Decompose complex questions into sub-questions

Break a hard forecasting question into smaller, estimable pieces and aggregate them.

Why it works

Forecasting a complex outcome directly invites an availability-based gut estimate, which is sensitive to framing and salient recent events. Decomposing into sub-questions forces the analysis to address each component independently, catches missed dependencies, and makes the estimate auditable. Aggregating well-calibrated sub-estimates produces more accurate total estimates than directly estimating the whole.

How to do it

1. Write the forecast question explicitly at the top of a page.
2. Ask: "What would need to be true for this to happen?" List three to five necessary components.
3. Estimate each component separately using available evidence and base rates.
4. Combine the sub-estimates: if each component is independent, the joint probability is the product of the sub-probabilities.
5. Note where the components interact or are correlated and adjust accordingly.

Evidence

Tetlock found decomposition to be a characteristic practice of superforecasters. Separately, Fermi estimation (a form of decomposition) has shown accurate results in contexts ranging from physics to population estimation. The mechanism — forcing component-level rigor — is well supported. (observational)

Decomposition errors can compound if sub-questions are not truly independent or if a key component is omitted; the quality of the decomposition matters as much as the act of decomposing.

Sources

• Tetlock & Gardner (2015), Superforecasting

Common mistake

Decomposing into components that feel tractable rather than ones that are causally relevant, which produces a precise answer to the wrong question.