Contango and Backwardation: The Rollover Cost Futures Traders Pay Monthly

You bought one crude contract in January at 80.00. It’s now July, the front month settles at 80.10, and your account shows a loss of roughly 9%. Nothing went wrong with the trade. You paid that money away in six rollover transactions, and you’ll pay it again in the next six.

The gap between what happened to the price and what happened to your account has a name: roll cost. When the contract month you’re moving into trades above the one you’re leaving, closing the near month and opening the far one means selling low and buying high, deliberately, on a schedule. That’s contango. Backwardation runs the identical machinery in reverse and pays you for the same trade.

What roll cost actually is

Roll cost is the price gap between the contract you close and the contract you open, measured as a percentage of the position.

Stay with the January position. The February contract settles at 80.00. February expires in eight days, so you sell it and buy March at 80.60. The spread is 0.60 / 80.00, or 0.75%. Call the exposure $10,000 to keep the arithmetic readable.

At the instant of the trade, nothing happened to your account balance. You sold one thing and bought another at market. But you now own the same exposure at an entry 0.75% worse, and every percentage gain from here is measured from a base that has to travel further to get you back to even. Do that once and it’s noise. Do it twelve times and it’s a number that shows up in the annual statement.

Contango is when the far month sits above the near month, which hurts a long position on every roll and helps a short one. Backwardation is the opposite: far month below near, so the roll sells high and buys low. A flat curve does neither. The mechanics never change. Only the sign does.

The assumptions behind every number here

State these up front, because they’re what moves the answers.

One roll per calendar month, twelve per year. Front-month contracts on most liquid energy and agricultural markets do roll that often; index and rate futures usually roll quarterly, and that case is handled separately below. The roll spread is quoted as a percentage of the near-month price and applied at each roll. Spot is held completely flat, which is the only way to see roll cost in isolation. Commissions, exchange fees, and the bid-ask on the roll trade itself sit outside these figures and get quantified later. No margin interest, no separate storage or dividend adjustment beyond what the spread already prices in.

The last assumption is the one people get wrong: compounding is multiplicative. A 0.75% cost applied twelve times is 0.9925^12, not 0.75 × 12. Simple multiplication overstates a contango drag and understates a backwardation gain, and both errors grow as the spread widens.

A year of rolling, at every spread level

The annual figure is (1 - s)^12 - 1 for contango and (1 + s)^12 - 1 for backwardation, where s is the monthly spread as a decimal.

Two things fall out of this.

Below about 0.30% a month, the shortcut is fine. Multiply by twelve and you’re within a tenth of a point. Don’t bother with the exponent.

At large spreads the shortcut is wrong in the frightening direction. A 3% monthly contango does not eat 36% of the position. It eats 30.6%, because each month’s loss applies to a smaller base than the one before. Still ugly. Just less ugly than the arithmetic people do in their heads.

The $10,000 walked through, roll by roll

Back to your position, twelve rolls at 0.75%, spot pinned.

Roll one takes $10,000 to $9,925.00 and costs $75.00. Roll two takes that to $9,850.56. By roll six you’re at $9,558.35, down 4.42%, which is roughly where the July statement in the opening lands. Roll twelve takes $9,205.24 to $9,136.20 and costs $69.04.

Ending value $9,136.20 on $10,000. Loss $863.80, or 8.63%, matching the table.

Look at the two roll costs side by side. Roll one cost $75.00, roll twelve cost $69.04, and the percentage never changed. The base shrank. That $5.96 difference, repeated across the year, is the entire source of the gap between the naive 9.00% and the real 8.63%.

How the answer moves with holding period

Spread is one axis. Time held is the other. Cumulative contango drag, as a percentage:

The three-year column is the one long-horizon holders should sit with. A persistent 1.5% monthly contango removes 42% of the position over thirty-six rolls with spot exactly where it started. The commodity would have to rise nearly 72% over those three years for you to break even.

There’s a fair objection here: curves don’t stay in contango for three straight years. Mostly true. Curves flip, and a steep contango year can be followed by a backwardation year that hands a good chunk of it back. This table isn’t a forecast of anything. It’s the answer to “what if this condition persists,” which is the number you need to judge whether a position survives its bad case.

Quarterly contracts change the exponent, not the method

Equity index and rate futures typically roll four times a year. Same formula, different power. A 1.50% spread rolled quarterly costs 5.87% a year. The same 1.50% rolled monthly costs 16.59%. Frequency matters roughly as much as size, and it’s the variable most people ignore, because it lives in the contract specification rather than on a quote screen.

One caution on that comparison. A quarterly roll usually carries a wider spread than a monthly one, since it’s covering three months of carry instead of one. Setting 0.75% quarterly against 0.75% monthly isn’t a like-for-like test. The honest version is 0.75% monthly against roughly 2.25% quarterly, which land at -8.63% and -8.70%. Nearly identical, as they should be. It’s the same carry, sliced differently.

The move you need to break even is bigger than the drag

If contango costs 8.63%, spot rising 8.63% does not put you back to flat. The gain has to be large enough to restore the original from a reduced base, so the break-even is 1 / (1 - drag) - 1.

Check it on the position. You started at $10,000 and roll drag left you at $9,136.20. For that to become $10,000 again it must gain $863.80, and $863.80 / $9,136.20 is 9.45%. So 8.63% of drag needs a 9.45% spot move.

The gap widens fast. An 11.36% drag needs +12.82%. A 16.59% drag needs +19.89%. A 30.62% drag needs +44.13%, which is the same arithmetic that makes a 50% loss require a 100% gain.

Run the question the other way and it’s more useful. Suppose you need the position up 15% after a year and you’re facing 0.50% monthly contango. Drag leaves you holding 0.9416 of what you started with. You need a final multiplier of 1.15, so spot has to deliver 1.15 / 0.9416 = 1.2213. That’s a 22.13% rise, not 15%. Sanity-check your thesis against 22.13%.

Two things people get wrong

“Contango is a fee, so the expense ratio covers it.” It isn’t and it doesn’t. An expense ratio is a disclosed, deducted cost. The SEC’s page on mutual fund fees and expenses covers management fees, 12b-1 fees, and other operating charges. Roll cost is none of those. It’s a market price differential that shows up in performance and nowhere else.

Numbers make the point better. A commodity fund charging 0.45% a year while holding front-month contracts through a 1.00% monthly contango carries 11.36% of roll drag on top of the fee, for a combined 11.76%. The undisclosed cost is twenty-five times the disclosed one. Read the fee table, stop there, and you’ve seen almost nothing of what the position costs.

“Contango and backwardation are symmetric, so it evens out.” The percentages aren’t symmetric, and the asymmetry doesn’t favour the long holder the way you’d hope. At 1.00% monthly, contango costs 11.36% and backwardation pays 12.68%. Alternate one of each and you get 0.8864 × 1.1268 = 0.9988, a 0.12% loss rather than zero. Run the same alternation at 3.00% monthly and it’s 0.6938 × 1.4258 = 0.9892, a 1.08% loss. Equal-magnitude years leave you slightly down, every time, because the geometric mean of two multipliers straddling 1.00 always sits below their arithmetic mean.

Where these figures break down

The spread isn’t constant. One number applied twelve times is a modelling convenience. Real curves shift week to week and season to season; natural gas has a structural winter shape that has nothing to do with its average. If your spread ranged from 0.2% to 1.8% across a year, the compounded result is worse than applying the 1.0% mean.

Trading costs sit on top. Each roll is two transactions. At $2 per side all-in — commission, exchange and clearing fees, half the bid-ask — that’s $4 per roll. On $50,000 of notional it’s 0.008% per roll and 0.096% a year, which is rounding error. On a micro contract with $2,500 of notional, the same $4 is 0.16% per roll and 1.9% a year. The fee is fixed; the notional isn’t.

Timing the roll changes the spread. Front-month prices converge toward spot as expiry nears, so rolling ten days out gives a different number than rolling two days out, sometimes materially, and which way depends on curve shape.

Deferred-month strategies change the exponent. Holding a contract six months out and rolling it once or twice a year means far fewer rolls, each at a wider spread. Frequency and size trade off, and the net is usually — not always — better.

Physical delivery distorts the near month. Storage, delivery logistics, and squeeze risk can push a physically-settled front month around in ways no smooth carry model captures. Cash-settled contracts behave more predictably into expiry.

Everything above is pre-tax. Futures have their own tax treatment, and whether you’re classified as an investor or a trader changes it. The IRS sets out that distinction in Topic 429 on traders in securities.

What this doesn’t tell you

These calculations do one thing: they compute what a given roll spread does over a given number of rolls.

They don’t tell you what the spread will be. Curve shape depends on storage costs, interest rates, inventory, and expectations, none of which comes out of a table.

They say nothing about whether contango predicts anything about future spot prices. That’s a long-running empirical argument and it isn’t settled here.

They exclude margin entirely. Futures are margined, so return on posted capital diverges wildly from return on notional. An 8.63% drag on notional is 86.3% of the margin posted if the position is margined at 10%. Leverage magnifies roll cost exactly as it magnifies everything else.

And they don’t say whether a position did its job. A hedger who paid 8.63% in roll cost to remove a price risk that would have cost 30% made a reasonable trade. Cost and mistake aren’t the same thing.

FAQ

How much does contango cost per month on average?

There’s no single average — it varies by commodity and by year, sometimes dramatically. What’s fixed is the arithmetic: a 0.5% monthly spread costs 5.84% over twelve rolls, 1.0% costs 11.36%, and 2.0% costs 21.53%. Pull the actual front-month and next-month settlement prices for your contract, compute (far - near) / near, and find your row.

Does contango affect commodity ETFs the same way as holding futures directly?

Yes, if the fund holds and rolls futures, which most commodity funds do. The fund performs the rolls internally and the cost surfaces as a gap between the fund’s return and the change in spot price. It never appears in the expense ratio. A fund charging 0.45% while facing 1.00% monthly contango absorbs 11.36% of roll drag on top of the fee.

What spot move do I need if contango is 1% a month?

12.82%. Twelve rolls at 1.00% leave you holding 0.8864 of what you started with, and 1 / 0.8864 = 1.1282. Spot rising 11.36% isn’t enough to get you flat, and that’s before commissions and taxes.

Is backwardation always good for a long position?

It’s good for the roll — worth +12.68% a year at 1.00% monthly. It says nothing about where spot goes. You can collect 12.68% of positive roll yield and still lose badly if the underlying falls 25%. Roll yield and price return are separate components that add together, sometimes with opposite signs.

How do I calculate roll cost for my own contract?

Four steps. Take the near-month and next-month settlement prices. Compute (next - near) / near; positive is contango, negative is backwardation. Count the rolls per year, which is twelve for most front-month energy and agricultural contracts and four for most index contracts. Then raise (1 - spread) to that power and subtract 1. Worked example: near at 78.40, next at 79.05, spread = 0.65 / 78.40 = 0.00829, and 0.99171^12 - 1 = -9.51% for the year.

Why does multiplying by 12 overstate contango but understate backwardation?

Because compounding works on a changing base. In contango each successive loss applies to a smaller amount, so twelve losses total less than twelve times the first. In backwardation each gain applies to a larger amount, so they total more. At 2.00% monthly the gap is 2.47 points on the contango side and 2.82 points on the backwardation side, and both widen as the spread grows.

What to do with this

Pull the settlement prices for the front two contract months on whatever you’re tracking, straight from the exchange that lists it, and compute the spread yourself. One subtraction, one division. Then check the contract specification for how many times it rolls in a year — that number isn’t on a quote screen and it changes the annual figure more than most people expect.

Then set the roll drag your contract implies against the price move you’d need for the position to accomplish what you bought it for. If those two numbers are close, the term structure is doing more work in your outcome than your price view is.

This article is general information, not financial advice. See our disclaimer.

More in Costs and Fees · all topics · calculators · how this was checked