Recency Bias Is Why Every Trend Looks Permanent While It Lasts
Recent events dominate expectations far beyond their informational value. It is why investors buy after strong performance and sell after weak, reliably and at scale.
The Pattern
Recency bias is the tendency to weight recent observations more heavily than older ones when forming expectations. What happened last quarter feels more informative about the future than what happened over the last twenty years.
Sometimes that is correct, since conditions do change and recent data can carry genuine signal. More often it substitutes a small recent sample for a large historical one, and small samples are dominated by noise.
The Flow Evidence
The most direct evidence is in where money goes. Fund flows follow performance with a lag: capital arrives after strong returns and departs after weak ones.
The consequence is a persistent gap between the return a fund reports and the return its average investor actually earned. The fund's number assumes holding throughout. The investor's number reflects buying after the rise and selling after the fall.
The published return belongs to the fund. The realised return belongs to the investor, and the difference between them is mostly a record of when people decided to arrive and leave.
Mistaking a Cycle for a Regime
The costlier version operates on longer horizons. A decade of low inflation makes inflation feel structurally dead. A decade of falling rates makes rising rates feel implausible. A long stretch of one asset class outperforming makes the alternative look permanently inferior.
Each of these has produced widespread confident error. Portfolios were positioned for a continuation of conditions that had persisted long enough to seem like properties of the world rather than phases of a cycle.
| Recent experience | Belief formed | What it was |
|---|---|---|
| Extended low inflation | Inflation is solved | A phase |
| Decades of falling rates | Rates only go down | A long trend |
| One region outperforming | The other is structurally worse | A cycle |
| Low volatility stretch | Risk has diminished | The calm phase |
Why Professionals Are Not Immune
Institutional processes amplify it rather than correcting it. Manager evaluation on three year performance windows encourages hiring after strong periods and firing after weak ones. Risk models calibrated on recent data show low risk precisely when recent conditions were calm.
Value at risk computed from a quiet trailing period understates exposure at exactly the moment position sizes were built on that understatement. The model is not broken. It is answering a question about the recent past that is being used as a forecast.
The Availability Overlap
Recency interacts with how easily examples come to mind. A recent, vivid, heavily covered event feels more probable than a statistically comparable one that happened earlier or received less attention.
After a market crash, investors overestimate the likelihood of another. After a long calm stretch, they underestimate it. The underlying probability moved far less than the perception did.
The Countermeasures
Look at long series deliberately. Any conviction about how markets behave should be checked against the longest available history, not the period you personally experienced.
Write down expectations with dates attached. Memory reconstructs past beliefs to match what subsequently happened, so an undated recollection of what you thought is unreliable evidence.
Rebalance mechanically. A rule that sells what has risen and buys what has fallen is structurally opposed to recency, which is why it feels wrong to execute and why executing it anyway is the point.
And treat any sentence beginning with the idea that things are different now as requiring evidence rather than supplying it.
The Bottom Line
Recency bias makes the last few years feel like a description of how the world works. It drives performance chasing, understates risk after calm periods, and turns cycles into apparent regimes. Long historical series, written and dated expectations, and mechanical rebalancing are the practical defences, and all three work by removing recent experience from the decision.