Which is the better equity investment strategy: (1) a consistent portfolio tilt toward one or a few factors widely accepted, based on linear regression backtests, as effective in selecting stocks with above-average performance (smart beta); or, (2) a more complex strategy that seeks to identify stocks with above-average performance via potentially dynamic relationships with a set of many proprietary factors? In their September 2014 paper entitled “Investing in a Multidimensional Market”, Bruce Jacobs and Kenneth Levy argue for the latter. Referring to recent research finding that many factors are highly significant stock return predictors in multivariate regression tests, they conclude that:
- A portfolio with exposures to a large number of factors diversifies opportunities and risks compared to a smart beta strategy based on one or a few factors, thereby offering more consistent performance.
- Some factor-return relationships vary predictably with economic or market conditions. A dynamic portfolio factor tilt, unlike the consistent tilt in a smart beta portfolio, can exploit such variations.
- Rigid, infrequent rebalancing practices limit the profit opportunities of smart beta strategies. Strategies that can shift as unexpected factor opportunities arise (as with earnings surprises and return reversals) have greater potential.
- Smart beta strategies generally rely on widely used factors susceptible to crowding, which can cause factor overvaluation and subsequent correction.
- By combining widely used factors and known rebalancing rules, smart beta strategies are vulnerable to front running and associated degradation of smart beta portfolio performance.
- Smart beta strategies shift decisions about factor selection and timing from investment managers to less astute investors, thereby probably degrading investment decisions.
In summary, the authors conclude that professionally managed strategies diversifying across many proprietary factors and allowing continuous adjustments to factor exposures have more to offer investors than smart beta strategies.
Cautions regarding conclusions include:
- The examples used appear to be equity-centric, not clearly considering diversification across asset classes.
- The authors do not support assertions with any new quantitative analyses. In other words, the paper is more a statement of beliefs than a test of beliefs.
- The advocated multidimensional strategies are complex and somewhat secretive. Hence:
- Assessing the investment process is difficult for investors.
- The process of constructing the strategies has considerable freedom for data snooping (incorporation of luck in backtests).
- The strategies may inherently drive high portfolio turnover and consequently bear high trading frictions.
- “Improving Established Multi-factor Stock-picking Models Is Hard” summarizes findings that success in multivariate regression tests may not translate well into multi-factor portfolio formation. Results indicate that researchers should go further than regression analysis in deciding how many and which factors to include in a stock screen.
