The Volatility Risk Premium (Short Volatility / Options Selling) — a Real Premium With a Peso-Problem Tail

Implied volatility trades above subsequent realized volatility on S&P 500 options most of the time — sellers of options and variance swaps collect that spread as the volatility risk premium (VRP). It is one of the better-documented premia in this wiki (positive academic literature plus a well-quantified crash record), but its return profile is a small steady credit funded by a rare, unlimited-looking debit: the Feb 2018 "Volmageddon" episode is the numeric anchor for what "tail risk" costs here.

Implied vs Realized Volatility Spread: Historical Average and VIX Premium

- Average spread since 1990: 2–4 percentage points on S&P 500 options, with positive outcomes in approximately 85% of non-overlapping monthly windows — https://volatilitybox.com/research/volatility-risk-premium/ - 1990–1996: VIX overshot realized volatility by 5–7 points, a wider spread than in recent decades — https://www.topstep.com/blog/implied-vs-realized-volatility-the-vix - 2008 is the one negative-spread exception since 1990: realized volatility exceeded the VIX July–November 2008 — https://www.topstep.com/blog/implied-vs-realized-volatility-the-vix - Distribution is right-skewed: calm-market spreads run 1–3 points; fear spikes push it to 8–15 points during VIX surges — https://www.topstep.com/blog/implied-vs-realized-volatility-the-vix - 2010–2019 was VRP harvesting's "golden era", an extended low-vol regime with Volmageddon (Feb 2018) as the only major disruption — https://www.topstep.com/blog/implied-vs-realized-volatility-the-vix - Structural explanation offered in the source: option sellers demand compensation for unlimited risk exposure, option buyers pay for defined, limited risk — this is the proposed reason IV persistently exceeds RV — https://www.topstep.com/blog/implied-vs-realized-volatility-the-vix

Return, Sharpe, and Tail Risk of Systematic Short-Vol Strategies

- Sharpe ratio by decade (VRP harvesting): 1990s = 0.72, 2000s = 0.41, 2010s = 0.81, 2020–2025 = 0.53; overall 1990–2025 = 0.62 — https://volatilitybox.com/research/volatility-risk-premium/ - Short variance swaps: Sharpe 0.38–1.47 depending on the underlying (S&P 500, DAX, crude oil, gold, currency pairs) — https://quantpedia.com/asset-risk-premiums-explained-by-skewness/ - Win rate ~85% of non-overlapping monthly windows, i.e. a high hit rate with losses concentrated in rare crashes — https://volatilitybox.com/research/volatility-risk-premium/ - Feb 5, 2018 "Volmageddon": VIX jumped from 17 to 37 in one day (+116% daily move), the largest percentage move since inception — https://www.sixfigureinvesting.com/2019/02/what-caused-the-february-5th-2018-volatility-spike-xiv-termination/ - XIV lost 96–97% in a single day; a $100,000 position was worth ~$2,700 at the intraday low — https://www.bitget.com/wiki/xiv-stock and https://www.sixfigureinvesting.com/2019/02/what-caused-the-february-5th-2018-volatility-spike-xiv-termination/ - SVXY lost 90–93% on the same day; ProShares then cut its leverage from -1.0x to -0.5x, limiting future maximum drawdowns — https://www.sixfigureinvesting.com/2019/02/what-caused-the-february-5th-2018-volatility-spike-xiv-termination/ - Credit Suisse terminated XIV effective Feb 15, 2018, liquidation value ~$5.99/share — https://www.bitget.com/wiki/xiv-stock - Mechanism of the blowup: falling XIV/SVXY values forced mechanical rebalancing buys of VIX futures, which pushed VIX futures prices higher and triggered further losses — a destructive feedback loop documented in the CFA Institute Financial Analysts Journal — https://rpc.cfainstitute.org/research/financial-analysts-journal/2021/volmageddon-failure-short-volatility-products - Short variance swaps carry highly negative skewness: -4.61 to -5.53 across underlyings — https://quantpedia.com/asset-risk-premiums-explained-by-skewness/

Path-dependent decay compounds the leveraged version

a -1x daily-reset fund down 10% in one day needs an 11.11% gain just to break even — a mathematical drag on top of the market loss, documented for XIV-style products — https://houndstoothcapital.com/2018/05/06/anatomy_of_a_blowup/

Drawdown Magnitude and Position Sizing: What Separates Survivors From XIV Holders

- Maximum drawdown, unfiltered VRP strategies: 28.3%; regime-filtered/hedged approaches cut this to 16.1% — https://volatilitybox.com/research/volatility-risk-premium/ - March 2020 COVID crash: simulated accounts sized aggressively (5% per trade) suffered 41.7% drawdowns; conservative sizing (2% per trade) produced 18.4% drawdowns — https://volatilitybox.com/research/volatility-risk-premium/ - Volmageddon, sizing comparison: leveraged products (XIV/SVXY) lost 93–96%; individual option traders using proper position sizing (2–5% per trade) experienced only 8–12% drawdowns — https://volatilitybox.com/research/volatility-risk-premium/ - Distributional shape: variance risk premium exhibits negative skewness with a long left tail — infrequent but large losses concentrated in stress periods — https://quantpedia.com/asset-risk-premiums-explained-by-skewness/ - Return and skewness move together: strategies with higher variance-premium returns show substantially more negative skewness, indicating the return is compensation for tail risk rather than a separate free lunch — https://quantpedia.com/asset-risk-premiums-explained-by-skewness/ - Crisis variance can run 50–100x normal-market variance, the scale that justifies the large negative convexity built into any short-variance position — https://quantpedia.com/volatility-risk-premium-effect/

What this means operationally

the same premium that destroyed XIV holders left disciplined, modestly-sized option sellers with single-digit-to-low-teens drawdowns in the identical event. The numbers above show sizing/leverage, not "having an edge," as the variable that determined survival in Feb 2018. See Computing a Strategy's Transaction-Cost Threshold — the four-step check that decides whether a documented edge is tradable for how to size a documented edge against realistic costs, and Overfitting and Data-Snooping in Backtests — why the Sharpe ratio you see is not the Sharpe ratio you get before trusting any in-sample Sharpe estimate like the ones quoted here.

Whether It Survives Costs and Is a Real Premium or a Peso Problem

- VRP is documented as persistent and pervasive across asset classes (stocks, bonds, commodities, currencies) and geographies — https://alphaarchitect.com/the-variance-risk-premium-is-pervasive/ - Predictive power: a Federal Reserve FEDS working paper reports the variance risk premium predicts aggregate market returns better than dividend-price and traditional valuation ratios, especially at quarterly-to-annual horizons, explaining >15% of quarterly excess market return variation over the 1990–2005 sample — https://www.federalreserve.gov/pubs/feds/2007/200711/200711pap.pdf - Why it's priced: realized variance is highly correlated with large negative jumps, and the Fed paper's reading is that investors specifically pay to hedge downside volatility exposure — https://www.federalreserve.gov/pubs/feds/2007/200711/200711pap.pdf - Costs bite hardest exactly when the premium is widest: bid-ask spreads in VIX futures and variance swaps widen sharply during volatility stress, directly cutting into premium capture in the periods that would otherwise pay the most — https://volatilitybox.com/research/volatility-risk-premium/ - Peso-problem framing: rare disaster events (October 1987, August 2008, February 2018) are read as the reason the premium exists at all — compensation for a tail risk that was always possible and materializes only intermittently — https://quantpedia.com/volatility-risk-premium-effect/ - Compensation, not free money: the premium is payment for a real, uncomfortable exposure — large losses concentrated in market stress, for which sellers are compensated — https://quantpedia.com/volatility-risk-premium-effect/ - Concentration/liquidity risk compounds the tail: XIV and SVXY together controlled a large share of VIX futures positioning, so their mechanical rebalancing needs amplified the very crash that triggered them — a market-structure risk layered on top of the pure volatility premium — https://rpc.cfainstitute.org/research/financial-analysts-journal/2021/volmageddon-failure-short-volatility-products

Where the evidence is weaker

The decade-by-decade Sharpe ratios and the 28.3%/16.1% and 41.7%/18.4% drawdown figures all come from a single commercial research source (volatilitybox.com), not a peer-reviewed study — treat them as practitioner-attested, not academic-grade (see Proof Regimes: Peer-Reviewed vs Practitioner-Attested vs Unverified — how to grade the source of a claimed edge). The academically-sourced claims here — pervasiveness across asset classes, predictive power for market returns, negative skewness of variance swaps, the peso-problem framing — rest on named sources (Federal Reserve FEDS paper, quantpedia's literature summaries, AlphaArchitect) but the notes do not give this wiki the underlying papers' own sample sizes or t-statistics for those specific numbers, so multiple-testing exposure (see Multiple Testing: Why t>1.96 Is Not Enough — the bar this wiki uses to grade a factor's significance) cannot be assessed from what's in hand. No number here has been checked against realistic net-of-cost backtests beyond the general statement that spreads widen in stress; a precise cost-survival threshold, comparable to what this wiki computes elsewhere (see Transaction Cost Accounting — the arithmetic that separates a real edge from a paper one), is not in the notes.

For the wiki's objective

VRP clears this wiki's bar unusually well on one axis and unusually poorly on another: the effect size, sign, and rough Sharpe range are documented from multiple independent sources across decades (concept + numbers present), but the proof regime is mostly practitioner-attested rather than peer-reviewed, and no explicit net-of-realistic-cost breakeven or multiple-testing accounting is available in these notes. It is a strong case study for why "documented effect" and "actionable edge" are different bars — see What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique — and its Feb 2018 numbers are this wiki's clearest quantitative illustration of a peso problem, comparable in kind (rare catastrophic tail funding an apparent steady return) to what shows up in The Carry Trade (FX and Commodities) — a real premium with a documented crash tail and a post-2008 fifty-percent haircut and in the stat-arb decay documented in Out-of-Sample vs Post-Publication Decay: The Two Numbers That Tell You If a Premium Is Real.

Related

- What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique — VRP is the sharpest example in this wiki of a technique whose *average* return is real and documented but whose survivability depends entirely on sizing against a fat left tail; the evidence bar has to separate "the premium exists" from "you can hold it through a crash." - Computing a Strategy's Transaction-Cost Threshold — the four-step check that decides whether a documented edge is tradable — the notes show costs widen exactly during the stress periods that matter most for VRP; this page has the method for turning that into a numeric breakeven, which is not yet computed here. - Proof Regimes: Peer-Reviewed vs Practitioner-Attested vs Unverified — how to grade the source of a claimed edge — most of the headline Sharpe/drawdown numbers on this page come from a single practitioner research source rather than peer review; use that page's weighting rules before trusting them at face value. - The Carry Trade (FX and Commodities) — a real premium with a documented crash tail and a post-2008 fifty-percent haircut — another premium whose return is widely read as compensation for rare crash risk (a structural peso problem), useful to compare tail magnitude and frequency against VRP's Feb 2018 episode. - Overfitting and Data-Snooping in Backtests — why the Sharpe ratio you see is not the Sharpe ratio you get — the decade-level Sharpe ratios quoted here (0.41 to 0.81) are in-sample descriptive statistics from one source, not a backtest audited for overfitting; read them with that page's cautions.