Lesson 19 of 20 · 6 min
Base rates, and why they get ignored
The general frequency of an outcome usually predicts better than the specific story about your case.
A base rate is how often something happens across all comparable cases. It is the least glamorous input to any judgement and usually the most predictive one.
The reason base rates get ignored is that specific information feels more relevant than general information. Presented with details about a particular situation — this company's position, this manager's record, this year's conditions — people update heavily on the details and lightly on the general frequency, even when the details have little predictive power. The general frequency is boring and it is usually the better starting point.
Applied here, several base rates are worth knowing. Broad equity markets have had a drawdown exceeding 20 percent roughly once a decade, and one exceeding 30 percent several times a century. Individual stocks fail at rates far above index-level statistics. And the frequency of large declines is higher than any model fitted to ordinary conditions predicts, which is the fat-tails lesson restated as a base rate.
The specific way base-rate neglect interacts with leverage is worth stating. A person setting a draw size is implicitly answering how far can this fall, and the natural anchor is recent experience — which, if the last several years were calm, is precisely the sample that excludes the events the threshold exists for. Recency and base rates pull in opposite directions, and the threshold does not care which one you used.
The base rate is the number to start from. Specific information should move you off it, and usually by less than it feels like it should.
Key terms
- Base rate
- The frequency of an outcome across all comparable cases. The correct prior before specific information is considered.
- Base-rate neglect
- Weighting specific details heavily and general frequencies lightly, even when the details have little predictive power.
- Recency bias
- Over-weighting recent experience. Especially damaging when the recent sample is calm and the threshold exists for turbulence.
Worked example
Two ways to size the same decision
A borrower deciding how much headroom to leave, in a period whose last five years contained no drawdown beyond 15%.
- Recency anchor: worst decline in the last five years
- 15%
- Draw implied by budgeting for 15%
- Headroom set near 15%
- Base rate: drawdowns beyond 20%, broad equities
- Roughly once a decade
- Base rate: drawdowns beyond 30%
- Several times a century
- Trigger drawdown at a 28.6% headroom
- 28.6%
- Which anchor the threshold responds to
- Neither — it responds to what happens
A five-year sample and a century-long base rate give very different answers to the same sizing question. Only one of them contains the events thresholds exist for.
Check yourself
The last five years contained no decline worse than 15%. Why is that a poor basis for deciding how much headroom to leave?
Saved in this browser only. No account, nothing sent anywhere.
Educational only. Nothing in this lesson is investment, legal or tax advice, or a recommendation to enter into any transaction. Figures are illustrative and computed from the inputs stated in the lesson.