Recency Bias in Stocks — Avoiding the Short-Memory Mistake
The S&P 500 is up 24% over the last 12 months. Your gut says: “stocks are safe.” In March 2020 the S&P had lost 34% in four weeks. Back then your gut said: “stocks are doom.” Both guts were wrong — they over-weighted the last few weeks. That is recency bias: your brain extrapolates the recent past into the future. Here are the mechanics, the most expensive recency mistakes of the past 25 years, and a 4-step routine that breaks the pattern.
What recency bias is
Tversky/Kahneman 1973 (“Availability Heuristic”) showed: people estimate event probability higher when examples come to mind easily. Since recent events are more available, they get systematically overweighted. In stocks: you extrapolate the past 30-90 days into the next 12 months — even though statistically the opposite is closer to truth (mean reversion).
Studies (Greenwood/Shleifer 2014, Choi/Robertson 2020): retail investors estimate the next 12-month return at roughly 1.5 to 3 times the last 90-day return. Statistically wrong — the correlation between 3-month and next-12-month return is on average weakly negative (mean reversion).
Five expensive recency mistakes from 25 years
- Tech mania 1999-2000: Cisco had compounded 50%+ p.a. for five years. Investors projected linear continuation. Cisco lost 90% over two years. Recency had ignored the 25-year average tech cycle (~8% p.a.).
- Real estate 2007 (US): US homes only went one direction 1995-2006 — up. Famous consensus: “home prices don’t fall nationally.” That was an 11-year recency claim. 2007-2012, home prices fell 27% nationally.
- Bond recency 2009-2020: Eleven years of falling rates → bonds delivered ~5% p.a. total return. Retail extrapolated into the 2020s. Reality: 2022 saw 10-year US Treasuries lose 17% — worst year since the 1920s.
- NASDAQ 2021 vs 2022: 2020-2021 NASDAQ ran +88% over two years. Recency investors raised tech allocation at the highs. 2022 NASDAQ −33%. Recency trade typically cost 25-40% on the tech sleeve.
- Crypto 2021-2022: Bitcoin from $4,000 to $69,000 in 18 months. Recency trades bought at $60,000. Bitcoin fell to $16,000 — the previous 18 months were entirely irrelevant for the next 12.
How much does recency cost you?
| Study | Finding | Return drag |
|---|---|---|
| Dalbar QAIB 2024 | US retail investors underperform market by 1.7% p.a. on average | −1.7% p.a. |
| Morningstar Mind-the-Gap 2024 | ETF investors lose ~1% p.a. through bad timing | −1.0% p.a. |
| BMI data 2024 | 87% of allocation switches follow 90-day performance | structural −1 to −3% p.a. |
| Greenwood/Shleifer 2014 | Expected returns most correlated with recent return, not future | negative predictive power |
Compounded over 30 years, 1.5% p.a. recency drag amounts to roughly 56% less terminal wealth. Recognising and neutralising recency is one of the largest behavioural edges available.
The anti-recency routine in 4 steps
- 10-year-chart obligation: Before any stock analysis, open the 10-year chart, not the 1-year chart. You see full range, not just the recent trend. Tilray, Wirecard, Bayer look very different on a 10-year chart vs a 1-year chart.
- Mean-reversion check: Test whether the stock/market is significantly above or below the 10-year average. P/E vs 10-year median, drawdown vs mean. Currently more than 1 standard deviation above median = recency phase likely.
- Contrarian read: Actively read 1 pro-thesis and 1 contra-thesis per position. Recency mode confirms only the pro-thesis. Contrarian reading forces the brain out of the tunnel.
- Allocation brake: Wait 14 days before any larger allocation change and re-check. If you still hold the same view, it’s not a recency reaction. If the urgency fades — you were in recency mode.
Recency in your own strategy
Recency hits not only stock selection but strategies themselves. After five years of growth-stock outperformance, investors don’t switch to value because “the strategy works”. Until it doesn’t. Classic examples:
- Growth vs value: 2015-2020 growth +120% vs value. Investors pushed money into growth. 2021-2022 value outperformed by 25 points.
- US vs international: 2010-2021 US outperformed international by 7% p.a. “US dominance” became consensus. 2022-2024 international caught up by 3% p.a.
- 60/40 vs 100% stocks: 2009-2021 100% stocks beat 60/40 strongly. Investors cut bonds. 2022 100% stocks −18%, 60/40 only −10% — bonds delivered the drawdown buffer.
Rule of thumb: don’t switch strategies based on the last 5 years’ performance — use 15-30 year median performance instead. The latter changes slowly, the former constantly.
Pros & cons: switching recency off entirely
- Avoids top buys and bottom sells
- Stabilises allocation against market mood
- Enables counter-cyclical rebalancing
- Real trends (Apple 2010s, NVIDIA 2020s) get filtered out by mean-reversion
- Some sector shifts are structural, not mean-reverting (Industry 4.0)
- Pure mean-reversion strategy misses the compounders
Solution: mean-reversion at the market level (allocation), trend-following at the stock level (single positions). That neutralises recency on allocation and uses it for stocks.
Common questions
Is recency bias the same as trend-following?
No. Trend-following is rule-based with stops. Recency is intuitive following without stops. Both start similarly, but trend-following has exit logic (e.g. 200-day moving average), recency has none.
Does dollar-cost averaging help?
Yes, very strongly. Savings plan = mechanical buying regardless of market moves. Immune by construction to recency. Weakness: investors pause savings plans in crises. So: once set up, never pause.
What is the recency filter for ETFs?
No per-stock recency on ETFs. But investors switch ETF styles based on 90-day performance (tech ETF 2021, bank ETF 2024). Filter: never switch into the style that outperformed last quarter. If anything — into the one that underperformed.
How long is “recent past” in recency terms?
Empirically 30-90 days. 30 days: strong. 90 days: medium. 12 months: weaker. 3 years: barely. So: checking 3-year charts before decisions removes ~80% of recency.
Recency in sector rotation?
Very strong. Buying only AI stocks in 2024 because they’re up 60% in 2023-2024 is recency mode. Real sector trades have a fundamental rationale (cashflows, market growth), not last-12-month performance.
What about crypto recency?
Maximum recency effect. Bitcoin correlation data: 90-day performance explains 75-85% of retail sentiment. The famous 4-year cycles aren’t mystic — they’re pure recency-driven buyer waves.
Related Hubs: Investor Glossary | Legendary Quotes
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.