Skip to main content

Contango & Backwardation Roll Cost

Updated Jun 24, 2026 at 2:35pm

Research Draft Medium 1,189 words

Contango and backwardation describe the shape of a futures curve — whether contracts dated further out trade above (contango) or below (backwardation) nearer-dated contracts. For VIX and volatility products this shape is not a side detail; it is the dominant driver of return. Because the cash VIX index cannot itself be held, every "volatility ETP" (VXX, UVXY, the former XIV/SVXY) is built on futures that must be continually rolled forward as they expire. The roll itself transfers money: a holder repeatedly trades cheaper expiring contracts for more expensive longer-dated ones (in contango) or vice-versa (in backwardation). That recurring transfer — the roll cost or roll yield — accrues regardless of where spot VIX ends up, which is why a long-volatility ETP can bleed for years even though the VIX it tracks is roughly flat. The core tension: the curve shape that makes long-vol expensive to hold is the same shape that makes it a profitable but catastrophically tail-exposed thing to short.

How it's formed and calculated

A VIX-futures ETP targets a constant maturity (VXX/SVXY track the S&P 500 VIX Short-Term Futures Index, SPVXSTP, ~30-day average of the first two months). To keep maturity constant, the index sells a slice of the front-month contract and buys the second-month contract every trading day, mechanically "rolling down" the curve.

The economic drag per roll is the curve slope. A commonly used daily approximation is: Daily roll yield ≈ ((M1 − M2) / M1) / (days in the roll period)

With a front month M1 = 17.5 and second month M2 = 19.0 over a ~30-day roll, that is roughly −0.29% per day (example figures from Volatility Box / VolatilityTradingStrategies). In contango M2 > M1, so the term is negative — a continuous cost to longs. In backwardation M2 < M1, the sign flips and longs earn positive carry. Drivers differ by asset class: in storable commodities contango reflects cost-of-carry (storage + financing); in VIX the curve is shaped instead by the variance risk premium and mean-reversion expectations — markets normally price near-term vol below the longer-term equilibrium, producing persistent upward slope. The spot-vs-futures definition (futures above/below spot) and the month-to-month slope definition usually agree but can diverge; for ETP roll cost it is the slope between the contracts actually held that matters.

How it's used in practice

Three practical uses dominate:

1. Performance attribution. Before trading any vol ETP, practitioners decompose expected return into the spot-VIX move plus the roll yield. In quiet markets the negative roll usually swamps small spot moves, so VXX drifts down even on flat VIX. 2. Regime read. The front-to-back slope is read as a sentiment gauge: steep contango = complacency/calm; flattening or inversion into backwardation = stress, often coinciding with sharp equity drawdowns. Traders watch the M1/M2 ratio (e.g. the "VIX contango" figures on vixcentral.com) as a regime flag. 3. Carry harvesting. The persistence of contango motivates short-volatility / inverse products and outright short-VXX positions to collect the negative roll as positive carry. Academic strategies (Simon & Campasano) short VIX futures when daily roll exceeds a threshold and buy when in backwardation, often S&P-hedged.

Adoption, debate & evidence

That the VIX curve is usually in contango is well documented: multiple sources put contango at roughly 80–85% of trading days since the mid-2000s (commonly cited 84–85%; treat exact figures as estimate-dependent). The cause — a negative variance risk premium, where options/VIX systematically price implied vol above subsequently realized vol — is a mainstream, peer-reviewed finding (e.g. the VIX-premium literature in Review of Financial Studies; Johnson, JFQA 2017, on the VIX term-structure risk premium). On that point there is broad agreement.

What is contested is whether roll cost is a harvestable edge. Carry strategies look spectacular in-sample — Quantpedia's writeup of the Simon-Campasano basis strategy cites ~19.7% annualized over 2007–2011 — but the same source flags that out-of-sample performance turned slightly negative and "the strategy's alpha is deteriorating." Crucially, the return profile is not a free premium; it is compensation for bearing crash risk. The collapse of XIV in February 2018 is the canonical evidence: VIX roughly doubled in a day (close ~17 to ~37), the curve snapped into backwardation, and the inverse product lost ~96% overnight and was terminated. Short-vol carry is best understood as selling insurance — steady premiums punctuated by ruinous payouts.

Folklore vs. measured: the often-repeated "VXX loses ~X% a year to contango" (cited ranges of ~40–55% for VXX, more for leveraged UVXY) and "VXX is down ~99.9% since inception" are real outcomes but are path- and reverse-split-dependent; they should be read as illustrative of long-run drag, not a fixed annual decay rate.

Strengths & limitations

When the concept works: roll yield is a genuinely reliable explanatory variable — it predicts the structural drag on long-vol ETPs better than spot VIX does, and the contango/backwardation flip is a respected stress signal. As an analytical lens it is hard to beat.

When it fails: as a trading edge it is regime-fragile. Contango can vanish without warning; the carry that accrued slowly for months can be erased in a single backwardation spike. Leverage (UVXY 1.5x, daily-rebalanced inverse products) compounds both the roll drag and the path-dependency.

The #1 misuse: treating short-vol roll carry as "passive income." Its smooth equity curve masks a deeply negatively-skewed, fat-tailed distribution; sizing it like a normal asset is how accounts get destroyed. The second-most-common error is buying VXX as a "VIX tracker" and being surprised by the contango bleed — VXX tracks VIX futures, not spot VIX.

Sources

Disputed / flagged: contango-frequency percentages (80–85%) and ETP annual-decay figures (40–75%) are estimate- and period-dependent — treated as qualified ranges, not exact constants. Whether roll-yield carry is a persistent edge is genuinely contested (in-sample strong, OOS deteriorating); presented as crash-risk premium, not free money.