Skip to content

New to Omni?·Get 20% off your first purchase with code WELCOME20 at checkout.See plans

CAIA® Level I: what do credit models and equity regression show?

Credit models estimate default risk; equity regression estimates market exposure. Learn what each output means, then test the limits of an illustrative case.

ConceptOmni Curriculum Team6 min read

A smooth unlabelled wire line rises over clear acrylic layers beside a blank navy notebook.
On this page

Credit models and equity regression both turn assumptions and observations into estimates, but they answer different questions. A credit model estimates the chance or cost of a borrower failing to pay; an equity regression estimates how a return series moves with a chosen factor. Read each output through its inputs and assumptions before treating it as a forecast or an explanation of cause.

Start with the event each model tries to describe

For credit, the event is default over a defined horizon. A useful loss estimate combines the chance of that event, the balance outstanding at default and the portion that cannot be recovered. These components are conceptually different. A borrower could have a low chance of failure but produce a large loss if the exposure is substantial; strong collateral may reduce the loss if default occurs without changing the event itself. Mixing probability with severity is an easy way to misread a credit result.

Structural models describe default through the value of a firm's assets relative to its obligations. Shareholders retain the residual value after debt claims, so limited liability makes equity resemble an option on the assets. This is an economic story about why default can happen. It also creates a practical difficulty: market observers can see share prices, but cannot directly see the total market value and volatility of every firm asset. Estimates of these inputs carry model risk.

Reduced-form models take another route. They model default as an arrival event and infer a risk parameter from traded credit prices or spreads. That can be useful when a liquid security offers a market anchor for valuing a less liquid claim. The implied default parameter is fitted to prices, however, so it reflects pricing assumptions and compensation for risk. It should not be relabelled as a direct count of how often similar borrowers will actually fail. The CAIA curriculum overview places financial economics foundations within Level I's introduction to alternatives.

Make an illustrative expected-loss calculation

Imagine a fictional one-year loan with £1 million outstanding at the point of default. Suppose the estimated real-world probability of default over that year is 2% and the fraction recovered after default is 40%. These are invented teaching inputs, not a published rate or a particular borrower's estimate. The loss if default occurs is the remaining 60% of the exposure, or £600,000. Weighting that conditional loss by the 2% probability gives an expected loss of £12,000. The arithmetic can be checked as £1,000,000 times 0.02 times 0.60.

The £12,000 result is an average across the assumed distribution, not the amount the lender would necessarily lose in a single outcome. If there is no default, the credit loss in this simplified example is zero; if default occurs and recovery matches the assumption, the loss is £600,000. The expected figure also says nothing by itself about the timing of recovery, expenses of pursuing collateral or dependence between default and recovery. Clarify what is inside the exposure and recovery inputs before applying the multiplication.

A traded bond spread adds a different perspective. A wider spread can reflect expected default losses, but also uncertainty, liquidity, market conditions and risk compensation. In a deliberately simplified one-period risk-neutral model, spread and expected loss severity are linked approximately. The approximation relies on its horizon, discounting and recovery assumptions. It is useful for recognising why lower expected recovery can widen the required spread even when the issuer's default event is unchanged. It is not a licence to call the spread an observed physical default frequency.

When comparing two claims on the same borrower, hold the default event separate from the recovery waterfall. A senior claim and a junior claim may face the same company failure but receive different amounts after it. It follows that a difference in their spreads need not imply two different probabilities of the issuer failing. A simplified model can vary recovery while keeping the event parameter common; in practice, seniority, collateral terms and market liquidity still need checking before that comparison is persuasive. This is why a model output should be labelled with the claim and assumptions it describes.

Read equity regression as exposure and residual

A single-factor equity model asks how a security's excess return responds to the market's excess return. The slope, commonly called beta, measures the estimated change in the security's excess return associated with a change in the chosen factor. The intercept, often called alpha in this context, is the average part not explained by that linear factor relation. The residual is the period-specific difference between the model's fitted value and the observed return. These are statistical descriptions of the sample; none is automatic evidence of manager skill.

Suppose a purely illustrative model has a beta of 1.2 and an intercept of 1 percentage point for a particular period convention. If the market excess return in a period is 5 percentage points, the fitted security excess return is 7 percentage points: one point plus 1.2 times five points. If the observed security excess return is 6 points, the residual is minus 1 point. Those values are invented to show how fitted return and residual differ; they are not a realised market observation or a claim that the intercept will recur.

Now change the market excess return in the same illustrative model to minus 5 points. The fitted security excess return becomes minus 5 points, from one point plus 1.2 times minus five. The example makes the slope intuitive: the fitted return changes by 12 points when the factor changes by 10 points. It also shows why a positive intercept does not prevent a negative fitted return. Do not interpret beta as the entire return or alpha as a cash amount; both depend on the regression specification and return convention.

Ask what the data can and cannot establish

Before interpreting beta, check which factor was selected, whether returns were measured over comparable intervals, and whether the relationship appears reasonably stable. A market-only regression may omit other exposures. Smoothed valuations in some alternative assets can make returns look less responsive than the underlying economics. Extreme observations can pull a fitted line; serial dependence or changing residual variability can make standard error based claims less reliable. A tidy coefficient table should therefore lead to diagnostic questions, not a stronger conclusion than the data support.

The two modelling families are easy to confuse because both may use market prices. A structural credit model works from a story about asset value and debt obligations. A reduced-form credit model fits default-related pricing to traded claims. A return regression estimates a historical relationship with a factor. Each can inform a different decision, and none transforms an uncertain future into a known outcome. This is the central distinction to retain when the wording of an exam item switches from a borrower's loss to a security's systematic return exposure.

A quick diagnostic is to name the unit of the requested answer before calculating. Probability is a fraction over a stated horizon; expected loss is a currency amount when exposure is stated in currency; a spread is a return difference; beta is a slope relating two return series. If an answer appears in the wrong unit, the formula or interpretation has slipped. If an estimate has no horizon or factor definition, ask what information was left implicit. This simple check helps distinguish a sound calculation from a plausible-looking number attached to the wrong question.

CAIA Association says its current official curriculum is the source of truth for exam alignment and that prior editions should not be used as the study anchor. Check the current curriculum overview and the candidate study resources for the applicable exam cycle. For revision, rebuild both invented calculations from their inputs and then write one sentence identifying what each output does not measure. The CAIA program page and how it works page can help place that exercise within a broader plan.

Sources

  1. Explore the Curriculum | CAIA Association · caia.org · Retrieved
  2. Curriculum and Study Tools | CAIA Association · caia.org · Retrieved
  3. Frequently Asked Questions | CAIA Association · caia.org · Retrieved

Explore CAIA study resources

Educational content, not investment advice. This article is written to help candidates prepare for professional exams. It is not a recommendation to buy, sell or hold any security, and it does not take your personal circumstances into account.

Omni Finance Academy is an independent exam-preparation provider and is not affiliated with or endorsed by CFA Institute, GARP, CAIA Association or CFP Board. CFA®, FRM®, CAIA® and CFP® are trademarks of their respective owners. Full disclaimer.