Fixed income and credit analytics, explained

Fixed income and credit analytics covers two questions that turn out to be the same question asked from opposite sides of a loan: what is this debt worth, and how likely is it to be repaid. A bond trader pricing a corporate note and a bank underwriting a mortgage are both, at bottom, measuring credit risk and the compensation demanded for carrying it — one does it continuously in a market, the other does it once at origination and then monitors it. The methods differ in sophistication, but the underlying question, and much of the vocabulary, is shared.

The questions people actually ask

A bond portfolio manager asks how much a position will lose if interest rates move, and how much extra yield is fair compensation for the risk that an issuer defaults rather than a risk-free government paying it back. A bank credit officer asks whether to approve a loan, at what rate, and how big a loss reserve to hold against the possibility the borrower doesn't pay. A risk team at either institution asks how the whole portfolio behaves in a downturn, not just any single position. All of it comes down to pricing uncertainty about repayment and about the path of interest rates.

The data it runs on

• Prices and yields across the curve. Bond prices (or their equivalent, yields) at each maturity, refreshed continuously in liquid markets and far less often in illiquid corners of the credit market — a distinction that matters more here than almost anywhere else in finance.
• Issuer and borrower financial data. Corporate financial statements and credit ratings for bond issuers; income, existing debt and repayment history for individual borrowers. This is the raw input to any credit model, whether it prices a bond or underwrites a loan.
• Macroeconomic and scenario data. Interest-rate paths, unemployment and other macro variables that both bond risk models and bank stress tests run against, because credit losses cluster with the economic cycle rather than occurring at a steady background rate.

Core metrics and how to read them

Duration measures a bond's price sensitivity to interest-rate changes, expressed in years: a bond with a duration of 7 will fall in price by roughly 7% for a 1-percentage-point rise in rates, and rise by roughly the same for a 1-point fall. It is the single most-used risk number in fixed income precisely because it compresses an otherwise complex price-rate relationship into one comparable figure — with the caveat that the relationship is only approximately linear, and breaks down for large rate moves or bonds with embedded options.

The yield curve plots yield against maturity for otherwise comparable bonds, usually government debt. Its shape carries information well beyond any single price: a normal upward-sloping curve reflects investors demanding more yield to lend for longer, while an inverted curve — short-term yields above long-term ones — has preceded most recent recessions closely enough that it is watched as a standing economic signal, not just a pricing curiosity.

The credit spread is the extra yield a bond pays over a comparable-maturity risk-free benchmark, compensating an investor for default risk, liquidity risk and structural complexity. Spreads widen when investors grow more worried about default across a sector or the whole market, and narrow when confidence returns — making the spread itself a real-time read on credit sentiment, independent of the underlying rate environment.

Probability of default and loss given default are the two building blocks of expected credit loss: expected loss ≈ PD × LGD × exposure at default. PD estimates how likely a borrower or issuer is to default within a given horizon; LGD estimates what fraction of the exposure is actually lost when default happens, after any recovery. Banks hold loss reserves and price loans against this product, not against PD alone — a high-PD loan secured by strong collateral can carry less expected loss than a lower-PD loan with none.

Credit scoring is the applied, consumer- and small-business-facing version of the same idea: a model condensing income, repayment history and other borrower data into a score used to approve, price or decline a loan. The debt-to-income ratio — total debt obligations divided by income — remains one of the simplest and most heavily weighted inputs into that decision, precisely because it is a direct, hard-to-game measure of repayment capacity.

How the work is done in practice

On the markets side, fixed-income desks work primarily on cross-asset terminals that put live prices, yield curves and issuer credit data in one place, because a bond price is only meaningful next to its curve and its comparable spreads. Bloomberg Terminal is the dominant platform for this: alongside its equity and FX coverage, it carries dedicated fixed-income analytics — yield-curve construction, spread analysis and bond pricing tools — that desks treat as inseparable from the rest of their cross-asset workflow, sold under a negotiated enterprise contract rather than a public price list.

On the lending side, the work is model-building and monitoring rather than real-time trading. Moody's Analytics supplies the credit-risk models and macroeconomic scenario data that bank risk and finance teams use to underwrite loans, estimate probability of default and loss given default across commercial and retail portfolios, and run the regulatory stress tests supervisors require — the institutional-scale version of the PD/LGD framework above. Zest AI approaches the same underwriting problem from a different angle: a machine-learning platform that helps lenders build credit scoring models more predictive than a traditional scorecard, while keeping them explainable enough to satisfy fair-lending compliance — a real tension, since the most flexible models are often the hardest to explain to a regulator or a declined applicant.

Common mistakes and misreadings

Treating duration as exact rather than an approximation. Duration assumes a roughly linear price-rate relationship; for large rate moves, or bonds with embedded call or put options, the actual price change diverges from the duration estimate, which is what convexity measures correct for.

Reading a widening credit spread as only about one issuer. Spreads move with sector- and market-wide sentiment as much as with issuer-specific news; a spread move needs to be compared against a sector benchmark before it is read as company-specific deterioration.

Using probability of default in isolation. A loan's expected loss depends on PD and LGD together — a high default probability against strong, liquid collateral is a different risk than the same probability against none.

Inverting cause and effect on the yield curve. An inverted curve is a widely watched recession signal, but it reflects aggregated market expectations, not a guaranteed mechanical trigger; treating it as a precise timing tool for a downturn oversells what it actually measures.

Over-relying on a single score without the underlying factors. A credit score or a debt-to-income figure summarizes risk but discards the detail behind it; two borrowers with the same score can have very different real repayment risk once the underlying factors are examined.

For the full landscape of platforms behind this work, see every tool in this category and every credit-risk tool in this category.