The first-order approach to principal-agent problems involves relaxing the constraint that the agent choose an action which is utility maximizing to require instead only that the agent choose an action at which his utility is at a stationary point. Although more mathematically tractable, this approach is generally invalid. This paper identifies sufficient conditions--the monotone likelihood ratio condition and convexity of the distribution function condition--for the first-order approach to be valid. The Pareto-optimal wage contract is shown to be nondecreasing in output under these same conditions.