What is recency bias?
How strongly the latest information pulls on a verdict, where the pull holds firm, and what removed it in one study of auditors.
Answer to the question
What is recency bias?
Recency bias is the tendency to give the most recent information more weight than it deserves when judging a whole record. In a 2025 study of running averages, people weighted recent items strongly and early ones less, and in a study of staff auditors a simple self-review step eliminated the effect.
Key takeaways
- In a 2025 study by Minhee Yoo and colleagues, people reporting a running average gave recent items a strong weight and early items a smaller one, across perceptual and value-based tasks.
- Recency bias is about the weight given to the latest information when a record is summed up, which is related to but different from better recall of the last items in a list.
- Experience did not reduce order and recency bias under heavy information load in two experiments with professional accountants, including 87 insolvency practitioners (Arnold and colleagues, 2000).
- A simple self-review technique eliminated recency in the going-concern judgments of staff auditors (Ashton and Kennedy, 2002).
- After a rare and distressing event, people in a field study judged the risk as lower yet behaved more cautiously, so estimates and choices moved in opposite directions (Barron and Yechiam, 2009).
People keeping a running average give the last items far more weight than the first, though an average treats every item alike. A recent study by Minhee Yoo and colleagues found this in both perceptual and value-based tasks. That tilt toward the latest information, applied to a whole record, is the core of recency bias.
The pull is easy to miss because recent information often does deserve attention. The trouble starts when it displaces what came before, as when a year-end review is written mostly from the final few weeks, or a product is judged from its newest comments alone. Those are illustrations, not cases, and the studies below show how the pattern behaves when it is tested.
What recency bias is
Recency bias is the habit of letting the most recent information outweigh the rest of the record when a person sums that record up into one judgment, such as an average, a rating or a risk estimate. It sits beside the [recency effect](https://thewhyweproject.com/articles/what-is-the-serial-position-effect) in memory, the finding that the final items of a list are recalled well, but the two answer different questions. The recency effect asks what is remembered from a sequence. Recency bias asks what is weighted when the sequence is turned into a verdict. A memory that favours the end can feed a judgment that favours the end, which is why the terms travel together.
Something that happened last week comes to mind faster than something that happened three years ago, so a verdict built from what is in view leans toward the recent. Daniel Kahneman's Thinking, Fast and Slow supplies the route: people judge how common or how likely something is by how easily examples come to mind, which is the availability heuristic, and the mind builds its verdict from the evidence in view without asking what is missing. That would produce exactly the tilt that experiments measure.
A trip is remembered through its final days, as set out in Why You Remember the End of a Holiday, while the early items in a sequence have their own advantage, explored in Why the First Option Gets Picked. Recency bias is the late half of the order effects, with the early half pulling the other way.
The latest weighs most
That study asked participants to report continuously the average strength of a series of stimuli as it unfolded. Some tasks were perceptual and others were value-based. The weighting curves, which show how much each moment of the sequence counted in the running report, looked highly similar across both kinds of task: a strong recency bias alongside a smaller primacy bias. Averaging is the job in which every moment ought to count equally, so a steep rise in weight toward the end is a tilt, and the authors conclude that people consistently give recent information too much weight when judging the past.
The authors draw a line between two kinds of judgment. A summary of the past is usually meant to treat early and late information as equally important, while a forecast of what comes next leans more on recent information. The tilt they measured sits in the first kind, where an even weighting is the usual aim.
The brain-imaging part of the study followed where evidence was tracked. Evidence-tracking activity appeared in the cuneus in both tasks, and in the left dorsolateral prefrontal cortex in the value-based task. Participants whose primacy bias was stronger than their recency bias showed more activity in regions involved in cognitive control, which led the authors to suggest that overcoming the recency tilt may depend on cognitive control.
For everyday summaries the finding is concrete. Someone asked for the average of twelve monthly figures would probably produce a number closer to the last few months than the arithmetic gives, and the same drift would apply to a score for a season, a project or a supplier. The size of that drift outside the laboratory is not measured by this task, but the direction of the pull is clear, and it runs toward whatever happened last.
When the tilt gives way
The strongest test of whether expertise protects against the tilt came from professional accounting. The belief-adjustment model predicts that recency effects will always appear in complex decision domains, whatever a person's experience. Vicky Arnold and colleagues built complexity from heavy information load while keeping task familiarity high, then ran two experiments: a going-concern decision, a judgment about whether a business can keep operating, with experienced partners and managers, and an insolvency case with 87 insolvency practitioners. Experience did not reduce order and recency bias under heavy information load.
Lighter conditions change the picture. Robert Ashton and Jane Kennedy found recency in the going-concern judgments of staff auditors, and a simple self-review technique eliminated it. Auditors who reviewed their own reasoning were also less likely to choose report options that conflicted with their judgment.
Greg Barron and Eldad Yechiam ran three laboratory experiments and a field study on rare events. Earlier work showed that people overestimate rare events when they make judgments and underweight them when they decide from experience, and their results suggest both can happen at once. Part of the gap, they propose, is a contingent recency effect: estimates showed negative recency, so a recent rare event lowered the judged risk, while choices showed positive recency. In the field study, right after a rare and distressing event, people judged the risk as lower yet became more cautious in their behaviour.
Recency bias behaves like a default weighting and not a law. It is strong when a record is summarised, it persists for experienced people under heavy load, a deliberate second look can remove it, and it can point one way in an estimate and the other in a choice.
Conditions that move it
| Condition | What happened to the recency tilt | Study |
|---|---|---|
| Running average of a sequence | Strong weight on recent items, smaller weight on early ones | Yoo and colleagues, 2025 |
| Heavy information load, experienced professionals | Experience did not reduce the tilt | Arnold and colleagues, 2000 |
| Staff auditors with a self-review step | Tilt eliminated | Ashton and Kennedy, 2002 |
| Estimate of risk versus choice after a rare event | Estimate moved one way, behaviour the other | Barron and Yechiam, 2009 |
Read down the rows, the tilt appears when a record is summed up, persists for experts under heavy load and is absent after a deliberate second look, while the fourth row differs because it compares two outputs from the same people, an estimate and a choice. Ashton and Kennedy's result is the one that points to a habit: going back over the evidence once, deliberately, removed the tilt in their auditors. For a review or a rating, the everyday form of that step is rebuilding the verdict from the whole record. Each period then counts equally.
Common questions
What is recency bias?
Recency bias is the tendency to give the most recent information more weight than it deserves when judging a whole record. In a 2025 study of running averages, people weighted recent items strongly and early items less.
Is recency bias the same as the recency effect?
They are related but not identical. The recency effect describes better recall of the last items in a list, while recency bias describes extra weight on the latest information when a verdict is formed.
Does recent news always push judgment in the same direction?
No. After a rare and distressing event, people in one field study judged the risk as lower while behaving more cautiously, so estimates and choices moved in opposite directions.
Can experience or a second look reduce recency bias?
Experience alone did not reduce it under heavy information load in professional accounting. A simple self-review step eliminated it in staff auditors in a separate study.
Sources
Kahneman, D. Thinking, Fast and Slow. Seed book; vault hub Books/Thinking Fast and Slow.
Yoo, M., Bahg, G., Turner, B. M., & Krajbich, I. (2025). People display consistent recency and primacy effects in behavior and neural activity across perceptual and value-based judgments. Cognitive, Affective, & Behavioral Neuroscience, 25(4), 923-940. doi:10.3758/s13415-025-01285-1 https://doi.org/10.3758/s13415-025-01285-1
Arnold, V., Collier, P. A., Leech, S. A., & Sutton, S. G. (2000). The effect of experience and complexity on order and recency bias in decision making by professional accountants. Accounting & Finance, 40(2), 109-134. doi:10.1111/1467-629X.00039 https://doi.org/10.1111/1467-629X.00039
Ashton, R. H., & Kennedy, J. (2002). Eliminating recency with self-review: The case of auditors' going concern judgments. Journal of Behavioral Decision Making, 15(3), 221-231. doi:10.1002/bdm.412 https://doi.org/10.1002/bdm.412
Barron, G., & Yechiam, E. (2009). The coexistence of overestimation and underweighting of rare events and the contingent recency effect. Judgment and Decision Making, 4(6), 447-460. doi:10.1017/S1930297500003983 https://doi.org/10.1017/S1930297500003983