Bachelor Thesis
The inclusion effect in the EURO STOXX 50 index
Author: Wouter Pigge Student number: 10353518
Specialization: Finance and Organization Field: Finance
Supervisor: R.C. Sperna Weiland
June 29, 2015
Verklaring eigen werk
Hierbij verklaar ik, Wouter Pigge, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.
Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties
worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.
Abstract
In this paper we examine the inclusion effect in the EURO STOXX 50 index. Previous studies have reported positive abnormal returns for stock inclusions in the S&P500. To investigate whether an inclusion effect is present in the Euro Stoxx 50 index, we use a standard event study with data from September 1999 until September 2014. Two event windows are examined, one around the announcement date and one around the effective date of a stock inclusion. We observe a 1.206% positive abnormal at a 10% significance level two days prior to the effective date. This could be evidence of price pressure caused by index funds, but the lack of other significant observations makes it unclear if this abnormal return is purely driven by the inclusion effect. We address the lack of significant abnormal returns by a higher awareness of inclusion effects. Our findings could be seen as evidence that demand curves for stocks are elastic and as disproof for the existence of an inclusion effect for the EURO STOXX 50 index.
Table of Contents
Abstract………..………3 1. Introduction………..…….5 2. Academic Framework………..………6 3. Related Literature……….………..……..8 4. Methodology………...…….………104.1 Event study methodology ..………10
4.2 Description of event………....…..………..15 4.3 Hypothesis………..……….………15 5. Data description …...………..…17 6. Results ...………..……19 6.1 Empirical results………..19 6.2 Analysis of results………...20 7. Conclusions………...………..….22 References……….………..……….23 4
1. Introduction
The inclusion effect refers to the positive price reaction stocks experience when they become included in a market index. Index funds try to mimic the performance of an index. When an index is tracked by index funds, fund managers must buy the stock that becomes included in the index. After the news of a new inclusion is out, investors make a profit with buying shares of a stock that becomes included ahead of index funds and sell them after the index fund demand is satisfied. Previous studies have focused on the S&P500 index and found on average a 3% significant abnormal return immediately after an inclusion is announced.
The existence of the abnormal return is explained by several conflicting hypotheses. These hypotheses are the downward sloping demand curves for stocks, temporary price pressures caused by the demand of index funds, the fact that an inclusion contains new positive information and lower trading cost resulting from the fact that a stock becomes more liquid.
Majority of the studies investigate stock inclusions from the late seventies until the first half of the nineties. Awareness for the inclusion effect must have grown since then, which could have attracted more investors and diminished the magnitude of the inclusion effect. On the contrary, investments in index funds have grown since then, which must enlarge the inclusion effect.
In this paper, we examine the inclusion effect in the EURO STOXX 50 index from September 1999 until September 2014. The EURO STOXX 50 index is the leading blue-chip index for the Eurozone and serves as underlying for a wide range of investment products (STOXX Ltd., 2015). The aim of this study is to analyze whether there are abnormal returns observed, driven by the inclusion effect, around index inclusions in the EURO STOXX 50 index. We do not find clear evidence that this is the case.
This study is organized as follows. In section 2 we will discuss several hypotheses, which could explain the inclusion effect. In section 3 we give an evaluation of related literature. Further, in section 4 we will describe the event study methodology, specify our event of interest and discuss our hypothesis. Section 5 is a summary of the data we use. In section 6 we summarize our findings and analyze them. At last, in section 7 we will present our conclusions and recommendations for future studies.
2. Academic Framework
A lot of research has been done on the effects that occur when a stock becomes included in an index. Hypotheses, which could explain the inclusion effect, have been put forward in the literature. Shleifer (1986), for example, examines the downward sloping demand curve for stocks hypothesis (the DS hypothesis), also referred to as the imperfect substitute hypothesis. The DS hypothesis states that demand curves for stocks are not perfectly elastic. This is could be observed when stocks do not have close substitutes.Index funds try to track the
performance of an index by holding the securities that are included in the index.After an announcement of a new stock inclusion, index funds will start to rebalance their portfolios. They buy the shares of the new included stock. The index funds use a buy-and-hold strategy. The shares bought are hold for as long as the stock is quoted on the index and are no longer available in the market for trading. If the stock has no close substitutes, investors will also hold on to their shares, because they cannot find other stocks with approximately the same expected return and similar risk. If the decreasing quantity of shares in the market is accompanied with an increase in share price, the demand curve for stocks must slope
downward. After the demand of the index funds is satisfied, the prices and quantities of shares represent a new equilibrium. Therefore, the inclusion effect must be permanent.
Harris and Gurel (1986) examine the price pressure hypothesis (PP hypothesis). The PP hypothesis states, in contrast to the DS hypothesis, that the demand curve is perfectly elastic in the long-run only not in the short-run. In the short-run, an immediate increase in price is necessary to induce current shareholders to sell their shares to the index funds. After the index funds are rebalanced the price pressure will be over and prices will drop to their pre-announcement levels. The short increase in price is explained by index funds paying a
premium to compensate the previous stockholders for their liquidity service. There is no permanent effect.
The DS hypothesis and the PP hypothesis assume that the inclusion or exclusion of a stock in an index is an information free event. The information hypothesis states that an inclusion of a firm in an index contains relevant information to investors. Denis, McConnell, Ovtichinnikov and Yu (2003) have a few suggestions why the information hypothesis would hold. For instance, a stock inclusion in an index could lead to greater monitoring of
management by investors, so that management would respond with greater effort. Or that the managers of a firm that is included in an index work harder because their cost in managerial
reputation is greater than if the same firm where not included in the index. Both actions could lead to higher future earnings. So an inclusion may signal extra good news about the future performance of a firm. Instead of price-pressures or downward sloping demand curves, this good news could explain the increase in stock prices. This information will be priced in the stock and represents a new equilibrium. Like the DS hypothesis and in contrast to the PP hypothesis, the information effect must also be permanent.
A fourth hypothesis, proposed by Amihud and Mendelson (1986) is the liquidity hypothesis. The liquidity hypothesis states that investors require a higher expected return for higher trading costs. Inclusion in an index could lead to higher trading volumes. Higher trading volumes could lead to a lower bid-ask spread and lower the overall trading costs. This effect lowers the cost of capital of a firms stock and thereby increases the stock prices until a new equilibrium is reached. The price increase caused by the liquidity effect is also
permanent.
3. Related Literature
A lot of research has been done on the effects that occur when a stock becomes included in an index. Most of these studies focus on the U.S stock markets, mainly on the S&P500 index. A majority of the studies find positive abnormal returns for stock inclusions. There are
conflicting results about the persistency of these returns.
The study of Harris and Gurel (1986) support the PP hypothesis. They found a 3.13 percent cross-sectional mean abnormal return on the first day after the announcement for the period 1978-1983. Volumes increased significantly, suggesting a shift in demand. Harris and Gurel (1986) state that this shift in demand must be caused by index funds. Prices tended to return to their pre-announcement levels after three weeks, which is not in line with the DS, liquidity and information hypotheses.
Schleifer (1986) finds approximately the same abnormal return as Harris and Gurel (1986), only with a persistence of at least 10 to 20 days for the same period. Schleifer (1986) finds empirical evidence on an increase in volumes and he supports, like Harris and Gurel (1986), that this must be caused by the demand of index funds. He states that the persistence of the abnormal return and the increase in demand is evidence for the DS hypothesis.
Both Schleifer (1986) and Harris and Gurel (1986) suggest in their studies that the abnormal returns are caused by index funds. They do, however, not provide real evidence for this thought . Pruitt and Wei (1989) examine the changes in institutional ownership following from additions and deletions from the S&P 500 in search for evidence for the suggestion made by Schleifer (1986) and Harris and Gurel (1986). They found a positive correlation between the abnormal returns and changes in institutional security holdings in reaction to additions to the S&P500. Pruitt and Wei (1989) state that the increase in institutional
ownership is relatively small to the observed increase in trading volumes by Harris and Gurel (1986). They suggest that this small change means that non-index institutions supply the required liquidity coming from the increased demand by index funds for a stock that will be listed on the S&P500.
After October 1989, the S&P changed the announcement and implementation policy. Before 1989, the announcement of a change in the index composition where made at the end of a trading day and the change became effective on the next trading day. The policy since October 1989 has been to announce a change one week before the change in the index composition becomes effective (Lynch and Mendenhall, 1997).
Baneish and Whaley (1996) examined the effect of this policy change. During the period from January 1986 through September 1989 they found on average a 3.7 percent abnormal return on the day after the announcement. This increase, in contrast to the abnormal returns found by earlier studies, is consistent with the growth in index funds. After the policy change, they found a lower abnormal return of 3.1 percent (for the period October 1989 through 1994). Furthermore, they found that the stock price increased by another 4.1 percent by the effective date. Baneish and Whaley (1996) suggest that the index funds appear to wait with rebalancing their portfolios until the effective date and that the increase after the
announcement date is partially caused by arbitrage traders.
Lynch and Mendenhall (1997) found a 3.158 percent significant abnormal return at the announcement date for the period March 1990 through April 1995, almost consistent with the result of Baneish and Whaley (1986). Through the announcement and the effective date they find a positive cumulative abnormal return of 3.8 percent, this could be some support for the DS hypothesis. After and following the effective date they, however, observe a price reversal, evidence for the PP hypothesis. They suggest that the 3.8 percent cumulative abnormal return is the premium paid by index funds to compensate the previous stockholders, consistent with the PP hypothesis theory. Lynch and Mendenhall (1997) find no evidence to rule out the possibility that the information and liquidity hypotheses contribute to the abnormal returns after the announcement date, while they do find evidence to rule out the information and liquidity hypotheses as complete explanations for the observed price changes.
In contrast to other studies, Denis et al. (2003) find that the event of inclusion in the S&P500 index contains information. They look at analysts’ earnings per share forecasts around index inclusions before and after inclusion. After inclusion companies experience a significant increase in the analysts’ earnings per share forecasts and significant improvements in realized earnings, relative to benchmark companies who did not got included. There results show that analysts’ earnings expectations rise when a company becomes included in the S&P500. Denis et al. (2003) state that for the rise in earnings expectations to be rational they must be based on new information about the performance of the company. Denis et al. (2003) do not rule out the existence of a downward sloping demand curve effect as well, but they do conclude that additions in the S&P500 index is not an information free event.
4. Methodology
In our study, we are investigating whether an inclusion effect is present in the Euro Stoxx 50 index. To do this, we will use standard event study explained by MacKinlay (1997). In section 4.1, we will discuss the event study methodology in general. In section 4.2, we will specify our event definitions in more detail and in section 4.3. we will discuss our hypotheses.
4.1 Event study methodology
An event study assesses the impact of an event on the value of a firm and is widely used in finance research. The goals is to check whether there are abnormal returns driven by the event being studied. Mackinlay (1997) defines an abnormal return as the realized post return of a security over an event window minus the actual normal return of that security over the event window. Normal returns are defined as the expected returns of a security without the event taking place.
In order to construct an event study, MacKinlay (1997) states that we first need to define the estimation window and the event window. The event window is the period over which the stock prices of the firms in the event will be examined. Second, the estimation window is the period used to estimate normal returns. The two periods are illustrated with a time line in Figure 1.
Figure 1. Time line for an event study.
The notations in Figure 1 are intended to facilitate the measurement and analysis of abnormal returns. The returns will be indexed in event time using τ. The event date is defined as τ = 0. The estimation window is the period from τ = T0 to τ = T1 and the event window is
represented by the period from τ = T1 until τ = T2. The length of the estimation window is
given by L1= T1-T0 and the length of the event window is defined as L2=T2-T1.
The hypotheses that are tested for the event are:
H0: There is no significant abnormal during the event window.
H1: There is a significant abnormal return during the event window.
The formula for calculating the abnormal return for a stock i on event date τ is given in Equation 1:
ARi,τ= Ri,τ− E�Ri,τ�Xτ�, (1)
where ARi,τ is the abnormal return, Ri,τ the actual return and E�Ri,τ�Xτ� is the expected normal return of stock i on event date τ.
Two common choices for modeling normal returns, mentioned by MacKinlay (1997), are the constant mean return model and the market model. The constant mean return model assumes that the mean return of a security is constant over time, the market model assumes a stable linear relationship between the market return and the security’s return. Brown and Warner (1985) state that the market model outperforms the constant mean return model. MacKinlay (1997) remarks that the market model yields a lower variance of the abnormal returns than the constant mean return model. For these reasons, we are more confident using the market model in this study.
The market model is a statistical model, which assumes that asset returns are jointly multivariate normal and independently and identically distributed through time. Brown and Warner (1985) find that non-normality of daily returns does not have a serious impact on event study methodologies. Mackinlay (1997) comments that in practice a violation of this assumption does not leads to a problem because the assumption is empirically reasonable and inferences using the market model tends to be robust to deviations from the assumption.
The market model is expressed in Equation 2.
Ri,t = αi+ βiRm,t+ εi,t (2)
with E�εi,t� = 0 and VAR�εi,t� = σ2εi
Here Ri,t is the return of stock i on time t and Rm,t the market return at time t. αi and βi are the parameters of the model which will be estimated using ordinary least squares (OLS) method. εi,t is defined as the zero mean residual and σε,t2 is the variance of the residuals.
Under the assumptions that asset returns are jointly multivariate normal and
independently and identically distributed through time, the OLS is an efficient and consistent estimation procedure for the market model parameters (MacKinlay, 1997).
After estimating the market model parameters we can measure the abnormal returns. When we let ARi,τ, τ = T1, . . . , T2 be the sample of L2 abnormal returns for stock i in the
event window, the abnormal returns can be calculated using the sample formula in Equation 3 (MacKinlay, 1997).
ARi,τ = Ri,τ− α�i− β�iRm,τ (3)
Here Ri,τ and Rmt,τ are the actual stock returns of stock i and the actual market returns in the event window for τ = T1, . . . , T2, respectively. The terms α�i and β�i are the estimated
parameters of the market model of stock i and assumed to remain constant in the event window.
Under the null hypothesis, conditional on the event window market returns, the abnormal returns will be jointly normally distributed with a zero conditional mean and
conditional variance ARi,τ ~ N(0, σ2(ARi,τ)). The conditional variance is stated in Equation 4. The first part of Equation 4 is the variance of the residuals from Equation 2, the second part of the equation is the additional variance due to the sampling error in αi and βi.
σ2�AR i,τ� = σε2+ L1 1�1 + (Rm,τ− µ�m)2 σ�m2 � (4) with τ = T1, . . . , T2
Here σε,t2 is the variance of the residuals from Equation 2, L1 is the length of the
estimation window, Rm,τ is the market return on day τ in the event window, σ�m2 is the
variance of the market in the estimation period and µ�m is the average market return during the estimation period.
The sampling error is caused by the fact that the αi and βi are calculated with data from the estimation period and that the variance of the abnormal returns contains data from
the event window. This could lead to serial correlation of the abnormal returns. By extending the length of the estimation window L1, the second part of the equation approaches zero and
the sampling error disappears.
To draw overall conclusions for the event of interest, we need to aggregate the abnormal return observations. The abnormal return observations can be aggregated through time and across securities. First, the aggregation through time for an individual security is described. The idea of a cumulative abnormal return (CAR) is fundamental to understand a multiple period event window. CARi(τ1, τ2) is the sample CAR from event date τ1to τ2 within
the event window where T1 < τ1 ≤ 𝜏𝜏2 ≤ T2 . The CARi(τ1, τ2) is the sum of the included
abnormal returns from τ1to τ2 for stock i,
CARi(τ1, τ2) = � ARi,τ τ2
τ=τ1
(5)
with asymptotic (as L1 increases) variance:
σi2(τ1, τ2)= (τ2− τ1+ 1)σε2i (6)
This sample estimator of the variance can be used when L1 is large. When the L1 is
small this variance needs to be adjusted for the effects of the estimation error in the normal model parameters, involving the second part of Equation 4 and a further related adjustment for the serial covariance of the abnormal return (MacKinlay, 1997). The CAR is normally distributed under the null hypothesis i.e., CARi(τ1, τ2) ~ N[0, σi2(τ1, τ2)].
To check whether there is on average an abnormal return on one of the event dates, we need to aggregate the abnormal returns across securities for that date and divide the
aggregated abnormal returns by the number of securities in the sample. In order to do so, the abnormal returns of individual securities, calculated with Equation 3, are aggregated and divided by the number of securities in the sample to get the sample average abnormal return.
AR ����τ= 1 N � ARi,τ N i=1 (7) 13
with VAR(AR����τ)= 1
N2� σε2i
N i=1
Equation 7 gives the sample average abnormal return (AR����τ) on day τ in the event window, with τ = T1, . . . , T2 and N is the number of securities in the sample. It is assumed
that there is not any overlap in the event windows of the securities in the sample. Under the distributional assumptions and in the absence of overlap in the event windows, the abnormal returns and the average abnormal returns will be independent across securities so that the covariance terms are zero for the variance estimators (MacKinlay, 1997).
In order to draw overall conclusions for the total event observations across securities and across time, the sample average abnormal returns are aggregated over the event period to get the sample average cumulative return. Equation 8 represents the formula to calculate the sample average cumulative return.
CAR
������(τ1, τ2) = � AR����τ 𝜏𝜏2
𝜏𝜏=𝜏𝜏1
(8)
The CAR������(τ1, τ2) is defined as the sample average cumulative return (CAR������) and set from day τ1to τ2 within the event window with T1 ≤ τ1 ≤ 𝜏𝜏2 ≤ T2 . The 𝐶𝐶𝐶𝐶𝐶𝐶������ from τ1to τ2 is
the sum of the sample average abnormal returns. The variance of the CAR������ is given by equation 9.
VAR�CAR������(τ1, τ2)� =N12� σi2 N i=1
(𝜏𝜏1, 𝜏𝜏2) (9)
with σi2(τ1, τ2)= (τ2− τ1+ 1)σε2i
The assumption that there is no overlap of events is used to set the covariance terms to zero for the variance estimators, similar as for the VAR(AR����τ). Under the null hypothesis the 𝐶𝐶𝐶𝐶𝐶𝐶
������ will be jointly normally distributed with a zero conditional mean and conditional variance of VAR�CAR������(τ1, τ2)�, CAR������(τ1, τ2) ~ N�0, VAR�CAR������(τ1, τ2)��.
The null hypothesis, that the sample average cumulative return is zero, could be tested using a standard procedure Equation 10 (MacKinlay, 1997).
θ1 = 𝐶𝐶𝐶𝐶𝐶𝐶������(𝜏𝜏1, 𝜏𝜏2)
VAR�𝐶𝐶𝐶𝐶𝐶𝐶������(𝜏𝜏1, 𝜏𝜏2)� 1 2
~ N(0,1) (10)
The null distribution of θ1 is standard normal. For a two-sided test with a significant level α, the null hypothesis will be rejected if θ1 is in the critical region. The formula to calculate the critical values is stated in Equation 11,
θ1 < 𝑐𝑐 �α2� or θ1 > 𝑐𝑐 �1 −α2� (11)
where c(x) = ф-1(x). ф( ּ◌) is the standard normal cumulative distribution function, the
term α is the level of significance. The critical values for a 10% significance level (α=0.10) are -1.645 and 1.645, for a 5% significance level (α=0.05) the critical values are -1.96 and 1.96 and for a 1% significance level (α=0.01) the critical values are -2.575 and 2.575, respectively.
4.2 Description of event
In our study, we will examine two types of events, since it is not a priori clear whether we should view at the announcement date or the effective inclusion date as the real event. We define our first event window as three days before the announcement date (AD-3) until the three days after the announcement date (AD+3) and our second event as the three days before the effective date (ED-3) until the three days after the effective date (ED+3). The announcement date is around 15 trading days before the effective date, depending on fast entries or exits and trading period. As suggested by MacKinlay (1997), the estimation window is set to 120 trading days prior to the event (AD-124). The event period itself is not included to prevent the event from influencing the normal performance model parameter estimates.
4.3 Hypothesis
The stocks quoted on the EURO STOXX 50 index are from very large and already well-known companies. We expect, therefore, that the stocks have more close substitutes and an
inclusion effect caused by a downward sloping demand curves for stocks will not be observed.
All information about the selection procedures for index inclusions on the EURO STOXX 50 index is publically available. The stocks who wait for inclusion are already traded on the large National Stock Exchanges and only become included if they satisfy a high degree of liquidity. For these reasons it seems illogical to expect positive abnormal returns caused by the information and liquidity hypotheses.
If it is profitable to trade on inclusions, we expect that awareness for the inclusion effect must have grown over time. This should have drawn more investors and diminished the magnitude of the abnormal returns. The markets should have become more efficient, for that reason it is possible that we do not observe any abnormal returns driven by the inclusion effect.
If the inclusion effect follows the increase in index funds, we do expect an inclusion effect. The inclusion effect must then be caused due price pressure from index funds. The EURO STOXX 50 index is tracked by a large number of index funds and as stated in the index guide, the index is used as underlying for a wide range of investment products including Exchange Traded Funds (EFT).
5. Data Description
The information about the EURO STOXX 50 index is provided by STOXX Ltd. (2015). The EURO STOXX 50 index was introduced in February 1998. The index is derived from the EURO STOXX index and is the leading Blue-chip index for the Eurozone serving as underlying for a wide range of investment products. It provides a representation of the performance from 50 supersector leaders in terms of free-float market capitalization in 12 Eurozone countries. The free-float market capitalization is the amount of shares of stocks of a firms total market capitalization that is available for trading. Amongst the 12 Eurozone countries are: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal and Spain.
The selection criteria are as follows: All the current EURO STOXX 50 stocks are added to a selection list. Also added are the largest stocks from each of the 19 EURO STOXX Supersector indices until they cover approximately 60% of the free-float market cap of the EURO STOXX TMI Supersector index. The 40 securities with the largest free-float market cap are selected from the list, the other 10 stocks are selected from the remaining current stocks on the list between 41 and 60. If the total number of stocks added is still below the required 50, then index is completed with the stocks having the highest free-float market cap. The index gets annually reviewed in September. The review will be based on the selection list created at the last trading day of August. The selection lists are updated monthly.
Additions could also follow from the fast entry rule. A stock is added to the index following the fast entry rule if it qualifies for the latest selection list, created end of February, May, August or November, and rank in the lower buffer of stocks between 1-25. The change will be announced on the first trading day of the month after the markets close, the smallest stock in the EURO STOXX 50 index will be replaced by the new stock. In opposite to the fast entry rule, there is also the Fast exit rule. A stock is deleted from the index following the fast exit rule if it ranks 75 or below on the monthly selection and on the selection list of the
previous month. The change will also be announced on the first trading day of the month after the markets close, the fast exist stock will be replaced by the highest-ranked non-component on the monthly selection list and the replacement will be effective the next trading day.
To carry out the event study a list of firm names, announcement and effective dates of stock inclusions for the EURO STOXX 50 index, from September 1999 until September
2014, is put together from press releases of STOXX Ltd. The daily stock prices of the firms that were included are retrieved from DataStream.
The benchmark used for calculating the expected returns with the market model is the EURO STOXX index. This index is a broad index representing companies with a large, mid and small capitalization of the same Eurozone countries as the EURO STOXX 50. Because this index represents a large group of different sized firms in the Eurozone it is considered as an excellent index to track the performance of the Eurozone and the market segments. For this reason the EURO STOXX index is chosen as the market index. The daily returns of the EURO STOXX index are also retrieved from DataStream.
The number of observations on inclusions is 32, from these 32 observations fourteen had to be excluded, remaining 17 observations for the sample.
First, seven firms that were included on September 20, 1999, are excluded to prevent a clustering bias. At that date the index underwent a major revision. Second, one firm is
excluded because the event windows overlap for this firm. Third, three companies are excluded because a lack of data. Finally, the other firms were excluded because there inclusions where the result of corporate events, as mergers and acquisitions. These events could lead to abnormal returns caused by other effects than purely driven by the inclsuion effect. Other studies mentioned earlier, e.g. Harris and Gurel (1986), Schleifer (1986), Pruitt and Wei (1989) etc., use the same adjustment method.
6. Results
In this chapter, we will describe the results of our event study, whether an inclusion effect is present in the Euro Stoxx 50 index. In section 6.1 we will present the results of the event study, in section 6.2 we give an analysis of our results.
6.1 Empirical results
As explained in section 4.3 we investigated two types of events. The results are obtained by following the standard event study methodology from section 4.1. Equation 10 from section 4.1 is used to test if the results are significant different from zero. We will first present the results of the first event window and later we will discuss the results from the second event window.
The first event window was set as three days before the announcement date (AD-3) until three days after the announcement date (AD+3). Table 1 gives the results of the average abnormal returns and the average cumulative abnormal returns for the first event window.
Event
day 𝐀𝐀𝐀𝐀����𝛕𝛕 std. Dev. 𝛉𝛉𝟏𝟏 𝐂𝐂𝐀𝐀𝐀𝐀������(𝛕𝛕𝟏𝟏, 𝛕𝛕𝟐𝟐) std. Dev. 𝛉𝛉𝟏𝟏
AD-3 -0.079% 0.007 -0.113 -0.079% 0.007 -0.113 AD-2 0.117% 0.007 0.167 0.038% 0.010 0.038 AD-1 0.622% 0.007 0.889 0.659% 0.012 0.544 AD -0.583% 0.007 -0.834 0.076% 0.014 0.054 AD+1 0.049% 0.007 0.069 0.125% 0.016 0.080 AD+2 -0.229% 0.007 -0.327 -0.104% 0.017 -0.061 AD+3 0.292% 0.007 0.417 0.188% 0.019 0.101
Table 1: Average and average cumulative abnormal returns with standard deviations and test statistic θ1from AD-3 until AD+3.
Table 1 shows that there are no significant average abnormal returns nor average cumulative abnormal returns observed for the first event window, AD-3 until AD+3.
The second event window was set three days before the effective date (ED-3) until three days after the effective date (ED+3).
Event
date 𝐀𝐀𝐀𝐀����𝛕𝛕 std. Dev. 𝛉𝛉𝟏𝟏 𝐂𝐂𝐀𝐀𝐀𝐀������(𝛕𝛕𝟏𝟏, 𝛕𝛕𝟐𝟐) std. Dev. 𝛉𝛉𝟏𝟏
ED-3 -0.241% 0.007 -0.344 -0.241% 0.007 -0.344 ED-2 1.206%* 0.007 1.724 0.965% 0.010 0.975 ED-1 -0.278% 0.007 -0.398 0.687% 0.012 0.567 ED -0.220% 0.007 -0.315 0.466% 0.014 0.333 ED+1 -0.399% 0.007 -0.570 0.068% 0.016 0.043 ED+2 -0.076% 0.007 -0.108 -0.008% 0.017 -0.004 ED+3 -0.489% 0.007 -0.699 -0.497% 0.019 -0.268
Table 2: Average and average cumulative abnormal returns with standard deviations and test statistic θ1from ED-3 until ED+3. * significantly different from zero at 10%.
In Table 2 are the results presented for the second event window. We observe two days prior to the effective date a 10% significant positive average abnormal return of 1.206%. We do not observe other average abnormal returns or average cumulative abnormal returns.
6.3 Analysis of results
In contrast to previous studies, we do not find positive abnormal returns around the
announcement date of a stock inclusion. An explanation for this result could be found by the suggestion of Baneish and Whaley (1996). If index funds appear to wait with rebalancing their portfolios until the effective date and that an increase after an announcement date is partially caused by arbitrage traders, it could mean by growth of arbitrage traders the
abnormal returns diminished. The majority of the studies investigated stock returns from the late seventies until the first half of the nineties. Awareness of the positive price reactions of stocks to news that they will become included in an index must have grown since then. This must have attracted more arbitrage investors and diminished the magnitude of the abnormal returns.
Results obtained from Table 2 are set out in Figure 2. We have to be cautious drawing conclusions on this data, because only the average abnormal return at date ED-2 is on a 10% level significant from zero. The results in Figure 2 indicate that there could be a price pressure effect on ED-2, caused by index funds. After ED-2 the index funds are rebalanced, the price
pressure is over and the positive average cumulative abnormal return decreases to zero over the next four days. Because the positive abnormal return at ED-2 does not seem to be persistent it is unlikely that the DS, information or liquidity hypotheses explain it. The price reversal is consistent with the PP hypothesis, but since other than on ED-2 the abnormal returns are not significant different from zero our data cannot strongly support or contradict the PP hypothesis.
Figure 2: Average and average cumulative abnormal returns from ED-3 until ED+3.
Taken together, it is difficult to conclude whether an inclusion effect is present in the EURO STOXX 50 index. What is interesting is that we do observe a significant positive average abnormal return of 1.206% two days prior to the effective date. Nonetheless, the lack of significant observations makes it unclear to find a significant relationship between this abnormal return and the inclusion effect.
Concluding, we can explain the lack of significant observations by the fact that the awareness for the inclusion effect has risen. If it is profitable to trade on index inclusions, it should have drawn more investors over time. The investors try to buy the stock before the other investors, which drives prices over a longer time period ahead of the announcement and/or effective date. Furthermore, index funds should have become aware of the premium they pay as a result of waiting until the announcement or effective date. They too should initiate their buying earlier before the announcement or effective date and spread their buying over a longer period, which also diminishes the abnormal returns. Moreover, we can explain the lack of significant abnormal returns by the limited data sample we have.
7 Conclusions
In this study we examined whether an inclusion effect is present in the EURO STOXX 50 index. The objective of this study was to examine if there are abnormal returns present around the announcement and/or effective date of a stock inclusion in the EURO STOXX 50 index driven by the inclusion effect. An event study approach was used with data from September 1999 until September 2014.
Previous studies suggest that the inclusion effect is caused by the demand of index funds, who aim to replicate the movements of an index by holding the securities that are included in an index. We argued that an inclusion effect in the EURO STOXX 50 index could be present since, stated by STOXX Ltd. (2015), the EURO STOXX 50 index is the leading blue-chip index for the Eurozone and heavily tracked by index funds.
In contrast to previous studies, we observe only two days prior to the effective date a 1.206% positive abnormal at a 10% significance level. This could be evidence of price pressures, unfortunately the lack of other significant observations makes it unclear to state that this abnormal return is purely driven by the index effect. The lack of significant abnormal returns could support our idea that the awareness for the inclusion effect has risen, attracting more investors which in turn diminishes the abnormal returns. Our findings could not be seen as real disproof for the existence of an inclusion effect in the EURO STOXX 50 index. Despite its real exploratory nature, this study offers more insight in the existence of inclusion effects and adds contradicting findings to a growing body of literature.
This study was limited by the size of our data sample, which could have influenced the statistical significance of our results. We had to exclude almost half of our original sample as a result of corporate events, lack of data and to prevent clustering problems. Unfortunately, it was not able to improve the analysis since there are not more inclusions in the EURO STOXX 50 index during our period of interest.
We did not address the issue whether an inclusion effect is present for exclusions from the EURO STOXX 50 index because due a lack of a enough clean observations. Exclusions are often caused by corporate events. Previous studies experience the same problem for exclusions. We recommend further research on index exclusions. We also recommend further research on if our findings are consistent for other indices. At last we recommend further research whether the inclusion effect is diminishing as capital markets become more efficient or that the inclusion effect follows the increase in index funds.
References
Amihud, Y., & Mendelson, H. (1986). Asset pricing and the bid-ask spread. Journal of
financial economics, 17, 223-249.
Beneish, M. D., & Whaley, R. E. (1996). An Anatomy of the "S&P Game": The Effects of Changing the Rules. The Journal of Finance, 51(5), 1909-1930.
Brown, S. J., & Warner, J.B. (1985). Using daily stock returns: The case of event studies.
Journal of Financial Economics, 14(1), 3-31.
Denis, D. K., McConnell, J. J., Ovtichinnikov, A. V., & Yu, Y. (2003). S&P 500 Index Additions and Earnings Expectations. The Journal of Finance, 58(5), 1821-1840. Harris, L., & Gurel, E. (1986). Price and volume effects associated with changes in
the S&P 500: New evidence for the existence of price pressures. Journal of Finance, 41(4), 815-829.
Lynch, A. W., & Mendenhall, R. R. (1997). New Evidence on Stock Price Effects Associated with Changes in the S&P 500 Index. The Journal of Business, 70(3), 351-383.
MacKinlay, A. C. (1997). Event Studies in Economics and Finance. Journal of Economic
Literature, 35(1), 13-39.
Pruitt, S. W., & Wei, K. C. J. (1989). Institutional Ownership and Changes in the S&P 500. The Journal of Finance, 44(2), 509-513.
Shleifer, A. (1986). Do demand curves for stock slope down? The Journal of Finance, 41(3), 579-590.
STOXX Ltd.. (2015). STOXX Index Methodology Guide. Retrieved from http://www.stoxx.com/download/indices/rulebooks/stoxx_indexguide.pdf
