Backward Stochastic Differential Equations

An ordinary stochastic differential equation (SDE) specifies a process forward: you know where it starts today, and randomness pushes it around going forward in time. Many finance problems actually run the other direction — you know the payoff of a derivative at expiry (a known future condition), and you want to work out what the price must be today and along the way, consistent with that known ending. A backward stochastic differential equation (BSDE) is built exactly for this: it specifies a process by its terminal value and an equation it must satisfy, and the solution is found working backward from the endpoint.

A BSDE's solution is actually a pair of processes: one that gives the value at each point in time (like a derivative's fair price along the way), and a second, less intuitive process that captures the process's sensitivity to the randomness driving it — in an option-pricing context, this second piece corresponds to the hedge ratio needed to replicate the payoff. This pairing is what makes BSDEs the natural mathematical object underlying no-arbitrage pricing: solving the BSDE for a given payoff simultaneously produces the price and the hedge, rather than requiring two separate calculations.

BSDEs generalize the Black-Scholes framework to settings ordinary PDEs handle awkwardly — pricing under trading constraints, counterparty default risk, or nonlinear funding costs — because the equation can be made nonlinear in ways a standard PDE-based derivation resists.