Statistical frequency of events in a reference class that provides the starting point for predictions before considering specific case details
Base rates are the underlying statistical frequencies or prior probabilities of events in a population or reference class. They represent "what usually happens" before you know any specific details about a particular case. Base rate thinking is a cornerstone of rational decision-making because it anchors predictions in empirical reality rather than intuition or vivid examples.
Discovered and formalized by Daniel Kahneman and Amos Tversky in their research on judgment under uncertainty, base rate neglect is one of the most common and costly cognitive biases. People systematically underweight or ignore statistical information (base rates) in favor of specific, individuating information—even when the base rate is more predictive.
The classic illustration: If you're told someone is shy, loves puzzles, and dislikes crowds, is she more likely a librarian or a farmer? Most people say librarian, but there are 20× more farmers than librarians in the U.S., making farmer far more probable even given the personality description.
Using base rates properly means starting with the outside view (what usually happens in similar cases) before adjusting for the inside view (what's special about this case). This produces dramatically more accurate predictions across domains from medical diagnosis to startup success to hiring.
Key insight: Specific, vivid details feel more important than statistical base rates, but base rates are usually more predictive. Start with "what usually happens" before considering "what's special here."
Apply base rate thinking in these situations:
Trigger question: "What usually happens in cases like this?" or "What's the historical success rate for this type of situation?"
Determine what category or population this case belongs to. The reference class provides the base rate.
Action: Write down 2-3 possible reference classes, from most specific to most general.
Find statistical data on success/failure rates, typical outcomes, or frequency in the reference class.
Sources to check:
Action: Document the base rate as a percentage or ratio with source citation.
Evaluate how reliable and relevant the base rate is for your situation.
Quality factors:
Action: Rate base rate quality as High/Medium/Low and note any concerns.
Use the base rate as your initial probability estimate before considering case-specific details. This is your "outside view."
Action: Write down "Initial estimate based on base rate: X%"
List specific details about this case that might justify adjusting away from the base rate.
Relevant individuating information:
Action: List 3-5 case-specific factors and assess whether each is genuinely unusual or just feels special.
Update your estimate using specific information, but make smaller adjustments than feel natural. Research shows people over-update on specific details and under-weight base rates.
Adjustment principles:
Action: Calculate adjusted probability and document reasoning for each adjustment increment.
Review your final estimate to ensure you haven't abandoned the base rate in favor of compelling specifics.
Warning signs of base rate neglect:
Action: If final estimate deviates significantly from base rate, re-examine whether adjustment is justified.
Scenario: Your company is considering hiring a VP of Sales who has impressive credentials: graduated from Stanford, worked at two successful startups, excellent references, and compelling interview.
Base rate thinking in action:
Reference class: VP-level sales hires at B2B SaaS companies
Base rate research:
Base rate quality: Medium-High
Start at 65%: Before considering anything specific about this candidate, the base rate suggests 65% chance of success.
Individuating information:
Conservative adjustment:
Check for neglect:
Decision: Hire the candidate but design onboarding assuming 75% success rate. Have a 30/60/90 day performance review with clear metrics, because there's still a 25% chance this doesn't work out despite excellent credentials.
Contrast with base rate neglect: Without base rate thinking, you might estimate 90%+ success given the impressive profile, leading to insufficient onboarding support and delayed recognition if issues emerge.
Ignoring base rates entirely: Jumping straight to evaluating specific details without establishing what usually happens. This guarantees overconfidence and poor calibration.
Cherry-picking reference classes: Choosing a reference class that confirms your intuition rather than the most appropriate comparison group. If you want to be optimistic, you'll find an optimistic reference class.
Treating all cases as special: Believing "this time is different" or "this case is unique" so base rates don't apply. Most cases that feel special aren't, and even truly special cases should start with base rates.
Over-adjusting on vivid details: Making large probability shifts based on emotionally compelling but statistically weak information (story about one success, charismatic founder, compelling pitch).
Using base rates without understanding: Mindlessly applying statistics without considering whether the reference class is appropriate or the data is reliable.
Confusing individual prediction with group prediction: Base rates tell you about populations, not individuals with certainty. A 70% base rate means 3 in 10 fail—it doesn't tell you which specific one will fail.
Failing to update base rates: Using outdated statistics when the world has changed (e.g., startup success rates pre- and post-2008 are different).