The Four-Part Risk Underwriting Model
Analyze, diversify, and price known risks instead of rejecting them outright.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 99%
The model treats risk as something to underwrite rather than reflexively avoid. Howard Marks compares high-yield investing with life insurance: insurers know every insured person will eventually die, but they can still operate profitably because mortality is recognized, studied, diversified, and priced. Applied elsewhere, the decision-maker first names the risk so that its occurrence will not be a surprise. Next, evidence is used to estimate likelihood and severity. Exposure is then spread across independent cases rather than concentrated in one type of borrower, customer, geography, or failure mode. Finally, the offered yield, premium, discount, or strategic reward must more than compensate for expected losses and uncertainty. The output is a disciplined accept-or-reject decision with an explicit margin of safety.
Origin
Howard Marks articulated the model while explaining why Oaktree could invest safely in low-rated bonds despite knowing that some issuers would default.
Core principles
- 01Risk is not automatically unacceptable.
- 02Known risk is safer than unexamined risk.
- 03Analyze the probability and severity of loss.
- 04Diversification limits exposure to any single failure mode.
- 05Demand compensation greater than expected losses.
How to run it
- 1
Recognize the Risk
State the adverse event explicitly and determine whether it is an expected feature of the activity or an overlooked surprise.
Pro tip Write the risk in concrete event terms rather than vague labels such as dangerous.
Watch out Unacknowledged risk cannot be deliberately managed.
- 2
Analyze the Exposure
Estimate how often the event may occur, how severe the loss could be, and which observable factors change those estimates.
Pro tip Use base rates and case-specific evidence together.
Watch out Do not confuse a promised return with the return likely to be realized after losses.
- 3
Diversify Failure Modes
Spread exposure across independent borrowers, sectors, geographies, or other risk drivers.
Pro tip Test whether supposedly different positions would fail for the same underlying reason.
Watch out A large portfolio is not diversified when its positions share one hidden dependency.
- 4
Price the Risk
Compare the yield, premium, discount, or other reward with expected losses and uncertainty.
Pro tip Require enough excess compensation to create a margin of safety.
Watch out A risky asset is not attractive merely because its quoted return is high.
- 5
Monitor the Underwriting Case
Track whether loss probabilities, correlations, or compensation change after the decision.
Pro tip Reduce exposure when the original margin of safety disappears.
Watch out A sound initial analysis does not make the position permanently safe.
In the wild
A credit manager expects some single-B issuers to default. Instead of rejecting the category, the manager analyzes each issuer, limits position sizes, diversifies across industries, and compares the portfolio's excess yield with expected credit losses. The manager invests only when the excess yield provides a meaningful cushion.
→ Defaults can occur without destroying the portfolio's positive realized return.
An insurer estimates mortality using medical evidence, avoids insuring only people exposed to the same extreme hazard, and prices policies on conservative assumptions. Individual deaths are expected rather than treated as evidence that the model failed.
→ Premiums and diversification cover claims while preserving an underwriting profit.
Common mistakes
Rejecting Risk Regardless of Price
An asset can be attractive when the price more than compensates for an acknowledged and analyzable risk.
Trusting the Promised Yield
Quoted yield ignores defaults and recoveries. Focus on expected realized return after losses.
Diversifying by Position Count
Many positions can still share one correlated source of failure. Diversify the actual risk drivers.
Is it for you?
Best for
It is best for credit, insurance, portfolio, and operating decisions involving recurring and estimable losses.
Not ideal for
It is not ideal for unknowable, catastrophic, or highly correlated risks that cannot be diversified or sensibly priced.
From the transcript
“number one it's risk they're aware of it doesn't come as a shock when somebody dies”
“number three it's risk you can diversify”
“number four it's risk they're well paid to take”
From the episode
Howard Marks & Andrew Marks: Something of Value
Howard Marks