-event period experiences two distinct hikes immediately prior to the day firms are deleted from the FTSE 100. Firms deleted from the FTSE 100 experience a significant price decline in the run up to being removed from the index. The pre-Change CAR if -0.0515 per cent and significant. The release CAR exhibits a price reversal at 0.0148 per cent, which again is significant. This 20
positive price reversal persists; the Post-Release CAR is 0.0324 per cent and significant. The Combined-CAR suggests that the net effect is negative whereas the Permanent-CAR suggests that overall there is a net positive effect which is statistically insignificant. Figure 8 provides a plot of cumulative abnormal trading volume for deletions. This suggests a permanent increase in trading volume. While the results from analyzing volume indicate a significant increase in trading volume over the Pre-change day CAR period, even though the Release CAR was insignificant, the Combined CAR was significant whereas the permanent CAR was insignificant. Trying to explain the price and volume patterns associated with firms being deleted from the FTSE 100 is more challenging than for additions. If we were to focus on the short-term it appears that pre-event price decline, followed by a price reversal on the day the stock exits the FTSE 100, is consistent with a temporary liquidity effect with speculators entering the market. What is then experienced is a significant price increase over the long-term. However we don’t have any prior theories which try to explain long-term effects. The model proposed by Cooper and Woglom (2003) was developed for additions to the S&P 500. If we were to assume symmetry, with respect to the Price-Volatility-Hypothesis, it may help explain the UK experience for deletions8. This is a tentative conjecture. An obvious issue regarding these findings is what appears to be an anomalous violation of semi-strong-form market efficiency. In order to determine if this is the case investors would need to be able to earn economically significant abnormal from a trading strategy base on this public information. In this paper we document this apparent anomaly, but don’t test it. This paper only begins to examine the effect upon stocks when they are either included to, or deleted from, the FTSE 100. In the US considerable efforts have been made, and are being 21
made, to better understand ‘S&P’ effects, as they are referred to in the literature. There is an obvious dearth of both theoretical and empirical literature for the UK environment. Both this paper and recent US contribution have extended the post event analysis to unearth previously ignored price patterns. In the context of Event Study Methodologies, it is unlikely that further significant contribution can be made due to associated econometric weakness with extending the post event period any further. Therefore, an obvious extension of this paper would be to conduct a long-run performance approach. Also, we document a significant liquidity effect. A market microstructure analysis has the potential to yield insights into the effect inclusion in (deletion from) the FTSE 100 has on both the direct and indirect costs of trading. Finally, we appear to have unearthed an anomalous violation of semi-strong-form market efficiency. Whether this is economically significant remains to be determined. 7. CONCLUSION This paper we examines the price and volume effects when stocks were included in the FTSE 100 or deleted from it. The bulk of the theoretical and empirical literature has focused on examining the S&P 500 index and trying to explain the underlying economic relationships. The literature provides a number of competing hypotheses. These include the Price-Pressure- Hypothesis, Imperfect-Substitutes-Hypothesis, Price-Volatility-Hypothesis, Liquidity- Hypothesis, Information-Hypothesis and arguments in relation to Agency-Costs. We examine a sample of stock additions and deletions to and from the FTSE 100 from 1984 to 1999. Our empirical analysis provides evidence of significant price and volume effects, both for stock which were added to the FTSE 100 and for stocks which were deleted from it. Our analysis suggests that over the short-run, from twenty-one days before addition (deletion) to 22
twenty-one days after, both additions and deletions experienced virtually the same price and volume patterns. In a world with no arbitrage index-funds would purchase the stock at the price it goes into the index just prior to the stock’s inclusion to minimise tracking error. The significant cumulative abnormal returns in the twenty-one day pre-event period suggest the intervention of risk-arbitrage traders in the market. In the release period additions to the FTSE 100 experienced a price reversal, and vice versa for deletions. For a similar holding period a number of contributions to the US literature report a permanent effect (Beneish and Whaley, 1996; Mendenhall and Lynch, 1997). When we examine the longer-term we report a significant price decline for additions and a significant price increase for deletions. This finding for additions is consistent with more recent contributions to the US literature (Malkiel and Radisich, 2001; Cooper and Woglom, 2003). The only asymmetric result was that the magnitude of the price decline for additions was much greater than the magnitude of the price increase for deletions. The only hypothesis which is consistent with the hypotheses we developed, was the Price-Volatility-Hypothesis. Returning to our original proposition, does inclusion in (deletion from) the FTSE 100 have a transitory of permanent effect? It would appear from our analysis that, at least over the period we studied, the effects are permanent. Probably the most important findings from this paper is the significant gaps in the UK literature. We feel that significant contribution could be made in relation to theoretical modelling, empirical analysis focusing on market microstructure and reconciling apparently anomalous violations of semi-strong-form market efficiency. 23
NOTES 1 In this study we centre our analysis on the change day, which is the day the stock is added (deleted) to (from) the FTSE-100. Typically, US studies examine two key days, the Announcement Day, which is equivalent to the steering committee meeting for the FTSE-100, and the change day. In the US the market becomes aware of stocks selected for inclusion (deletion) to (from) the S&P-500 after close of business on the announcement day. In the context of the FTSE-100 the inclusion (deletion) criteria is mechanical and known to market participants. Consequently, market participants trading behaviour around this time is not conditional upon the Steering Committee Meeting. 2 The volume is estimated to have returned to its normal post-change level on the earliest day after the change day with mean abnormal trading volume (MATV) that is lower than the average MATV for all later days through to +21 days. We also computed the release ending day on the basis of 10 days after the change day and which was the same. 3 Tracking error is the difference between the fund’s return and the return on the tracked index over a given period 4 Cooper and Woglom (2003) point out that the standard error of the CAR is proportional to the square root of the length of the event window. According to the efficient markets hypothesis the expected value of the CAR does not change with the length of the event window. Therefore, statistical tests about the implications of the CARs lose their power as the length of the event window increases. Consequently, it is not practical to infer results from CARs cumulated for periods greater than sixty days. 5 The single index market model was used returns generating model for both the conventional event study methodology and the ZD test. No adjustment for asynchronous trading is made as Hillier and Yadav (1997) demonstrate that this adjustment can at worst lead to a serious misspecification and bias in the market model prediction errors and at best reduce the efficiency of the methodology. Also, as the sample consists of the largest stock on the London Stock Exchange it would be expected that they would least be expected to be affected by thin trading. 6 The results for additions from Corrado’s (1989) non-parametric rank test are consistent with those reported for the market model. Days -2, -1, 0, 1, 3 and 4 were significant at the 5 per cent level or less. 7 The results for deletions from Corrado’s (1989) non-parametric rank test are consistent with those reported for the market model. Days -1, 0, 1 and 5 were significant at the 5 per cent level or less. 24
8 Cooper and Woglom (2003) don’t study deletions from the S&P 500 as they regard deletion as being largely irrelevant due to the fact that most companies are removed from the S&P 500 due to mergers, bankruptcies or other corporate transformations. 25
REFERENCES Amihud, Y. and H. Mendelson (1986),’Asset Pricing and the Bid-Ask Spread’, Journal of Financial Economics, Vol. 1, pp. 223-249. Beneish, M. D. and R. E. Whaley (1996),’An Anatomy of the “S&P Game”: The Effects of Changing the Rules’, Journal of Finance, Vol. LI, No. 5. pp. 1909-1930. Cooper, D. and G. Woglom (2003),’The S&P 500 Effect: Not Such Good News in the Long- Run, Federal Reserve Board, Research and Statistics, Washington DC, FEDS Working Paper No. 2002-48. Corrado, C. (1989),’A Non-Parametric Test for Abnormal Security Price Performance in Event Studies’, Journal of Financial Economics, Vol. 23, pp. 385-395. Dhillon, J. and H. Johnson (1991), ‘Changes in the Standard and Poor's List’, Journal of Business, Vol. 64, No. 1, pp. 75-85. Hamill, P. A. Opong, K. K. and McGregor, P. (2002), ‘Equity Option listing in the UK: a Comparison of Market-Based Research Methodologies’, Journal of Empirical Finance, Vol. 9, pp. 91-108. Harris, L. and E. Gurel (1986), ‘Price and Volume Effects Associated with Changes in the S&P 500 List: New Evidence for the Existence of Price Pressures’, Journal of Finance, Vol. XLI, No. 4, September, pp.815-829. Hillier, D. and P. K. Yadav (1997),’The Effect of Price Adjustment Delays on the Specification of Event Studies’, Working Paper (University of Strathclyde). Keim, D. B. and A. Madhavan (1996),’The Upstairs Market for Large-Block Transactions: Analysis and Measurement of Price Effects, The Review of Financial Studies, Spring, Vol. 9, Vo. 1, pp. 1-36. Lynch, A. W. and R. R. Mendenhall (1997),’New Evidence on Stock Price Effects Associated with Changes in the S&P 500’, Journal of Business, Vol. 70. No. 3. pp. 351-382. McIlkenny, P., K.K. Opong and I. Watson (1996), ‘Changes in FTSE-100 Index and Shareholders’ Returns’, Irish Accounting Review, Vol. 3 No. 1, Spring, pp. 91-110. Malkiel, B. G. and A. Radisich (2001),’The Growth of Index Funds and the Pricing of Wquity Securities’, The Journal of Portfolio Management, Winter, pp. 9-21. Mills, T. C. (1993),’The Econometric Modelling of Financial Time Series’, Cambridge University Press, Cambridge. Shleifer, A. (1986),’Do Demand Curves for Stocks Slope Down?’,Journal of Finance, July, pp. 579-590. 26
Table 1 Hypotheses Development CD-1 Pre-Change Permanent Combined Release Post- Hypotheses Day CAR CAR CAR CAR release CAR Price-Pressure/no interaction Largest MAR in Zero Zero Negative Negative Zero run-up Price-Pressure/ with interaction Largest MAR prior Positive Zero Zero Negative Zero to CD-1 Imperfect-Substitutes-Hypothesis/no Largest MAR in Zero Positive Positive Zero Zero interaction run-up Imperfect-Substitutes-Hypothesis /with Largest MAR prior Positive Positive Positive Zero Zero interaction to CD-1 Price-Volatility Hypothesis/no interaction Largest MAR in Zero Negative Positive Zero Negative run-up Price-Volatility Hypothesis/with interaction Largest MAR prior Positive Negative Positive Zero Negative to CD-1 Liquidity Zero Zero Positive Zero Zero Zero Notes: The hypotheses developed in this table are for firms added to the FTSE-100. These predictions are symmetric, the reverse is the case for firms deleted from the FTSE-100. CD is the change day, which is the day the stock is added to (delete from) the FTSE-100. CD-1 is the day before the stock is added to (delete from) the FTSE-100. The Pre-Change CAR covers the period from CD-21 to CD-1. The Release CAR runs from the CD to CD+3 for additions and CD+4 for deletions. The Post-Release CAR cover the period from CD+4 to CD+60 for additions and CD+5 to CD+60 for deletions. The combined CAR is a composite of the Release CAR and Pre-Change Day CAR for both additions and deletions. The Permanent CAR extends over the period from CD-21 to CD+60 for additions and deletions.
Table 2 Abnormal Returns and Trading Volume Surrounding Stock Additions to the FTSE 100 Abnormal Returns test- Abnormal test- statistic Volume statistic Panel A: Mean Abnormal Returns and Trading Volume Days relative to compositional change [-5] -0.0025 -1.55 -0.0730 -1.12 [-4] 0.0009 0.56 -0.0333 -0.51 [-3] 0.0010 0.63 -0.0420 -0.64 [-2] 0.0031 1.97 0.1013 1.55 [-1] 0.0145 9.13** 1.4449 22.10** Index compositional change [0] -0.0040 -2.50* 0.1917 2.93** [1] -0.0079 -4.98** 0.0942 1.44 [2] -0.0037 -2.34* 0.0557 0.85 [3] -0.0063 -3.97** -0.0014 -0.02 [4] -0.0032 -1.99* 0.1859 2.84** [5] -0.0018 -1.10 -0.0676 -1.03 Panel B: Cumulative Abnormal Returns and Trading Volume Pre-change day CAR [-21,-1] 0.0332 3.63** [-21,-1] 1.9722 3.86** Release CAR [0, 3] -0.0224 -7.12** [0, 3] 0.3402 1.65 Post-release CAR [4, 60] -0.0463 -3.22** [4, 60] -0.3481 -0.41 Combined CAR [-21,3] 0.0109 1.04 [-21, 4] 2.3124 4.14** Permanent CAR [-21, 60] -0.1821 -3.10** [-21, 60] 0.2130 0.21 Notes: The sample consists of 162 additions to the FTSE-100 from 30th March 1984 to 20th December 1999 for abnormal returns calculations. The trading volume sample consists of 132 announcements of firms added to the FTSE 100. Mean abnormal returns are estimated using the standard market model procedure. The cumulative abnormal returns were tested for significance using the robust variance estimate from the ZD test. Day ‘0’ is the day the stock is added to the FTSE 100. * represents significance at the 5% level, ** represents significance at the 1% level or less, using a two-tailed test.
Table 3 Abnormal Returns and Trading Volume Surrounding Stock Deletions from the FTSE 100 Abnormal Returns test- Abnormal test- statistic Volume statistic Panel A: Mean Abnormal Returns and Trading Volume Days relative to compositional change [-5] -0.0001 -0.04 0.4354 7.85** [-4] -0.0031 -1.83 0.1581 2.85** [-3] 0.0017 0.99 0.1022 1.84 [-2] -0.0039 -2.32* 0.1396 2.52** [-1] -0.0111 -6.55** 0.7205 12.98** Index compositional change [0] 0.0072 4.28** 0.0841 1.52 [1] 0.0061 3.60** 0.0487 0.88 [2] -0.0014 -0.85 0.0568 1.02 [3] 0.0009 0.53 0.0249 0.45 [4] 0.0021 1.22 0.0006 0.01 [5] 0.0040 2.39* 0.0187 0.34 Panel B: Cumulative Abnormal Returns and Trading Volume Pre-change day CAR [-21,-1] -0.0515 -5.41** [-21,-1] 2.3002 5.31** Release CAR [0, 4] 0.0148 3.88** [0, 4] 0.2144 1.22 Post-release CAR [5, 60] 0.0324 2.28* [5, 60] 0.0502 0.07 Combined CAR [-21,4] -0.0367 -3.30** [-21, 4] 2.5147 5.30** Permanent CAR [-21, 60] 0.0216 0.38 [-21, 60] 0.0245 0.03 Notes: The sample consists of 160 deletions from the FTSE-100 from 30th March 1984 to 20th December 1999 for abnormal returns calculations. The trading volume sample consists of 123 announcements of firms deleted from the FTSE 100. Mean abnormal returns are estimated using the standard market model procedure. The cumulative abnormal returns were tested for significance using the robust variance estimate from the ZD test. Day ‘0’ is the day the stock is added to the FTSE 100. * represents significance at the 5% level, ** represents significance at the 1% level or less, using a two-tailed test. 29
Figure 1 Estimation and event periods Time = 1 T T+1 e T+m 30
Figure 2 Mean Abnormal Returns for Firms Added to the FTSE-100 0.015 0.010 Mean abnormal returns 0.005 0.000 -0.005 -0.010 -21 0 50 100 150 Days relative to additions date 31
Figure 3 Mean Abnormal Trading Volume for Firms Added to the FTSE 100 1.5 Mean abnormal trading volume 1.0 0.5 0.0 0 50 100 150 Days relative to additions date 32
Figure 4 Cumulative Abnormal Returns for Changes in FTSE 100 Constituents Cumulative abnormal returns 0.00 Deletions -0.05 Additions -0.10 -0.15 -21 -7 0 20 40 60 80 100 120 140 Days Relative to Announcement of FTSE-100 Changes 33
Figure 5 Cumulative Abnormal Trading Volume for Firms Added to the FTSE 100 Cumulative abnormal trading volume 3 2 1 0 -21 0 20 50 100 150 Days relative to additions date 34
Figure 6 Mean Abnormal Returns for Firms Deleted from the FTSE-100 0.005 Mean abnormal returns 0.000 -0.005 -0.010 -21 0 50 100 150 Days relative to deletions date 35
Figure 7 Mean Abnormal Trading Volume for Firms Deleted from the FTSE 100 0.8 0.7 M ean abnormal trading volume 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -21 0 20 50 100 150 Days relative to deletions date 36
Figure 8 Cumulative Abnormal Trading Volume for Firms Deleted from the FTSE 100 3 Cumulative abnormal trading volume 2 1 0 -21 0 20 50 100 150 Days relative to deletion date 37
38
You can also read