Tail hedging: what put protection costs, when it pays off and what short-vol strategies teach

Advanced12 min read · 11 sources

Tail hedging means deliberately protecting a portfolio against rare, severe crashes, usually with deep out-of-the-money puts. In a crash, the hedge pays out many times its price, but in every other year it costs premium. Whether that is worthwhile depends on the volatility risk premium, on strike and term, and on timing. Here you will work through the costs and payoffs, get to know the debate between advocates and critics, and see from February 5, 2018 how the other side of the trade can fail.

What tail hedging is and what it costs

A tail hedge is not meant to smooth every fluctuation but only to cut off the left edge of the distribution: losses beyond a threshold. The classic tool is the protective put, a purchased put on the index, along with variants such as the put spread (buy a put, sell a lower put) and the collar (buy a put, sell a call). How the building blocks are priced is explained in the article option pricing; how the volatility priced into options is measured, in the article VIX.

The price shows up as negative carry: as long as nothing happens, the put expires and the premium is lost. If you hedge again every year, you pay this premium year after year. So the question is not whether the hedge helps in a crash (it does), but whether the sum of the premiums over many calm years is smaller than the payoff in the few bad ones.

Worked example: put costs vs. the payoff at −30%

The prices in the example are Black-Scholes model values with a 2% interest rate, no dividends and a volatility per strike that mimics the typical skew: deep out-of-the-money puts are priced with higher volatility than out-of-the-money calls. These are illustrative figures, not market prices.

A second lesson lies in the term. Instead of a one-year put, you can buy a put with three months to expiration and a strike of 90% four times a year. With σ = 24%, each one costs 1.10, so about 4.39 a year, which is more than the one-year put. And it protects differently: if the index falls 30% within one quarter, the current put pays 20 points. If, on the other hand, it falls 8.5% four times, the portfolio loses 29.9% over the year, yet not a single quarterly put ends up in the money. Israelov (2018) shows exactly this problem: without lucky timing of purchase and term, puts protect hardly better than a permanently lower equity allocation.

The volatility risk premium: why hedging loses on average

Options contain a premium for allowing their buyer to shed volatility risk. Carr and Wu (2009) measure it directly: they replicate the fair rate of a variance swap from a portfolio of options and compare it with the variance realized later, for five stock indices and 35 individual stocks. The variance risk premium is the difference:

• Delta-hedged, the buyer loses: Bakshi and Kapadia (2003) find for options on the S&P 500 that purchased, delta-hedged options stay below zero on average, less so for deep out-of-the-money options and more so in periods of high volatility. This is evidence of a negative market price of volatility risk.
• Puts below the risk-free rate: according to Coval and Shumway (2001), the expected returns of puts are below the risk-free rate, and beta-neutral at-the-money straddles lost about 3% a week on average.
• Selling insurance is rewarded: Ilmanen (2012) finds positive long-term premiums for selling insurance in many markets and weak results for buying catastrophe protection.
• Overpriced puts: Bondarenko (2014) shows that the historical prices of puts on the S&P 500 are not consistent with any model from a broad class of equilibrium models.

For the tail hedger, this means: in normal times, you are buying expensive insurance. That is not a mistake if the protection is worth more than the expected return it costs. But you need to know it. How large the priced-in premium currently is can be seen from the volatility premium derived from the VIX and realized volatility.

The debate: Spitznagel vs. AQR

Among practitioners, it is disputed whether tail hedging pays off. On one side is Spitznagel (2021), manager of a hedge fund specializing in tail protection. He asks whether risk reduction can increase returns and judges safe havens by how they change the compound growth of the whole portfolio over time, not by their own return. The idea behind this: large losses have a disproportionate effect on the geometric return (see volatility drag). A small, highly convex hedge that explodes in a crash can therefore add more growth than its premiums cost.

On the other side are the critics. Israelov (2018) finds that puts, as typically used, reduce drawdowns hardly better than a permanently lower equity allocation and can even increase drawdowns and volatility per unit of expected return. Ilmanen et al. (2021), whose study the asset manager AQR also published itself, compare regularly purchased index puts with trend following across many asset classes: the widespread view that puts cost more but protect more reliably has a kernel of truth but falls short; on costs, trend following comes out ahead. Who is right depends on assumptions about the frequency and depth of crashes. A simple calculation shows this:

The example is simplified because the put costs the same every year; in reality, protection becomes more expensive after a crash. What matters most for you is the honesty of the calculation: if you justify a tail hedge with a backtest, you should show how much of the result depends on individual crashes.

Short volatility and February 5, 2018

If you sell protection, you earn the premium on average and bear the risk the buyer wants to shed. How quickly this can turn is shown by February 5, 2018, which traders call Volmageddon. At the time, there were exchange-traded products that delivered the daily return of short-term VIX futures with leverage or inversely. A product with factor k has to rebalance every evening, and in the same direction the market just moved:

This is exactly what the BIS describes for February 5, 2018 (BIS 2018). The S&P 500 fell 4.2% that day, a move of 3.8 standard deviations, and the VIX jumped 20 points, its largest daily rise since the 1987 crash. The assets of selected leveraged and inverse volatility products had grown to about $4 billion by the end of 2017. Both types had to buy VIX futures shortly before 16:15 New York time, when they publish their net asset value; at 16:08, 115,862 contracts traded within a single minute. The inverse product XIV lost 84% and was shut down. Another inverse product halved its factor from −1 to −0.5 at the close of trading on February 27, 2018 (ProShares 2026).

Augustin et al. (2021) attribute the collapse to the same feedback loop: buying futures to hedge and to maintain leverage drove prices even higher in a concentrated market. The lesson goes beyond a single product. Short volatility earns steadily for a long time and rarely loses, but then all at once. The result is a distribution with strongly negative skewness that flatters metrics such as the Sharpe ratio.

Alternatives to the put

• Fewer stocks: the simplest hedge is a lower equity allocation. It costs expected return but no volatility premium; Israelov (2018) uses it as the benchmark.
• Trend following: managed futures have often gained in prolonged downturns and cost less than puts on average (Ilmanen et al. 2021); fast crashes after long rallies, however, catch them long (see CTAs and managed futures).
• Government bonds: inexpensive protection in growth crises, but not reliable in inflation shocks (see stock-bond correlation).
• CPPI and portfolio insurance: a rule instead of an option; it pays no premium but loses in choppy markets and carries gap risk (see CPPI).
• Long volatility via VIX futures: rises sharply in a crash but loses steadily in calm periods from rolling in contango (see VIX).
• Combinations: several imperfect hedges with different weaknesses, such as a small put allocation and trend following, often complement each other better than a single large one.

Advanced: decomposing the premium of a put

How much of the put price is expected loss? With assumptions about the actual distribution, this can be decomposed. Under a lognormal distribution with expected return μ and volatility σ_P, E^P[max(K − S_T, 0)] = K N(−d₂) − S e^(μT) N(−d₁), with d₁ = [ln(S/K) + (μ + σ_P²/2) T] / (σ_P √T) and d₂ = d₁ − σ_P √T.

Two caveats belong here. First, the calculation depends on μ and σ_P, which you do not know; with fat tails, a fall to 70 is much more likely than 0.58% a year (see skewness and kurtosis). Second, the put is a position in volatility: delta-hedged, the buyer earns roughly ½ Γ S² (σ²_real − σ²_impl) Δt between two rebalancings. So the buyer wins when realized volatility exceeds the priced-in volatility: usually by a lot in a crash, and usually not otherwise. Many tail hedgers also use monetization rules: they sell a put that has risen sharply in value and buy new protection instead of waiting until expiration.

Limitations and common mistakes

• Backtests with few crashes: the result hinges on two or three events. A different period can reverse the verdict.
• Timing: term, strike and purchase date determine whether a hedge pays at the decisive moment. Slow bear markets erode the value of short-dated puts.
• Overlooking skew: deep out-of-the-money puts are relatively expensive; low absolute prices do not mean cheap insurance.
• Costs on the other side: if you sell options instead, you collect premium but bear the crash risk, including margin requirements that rise precisely under stress.
• Basis risk: a put on a benchmark index only roughly hedges a broader or differently composed portfolio (see short selling and hedging).
• Taxes and trading costs: spreads on deep out-of-the-money options and the taxation of premiums and gains change the result. This article is not investment or tax advice.

How Portfolio Lab calculates

The option pricer values puts, calls and combinations with the Black-Scholes-Merton model and shows the payoff profile at expiration. Templates include the protective put (put at 95% of the price), the collar (put 90%, call 110%) and the bear put spread; custom strategies combine up to eight legs, each strike with its own volatility. The Volatility dashboard shows the volatility premium as the VIX minus the realized volatility of the last 21 trading days.

The app does not simulate a rolling put strategy with actual option prices. For backtests, approximations are available as Asset classes in euros: “Tail-risk hedging” (@TAIL_RISK), long-dated puts on US stocks plus government bonds, since 2017; “Volatility long (VIX futures)” (@LONG_VOLA) and “Volatility short (VIX futures)” (@SHORT_VOLA), the latter only with half the inverse factor since 2018; plus “Put-write S&P 500” (@PUT_WRITE) and “Covered call S&P 500 (buy-write)” (@COVERED_CALL) as sellers of insurance. In the Backtest, you blend them into an equity portfolio; alternatives such as trend following can be found under Tactical, and portfolio insurance under CPPI. The formulas are in the option pricer methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.