Correct for the recency amplification of availability

Recent events feel more probable than they are — apply an explicit recency discount.

Why it works

Recent events are easier to retrieve because they are stored in more active, higher-fidelity memory traces. This makes them feel more common and more predictive than they are. After a market crash, people overestimate the probability of another crash; after a car accident, they overestimate driving risk. Applying an explicit recency discount asks whether the event’s frequency estimate would be the same if it had happened three years ago instead of last week.

How to do it

1. When an estimate is informed by a recent event, ask: "If this had happened three years ago instead of last week, would I still estimate the same probability?"
2. If the answer is no, your estimate is partially a recency artifact — adjust toward the longer-run frequency.
3. Check: is the recent event representative of a real trend, or an outlier that was followed by regression to the mean?
4. Give extra weight to multi-year base rates over single salient events when estimating future probabilities.

Evidence

Research on investor behavior after market crashes, crime estimates after news coverage, and insurance purchases after natural disasters all document availability-driven overestimation of probability in the period immediately following a vivid event. (observational)

Not all recency effects are biases — recent data sometimes genuinely updates the probability of a changing-rate process. The question is whether the trend is real or driven by a one-time event.

Sources

• Hertwig et al. (2004), decisions from experience and the effect of rare events, Psychological Science

Common mistake

Dismissing the recency artifact while still building plans around the recent event — the explicit acknowledgment of the bias does not automatically correct the underlying estimate.