Corporate Bond Investing 101: The Market, Credit Spreads, and Default Risk

Book: Systematic Fixed Income: An Investor’s Guide Author: Scott A. Richardson, Ph.D. ISBN: 9781119900139 Publisher: John Wiley & Sons, 2022

Chapter 6 tackles corporate bonds, and it’s a big one. So I’m splitting it into two posts. This first part covers the lay of the land: how big the market is, what corporate bonds look like, and the key concept of default risk. The second part will get into the actual investment signals and performance results.

The Corporate Bond Market Is Massive

As of December 2020, the total market cap across four major corporate bond universes was about $14 trillion:

• US Investment Grade (IG): $8.12 trillion, about 1,206 issuers, 8,802 bonds
• European IG: $3.84 trillion, about 602 issuers, 3,443 bonds
• US High Yield (HY): $1.55 trillion, about 860 issuers, 2,030 bonds
• European HY: $0.54 trillion, about 384 issuers, 808 bonds

These markets have grown enormously over the past two decades. More companies are tapping public bond markets for financing.

The average US IG issuer has about seven bonds outstanding. The average HY issuer has about two. So there’s room for both across-issuer and within-issuer selection, but the real breadth of opportunity is in picking between issuers, not between different bonds from the same company.

Duration and Spread Profiles

Corporate bonds have shorter duration than government bonds. US IG averages about 8.2 years of duration, while US HY is just 3.6 years. This makes sense. Lenders rationally lend to riskier companies for shorter periods.

Credit spreads are strongly countercyclical. They widen fast during economic stress (dot-com bust, 2008 financial crisis, COVID) and compress during good times. HY spreads are much wider than IG spreads, obviously, because those companies are riskier.

One useful feature: the return per unit of risk tends to be higher at the front of the credit curve. Shorter-duration credit has historically offered better risk-adjusted returns. This will matter when we talk about the “defensive” investment theme.

Many Issuers Are Private

Here’s something that surprises newcomers. A significant portion of corporate bond issuers don’t have publicly listed equity. This is especially true in Europe. If your investment signals rely on stock prices or equity market data (like financial statements filed with stock exchanges), you’ll miss these private issuers entirely.

This doesn’t mean you should avoid private issuers. But it does mean you need to be thoughtful about data sourcing. Your investment opportunity set shrinks when you can only analyze public companies.

Across-Issuer Selection Is What Matters Most

Richardson does a neat decomposition of credit spread changes into two components:

• Level: the average spread change across all bonds of an issuer (parallel shift in the credit curve)
• Slope: how longer-term bonds moved relative to shorter-term ones (flattening or steepening of the credit curve)

The level component alone explains about 60% of credit excess return variation for IG and 77% for HY. Adding slope gets you to 75% for IG and 90% for HY. The takeaway: if you could predict the average credit spread change for each issuer, you’d capture most of the return opportunity.

So across-issuer selection is the primary game in corporate bonds.

Credit Investing Is NOT Equity Investing

Richardson is emphatic about this point. Even academics sometimes just copy their equity analysis into credit markets and expect the same results. That’s wrong.

An equity investor participates in all future upside. Equity returns are driven by earnings and earnings growth, with no cap on participation. A credit investor gets their coupons and principal back (hopefully), with strictly limited upside. Credit returns are driven by changes in default expectations and recovery rates. It’s fundamentally about downside risk.

The correlation between stock and bond returns for the same company depends on how risky the company is. For safer companies, equity and credit claims are less correlated. For riskier companies, the claims move more in sync because both become more sensitive to changes in enterprise value.

And watch out for agency conflicts. Things that are great for equity holders (like stock buybacks or leveraged acquisitions) can be terrible for creditors because they increase leverage and reduce the distance to default.

Understanding Default Risk

Credit spreads are approximately equal to expected loss given default, which breaks down into:

Spread ≈ E[PD] x E[1 - Recovery Rate]

Where E[PD] is the expected probability of default. So to understand credit spreads, you need to understand default risk.

Distance to Default: The Merton Model

Richardson explains the Merton model (1974) for measuring credit risk, and it’s actually really intuitive. Picture it like this:

A company has assets worth some amount (enterprise value) and debts it owes (the default barrier). Right now, assets are worth more than debts, so the company isn’t in default. But enterprise value is uncertain. It could go up or down.

Distance to default (D2D) measures how many standard deviations of asset value decline it would take to push the company into default. It’s basically a t-statistic where the null hypothesis is that asset value equals total debt.

The formula considers:

• Enterprise value vs. debt: more leverage means closer to default
• Volatility: more uncertain businesses are riskier
• Drift: expected return on assets
• Time: how long until debts mature

If D2D is high, the company is far from default. If it’s low, trouble could be close.

One important caveat: don’t use a normal distribution to map D2D to actual default probabilities. The real world has fatter tails. A company with D2D of 5 has a normal-implied default probability of basically zero (0.00015%), but empirically, companies at that level still default at meaningful rates. You need to calibrate against actual default data.

Default Forecasting Methods

Richardson surveys several approaches:

• Basic probability models (Beaver 1966, Altman 1968, Ohlson 1980)
• Structural models (Merton 1974)
• Hybrid approaches combining both
• Machine learning like random forests

The critical requirement: all of this must be done out-of-sample. You can’t train your model on 50 years of data and then go back and test it on the same data. Your model parameters need to be point-in-time, just like your data inputs.

ROC curves (receiver operating characteristic) are the standard way to evaluate these models. They plot true positive rates against false positive rates at various thresholds and give you a visual sense of how well your model discriminates between companies that will default and those that won’t.

Building the Value Signal

The simplest value measure would be the ratio of credit spread to your default forecast. A “cheap” bond is one where the spread is wide relative to what your default model says the risk actually is.

In practice, you typically run a regression of credit spreads on your default forecast plus other variables (like recovery rates and duration), and the residual becomes your value signal. If this residual is strongly mean-reverting (it closes over time), you’ve got a good signal.

Richardson uses two value measures in combination: one based on the Merton/structural approach (distance to default regressed against spreads) and one based on a simpler linear model using credit ratings, spread duration, and return volatility.

That covers the foundations. In the next post, we’ll look at the momentum, carry, and defensive signals for corporate bonds, plus the actual performance results and some interesting extensions around liquidity and machine learning.

Previous: How to Pick Government Bonds Systematically | Next: Credit Bond Selection Signals and Models