Chapter 2: Credit Ratings and Default Statistics#

Think of a credit rating as a summary health report for a borrower. Just as a school grade tells you how a student is doing, a credit rating tells a lender how likely a company (or a bond) is to repay its debts. But no summary is perfect. In this chapter we’ll see what ratings really mean, how often defaults actually happen, and how much money lenders can expect to get back when things go wrong. By the end, you’ll understand why the raw numbers are tiny in a single year but add up to a lot over a longer period, and you’ll have a simple tool for modelling the messy world of recovery rates.

The Big Picture#

Credit ratings and hard default data form the foundation of credit analysis. Ratings give a forward‑looking opinion about whether a borrower can repay, while default statistics show what actually happened. Together, they help investors, banks, and regulators answer one central question: “How much should I worry, and how much should I charge for taking this risk?” This chapter looks at the rating scales, the skewed shape of historical defaults, and a practical way to model the losses left after a default. Everything we discuss here will appear again — in pricing bonds, valuing credit derivatives, and building loan portfolios.

Credit Rating Agencies: What They Do and Why We Use Them#

Before we dig into default numbers, let’s meet the companies that create ratings and understand their goals. Three large, independent firms — called Credit Rating Agencies (CRAs) — dominate the global market: Standard & Poor’s (S&P), Moody’s, and Fitch. Their job is to assess an issuer’s creditworthiness: its ability and willingness to make all promised payments, on time and in full.

Credit Rating: A symbol (such as AAA or BB) that sums up an agency’s forward‑looking opinion on the relative chance that an issuer will fail to meet its financial obligations.

There’s a crucial point that’s easy to miss: a rating is not a recommendation to buy or sell, nor is it a guarantee. It is an expert’s opinion based on public and private information — financial statements, industry outlook, management quality, and the specific terms of the debt. Over time, these opinions have become embedded in regulations and investment mandates. For example, many pension funds are only allowed to hold bonds rated above a certain level.

Credit Rating Agency (CRA): An independent company that analyses the financial health of borrowers and assigns ratings used by investors and regulators to gauge credit risk.

Why do we rely so heavily on these agencies? It boils down to information and efficiency. A single investor cannot possibly analyse thousands of bonds across dozens of countries. CRAs pool that analytical effort, giving the market a common language. That language is the rating scale.

📝 Section Recap: Rating agencies give us a shared shorthand for creditworthiness. Their ratings are long‑term opinions, not guarantees, and they underpin most of the world’s debt markets.

Rating Scales and the Cost of Borrowing#

The scale used by S&P and Fitch runs from AAA down to D, while Moody’s uses a similar scale with letters and numbers (Aaa, Aa1, etc.). To keep things simple, we’ll use S&P‑style notation. Bonds rated BBB‑ or higher are called investment grade; everything below (BB+ and lower) is speculative grade or high‑yield.

A rough rule of thumb: as you move one notch down the scale, the extra yield (spread) that investors demand over a default‑free government bond rises by a noticeable step. For a healthy company a few notches into speculative grade, that extra spread might be around 2–3% per year; for the riskiest ones it can be 10% or more. The exact spread depends on market conditions, but the pattern is always the same — lower rating, higher borrowing cost.

Here is a simplified rating scale with a broad feel for the kind of average credit spreads a company might face in a normal market environment (the actual numbers vary daily):

The spread isn’t random — it’s the market’s attempt to charge a fair fee for the expected loss and for the uncertainty surrounding that loss. To see if that fee is fair, we need to look at what actually happens to rated bonds: their default rates and loss severity.

📝 Section Recap: Ratings group issuers from virtually risk‑free (AAA) to already in default (D). Each step down the ladder means a higher interest spread for the borrower.

Default Statistics: Annual Rates and Cumulative Risks#

If you scan a long history of bond defaults, the first thing that jumps out is how lopsided the data are. Most years, very few bonds default. Then, every so often, a recession or a crisis hits and defaults shoot up sharply. This asymmetry is key: default risk is not a smooth, steady process — it’s a process where quiet times are interrupted by bursts of trouble.

Annual Default Rates#

For investment‑grade bonds, yearly default rates are tiny. Over the course of a decade, it’s not unusual for no AAA‑rated company to default at all. Even for BBB‑rated issuers, the yearly percentage of defaults rarely exceeds 0.3–0.5% in normal times.

For speculative‑grade bonds the numbers are higher, but still far smaller than many people expect. In a typical year, fewer than 2–3% of all high‑yield bonds default. When the economy hits a rough patch, that rate can climb to 8–12%. Over the very long run — spanning many credit cycles — the average annual default rate for the whole speculative‑grade universe sits around 3–4%.

Why are these numbers so low in an average year? Because most companies can keep servicing their debt even when profits dip, and because a default is a last resort after all other options (refinancing, selling assets, cutting dividends) have been tried. Still, don’t let a small one‑year number lull you into thinking the risk is trivial. Time is a powerful magnifier.

Cumulative Default Rates#

Imagine you buy a bond rated BB today. Suppose its annual default chance is a constant 2.5% each year. That doesn’t mean your chance of seeing a default over five years is 5 × 2.5% = 12.5%, because each year you survive shrinks the pool of remaining bonds. The correct way uses survival chances:

If the annual default probability is , the chance of surviving one year is . Over years, the probability of surviving all of them is the product of the yearly survival chances. If stays constant, the cumulative default probability is

For a 2.5% annual probability, over five years that works out to roughly

So even a modest one‑year risk grows into a real multi‑year number. But reality is often harsher than this constant‑rate math. A bond’s rating can change over time — a BB company might get downgraded to B, making its future default chance higher. Because of this rating migration, actual cumulative default rates over long stretches are often larger than the simple formula suggests.

More importantly, the initial rating’s power to predict the future gets weaker as the horizon lengthens. A rating assigned today best captures the agency’s view of the next few years. Ten or fifteen years out, companies can look very different — mergers, new technologies, or regulation changes that no one saw coming. As a result, the gap in cumulative default rates between, say, a BB and a B rating shrinks over very long periods; the labels grow noisier with time.

A stylised illustration helps:

Notice that the difference in default rates between BB and B is about 5 percentage points in year 1, widens to 12 points after five years, then stands at around 17 points after ten years. But as the overall risk level rises, that gap becomes a smaller fraction of the total — ratings become less discriminating.

The skewed nature of defaults also means that modelling them with a smooth bell curve is misleading. Later chapters will tackle better models, but for now the key takeaway is: defaults are rare but clustered events, and the longer your time frame, the more clustering and rating drift matter.

📝 Section Recap: Annual default rates are tiny for investment‑grade bonds and stay in the low single digits for high‑yield in average years, but they compound into substantial multi‑year risks. Initial ratings lose predictive accuracy over longer horizons as companies evolve.

Loss Given Default and Modelling Recovery Distributions#

A default does not mean a total loss. When a company fails to pay, its assets are sold (or restructured) and the proceeds are distributed to creditors according to a priority of claims — essentially a pecking order. How much you get back, expressed as a percentage of the bond’s face value, is the recovery rate.

Recovery Rate: The share of the bond’s face value that you eventually get back after a default, often quoted as a percentage.

The flip side of recovery is Loss Given Default (LGD):

If a bond ultimately pays back 60 cents on the dollar, the recovery rate is 60% and LGD is 40%. Across all corporate bonds, the long‑run average LGD has been close to 40%, meaning the typical recovery is about 60%. But that average hides enormous variation.

What Drives LGD?#

Three big factors matter:

1. Seniority. Senior secured bonds — backed by specific assets like buildings or equipment — often recover 70–80% (LGD 20–30%). Unsecured subordinated bonds may recover only 20–30% (LGD 70–80%). The more layers of protection you have, the bigger your share of the liquidation pie.

2. Economic cycle. In a deep recession, asset prices are depressed and there are fewer buyers for a bankrupt company’s assets. Unsurprisingly, average recoveries fall just when defaults rise, amplifying losses. A default during a boom may give back 70%; the same bond defaulting in a bust might return only 40%.

3. Sector. Industries with lots of tangible, easy‑to‑sell assets (utilities, heavy manufacturing) tend to show higher recoveries. Technology or service companies, whose value lies mainly in intangibles like patents or customer relationships, often leave creditors with much less.

The Beta Distribution for Recovery Modelling#

Because recovery rates always lie between 0% and 100%, we need a model that respects these boundaries. A natural choice is the beta distribution. It lives on the interval [0, 1] and can take many shapes — bell‑shaped, U‑shaped, or skewed — by tuning just two parameters, (alpha) and (beta).

Beta Distribution: A probability distribution on the range (0,1) that can be shaped by two numbers, and ; widely used for variables that are naturally bounded between 0% and 100%.

The probability density function is

where is a normalising constant (the beta function). In plain English, and push the mass toward 0 (low recovery) or toward 1 (high recovery), and they control how peaked the distribution is near its mean.

When we fit a beta distribution to historical recovery data for a given seniority class, we get specific and numbers. For example, senior secured bonds might have a mean recovery around 75% and a slightly left‑skewed shape (). A subordinated bond might have a mean near 30% and a right‑skewed shape ().

Why does this matter? When a bank or investor calculates how much capital to hold against a loan portfolio, they must simulate many possible default and recovery scenarios. A beta distribution lets them do that realistically, keeping the natural floor of 0% and ceiling of 100% while capturing the tendency for recoveries to cluster around a typical value but occasionally fall very low. It connects messy real‑world data to a simple, workable formula.

📝 Section Recap: Loss Given Default averages about 40% but varies a lot with seniority, the economy, and industry. The beta distribution gives us a flexible and accurate way to model recovery rates inside their natural 0%–100% bounds.

Summary#

You have now seen that credit ratings give a universal shortcut for risk. Behind the symbols lies a wealth of real‑world default and loss experience. The numbers teach us that a single year’s default risk may look tiny, but over many years — with possible rating changes — it builds into a much larger threat. When defaults happen, losses are rarely total: recoveries are predictable on average but wildly uncertain for any single case, which is why we model them with a beta distribution. Together, these ideas form the first solid layer of credit risk analysis.