Objective: Various propensity score matching techniques are used in observational studies to reduce selection and confounding bias. The purpose of this study was to evaluate the four most popularly used matching methods  namely Mahalanobis metric matching within calipers of propensity scores, caliper matching with and without replacement and greedy matching for making elderly antipsychotic users comparable.       Methods: IMS LifeLink™ Claims were utilized to identify elderly patients using atypical and typical antipsychotics. Eighty covariates including demographics, hospitalization, co-morbidities and co-medications were used to match typical and atypical antipsychotic users using propensity scores matching. Propensity matching methods were evaluated on the basis of following criteria: (1) Number of variables which remains significant after matching using t-test and chi-square; (2) Percentage bias reduction for the variables which remained significant after matching; (3) Mean difference in propensity scores as a percentage of average standard deviation (SD); and (4) Density estimates of the propensity scores of the two groups using the Kolmogorov-Smirnov test. Results: The four matching methods reduced bias by making two groups comparable. However greedy matching yielded the best results when  the four criteria were applied. Only 5 explanatory variables remained significant after greedy matching compared to 36, 43 and 9 with Mahalanobis metric matching, and caliper matching with and without replacement, respectively. More than 90% bias reduction was obtained through all the matching methods. Mean difference as a percentage of the average SD was 0% with greedy and caliper matching with replacement and these were the only techniques that produced propensity scores densities with insignificant differences. Conclusion: The greedy matching technique was found to be efficient in matching different classes of antipsychotic users. Although the efficiency of matching methods could differ based on the study sample and availability of covariates, a priori criteria can be useful in selecting the most appropriate matching technique.