A simulation study is performed to examine the precision of design effect in single-stage and two-stage cluster sampling without stratification and equal cluster sizes. Interest is focused on the effect on precision of such factors as sample size, number of stages of the sample, and intra-cluster correlation. The precision of simple random sample formula relative to that of the proper formula in estimating variance of the mean is also investigated. It is shown that simple random sample estimator, though biased, is more precise than the proper estimator for all cases considered. Some results concerning the distribution of design effect are also given.