Although research has shown that propensity score (PS) stratification is an effective method for reducing selection bias in quasi-experiments, many researchers tend to favor other PS methods over stratification. Part of this may be due to the unreliability of effects found across studies that use PS stratification. One reason for these inconsistencies may be due to the way the stratification procedure is done. Two common approaches are to (1) compute a t-test for each stratum and average effect sizes across strata or (2) use a treatment-by-strata factorial analysis of variance to estimate the treatment effect. This study compared how these two methods; as well how the number of strata (3, 5 or 10) and sample size ratio (equal n, 1:2 or 1:3); affect common support, covariate balance, and bias reduction.