Institutional flows and equity style diversification

Abstract

This article examines the composition of style-diversified portfolios and the influence of institutional trading on style performance over the period 1979 to 2004. We employ a methodology to identify possible cointegrating relationships among four equity styles and to determine styles necessary to a well-diversified portfolio. Two seemingly dissimilar styles, large value and small growth, are cointegrated and hence, redundant diversifiers. We show an optimized three-style portfolio that omits one cointegrated style, improves performance and lowers market risk, demonstrating the importance of allocation to style diversification. We also find evidence the trading behaviour of institutional investors explains, in part, the relationship among and between the cointegrated and the independent styles.

Notes

¹Pension plans, in particular, have diversification requirements specified by the Employee Retirement Income Security Act of 1974 (ERISA).

²The largest 750 are large caps, stocks ranked 501–1000 comprise mid-caps and the stocks ranked 751–2500 constitute small caps. Stocks ranked lower than 2500 are assigned to the micro-cap segment that is not further styled.

³Stocks with characteristics not conforming to style criteria are eliminated from style membership. Stocks with less than 5 years history, high dividend payout, low growth, low price-to-book and low ROE are removed as potential growth style members. Those with high relative P/E and/or P/B and high relative dividend yield are eliminated from potential value membership (Brown and Mott, 1997). Detailed information on the construction of the Wilshire style indexes can be obtained from the Wilshire website (http://www.wilshire.com/Indexes/Wilshire).

⁴Stocks ranked in the middle 40th percentile range are categorized as style neutral.

⁵More information on the construction of the FF portfolios and factors can be obtained from the website of Professor Kenneth French (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html)

⁶The MMMF data are collected by the Investment Company Institute and provided to the Federal Reserve Board. Retail MMMFs are included in the M2 money measure; institutional MMMFs are included in M3. See http://www.stls.frb.org/fred/index.html for more information.

⁷Conversion to natural logarithms makes the first differences of index prices, the percentage change or return, conceptually more meaningful in the vector autoregressive representation than the absolute changes of index prices.

⁸Previous literature (e.g. Masih and Masih, 2001; Chen et al., 2002) provides consistent evidence that index price series are nonstationary.

⁹The ADF unit-root tests were conducted prior to cointegration rank tests. The results indicate that all included price series are nonstationary or I(1) over the sample period and thus, merit the cointegration rank tests. For brevity, the unit-root test results are not tabulated, but are available upon request.

¹⁰The cointegration results are dependent on lack of a structural break in the time series during the sample period. We perform the Zivot–Andrews (1992) unit-root test on each style benchmark price series and find no structural breaks. Further, cointegration results in two equal 1979–1991 and 1992–2004 inter-temporal periods are substantively identical to those in the full period.

¹¹To confirm, the performance results are not an artifact of a specific period, we also compute portfolio performance in 1979–1991 and 1992–2004 inter-temporal periods. Results, available from the authors upon request, support those of the full sample period.

¹²The portfolio assets in the B–S model are fixed, so that any new funds directed to in-favour styles are withdrawn from the poorly performing styles. Thus, the enhanced performance of in-favour styles comes at the detriment of the out-of-favour styles.

¹³The MMMF data are collected by the ICI and provided to the Federal Reserve Board. Retail MMMFs (deposits less than $100 000) are included in the M2 money measure; institutional MMMFs are included in M3. See http://www.stls.frb.org/fred/index.html for more information.

¹⁴Because the Wilshire style indices are employed in the cointegration tests, we create the independent portfolio from the Wilshire large growth and small value styles to standardize the data source across the portfolios.