Heterogeneous effect of coinsurance rate on the demand for health care: a finite mixture approach

The paper exploits finite mixture (latent class) models to account for consumer heterogeneity in estimating the effect of coinsurance rate on the demand for health care. The parametric analysis employs a two-part model, a Tobit model, and generalized linear models with latent classes. The non-parametric analysis uses matching estimators in each latent class to construct a control group of consumers and measure the average treatment effect of the natural experiment with a rise in nominal coinsurance rate. The paper exploits the 2000-2008 data of the Japanese Panel Survey of Consumers and the 2008 data of Japan Household Panel Survey. The estimations demonstrate a significant negative effect of nominal coinsurance rate on the demand for health care. The effect is primarily noticeable in the latent class of consumers with high health care demand, who constitute 21% of the sample. Our finding with the latent class models with Japanese data, where the assignment of insurance plans is exogenous, is similar to the results with the RAND Health Insurance Experiment data, where the assignment was randomized.