Recency Bias Investing: Why “What’s Working Now” Is the Worst Signal You Can Follow
Between January 2016 and December 2025, U.S. mutual funds and ETFs returned 9.9% a year. The average dollar actually invested in those same funds earned 8.7%. Morningstar’s 2026 Mind the Gap study put the decade-long shortfall at roughly 12% of total returns — about $3.8 trillion in aggregate, across nearly 23,000 funds. Nobody lost that money to fees or bad funds. They lost it to timing.
Recency bias investing is the mechanism behind most of that gap: the assumption that whatever has happened lately is the best available forecast of what happens next. It feels like paying attention. This post takes the belief apart — what the persistence data actually shows, what the mistiming costs in measurable percentage points, four structural defenses that work better than willpower, and the narrow set of cases where recent information genuinely does carry signal.
The belief: “recent performance is information”
Almost nobody states it this baldly, which is exactly why it survives. It shows up in softer forms: this fund has been on a tear. That sector has been dead money for three years. International has underperformed for a decade, why bother. Each of those sentences quietly converts a historical fact into a forecast.
The psychological machinery is well established. The recency effect — the tendency for the most recently encountered items in a sequence to dominate recall — is one of the oldest reliable findings in memory research, documented in serial-position experiments going back to the 1960s. Your brain isn’t weighting the last twelve months more heavily because it evaluated the evidence and concluded they matter more. It’s weighting them more heavily because they’re easier to retrieve.
And the belief has a plausible-sounding defense: momentum is real, trends persist, “the trend is your friend.” That’s the strongest version of the argument and it deserves a serious answer rather than a dismissal. The answer is that documented momentum effects operate on specific horizons, in specific asset classes, with meaningful transaction costs — and the informal version most individual investors run, which is “buy what’s been good over the last year or two and sell what hasn’t,” is not that. It’s something much closer to noise-chasing.
Why recency bias investing fails: the persistence data
The cleanest test of whether recent performance predicts future performance is to take the winners and check on them later. S&P Dow Jones Indices does exactly this, every year, in its U.S. Persistence Scorecard.
The result is not “weak persistence.” It’s closer to none. Of the top-quartile U.S. large-cap funds as of 2020, not a single one remained in the top quartile through the end of 2024 — a 0% persistence rate over four years. Widening the bar all the way out to merely staying in the top half over five years, only about 4.2% of U.S. funds managed it. Random chance would put that number closer to 6%.
Sit with that for a second. The screen most investors run — sort by recent returns, buy from the top — is selecting on a characteristic with roughly zero demonstrated persistence, and possibly negative. You are not identifying skill. You are identifying which strategy just had its moment, which is a different thing and usually a worse entry point.
| What recency tells you | What the evidence shows | Source |
|---|---|---|
| “This fund has been a top performer, so it’s a good fund.” | 0% of 2020’s top-quartile large-cap funds were still top-quartile four years later. | S&P DJI U.S. Persistence Scorecard |
| “Moving when conditions change is active management.” | Investor dollars earned 8.7%/yr vs. 9.9%/yr for the funds themselves, 2016–2025. | Morningstar Mind the Gap 2026 |
| “In a real crash, sitting still is denial.” | Over 80% of investors who fled to cash in early 2020 would have done better staying put. | Vanguard, Cash Panickers (2020) |
| “Everyone repositions when things get volatile.” | 83% of self-directed retail households didn’t trade at all, Feb 19–May 31, 2020. | Vanguard, Cash Panickers (2020) |
The cost, measured in percentage points
A 1.2-percentage-point annual gap sounds survivable. It is not, because it compounds against you for the entire life of the portfolio.
Run it forward on a $100,000 balance held for thirty years with no additional contributions. At 9.9% you finish at about $1.70 million. At 8.7% you finish at about $1.22 million. The gap — roughly $476,000, or 28% of the ending balance — is the price of a behavior that, in the moment, felt like being responsive to new information.
What makes this so hard to see is that the loss never appears as a loss. There’s no line on any statement labeled “cost of repositioning in 2022.” Each individual decision looked defensible at the time, which is precisely the fingerprint of hindsight bias reshaping the story after the fact — you remember the reasoning as sounder than it was, so the pattern never gets flagged for review.
Want to see what a 1.2-point drag does to your own numbers over 20 or 30 years?
What to do instead: four structural defenses against recency bias investing
Telling yourself to ignore recent performance does not work — the recency effect is a property of how memory retrieves, not a lapse in discipline. What works is changing the structure so recent performance never becomes an input in the first place.
1. Rebalance on a date, not on a feeling. Pick a calendar trigger — annually, or a fixed drift band like five percentage points — and write it down before you need it. This inverts recency mechanically: a date-based rebalance forces you to sell whatever has run and buy whatever has lagged, which is the opposite of what recent performance is telling you to do. The trade is identical in either framing; only the trigger differs, and the trigger is the whole thing.
2. Automate contributions so no decision is required. A fixed monthly transfer that executes without your involvement removes the moment where recency operates. This is the practical case underneath the dollar cost averaging versus lump sum debate — lump sum wins on expected return in the historical data, but automated contributions win on the far more common failure mode, which is not investing at all because now feels like a bad time.
3. Lengthen every chart you look at. If your default view is one year, you are feeding the bias directly. Set every performance screen to ten years or the maximum available. This costs nothing and changes what your intuition is reacting to, which matters more than any amount of resolve.
4. Write the sell rule before you own the asset. One sentence: “I will sell this if X happens,” where X is a fact about the holding — a fee change, a mandate change, a merger — and never a return figure. A pre-written rule is the only version of your judgment that isn’t contaminated by whatever the market did last quarter. Without one, you’ll construct a justification after the fact, and it will feel like analysis. That’s the same trap as confirmation bias, where more research makes you more certain rather than more accurate.
When recency actually is a signal
The myth-busting shouldn’t overshoot. There is a real category of recent information you should act on, and it has a clean defining property: it’s a fact about the vehicle, not about the return.
Fund closures qualify. The same S&P persistence research found roughly 10% of top-quartile funds were merged or liquidated within the following five years — which is both a reminder that past winners are fragile, and a genuine trigger for action when it happens to something you hold. Expense ratio changes qualify. A shift in stated mandate or a strategy drifting away from what you bought qualifies. A tax-relevant change in your own situation qualifies.
What doesn’t qualify: three quarters of underperformance, a sector “having a moment,” a headline about a rotation. The test is simple — if you removed the price chart entirely, would this still be a reason to trade? If not, it’s recency wearing a costume.
There’s also a legitimate version of paying attention to the recent past that has nothing to do with performance chasing: the timing of returns genuinely does matter enormously if you’re drawing down a portfolio rather than building one. That’s sequence of returns risk, and it’s a real structural concern rather than a bias — worth separating carefully from the pattern this post is about.
I’ve been investing in index funds and tax-advantaged accounts for years now without an advisor, and the change that moved the needle most wasn’t learning more — it was writing my rebalancing rule into a calendar entry and refusing to touch the portfolio outside those dates. Writing software for a living probably helped here: my instinct with any system is to look for the state where a bug can occur and eliminate it, and in a portfolio that state is “human is looking at a one-year chart with the ability to trade.” I removed the state. What genuinely surprised me was how often my post-hoc reasoning for a trade I wanted to make was airtight-sounding and, checked six months later, wrong.
Frequently asked questions about recency bias investing
Is momentum investing the same thing as recency bias?
No, though they’re easy to confuse. Academic momentum strategies are formally defined — specific lookback windows, specific holding periods, systematic rules applied across a broad universe, with transaction costs accounted for. Recency bias is unstructured and reactive: you notice something has done well and you buy it. The distinction that matters is whether the rule existed before the observation. If you defined it in advance and apply it mechanically, it’s a strategy. If the recent performance is what generated the idea, it’s the bias.
How long a track record should I look at before judging a fund?
For predicting future performance, the honest answer from the persistence data is that no length of track record does the job well — 0% of 2020’s top-quartile large-cap funds were still there four years later. Track records tell you what a fund did, not what it will do. The attributes with actual predictive power are structural and boring: expense ratio, turnover, tracking difference, and whether the mandate matches what you want to own. Those you can evaluate on day one.
If markets drop 30%, isn’t selling just risk management?
It can be, if the sale executes a rule you wrote before the drop, or if your actual time horizon has changed. Otherwise it’s recency with better vocabulary. Vanguard’s data from the February–May 2020 decline is direct on this: under 0.5% of its clients moved to cash, and more than 80% of those who did would have finished better off staying invested — markets rose 36% off the March lows by the end of May. The people who “managed risk” mostly locked in the loss and missed the recovery.
Key takeaways
- Recent performance has close to zero demonstrated persistence — 0% of 2020’s top-quartile U.S. large-cap funds remained top-quartile through 2024.
- The measured cost of mistimed moves is roughly 1.2 percentage points a year: 8.7% earned by the average dollar versus 9.9% earned by the funds, 2016–2025.
- On a $100,000 balance over thirty years, that gap is roughly $476,000 — about 28% of the ending value.
- Structural defenses beat willpower: date-based rebalancing, automated contributions, ten-year default chart views, and a sell rule written before you buy.
- Recent information about the vehicle — fees, mandate, closure, merger — is actionable. Recent information about the return generally is not.
Photo by Frames For Your Heart on
Unsplash
