بیمه سلامتی و پس انداز احتیاطی: تجزیه و تحلیل ساختاری

Starr-McCluer (1996) documented an empirical finding showing that US households covered by health insurance saved more than those without coverage, which is inconsistent with the standard consumption–saving theory. This study conducts a structural analysis and suggests that institutional factors, particularly, a social insurance or safety net system and an employment-based health insurance system, can account for this puzzling finding. A dynamic equilibrium model is built that combines these two institutions with heterogeneous agents making endogenous decisions regarding saving, the labor supply and health insurance when they are young. The model, in which agents save in a precautionary manner, can generate Starr-McCluerʼs empirical finding. The result implies that Starr-McCluerʼs results are not inconsistent with the standard theory of saving under uncertainty, but it does indicate that the standard saving regression model is unable to reveal the precautionary saving motive. Counterfactual experiments are performed to provide implications for empirical analyses.

Starr-McCluer (1996) argued that the standard consumption–saving theory, which implies that more insured households should save less due to precautionary motives, is inconsistent with the empirical finding that US households covered by private health insurance save more than comparable uninsured households. This paper takes a structural approach to revisit this issue and suggests that two institutions in the US, means-tested social welfare and employment-based health insurance (EHI), can account for the puzzling empirical finding documented by Starr-McCluer. Starr-McCluer studies the impact of private health insurance on household savings in the American working-age population and tests the precautionary saving hypothesis. Although several econometric methods are applied to control for other household characteristics and factors that also affect saving, the results indicate that health insurance coverage has a significant and positive effect on savings. Table 1 presents part of the empirical results. The coefficients of health insurance coverage (labeled ‘PHI coverage’) are significant and positive in all the three regressions regardless of the measure of assets. To conduct a structural analysis, I build a dynamic stochastic general equilibrium model where heterogeneous agents face uncertain retirement or death, idiosyncratic income and medical shocks, and make decisions regarding saving, the labor supply and health insurance. Markets are incomplete, and thus risk-averse agents in the economy have an incentive to save in a precautionary manner and to purchase health insurance. Two institutions, a social insurance (safety net) system and an employment-based health insurance system, are incorporated as key factors to examine the insurance–savings correlation. Although this study focuses on the saving–insurance decisions of the working-age population, social security and Medicare are also incorporated because they affect younger agents’ expectations of their lives after retirement

Considering a standard model of precautionary saving, a puzzling positive correlation between private health insurance coverage and household asset holdings in the US has been reported by Starr-McCluer (1996). This study suggested that those findings can be explained by the existence of two institutions: a large social insurance (safety net) system and an employment-based health insurance system. To analyze this issue, I built a dynamic stochastic general equilibrium model that incorporates these two institutions with heterogeneous agents making decisions regarding saving, labor supply, and health insurance endogenously. I showed that the model can generate the same empirical finding as found in Starr-McCluer (1996). This result does not depend on the assumptions of heterogeneity of risk aversion or preferences. These findings call into question the appropriateness of those empirical approaches that are directly based on the standard consumption–saving theory for testing the precautionary saving hypothesis. Applying the empirical approach in the model economy, the regression result is consistent with Starr-McCluer’s empirical finding, but the positive coefficient of PHI coverage does not imply the nonexistence of precautionary saving. In fact, households do have a motive to engage in precautionary saving in the model economy. The regression results must be interpreted carefully because the regression model captures the net saving effect of health insurance rather than the effect on precautionary savings exclusively. To appropriately test for the existence of precautionary saving or to investigate the substitution effect of insurance on precautionary saving, it is necessary to control for the effects caused by these institutional factors. The counterfactual experiments have demonstrated that if either the social insurance or the EHI system is eliminated, the regression would generate a negative coefficient of PHI coverage that supports the precautionary saving hypothesis. The analysis in this study indicates that an economy with a low-level social assistance and without the feature of employment-based insurance coverage would be a better environment for conducting an empirical study regarding this issue. Nevertheless, in practice it is generally difficult to distinguish the effects on saving caused by the institutions from the effect of insurance on precautionary saving. Further research is required to resolve this difficulty.