Credit risk is more likely to do poorly when stocks do poorly. However, from what I see, credit risk does not become more sensitive to the stock market and stock-like during bear markets overall. It may fall to some degree with stocks, but not by more than it gains to some degree with stocks. Overall, credit risk is only somewhat correlated with equities. If credit risk is rewarded on average (positive excess returns, which a recent paper argues is true but some previous studies did not find evidence of), then given its properties it seems worth consideration in portfolio construction.
Update with a few more graphs and minor things of interest from live funds the last ten years available in two subsequent posts further down in the thread:
Opinions are divided on the role and suitability of corporate bonds in asset allocation. Some experts, such as David Swensen, Bill Bernstein, and Larry Swedroe recommend sticking to quality bonds and taking more risk elsewhere. On the other hand, one Jack Bogle himself has suggested that the broad bond index (Bloomberg Barclays Aggregate and variants thereof) and such funds following it such as Vanguard Total Bond Index contain too many government bonds, and he'd like to see more corporate bonds represented. Rick Ferri prefers supplementing total bond with both TIPS and junk bonds. Some fund companies via target date funds include overweights to credit risk, and popular robo advisor platforms like Betterment and Wealthfront do the same.
One of the main criticisms of credit risk from the skeptics is that credit risk shows up at the wrong time, when stocks are tanking, and corporate bonds (especially high yield, but even investment grade) start to act more stocklike when the going gets tough. Is this true? Let's find out.
Disclaimer: Any difference in fixed income, particularly between investment-grade choices and outside of the long end of the yield curve, is unlikely to make much of an impact in terms of overall portfolio behavior. Most allocations are dominated by equity performance unless they own very little in stocks, well under 30%, which almost nobody here does. Also, individual investors can frequently get better deals outside of the bond market in the form of FDIC or NCUA insured CDs.
Data and background
I use data from "The Credit Risk Premium" by Attakrit Asvanunt and Scott Richardson. The main observation motivating this paper is that many previous analyses of corporate and government bonds, including Fama and French back in 1993, examine bonds of equivalent maturity rather than equivalent duration (or really, option-adjusted duration), leading to a mis-measurement of credit risk. Corporate bonds of equivalent maturity will have higher yields and thus lower durations. The paper attempts to adjust older historical data to account for this, looking over the history of US investment-grade corporate bonds available, 1926-2014. They also examine shorter histories of high yield bonds, corporate bonds in other markets, and credit default swaps to evaluate credit risk in as many markets and scenarios as possible. The paper is worth reading in full, but to summarize, they do find a persistent credit risk premium and show that it can improve the efficiency of portfolios containing equity and term risk.
The main data, for investment-grade corporate bond returns in the US from 1926-2014, is available for download here.
They provide three monthly data series in the spreadsheet: CORP_XS, GOVT_XS, and SP500_XS. GOVT_XS is the difference between government bond and 30-day T-Bill returns, while SP500_XS is the return of the S&P Composite/S&P 500 over 30-day T-Bill returns. (The XS means excess, as in excess returns of the risk-free rate, in this context.) The data series of most interest here is CORP_XS, which is the excess return of corporate bonds over a duration-matched benchmark of government bonds. The data prior to 1988 is less reliable, reflecting some best guesses of duration with all the stipulations and procedures you can read for yourself in the paper. Additionally, they note that the nature of the indexes has changed over time, and the oldest data reflects mostly longer-term bonds of relatively high (AA/A) credit quality.
Does credit risk act more stock-like when stocks are down?
It is often said that corporate bonds tend to be more likely to underperform during bad equity markets, which is intuitive economically. Stress and uncertainty in corporate fortunes should impact both the equity and debt of a given company. This is clear from any look at the data. For example, in the bottom decile of monthly stock returns (the worst 10% of months from 1926-2014), the arithmetic average excess credit return is -0.27%. Over the full set, the credit return has an average of 0.18%.
But does this simply reflect a constant equity component to credit over all market conditions, or do the equity-like characteristics phase in and out? For reference, the correlation (Pearson's r) of monthly returns over the full data between credit excess returns and S&P 500 excess returns is 0.22 (95% CI between 0.16 and 0.28), so there is a relationship, but credit risk is distinct.
If credit risk truly behaves more like equity risk at certain times, this should be reflected in figures for equity market beta. Here is the rolling 3-year (36-month) equity market beta for the excess credit return. For each date, the figure given is the estimate for beta over the preceding 36 months. The standard error on the estimates averages under 0.04, so while some of the fluctuations likely reflect noise, the larger trends are probably significant. For reference, the S&P 500 rolling 3-year excess return over T-Bills and credit excess return over duration-matched government bonds are also shown below so equity bear markets (in nominal terms) can be identified.
Other than the spike in equity beta for credit risk around the Great Recession, there doesn't appear to me to be much of any relationship between stock market performance and the sensitivity of credit risk to the equity market. Notably, the beta was close to zero through the Great Depression, though again, the data quality that far back is low and the corporate bonds of the day may have been higher quality. I'd also expect that trading on the stock and bond markets is more integrated and related these days.
Now let's focus on data since 1950, as we've clearly seen high stock market volatility and little relationship with credit risk over the beginning period. Maybe that's an extended outlier or represents an era that no longer exists. In these remaining months, if credit risk becomes more equity-like during bear markets, then presumably we should expect higher market beta during months of downwards stock movements than months of upwards stock movements, right? In fact, the data indicates the opposite: 0.056 equity beta for credit risk under 316 negative months and 0.110 beta for the remaining 464 positive and zero excess return months for stocks.
Over the period since 1950, the equity beta is 0.081. If we subtract out 0.081 times the equity return from credit risk, how does the residual do? Again, if credit risk is worse when it counts the most at a level beyond its typical relationship with stocks, then the residual [excess credit return minus its equity market beta influence] should be worse than average at these times. The arithmetic average residual over the period since 1950 is 0.082 (a positive return on average above any amount explained by the equity beta). For the worst decile of stock returns over the period, the arithmetic average residual is 0.162, even higher, rather than being negative.
Enough theory. How does this work with real funds and asset allocations?
The trouble with portfolio evalution using live trading data from actual mutual funds is that virtually everything we have comes from the bond bull market era the last 35 years, where term risk was overall richly rewarded. Any comparison using corporate bonds of lesser duration than government bonds will inadvertently capture the combined effect of lower term risk along with the credit risk component. This is common, as corporate bonds of equivalent maturity as government bonds will have higher yields and thus lower duration (though this is further complicated by corporate bond call options).
Additionally, most older bond funds are actively managed, with potentially varying credit and term risk exposures. The broad index itself has changed over the years.
One way to minimize the impact of diverging term risk in our comparisons is to restrict bonds to relatively short, constrained maturities. The best two funds I found for this purpose to compare government and corporate bonds are iShares 1-3 Year Treasury Bond (SHY) and iShares 1-3 Year Credit Bond (CSJ), which currently have effective durations of 1.88 and 1.90 years respectively. The expense ratios are 0.15% and 0.20%, an edge to the government bonds.
Here is the performance of three asset allocations over Feb 2007-Dec 2016.
Code: Select all
IVV - iShares Core S&P 500; SHY - iShares 1-3 Year Treasury Bond; CSJ - iShares 1-3 Year Credit Bond Portfolio 1: 60% IVV, 40% SHY, 0% CSJ Portfolio 2: 55% IVV, 0% SHY, 45% CSJ Portfolio 3: 57% IVV, 19% SHY, 24% CSJ
Over this period, which includes the financial crisis, having corporate bonds actually helped. Portfolios 2 and 3 were slightly superior to Portfolio 1 despite the slightly extra cost and all worries about correlations or betas spiking during the worst of the period. Furthermore, only US stocks were used and US stocks outperformed ex-US stocks over the period, so a similar analysis using a global stock portfolio would have shown an even greater benefit for including corporate bonds (notably, over 35% of the bonds in CSJ are from outside the US such as from KfW Bankengruppe, just denominated in USD). However, the scope of this investigation is limited, looking only at short-term bonds and over less than a decade of returns.
I also attempted to compare corporate and government bonds over longer periods of time, using other funds. The caveats about duration/yield curve positioning mismatch and potential risk exposure drift apply here, so I would not trust this analysis as much as the one above.
In the first case, the bond funds are iShares 7-10 Year Treasury Bond ETF (IEF) and iShares iBoxx $ Investment Grade Corporate Bond ETF (LQD), evaluated over Aug 2002-Dec 2016. Corporate bonds here detracted from portfolio efficiency, reducing return and increasing risk. Notably, intermediate-to-long Treasuries were extremely good over this time period, having negative correlation with equities. In the second, the bond funds are Vanguard Intermediate-Term Treasury Fund (VFITX) and Vanguard Intermediate-Term Investment Grade Fund (VFICX), evaluated over Nov 1993-Dec 2016. Corporate bonds again detracted, but not by much.
PV link for IEF/LQD
PV link for VFITX/VFICX
edit: I just realized that the Nov 1993 - Dec 2007 annualized returns for Vanguard Intermediate-Term Treasury Fund (VFITX) and Intermediate-Term Investment Grade Fund (VFICX) were 6.17% and 6.16% respectively. Over this period, the CORP_XS series indicates well over a 1% pa return for credit, so... I'm going to have to assume this is not a fair comparison with respect to term risk, especially considering that the standard deviations were 4.88% and 4.42%.
I find that historically, credit risk probably does not act significantly more like stocks and misbehave in equity bear markets. The possible exception is with our recent experience during the Great Recession.
Corporate bonds contain both credit risk and term risk. Therefore, if credit risk is additive to returns* and sufficiently distinct from equity and term risk, they should be a useful building block for asset allocation. However, there are challenges in backtesting bond allocations with respect to properly adjusting for and equalizing term risk exposure, as covered in the previous section. The results were encouraging using the better duration-matched funds but less so with the other two comparisons (that are on shakier ground with respect to matching fund durations).
*They are not under Fama and French's finding of 2 bp per month of excess return over government bonds, which as you'll recall was not duration matched. If you assume a near-zero or negative forward return for credit risk, there's no reason to include it.
Factor investors typically attempt to gain exposure to multiple factors in a search for multiple sources of potential return. Some formulate this as gaining exposure to (historically) rewarded risk factors, though others these days drop the conception of factor returns being purely explained by risk. Regardless, there is a clear story behind credit risk. It should not be considered anomalous if credit risk carries a positive return on average.
In light of this factor investing framework, I suggest that credit risk may be another factor worth pursuing. Like SMB (size), it is not expensive to access: with corporate bonds, you load on both term and credit risk simultaneously, and they are not drastically more expensive to own than government bonds. Like SMB, there is some modest correlation with equities, but it remains a distinct factor.
This is not a recommendation to abandon high quality bonds but rather a suggestion to not automatically run away from corporate bonds. The highest quality bonds still are most likely to hold up or even gain in a flight to quality, and you may have spending needs sooner than you might think. As noted near the very top, sometimes CD excess yield over Treasuries, tax considerations, and other factors may make corporate bonds less appealing. Others may be turned off by the convexity and call features of some longer-dated corporate bonds. That is fine. But for the investor mulling bond fund choices in a 401(k), a core bond fund containing significant credit risk may not be such a bad idea. Just be sure to account for the presence of riskier bonds by adjusting the asset allocation overall.