The California Medicaid program added olanzapine and risperidone to its formulary in October 1997. This resulted in an immediate but temporary increase in the total number of patients initiating episodes of therapy (access effect) and the substitution of second-generation antipsychotics for older drugs. OBJECTIVE: Model how the formulary expansion altered drug selection decisions and then use these models to match patients for the purpose of estimating the cost impact of the formulary expansion. STUDY POPULATION: 19,221 olanzapine, 25,252 risperidone and 1,22,547 patient episodes using traditional drugs were classified into three groups: patients with no previous antipsychotic use (new), patients re-starting drug therapy and patients switching between antipsychotic medications without a break in therapy. METHODS: Multinomial logistic regression models of the drug selection process were estimated then used to calculate propensity scores for olanzapine and risperidone use using the post-expansion clinical decision criteria. Health care costs were then compared within quintile bins for these 3 treatment episode types. RESULTS: The formulary expansion significantly improved access for minorities, women and urban residents. However, the propensity-score matching of olanzapine and risperidone patients with patients using traditional antipsychotics (quintile bins) did not result in highly congruent patient populations. Patients using traditional antipsychotics consistently exhibited higher levels of health care use prior to the start of the treatment episode. Simple within-bin comparisons of health care costs found lower costs for olanzapine and risperidone patients. Differences based on multivariate cost models within each quintile bin greatly reduced these differences. However, continuous days of therapy were consistently higher for risperidone and olanzapine patients. CONCLUSIONS: Patients with schizophrenia treated with either risperidone or olanzapine achieved better drug therapy outcomes after the formulary expansion. The higher cost of drug therapy was largely offset by reductions in the use of other services.