Impact of sector versus security choice on equity portfolios

Abstract

We measure the contribution of industry sector choice and individual stock selection to the performance of 3350 United States’ equity funds from 1980 to 2005. First, we demonstrate that sector choice makes a relatively greater contribution to portfolio variance, holding constant manager skill in identifying mispriced securities, and correcting for a bias in research methods previously applied to this issue. Second, using managers’ reported stock holdings, we estimate the actual contribution of industry sector versus security choice to portfolio returns. Active managers of funds with a preference for small stocks with a style preference – value or growth – generate abnormal returns above those achieved by less active managers, consistent with managers having style-specific investment skills. This relative performance is attributed to the incremental returns generated from security selection over sector choice. Security selection also explains significantly more variation in returns across funds. These results imply that active managers make relatively greater use of security selection in forming portfolios which differ from benchmark.

Keywords:

• portfolio choice
• asset allocation
• sector choice
• security selection
• top-down
• bottom-up

JEL Classification:

• G11
• G23

Notes

1 Cremers and Petajisto (2009) treat the increased diversification associated with stock-picking as a given, and use this view to motivate their analysis of active versus passive management.

2 While Kritzman and Page (2002, 2003) contrast the impact of asset allocation amongst asset classes versus security selection, rather than sector allocation versus security selection, the issues are the same.

3 Our decomposition of sector and security returns does not involve a cross-product term, ∑_i=1 ⁿ (w_i ^p − w_i ^b ) × (r_i ^p – r_i ^b ), facilitated by the use of portfolio sector weights (w_i ^p ) within the computation of security returns. Our security returns can therefore be interpreted as the impact on portfolio returns from security selection, conditional upon sector weights in the portfolio. An alternative benchmarking technique is to use historical fund weights as the benchmark, rather than market weights (Kacperczyk et al., 2005; Wermers, 2006). This is likely to understate the contribution of managers’ long-term strategic choices.

4 Fund size is not available for all funds so this is the only analysis performed on a value-weighted basis.