Guide

Variance risk premium explained

Harbor Capital's balanced sleeve earned steady carry from bond coupons and equity dividends, but the risk team noticed something else in the data: S&P 500 implied volatility from one-month at-the-money options averaged 3.2 percentage points above realized volatility over rolling 21-day windows from 1990 through 2019. That gap is the variance risk premium (VRP) — compensation for sellers who supply crash insurance when investors overpay for protection. Harbor allocated 3% of the portfolio to a disciplined short-volatility overlay (delta-hedged strangles on the index ETF). The sleeve added roughly 1.1% annualized before costs over the next four years, then lost 18 months of gains in three weeks during March 2020 when realized vol spiked above 80% while implied had already repriced. The refactor: cap overlay at 2% risk budget, add a crisis regime kill switch, and pair short vol with long VIX call convexity rather than naked premium harvesting.

The VRP is one of the most studied return premia in empirical finance: on average, options imply more variance than equities subsequently realize. It is not a free lunch — tail events can overwhelm years of carry — but understanding implied vs realized measurement, why the premium exists, and how to harvest it with explicit risk budgets separates systematic allocators from yield-chasers. This guide defines variance and volatility formulations of the premium, explains measurement with VIX and variance swaps, walks through Harbor Capital's overlay refactor, compares harvest techniques in a decision table, lists pitfalls, and ends with a production checklist alongside our options greeks, GARCH modeling, and covered calls guides.

Implied vs realized volatility: the premium defined

Realized volatility (historical vol) summarizes how much an asset actually moved over a past window — typically annualized standard deviation of log returns. Implied volatility is the breakeven vol input to an option pricing model (often Black-Scholes) that matches the option's market price. It is forward-looking: traders' consensus about uncertainty until expiry.

The volatility risk premium at horizon T is commonly stated as:

VRP_vol = IV_T − RV_T

where IV is implied vol from options expiring in T and RV is realized vol over the same forward window (measured ex post for backtests). Researchers often work in variance units because variance is additive across time under idealized assumptions:

VRP_var = IV² − RV² (both annualized).

Positive VRP means implied exceeds realized — option sellers earned more premium than justified by subsequent moves. The premium is conditional: it is positive on average across many months but sharply negative during crises when realized vol explodes past implied.

Why the premium exists

Several mechanisms reinforce a positive average VRP:

Insurance demand — pension funds, structured products, and risk-averse investors buy puts and calls for protection, bidding up implied vol beyond actuarial fair value.
Leverage and margin constraints — investors who cannot lever equities directly use options; dealers hedge and charge a spread.
Volatility clustering — as documented in GARCH models, high-vol periods cluster, so implied vol rises ahead of realized vol but mean-reverts slower than spot prices.
Jump risk and skew — index options embed negative skew; out-of-the-money puts trade rich, lifting index implied vol above subsequent realized moves that exclude tail jumps.

None of these guarantee profits: the premium compensates for bearing negative skew — many small wins, occasional catastrophic losses.

Measuring the VRP in practice

Academic studies use different proxies; production systems should fix one methodology and stick to it.

VIX minus realized S&P 500 vol

The VIX is a 30-day implied vol index from S&P 500 options. A simple signal compares VIX to 21- or 30-day realized vol on SPX or SPY. Caveats: VIX is 30-day; realized windows must align; VIX uses a model-free blend of strikes, not a single ATM quote; roll periods matter for P&L but less for signal research.

Variance swaps and vol swaps

A variance swap pays the difference between realized variance and a fixed strike set at trade inception. The strike embeds the market's implied variance; settlement reveals realized. Portfolio managers use variance swap marks for clean VRP attribution without delta-hedging noise from options.

Cross-sectional and term-structure VRP

VRP varies by tenor: one-month index options often show a larger premium than six-month. Single-stock implied vol can exceed realized while index VRP is flat — idiosyncratic earnings risk inflates single-name IV. Document whether your mandate harvests index VRP, single-name richness, or term-structure carry (selling front-month, buying back month).

Estimation hygiene

Use overlapping windows consistently or non-overlapping months to avoid autocorrelation in t-stats.
Annualize realized vol with the same convention as implied (252 trading days, log returns).
Exclude crisis windows from “average premium” marketing if your risk budget must survive them.

Harvest strategies: how allocators capture VRP

Selling volatility is economically equivalent to collecting the VRP, but implementation choices change skew exposure, capital efficiency, and tail risk.

Covered calls and put writing — retail-accessible; embed equity delta. See our covered calls guide. Premium is partly VRP, partly income for capping upside.
Short strangles / iron condors on indices — purer vol bet with defined wings; needs margin and active gamma/vega management.
Delta-hedged option selling — Harbor's approach: sell strangles, hedge delta daily, isolate vega exposure. Reduces directional bleed but increases transaction costs.
Short volatility ETPs (VXX, SVXY) — packaged exposure with roll and contango costs; convenient but path-dependent decay is not the same as steady VRP.
Risk-managed vol targeting inverse — scale short-vol notional inversely to recent realized vol (see vol targeting); reduces size into spikes.

Systematic CTA and multi-strategy funds often allocate 5–15% risk budget to vol premia sleeves with hard drawdown stops. Retail portfolios more commonly use overwrite (covered call) indices at 30–50% notional.

Worked example: Harbor Capital short-vol overlay refactor

Harbor's policy portfolio was 55% global equities, 35% bonds, 10% alternatives. The team sought diversifying return without adding equity beta.

Version 1 (2016–2019)

3% allocation to delta-hedged one-month SPX strangles, 15-delta wings, rolled monthly.
Entry filter: sell only when VIX > 21-day realized vol by at least 2 vol points.
Result: +1.1% CAGR contribution, Sharpe ~0.9 on sleeve, max sleeve drawdown −6%.

March 2020 stress

Realized vol exceeded 80%; short gamma losses overwhelmed two years of carry in three weeks.
Implied had repriced higher, but delta-hedge slippage and gap risk broke the model.
Lesson: average VRP statistics hide conditional left-tail exposure.

Version 2 (post-2020)

Reduce sleeve to 2% risk budget; vol-scale notional by inverse 60-day realized vol.
Crisis kill switch: flatten short vol when filtered crisis probability > 0.5 or VIX > 40.
Spend 0.3% p.a. on far OTM VIX calls (3-month, 30-delta) as convexity offset.
Cap single-month sleeve loss at −4% of portfolio; auto-delever if breached.
OOS 2021–2025: sleeve contribution +0.6% CAGR, max drawdown −3.2%, one near-miss in Aug 2024 vol spike where kill switch fired early.

Technique decision table

Approach	VRP exposure	Best when	Watch out for
Delta-hedged short strangles	Pure index vega	Institutional vol premia sleeve with hedge desk	Gap risk, hedge slippage, March 2020-style spikes
Covered call overwrite	VRP + equity upside cap	Equity-heavy portfolios seeking income	Bull market participation forgone; not pure vol bet
Short vol ETPs	Packaged roll yield	Tactical trades, small satellite allocation	Contango decay, product redesign risk, path dependence
Long VIX calls (hedge)	Negative carry, tail convexity	Offset short-vol book or equity crash insurance	Steep cost drag in calm years
Vol targeting (long assets)	Indirect; scales exposure down in high RV	Core portfolio risk control	Does not harvest VRP; complements short-vol sizing
GARCH timing signal	Conditional entry/exit on vol forecast	Filter short-vol entries when RV forecast high	Model risk; false signals in regime shifts

Common pitfalls

Treating average VRP as guaranteed return — the premium is earned in calm years and repaid in crises; size for tail loss, not mean alone.
Ignoring negative skew — high Sharpe on short vol backtests often reflects selling lottery tickets; report worst-month and crisis-path stats.
Mismatched horizons — comparing 30-day VIX to 10-day realized vol inflates apparent premium.
Survivorship in vol-selling funds — delisted short-vol ETNs and blown-up funds disappear from databases (see survivorship bias guide).
Over-leveraging — short gamma scales nonlinearly; a 3% allocation can become 15% vega equivalent in a spike without caps.
No crisis protocol — discretionary “we'll close when it feels bad” fails under stress; automate kill switches.
Confusing roll yield with VRP — VIX futures contango benefits short-VIX products separately from SPX option VRP.
Transaction cost neglect — delta hedging and wide spreads erode retail-sized accounts; model costs in backtests.

Production checklist

Fix VRP definition (vol vs variance units) and measurement window (21 vs 30 days).
Compute historical VIX − realized series; report mean, median, and 5th percentile.
Choose harvest vehicle (overwrite, strangles, variance swaps) matching mandate and margin.
Set risk budget (% portfolio) and max sleeve drawdown stop.
Implement vol-scaling or entry filter (sell when IV − RV > threshold).
Define crisis kill switch (VIX level, regime probability, or RV spike).
Budget convexity hedge cost (OTM VIX calls or long puts) and disclose drag.
Stress-test 2008, 2011, 2015, 2018, 2020 paths on sleeve P&L.
Monitor live vega, gamma, and delta daily; document hedge frequency.
Report sleeve attribution separately from equity beta in LP updates.

Key takeaways

The variance risk premium is the gap between implied and realized volatility — positive on average because investors overpay for crash insurance.
Harvesting VRP means selling volatility in some form; returns look smooth until tail events erase years of carry.
Measure consistently (VIX vs aligned realized vol, or variance swaps) and size with crisis kill switches, not mean premium alone.
Harbor Capital's refactor cut overlay size, added regime-based flatten rules, and bought cheap convexity after March 2020.
Pair short-vol sleeves with vol targeting on the core book and explicit tail hedges — VRP is a risk premium, not arbitrage.