Finance theory sets constraints on option pricing models. Yet empirical option pricing models either implicitly ignore some theoretical constraints or impose possibly misspecified parametric structure on them. Previous specification tests of such parametric structure often require the solution of certain density function which is difficult to obtain in close-form. In contrast, this paper proposes a new specification test without such requirement. I show how to derive a testing equation from the nonparametrically constructed theoretical constraints. Next, I provide a delta-hedging interpretation of this testing equation and show that when the option pricing model is correctly specified, the hedging error estimated from times series data should be consistent with the risk premia obtained from cross-sectional options data. I then investigate the statistical properties of the components of my testing equation and discuss how to adapt the testing procedure to specific parametric assumptions. A Monte Carlo example shows that the test has reasonable size and power. I also provide other empirical implications of the test statistics.

Field: Econometrics