Static Replication Of Payoffs

Delta-hedging an exotic option means rebalancing constantly, and every rebalance depends on a volatility assumption that might be wrong. There's a quieter alternative for a whole class of payoffs: build the hedge once, out of a fixed basket of ordinary options, walk away, and let it replicate the exotic payoff exactly at expiry — no model of future volatility required at all.

A vending machine versus a cashier

A dynamic hedge is like a cashier who has to keep re-pricing an item every time the exchange rate ticks, all day, every day, and who is wrong whenever they mis-time an update. Static replication is a vending machine: you design the mechanism once — the right combination of coin slots and levers — and from then on it does the right thing automatically, for any input, with no further attention. For payoffs that only depend on the terminal price of the underlying (not the path it took to get there), a basket of plain vanilla options struck at every price level, held fixed from today until expiry, can reproduce that terminal payoff exactly. You never touch the position again.

The formula

Any twice-differentiable payoff function $f(S_T)$ can be decomposed as:

$$f(S_T) = f(S_0) + f'(S_0)(S_T - S_0) + \int_0^{S_0} f''(K)\,(K - S_T)^+\, dK + \int_{S_0}^{\infty} f''(K)\,(S_T - K)^+\, dK$$

In plain English: any payoff, however oddly shaped, can be rebuilt from four ingredients held today — cash equal to the payoff's value at today's price $f(S_0)$, a forward position sized by the payoff's slope $f'(S_0)$, a continuum of puts struck below today's price weighted by the payoff's curvature $f''(K)$ at each strike, and a continuum of calls struck above weighted the same way. The curvature term is the interesting one: wherever the payoff bends, you need options struck at that exact point, in a quantity proportional to how sharply it bends there. A straight-line payoff needs no options at all — its curvature is zero everywhere.

Worked example 1: replicating a simple digital-like kink cheaply

You want to hand a client a payoff that is flat at $0 below $100, then rises linearly by $1 for every $1 the stock is above $100, capped once it reaches $110 (a "capped call," paying $0 to $10). This has two kinks: one at $100 (where the payoff starts rising) and one at $110 (where it stops). By the decomposition above, that's exactly a bull call spread: buy one call struck at $100, sell one call struck at $110. Suppose the $100 call costs $6.50 and the $110 call costs $2.10. The static replication costs $6.50 - 2.10 = 4.40$, i.e. $4.40, today, and at expiry it pays exactly $0 below $100, ramps to $10 at $110, and stays at $10 above — matching the target payoff at every single point, with zero rebalancing between now and expiry.

Worked example 2: replicating a variance swap's payoff

A variance swap pays the difference between realized variance and a fixed strike, and the celebrated Carr-Madan result is that this payoff — despite depending on the whole path, not just the endpoint — can be statically replicated using a log-payoff, $-\ln(S_T/S_0)$, which itself decomposes into a strip of puts and calls at every strike, weighted by $1/K^2$. Concretely: to replicate $1 of variance exposure with strikes spaced $5 apart between $80 and $120 around a $100 spot, a put struck at $90 gets weight $1/90^2 \times 5 = 0.000617$, while a put struck at $95 gets weight $1/95^2 \times 5 = 0.000554$ — the weighting shrinks smoothly as strikes move away from spot, which is exactly why variance swap trading desks buy the whole visible strip of listed options rather than a handful of strikes.

Drag the strike and premium on this single-leg view first to see one piece of the puzzle — then picture stacking a second, opposite-signed leg at $110 on top of it, which is exactly the bull spread from Example 1.

What this means in practice

Static replication underlies how banks manufacture structured products (capped notes, barrier-adjacent payoffs, digital-like coupons) and hand the resulting basket to a trading desk that hedges it once rather than continuously — sharply reducing model risk, since no assumption about future volatility is needed for a purely static hedge. It is also the theoretical backbone of the Breeden-Litzenberger Formula and of how variance swaps and the VIX itself are actually constructed from a strip of listed options.