Adverse selection in a voluntary Rural Mutual Health Care health insurance scheme in China $

Transcription

1 Social Science & Medicine 63 (2006) Adverse selection in a voluntary Rural Mutual Health Care health insurance scheme in China $ Hong Wang a,, Licheng Zhang b, Winnie Yip c, William Hsiao c a Yale University School of Medicine New Haven, CT 06520, USA b Beijing University School of Public Health, China c Harvard University School of Public Health, USA Available online 24 April 2006 Abstract This study examines adverse selection in a subsidized voluntary health insurance scheme, the Rural Mutual Health Care (RMHC) scheme, in a poor rural area of China. The study was made possible by a unique longitudinal data set: the total sample includes 3492 rural residents from Logistic regression was employed for the data analysis. The results show that although this subsidized scheme achieved a considerable high enrollment rate of 71% of rural residents, adverse selection still exists. In general, with worse health status are more likely to enroll in RMHC than with better health status. Although the household is set as the enrollment unit for the RMHC for the purpose of reducing adverse selection, nearly 1/3 of enrolled are actually only partially enrolled. Furthermore, we found that adverse selection mainly occurs in partially enrolled. The non-enrolled in partially enrolled have the best health status, while the enrolled in partially enrolled have the worst health status. Pre-RMHC, medical expenditure for enrolled in partially enrolled was yuan per capita per year, which is 1.7 times as much as the pre-rmhc medical expenditure for non-enrolled in partially enrolled. The study also reveals that the pre-enrolled medical expenditure per capita per year of enrolled was 9.6% higher than the pre-enrolled medical expenditure of all residents, including both enrolled and non-enrolled. In conclusion, although the subsidized RMHC scheme reached a very high enrollment rate and the household is set as the enrollment unit for the purpose of reducing adverse selection, adverse selection still exists, especially within partially enrolled. Voluntary RMHC will not be financially sustainable if the adverse selection is not fully taken into account. r 2006 Elsevier Ltd. All rights reserved. Keywords: Voluntary health insurance; Community health insurance; Adverse selection; Rural; China $ The data used in this study is from a Rural Mutual Health Care (RMHC) social experimental study carried out by a Harvard-led research team. Corresponding author. Tel.: ; fax: addresses: hong.wang@yale.edu (H. Wang), pkuzlc@hotmail.com (L. Zhang), wyip@hsph.harvard.edu (W. Yip), hsiao@hsph.harvard.edu (W. Hsiao). Introduction One of the major concerns about voluntary health insurance is adverse selection. In the voluntary health insurance market, there is asymmetric information between the seller of insurance and the buyer. In particular, potential consumers know /$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi: /j.socscimed

2 H. Wang et al. / Social Science & Medicine 63 (2006) their own risk levels but insurers are unable to distinguish among risks. Therefore, high-risk consumers are able to purchase insurance at a premium that is based on a lower-risk group (or the average risk of the group). This situation is referred to as adverse selection (Rothschild & Stiglitz, 1976). When adverse selection exists, premiums set according to the average risk of the general population will not be sufficient to cover claims, therefore the insurance will not be financially sustainable (Cutler & Zeckhauser, 2000; Feldstein, 1993; Pauly & Nicholson, 1999). There is substantial literature examining adverse selection in insurance markets in developed countries. These studies have been well summarized in the Handbook of Health Economics (Cutler & Zeckhauser, 2000). The list of literature on adverse selection continues to grow due to continuing efforts to study existing and newly developed insurance schemes (Batavia & DeJong, 2001; Finkelstein, 2004; Jack, 2002; Sapelli & Vial, 2003; Savage & Wright, 2003; Simon, 2005). This literature is categorized into one of following categories based on its analytical strategy: (1) examining enrollment conditional on health status, (2) examining enrollment conditional on health service utilization, (3) examining health status conditional on health insurance, and (4) examining health service utilizations conditional on insurance after controlling for moral hazard effects (Cutler & Zeckhauser, 2000). Regardless of the analytical strategy used, these studies provide evidence that adverse selection is quite significant. Adverse selection has also been discussed extensively in Community-Based Health Insurance (CBHI) schemes in developing countries (Atim, 1998; Carrin, 2003a; Jakab & Krishnan, 2004). However, the results of empirical studies of adverse selection are mixed. A study from Zaire revealed that there was significant adverse selection for women of productive age who enrolled in a prepayment scheme for hospital care, and therefore used more hospital delivery services (Noterman, Criel, Kegels, & Isu, 1995). The results from a study in Burundi demonstrated that adverse selection was not a major problem at the individual level since the illness occurrence as measured by episode of illness per person was almost identical for household that purchased CAM and those that did not (CAM is a national health card insurance implemented by the government of Burundi beginning in 1984). However, there is household adverse selection since large are more likely to join CAM as it has a fixed price irrespective of household size (Arhin, 1994). One recent study from the Philippines showed that the adverse selection does not exist in their Micro Health Insurance Units (MIUs) because the morbidities of insured and noninsured groups did not differ from each other (Dror et al., 2005). In addition, although a series of methods have been utilized to minimize adverse selection, such as making the enrollment unit the household rather than the individual and imposing a waiting time before obtaining the benefit from the insurance scheme after enrollment (Atim, 1998), the effects of these methods on the reduction of adverse selection have not been assessed empirically. The objective of this study is to examine adverse selection in a subsidized voluntary Rural Mutual Health Care (RMHC) insurance scheme in a poor rural area of China (Hsiao et al., 2004). The hypothesis of this study is that adverse selection would not be a major problem in this scheme for two reasons. First, RMHC is a subsidized CBHI scheme. In October 2002, China announced a new funding strategy for a newly established CBHI. The government would encourage farmers to participate in the new CBHI by providing each participant with an annual subsidy of yuan ($ USD) (Liu, 2004). In order to be consistent with this government policy, RMHC was also subsidized 20 yuan per participant per year, in addition to the premium for enrollment (Hsiao et al., 2004). With this subsidy, we expect that the enrollment would be high and the scheme would be not only attractive to the high-risk residents, but also to the low-risk. Therefore, adverse selection would be minimized. Second, in order to avoid/reduce adverse selection, the RMHC enrollment unit was set at the household level rather than the individual level. We expect that adverse selection might be reduced by enrolling with mixed health status if this policy could be fully implemented. Methods Background of RMHC and source of data The RMHC is a voluntary CBHI scheme, which was established by a Harvard led research team in Fengshan Township, Kaiyang County, Guizhou Province in China in According to government statistics, the population size was about 37,000 and the annual income per capita was about 2000

3 1238 ARTICLE IN PRESS H. Wang et al. / Social Science & Medicine 63 (2006) Chinese yuan (275 US dollars) in This township has one township hospital and 56 village doctors who serve in village health posts across the whole township. The purpose of this study is to examine if a subsidized community-based rural health insurance is a viable health care financing strategy in rural China since many efforts undertaken in the 1990s to reestablish CBHI failed (Bloom & Gu, 1997; Carrin et al., 1999; Hsiao, 1995; Hsiao & Liu, 1996; Liu, Hsiao, Li, Liu, & Ren, 1995). In addition to the individual s premium contribution of yuan (depending on which of the three benefits packages the participants selected), each participant will receive yuan from RMHC study to subsidize their premium to join the RMHC. Therefore, the participant s average total premium was yuan, which is twice as much as their own contribution. In order to avoid the exclusion of the poor from scheme, local government paid full premium to those who are identified as the poor (about 5% of total residents) using the criteria set by the local Civil Affair Agency (Hsiao et al., 2004). The scheme also made the household the enrollment unit in order to minimize adverse selection (Hsiao et al., 2004). This study was made possible by a unique data set, which combines the measures of health status, medical expenditure, and other socio-economic and demographics variables from a baseline survey, conducted before RMHC was established, and measures of enrollment status from a first year evaluation survey, conducted as part of this study 1 year after RMHC was implemented. The baseline survey was conducted during the scheme s design period at the end of The study population was chosen through a multistage sampling process. In the first stage, random sampling was used to select 6 villages in the study area. In a second stage, all with family members in the high-risk population, as well as about one out of three other chosen at random, were selected. High risk was defined as: single elderly, the disabled, those with dementia, women who were pregnant in the previous year or at the time of interview, those admitted to the hospital in the previous year, or those with a severe health condition as diagnosed by the village doctor. The entire sample includes 1173 with 4160 residents from 6 sampled villages. One year after the implementation of RMHC, a follow-up evaluation survey was conducted in the end of Among residents participating in the baseline survey, 3492 rural residents in 1020 were successfully followed-up. The individual follow-up rate is 83.9% and the household follow-up rate is 87.0%. The sample used in this study is all rural residents who completed follow-up. Analytical model The choice of rural residents to join or not to join CHI is a discrete decision process consistent with qualitative choice models (Manshi & Mcfadden, 1981). From an economic perspective, the farmers choice of joining CHI is grounded in the comparison of the expected utility of having health insurance versus having no health insurance. If the expected utility to be derived from joining CHI at the stated premium is greater than having no CHI, an individual will opt to join, and pay for, the scheme. In the context of our study, all rural residents in the study area can be categorized into two groups based on their RMHC enrollment status: enrolled in and non-enrolled. Based on the household enrollment status, however, all can be categorized into four groups: enrolled in a fully enrolled household, non-enrolled in a nonenrolled household, enrolled in a partially enrolled household, and non-enrolled in a partially enrolled household (Fig. 1). Here, the fully enrolled are those where all members in the enrolled in RMHC. Non-enrolled are those where no members of the household enrolled in the RMHC. Both these two types of followed the rule that enrollment unit is at the household level rather than at the individual level. The partially enrolled are those that only selected household members enrolled in RMHC. These actually violate enrollment rules but are not identified when they enroll in the scheme. Since the dependent variables are binary choice variables, we adopted the logistic regression for this study (Stata Corporation, 2003). Three logistic regression analyses were conducted. The first one is full sample model. The individual enrollment status is the dependent variable and health status variable is the independent variable to observe adverse selection while controlling for all other socio-economic and demographic indicators. Furthermore, in order to investigate whether the adverse selection comes from fully enrolled or partially enrolled, two separate

4 H. Wang et al. / Social Science & Medicine 63 (2006) Fullyenrolled Enrolled Households in the community Enrolled Households Partiallyenrolled Enrolled Non-enrolled Non-enrolled Non-enrolled Fig. 1. Households/ enrollment status in RMHC in Fengsan. Table 1 RMHC enrollment status in Fengsan township Indicators Households Individuals Number of Percentage (%) Number of Percentage (%) Individual enrollment status Enrolled (70.2) Non-enrolled (29.8) Subtotal (100.0) Household enrollment status Full-enrolled (51.7) (51.7) Non-enrolled (18.2) (18.2) Partial-enrolled (30.1) Enrolled (18.6) Non-enrolled (11.5) Subtotal (100.0) (100.0) Note: weighted analysis results are in parentheses. restricted logistic models are applied to two subsamples analyses; the first one includes sub-sample of in fully enrolled and non-enrolled, and the second one includes sub-sample of in partially enrolled. Since we over-sampled the high-risk population, a sample weight variable was created based on different sampling probabilities for the with family member in the high-risk population and other for weighted analysis (Palta, 2001), and robust estimation was used in the regression analyses to control for potential cluster effects within. Dependent variables Enrollment status: Three binary choice variables (with enrolled ¼ 1 and non-enrolled ¼ 0) are created for the full sample and sub-sample analyses that are described in analytical model section. The results of a descriptive analysis of enrollment status are listed in Table 1. The results from Table 1 show that about 71% enrolled in the RMHC. These results also show that although about 82% enrolled in RMHC, only 57% are fully enrolled. One-quarter of enrolled violate the enrollment rules and are

5 1240 ARTICLE IN PRESS H. Wang et al. / Social Science & Medicine 63 (2006) categorized as partially enrolled. This figure is even higher if we take the sampling framework into consideration (weighted result). Independent variables Health status: For the purpose of examining adverse selection, we focused on estimating the effects of health status on enrollment in RMHC. Health indicators used in this analysis come from two questions in the baseline survey. The first question was Did you suffer from any health problem in the month prior to the interview? The second question was If you did suffer from any health problems in the month prior to the interview, how severe was this health problem? with the responses of not severe or severe. Health status is categorized as good health (G-health), moderate health (M-health), and poor health (P-health) groups. The description and descriptive analysis of health status indicator is listed in Table 2. Other control variables: Recognizing that health insurance enrollment may differ for reasons other than residents health status, we controlled for observable socio-economic and demographic variables. Specifically, we controlled for age, gender, household size, marital status, level of education, income, and the distance from the respondent s home to village, township, and county centers. It is worth mentioning here that the income used in this study is measured by self-reported, post-direct tax expenditure rather than by self-reported income because income is more likely to be misreported and actual expenditure is a better proxy for resources available for spending (Meyer & Sullivan, 2003). Instead of using a continuous variable, we created categorical income variables in the analysis in order to capture nonlinear effects. Table 2 describes these variables and provides summary statistics for each. Results The results in Table 2 show that non-enrolled have better health status than enrolled. Furthermore, the non-enrolled in partially enrolled have the best health status, while enrolled in partially enrolled have the worst health status. The descriptive results in Table 2 also reveal that rural residents socio-economic and demographic characteristics vary by enrollment status. For example, non-enrolled in partially enrolled are more likely to be younger, male, unmarried, and have a relatively high level of education. Recognizing these differences, we conducted multiple regression analysis to control for observable socioeconomic and demographic variables. Effects of health status on enrollment status The results of logistic regression analyses are presented in Table 3. Odd Ratio (OR) is used for the interpretation. The results in the full sample model display that the OR for the P-health is 1.6, which indicates that the individual with P-health status is almost 1.6 times as likely to enroll in RMHC as an individual with G-health status, after controlling for all other variables. This result is statistically significant at p ¼ 0:01 level. The OR for the M- health in the full sample model is also larger than 1. This result, however, is not statistically significant. In the sub-sample analysis, the OR for the P- health in the sub-sample model of in both fully and non-enrolled is This result, however, is not statistically significant. The ORs for the P-health and M-health in the subsample model of in partially enrolled is 2.59 and 1.26, respectively. However, only the OR for the P-health is statistically significant. The results of sub-sample analyses indicate that the adverse selection mainly comes from partially enrolled. Other variables that are related to the enrollment status In addition to the effect of health on the enrollment status, age, gender, household size, and education display significant correlations with enrollment status (Table 3). In the sub-sample model of in partially enrolled, the ORs for those aged and aged 55+ are all significantly larger than 1, which implies that the older residents are more likely to enroll in RMHC than in young residents. The OR for gender is significantly smaller than 1, which implies that female residents are more likely to enroll in RMHC than male residents in partially enrolled. The OR for household size is significantly larger than 1, which implies that the residents with a large family are more likely to be enrolled in RMHC than the residents with a small family. The OR for primary education is significantly smaller than 1, which implies that the residents with relatively

6 H. Wang et al. / Social Science & Medicine 63 (2006) Table 2 Description of variables Variable Description Total All enrolled All nonenrolled Enrolled in full-enrolled Non-enrolled in non-enrolled Enrolled in partialenrolled Non-enrolled in partialenrolled G-health 1 if no health problem in the previous month of interview, 0 otherwise. Omitted group M-health 1 if there is a minor health problem, 0 otherwise P-health 1 if there is a severe health problem, 0 otherwise Age if age between 0 and 14, 0 otherwise. Omitted group Age if age between 15 and 24, 0 otherwise Age if age between 25 and 34, 0 otherwise Age if age between 35 and 44, 0 otherwise Age if age between 45 and 54, 0 otherwise Age if age between 55 and above, 0 otherwise Female 1 if female, O otherwise. Omitted group Male 1 if male, 0 otherwise Hsize Mean number of household members Unmarried 1 if unmarried, 0 otherwise. Omitted group Married 1 if married, 0 otherwise Divorced 1 if divorced or separated, 0 otherwise Illiterate 1 if illiterate and agex6, 0 otherwise. Omitted group Elementary 1 if elementary school, 0 otherwise Primary 1 if primary school, 0 otherwise L-income 1 if post-direct tax expenditure is at the low 25% percentile, 0 otherwise. Omitted group. M-income 1 if post-direct tax expenditure is between 25% and 75% percentiles, 0 otherwise. H-income 1 if post-direct tax expenditure is at the high 25% percentile, 0 otherwise D-village Mean distance (km) from home to village health post. D-township Mean distance (km) from home to township hospital/health center. D-county Mean distance (km) from home to country hospital Total Note: All numbers here are weighted estimation.

7 1242 ARTICLE IN PRESS H. Wang et al. / Social Science & Medicine 63 (2006) Table 3 Logistic regression results (OR): the effects of health status on RMHC enrollment Variables Full sample model Individuals in both fully and non-enrolled Individuals in partial-enrolled M-health (0.1544) (0.1888) (0.2639) P-health (0.2593)*** (0.2577) (0.7217)*** Age (0.0741)*** (0.1499)* (0.1114)*** Age (0.1145)*** (0.1611)** (0.2159) Age (0.1526)** (0.1571)** (0.6836) Age (0.2490) (0.1794)** (3.2210)*** Age (0.3675) (0.3694) (1.8527)** Gender (0.0611) (0.0742) (0.1020)*** Hsize (0.0768)*** (0.1081)** (0.0801)*** Married (0.5895)*** (0.7376)** (0.5770) Divorced (0.6929)** (0.9546)** (0.4393) Elementary (0.1112) (0.1250) (0.2154) Primary (0.1288) (0.1793) (0.1654)** M-income (0.2309) (0.3565)* (0.1927) H-income (0.2543) (0.3918) (0.2829) D-village (0.0292)*** (0.0392)*** (0.0327) D-township (0.0053) (0.0049) (0.0154) D-county (0.0078) (0.0109) (0.0083) N Robust standard errors in parentheses. * Significant at 10%; ** significant at 5%; *** significant at 1%. higher education are more likely to enroll in RMHC than the residents with lower education. Pre-RMHC medical expenditures under different enrollment scenarios In order to examine the potential effect of adverse selection on the financial sustainability of the RMHC scheme, we estimate the pre-rmhc medical expenditures under different scenarios of enrollment. Since these expenditures occurred before the establishment of RMHC, they are not contaminated by the moral hazard effect. If RMHC were compulsory insurance scheme, every rural resident would enroll in RMHC and therefore no adverse selection would exist. Under this scenario, the pre-rmhc medical expenditure would be yuan per capita per year (Table 4).

8 H. Wang et al. / Social Science & Medicine 63 (2006) Table 4 Pre-RMHC medical expenditure per capita per year in different enrollment scenarios Enrollment status Number of Pre-RMHC medical expenditure (yuan) Total residents Actual enrollment status All enrolled All non-enrolled Enrolled in full-enrolled Non-enrolled in non-enrolled Enrolled in partial-enrolled Non-enrolled in partial-enrolled If all in partial-enrolled do not enroll in RMHC Enrolled Non-enrolled If all in partial-enrolled enroll in RMHC Enrolled Non-enrolled All numbers here are weighted estimation. Under the voluntary scheme, however, adverse selection exists, as identified in the previous section. The results in Table 4 demonstrate that the effect of adverse selection on medical expenditure is quite substantial. Pre-RMHC medical expenditure for enrolled was yuan per capita per year, which is 1.4 times as much as the pre-rmhc medical expenditure for non-enrolled (202.5/142.7). The largest difference is between enrolled in partially enrolled and non-enrolled in partially enrolled. Pre-RMHC medical expenditure for enrolled in partially enrolled was yuan per capita per year, which is 1.7 times as much as the pre-rmhc medical expenditure for non-enrolled in partially enrolled (206.6/123.7). Although the medical expenditures are very different among enrolled and non-enrolled residents, due to the high enrollment rate, the annual per capita pre-rmhc medical expenditure for the enrolled group is not dramatically different from the annual per capita medical expenditure for overall residents. The results in Table 4 show that the annual per capita pre-rmhc medical expenditure for the enrolled was yuan, which is about 9.6% higher than the pre-rmhc medical expenditure of overall residents. There are two extreme scenarios if the enrollment rules of household as enrollment unit were fully enforced. The first scenario is that all who are in partial-enrolled would be excluded. In this case, we would be able to eliminate adverse selection in partially enrolled. However, we would also lose healthy in those partially enrolled and the enrollment rate would drop. The annual per capita pre- RMHC medical expenditure for the enrolled group would be yuan, which would still be about 8.8% higher than the pre-rmhc medical expenditure of overall residents, which was yuan. The second scenario is to force all who are in partially enrolled to participate in RMHC. In this case, we would also be able to eliminate adverse selection in partially enrolled and increase enrollment as well. The annual per capita pre-rmhc medical expenditure for the enrolled group would be yuan, which would be about 3.6% higher than the pre-rmhc medical expenditure of overall residents. Discussion and policy implication Adverse selection is one of the major concerns for any voluntary health insurance scheme. Using a unique follow-up data set, we were able to examine how the pre-enrollment health status of rural residents influences their decisions to enroll in a subsidized, voluntary RMHC insurance scheme in a poor rural area of China.

9 1244 ARTICLE IN PRESS H. Wang et al. / Social Science & Medicine 63 (2006) The results from this study display that even with the premium subsidy, 29% of residents still do not enroll in RMHC, which verified the conclusion that the universal coverage will not be reached by reduction in health insurance premiums (Chernew, Frick, & McLaughlin, 1997). The results of this study further display that adverse selection in RMHC is very significant, especially in partially enrolled. In order to avoid/reduce adverse selection, the RMHC enrollment unit was set at the household level rather than the individual level. We expect adverse selection might be reduced by enrolling residents with mixed health status if this policy was fully implemented. However, the results from this study demonstrate that this enrollment rule is not well enforced in RMHC. Although about 82% enrolled in the RMHC, only 52% of are fully enrolled. Over 30% of violate enrollment rule and are categorized as partially enrolled, which means that only selected household members enroll in RMHC. These enrolled are more likely to have P-health status and to be elderly. Their pre-rmhc medical expenditure is 1.7 times as much as the pre-rmhc medical expenditure of those who are in the same, but not enrolled in RMHC. The results of this study also demonstrate that adverse selection also exists between enrolled in fully enrolled and nonenrolled in non-enrolled. However, the adverse selection in these two groups is relatively smaller than that in partially enrolled. People who have P-health status are slightly more likely to be enrolled in fully enrolled than non-enrolled in non-enrolled. The pre-rmhc medical expenditure of enrolled in fully enrolled is 1.3 times as much as the pre- RMHC medical expenditure of non-enrolled in non-enrolled. Due to the high enrollment rate, the impact of adverse selection on medical expenditure was not as significant as we expected. However, it is still a major threat to the financial sustainability of RMHC, if this effect is not fully taken into account. If we assume that the pre-rmhc medical expenditures differences in different groups with different enrollment status represent the post- RMHC medical expenditures (after RMHC is implemented) differences (which means that the other effects, such as moral hazard, are controlled), then RMHC would have about 10% deficit, which could lead to bankruptcy. The results of this study show that this deficit can be reduced to 3.6% by an extreme case, which is to force all in partially enrolled to enroll in RMHC. However, this is unlikely to happen, since once this household enrollment rule is enforced rigorously, some of those partially enrolled may drop out of RMHC. Our results show that once those partially enrolled drop out, healthy in these also drop out, thus the deficit due to adverse selection may not be significantly reduced. This result is consistent with analytical results from India (Ahuja & Jutting, 2003). Even if RMHC is able to enroll in all in partially enrolled, the deficit due to adverse selection will not be eliminated according to these study results. This study was made possible by a unique data set, which combines measures of health status and medical expenditure from a baseline survey, conducted before RMHC was established, and the measures of enrollment status from the first year evaluation survey, conducted 1 year after RMHC was implemented, taken by this study. The baseline survey sample includes 4160 residents, 3492 of whom were followed-up successfully in the evaluation survey. The follow-up rate is 84%. We examined the differences between those who completed follow-up and those who were lost to followup and found that the residents who were lost to follow-up were more likely to be healthy, unmarried, and farther from village center than the residents who completed follow-up. The other characteristics of these two groups, such as gender, age, education level, income, family size, the distance to township and county centers, are not different from each other. If these residents who were lost to follow-up were more likely to enroll in the RMHC, our results would overestimate the effects of adverse selection. If they were less likely to enroll in the RMHC, our results would underestimate the results of adverse selection. We also realized that it is possible that over time, residents learning and experience of the RMHC may change adverse selection. Therefore, the results presented here should be viewed in the context of the first year of our program. The results from this study raise the question on whether a voluntary-based CBHI is a vital approach in rural China. This question is also raised by World Health Organization documents (Office of the

10 H. Wang et al. / Social Science & Medicine 63 (2006) WHO Representative in China, 2004). Administratively, the cost of reducing adverse selection within will be very high since the in rural China are defined very vaguely, and the scheme may not able to obtain much benefit by enforcing this rule because healthy people in those partially enrolled may also drop out of the scheme. Therefore, the choices such as collective membership, a compulsory waiting period, increasing the enrollment rate by providing an even higher enrollment subsidy, or compulsory universal coverage might be the options of further reducing/ eliminating adverse selection (Atim, 1998; Carrin, 2003b; Office of the WHO Representative in China, 2004). References Ahuja, R., & Jutting, J. (2003). Design of incentives in community based health insurance schemes. New Delhi: Indian Council for Research on International Economic Relations. Arhin, D. C. (1994). The health card insurance scheme in Burundi: A social asset or a non-viable venture? Social Science & Medicine, 39(6), Atim, C. (1998). Contribution of mutual health organizations to financing, delivery, and access to health care: Synthesis of research in nine west and central African countries, Vol. 18. Bethesda, Maryland: Partnership for Health Reform, Abt Associates Inc. Batavia, A. I., & DeJong, G. (2001). Disability, chronic illness, and risk selection. Archives of Physical Medicine and Rehabilitation, 82, Bloom, G., & Gu, X. (1997). Health sector reform: Lessons from China. Social Science & Medicine, 45(3), Carrin, G. (2003a). Community based health insurance schemes in developing countries: Facts, problems and perspectives. Geneva: World Health Organization. Carrin, G. (2003b). Community based health insurance schemes in developing countries: facts, problems and perspectives. Geneva: WHO. Carrin, G., Ron, A., Yang, H., Wang, H., Zhang, T., Zhang, L., et al. (1999). The reform of the rural cooperative medical system in the People s Republic of China: Interim experience in 14 pilot counties. Social Science & Medicine, 48, Chernew, M., Frick, K., & McLaughlin, C. G. (1997). The demand for health insurance coverage by low-income workers: Can reduced premiums achieve full coverage? Health Services Research, 32(4), Cutler, D. M., & Zeckhauser, R. J. (2000). The anatomy of health insurance. In A. J. Culyer, & J. P. Newhouse (Eds.), Handbook of health economics (pp ). Amsterdam, The Netherlands: Elsevier Inc. Dror, D. M., Soriano, E. S., Lorenzo, M. E., Sarol, J. N., Jr., Azcuna, R. S., & Koren, R. (2005). Field based evidence of enhanced healthcare utilization among persons insured by Micro Health Insurance Units in Philippines. Health Policy, 73, Feldstein, P. J. (1993). The demand for health insurance. In P. J. Feldstein (Ed.), Health care economics (pp ). Albany, New York: Delmar Publishers Inc. Finkelstein, A. (2004). Minimum standards, insurance regulation and adverse selection: Evidence from the Medigap market. Journal of Public Economics, 88, Hsiao, W. C. L. (1995). The Chinese health-care system lessons for other nations. Social Science & Medicine, 41(8), Hsiao, W. C. L., & Liu, Y. L. (1996). Economic reform and health lessons from China. New England Journal of Medicine, 335(6), Hsiao, W., Wang, H., Wang, L. S., Gao, J. M., Zhang, L. C., & Xue, Q. X. (2004). Introduction of China rural mutual health care system. Chinese Health Economics(7), 5 8. Jack, W. (2002). Equilibrium in competitive insurance markets with ex ante adverse selection and ex post moral hazard. Journal of Public Economics, 84, Jakab, M., & Krishnan, C. (2004). Review of the strengths and weaknesses of community financing. In A. S. Preker, & G. Carrin (Eds.), Health financing for poor people: Resource mobilization and risk sharing (pp ). Washington, DC: The World Bank. Liu, Y. (2004). Development of the rural health insurance system in China. Health Policy and Planning, 19(3), Liu, Y. L., Hsiao, W. C., Li, Q., Liu, X. Z., & Ren, M. (1995). Transformation of China s rural health care financing. Social Science & Medicine, 41(8), Manshi, C., & Mcfadden, D. (1981). Structural analysis of discrete data with econometric applications. Cambridge, MA: MIT Press. Meyer, B. D., & Sullivan, J. X. (2003). Measuring the well-being of the poor using income and consumption. Working paper Cambridge, MA: National Bureau of Economic Research. Noterman, J. P., Criel, B., Kegels, G., & Isu, K. (1995). A prepayment scheme for hospital care in the Masisi District in Zaire: A critical evaluation. Social Science & Medicine, 40(7), Office of the WHO Representative in China. (2004). Implementing the new cooperative medical schemes in rapidly changing China: Issues and options. Beijing, China: Office of the WHO Representative in China. Palta, M. (2001). Advanced quantitative methods in population Health: Extensions of ordinary regression. Madison, Wisconsin, pp Pauly, M., & Nicholson, S. (1999). Adverse consequences of adverse selection. Journal of Health Politics, Policy and Law, 24(5), Rothschild, M., & Stiglitz, J. (1976). Equilibrium in competitive insurance markets: An essay on the economics of imperfect information. Quarterly Journal of Economics, 90, Sapelli, C., & Vial, B. (2003). Self-selection and moral hazard in Chilean health insurance. Journal of Health Economics, 22, Savage, E., & Wright, D. J. (2003). Moral hazard and adverse selection in Australian private hospitals: Journal of Health Economics, 22, Simon, K. I. (2005). Adverse selection in health insurance markets? Evidence from state small-group health insurance reforms. Journal of Public Economics, 89, Stata Corporation (2003). Stata Statistical Software Release 8.0. College Station, TX: Stata Corporation.