On the Effectiveness of Common Random Numbers

Abstract

The common random number (CRN) simulation technique is a variance reduction method in which policy alternatives are tested against the same random input streams. The CRN literature suggests that positively correlated input streams will generate positively correlated policy responses and, therefore, that the variance of CRN estimators of response differences will be smaller than the variance of independent sample estimators. This paper reports simulation experiments with a typical inventory model for which the CRN technique induces negative correlation and thus augments variance. The experiments also show that CRN designs which assign separate random streams to each stochastic input and hold all streams common across policy tests usually, but not necessarily, yield maximum variance reduction.