Subjective probability is an individual’s estimate of how likely an event is to occur based on personal judgment, experience, intuition or belief rather than on an objective, repeatable calculation from long-run data. It’s the kind of probability expressed when someone says “I think there’s a 60% chance this stock will rise” even though they haven’t computed a formal statistical model.
Key takeaways
– Subjective probability is personal and can differ widely between people; it reflects opinion, experience and intuition rather than strictly measured frequencies.
– It contrasts with objective (frequentist) probability, which is derived from repeatable experiments or long-run frequencies, and with model-based probabilities estimated from data.
– Subjective probabilities are useful when data are scarce or events are unique, but they are prone to bias (overconfidence, availability, anchoring, representativeness).
– Calibration, structured elicitation, aggregation, and updating (Bayesian updating) can improve the quality of subjective probability estimates.
How subjective probability works
– Source: Personal judgment. An individual forms a probability based on memory, interpretation of available evidence, rules of thumb, and gut feeling.
– Flexibility: The same person can give different probabilities for the same event depending on framing (e.g., “exact 25%” versus “25–30%”) or recent experiences.
– No single “correct” value: Unlike an objective probability that can be estimated by repeated trials, subjective probabilities are coherent only to the extent the person’s beliefs are internally consistent and can be updated rationally.
– Rational foundation (when used well): In decision theory and Bayesian statistics, subjective probabilities are legitimate inputs (priors) that can be updated with evidence using Bayes’ rule.
Important: when subjective probabilities are appropriate
– Limited historical data, unique events, or strategic games (e.g., business launches, geopolitical events, novel technologies).
– Early-stage forecasting where models are immature and expert judgment is the main information source.
– Personal decisions where preferences and private information matter.
Common pitfalls and biases
– Overconfidence: giving probability ranges that are too narrow or stating excessive certainty.
– Availability bias: over-weighting outcomes that are vivid or recently observed.
– Anchoring: initial numbers or suggestions unduly influence the estimate.
– Representativeness: assuming small samples look like large populations (e.g., 10 tails in a row changes belief about a fair coin).
– Confirmation bias: seeking evidence that supports the initial belief and neglecting counter-evidence.
Practical steps for producing and using subjective probabilities
1. State the question precisely
• Define the event, the timeframe, and the units (e.g., “Probability S&P 500 returns > 10% over next 12 months”).
2. Use reference classes and base rates (outside view)
• Before instincts, ask: “How have comparable projects/companies/teams performed historically?” This grounds judgments and reduces over-optimism.
3. Elicit an explicit probability, not just a qualitative judgment
• Force a numerical or a range estimate (e.g., 40% or 35–45%). Require a confidence level for the range.
4. Break the problem down
• Decompose a complex event into parts (P(A and B) = P(A) × P(B|A)) and elicit probabilities for sub-events when possible.
5. Use structured elicitation techniques
• Probability wheel, betting odds format, or asking for percentiles (e.g., 5th, 50th, 95th) helps make beliefs explicit and coherent.
6. Document assumptions and reasoning
• Record why you chose a probability, what evidence was used, and alternative scenarios. This makes later review and learning possible.
7. Aggregate multiple judgments
• Combine independent experts’ probabilities by simple averaging or weighted methods (e.g., weights based on track record). Aggregation usually outperforms individuals.
8. Score and calibrate your forecasts
• Use a scoring rule like the Brier score to measure forecast accuracy and use results to improve calibration over time.
9. Update with Bayes’ rule when new evidence arrives
• Treat your subjective probability as a prior and revise it formally as new data arrives to produce a posterior probability.
10. Run pre-mortems and counterfactual checks
• Imagine the event failed and list plausible reasons. That surface unknowns and reduces optimism bias.
11. Use ranges and stress tests
• Provide probability ranges and examine sensitivity to different assumptions rather than reporting a single point estimate.
12. Seek disconfirming evidence
• Actively try to falsify your own probability estimate to uncover blind spots.
Worked examples
A. Coin flips — intuition vs. correct reasoning
– Scenario: You flip a coin 10 times and observe 10 tails. Your initial subjective belief before any flips is 50% heads / 50% tails for the next flip if you assume a fair coin (objective frequency). After seeing 10 tails, the human instinct may shift the subjective probability (e.g., “now it’s 75% tails”), but rational updating depends on your model:
• If you are certain the coin is fair, the probability remains 50%.
• If you are uncertain about fairness, use Bayesian updating: assign a prior distribution over the coin’s bias and update that prior with the 10 tails data to compute a posterior probability for the next flip. The posterior will shift toward tails proportionate to your prior uncertainty.
B. Sports fandom (Yankees example)
– Question: “What is the probability the Yankees win the World Series this season?”
• Start with base rates: how often do teams with similar payrolls/rosters/history win?
• Add team-specific evidence: injuries, trades, schedule, pitching depth.
• Elicit a range to acknowledge uncertainty (e.g., 10–25%) and document the reasons for the upper vs. lower bound.
• Combine fan intuition with market odds (betting markets) and analytic forecasts to produce a blended probability that balances subjective passion and objective signals.
Improving quality of subjective probabilities — methods and tools
– Calibration training: exercises that force regular probabilistic answers and provide feedback improve long-term calibration.
– Aggregation and expert combination: simple averages of many judgments often beat single experts. More sophisticated methods (e.g., Cooke’s classical model) weight experts using performance scores.
– Scoring rules: use proper scoring rules (Brier score, log score) to incentivize honest, well-calibrated probability statements.
– Structured analytic techniques: pre-mortems, devil’s advocate, and red-team reviews reduce groupthink and anchoring.
– Bayesian frameworks: formalize subjective beliefs as priors and update them with evidence to produce disciplined revisions.
When not to rely on subjective probability alone
– When ample high-quality historical data exists (use objective models and statistical inference).
– For systems where repeatable experiments are possible and preferable (e.g., manufacturing defect rates).
– When decisions require fully auditable, reproducible risk estimates (regulatory or compliance settings may demand objective methods).
Summary checklist for producing a useful subjective probability
– Define the event and horizon precisely.
– Gather base rates and comparable-case data first.
– Elicit a numeric probability or a confidence range; capture your rationale.
– Break complex events into subcomponents if possible.
– Aggregate multiple independent judgments where feasible.
– Score and review forecasts to improve calibration.
– Update beliefs when new data arrive using a transparent rule (Bayes’ rule if appropriate).
– Be explicit about uncertainty and avoid unwarranted precision.
References and further reading
– Investopedia — “Subjective Probability” (Paige McLaughlin);
– Kahneman, D. (2011). Thinking, Fast and Slow. (Covers biases and heuristics that affect subjective judgments.)
– de Finetti, B. (1974). Theory of Probability (A foundational work on subjective probability and coherence).
– Cooke, R. M. (1991). Experts in Uncertainty: Opinion and Subjective Probability in Science. (Methods for combining expert judgments.)
Editor’s note: The following topics are reserved for upcoming updates and will be expanded with detailed examples and datasets.