Although allocation by asset class is still the dominant approach, factor investing is gaining momentum. A recent survey showed that 30% of institutional investors now favor factor investing. ^[1]

Little surprise, then, that factor investing has attracted significant attention from both the academic and the practitioner communities. These communities have documented at least 300 factors that are believed to outperform, and newly discovered factors are regularly added to the list.

It is impossible that every one of these factors can contribute meaningfully to returns, the authors argue, so they set out to find out which factors actually do.

The proliferation of documented sources of outperformance has been labelled by some academics as a “zoo of factors” and may be the result of massive data mining. That is, many researchers mine data to identify sources of excess returns and then develop an argument around their findings, claiming discovery of a new factor. The authors note that if each of the thousands of professors, graduate students, and quantitative analysts were to backtest a single strategy every year, some would inevitably discover what appears to be a winning strategy. But the vast majority of these supposedly winning strategies would be flukes.

Recent research exploring the persistence of factor premiums has found that many cannot be replicated or that premiums have been exaggerated.

The authors devised a four-part approach to try to identify “true” factor premiums and expose “false” ones.

The first part was to identify factors that are well established in the academic literature—a factor strategy that does not attract follow-on research is probably unreliable, reasoned the authors. This led to the identification of six potentially reliable factors: illiquidity, low beta, value, momentum, size, and quality.

Second, these six factors were tested for robustness across many definitions. Using both qualitative and quantitative procedures, the authors applied a groundbreaking multi-dimensional analysis, which they believe is more informative than a purely statistical approach.

Third, factors were tested for robustness across geographies, thus providing out-of-sample verification for factor strategies that have typically been documented using US data only.

Fourth, the authors took trading costs into account. Although most academic research ignores implementation costs, these costs are a considerable performance drag for investors and should, therefore, be considered.

Most of the 300-odd documented factors are not covered extensively in the literature, suggesting that they have low reliability or replicability. Of the six factors that have wide coverage, the authors found that two popular ones lack robustness: Size and quality exhibit only weak premiums. In contrast, value, momentum, illiquidity, and low beta are more robust.

The authors’ findings on each factor can be summarised as follows:

•        Size. No definition of size produces small-cap portfolios that deliver significant risk-adjusted returns. Small-cap portfolios also exhibit a value bias: When the small-cap excess return was adjusted for the value effect, the size premium fell close to zero.

•        Quality. There were few signs of a quality premium across multiple definitions of quality.

•        Low beta. Low beta is a robust source of outperformance for investors, even if the tracking error is large. It does, however, have the potential for extended periods of underperformance.

•        Value. All definitions of value deliver significant risk-adjusted returns. The only outlier is the UK large-cap universe.

•        Momentum. Momentum is a source of outperformance and is most reliable in the small-cap universe.

•        Illiquidity. Whereas the illiquidity premium is strong in the United States, it is not outside the United States.

Unsurprisingly, trading costs were found to have a major effect on two of the six factor strategies: Momentum and illiquidity lose their attractiveness when trading costs are taken into account.

The authors found differences in the ability of factor strategies to offer downside protection. All strategies had their largest drawdown during either the global financial crisis or the recession of the early 1970s. Most drawdowns were similar in magnitude and more severe than the markets’ drawdowns. The exception is low-beta strategies, which experienced only 56% of the market loss. The authors also found that value and illiquidity tend to limit the downside in falling markets and fully capture the upside in rising markets.

First, some investors may be surprised that size and quality, two widely used factors, do not perform well. So, investors in factor-based strategies may wish to ask more questions of providers about how each strategy is expected to perform and why.

Second, certain factors seem to work well in some geographies but not in others. Investors should be wary of implementing global factor-based strategies when there is little evidence that the factor is robust across geographies.

Third, investors who prioritise avoiding volatility will be interested to see confirmation that low-beta strategies offer far greater downside protection than other factor-based strategies.

Fourth, investors should ask themselves whether a factor-based strategy is the best approach to capture premiums when trading costs are high. Active managers may have more flexibility to reduce implementation costs for both momentum and illiquid strategies.

Fifth, a key question for fund managers is how to combine factors in a portfolio. The simplest approach assigns equal weights to each selected factor, providing diversification and ease of governance. But some managers may favor combining negatively correlated factors to improve the portfolio’s risk–return characteristics.

Finally, it is worth highlighting that the increase in factor-based investing has given rise to an entire subset of the investment industry: smart beta. In light of this study, it will be interesting to see how smart-beta strategies are refined and, possibly, tilted towards those factors that demonstrably contribute to excess equity returns.