Picking Corporate Bonds: Credit-Sensitive Security Selection (Part 1)

Government bonds had 13 sovereign issuers to choose from. Corporate bonds? Try over 3,000 issuers and 15,000 individual bonds across developed markets alone.

That’s the scale shift in Chapter 6 of Systematic Fixed Income. We’re leaving government bond territory and moving into credit. And the rules of the game are different here. Richardson walks through how the corporate bond universe is structured, why credit investing is not the same as stock investing, and how to build value signals for picking winners.

This chapter is long, so we’re splitting it into two posts. Part 1 covers the corporate bond landscape, the importance of across-issuer selection, and the value investment theme. Part 2 will cover momentum, carry, defensive signals, and combined strategies.

The Corporate Bond Universe

Richardson uses ICE/BAML indices to map out the corporate bond market. He breaks it into four buckets: US investment grade (IG), US high yield (HY), European IG, and European HY.

As of December 2020, the numbers looked like this:

• US IG: $8.12 trillion, 1,206 issuers, 8,802 bonds
• European IG: $3.84 trillion, 602 issuers, 3,443 bonds
• US HY: $1.55 trillion, 860 issuers, 2,030 bonds
• European HY: $0.54 trillion, 384 issuers, 808 bonds

Total market cap: $14.05 trillion. And it’s been growing fast. More companies are tapping public bond markets to raise debt, so both the number of issuers and the number of bonds have climbed steadily since the mid-1990s.

The average US IG issuer has about seven bonds outstanding. The average US HY issuer has about two. That’s way fewer than the 100+ bonds a single government can have, which means there’s less room for within-issuer selection (picking one bond over another from the same company). The real action is across-issuer selection: choosing Company A over Company B.

Duration and Credit Spreads

Corporate bonds have shorter durations than government bonds. The average US IG bond had a duration of 8.21 years, while US HY came in at just 3.55 years. That gap is not random. Lenders give riskier borrowers shorter ropes. Nobody wants to lend a shaky company money for 30 years.

There’s also a big difference in the long end. US IG has a heavy tail of long-duration bonds because American pension plans love long-dated safe income for matching their liabilities. This creates demand, and US corporations are happy to issue into it.

Credit spreads follow a strong countercyclical pattern. They stay relatively tight during good times, then blow out fast during crises. The dot-com bust in 2000-2001, the 2008 financial crisis, and the COVID shock in early 2020 all show this pattern clearly. HY spreads are much wider than IG on average, which makes sense since HY issuers carry more default risk.

One important wrinkle: many corporate bond issuers don’t have publicly traded stock. This is especially true in Europe. That matters because a lot of useful data (quarterly financial statements, analyst coverage, equity prices) comes bundled with being a public company. If an issuer is private, your data sources shrink.

Across-Issuer vs. Within-Issuer Selection

Richardson demonstrates that across-issuer selection is where the money is. He decomposes credit spread changes into two pieces:

• Level: the average spread change across all bonds from the same issuer (a parallel shift in the credit curve)
• Slope: how spread changes differ between short and long maturity bonds for the same issuer

The level component alone explains about 60% of credit excess return variation for IG corporates and 77% for HY. Adding slope bumps those numbers to 75% and 90%.

The takeaway is clear. If you had a crystal ball, you’d want to know which issuers’ spreads are going to tighten or widen. That across-issuer call is the primary driver of credit returns. Within-issuer curve trades matter, but they’re secondary.

Credit vs. Equity: Not the Same Game

Richardson makes a strong point here. A lot of people, including some academics, treat credit investing as a copy-paste of equity investing. It’s not.

An equity investor owns the residual claim. After all other stakeholders get paid, whatever is left belongs to shareholders. The upside is unlimited. Equity investing is fundamentally about earnings and earnings growth.

A credit investor owns a senior claim with a capped upside. You get your coupons and principal back, and that’s it. What drives credit returns is not earnings growth. It’s the probability of default and how much you’d recover if default happens. Credit investing is about the downside.

This difference has real consequences. Equity market signals don’t automatically translate to credit markets. Richardson cites a sharp observation from his own research: just because factor X predicts stock returns, and stock returns are correlated with bond returns, does not mean factor X is useful for credit investing. That logic is lazy and misses the structural differences between the two claims.

There’s also an agency problem. Companies are managed for the benefit of equity holders. Actions like leveraged buyouts or stock buybacks can be great for shareholders but terrible for bondholders. A buyback reduces cash on the balance sheet and pushes the company closer to its default threshold. Good for stocks, bad for bonds.

The Value Signal: Spread vs. Default Risk

So how do you actually pick corporate bonds systematically? Richardson starts with value.

The core idea is simple. Credit spreads reflect the market’s expectation of default losses. If you have your own estimate of default probability, you can compare it to what the market is pricing. The “gap” between the two is your value signal. A bond is cheap when its spread is wide relative to your fundamental view of its default risk.

In formula terms: Spread is roughly equal to expected default probability times expected loss given default. If you can model default probability independently, then the regression residual (actual spread minus model-implied spread) tells you which bonds are mispriced.

Distance to Default: The Key Input

To build a default probability model, Richardson focuses on “distance to default” (D2D), a concept from Merton’s 1974 structural model.

The idea behind D2D is intuitive. Imagine a company with a certain enterprise value (the total market value of all its claims). Now look at its contractual commitments (total debt obligations). Right now, enterprise value exceeds debt, so the company is solvent. The question is: how likely is it that enterprise value will drop below the debt threshold in the future?

D2D quantifies this by measuring the gap between current enterprise value and the default threshold, scaled by the volatility of the firm’s assets. Think of it like a t-statistic. A D2D of 5 means you’d need a 5-standard-deviation shock to push the company into default. That sounds super safe, but Richardson warns against using a normal distribution to map D2D to default probability. Real-world defaults are fat-tailed. A D2D of 5 under a normal distribution gives you a default probability of 0.00015%, which is basically zero. But empirically, firms with D2D of 5 default at meaningfully higher rates.

That’s why serious credit investors calibrate their models to actual default data rather than assuming normal distributions. Getting access to quality default data going back decades is a big deal here.

Three things push D2D down (meaning default risk goes up):

1. More debt relative to enterprise value (higher leverage)
2. Higher volatility of the firm’s assets (more uncertainty)
3. Shorter maturity on the debt (less time to recover)

There’s a catch, though. D2D requires publicly traded equity to estimate enterprise value. So it only works for issuers with listed stock. For private issuers, Richardson uses a second model: a cross-sectional regression of credit spreads onto credit ratings, spread duration, and return volatility. The final value signal blends both approaches equally.

Checking Your Work

Richardson emphasizes a practical validation step. If your value signal is good, it should be mean-reverting. When a bond looks cheap (positive residual), spreads should tighten over the following months, closing the gap. You can test this with simple regressions of future spread changes on the current value signal. The coefficient should be negative (cheap bonds tighten, expensive bonds widen), and the cumulative coefficients should approach negative one over time (full closure of the valuation gap).

You also need to verify that it’s the spread leg doing the converging, not your model drifting. Both checks should pass for a robust value signal.

And critically, all of this must be done out of sample. You can’t train your model on the full time series and then go back to test it. Many academic papers fail this test and are only useful for descriptive purposes. If you’re building a real investment process, your model parameters need to be point-in-time, just like your data.

What’s Next

That covers the corporate bond landscape and the value investment theme. In Part 2, we’ll look at momentum signals, carry, defensive themes, and how combining everything into a multi-signal approach performs across developed credit markets.

Previous: Selecting Government Bonds Systematically

Next: Corporate Bond Selection: Momentum, Carry, and Machine Learning (Part 2)

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