Valuation guides

PDE option pricing explained

What PDE pricing is, why finite differences are used, an example and how any payoff is valued on a grid.

Definition

Option prices satisfy a partial differential equation, the Black-Scholes PDE or its extensions. PDE pricing solves it numerically on a grid of share prices and times, starting from the payoff at expiry and stepping back to today. Early exercise and barriers are applied as conditions on the grid, which makes PDEs fast and accurate for one or two factors.

History

Brennan and Schwartz (1977) first used finite differences to value American options. Crank-Nicolson, a time-stepping scheme from 1947, became standard, with Rannacher (1984) start-up steps to handle kinked payoffs. Alternating direction implicit (ADI) schemes such as Craig-Sneyd (1988) made two-factor models like Heston practical.

Examples

A capped call

A payoff of min(max(S − 100, 0), 30) with an American exercise right and a knock-out at 150 is written as an expression and solved on a grid in seconds, without any closed form.

Variations

PDE option
Any one-asset payoff, European, American or Bermudan, with barriers.

Valuation methods

The PDE is solved by Crank-Nicolson under Black-Scholes, CEV or local volatility, and by Modified Craig-Sneyd ADI under Heston. Early exercise is applied at each step and barriers as boundary conditions.

Black-Scholes PDE

∂V∂t+12σ2S2∂2V∂S2+(r−q)S∂V∂S−rV=0\frac{\partial V}{\partial t} + \tfrac12\sigma^2 S^2\frac{\partial^2 V}{\partial S^2} + (r-q)S\frac{\partial V}{\partial S} - rV = 0

V(S,t) is the option value.

Models on RiskAtoms

InstrumentPricing models
Custom payoff option (PDE)
Finite-difference PDE (Crank-Nicolson)
Black-Scholes
CEV (constant elasticity of variance)
Local volatility (Dupire PDE)
Heston PDE (Modified Craig-Sneyd ADI)