Long-short factor portfolios are constructed on the assumption that the two legs are complementary drivers of factor premiums. In our research paper,1 we critically assess this notion by breaking down the Fama–French style portfolios – low risk, momentum, value, profitability and investments (the latter two are often referred to as quality) – into their long and short legs. Although we find that factor premiums originate in both, it is evident that most of the added value tends to come from the long positions.
Moreover, our research shows that the long legs of factors offer more diversification than the short ones, while the performance of the shorts is generally subsumed by the longs. This outcome is particularly driven by a high common risk in short positions. This can be ascribed to the relatively high correlations between the shorts legs of various factors as well as between the shorts and longs of individual factors. As a result, we observed that the Sharpe ratios have been highest for the long legs of factors and lowest for the short legs.
The Sharpe ratios have been highest for the long legs of factors and lowest for the short legs’.
These results are confirmed in a series of robustness checks. They hold across large and small caps, as in both cases, the long positions contribute the most to factor premiums. They are robust over time and across international equity markets, and cannot be attributed to differences in tail risk. Thus, the long legs are crucial for understanding factor premiums, while no essential information is lost by dropping the short legs. All in all, the shorts offer essentially the same exposure as the longs, but with lower rewards and more risks.
Therefore, factor premiums can be more efficiently harvested by dropping short positions and focusing only on the long ones. Our findings do not even account for the substantially higher implementation costs borne by short positions compared to long ones. Additionally, short selling faces the impact of liquidity and capacity limits, as well as other risks, like the potential of unlimited losses, ‘short squeeze’ scenarios, counterparty risk and reputational risk.
Finally, our research also challenges recent claims that the low risk and value factors are subsumed by the new Fama-French factors, as this hypothesis holds for the short positions of these factors but breaks down for the long ones. To this end, we establish that the long legs of both the low risk and value factors offer distinct premiums that cannot be explained by the longs of other factors.