Changes In Health Insurance At Retirement

The mandatory health insurance UEBMI does not shift when workers enter retirement; however, it should be noted that the cost of health insurance drops distinctly after they retire. Workers who have paid the premia of their health insurance for at least 15 years cease having to do so after they retire. As shown in Table 1, retirees in Shanghai, for example, also pay a lower deductible.

Insurance policies vary in coinsurance rate depending on the specific tiers that the hospital where the service was provided utilizes. Coinsurance incentivizes patients to seek health care at a lower tier hospital, unless it is necessary to move up to a higher tier. Table 2 shows the comparison of coinsurance rates pre- and post- retirement. Retirees pay about a 5% lower coinsurance rate for outpatient service, and a 7% lower rate for inpatient service.


According to the notation of the potential outcome framework, pY0, Y1q are the two potential outcomes a household may experience. Y0 is the measure of interest when an individual is not retired, and Y1 is the measure of interest when one is retired. In this case, D = 1 represents an individual who is retired, and D = 0 indicates that an individual is not retired. Therefore, an actual measure of medical expenditure Y can be represented by

In this paper, there are three outcome variables of interest. For i-th individual in the sample, Y1i is defined as total medical expenditures. Y2i is defined as medical expenditures that are covered by health insurance. Y3i is defined as out-of-pocket medical expenditures. Di is a binary variable indicating if an individual is retired or not. A discontinuity design (Thistlethwaite and Campbell (196o)) arises when Di depends on an observable variable Si, where Si measures the age of an individual, and, in support of Si, the probability of being retired changes discontinuously at the threshold s. In other words,

where s’ and s“ refer to individuals who are marginally above and below s, respectively.

According to Trochim and Thochim (1984), the distinction between sharp and fuzzy re­gression discontinuity designs depends on the size of the probability jump at the threshold. A sharp regression discontinuity design occurs if the retirement status deterministically depends on whether an individual’s age is above s; in other words, Di 1 (Si sq. A fuzzy design is applicable ifthe size of discontinuity at s is smaller than one.

Urban residents in China follow mandatory retirement policies, which are themselves pri­marily based on age requirements. The normal retirement age for females are either 5o or 55 as a cadre, while for males the normal retirement ages are 55 or 6o. In addition, a policy that was estab­lished in 1994 during the SOE (state-owned enterprises) reform in the 1990s states that a female can retire at age 45, 50 or 55 as a cadre, while a male can retire at age 50, 55 or 60.

In this paper, I focus on the peak age of retirement for females (age 50) and males (age 60); other retirement-age thresholds are not considered. A fuzzy design is used in this paper because the size of the discontinuity at age thresholds (again, age 50 for female, age 60 for male) is smaller than one, as reaching the age threshold does not necessarily imply that individuals are actually retired. Though the mandatory rule induces a higher probability of being retired, some individuals may choose not to comply with the retirement rule.

The local average treatment effects of retirement on medical expenses can be recovered for retired individuals around s.