We propose a generalized look-ahead estimator for computing
densities and expectations in economic models. We provide conditions
under which the estimator converges globally with probability
one, and exhibit the asymptotic distribution of the error. Our estimator
is more efficient than other Monte Carlo based approaches.
Numerical experiments indicate that the estimator can provide large
increases in accuracy and speed relative to traditional methods. Particular
applications we consider are the stochastic growth model and
an income fluctuation problem.
|