Recency bias: why investors keep buying at the top

Recency bias is the mind's habit of treating the recent past as the best available forecast of the future. It is not stupidity — it is an efficient rule for most of life, where the weather tomorrow really does resemble today. Applied to markets, where returns mean-revert, it is precisely backwards.

The machinery that turns it into a loss

Step 1 — a category performs. Small caps, or a sector theme, run for two or three years.

Step 2 — the numbers become visible. The trailing three-year return in every table now looks extraordinary. Star ratings, which are computed from past returns, rise. See star rating.

Step 3 — the story arrives to explain it. A narrative forms that makes the past returns feel inevitable and the future obvious.

Step 4 — products are launched into the demand. New schemes are marketed into the theme, at exactly the point where the run is most visible. This is why NFO launches cluster around peaks — they are a coincident indicator of the top, not a signal of opportunity.

Step 5 — money arrives. Retail flows into the category surge.

Step 6 — the cycle turns, the category spends years going nowhere, and the same investors exit near the bottom to buy whatever has performed recently.

Every step is rational in isolation. The outcome is buying high and selling low with total conviction.

The evidence, in one number

The industry measure of this is the gap between investor return and fund return. The fund's return is what a unit held for the whole period earned. The investor return is what the average rupee in that fund actually earned, weighted by when it arrived.

Investor return is persistently lower, and the gap is widest in the most volatile categories — sectoral, thematic and small-cap funds. That gap is not fees or taxes. It is timing, and it is the cost of recency bias measured in rupees.

You can observe your own version of it directly: your personal XIRR versus the fund's CAGR. Some of the difference is simply that your later instalments have had less time. Some of it is when you chose to add money.

Where it shows up in your decisions

• Choosing a fund from a one-year or three-year league table. The window is the finding, not the fund. See rolling vs trailing returns.
• Increasing the SIP into whatever just ran. Feels like conviction, is actually chasing.
• Abandoning an underperforming fund after three bad years — which is frequently the moment a sound style-based strategy is about to work. See contra funds.
• Believing the market's current direction will continue. In a bull market, every fall looks like a buying opportunity. In a bear market, every rally looks like a trap. Neither belief has predictive content.
• Extrapolating a return assumption from the last five years. A retirement plan built on the last five years' returns is a plan built on a regime.

What actually reduces it

1. Pre-commitment. The most effective defence, because it moves the decision to a moment when you are calm. A written investment policy — target allocation, rebalancing bands, the conditions under which you would sell — takes fifteen minutes and removes most of the opportunities to act on recency.

2. Automation. A SIP invests the same amount regardless of the last three years' returns. That is not merely convenient; it is a structural defence against your own extrapolation. See rupee-cost averaging.

3. Rebalancing on bands. Mechanically sells what has run and buys what has not — the exact opposite of recency, executed without requiring you to feel contrarian. See rebalancing.

4. Look at longer windows. Ten-year and fifteen-year rolling returns show a distribution rather than a point, which makes the last three years look like what they are: one sample.

5. Read your own history. Keep a note of what you thought at the time of each significant decision. Nothing calibrates faster than reading your own March-2020 note in 2026.

6. Change the question. Instead of "how has this performed?", ask "what would have to be true for this to keep performing?" The second question is answerable and usually deflating.

The version that catches experienced investors

Recency bias is not only about chasing winners. Its subtler form is assumption drift: the return assumption in your retirement spreadsheet creeping up after a good decade, the equity allocation you are "comfortable" with rising after three good years, the risk you believe you can tolerate being re-estimated from a period in which nothing bad happened.

The test: what number were you using five years ago, and why did it change? If the answer is "the market did well", you have found it.

Pitfalls to avoid

• Choosing funds by recent rank. The strongest single expression of the bias.
• Adding money to whatever just ran. Especially into thematic funds, where the launch cycle is built on this.
• Selling a fund purely for three bad years. Distinguish "the fund changed" from "the market changed" — only the first is a reason. See when to sell.
• Raising your return assumption after a good run. Quietly the most expensive version.
• Treating a five-year track record as a full picture. It may contain no drawdown at all.
• Confusing a story with evidence. The narrative always arrives after the returns.

Key takeaway

Recency bias makes the last three years feel like a forecast, and the entire industry's product cycle is built around it — themes are launched after they have run, tables rank by past return, and the money arrives at the top. The measurable cost is the persistent gap between what funds return and what investors in them return, widest in exactly the categories that run hardest. The defence is structural, not emotional: automate the contributions, rebalance on written bands, judge funds on rolling rather than trailing returns, and check whether your own return assumptions have quietly drifted upward since the last good decade.