Table 1.

Simple Example of the Propensity Score Matching Paradox

Sex and Racea	Index Exposure (n = 4)	Reference Exposure (n = 8)	Total (n = 12)	Stratum PS
Male
White	1	2	3	0.3
Not white	1	2	3	0.3
Female
White	1	2	3	0.3
Not white	1	2	3	0.3

Abbreviation: PS, propensity score.

^a The example population represented in this table contains index and reference exposure groups that are perfectly balanced on sex and race. The propensity score values for all 12 units are equal. One-to-one propensity score matching without replacement would be expected to increase the underlying covariate imbalance in the matched data set, compared with the prematched data set.