The Price Dynamics Around Sensex Reconstitutions Vijaya B Marisetty*, AV Vedpuriswar** The price dynamics around index reconstitutions has been tested for an emerging market. Unlike developed markets like the US, market efficiency during the announcement and effective dates is much poorer in India. We found significant price dynamics for firms added to the sensex but not for the firms deleted. However, we found weaker evidence of permanent downward price effects for firms deleted from the sensex. Introduction Section I The Bombay Stock Exchange s (BSE) Sensitivity Index (Sensex) is one of the most popular indexes used by practitioners and researchers, to have a bird s eye view of the Indian economy. The 30 constituent firms of the index represent almost 50% of the BSE total market capitalization. However, they are not even half percent in terms of number of shares in the BSE. BSE has preset eligibility criteria for firms to be part of this elite 30 member group. The guidelines basically reflect firm performance and investors interest. The board of BSE meets quarterly to decide whether to delete the existing members by adding new members based on four major criteria namely, (1) the market capitalization of the members should be always above 5% of the total market capitalization of the sensex, (2) they should have been traded continuously during the last one year, (3) the volume should not be less than top 150 firms volume during the last one year and (4) by adding new members the correlation between old and new index should not be lower than 0.98 in order to maintain the continuity of the index and, apart from that, they also consider the track record and some other undisclosed qualitative criteria. With all these stringent criteria, BSE aims to represent the most active and valuable scrips of the stock exchange. In other words, it minimizes the efforts of individual and institutional investors to identify the most valuable scrips. One can argue this from the other way most valuable scrips identified by the investors will eventually become part of sensex. No matter how it happens, the investor reactions cause sharp changes in the prices of the firms deleted or added. Investors decisions for the changes will have a severe impact on their wealth. With the clearly defined criteria, it should be easy enough for the investors to anticipate which firms may be excluded and which firms may get included into sensex long before BSE announces such information. In fact, index funds and other mutual funds * Department of Accouting and Finance, Post box: 36, Menzies Building, Monash University, Clayton, Australia- 3800. Phone no: 0061-3-99058014; Fax: 0061-3-99055475; E-mail: Vijay.Marisetty@buseco.monash.edu.au. ** Dean, ICFAI Knowledge Center, Hyderabad, India. The Price Dynamics Around Sensex Reconstitutions 2003, The ICFAI Journal of APPLIED FINANCE. 5
should anticipate more accurately than retail investors. For instance, in case of individual investors, if an investor anticipates an addition in advance before the announcement, he/she can buy the shares of the firms for far lesser premium than after the announcement. After the announcement, index funds buy the added firm s shares in large quantities to reconstitute their portfolios creating an upward price pressure. Even though such a temporary price pressure results in an immediate price reversal after the effective date, the addition to the Sensex itself may bring some permanent positive increase to the share price of the newly added firm. A study on the investors reaction to Sensex reconstitution would help to analyze the wealth effects for retail and institutional investors. This section is followed by theories explaining the effects of inclusion and exclusion of firms from an index in section two. Section three describes the data used and the methodology adopted for this paper. The results are reported in section four. Section five concludes the paper. Section II Literature Although, extensive research work has been done in the US on stock switches from an index, there is no evidence of any research on Indian capital market. Researchers have proposed different theories to explain the abnormal returns associated with index reconstitution. Lynch and Mendenhall (1997) divided these theories broadly as follows: Price reversal effect. This Effect which was initially proposed by Harris and Gurel (1986). Harris and Gurel found that the price effect on the announcement and effective days is gradually but completely reversed in the period after addition and deletion. This has been attributed to the trading of index funds to reconstitute their portfolios that results significant but temporary shift in the equilibrium value of the shares. This theory is consistent with the block trading literature. Chan and Lakonishok (1993) and Keim and Mahadevan (1996) found temporary price shifts due to block trading of large investors. Permanent price shift effect. This Effect was observed by Shleifer (1986) and Dhillon and Johnson (1991). As per permanent price shift hypothesis, the temporary shift in the equilibrium persists, bringing the share price to a new equilibrium. While this theory contradicts price reversal effect, the reason behind permanent shift has been attributed to the changes in the investor revaluation. Information content effect. This effect was found by Jain (1987), where the market perceives addition of a firm in the index as good news through significant positive abnormal returns and deletion as bad news through significant negative abnormal returns, may support the revaluation argument of permanent price shift effect. Liquidity effect. This Effect was found by Amihud and Mendelson (1986) is another theory that explains price shifts. If being a member of an index increases (decreases) firms liquidity through the increase (decrease) in the trading volume, then the announcement of addition (deletion) results in increase (decrease) in the price. 6
Table1: Predictions of Hypotheses for Event Days and Specific Event Windows for Additions (Deletions) in the SENSEX Event Day Lynch and Mendenhall (1997) tested all the four theories on 34 additions and 33 deletions in S&P 500 index during 1990 and 1995. They found strong evidence supporting price reversal hypothesis both in additions and deletions. Weak evidence was found supporting permanent price shifts in additions compared to deletions. They also found that volume is abnormal between announcement date and the effective date. The largest abnormal volume was found a day before effective date. This was attributed to the index fund reconstitution just a day before the effective date to obtain the least tracking error. We use Lynch and Mendenhall (1997) framework to tests the above competing hypotheses. However, it differs in terms of the market conditions. This paper looks at an emerging market where the information dissemination process is assumed to be less efficient. The paper also helps the policy makers to understand the information dissemination process in such market conditions. Due to unavailability of daily trading volume data, the liquidity effect is not pursued in this paper. Data and Methodology Specific Event Windows AD Run-up Post-AD Total Release Post-release Hypotheses Permanent Permanent Price pressure zero positve zero zero negative zero (no anticipation) (negative) (positive) Price pressure positive zero negative zero negative zero (full anticipation) (negative) (positive) (positive) Downward sloping demand zero positive positive positive zero zero (no anticipation) (negative) (negative) (negative) Downward sloping positive zero zero positive zero demand (negative) (negative) (full anticipation) Information positive zero zero positive zero zero (negative) (negative) AD: Announcement date The predictions in the parentheses are for the firms deleted from sensex. Section III Data The data has been obtained primarily from BSE website and PROWESS database. BSE has listed the additions and deletions from sensex from 1986 till to date. The data includes the company names and the effective date of addition or deletion. BSE did not mention the exact announcement dates. However, they provided effective dates and also mentioned announcement is done six weeks before effective date. The daily stock prices data and daily market bench-mark (NSE DIFTY) data are sourced from PROWESS database from the Center for Monitoring Indian Economy. Table 2 summarizes the sample used for the study. The table includes 52 firms which have been classified as additions and deletions of 26, each along with the effective date. The Price Dynamics Around Sensex Reconstitutions 7
Table 2: List of Firms Added to and Deleted from BSE Sensex During 1996-2002 No Effective Date Firms Added Firms Deleted 1 08.19.1996 Arvind Mills Ballarpur Inds. 2 Bajaj Auto Bharat Forge 3 BHEL Bombay Dyeing 4 BSES Ceat Tyres 5 Colgate Century Text. 6 Guj. Amb. Cement GSFC 7 HPCL Hind. Motors 8 ICICI Indian Organic 9 IDBI Indian Rayon 10 IPCL Kirloskar Cummins 11 MTNL Mukand Iron 12 Ranbaxy Lab. Phlips 13 State Bank of India Premier Auto 14 Steel Authority of India Siemens 15 Tata Chem Voltas 16 11.16.1998 Castrol Arvind Mills 17 Infosys Technologies GE Shipping 18 NIIT Ltd. IPCL 19 Novartis Steel Authority of India 20 04.10.2000 Dr Reddy s Laboratories IDBI 21 Reliance Petroleum Indian Hotels 22 Satyam Computers Tata Chem 23 Zee Telefilms Tata Power 24 01.08.2001 Cipla Ltd. Novartis 25 01.07.2002 HCL Technologies Ltd. NIIT Ltd. 26 Hero Honda Ltd. Mahindra & Mahindra Competing hypotheses Table 1 represents various hypotheses being tested and the possible predictions to hold the various theories discussed in section II. The data is primarily divided into two event windows, one for the announcement date and the other for the effective date. Both windows have 10 days before and after event date (AD -10 TO AD +10 and ED-10 to ED +10). There is a controversy with regard to the number of days to be included around the event date. But there is a consistency of 10 days for most of the studies involved with index reconstitution. 8
Table 3: Daily Market-adjusted Abnormal Returns for Firms Added to and Firms Deleted From BSE Sensex During 1996-2002 Firms Added Firms Deleted Event Day N AAR AAR(t-value) AARAAR(t-value) Announcement Date Window -10 52 0.005228 3.614075 0.009959 5.627868-9 52 0.007248 5.010561 0.004571 2.583282-8 52 0.003324 2.297902-0.0012-0.67598-7 52 0.004697 3.247246 0.000131 0.074066-6 52 0.002438 1.685413-0.00498-2.81285-5 52 0.006938 4.795978-0.00646-3.65214-4 52 0.010998 7.602691-0.00836-4.72223-3 52-0.00587-4.06073 0.011429 6.459124-2 52 0.013167 9.101918 2.92E-05 0.016483-1 52-0.00143-0.98541 0.005137 2.903071 0 52 0.007071 4.887738 0.005712 3.227901 1 52 0.002971 2.053626-0.00978-5.52596 2 52 0.012045 8.32663 0.007125 4.026429 3 52-0.00293-2.02545-0.02448-13.8358 4 52-0.00341-2.36007-0.00192-1.08357 5 52 0.008073 5.580812 0.001092 0.61726 6 52-0.00527-3.64626-0.00468-2.64687 7 52-0.01127-7.78933 0.007944 4.48952 8 52 0.004232 2.925371 0.002288 1.293121 9 52-0.0021-1.44921-0.00017-0.09706 10 52-0.0059-4.07796-0.00583-3.29372 Effective Date Window -10 52 0.004157 2.510585-0.01216-6.89755-9 52-0.00021-0.12841-0.00839-4.75967-8 52-0.00136-0.82131-0.01323-7.50135-7 52 0.002741 1.65504-0.00095-0.54054-6 52 0.000187 0.113042 0.009967 5.65273-5 52-0.0165-9.96675 0.000856 0.48554-4 52-0.00338-2.03917 0.005809 3.294746-3 52-0.00746-4.50477 0.00476 2.699595-2 52-0.00158-0.95362-0.00145-0.82351-1 52 0.001877 1.133277-0.00317-1.79681 0 52 0.010578 6.388032-0.02114-11.9911 1 52 0.007342 4.433884-0.00346-1.964 2 52 0.005973 3.607154 0.001596 0.905221 3 52-0.00995-6.0076 0.005814 3.297199 4 52 0.006944 4.193706-0.00182-1.02997 5 52-0.00224-1.35532 0.003008 1.705839 6 52-0.00852-5.14382 0.002303 1.30606 7 52-0.00062-0.37742-0.00503-2.85363 8 52-0.00153-0.92173 0.008348 4.734523 9 52 0.014324 8.650385 0.005589 3.16974 10 52-0.01232-7.43738-0.0128-7.26076 N: Number of firms. AAR: Average Abnormal Return of the 52 firms (Significance at 95% confidence level). The Price Dynamics Around Sensex Reconstitutions 9
Table 4: Specific window statistics for firms added to and deleted from BSE Sensex during 1996-2002 FIRMS ADDED FIRMS DELETED Specific Event Window Event Days CAAR CAAR(t-value) CAAR CAAR(t-value) Run-up AD+1,ED-1 0.02509-0.42039-0.04066-0.55896 Post-AD permanent AD+1,ED+10-0.01511-0.22923-0.06397-0.61321 Total permanent AD, ED+10-0.00804-0.12169-0.05838-0.51253 Release ED,ED+7 0.009985 0.517875-0.06397-3.39344 Post-release ED+8,ED+10-0.01384-1.03044-0.06397-5.75455 ED: Effective Date. CAAR: Cumulative Average Abnormal Return of the 52 firms (Significance at 95% confidence level). The paper tests three of the four hypotheses discussed in section II. The event windows are sub-divided to allow us to test these hypotheses. Apart from the Announcement Day (AD) there are four event specific windows, as shown in Table 1. The run-up window ranges from a day after the announcement day and ends at a day before effective day (AD to ED-1). Run-up window looks at the abnormal return after the announcement date. Run-up window can be used to explain price pressure effect hypothesis. For instance, if the investors have no anticipation of the additions/deletions then there should be no significant abnormal return till the announcement date. However, abnormal return should be prominent during the run-up window and especially on the last day before the effective date. The largest abnormal return should occur just before the effective date as the index funds rebalance their portfolios to avoid tracking error. If there is full anticipation of the changes, then the abnormality should be found on the announcement date and before the announcement date. One can also interpret the results of run-up window to explain information hypothesis. For instance, if additions/deletions have value-based information and no price pressure associated with it, then there should be abnormal return on the day of announcement and there should not be any abnormality when it reaches effective date. In other words, there should be no abnormal return on the last day of the run-up window. The next window is the post announcement date permanent window, which ranges from a day after announcement day to the 10 th day from the effective date (AD +1 to ED +10). This window captures the price reversal and permanent price effects. Again, if the investors have full anticipation of the changes to the index and there are any price reversals to offset the abnormality, then there should be abnormal negative (positive) returns in case of additions (deletions). Total permanent window starts from the effective date and ends on the 10 th day after the effective date looks at the persistence in the abnormal returns started on the announcement date. The persistence in the abnormal return is attributed to the permanent price effect. Release window and post release window represents the abnormal returns after the change has occurred in the index. Release window is from the effective date to the seventh day after effective date and the post-release window follows till the 10 th date after the effective date. These windows capture the price reversal effect that arises due to temporary price pressure. There should be a complete reversal at the end of release window and there should be no abnormality in the post release window if markets are semi-strong efficient (Fama, 1992). 10
Graph 1: Deleted Firms Announcement Date Window Average Abnormal Returns Graph 2: Deleted Firms Effective Date Window Average Abnormal Returns Graph 3: Added Firms Announcement Date Window Average Abnormal Returns The Price Dynamics Around Sensex Reconstitutions 11
Methodology We use event study methodology using the conventional Brown and Warner (1985) event study method in order to test the price pressure, permanent effect and information hypotheses as discussed in the above sections. Measuring abnormal return There are several methods to measure the abnormal returns in the event-study literature. We got almost similar results for conventional risk-adjusted, mean-adjusted, market adjusted and market and risk adjusted models. The market-adjusted model results are reported for interpretational purposes. In a market adjusted method, Abnormal Return (ARi) of firm i is: AR it = R it R mt where Rit is the continuously compounded daily return of firm i and R mt is the continuously compounded return of the market index. Cumulative Abnormal Return (CAR i ) varies with the event specific windows. For instance, the CAR i for the run-up window is: CAR i = S t= ad+1 to ED - 1 AR i. The t-statistic to test the significance of AR it is as follows: t = AR it /SEE, where SEE = sqrt (var(ar it )/N) where sqrt = square root ; Var= variance and N = number of observations. The t-statistic to test the significance of CAR it is as follows: t = CAR it /SEE, where SEE = sqrt (T*var(AR it )/N) where t is the number of days in each event specific window and AR it is the AR during that event specific window. Section IV Results TableS 3 and 4 provide summary results of the study. Graphs 1 to 4 support the findings reported in tables 3 and 4. Table 3 reports the average abnormal return of 52 firms along with their t-values. The average abnormal returns are reported for the announcement date window and effective date window and they represent both additions and deletions to sensex during 1996 to 2002. The event date has significant AR s for both additions and deletions on announcement and effective dates with the t-values ranging from 3 to 11. Addition to sensex has provided a positive AR during both on the announcement date and effective dates. However, for deletion from sensex has provided a positive abnormal return on the announcement date and a negative abnormal return on the effective date. The positive abnormal return on the announcement due to addition may imply that investors consider and value additions as good news. The results also support full anticipation hypothesis. There are signs of price reversal two days after the effective date. However, investors received 7% and 10% daily returns during the announcement day and effective date for the firms added to sensex. The results violate market efficiency hypothesis and represent a slow adjustment process to the information. 12
The deleted firms exhibit surprising results. The deleted firms exhibited significant positive results on the announcement day. However, on the effective day they have significant negative abnormal return. The announcement information, if not covered extensively in the media, may lead to inconsistent reactions on the AD but the real reaction might occur on the effective date. Table 4 reports the cumulative average abnormal return for different event specific windows for both firms added and deleted from sensex. The results in table 4 show that the only significant cumulative average AR are for the deleted firms in the release and post-release windows. The deleted firms exhibit significant negative cumulative AR s during the release and post-release windows. These results support slow adjustment process as discussed above. The investors react more after the effective date than on the effective date. In an efficient market where the information dissemination is fast, the negative returns would have occurred before and on the announcement date. The results also contradict the price reversal effect for deleted firms. A deleted firm s temporary decrease in the price should reverse back during the release window. This may suggest that the deleted firm settles at a new equilibrium as market re-values the worth of the deleted firm. This argument supports permanent price shift in case of deleted firms, even though there is no significant cumulative average AR in the permanent post AD window and total permanent window. Graphs 1 to 3 report significant patterns in the return generation process. For instance, graph 2 supports the argument that the deleted firms reach a lowers equilibrium due to the permanent shift after the effective date. Graph 3 shows the increasing momentum in the prices of the added firms before the announcement date. This compliments with the full anticipation of additions by the investors before the announcement date. However, graph 3 indicates that price reversal happens far before the effective date. This may be due to the fact that unlike in the US, the index funds price pressure is low in India or there are only a few index funds based on BSE index during our study period. Section V Concluding remarks We analyzed the price dynamics around the sensex reconstitution using three competing hypotheses namely, price pressure hypothesis, permanent shift hypothesis and information hypothesis. We found no consistent results that can strongly support any one of these hypotheses. These results might be due to the low index funds activities on the sensex index during the study period. Most of the earlier studies are based on the US stock market where the influence of index funds trading during index reconstitutions is substantial. There is evidence of investors anticipating firms being added to sensex. However, they fail to anticipate the firms being deleted from sensex. There is considerable evidence that the information arrival and adjustment process in BSE is slow. There is weak evidence that the deleted firms will have permanent price effect. A trading strategy to profit from the announcements would yield a significant positive market-adjusted return. Reference # 01J-03-07-01 The Price Dynamics Around Sensex Reconstitutions 13
References 1. www.bseindia.com 2. Amihud Y, and Mendelson H, 1986. Asset pricing and bid ask spread, Journal of Financial Economics, 17: 223-49. 3. Brown S and Warner J, 1985. Using daily stock returns: The case of event studies, Journal of Financial Economics, 14: 3-31. 4. Chan L and Lakonishishok J, 1993. Institutional trades and intra-day stock price behavior, Journal of Financial Economics, 33: 173-200. 5. Dhillon U and Johnson H, 1991. Changes in S$P s 500 list, Journal of Business, 64: 75-85. 6. Fama E 1991. Efficient capital markets II, Journal of Finance, 46: 1575-1618. 7. Harris L and 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 Financial Economics, 26: 815-29. 8. Jain P 1987. The effect on stock price from inclusion and exclusion from the S&P 500, Financial Analyst Journal, 43: 58:65. 9. Keim and Mahadevan A, 1996. The upstairs market for large block transactions: analysis and measurement of price effects, Review of Financial Studies, 9: 1-36. 10. Lynch WA and Mendenhall RR, 1997. New evidence on the stock price effects associated with changes in the S&P 500 index, Journal of Business, 3: 351-383. 11. Shleifer A 1986. Do demand curves for stocks slope down?, Journal of Finance, 41: 579-90. 14