WU Wien. November 23, 2012 AWG Innsbruck. Price and Dividend Implications. of Index Composition Changes. Georg Cejnek, Otto Randl. WU Wien.

Size: px

Start display at page:

Download "WU Wien. November 23, 2012 AWG Innsbruck. Price and Dividend Implications. of Index Composition Changes. Georg Cejnek, Otto Randl. WU Wien."

Monica Randall
5 years ago
Views:

1 November 23, 2012 AWG Innsbruck 1/33

2 Agenda (Euro Stoxx 50) 2/33

3 Stock market indices are extremely important in practice Huge market share of passive investing (ETFs) Underlying for derivatives Development of dividend indices Growing importance of long term dividend derivatives Investment bank sales argument positive survivorship bias on index compositon emphasizes single stocks Only few papers look into long term effects on index itself These papers are partly contradictory Focus is almost entirely on prices Needed: Consistent methodology to investigate both prices and dividends on the index level 3/33

4 Euro Stoxx 50 s jan jan jan jan jan jan2012 4/33

5 (1) Abnormal return studies Downward sloping demand curve If securities are not perfect substitutes for each other, additional demand from index inclusion will create permanent price increase, e.g. Shleifer (1986) Price pressure Temporarily, less than perfectly elastic demand curve as liquidity providers want more compensation. Price pressure reverses over time, e.g. Harris and Gurel (1986), Lynch and Mendenhall (1997) Liquidity hypothesis Index inclusion leads to more liquidity. As required liquidity premium shrinks, there is a corresponding (permanent) price increase, e.g. Edmister, Graham and Pirie (1996), Erwin and Miller (1998) 5/33

6 (2) Abnormal return studies (cont.) Information content hypothesis Inclusion of a stock in the index reveals important information on the quality of the stock which should have (permanent) price impact, e.g. Denis, McConnell, Ovtchinnikov and Yu (2003), Jain (1987), Kappou, Brooks and Ward (2008) Long term index studies Siegel and Schwartz (2006) find the orginal S&P 500 companies from 1957 to outperform the rebalanced index by 55bp p.a. Cai and Houge (2008): passive buy-and-hold portfolio outperforms Russell 2000 small cap index ( ) by 222bp p.a. Ranaldo and Häberle (2008) argue that indices are forms of active investment management. Exclusive indices outperform in bull markets. 6/33

7 (3) Momentum and mean reversion of stock prices e.g., Balvers, Wu, and Gilliland (2000) and Balvers and Wu (2006) find mean reversion for stock indices, resp. in combination with momentum effects. payouts Life cycle theory of dividends. Evidence that bigger (book value) and more mature companies pay out more. e.g., Fama and French (2001), Eije and Megginson (2008), Denis and Osobov (2008). 7/33

8 How does periodic rebalancing (together with momentum and mean reversion) affect price indices? Does this also affect dividend indices? Are these effects stronger for more exclusive indices? 8/33

9 Eurostoxx 50 price index and Eurostoxx 50 dividend index from 01/ /2011 Prices and dividends of all member companies (at least at one point in time) and direct descendants Member weights, index divisor Information on corporate actions and stock price adjustments Additional data for added and deleted companies (e.g. book value) 9/33

10 (1) Calculate number of shares: n i,t = weight i,t index t divisor t unadjusted stockprice i,t The stock index can be replicated by index t = 50 i=1 n i,t unadjusted stockprice i,t divisor t Each time when active index composition changes occur, we start construction of a passive portfolio: passivepf t = 50 i=1 n i,t unadjusted stockprice i,t divisor t 10/33

11 (2) Similarly, for dividends the index is constructed: DP t = 50 i=1 n i,t unadjusted dividends i,t divisor t We aggregate daily dividend points (DP t ) over each calendar year: T DI t = DP t t=t 0 As for the price index, we also calculate for the dividend index passive portfolios for each composition change. 11/33

12 (3) Event dates: Whenever at least one company gets replaced by another stock in the index. For each event date, we compare the price and dividend performance of the actual indices and the passive buy-and-hold portfolios, 1, 2 and 3 years following the event. In addition, we perform pairwise comparisons (dividends, prices) and calculate alphas from market factor and Fama French regressions. 12/33

13 Index Passive PF: Summary Statistics Panel A. s. year +1 year +2 year +3 Mean Median Min Max Nr of obs Panel B. Price Index. year +1 year +2 year +3 Mean Median Min Max Nr of obs /33

14 s vs. Price Performance s vs. Price Performance Average Effect Across Events year +1 year +2 year +3 s Stock Index 14/33

15 Index Passive PF: t-tests t-statistics year +1 year +2 year +3 effect Price index effect effect Price index effect Degrees of freedom /33

16 Pairwise Comparison: Summary Statistics Panel A. s. year +1 year +2 year +3 Mean Median Min Max Nr of obs Panel B. Prices. year +1 year +2 year +3 Mean Median Min Max Nr of obs /33

17 Pairwise Comparison: t-tests t-statistics year +1 year +2 year +3 effect Price index effect effect Price index effect Degrees of freedom /33

18 Event Fundamentals Event Day Added Company Deleted Company Total Ratio Ret. Market Equity Earnings / Tot. Equity Cap Lafarge PPR % Iberdrola Bayrische HVB % SAP Aventis % Credit Agricole Volkswagen % Allied Irish Banks Telecom Italia M % Renault Royal Dutch % Vinci Lafarge % Schneider Electrics Allied Irish Banks % Arcelormittal Ahold % Volkswagen Endesa % Deutsche Börse ABN Amro % Alstom Alcatel Lucent % CRH PLC Renault % Anheuser Busch Fortis (Ageas) % Unibail Rodamco Volkswagen % BMW Aegon % Inditex Credit Agricole % Volkswagen Pref Alstom % Mean % 1.94 Median % /33

19 Market Beta Regressions Start date Beta t (β 0) t (β 1) Alpha t-stat 22. Sep Sep Jul Sep Jun Jul Sep Okt Okt Sep Sep Feb Sep Sep Mean /33

20 Fama French Regressions Start date Beta t (β 1) SMB t-stat HML t-stat Alpha t-stat 22. Sep Sep Jul Sep Jun Jul Sep Okt Okt Sep Sep Feb Sep Mean /33

21 Simulate dividends: d t+1 = d t e ηt ( η t N 1 ) 2 σ2 η,ση 2 Simulate price-to-dividend ratios: PD t+1 = PD t e λt λ t = (1 δ) (lnpd lnpd t 1 )+ρ λ t 1 +ǫ t ( ǫ t N 1 ) 2 σ2 ǫ,σǫ 2 21/33

22 (2) Calculate prices as P t = d t PD t Construct market cap weighted price indices Construct dividend indices with same weights as price indices All indices with periodic rebalancing and without rebalancing Compare differences over 10 years 22/33

23 Variants sim 1 sim 2 sim 3 sim 4 nr. sim. runs nr. of companies nr. of index members nr. of periods starting period σ η δ σ ǫ ρ /33

24 : base case (1), price index apply(ergebnisse[, 12:21] ergebnisse[, 2:11], 2, mean) Index 24/33

25 : base case (1), dividend index apply(ergebnisse[, 32:41] ergebnisse[, 22:31], 2, mean) Index 25/33

26 : small index (2), price index apply(ergebnisse[, 12:21] ergebnisse[, 2:11], 2, mean) Index 26/33

27 : small index (2), dividend index apply(ergebnisse[, 32:41] ergebnisse[, 22:31], 2, mean) Index 27/33

28 : more momentum and less mean reversion (3), price index apply(ergebnisse[, 12:21] ergebnisse[, 2:11], 2, mean) Index 28/33

29 : more momentum and less mean reversion (3), dividend index apply(ergebnisse[, 32:41] ergebnisse[, 22:31], 2, mean) Index 29/33

30 : more momentum and less mean reversion (4), small price index apply(ergebnisse[, 12:21] ergebnisse[, 2:11], 2, mean) Index 30/33

31 : more momentum and less mean reversion (4), small dividend index apply(ergebnisse[, 32:41] ergebnisse[, 22:31], 2, mean) Index 31/33

32 - Summary of Eurostoxx 50 price and dividend index On average, Eurostoxx 50 price index outperforms the buy-and-hold portfolios. However, barely statistically significant and decreasing magnitude over time. For dividend index: negative sign, no statistical significance Pairwise comparisons of companies added/removed: no statistical significance for prices and dividends; inconclusive with respect to proxies for life cycle theory of dividends. Market beta regressions: betas close but not equal to one. There seem to be systematic effects left. Fama French regressions capture most of that. results Momentum is positive for rebalanced price index. However, rebalancing generally leads to lower dividends. Exclusive indices tend to exhibit stronger patterns. 32/33

33 - Takeaways Momentum effects can lead rebalanced price indices to outperform buy and hold; mean reversion to underperformance. In contrast to conventional wisdom, rebalancing generally leads to relatively lower levels of dividend indices. Overall, magnitudes are rather small. Exclusive indices tend to exhibit stronger patterns. 33/33

S&P 500 INDEX RECONSTITUTIONS: AN ANALYSIS OF OUTSTANDING HYPOTHESES. Lindsay Catherine Baran

S&P 500 INDEX RECONSTITUTIONS: AN ANALYSIS OF OUTSTANDING HYPOTHESES by Lindsay Catherine Baran A dissertation submitted to the faculty of The University of North Carolina at Charlotte in partial fulfillment