One of the best-known phenomena is an excess level of reward for having exposure to smaller companies in an equity portfolio, leading to substantially higher cumulative returns for small-cap stocks than for large caps over the past century. Still, this effect has also disappeared for substantial periods at times, so timing is also important, as the small-cap premium is highly cyclical and only positive only half of the time. The ability to time the factor is therefore a valuable source of alpha, and our research shows how this can be done systematically.
As mentioned above, the cumulative return of small-cap stocks has significantly exceeded that of large caps over the past 100 years. The expectation for small caps to outperform relative to large caps is often dubbed the “size premium.” A premium can arise for both rational economic reasons and behavioural considerations. Economic reasons for the premium include liquidity, less analyst coverage, which may increase the risk of mispricing, and non-inclusion in indices, leading to stocks being overlooked by passive investors. Behavioural finance also provides us with an explanation for the anomaly: it exists because investors prefer positively skewed assets, investments that provide a small chance of a large payoff. It is worth noting that all investors cannot simultaneously overweight small caps. While we cannot predict future returns, we can look at historical stock returns for evidence of a small-cap premium. A study by professors Eugene Fama and Kenneth French measuring the performance of U.S. small-cap stocks relative to U.S. large-cap stocks showed that over an 87-year period, between 1926 and 2013, small caps outperformed large caps by over 2% per year1.
However, the premium is positive in only a little more than 50% of calendar years, making small-cap outperformance in any given year more akin to the toss of a coin. More importantly, small caps have gone for long periods of time without evidencing a premium. For example, in the 32-year period between 1982 and 2013, following the publication of Rolf Banz’s research on the relationship between return and market value in 19812, the premium was virtually non-existent in the U.S. stock market.
This might stem from small stocks’ higher sensitivity to the cost of capital than large stocks. At the same time, the share prices of small stocks are mostly driven by bottom-up investors, who closely watch the underlying fundamentals of the companies involved. It therefore seems fair to assume that these bottom-up investors do not take a sufficiently external view when considering the financial conditions of the specific companies that they follow. Hence, we hypothesized that spread tightening/widening could help to predict the future level of the size premium. We tested this assumption using option-adjusted BBB credit spreads. Our research shows that timing the size factor could enhance risk-adjusted returns.
Timing the size factor based on credit spreads dramatically outperforms a static size premium strategy, both in nominal and risk-adjusted returns. Using a global universe of around 1,600 stocks, our switching strategy has returned an excess return of 3.3% per year, as opposed to 2.2% for a simple small-cap versus large-cap strategy over the sample period 1997-2018. The timing strategy also increases the information ratio from 0.54 to 0.76.
There is sound empirical evidence that favours small-cap over large-cap stocks on a long-term investment horizon. However, small-cap stocks have tended to be riskier than large caps, and the small-cap premium is highly volatile and therefore may not be positive even over long periods. We therefore advocate a more dynamic approach to the weighting of small-cap stocks in investment portfolios. This approach follows the economic rationale that the small-cap premium is related to risk aversion; hence, a strategy of overweighting small-cap stocks when credit spreads tighten and reducing positions when spreads widen is warranted. In this manner, the small-cap premium can be harvested systematically, as NN IP currently does in the AI equity range.
1https://www.sciencedirect.com/science/article/pii/S0304405X12000931