This paper investigated the performance of a propensity score analysis by a simulation study that involved using 12 covariates consisting of 3 different variable types. The three types of covariates included covariates related only to the outcome measure, covariates related only to assignment of treatment, and covariates related to both outcome and assignment. Logistic regression was used as the propensity score computation stage, and propensity score matching was used as the treatment effect estimation stage. The study revealed that it is paramount to include covariates related to both outcome and assignment. On the other hand, covariates only related to outcome or assignment did not help estimate the treatment effect better.