My variable of interest is ethnicity. Ethnicity has two categories: natives or migrants. I want to see whether ethnicity affects charitable behavior. Since the migrant status is not random, therefore I want to perform a PSM to check the balancing of observables. Even though t test is showing insignificance between treated (Here, migrants) and untreated (Natives), the % bias is more than 5% for some covariates. Is that a problem? Or should I only focus on the t test?