A simple Monte Carlo calibration approach is implemented in a GE model with uninsurable employment risk to quantitatively study the optimal replacement rate of a public unemployment insurance (UI) scheme. The optimal UI sampling distribution is found to be bimodal.