The Hedge That Eats Itself: Leland on Options and Transaction Costs

There is a devastating objection to Black-Scholes that anyone can understand in ten seconds, and which the theory has no answer to.

The model says: to replicate an option, adjust your stock hedge continuously. Rebalance at every instant. In the limit, you trade infinitely often.

Every trade costs money. Infinitely many trades cost infinitely much money.

So the replicating strategy that supposedly proves the Black-Scholes price is correct would, executed as specified, bankrupt you. The proof requires an action that is not merely difficult but economically impossible. Hayne Leland faced this squarely in 1985, and produced a fix that traders have used ever since.

The problem: the arbitrage argument requires a strategy nobody can execute

It is worth being precise about why this is not a small technicality.

Black-Scholes is not a statistical model. It does not say "options tend to be worth about this much." It makes a much stronger claim: that the option price is forced by arbitrage, because you can manufacture the option exactly by trading the stock. If the manufacturing process is infinitely expensive, the arbitrage argument does not hold, and the entire logical foundation of the price wobbles.

Worse, the problem does not go away if you rebalance less often. Rebalance rarely, and you save on costs but your hedge drifts away from the option and you take real losses from being mishedged. Rebalance often, and your hedge tracks nicely but the fees eat you alive. You are squeezed from both sides, and the model offers no guidance on where to sit.

There is also an asymmetry that makes the position genuinely dangerous. If you have sold an option, your delta hedge requires you to buy the stock as it rises and sell as it falls: you are always trading in the direction the market just went. Buy high, sell low, over and over, all day. Every one of those trades crosses the spread. Selling options and hedging is a business of accumulating a long stream of small, guaranteed losses in exchange for the premium. If the premium does not cover the losses, you lose. Leland's paper is, at bottom, about how much premium you need.

The key idea via analogy: pay for the friction with a fatter volatility

Leland's solution is beautifully pragmatic.

The costs of hedging are, ultimately, driven by how much you have to trade. How much you have to trade is driven by how much the stock moves. And how much the stock moves is measured by volatility. So the cost of hedging is essentially a volatility-like quantity. If it walks like volatility, treat it like volatility.

Leland showed that if you rebalance at fixed intervals (say once a day) rather than continuously, and you face proportional transaction costs (a percentage fee on each trade), then you can go on using the ordinary Black-Scholes formula, provided you plug in a modified volatility that is bumped up to account for the friction.

The adjustment has an intuitive shape:

• It increases with the size of the transaction cost. Obviously.
• It decreases with how much time you leave between rebalances. Trading less often costs less.
• It is larger for shorter-dated options. A one-week option needs a lot of frantic re-hedging per unit of time, so its friction cost is proportionally higher.

And crucially, the adjustment applies asymmetrically depending on which side you are on. If you are selling an option, your true cost of manufacturing it is higher than Black-Scholes says, so you must charge a higher price, which means using a higher volatility. If you are buying an option and hedging it, the frictions work in your favour on the price you can afford to pay, so the adjustment goes the other way, with a lower volatility.

That means there is no single fair price any more. There is a bid price and an ask price, and the gap between them is the cost of friction. Leland's paper is, in a real sense, the theoretical origin of the option bid-ask spread as an economic quantity rather than a market-microstructure accident.

Why it mattered

• It is used in practice, right now. Options desks routinely mark their books with a volatility that is adjusted for hedging costs, and they widen their quotes for less liquid underlyings where hedging is expensive. The instinct that "I need to charge more vol on this name because the stock is a pig to trade" is Leland's result, internalised.
• It explains the bid-ask spread in options. Why is there a two-way market rather than a single price? Because the cost of replicating the option differs depending on which way you are replicating it. This gives the spread a reason, tied to the underlying's liquidity, the option's maturity and its gamma.
• It killed the fantasy of the riskless hedge, gently. Leland does not blow up Black-Scholes. He shows how to keep using it while being honest that the hedge is imperfect and costly. The hedging error does not vanish, it just becomes small and, importantly, uncorrelated with the market, so it is diversifiable across a book of many positions. That is a much more realistic description of how a real options desk actually makes money.
• It opened a research field. The question of how to hedge optimally under transaction costs became a serious literature, producing the idea of no-transaction bands: rather than hedging back to a target delta, you let your delta drift within a tolerance band and only trade when it escapes. Nearly every real trading system does something like this, and it descends from the problem Leland posed.

The honest limitations

• The approximation degrades under exactly the conditions that matter. Leland's adjustment works well when transaction costs are small and rebalancing is reasonably frequent. Later analysis showed that as you rebalance more and more often, the approximation does not converge as cleanly as one might hope, and the hedging error does not vanish the way the original argument suggested. The result is a good practical rule, not an exact theorem.
• It assumes proportional costs, which is only part of the story. Real trading costs include the bid-ask spread, but also market impact, which grows with the size you trade and is nonlinear. Hedging a large option position moves the underlying, which changes your hedge, which moves the underlying again. Leland's framework does not capture this feedback, and it is precisely the mechanism behind portfolio insurance's role in the 1987 crash, an irony given the timing.
• It fixes the rebalancing schedule in advance. Trading at a fixed clock interval is not optimal. The right time to trade depends on how far your hedge has drifted, not on what time it is. The no-transaction band literature exists because Leland's fixed-interval assumption is a simplification.
• The hedging error is small on average, but it is not small in the tail. Leland's comfort, that errors are uncorrelated with the market and shrink with frequent rebalancing, is a statement about typical conditions. In a crash, when the stock gaps and liquidity evaporates, both assumptions fail simultaneously. The costs spike and the hedging error becomes both large and, across your whole book, highly correlated.
• It does not tell you what to do when you cannot trade at all. Sometimes the market is closed, or the stock is halted, or nobody will take the other side. That is a hedging cost of a different kind.

The one-line takeaway

Leland confronted the fact that Black-Scholes' continuous hedge would cost an infinite amount in a world with real trading fees, and showed that you can keep using the formula if you bump volatility up when you sell and down when you buy, which quantifies the option bid-ask spread, explains why illiquid underlyings carry fatter option premiums, and quietly retires the idea that a hedged options position is ever truly riskless.