Volatility & Dispersion Research/Cboe DSPX × S&P 500
For a decade the S&P 500 Dispersion Index behaved like every other fear gauge — high in the wreckage, quiet at the peaks. Since 2024 that rule has broken. The index sits near record highs and so does expected single-stock dispersion, a pairing that had essentially never occurred before.
01 The thesis
The Cboe S&P 500 Dispersion Index (DSPX) measures the market’s 30-day expectation of how far individual constituents will scatter around the index — option-implied idiosyncratic risk, effectively the inverse of implied correlation. High DSPX means traders are paying up for big single-name moves even if the index itself stays calm.
Historically that reading spiked in stress: the constituents come apart when something breaks. It faded to the high-teens whenever the S&P ground higher on low, correlated, orderly gains. Today both are extended simultaneously — the index a whisper from its high, DSPX in the 99.8th percentile of a twelve-year record. Your read is correct: this configuration is close to unprecedented.
02 The relationship
Plotted side by side, the divergence is visible without any statistics. From 2016 to 2023 DSPX (amber) chopped sideways in a 15–35 band while the S&P (blue) trended up — the two lines crossing, uncorrelated in level. From early 2024 they begin to ascend together, DSPX stair-stepping to fresh records right alongside a record index.
But you asked specifically for the rolling correlation of returns, and it tells a subtler, complementary story. Day to day, DSPX still behaves like a volatility gauge: when the S&P sells off, dispersion tends to pop. The 60-day return correlation spends most of its life negative — averaging about −0.21 since 2016 — and lurches to −0.8 in every real drawdown (2019, 2022, the April 2025 tariff shock).
So both facts hold at once. On any given day dispersion still trades inversely to the tape. Yet the resting altitude of both series has drifted up together. The regime shift lives in the levels, not the day-to-day beta.
03 The break
Define the joint state cleanly: days where the S&P is within 2% of its running all-time high and DSPX sits in the top quartile of its full-history distribution (≥ 29.5). Count how many of each year’s trading days qualify.
The companion number is the amber line: what level does dispersion hold when the index is at its high? In 2017 it was 18.5. In 2026 it is 38.1 — the price of expected single-stock chaos at the top has roughly doubled even as the index looks placid.
| Year | Avg DSPX at highs | Joint-high days | % of year |
|---|
04 Why
Mechanically, elevated DSPX at a calm index requires one thing: high implied single-stock volatility paired with low implied index correlation. Stocks are expected to move violently, but not together — so the moves net out at the index level. Several structural shifts since 2023 all push in exactly that direction. These are hypotheses, ordered by how directly they bear on the option-implied inputs DSPX is built from.
Leveraged and inverse single-name ETFs (2×/3× on Nvidia, Tesla, and dozens more), plus the explosion of single-stock and 0DTE options, mechanically bid up individual-name implied vol. DSPX is long single-stock vol and short index vol by construction — this is the most direct pipe into a higher reading.
A handful of AI/semiconductor names now drive enormous, stock-specific gaps on earnings and headlines. That is idiosyncratic risk by definition — huge single-name variance that partially offsets at the index because leadership rotates. Precisely the dispersion signature: big names moving big, but not in lockstep.
Selling index volatility to buy single-stock volatility has gone from niche to mainstream desk strategy. The flow itself lifts single-name IV relative to index IV — a partly self-reinforcing bid under the exact spread DSPX measures. Popularity of the trade can keep the index elevated well beyond fundamentals.
Systematic vol-selling, heavy index-level premium harvesting, and low realized correlation keep headline SPX/VIX calm. When constituents move independently, their vols diversify away at the index — so the index can print serene while the dispersion premium stays rich. Calm index and high dispersion are two faces of low correlation.
Money increasingly expresses views through thematic baskets and sector rotation rather than beta. That churns relative performance — winners and losers within the index — without moving the aggregate much, feeding realized and implied dispersion while leaving the index trend intact.
Each of these keeps the numerator (single-stock implied vol) high while capping the denominator (index implied vol). None of them requires the market to fall — which is exactly why dispersion can now sit at a record with the S&P at one too.
05 What it means
The old playbook — “high DSPX = something is breaking” — is miscalibrated for this regime. Elevated dispersion at the highs is no longer a straightforward risk warning; it is partly a structural artifact of how the market is now traded: through leverage on single names, through the AI complex, and through a dispersion trade that has become consensus.
Two watch-items follow. First, a genuinely useful stress signal now needs a level-change or the return-correlation collapse in Exhibit 2, not the level alone. Second, the flip side of a crowded, self-reinforcing dispersion bid is fragility: if the trade unwinds — index correlation snapping back toward 1 in a broad de-risking — both series could reprice hard and fast, and the comfortable joint-highs regime would end abruptly. It has happened in miniature already, every time the amber line in Exhibit 2 dives to −0.8.