The MSCI-Canada index rebalancing and excess comovement

Abstract

Major changes to the MSCI Canada Standard Country index were announced and implemented in May 2000. This rebalancing involved the addition of some 17 and deletion of 13 stocks and had the net effect of increasing the market capitalization by US$50 billions. We investigate the associated changes in stock return comovement around this event on the Toronto Stock Exchange, the third largest North American exchange. We find that the average beta of the added stocks increases by as much as a factor of 1.6 while the average R ² increases by up to 5%. Robustness tests indicate the results are not driven by nonsynchronous trading.

Acknowledgements

We are grateful to Chris Brooks, the Editor Mark Taylor and an anonymous referee for helpful comments. Kougoulis is grateful for financial support provided by the Greek State Scholarship Foundation (IKY).

Notes

¹ It might also be the case that the irrational trades occur randomly and, by cancelling each other out, their net effect on prices is zero.

² Harris and Gurel (1986), Shleifer (1986) and Lynch and Mendenhall (1997) establish significant price effects for S&P 500 inclusions. Wurgler and Zhuravskaya (2002) support this view by providing evidence that demand curves for stocks slope downwards which implies limits to arbitrage. See Mase (2006) and Mazouz and Saadouni (2007) for studies of the price effects of FTSE 100 index changes.

³ See McMillan 2003 2004 for further evidence of behavioural influences in the UK stock market.

⁴ Greenwood (2007) extends existing comovement studies in a novel direction. He documents a strong positive relation between overweighting of a stock and its comovement with other stocks in the Nikkei index. Such an approach is beyond the scope of this study.

⁵ The Toronto Stock Exchange (TSX) 300 was the broad-based Canadian national index at that time. This was subsequently replaced on 1 May 2002 by the S&P/TSX Composite Index which is a list of the largest TSX companies by market capitalization.

⁶ Index tracking is a low cost way of finding a proxy for the market portfolio and, under an efficient market framework, is a rational investment strategy. On the other hand, in a market with frictions and limits to arbitrage, index trackers who move funds in and out of the index can be seen as acting irrationally as long as they invest irrespective of the fundamental values of the underlying assets.

⁷ In December 2000, MSCI announced that it would adjust all of its equity indices for free float and increase the target market representation of its Standard Index Series from 60% of total market capitalization to 85% of free float-adjusted market capitalization. The combined changes were to be implemented in two separate phases. The first phase was to be implemented as of the close of 30 November 2001, and the second phase was as of the close of 31 May 2002. The phased implementation of the Enhanced Methodology in the Standard Index Series is designed to spread out the turnover in this series and its potential market impact and to allow ample time for investors to adapt to the changes. Another important feature of MSCI's implementation plan is the calculation, beginning on 31 May 2001 of the MSCI Provisional Index series based upon MSCI's Enhanced Methodology.

⁸ In checking for an equally weighted remainder index, we constructed the weights given by the market values Datasteam provides and not the original MSCI weights.

⁹ Only 12 out of the 13 deleted firms were considered in the analysis since the merger of Newbridge Networks Corporation on 25 May 2000 resulted in insufficient data for the post-event period.

¹⁰ In particular, the representation target increased from 60 to 85% of industry market capitalization in the Enhanced MSCI index methodology.

¹¹ Greenwood and Sosner (2002) and Coakley et al. (2004) also use the SUR procedure to deal with cross-sectional dependence while BSW (2005) use simulation methods to calculate appropriate SEs.

¹² The VARHAC procedure suggested by Den Haan and Levin (1997) has been applied in estimating C semi-parametrically.

¹³ While the z-statistics are corrected for heteroskedasticity, they are not corrected for cross-sectional dependence. This may imply that the z-statistics in Table 2 are biased upwards.

¹⁴ Note that in our regressions, returns are regressed on an equally weighted index of the stocks remaining in the MSCI Canada index rather than the market index.

¹⁵ It makes sense to check for nonsynchronous trading effects only for the beta changes of the added stocks since the corresponding figures for the deleted securities are lower in magnitude and statistically insignificant.