Nonfundamentals and value returns

Abstract

The study examines additions to and deletions from the Russell Value Index and the Russell 2000 Growth Index. The study documents stronger comovement in value reconstitutions relative to growth reconstitutions. This result is consistent with the hypothesis that nonfundamental comovement is related to the common factor in value stock returns. The mechanism of causality is difficult to determine, however; trade demand, firm characteristics and information diffusion are presented as potential sources that could explain why comovement of small to mid-cap value stocks is greater than the comovement of small to mid-cap growth stocks.

Acknowledgements

The authors thank Ji-Chai Lin, Mark Taylor (the editor), an anonymous referee and the participants at the University of Oregon research seminar for their helpful suggestions.

Notes

¹ Asgharian and Hansson (2005), Chan and Chen (1991), Chen and Zhang (1998), Dichev (1998) and Griffin and Lemmon (2002) address different aspects of the common problem in relating the firm's fundamental value with factor selection.

² In an international setting, Cheung and Lai (1999) and Dickinson (2000) found that international return comovement is related to macroeconomic comovement.

³ Zhang (2005) shows that costly reversibility and countercyclical price of risk has potential in explaining value returns in the neoclassical framework with rational expectations.

⁴ For example, Lee et al . (1991) argue that irrational sentiment simultaneously moves the prices of closed-end funds and small stocks while the net asset values of the funds and the fundamental values of small stocks have no direct relationship. Froot and Dabora (1999) examine Royal Dutch/Shell twin shares that have claims to identical cash flows and are traded in US and UK, respectively. They find that Royal Dutch comoves with the S&P 500 index whereas Shell comoves with the FTSE index. Rashes (2001) finds a significant correlation in returns, volumes and volatility at short frequencies between MCI Communications (under ticker symbol MCIC) and Massmutual Corporate Investor (under ticker symbol MCI) while the former which is telecommunication stock and the latter which is a closed-end fund share no fundamental commonality.

⁵ Examples include Barberis et al . (1998), Daniel et al . (2001), De Long et al . (1990) and Hong and Stein (1999).

⁶ Nevertheless, this reconstitution effect is also consistent with a friction-based view that the S&P 500 member stocks, as a group, are quicker in incorporating news about aggregate cash flows because these lager stocks are associated with lower transaction costs and higher institutional ownership.

⁷ This group of stocks satisfies the three requirements suggested by BSW. First, the Russell 2000 Growth and Value are widely used by institutional investors as the benchmarks for small growth and small value styles. Their popularity makes them a natural category and/or a preferred habitat. Second, there exist identifiable changes in group membership over time in the Russell 2000 Growth and Value, and the information is widely available. Third, no more and no less, the stated goal of the Frank Russell Company is to make the indices representative of the US market in terms of capitalization and style.

⁸ In the following discussion, the study uses the term ‘reconstitution effect’ to describe the changes in the slopes of the regressions on event returns, as documented by BSW.

⁹ Denis et al . (2003) find that S&P 500 index additions are associated with increases in earnings.

¹⁰ Most deletions from S&P 500 are due to poor firm performance. As a result, the majority of the sample events in BSW are index additions.

¹¹ The S&P 500 has a relatively long history, is widely used to represent the capitalization weighted return of large stocks, and 97% of US managers and pension plan sponsors use the index for benchmarking, performance evaluation, or passive management. As of April 2003, more than $1 trillion was indexed to the S&P 500. This is a nontrivial sum considering the total capitalization of the 500 stocks was approximately $8.3 trillion as of April 2003.

¹² Madhavan (2003) documents abnormal returns for the annual reconstitution of the Russell 2000.

¹³ The Frank Russell Company is one of the earliest advocates for recognizing the importance of diversifying across value and growth investment styles. Since then, these indices have become widely used by institutional investors for benchmarking small value and small growth styles. Russell 2000 Value and Growth exchange traded funds (ETFs) were introduced on 24 July 2000, and Russell 2000 futures contracts have traded on the Chicago Mercantile Exchange (CME) since 4 February 1993. As of 30 June 2001, over $16 billion was passively invested in the Russell 2000.

¹⁴ The Frank Russell Company had rebalanced its indices on a quarterly or semiannual basis.

¹⁵ Currently, the ranking date is 31 May.

¹⁶ For 2002, the index goes into effect on 1 July.

¹⁷ For 2002, final membership list is posted on 9 July.

¹⁸ The Russell 2000 consists of generally younger and smaller firms than the S&P 500. As young firms progress along their business life cycle, this causes their return styles to change over time. As a result, one would expect weaker effects from our data in longer horizons. For this reason, the study operationally defines the pre-event and post-event windows to be 12-month.

¹⁹ In the pre-Great Depression era, value investing was the dominant investing philosophy. During this period many security analysts believed that current (not future) earning power is the basis for stock valuation. Graham and Dodd (1934) echoed this tradition and argued that ‘investment values can be related only to demonstrated performance.’ One surmises that the value style is more popular than the growth style because of its rich history.

²⁰ BSW demonstrate that the bivariate specification is more powerful than the univariate specification of Vijh (1994). The study confirms this. While not reported, the empirical results based on the univariate specification are weaker, but largely consistent with the bivariate results.

²¹ Industry matching is not always available for small event stocks. As a result, the matching sizes are smaller. The sample sizes of matching portfolios for value additions, growth additions, value deletions and growth deletions are 270, 567, 327 and 489, respectively.

²² Increasing turnover is observed for 130 value additions, 281 growth additions, 163 value deletions, and 175 growth deletions increase their turnover, while decreasing turnover is observed for 150 value additions, 296 growth additions, 180 value deletions and 331 growth deletions.