Does Buying the S&P When Fear Spikes Actually Work? 13 Years Say Yes
Avg forward 20-day return (%)
Every instinct in a volatility spike tells you to get smaller. The screens are red, the VIX is up, the headlines are ugly, and selling feels like risk management. We tested that instinct against 13 years of daily data — roughly 3,400 sessions from 2013 to 2026 — and it is, on average, exactly wrong.
Sort the market by fear and the pattern is monotonic
We ranked every session by how much the S&P 500 had actually been moving (20-day realized volatility — a clean, model-free fear gauge) and split them into five equal buckets. Then we measured the average return over the next 20 trading days from each bucket.
| Volatility regime | Forward 20-day return | Win rate |
|---|---|---|
| Calm (lowest fifth) | +0.32% | 63% |
| Normal (middle) | +1.43% | 75% |
| Panic (highest fifth) | +2.06% | 72% |
Each bucket holds hundreds of sessions, so this is not a small-sample artifact. The forward return in the most panicked fifth is roughly six times the calm fifth — and the win rate goes up, not down. Buying when it feels worst has paid better than buying when it feels safe.
This is the volatility risk premium, and it is well established
None of this is a quirk of one broker's feed. When we re-ran it against the real VIX (implied volatility) instead of our realized-vol proxy, the two agreed tightly — they correlate about +0.81 — and the pattern sharpened: sessions with the VIX near its calm floor (~12) returned about +0.55% over the next month, while sessions with the VIX in panic territory (~28) returned about +2.41% at a 72% win rate, climbing monotonically in between. This is the volatility risk premium: markets overpay for protection when they are scared, and the buyer of that fear is compensated over the following weeks.
The honest caveat: you are buying into drawdowns
Read the caveat as carefully as the edge. A high-fear regime is where the largest single-day losses live. "Buy fear" works on average across many occurrences — it does not mean the next panic won't get worse before it turns. That is the whole nature of the premium: you are paid precisely because it is uncomfortable and occasionally punishing. Practically, that makes this a regime and sizing overlay, not an entry signal:
- Treat elevated fear as permission to scale up into longs you were already going to take — not as a reason to buy blindly.
- Pair it with a price trigger (a capitulation candle, a reclaim of a level) so you are entering strength off the low, not catching every knife.
- Respect the fat left tail with position size. The edge survives drawdowns only if you do.
Two things that look like timing signals but aren't
While measuring this, we checked the popular macro fear gauges — and two of them failed as short-horizon signals, which is worth knowing so you don't build on them:
- Yield-curve inversion does not time a 20-day horizon. Inverted-curve sessions actually returned more over the next month than normal-curve ones. Inversion leads recessions by many months; stocks routinely melt up after it inverts. It is not a sell signal on any tradeable horizon.
- Credit stress confirms, it doesn't lead. Credit ETFs move tightly with equities the same day (correlation strengthens in stress), but deteriorating credit did not predict lower forward index returns. Use it as a same-day confirmation, never as a forward short signal.
Null results like these are as valuable as the edge — they stop you from paying for information that isn't there.
What this means
The exact volatility thresholds, sizing curve, and trigger we run stay in the engine — a published edge is a dead edge. But the durable, transferable lesson is clean:
- Fear is priced, and the price is usually too high. The crowd de-risks into volatility; the premium accrues to whoever is willing to add.
- It's an overlay, not an entry. Fear tells you when the odds tilt your way, not where to buy — the price trigger still does that job.
- Not every scary gauge is a signal. Realized/implied volatility carries a real forward edge; curve inversion and credit stress, on this horizon, do not.
The uncomfortable trade is the paid trade. That is not a slogan here — it is what 13 years of forward returns actually show.
Studies like this become the filters inside a documented playbook — the research → playbook → backtest → live loop, locked to one instrument at a time, rather than a scanner firing on everything.