The cost-of-carry argument gave a clean equality, F_0 = S_0 e^{(r-q)T}, and it did so by appealing to two arbitrage strategies running in opposite directions — one that kicks in if the forward is too expensive, another if it's too cheap. Try to run that same argument on a barrel of crude oil or a bushel of wheat, and it quietly breaks. A refinery holding crude isn't holding it as a pure financial investment — it's holding it because it needs to keep feeding a production line today. Ask that refinery to lend you its oil so you can short-sell it, the way you might borrow shares of a stock, and the answer is almost always no. One whole half of the arbitrage argument quietly vanishes, and with it the tidy equals sign.

Why the equality collapses into an inequality

Call an asset held mainly to be used — as fuel, as feedstock, as food — a consumption asset, in contrast to an investment asset like gold, a stock, or a currency, held (at least by enough market participants) purely for its financial return. Redo the cash-and-carry argument for a consumption asset with a proportional continuous storage cost u (warehousing, insurance, spoilage — a negative yield, since it costs money rather than paying it):

Direction 1 still works. If F_0 > S_0 e^{(r+u)T}, anyone can still borrow cash, buy the commodity spot, pay to store it, and sell it forward at the rich price F_0 — a riskless profit F_0 - S_0 e^{(r+u)T} > 0, exactly as before. Buying spot and storing a commodity you don't currently need is always possible, so this half of the argument survives intact and still forces

Direction 2 breaks. The reverse trade needed F_0 < S_0e^{(r+u)T} arbitrage to short-sell the commodity — borrow it from an owner, sell it, and buy it back later. But the people who own physical crude oil or wheat mostly hold it because they're about to use it: a refinery needs that crude running through its plant, a mill needs that wheat in its silo. They won't lend it out for a token fee and risk a stock-out, no matter how attractive the arbitrage looks on paper. With too few lenders, the reverse cash-and-carry trade can't be executed at scale, so nothing stops F_0 from sitting below S_0 e^{(r+u)T} indefinitely. Only one direction of arbitrage survives contact with a real commodity market — which is exactly why only an inequality, not an equality, is all that no-arbitrage can guarantee.

Convenience yield: pricing the missing half

Markets don't like leaving a gap unexplained, so practitioners give the shortfall a name and a symbol. Define the convenience yield y \ge 0 as exactly the number that turns the inequality back into an equality:

y isn't a cash payment anyone actually receives — it's a stand-in for the real, non-financial benefits of holding the physical commodity itself rather than a promise of future delivery: insurance against a stock-out that would halt production, the flexibility to sell locally at a spike in demand, simply not having to scramble for feedstock next week. When inventories are ample, that convenience is worth little and y stays small — the forward curve slopes up, a pattern called contango. When inventories run tight, physical barrels in hand become precious and y can grow large enough to overwhelm r+u entirely, tipping the curve down — a pattern called backwardation, where the forward price sits below today's spot price.

Worked example: backing out the convenience yield

Spot crude oil trades at S_0 = \$75.00 per barrel. The continuously-compounded risk-free rate is r = 5\%, and storage (tank rental, insurance) costs u = 3\% per year, continuously compounded. The no-convenience-yield ceiling for a six-month forward (T = 0.5) is

But the market actually quotes the six-month forward at only F_0 = \$76.50 — comfortably below the ceiling, so no arbitrage opportunity exists, but a gap remains to explain. Solve F_0 = S_0 e^{(r+u-y)T} for y:

an implied convenience yield of about 4.04\% per year. The market is telling us that, right now, holding a physical barrel of oil in a tank is worth roughly four percent a year more than the pure financing-and-storage argument alone would justify — presumably because inventories are a bit tight and refiners value the certainty of supply in hand.

Dragging the curve from contango into backwardation

Fix S_0 = 100, r+u = 8\% per year, and plot the forward price against maturity T. The dashed line is the hard ceiling S_0e^{(r+u)T} from Direction 1's arbitrage — it never moves, because that half of the argument always holds. The solid curve is the actual forward price with a convenience yield y subtracted off. At y=0 the two coincide — no convenience yield, full contango, same as an investment asset. Drag y up past 8\% and the solid curve peels away below the dashed ceiling and starts sloping down: backwardation.