Stochastic Problems in the Simulation of Labor Supply

Modern work in labor supply attempts to account for nonlinear budget sets created by government tax and transfer programs. Progressive taxation leads to nonlinear convex budget sets while the earned income credit, social security contributions, AFDC, and the proposed NIT plans all lead to nonlinear, nonconvex budget sets. Where nonlinear budget sets occur, the expected value of the random variable, labor supply, can no longer be calculated by simply 'plugging in' the estimated coefficients. Properties of the stochastic terms which arise from the residual or from a stochastic preference structure need to be accounted for. This paper considers both analytical approaches and Monte Carlo approaches to the problem. We attempt to find accurate and low cost computational techniques which would permit extensive use of simulation methodology. Large samples are typically included in such simulations which makes computational techniques an important consideration. But these large samples may also lead to simplifications in computational techniques because of the averaging process used in calculation of simulation results. This paper investigates the tradeoffs available between computational accuracy and cost in simulation exercises over large samples.