We establish a general Lagrangian for the moral hazard problem which generalizes the well
known first-order approach (FOA). It requires that besides the multiplier of the first-order
condition, there exist multipliers for the second-order condition and for the binding actions of
the incentive compatibility constraint. Some examples show that our approach can be useful
to treat the finite and infinite state space cases. One of the examples is solved by the second-
order approach. We also compare our Lagrangian with Mirrlees'.