The demand for medical care in Japan: initial findings from a Japanese natural experiment

Abstract

This study examines the effect of the 1997 increase in the coinsurance rate for household heads on the demand for medical care and estimates the price elasticity of demand using the change as a natural experiment. It analyses both outpatient and inpatient utilization by using health insurance claim data from 111 insurance associations. A differences-in-differences type estimator is employed with household heads as the treatment group and dependents as the control group. This represents the first comprehensive analysis of medical care demand in Japan using a natural experiment. The results indicate a price elasticity of outpatient care ranging from −0.05 to −0.06 but no significant effects on inpatient care of the increase in cost sharing. The price elasticity for outpatient care is lower than those from previous studies that have used observational comparisons of individuals in Japan and also smaller than those derived from a randomized experiment in the USA.

Acknowledgements

We are grateful to the Ministry of Health, Labour and Welfare for permitting us to use their micro data. We would like to thank Charles Y. Horioka, Colin R. McKenzie, Shizuka Sekita and the members of the graduate seminar at Osaka University for their helpful comments.

Notes

¹ For a comprehensive survey, see Zweifel and Manning (2000).

² Refer to Bhattacharya et al. (1996) for details of the health insurance system in Japan.

³ Refer to Yoshida and Takagi (2002) for details of the reform of 1997. Although individuals’ coinsurance rates for the employees’ health insurance scheme are designated by law, there is considerable variation within this group due to the additional reimbursement of costs above the monthly ceiling on out-of-pocket payments. In addition, the introduction of a fixed co-payment for prescription drugs as part of the reform of 1997 led to an increase in out-of-pocket payments for both household heads and dependants.

⁴ Expenditures are expressed in terms of points. One point is equivalent to 10 Japanese Yen. As is customary in Japanese studies of medical care expenditure, this paper treats points as the object of analysis.

⁵ Since a visit is classified as an initial visit if the individual has not visited a physician's office for more than three months, three months represents one period. The unit of analysis is therefore a person-period, of which the sample comprises 289 448 (2 year data) and 434 172 (3 year data), respectively.

⁶ The significantly positive coefficient of HeadXAfter in the case of the hospital admission rate in Table 3 suggests supplier-induced demand. This should be examined in future research.

⁷ Yoshida and Takagi (2002) concluded that only dependants were affected by the reform because the increased coinsurance rate for household heads affected overall household expenditure through the household budget constraint. Our coefficient for ‘After’ (representing the impact on the revision on dependants’ behaviour) is significantly negative in the case of outpatient visits but not significant in the case of expenditure per visit based on the two-year data. Therefore, the relationship between the behaviour of dependants and that of household heads remains unclear.

⁸ See Newhouse and the Insurance Experiment Group (1993).