Exploring Index Effects - Tamas Toth
- 1. Exploring Index Effects A look at index and index constituent returns around rebalancing dates Tamas Toth, CFA
- 2. 1 Table of Contents Introduction.............................................................................................................................................2 Financial Environment.........................................................................................................................2 Principal-agent problem.......................................................................................................................4 Index reconstitution.............................................................................................................................4 Impact on market efficiency.................................................................................................................5 Hypotheses for Index Effects...............................................................................................................5 Data and methodology.............................................................................................................................7 Evaluation of the Index Effect in Canada.............................................................................................7 Data Sources .......................................................................................................................................8 Investigation of the existence of index effect in the SP/TSX Composite ...........................................8 Methodology.......................................................................................................................................8 S&P/TSX Composite...........................................................................................................................9 Eligibility ........................................................................................................................................9 Index and Stock Price Data................................................................................................................10 Cumulative Abnormal Returns...........................................................................................................10 Additions.......................................................................................................................................10 Deletions .......................................................................................................................................12 Findings ................................................................................................................................................13 Extensions.............................................................................................................................................14 Appendix I – S&P/TSX Composite Rebalancing Methodology..............................................................15 Works Cited..........................................................................................................................................16
- 3. 2 Introduction Financial Environment The current dynamics in financial markets made passive investment strategies very attractive for a broad range of investors. Passive investment strategies provide exposure to a specific market, sector or style by attempting to replicate the returns of a market index. One of the main drivers of growth in financial assets is the aging population in developed countries. Figure 1 shows that Canada’s population of 65 years and over is expected to grow from 14.4 percent in 2011 to 22.8 percent in 2031. Figure 1 - Canada's Aging Population Source: http://www4.hrsdc.gc.ca/.3ndic.1t.4r@-eng.jsp?iid=33 As people get older, a larger portion of their total wealth is stored in financial capital (investments) than in human capital (future income). Figure 2 graphically displays this trend of shifting source of wealth.
- 4. 3 Figure 2- Source of Wealth in Life's Stages While this trend supports the growth of all financial assets in the market, there has been a trend that tilted the growth of passive investments from those of active ones. Passive investments differ from active investments in that there is no stock-picking or timing ability is expected by the investor. The rationale for investing in passive funds over actively manages ones is that study after study shows that only very few active managers can outperform their benchmarks (Walker, 2013). Further, since no active management is required, fees in passive funds tend to be much lower. Competition and economies of scale have been driving fees lower every year. While Vanguard has earned the reputation of offering the lowest fees, Charles Schwab grabbed headlines in September of 2013 by cutting its fees on its US Broad Market ETF from 6 basis points to 4 bps (Walker, 2013). Figure 3 - Share of Equity Index Mutual Funds over Time
- 5. 4 Source: http://www.icifactbook.org/pdf/2013_factbook.pdf Figure 3 shows the rise of the share of equity index mutual funds in the United States (Investment Company Institute, 2013). The trend has been study in growth of passive funds from 1998 to 2007, but a significant jump can be observed in 2008. The financial crisis of 2008 prompted many investors to re-evaluate their investment portfolios and the returns they were receiving (or not receiving) for the fees they were paying for active management. Confirming the trend in Figure 3, Lipper also reported that since 2004, assets held in passive funds – excluding ETFs – as a portion of all equity products increased from 12 percent to 21 percent (Walker, 2013). This trend resulted in the doubling of assets held in passive investment funds over the past four years, ending 2013 with $1.3 trillion of US investors’ assets (Walker, 2013). In 2013 specifically a record $115 billion has been piled into index mutual funds, which is nearly 4 times the $38.3 billion invested in actively managed funds (Condon, 2014). Principal-agent problem The financial crisis not only reminded investors of the high costs they were paying for no apparent superior performance, but also of the lack of downside protection active investors are supposed to provide their investors. There have been multiple studies exploring the principal-agent problem in the money management framework. One way investors can overcome this issue is to constrain managerial risk by tying the performance of the fund to a benchmark. The effectiveness of the manager can then be evaluated based on the “absolute difference between each month’s index return and the fund return summed over the time frame in question” (Chen, Noronha, & Singal, 2006), which is referred to as the fund’s tracking error. By monitoring the tracking error of the fund, the investor can evaluate how well the manager is performing his mandate. Index reconstitution When an investor selects a benchmark they would like to gain exposure to through a passive investment, they consider the market or style the benchmark is set out to represent. For example, the “S&P/TSX indices are designed to be both representative of the Canadian equity market and its sectors, and liquid, to support investment products such as index mutual funds, exchange traded funds, index portfolios, and index futures and options.” (McGraw Hill Financial, 2013). When the securities included in the index are no longer the best representation of the market or style, the index has to be reconstituted. Index providers vary in the ways they maintain their indexes. S&P uses a committee to select the best fitting securities for their equity indexes, while Russell uses a straight forward methodology to mechanically select its constituents. These reconstitution times require passive investment managers to update their portfolios so that their returns reflect the returns of the updated index, i.e. minimize the tracking error. For institutional clients, such as large pension fund sponsors, tracking error in excess of 0.10 percent a year is unacceptable (Chen, Noronha, & Singal, 2006). Research has shown that that the trading volume on the effective date of reconstitution is several times the normal daily volume, indicating that mangers rebalance their portfolios on these dates (Honghui, Noronha, & Singal, 2004).
- 6. 5 Impact on market efficiency In the semi-strong form of efficient capital markets, current asset prices should reflect all publicly available information (Fama, Efficient capital markets: A review of theory and empirical work, 1970) (Fama, Efficient capital markets: II, 1991). Of course, in order for asset prices to reflect all publicly available information, the public has to react by trading on new information. It is therefore obvious to see how markets would not follow the semi-strong form of efficient capital markets in the extreme where all investors are passive investors. Even if this extreme is unlikely to materialize, large allocations to passive investments may create inefficiencies in the markets. A case in point for these inefficiencies is the constraint of the tracking error imposed on investment managers. Since index providers announce the changes to the indexes in advance, but passive managers will only rebalance their portfolios on the effective date, arbitrageurs can front-run those portfolio managers. It has been shown that the inclusion and exclusion of a stock in a benchmark causes abnormal returns to the involved stocks. For example, the Russell 1000 generate cumulative excess returns of 10.9% from 2 days before the rebalance dates while stocks deleted from Russell 2000 Growth Index suffer cumulative loss of 6.6% (Chen H.-L. , 2006). This paper will explore hypotheses for these abnormal returns in a later section. The combination of large amounts of assets in passive investments and the constraints of the tracking error introduce inefficiencies, which provide an opportunity for arbitrageurs to transfer wealth from passive investors to themselves. This wealth transfer has been estimated to cost passive investors investing in the S&P 500 and Russell 2000 Index between $1.0 billion and $2.1 billion a year (Chen, Noronha, & Singal, 2006). This paper aims to explore whether this trading opportunity still exists and the proposed hypotheses for its existence. Further, it will explore whether this trading strategy can be automated and turned into an investment vehicle that could be held in combination with passive investments. If such an opportunity exits, the current financial environment would predict a favorable reaction by the growing number of passive investors since it would eliminate the above mentioned wealth transfer. Pursuing a business based on this premise would then be of interest to providers of passive investment strategies. Hypotheses for Index Effects Abnormal returns attributable to index reconstitution have come to the attention of the investment community, both in practical sense by exploiting these inefficiencies and academically by exploring the reasons for their existence. The following six hypotheses have been proposed in various forms in different studies, but have been summarized by (Doeswijk, 2005) as follows: 1. Price Pressure Hypothesis In the price pressure hypothesis, and index revision causes a temporary price pressure, because traders who are willing to provide liquidity to passive fund managers require a premium for their services (i.e. liquidity). (Madhavan, 2003) describes the premium required by market makers with the following expression:
- 7. 6 Where DesInv is the market makers desired inventory level, Inv is the market makers current inventory and lambda is the coefficient that’s inversely related to market liquidity. Ultimately, large deviations from target inventory levels – driven by portfolio rebalancing activities by passive managers – will command a price that is higher or lower than their intrinsic values. 2. Imperfect substitutes hypothesis (downward sloping demand curves) As in any microeconomic model, a permanent increase (decrease) in demand for a stock, driven by its inclusion (exclusion) from an index, will shift the demand curve for that stock to the right in an environment where there are no close substitutes for that stock. Since tracking error is a major factor in selecting stocks for the portfolio, perfect substitutes are rare. Figure 4 - Shifting a downward sloping demand curve Source: http://mbaecon.wikispaces.com/Demand+(include+what+is+a+demand+curve+and+what+shifts+a+demand+curve) The change in price will permanently change as long as investors are required to hold the stock in their index replicating portfolios. In a joint test, (Chen H.-L. , 2006) finds that In the joint test, the price pressure hypothesis (1.) and the liquidity hypothesis (3.) explain the marginal index effect at most by 0.12% and 3.05%, respectively, while the imperfect substitutes hypothesis explains it at least by 9.21%. 3. Liquidity hypothesis The liquidity hypothesis postulates that index revisions affect the liquidity of a stock, which in turns reduces the liquidity premium required by investors. This line of reasoning would result in a one-time permanent jump in share price. As mentioned in (2.), (Chen H.-L. , 2006) also found support for this. 4. Information hypothesis
- 8. 7 The information hypothesis slightly reverses the causality relationship. This explanation states that the decision to include a stock in a benchmark reflects its improved future potentials, and therefore should command a higher price multiple. Consequently, while the stock remains in the index, an investor subscribing to this hypothesis would continue to attach a higher price multiple and the effect would then be permanent. 5. Attention hypothesis It is believed that an index reconstitution will receive media attention and that this attention will ultimately lead to an enlargement of the investor base for the mentioned company. Similarly to the downward sloping demand curve, the additional demand would lead to a permanently higher price. 6. Selection hypothesis Selection hypothesis argues against the previously mentioned ideas, and states that excess returns due to index reconstitutions are not robust and that the market displays a remarkable degree of liquidity and substitutability for replacement stocks. The original motivation for this paper is mostly in line with (1.). Since these hypotheses don’t appear to be mutually exclusive, the trading strategy focused on exploiting the index effect should focus on trading around the effective date to maximize the likeliness of profit – by capturing both temporary and permanent price changes. Data and methodology Evaluation of the Index Effect in Canada A large number of studies focusing on the abnormal returns due to index reconstitution have used popular indexes such as the S&P 500, Russel 2000 and the MSCI Country Indexes. (Chen, Noronha, & Singal, 2006) define the requirements for an index to be a good target for arbitrageurs as follows: • Index changes are transparent and known sufficiently in advance of the effective date • The index is heavily used by passive index funds • Fund managers are constrained to trade on the effective day by tracking error or other performance metrics While the constituents for the S&P 500 are decided by a committee – and therefore a less transparent than -, it is the index to which the largest amount of assets are indexed to in passive strategies. On the other hand the Russell 2000 is a prime candidate for index effect trading strategies. The constituents are decided mainly on market capitalization. The predictability of changes in the Russell 2000 are straight forward enough that brokerage houses now send reports to their clients on the expected changes to the index before the reconstitution date. These predictions have been shown to be 90-95% accurate.
- 9. 8 Accordingly, economically significant trading opportunities can be expected to be traded away by a large investor base. In order for this research to be valuable, an index that can reasonably be expected to offer index effect arbitrage opportunities should be used. The topic of discussion going forward is the S&P/TSX Composite index for the following reasons: • Index construction methodology is reasonably transparent (in line with S&P 500). See Appendix I • The index is broadly accepted as the benchmark representing the Canadian Equity Markets • Its subindexes are used for specific passive fund mandates Data Sources Investigation of the existence of index effect in the SP/TSX Composite The reliability of the findings in this study is highly dependent on the quality of data used. For this reason, the following data is sourced from Compustat: • Historical index constituent data • Historical share prices and dividends • Share adjustment factors to adjust for changes in shares outstanding The historical levels of the index are sourced from Yahoo! Finance. The time period for the study is between January 1 2010 and February 14 2014. Methodology It is common practice to use event studies in the investigation of index effects. This paper will comply with this practice specifically as outlined in (Glenn V. Henderson, 1990). The outline of this process is as follows: 1. Define the date upon which the market would have received the news (i.e. addition or deletion from index) 2. Characterize the returns on the individual companies in the absence of this news 3. Measure the difference between observed returns and “no-news” returns for each firm – the abnormal returns 4. Aggregate the abnormal returns across firms and across time 5. Statistically test the aggregated returns to determine whether the abnormal returns are significant a, if so, for how long. Henderson describes various options for each step outlined above. Specifically the determination of abnormal returns is of interest in this study. The no-news price in an event will be the daily index return and the excess return will be defined as the daily return on the stock (including dividends) minus the daily return for the index. The excess returns are then called market-adjusted returns. While there are more sophisticated techniques employed in practice to determine excess returns, (Glenn V. Henderson, 1990) has provided evidence that the power of this technique is comparable to that of the regression-based methodologies. This technique should therefore be sufficient for our preliminary analysis.
- 10. 9 Once abnormal returns are calculated, they will be averaged by other observations that are observed on the same day relative to the event. To understand the pre-event cumulative abnormal returns, a time series of cumulative returns can be constructed, which is called cumulative abnormal returns (CAR): Where T is the time of the event and t is all calendar days 40 days before the event. S&P/TSX Composite Eligibility The following criteria have to be met for a stock to be considered for inclusion in the S&P/TSX Composite (McGraw Hill Financial, 2013). • Market Capitalization o Based on the volume weighted average price (VWAP) over the last three trading days of the month-end prior to the Quarterly Review, the security must represent a minimum weight of 0.05% of the index, after including the Quoted Market Value (QMV) of that security in the total float capitalization of the index. In the event that any Index Security has a weight of more than 10% at any month-end, the minimum weights for the purpose of inclusion will be based on the S&P/TSX Capped Composite. o The security must have a minimum VWAP of C$ 1 over the past three months and over the last three trading days of the month-end prior to the Quarterly Review. • Liquidity o Total number of shares traded at Canadian trading venues in the previous 12 months divided by float adjusted shares outstanding at the end of the period has to be greater than 0.5. • Domicile o Issuers must be incorporated under Canadian federal, provincial or territorial jurisdictions and listed on the TSX. • Timing o Regular additions and deletions are made during the Quarterly Reviews. Securities under consideration for addition to or deletion from the index will be assessed by the Index Committee on the basis of the six-month data ending the month prior to the Quarterly Review. o The Quarterly Review months are March, June, September and December. o All additions, deletions and share changes are effective after the close of trading on the third Friday of the quarterly month. o Whenever possible, announcements of additions or deletions of stocks or other index adjustments are made five trading days before the adjustments are implemented. In those cases when it is not possible to trade a stock five days after an announcement,
- 11. 10 the announcement period may be shortened. However, the implementation of an index adjustment will never be earlier than the market close of the day following the announcement. Index and Stock Price Data The constituent data for the S&P/TSX Composite was acquired from Compustat for the period between January 1, 2010 and February 14, 2014. During this period 79 stocks were removed and 109 stocks were added. There were 241 constituents in the index at the end of the period. 105 of the 109 additions and 41 of the 79 deletions took place during the Quarterly Review. The following analysis will only refer to these additions and deletions when analyzing index effects. Daily index levels and constituent stock prices were also collected for the same period. Cumulative Abnormal Returns Cumulative abnormal returns (CAR) are examined for the stocks that were added or removed during the quarterly review. A time period of 40 calendar days are used as the CAR window. Day 0 is the effective day of the rebalance and day -39 is 39 calendar days before the effective date. Additions Abnormal returns are calculated for each of the to-be-included stocks based on their relative returns to the S&P/TSX Composite index. The significance of the abnormal returns is calculated using the Student’s t- distribution. Based on the sample data the returns on the following days are significant: -10, -20, -21 and - 25. The abnormal returns on day -10 are expected, because this is the fifth trading day before the index changes become effective and TSX aims to announce the changes to the index on this date. The abnormal returns are shown on Figure 5 – Additions Daily AR. Abnormal returns between day -25 and -20 are high on average and would likely be significant with a larger sample. An optimal trading strategy should therefore initiate the long positions 25 calendar days before the index rebalance.
- 12. 11 Figure 5 – Additions Daily AR Figure 6 – Additions Average CAR 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 9.0% 10.0% -39 -35 -33 -31 -27 -25 -21 -19 -17 -13 -11 -7 -5 -3
- 13. 12 Figure 7 - Additions Average CAR per year Deletions Similar to the additions, the abnormal returns and the t-statistics for each of the calendar days is shown on the figure below. An inconsistency in the data is present, because the additions are shown as effective as of the Monday following the index change announcements, while deletions are effective as of the Sunday following the announcement. The following days are significant in negative abnormal returns: -6, -10, - 26, -30, -34, -37. Figure 8 - Deletions Daily AR -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 -39 -35 -33 -31 -27 -25 -21 -19 -17 -13 -11 -7 -5 -3 2010 2011 2012 2013 -5 -4 -3 -2 -1 0 1 2 3 -5.0% -4.0% -3.0% -2.0% -1.0% 0.0% 1.0% 2.0% 3.0% -39 -37 -33 -31 -27 -25 -23 -19 -17 -13 -11 -9 -5 -3 t average
- 14. 13 Figure 9 - Deletions Average CAR Figure 10 - Deletions Average CAR per year Findings The study finds that the index effects due to the inclusion or exclusion of a stock in a major index generates abnormal returns around the date of index reconstitution. Recent data suggest that cumulative abnormal returns are positive for newly added constituents and negative for dropped constituents for the 40 calendar days leading up to the index reconstitution date. These effects have been shown consistently for index changes for the years 2010 – 2013. -0.25 -0.2 -0.15 -0.1 -0.05 0 -39 -37 -33 -31 -27 -25 -23 -19 -17 -13 -11 -9 -5 -3 CAR CAR -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 -39 -37 -33 -31 -27 -25 -23 -19 -17 -13 -11 -9 -5 -3 2011 2012 2013
- 15. 14 Extensions As outlined in the introduction of the paper, passive investors could benefit from an investment vehicle that produces the abnormal returns due to index reconstitutions. An extension of this study would explore the ex-ante trading strategy, which identifies the stocks which are added and removed from the index, before the announcement is made. The index construction methodology clearly outlines the eligibility of a stock for the index and the data is mainly publicly available. The most relevant data points are price, liquidity and relative market capitalization. Adjusting for float introduces some complexities.
- 16. 15 Appendix I – S&P/TSX Composite Rebalancing Methodology Rebalancing The index is reviewed quarterly and all Index Securities that, in the opinion of the Index Committee, do not meet the Buffer Rules are removed. Added securities, if any, are selected using the Eligibility Criteria. 1. Securities under consideration for addition to or deletion from the index are assessed by the Index Committee on the basis of the six-month data ending the month prior to the Quarterly Review. The Quarterly Review months are March, June, September and December. Publicly available information up to and including the month end preceding the Quarterly Review month, which pertains to shares outstanding (rounded to the nearest thousand), is considered in the Quarterly Review. Investable Weight Factor (IWF) updates are only made annually at the September Quarterly Review. All additions, deletions and share changes are effective after the close of trading on the third Friday of the quarterly month. For details regarding Investable Weight Factors (IWF) and Float Adjustment please refer to S&P Dow Jones Indices’ Float Adjustment Methodology available at www.spdji.com. Securities that the Index Committee determines meet the Eligibility Criteria are added to the index after the close on the third Friday of the Quarterly Review month. 2. Securities removed from the index as a result of the Quarterly Review are not eligible for re-inclusion for a period of 12 full calendar months following removal. The Index Committee may, nevertheless, add a security at an earlier date if, in the opinion of the Index Committee, the issuer’s business has been substantially restructured. 3. Index Securities that the Index Committee determines fail to meet Buffer Rules are removed from the index after the close on the third Friday of the Quarterly Review month. 4. A press release is issued to the market within 10 business days after the start of the rebalancing month announcing the additions and deletions, as well as the new float shares for index calculation. Frequency. Rebalancings occur quarterly. Intra-quarter changes are made on an as needed basis. Changes occur in response to corporate actions and market developments. The target announcement period is two- to-five business days, but exceptions may apply due to unexpected corporate activity.
- 17. 16 Works Cited Chakrabarti, R., Huang, W., Jayaraman, N., & Lee, J. (2005). Price and volume effects of changes. Journal of Banking & Finance. Chen, H., Noronha, G., & Singal, V. (2006). Index Changes and Losses to Index Fund Investors. Finanacial Analysts Journal. Chen, H.-L. (2006). On Russell index reconstitution. Rev Quant Finan Acc. Cohen, J. E. (2002). Tough game of profiting from index rebalancings. Pensions & Investments. Condon, C. (2014, January 30). Index Funds Hit Record Year as U.S. Savers Prefer Passive. Retrieved from bloomberg.com: http://www.bloomberg.com/news/2014-01-30/index-funds-hit-record-year- as-u-s-savers-prefer-passive.html Doeswijk, R. Q. (2005). The index revision party. International Review of Financial Analysis. Fama, E. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 383-417. Fama, E. (1991). Efficient capital markets: II. Journal of Finance, 1575-1617. Green, T., & Jame, R. (2011). Strategic trading by index funds and liquidity provision around the S&P 500 index additions. Journal of Financial Markets. Honghui, C., Noronha, G., & Singal, V. (2004). S&P 500 Indexers, Tracking Error, and Liquidity: A Complex Answer to Profiting. Journal of Portfolio Management, 1901-1929. Investment Company Institute. (2013). 2013 Investment Company Factbook. Retrieved from icifactbook.org: http://www.icifactbook.org/ Madhavan, A. (2003). The Russell Reconstitution Effect. Financial Analysts Journal. McGraw Hill Financial. (2013, 2014). S&P/TSX Canadian Indices: Methodology. Retrieved from S&P Indicies: http://ca.spindices.com/indices/equity/sp-tsx-composite-index Okada, K., Isagawa, N., & Fujiwara, K. (2006). Addition to the Nikkei 225 Index and Japanese market response: Temporary demand effect of index arbitrageurs. Pacific-Basin Finance Journal. Walker, O. (2013, June 23). Hello passive, goodbye active: fund investors make the switch. Retrieved from ft.com: http://www.ft.com/cms/s/0/3aa2cd62-d4da-11e2-9302-00144feab7de.html