Industry-Based Alternative Equity Indices

The authors examined five alternative equity indices (AEIs) in the United States using industries instead of individual stocks as building blocks to form portfolios and compared their performance with that of the capitalization-weighted equity benchmark for the period 1964–2011. The five AEIs had, ex post, lower risk and better returns than the cap-weighted benchmark. Net risk-adjusted returns of three AEIs were significantly positive when controlling for four risk factors.The authors examined five U.S. alternative equity indices (AEIs) using industries instead of individual stocks as building blocks to form portfolios and compared their performance with that of a cap-weighted equity benchmark. They chose industries instead of stocks for three reasons. First, industries are natural candidates to cluster stock constituents and to overcome the “curse of dimensionality” and the “error maximization” problem. Reducing the size of the covariance matrix mitigates these problems. Second, industries are important economic drivers that explain a significant part of the cross-sectional dispersion in active returns. The authors found that for the S&P 500 Index, the average cross-sectional dispersion in gross active returns from industry bets represents half of the dispersion in active returns from security bets. Third, although some stock constituent–based AEIs may outperform industry-based AEIs before any transaction costs, the latter make more sense in terms of capacity, liquidity, transparency, simplicity, and resultant transaction costs. Consequently, the outperformance of most constituent-based AEIs is largely driven by exposure to the small-cap factor. Using industries (capitalization-weighting scheme within the industry) instead of individual stocks as building blocks for portfolio formation allowed the authors to overcome this drawback.The five AEIs use alternative industry-based weighting schemes designed to improve on capitalization weightings by reducing risk without sacrificing return: (1) equally weighted (EW), (2) equally weighted among low-beta industries (EWLB), (3) equally weighted risk contribution (EWR), (4) minimum variance (MV), and (5) maximum diversification (MD). The authors used the U.S. industry total returns from Fama and French (Standard Industrial Classification, or SIC, and the broad universe of stocks) and focused on the period 1964–2011. The other contribution of this article is the detailed examination of the comparative performance of the five AEIs using the same dataset (the U.S. stock market from 1964 to 2011) and the same methodologies (shrinkage estimator for risk assessment and four risk factors for performance analysis). The authors also examined the robustness of these AEIs when using more granular building blocks, analyzing an out-of-sample period from 1931 to 1964, and considering different stock universes (the broad market versus the S&P 500) and different industry classifications (SIC versus the Global Industry Classification Standard, or GICS). Few papers have included this variety of in-depth comparative analyses and robustness tests.Our study yielded five main conclusions. First, each of the five alternative weighting schemes, compared with the cap-weighted scheme, had a lower risk ex post. This result is not surprising because all the AEIs are designed to, ex ante, reduce concentration, reduce systematic risk exposure, offer a better balance of portfolio risk contribution, reduce the absolute risk, or improve diversification. Each AEI except MV ranked first ex post against the performance metric derived from its respective stated objective. Interestingly, even though all five alternative weighting schemes had lower risk, they also had better returns. There is no theory that predicts, ex ante, that any of the AEIs will be more efficient than other portfolios. Second, the authors found that industry-based AEIs, like constituent-based AEIs, have a significant value tilt. However, unlike constituent-based AEIs, industry-based AEIs have a significant tilt toward large size and are consequently more scalable. Third, the equally weighted, equally weighted among low-beta industries, and equally weighted risk contribution AEIs displayed a statistically significant risk-adjusted outperformance (alphas from the Fama–French three-factor pricing model augmented by the Carhart momentum risk factor) net of costs. Fourth, these results are robust; they are not sensitive to the industrial granularity used (30 industries versus 10), the period examined (1931–1964 versus 1964–2011), the industry classification (SIC versus GICS), or the universe of stocks used (the S&P 500 versus the broad universe of stocks). Finally, the authors constrained the active risk of the five AEIs to mitigate any important deviation from cap weights, notably during industry bubbles, and they levered them up to match, ex ante, the risk of the cap-weighted benchmark. All five risk-constrained, levered AEIs significantly outperformed the cap-weighted benchmark, and except for MV, their risk-adjusted returns were significant and positive even after rebalancing costs.