Why Index Reviews Affect Stock Returns

There exist several indices describing all the stock markets around the world. Most of them are value weighted, so they are basically a weighted average of the prices of the stocks included in it, with weights proportional to their capitalizations. The purpose of these indices is to measure the performance of the whole market in a country or in specific regions. Among the most popular indices there are the MSCI Indices. They are computed by MSCI-Barra which developed the so called MSCI Global Standard Indices. They are a series of indices representing the stock market of almost all countries.
In order to maintain the composition of indices close to the actual one so to reflect the real performance of the market, these indices are periodically reviewed. In particular, MSCI does quarterly reviews of the Global Standard Indices. During these reviews, many stocks are removed and many are added.

In this article we investigate the effect of these reviews on the performance of the stocks that are added or removed. More specifically, the data suggest that the impact of deletions is clearer and more regular while that of additions tend to be more unpredictable. Therefore, we decided to focus on the effect of index reviews on stocks that are removed from the index.

Beside measuring the performance of the market, most of these indices are used as a benchmark by passive investing funds all around the world. This is the main reason why index reviews can affect stock returns. In fact, passive funds have to follow very closely the composition of the index used as benchmark because if they want to replicate the return of the index they must hold the same stocks included in the index and with the same weights.

So, for example every time a stock is added to an index these passive funds must add this stock to their portfolio and every time a stock is removed they must liquidate their positions in that stock.

MSCI release the reviews of its Global Standard Indices every 15th of February, May, August and November. The changes that this announcement bring to the composition of indices become effective the first trading day of the following month.

Looking at the data, it is very interesting to observe that the performance of the stocks that are removed from an index tends to follow a specific pattern. In fact, these stocks often have negative returns after that the announcement of the change is released but when the change becomes effective on the first day of the following month, they tend to have very good returns. One possible explanation of this phenomenon is the following: since passive funds have to follow closely the composition of the index, on the date of announcement they start selling the stocks that are removed. This goes on during the two weeks between the announcement date and the date in which the changes become effective. On the first day of the month following the announcement, these funds must already have the new and valid composition of the index in their portfolio, consequently they stop selling the stocks. In fact, we often see that in the two trading days after that changes become effective most of the stocks that were deleted rebound.

This can be explained by the fact that since the sales of stocks from the passive funds stop in the first day of the month after the announcement, the supply of stocks becomes very small in comparison to what it used to be in the previous two weeks when the passive funds were liquidating their positions. Therefore, if nothing happens to the demand, the reduction of selling pressure in the market make the price of these stocks increase.

A simple trading strategy which can be used to exploit this phenomenon is the following: you buy the stocks that are announced to be removed from the MSCI indices on the last trading day of the announcement month and then you sell them in the first two or three days of the following month, after the changes has become effective. According to the logic we explained before this trading strategy is likely to get a profit.

An Event Study on Market Reviews

In this part we conduct an event study on index reviews using historical data of stock returns. An event study is a statistical method used to investigate the relationship between security prices and economic events (Dyckman et al., 1984). These kind of studies are usually used to test whether the stochastic behavior of stock prices is affected by the disclosure of firm-specific events such as a merger or an acquisition. Similarly, we use the same approach to test whether stock returns are affected by the fact of being removed or added in a MSCI Index.
Our event study is basically a statistical hypothesis test with the following null and alternative hypothesis:

• H0: being removed from a MSCI Index does not affect the return of the stock
• H1: being removed from a MSCI Index has a significant impact on the return of the stock

So, our aim is to find enough statistical evidence in the data, which allow us to reject, with a high level of confidence the null hypothesis that index reviews doesn’t affect stock returns.

Before proceeding with the test we made some choices regarding the data to use. At first, as we already mentioned in the first section, the impact on the stock return of a deletion appears to be clearer to identify and more regular than the corresponding effect of an addition. Consequently, we decided to focus only on deletions. Another aspect we needed to consider before starting the test was choosing the MSCI index we wanted to focus on. As far as we were able to see, the impact of index reviews on stock returns tends to be larger for smaller markets such as Greece, Netherlands and Brazil. This could be easily explained by the fact that in small market there is usually lower liquidity and consequently the level of market efficiency tends to be lower too. However, in order to have results from our test that are significant from a statistical point of view we need samples of historical returns with a quite large number of observations. As you can imagine, if you consider a fixed time interval, the number of stocks that have been removed from an index representing a big stock market as the US or UK one is much higher than the corresponding number of deletions in smaller markets. Therefore, we decided to analyze the deletions in MSCI USA Index. Furthermore, if we are able to find empirical evidence that index reviews affect stock returns even in a large and efficient stock market as the US one, then for smaller market the effect will be probably greater.

MSCI makes additions and deletions in its indices quarterly, they release the names of the companies added and removed in February, May, August and September and the announced changes become effective in the first day of the following month (i.e. the changes announced in February are effective from March 1st). So, we downloaded all the announcements released by MSCI from 2015 to 2018 and we looked for the deletions in the MSCI USA Index. In the period we considered there were 72 deletions in MSCI USA and we were able to get the time series of historical prices of 53 out of them (we used Yahoo Finance and Reuters DataStream). Then, from historical prices we computed daily log returns as follows:

$R_{i,t}=ln \left( \dfrac{S_{i,t}}{S_{i,t-1}}\right)$

where Si,t is the price of the i-th stock at time t

At this point, in order to identify the effect that index reviews have on the stock returns we computed, we need a measure of it and the measure we used is the so called abnormal return. It is used in most of event studies to understand the effect of an event on stock prices. Abnormal returns are the difference between the actual return of stocks and the expected return of them in the same period:

$Abnormal \> Return_{i,t}=R_{i,t}-E(R_{i,t})$

where Ri,t is the log return of the i-th stock from t-1 to t

As far as the expected return is concerned, you can compute it in many different ways. We decided to compute it using a simple market model. According to a market model the expected return of a stock is a linear function of the return of the market measured in the same time interval:

$E(R_{i,t})= \alpha_i + \beta_i R_{M,t}$

where α and β are the parameters of the model for the i-th stock and RM,t is the log return of the market between t-1 and t. Since we are considering the US stock market, as a proxy of the return of the market we used the log return of S&P500.

In order to compute the expected and the abnormal returns for the 53 stock included in our sample we considered two different time intervals for each one of them, the estimation period and the test period (see the image below). For every stock in the sample the estimation period includes the 100 trading days before the announcement of the deletion of that stock from the MSCI USA. We used this time interval to estimate the parameters of the market model that we need to compute the expected returns. In particular, we estimated the parameters by running a linear regression where we considered the daily log return of the stock as the dependent variable and the daily log return of S&P500 as the independent one. Similarly, for every stock in the sample, the test period includes the two last days of the month of the announcement (we call them day -2 and -1) when that stock is still in the index and the first two days of the following month (we call them day 1 and 2) when the stock is no longer in the index because the changes have become effective. We used this time interval to compute daily abnormal returns using the parameters we estimated in the estimation period.

According to the logic explained in the first section, the return of deleted stocks should increase in the first days after the change has become effective because the selling pressure of passive funds disappear. Therefore, if what we said it is true then the stocks in our sample should show positive abnormal returns after the stock has been actually removed from the index.

In other words, we want to verify whether the abnormal returns after the deletion has become effective are significantly positive. In particular, we chose a high confidence level of 99%. Hence, the hypothesis test we introduced at the beginning of this section can be written more formally as follows:

H0: Abnormal Return after the decision = 0

H1: Abnormal Return after the decision > 0

In order to do the test, for each of the four days in the test period, we computed the average of the abnormal returns of all the 53 stocks in our sample.

$Average \> Abnormal \> Return_{t}= \dfrac{1}{N} \sum_{i=1}^{N}Abnormal \> Return_{i,t}$

where N=53 is the number of stocks in our sample and Abnormal Returni,t is the abnormal return of the i-th stock in day t.

Finally, in order to understand whether the average abnormal returns we computed are statistically significant or simply come from the sampling variability of our estimate, we computed the t-ratio for each of the four day in the test period:

$t-Ratio=\dfrac{Average \> Abnormal \> Return_t}{SD \left( Abnormal \> Return _{i,t} \right)} \sqrt{N}$

where SD(Abnormal Returni,t) is the standard deviation of the abnormal returns of day t

The results are summarized in the table below:

As you can see in day 1, when stocks are actually removed from the index we have an average abnormal return of about 1% which is statistically significant at a confidence level of 99%. In fact, its t-ratio is 2.84 which, according to a t-distribution with 52 (N-1) degrees of freedom, is equivalent to a p-value of 0.32%.
Therefore, we proved that being removed from a MSCI index has an impact on the return of the stock even in a large stock market as the US one. In the case of USA, the effect is not big, indeed the average abnormal return of the first day is only about 1%. However, as we said before, the impact of deletions is probably bigger in smaller markets.