The paper evaluates the German health care reform of 1997, using the individual number of doctor visits as outcome measure. A new econometric model, the Probit-Poisson-log-normal model with correlated errors, describes the data better than existing count data models. Moreover, it has an attractive structural interpretation, as it allows the reforms to have a different effect at different parts of the distribution. The overall effect of the reform was a 10 percent reduction in the number of doctor visits. The effect was much larger in the lower part of the distribution than in the upper part.