Recently we observe a shift towards factor investing, in which institutional investors strategically allocate their long-term investment portfolios to factor premiums. Well-known factors are Value, Momentum, Size and Low-Risk. Securities with these characteristics are known to deliver higher risk-adjusted returns, compared with investing in the broad market index. Moreover, these factors can explain a large part of the alpha of active equity portfolio managers. By explicitly allocating towards these factor premiums on a strategic level, investors can generate superior Sharpe ratios.
So far, most attention for factor investing has focused on equities, but the concepts and premiums carry over to other asset classes like credits. In this note, we address a specific factor premium in the credit market, namely the size premium. This premium relates to the effect that small caps tend to outperform large caps. In an earlier note we demonstrated the existence and robustness of the premium within the High Yield market. In this note, we expand upon the previous note by including results for Investment Grade and by testing whether the size premium is still present after taking transaction costs into account.
First we will show that small caps historically had a better Sharpe ratio than large caps in both the Investment Grade and the High Yield market and that this result is robust to numerous checks. Secondly, we will try to explain why the size premium exists. We find that credit and liquidity risk do not solely explain the size premium. Instead, we provide empirical evidence for an alternative explanation: small caps are overlooked by fundamental managers and ETFs. Given these findings, we do not expect the size premium to disappear in the near future. Thirdly, we show that the size factor also exhibits attractive Sharpe ratios in comparison with the market index after transaction costs. This suggests that investors should definitely consider the size premium when constructing their credit portfolios. For fundamental managers, implementation might be difficult due to the need of a very large team of analysts. At Robeco, we circumvent this by using a quantitative credit selection model. This allows us to reap the size premium in a cost-efficient way.
To measure the size premium, we use data on the constituents of the Barclays US Corporate Investment Grade Index (USIG) and US High Yield Index (USHY). The data start in January 1994 and end in November 2012. Each month we compute the size of each company as its total notional amount of debt in the Barclays indexes. Then we rank the companies on this size measure to create five quintile portfolios: the 20% companies with the smallest total notional amount are in Q1, the next 20% in Q2, etcetera. For these portfolios, we calculate return4 statistics and other characteristics like leverage.
Figure 1 shows the Sharpe ratios of the five size buckets for USIG and USHY. The Sharpe ratios decrease almost monotonically with size: small caps clearly outperformed large caps on a risk-adjusted basis. This improvement is mainly driven by higher returns, since the volatilities are similar (not shown here). Figure 2 shows the cumulative outperformance of small caps versus the market through time. One could view this as the gain of allocating towards the size factor instead of following the index. The annualized outperformance is 0.77% for USIG, and 1.88% for USHY. Note that these returns exclude transaction costs. In a later section we will show that the size premium can still be captured when transaction costs are taken into account.
The results are robust to numerous checks. We find that the size premium is not driven by sector allocations and exists for both companies with publicly listed equity and privately owned companies. Furthermore, the premium exists within each rating and maturity group, suggesting that credit risk is not the sole explanation. Finally, we find that the size premium also exists in the market for euro-denominated bonds, just like for US dollar-denominated bonds.
The question is why the size premium exists, and whether it will persist in the future. We argue that small caps are ignored by fundamental managers. Covering all index constituents would require a very large team of analysts, as the Barclays US Investment Grade Index contains about 600 companies, and the High Yield index contains about 1000. Figure 3 shows that by covering the 20% largest companies, over 60% of the total notional amount of the index is covered. On the other hand, the 40% smallest companies constitute at most 10% of the index. It is natural for managers to have their credit analysts primarily analyze the large caps to cover as much of the index market capitalization as possible. This keeps the costs at a minimum.
Furthermore, small caps are also ignored by ETFs. With their increasing popularity, ETFs have become a major factor in the corporate bond markets. Most of the ETFs follow a liquid benchmark, usually consisting of a limited number of large cap names. For example, the two most popular high yield ETFs, the iShares iBoxx $ High Yield Corporate Bond ETF (HYG) and SPDR Barclays High Yield Bond ETF (JNK), follow indices containing 300 to 400 companies, versus 1000 for the Barclays Capital Corporate High Yield index. This means that they ignore the small and mid-caps.
To illustrate that small caps are indeed overlooked, we obtain the number of sell-side analysts covering a name.5
Figure 4 shows the average number of analysts per size bucket: the analyst coverage shows a clear bias towards large caps. The number of analysts following the largest Investment Grade companies is more than threefold the number of analysts following the small companies. For High Yield, the smallest companies are hardly followed at all. We therefore conclude that in both Investment Grade and High Yield small caps receive relatively little attention.
Another reason for the existence of the size premium could be that the higher returns are a compensation for risk, e.g. market, credit or liquidity risk. We already mentioned that the return volatility of small caps is similar to that of large caps. In Figure 5, we look at credit risk by plotting the average leverage per size bucket, measured as the ratio of liabilities to assets.6 We find that small caps typically have lower leverage than large caps, so that the size premium is not obtained by a tilt towards riskier companies. Moreover, the size premium exists in every rating category.
Liquidity risk, on the other hand, might partly explain the size premium, as smaller bonds tend to be less liquid in general. In Figure 6, we first sort bonds on their age (a well-known proxy for liquidity7) into three buckets, and within each age bucket, we create three size buckets. Within each age bucket, we observe that small caps have the highest Sharpe ratio, both for Investment grade and for High Yield. Therefore, liquidity risk is not the only explanation, as even among the youngest bonds, which are typically the most liquid bonds, the size effect persists.
Investors may ask whether the size premium can actually be captured in practice, as transaction costs impact the final performance. Especially for the size premium, which purposely allocates to less liquid (smaller) bonds, this is a relevant question. To answer this question, we conduct a simulation study, in which we construct a buy-and-hold small cap portfolio consisting of 100 companies. This corresponds to about 20% of the universe, like in the previous results.8 If a bond drops out of the portfolio, because it matures, is called, defaults or no longer meets the index requirements, it is replaced by a new bond. We estimate transaction costs using the Barclays transaction cost model.9 Via this procedure, we obtain a net Sharpe ratio for the small cap portfolio. We follow the same procedure for the index. The results, shown in Figure 7 and Figure 8, indicate that small caps have higher transaction costs per bond on average, which results in a larger decrease from gross to net Sharpe ratio. However, the small cap portfolio still has a superior net Sharpe ratio in comparison with the index.
These results indicate that the size premium can be captured after transaction costs, even with a basic buy-and-hold strategy. In practice, one could improve the portfolio further, e.g. by applying issuer selection and controlling for exposures to other factors, such as Low-Risk, Value and Momentum, to prevent going against these other well-known factor premiums.
The final issue we address in this note is how portfolio managers can exploit the size premium in their portfolios. As already mentioned, due to the large number of names, covering the small and mid-cap market segment in a fundamental credit selection process would require a prohibitively large number of analysts, which is cost inefficient. However, a quantitative credit selection model does not have capacity constraints. On the contrary, the larger the universe, the more stable a model’s selection skills will be. Within Robeco’s High Yield Bonds fund, we use our quantitative corporate bond selection model.10 This model applies factors such as Value and Momentum to the public small and mid-cap segment of the High Yield market. The remaining part of the High Yield universe, i.e. the large names and the private names, are covered by a team of experienced credit analysts, who conduct thorough fundamental research on a company’s ability to meet its obligations. The reason why we apply our quantitative model to public (instead of all) small and mid-caps is that our model uses balance sheet and equity data; both require that a company is publicly listed. Reassuringly, as mentioned earlier, the size effect also holds when the universe is restricted to public companies. Using the quantitative selection model, we gain access to the size effect in an efficient way. Furthermore, by constructing a broadly diversified portfolio, with relatively small positions in individual names, we largely diversify away unwanted risks in the portfolio: smaller positions mean less default risk and also less liquidity risk, because it is easier to buy and sell smaller amounts than large amounts.
In this note we show that the size premium is strong, persistent over time, present after accounting for transaction costs and robust to a multitude of characteristics like sector, rating, maturity, age and leverage. We find that the premium is not driven by credit risk, as small caps actually tend to have lower leverage than large caps. Liquidity risk partly explains the premium, since smaller issues tend to be less liquid than larger ones. We feel that the main explanation for the existence of the size premium is the tendency of fundamental managers and ETFs to primarily cover large caps and to ignore mid and small caps. Given the large weight of a relatively small number of companies in the index, this is likely to continue. Therefore, we do not expect the size premium to disappear in the near future. To cover the public small caps within Robeco’s High Yield Bonds fund, we apply our quantitative credit selection model. By doing this, we efficiently obtain exposure to the size premium, while our team of experienced analysts can continue to focus on the large caps.
1 See the paper “Evaluation of Active Management of the Norwegian Government Pension Fund – Global” by Ang et al. (2009) or the Robeco research note “Efficient Factor-Investing Strategies” by Blitz, Huij, Lansdorp & van Vliet (2013).
2 See the paper “On Persistence in Mutual Fund Performance” by Carhart (1997).
3 See our Robeco research note “High Yield small and mid-caps: an attractive market segment” by Houweling, Haesen & Bus (2011).
4 We use excess returns over duration-neutral Treasury bonds to prevent the interest component of the return from interfering with our results.
5 I/B/E/S provides these data. The data are only available for companies with publicly listed equity. I/B/E/S provides the number of sell-side analysts, as data on buy-side analysts is not available.
6 Balance sheet data like liabilities and assets are not available for privately owned companies in our data set. The results shown are thus for the public companies only.
7 See for example the paper “Comparing Possible Proxies of Corporate Bond Liquidity” by Houweling et al. (2005).
8 To reflect the illiquidity of this market segment, we assume a 20% probability of actually being able to buy a particular bond. We try to buy the smallest companies first, but if not bought, we try to buy the second smallest company, etcetera. Hence, the gross premium captured possibly differs from the previous back-test results, which bought all small caps.
9 See “Liquidity Cost Scores for US Credit Bonds” by Barclays (2009).
10 See our earlier Robeco research note “Managing High Yield public small caps with Robeco’s corporate bond selection model COALA” by Bus, Haesen & Houweling (2010) for more information.