Health insurance and subjective well-being: evidence from integrating medical insurance across urban and rural areas in China

Abstract

Health insurance coverage and the risk protection it provides may improve enrollees’ subjective well-being (SWB), as demonstrated, e.g. by Oregon Medicaid’s randomized expansion significantly improving enrollees’ mental health and happiness. Yet little evidence from low- and middle-income countries documents the link between insurance coverage and SWB. We analyse individual-level data on a large natural experiment in China: the integration of the rural and urban resident health insurance programmes. This reform, expanded nationally since 2016, is recognized as a vital step towards attaining the goal of providing affordable and equitable basic healthcare in China, because integration raises the level of healthcare coverage for rural residents to that enjoyed by their urban counterparts. This study is the first to investigate the impact of urban–rural health insurance integration on the SWB of the Chinese population. Analysing 2011–18 data from the China Health and Retirement Longitudinal Study in a difference-in-difference framework with variation in the treatment timing, we find that the integration policy significantly improved the life satisfaction of rural residents, especially among low-income and elderly individuals. The positive impact of the integration on SWB appears to stem from the improvement of rural residents’ mental health (decrease in depressive symptoms) and associated increases in some health behaviours, as well as a mild increase in outpatient care utilization and financial risk protection. There was no discernible impact of the integration on SWB among urban residents, suggesting that the reform reduced inequality in healthcare access and health outcomes for poorer rural residents without negative spillovers on their urban counterparts.

China, life satisfaction, subjective well-being, universal healthcare coverage, urban–rural health insurance integration

Key messages

• We analyse insurance coverage and subjective well-being (SWB) based on a large natural experiment in China: the integration of the rural and urban resident health insurance programmes.

• This study is the first to investigate the impact of urban–rural health insurance integration on the SWB of the Chinese population.

• The integration policy significantly improved the life satisfaction of rural residents, especially among low-income and elderly individuals.

• The positive impact of the integration on SWB appears to stem from the improvement of rural residents’ mental health and associated increases in some health behaviours, as well as a mild increase in outpatient care utilization and financial risk protection.

Introduction

The primary role of insurance is to reduce individual risk through risk pooling and risk sharing (Zeckhauser, 1970). Exposure to risk can lead to worry and mental strain; therefore, it is not surprising that better health insurance can improve subjective well-being (SWB). Indeed, in 1946, the World Health Organization defined health as a state of physical, mental and social well-being, not merely the absence of disorders and infirmity (World Health Organization, 1946, 2020).

The improvement in well-being from insurance may be especially salient for low-income populations. For example, evidence from the randomized expansion of Medicaid in Oregon demonstrated that enrolment in Medicaid increased self-reported overall happiness 1 year later by about 32% (Finkelstein et al., 2012). Such effects may be even more prominent for low-income populations outside of high-income countries (Erlangga et al., 2019). Yet evidence about the impact of insurance on mental health and SWB is relatively limited despite rich and varied international experiences of expanding and integrating health insurance programmes to achieve universal health coverage (UHC), as in Japan, South Korea, Germany and Mexico (Busse, 2001; Kwon, 2003; Gakidou et al., 2006; Kondo and Shigeoka, 2013).

Raising the health insurance coverage of vulnerable groups is widely considered the key step towards achieving UHC, especially for low- and middle-income economies (World Health Assembly, 2005; Lagomarsino et al., 2012). China is an especially important case to study, given its large population and recent expansion of subsidized health insurance.

The Chinese government launched the urban–rural health insurance integration reform in 2016 to improve equity by pooling the New Rural Cooperative Medical Scheme (NRCMS for the rural population) and the Urban Resident Basic Medical Insurance (URBMI for non-working urban residents) to provide the same health service package and health insurance benefits for rural and urban residents. This reform constitutes an essential step in China’s health system reform to provide affordable and equitable basic healthcare, which is in line with the basic concept of UHC. As the integration reform is still in progress, there is an inadequate understanding of integration effects at the national level. This study aimed to investigate the effect of integration on the SWB of rural and urban residents.

This reform, expanded nationally since 2016, raises the level of healthcare coverage for rural residents to that enjoyed by their urban counterparts (those not covered by urban employee insurance). Improving health equality and the well-being of beneficiaries is at the core of the policy. As early as in 2009, the Chinese State Council promulgated the New Health Care Reform Plan, planning to gradually realize an integrated health insurance system. However, progress was slow, primarily due to lack of agreement on which governance structure would be suitable (Meng et al., 2015). A few localities, such as Shanghai, Chengdu, Chongqing and Tianjin, piloted integration of the rural and urban schemes before 2016. The integration reform was nationally started in 2016 after the State Council issued the ‘Opinions on Integrating the Basic Health Insurance System for Urban and Rural Residents’.¹

To our knowledge, this study is the first to provide national-level evidence of the impact of integration on residents’ SWB, which we measured by self-reported life satisfaction. Data come from five waves of the China Health and Retirement Longitudinal Study (CHARLS) between 2011 and 2020 (the latest 2020 data for robustness check).² Because different jurisdictions launched the integration policy at different times, we employed a difference-in-difference (DID) model with variation in the treatment timing to estimate impacts.

Previous studies of China’s insurance integration mainly focus on the impacts on healthcare utilization and physical health improvements with mixed findings (Ma et al., 2016; Shan et al., 2018; Yang et al., 2018; Li et al., 2019; Wang et al., 2019; Huang and Wu, 2020; Ren et al., 2022; Zhou et al., 2022). Apart from differences in population studied, time periods of data and analytic methods, these inconsistent findings may be due to a focus on the short-term impacts of the integration reform during the first few years of implementation, which may underestimate the longer-term overall effects of Urban–rural Residents’ Basic Medical Insurance (URRBMI). Most studies use data before 2016 to evaluate integration impacts. Since the State Council issued opinions on integrating urban and rural basic medical insurance systems in 2016, the integration reform has entered a period of rapid development nationwide. However, there is inadequate evidence to reflect the latest effects of integration reform.

Compared with previous studies, this study covers the longest time span—from 2011 to 2020—and provides distinctive evidence about dynamic impacts of the integration using a method combining an event study with DID. Our empirical results suggest that the integration policy significantly improved the life satisfaction of rural residents. The impact on life satisfaction is consistent and gradually increasing, at least within the following 4 years. Elderly people and lower-income residents in rural areas benefit most from the integration policy. The positive impact of integration on SWB appears to stem from the improvement in rural residents’ health status, especially mental health, as well as a mild increase in outpatient care utilization and health behaviours, a reduction in cost-sharing rates and an increase in willingness to choose health services at local government-owned facilities. There was no discernible impact of integration on SWB among urban residents, as expected. Rural residents with other chronic diseases (i.e. other than depressive symptoms) appear to experience minor improvements in well-being.

The findings overall imply that UHC may improve SWB, especially for vulnerable groups in low- and middle-income economies. Our results also suggest possibilities for further strengthening urban–rural health insurance in China, to enhance benefits for rural residents with other chronic conditions like hypertension and diabetes.

A brief review of literature

Health insurance plays a major role in reducing financial risk through the mechanism of risk pooling or risk sharing and decreased financial barriers to healthcare access. The association between health insurance and SWB (e.g. happiness, life satisfaction) has been of long-standing interest to researchers and policymakers seeking to understand the broader welfare impacts of gaining health insurance coverage (Kim and Koh, 2022).

Health insurance coverage may contribute to higher SWB through several channels. First, health insurance coverage can reduce worry about future healthcare expenditure risks, and thereby alleviate the concern that motivates precautionary saving (Arrow, 1963; Atella et al., 2006). As such, this ex-ante risk reduction can induce risk-averse individuals to be more satisfied with their lives. Second, health insurance can reduce out-of-pocket healthcare spending for patients when they utilize services, protecting them from heavy financial burden and even catastrophic medical expenditures, which can directly increase their satisfaction (Finkelstein et al., 2012; Baicker et al., 2013; Zhou et al., 2017). Third, health insurance can improve access to healthcare and reduce unmet need because of credit constraints or inability to pay, thus reducing underuse of beneficial services and improving health outcomes, all of which directly relate to personal happiness (Pan et al., 2016; Argys et al., 2020; Borgschulte and Vogler, 2020). Moreover, the increased human capital that insurance can support (e.g. through improved earnings of a healthier worker and better educational attainment of healthier children) may raise SWB, although insurance can only imperfectly address life-course exposures shaping the well-being of middle-aged and older adults (Huang et al., 2013). Whether the above channels work is of great interest for policymakers since expanding health insurance coverage, including subsidies for the poor, can create large fiscal pressure, while reforms may raise expectations (e.g. of parity with formal sector employee coverage) that cannot be completely fulfilled and may adversely impact SWB.

The integration reform of China’s urban–rural health insurance means that insured residents, especially rural residents, are covered by more generous health insurance with improved benefit packages. Several empirical studies have revealed the positive impacts of the integration policy on promoting access to healthcare (Li et al., 2019), and reducing inequality of healthcare and benefits (Yang et al., 2018; Wang et al., 2019). For instance, using CHARLS data from 2011, 2013, and 2015, Huang and Wu (2020) showed that integration significantly increases rural residents’ inpatient care utilization. Zhao et al. (2019) used the Fifth National Health Services Survey in 2013 to show evidence of the pro-poor contribution of integration for inpatient care utilization. Ren et al. (2022) used 2013 and 2015 CHARLS data to provide novel evidence of reducing pro-rich inequity in outpatient benefits. They found the reform of URRBMI integration had little effect on inpatient benefits. Also using CHARLS data from the early years of integration, Ma et al. (2016) find that the outpatient care utilization of rural residents significantly increased after integration, but inpatient care utilization did not. Zhou et al. (2022) also find that the integration had no impact on inpatient care utilization for rural and urban residents, based on the analysis of the 2015 and 2017 China Household Finance Survey. They find that integration significantly improved financial risk protection and the self-assessed health of rural residents, and these positive impacts were particularly salient among poor rural residents. Shan et al. (2018) investigated residents’ satisfaction with integration reform using a cross-sectional survey in three areas of China from 2014 to 2015. They found that nearly half of the respondents were dissatisfied with the integration reform, with dissatisfaction arising from perceived limitations of management system improvement, inequity reduction and actual coverage expansion.

Health insurance integration is not unique to China. There are rich and varied international experiences of integrating health insurance programmes into larger purchasing pools to establish a universal health insurance system, as in Japan (Kondo and Shigeoka, 2013), Mexico (Gakidou et al., 2006) and, much earlier, Germany (Busse, 2001). A reform implementing universal coverage in Thailand exerted positive impacts among low- and middle-income enrollees, improving the access to healthcare and financial risk protection (Limwattananon et al., 2015). Taiwan expanded the UHC through the establishment of National Health Insurance. Evidence in Chang (2012) suggests that the universal health insurance system in Taiwan contributed to a 24% reduction in the mortality rates of elderly people. The achievement of UHC in these economies faced non-trivial implementation challenges, as there were great differences in health and ability to pay between subpopulations pre-UHC. Their experiences suggest that China’s rural population could benefit disproportionately from insurance integration.

Previous studies of China’s insurance integration mainly focus on the impacts on healthcare utilization and health. Evidence about the association between the integration and SWB is scarce. This study seeks to help fill this gap. As China has the largest population covered by social health insurance in the world, and urban–rural health insurance integration reform is directed towards over two-thirds of China’s population, China’s integration reform goals, systemic strategies and evidence of impact might have important implications for other countries pursuing UHC.

Institutional background

China began to reform the medical and health system in the late 1990s to establish a government-led medical security system for all residents. By 2011, China had achieved universal health insurance coverage with the successful implementation of the three basic health insurance programmes. The first programme is the Urban Employee Basic Medical Insurance (UEBMI) programme for urban formal sector workers, built on the legacy of employment-based coverage in state-owned enterprises and organized at the municipality level; launched as UEBMI in 1998, it is financed by a payroll tax. The second programme, with more limited coverage and benefits, was the NRCMS for rural residents; launched in 2003, NRCMS built on the legacy of the original Cooperative Medical Scheme in the Mao era, with voluntary coverage subsidized by local and central government and risk pooling at the county level. The third programme, launched a few years later, is the URBMI for urban residents who are not covered by UEBMI, including elderly people, children and young students, unemployed, self-employed, temporary and part-time workers. In 2011 prior to integration reforms, more than 95% of China’s population was covered by the three basic health insurance programmes. The national coverage of basic health insurance signifies a key step towards UHC in China.

A rich literature documents the tangible benefits that China’s basic health insurance programmes provide to enrolled residents, including improved risk protection and access to care (Wagstaff et al., 2009; Eggleston, 2012; Fu et al., 2014; Liu et al., 2017; Meng et al., 2019; Yip et al., 2019), greater affordability (Fu et al., 2018; Fang et al., 2019; Ta et al., 2020) and improved health outcomes (Pan et al., 2016; Fan et al., 2019; Sun and Lyu, 2020) for all residents.

Despite supporting accessibility and affordability of basic services, this tripartite system—with disparate financing and coverage according to household registration (‘hukou’) and employment status—resulted in inequitable access to healthcare and financial protection, especially limiting the financial protection for rural residents under NRCMS (Fu et al., 2014; Yang et al., 2018) and for hundreds of millions of migrant workers (Meng et al., 2015).

In response, the State Council launched ‘Opinions of the CPC Central Committee and the State Council on Deepening the Health Care System Reform’ in 2009, which clearly stated that a unified urban–rural basic health insurance system should be gradually established to ultimately provide affordable and equitable basic health services for citizens. Some provinces and municipalities, such as Sichuan Province, Jiangsu Province and Tianjin City, have made a series of attempts to integrate NRCMS and URBMI programmes. However, the integration process was slow because of the lack of national guidelines and planning. In 2016, the State Council issued the ‘Opinions on Integrating the Basic Health Insurance System for Urban and Rural Residents’ (GF No. 2016-3) intended to ensure that both rural and urban residents would be covered by the same health insurance programme and receive equal benefit packages regardless of their place of residence. Central government required local governments to introduce a specific implementation plan by the end of December 2016, and launch the integration policy by 2020. The local governments have discretion for the timing of integration implementation.

The integration reform was nationally started by achieving integration in six key areas, including integration of health insurance coverage, fundraising policies, payment level, health insurance service list and formulary, and fund management and regulation. Specifically, the URRBMI will cover all the residents who participated in the NRCMS and URBMI before integration. Participants enjoy unified payment standards and benefits packages. After the integration, the benefit package of URRBMI was improved, especially for rural residents under the national guideline of ‘Providing for a wide benefits package, not for narrow, for high reimbursement rate, not for narrow’ (‘Jiu kuan bu jiu zai, jiu gao bu jiu di’). The URRBMI is regulated by the Ministry of Human Resources and Social Security and the local Bureau of Human Resources and Social Security for fund management and regulation. The URRBMI fund complies with a unified national fund management system.

These policy changes substantially increased the scope of benefits for over 800 million rural residents, in part because of additional government subsidies. In Beijing’s outlying rural areas, government subsidies increased by almost one-third (Table A1). The integration entailed essentially no change in coverage for the urban residents formerly covered by municipal URBMI programmes. Thus, we predict that the insurance integration likely improved the risk protection and well-being of rural and lower-income residents, with few discernible impacts on others, although we also empirically probe impacts on urban residents (e.g. any dissatisfaction if there were worsened crowding at local providers or any benefits from wider risk pooling).

Data

We analyse four waves of data—baseline (2011), 2013, 2015 and 2018—from the CHARLS, a detailed nationally representative survey of adults over the age of 45 years and their spouses; see  Appendix 1 for details about CHARLS. We also add 2020-round data for robust check. Focusing on respondents aged 40–89 years who were covered by NRCMS, URBMI or integrated insurance (URRBMI) in the pilot jurisdictions,³ and excluding missing values and three pre-2011 pilot areas (i.e. Tianjin, Chengdu and Chongqing), our analytic sample comprises 34 627 rural person-years and 14 118 urban person-years.

We collected integration information for each jurisdiction (e.g. municipality, county and prefecture) from their respective government bureau websites. The control group comprises all CHARLS respondents residing in jurisdictions that had not integrated insurance. Since survey respondents are asked to report data for the previous year for several variables—including household expenditures, total medical expenditures for hospitalization, the reimbursement rate of hospitalization and other inpatient care utilization—we measure whether the jurisdiction implemented the integration at the beginning of the previous year.⁴

The main outcome variable of interest was SWB as measured by life satisfaction. Respondents were asked, ‘Overall, how satisfied are you with your life?’ We analysed the answers in two ways. The first was to treat the response as a continuous variable, namely, ‘self-perceived life satisfaction’, with a value ranging from 1 to 5, corresponding to the five response options of ‘not at all satisfied’, ‘not very satisfied’, ‘average/general’, ‘very satisfied’ and ‘completely satisfied’, respectively. The second was to define the indicator variable ‘satisfied with life’ equal to 1 if the individual reported being completely or very satisfied with life, and zero otherwise.

We investigated seven potential mechanisms, including healthcare utilization (i.e. outpatient and inpatient care, respectively), financial risk protection (i.e. reimbursement rate, the ratio of out-of-pocket expenditures to family expenditures), selection of healthcare providers (i.e. facility ownership and level), health status (i.e. self-reported good health, mental health), health behaviours (i.e. physical activity, social activities), self-reported satisfaction with health and family economic benefits (i.e. family consumption, debts). The mental health index is constructed from the 10-item version of the Centre for Epidemiological Studies Depression Scale (CES-D). Following Andresen et al. (1994), we defined CES-D scores of 10 and above as having depressive symptoms and set a dummy variable with one as having depressive symptoms and zero otherwise. (See  Appendix 1. Data for definition details for all variables.)

We provide a brief description of life satisfaction and other main variables in Table A2. The average self-perceived life satisfaction score was 3.2 (out of a full score of 5). The proportion of individuals who were satisfied with their life was 32.3%. The average age of the respondents was ∼60 years, and 87% were married. Only 7.7% of the respondents had educational attainment of senior high school or above. Regarding working status, 43.2% of respondents were farmers, 18.4% were employed by firms or other organizations and 6.2% were self-employed. About 13.1% and 52.2% of the respondents reported poor and average childhood health status, respectively, and about 1/3 of the respondents had at least one physical disability, reflecting the older age profile of the sample. The average annual household per capita expenditure was about 11 887 RMB, a little lower than the national average of about 15 000 RMB during 2011–18.

Methods

Our analysis employed a time-varying DID (or heterogeneous-timing DID) approach to capture the effect of integration on SWB (Beck et al., 2010; Fadlon and Nielsen, 2019). Since different local governments introduced urban–rural health insurance integration reform in different years, subjects in the control group are constantly entering the treatment group every year. Following the literature (e.g. Beck et al., 2010; Fadlon and Nielsen, 2019), we adopted a time-varying DID for identification. Under the framework of two-way fixed effects (time and regional fixed effects), the general equation for time-varying DID is as follows:

$${y_{ijt}} = {\alpha _{}} + {\beta _{}}^{\ast} {\rm{polic}}{{\rm{y}}_{jt}} + {\mu _j} + {\lambda _t} + \eta^{\prime}{z_{ijt}} + {\pi _{ijt}},$$

(1)

where i, j and t are the individual, jurisdictional and time indices, respectively. Y_ijt represents life satisfaction variables. Policy_jt is a dummy variable marking the treatment status, which changes with time and region. We code it as 1 if the given jurisdiction j implements the integration policy during the research period (2010–17) and 0 otherwise.|${\mu _j}$| and |${\lambda _t}$| are jurisdiction- and year-fixed effects, respectively, and z_ijt is a vector of socio-demographic characteristics, including age, gender, marital status, educational level, working status, health status (i.e. childhood health status, physical disability), pension insurance and family economic status. |${\pi _{ijt}}$|is a random error term. The coefficient |$\beta $| is the key parameter of interest that captures the average treatment effect of integration on SWB. Linear and logit models with average marginal effects were employed for continuous and binary outcomes, respectively.

We find supportive evidence for parallel pre-reform trends in life satisfaction, based on an event study combined with a DID approach to test whether the six pre-treatment coefficients were statistically indistinguishable from zero, with four periods after the treatment year. This approach also helps in capturing the change in policy effects in the time dimension. The event study equation combined with the DID approach can be expressed as follows:

$$\begin{aligned}\mathop y\nolimits_{ijt} & = \mathop \delta \nolimits_0 + {\beta _1}{D_{jt}}( - 6) + {\beta _2}{D_{jt}}( - 5) + \ldots{\beta _{10}}{D_{jt}}( + 3) + {\beta _{11}}{D_{jt}}( + 4) \nonumber \\ & \qquad + \mathop \mu \nolimits_j + \mathop \lambda \nolimits_t + \mathop {\eta^{\prime}z}\nolimits_{ijt} + \mathop \pi \nolimits_{ijt}\end{aligned}$$

(2)

where D_jt represents the period (year) relative to the treatment year. According to our data, there are six pre-treatment periods, presented as ‘D(−1)’ to ‘D(−6)’, and four periods after the treatment year, presented as ‘D(+1)’ to ‘D(+4)’. We merge more than six periods before the treatment into the sixth relative period ‘D(−6)’, and merge more than four periods after the treatment into the fourth relative period ‘D(+4)’. We treated the sixth year of pre-treatment ‘D(−6)’ as the reference group. If parallel trends hold, the coefficients of ‘D(−1)’ to ‘D(−5)’ should be close to zero and insignificant.

We also used a placebo test and other methods to test the parallel trend assumption and performed sensitivity analyses by estimating individual fixed effects and marginal effects with an ordered probit regression, as well as examining heterogeneity among subgroups by gender, health status and economic status. We applied a time-varying DID model to explore the potential mechanisms whereby integration affects individuals’ SWB.

Results

The empirical results suggest that improved insurance raised life satisfaction among rural residents. Table 1 shows our time-varying DID estimates of the impact of the urban–rural health insurance integration on life satisfaction among rural and urban residents, respectively. We treat life satisfaction as a continuous variable in the first two columns (namely, ‘self-perceived life satisfaction’), and as a dummy variable in the last two columns (namely, ‘satisfied with life’). We conduct subsample regressions to analyse the heterogeneous impacts on life satisfaction among urban and rural residents. Columns 1 and 3 highlight that the reform significantly improved the life satisfaction of rural residents, with the probability of being very or completely satisfied with life increasing by 4.5 percentage points after integration. Results in Columns 2 and 4 indicate that there is no evidence of a significant impact of integration on life satisfaction among urban residents.

Table 1.

Effects of urban–rural health insurance integration on life satisfaction in China

	Self-perceived life satisfaction (continuous variable)		Satisfied with life (yes = 1)
	Rural	Urban	Rural	Urban
Variables	(1)	(2)	(3)	(4)
Policy	0.065***	0.000	0.046***	0.009
	(0.017)	(0.024)	(0.010)	(0.015)
Male (yes = 1)	0.045***	0.028*	−0.003	0.007
	(0.012)	(0.017)	(0.007)	(0.010)
Age	0.009***	0.007***	0.004***	0.003***
	(0.001)	(0.001)	(0.000)	(0.001)
Married (yes = 1)	0.109***	0.107***	0.024**	0.029*
	(0.019)	(0.027)	(0.010)	(0.015)
Educational level (reference group: ‘Illiterate’)
Primary school	−0.034**	−0.027	−0.047***	−0.052***
	(0.014)	(0.023)	(0.008)	(0.013)
Middle school	−0.045**	−0.048*	−0.081***	−0.088***
	(0.018)	(0.027)	(0.010)	(0.016)
High school and above	−0.081***	−0.083**	−0.103***	−0.109***
	(0.025)	(0.032)	(0.014)	(0.020)
Working status (reference group: ‘Unemployed’)
Farmers	0.060***	−0.011	0.024***	−0.006
	(0.014)	(0.022)	(0.007)	(0.013)
Retired	0.100***	0.026	0.021	−0.005
	(0.030)	(0.027)	(0.020)	(0.016)
Employed	0.091***	0.004	0.038***	−0.000
	(0.017)	(0.023)	(0.010)	(0.014)
Self-employed	0.102***	0.061**	0.036**	0.020
	(0.024)	(0.027)	(0.014)	(0.017)
Childhood health status (reference group: ‘poor’)
Average/normal	0.089***	0.098***	0.025***	0.022
	(0.017)	(0.027)	(0.009)	(0.014)
Good	0.166***	0.147***	0.079***	0.057***
	(0.019)	(0.028)	(0.010)	(0.015)
Physical disability (yes = 1)	−0.098***	−0.126***	−0.026***	−0.048***
	(0.012)	(0.019)	(0.007)	(0.011)
Pension insurance (yes = 1)	0.020*	0.032**	0.004	0.018*
	(0.012)	(0.016)	(0.007)	(0.010)
Family economic status (reference group: ‘The lowest 33%’)
The middle 33%	0.043***	0.053***	0.019***	0.017*
	(0.011)	(0.017)	(0.007)	(0.010)
The highest 33%	0.095***	0.130***	0.053***	0.071***
	(0.013)	(0.019)	(0.007)	(0.012)
Constant	2.373***	2.563***	0.012	0.085*
	(0.057)	(0.086)	(0.030)	(0.049)
Observations	31 769	12 975	31 769	12 975
The mean of the dependent variable	3.212	3.229	0.326	0.324

Note: Year and locality fixed effects are controlled for in the model. Standard errors are clustered at the individual level and are shown in parentheses.

*, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.

According to the estimated coefficients of the control variables in Table 1, married people, people having good and average childhood health status, those without disability and with better economic status are more satisfied with their lives. The younger the age, the lower the life satisfaction, perhaps because our research sample is restricted to middle-aged and elderly individuals >40 years of age. People in their forties and fifties may face multiple and greater pressures from work and family (e.g. career promotion and retirement, caring for their elderly parents, and children’s education, employment and marriage), adversely impacting their perceived well-being relative to retirees. Somewhat unexpectedly, people with more years of schooling report lower life satisfaction, perhaps reflecting that higher education may come with some negative non-pecuniary returns (Nikolaev, 2018). Veenhoven (2010) concludes that school education is the only capability that does not seem to make people happier.⁵

For rural residents, employees and self-employed people (including farmers) report much higher life satisfaction, compared with unemployed people; presumably this reflects the limited income sources of rural residents and thus that employment determines household living standards, which, in turn, affect SWB.

We use an event study combined with DID to explore the dynamic impacts of insurance integration and to test the assumption of parallel pre-trends in life satisfaction. Figure 1A for rural residents and Figure 1B for urban residents show that the coefficients of the interaction between the treatment indicator and relative year of pre-treatment are close to zero and not significant at the 5% level. Hence, there is no significant difference in life satisfaction between the treatment and control groups in the pre-treatment periods, suggesting reassuringly that the assumption of parallel trends holds.

Figure 1.

The dynamic impacts of integration policy on life satisfaction

Note: The figure plots the difference in life satisfaction between the treatment and control groups over time, along with their 95% CIs. Event study and DID approaches were utilized to explore dynamic impacts. The x-axis represents the interaction between the treatment indicator and the relative year of policy implementation. We consider an 11-year window, spanning from 6 years before integration until 4 years after integration. We treated the sixth year of pre-integration as the reference group. All estimations were adjusted for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status and local- and year-fixed effects.

The coefficients of the interaction between the treatment indicator and relative year post-treatment in Panel A indicate that the impact of the integration on improving life satisfaction among rural residents is persistent and gradually increasing, at least during the survey period. This finding is consistent with insured residents needing time to understand and benefit from insurance integration and for those benefits in turn to lead to improvements in internalized well-being. Thus, the integration impacts may have been underestimated in studies with a short follow-up period. Figure 1B confirms that integration has no significant effect on life satisfaction among urban residents, at least during the survey window.

Probing for heterogeneity, we find that insurance integration differentially benefits older, lower-income rural residents (Figure 2; Table A3). Based on the average annual household non-medical per capita expenditures, we classify the sample into three economic groups (i.e. the lowest 1/3, middle 1/3 and highest 1/3). The integration policy significantly improves life satisfaction among rural residents of lower economic means, with a 10.3 percentage point improvement in life satisfaction in the lowest tertile and a 7.5 percentage point improvement in the middle tertile. Integration has no significant effect on households in the highest tertile. These findings suggest that rural residents with lower economic status, who are more likely to be credit-constrained and thus find care less affordable, benefit more from the urban–rural health insurance integration.

Figure 2.

Insurance integration differentially benefits older, lower-income rural residents

Note: Figures based on Table A2. All models control for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status and year and local fixed effects. The dots mark the point estimates, and the lines denote 90% CIs.

In addition, older adults (aged 60–89 years) benefit more from the integration policy than residents aged 40–59 years, indicating that the integration policy improves the welfare of older adults who have greater demand for healthcare services and medical security. These positive impacts of integration on life satisfaction hold for both men and women in rural areas.

Furthermore, the reform improved the life satisfaction of rural residents with chronic disease by 2.3 percentage points (significant at the 10% level), much lower than that among rural residents without chronic disease (8.2 percentage points and significant at the 1% level). This finding probably reflects the fact that although chronic disease patients benefit from the reform, they do so relatively less than those without chronic disease given their higher frequency of outpatient visits and need for expensive longer-term medication, which continue to place a significant economic burden on them despite the reforms. The URRBMI payment for outpatient care in most areas is <1000 RMB (frequently 300 RMB). Therefore, patients requiring outpatient chronic disease care (e.g. medications for hypertension or diabetes) still face considerable out-of-pocket costs.

For urban residents, integration has no significant impact, as expected.

We test several mechanisms underlying these findings (see Table A4 and  Appendix 1, Mechanisms Analysis for more details). As shown in Figure 3, integration significantly increases outpatient care utilization of rural residents, leading to higher medical expenditures for outpatient care. The actual reimbursable rate for inpatient care increased by 8.4% (at a 1% significance level) after the integration. Thus, these improvements may provide direct benefits to rural residents. The self-assessed health status of rural residents significantly improved after integration, although with a lag. Importantly, the depressive symptoms of rural residents are significantly reduced after the integration. Additionally, rural residents are more likely to be satisfied with their health after integration. The above results support the vital mechanism of improved health status (especially reduced mental illnesses), probably reducing residents’ concerns over health and financial risk caused by illness.⁶

Figure 3.

Potential mechanisms (rural sample)

Note: Figure based on selections from Table A3. Potential mechanisms using the time-varying DID estimation. All models control for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status, year and local fixed effects. The dots mark the point estimates, and the lines denote 90% CIs.

We find that rural residents with an inpatient admission are more likely to be hospitalized in government-owned hospitals after the integration, meaning that their medical expenses are more likely to be reimbursed through subsidized health insurance. They are also more likely to go to county-/district-level hospitals for outpatient and inpatient care, instead of regional/city, provincial, ministry-affiliated or military hospitals. These findings indicate that the URRBMI programme supported rural residents’ accessing healthcare at local public healthcare providers, without exacerbating congestion at higher-level hospitals. In addition, insurance beneficiaries may have more ways to obtain health knowledge and advance health literacy, thus improving health behaviours. Intuitively, good health behaviours are positively related to residents’ well-being. Indeed, we find that the probability of walking activity and taking part in a community-related organization increased after integration.

In summary, we explore several channels through which integration might affect life satisfaction among rural residents. The findings indicate that integration significantly reduced rural residents’ mental illnesses, as well as strengthening financial risk protection, increasing some healthy behaviours, outpatient care utilization, and visits and admissions at county-/district-level public hospitals; these in turn lead to better self-reported health and satisfaction with health, all of which boost SWB. The findings are supported by several robustness checks (see  Appendix 1. Robustness Checks and Table A5).

Discussion

Health insurance expansion is regarded as an important part of achieving the goal of UHC (Lagomarsino et al., 2012). On the one hand, health insurance expansion could help enrollees gain better access to healthcare and reduce the uncertainty of future medical expenditure, all of which would promote SWB. However, it is also possible that the health insurance policy fails to meet the expectations of residents and thus lowers their SWB. In other words, whether the expansion of health insurance improves respondents’ SWB is an empirical question.

This study examines the case of China’s integration of urban–rural insurance programmes covering over one billion residents, substantially increasing the coverage for rural residents previously covered by the NRCMS programme. We find that the integration policy significantly improves the life satisfaction of rural residents, with the probability of being satisfied with life increasing by 4.6 percentage points after integration. The positive impact on life satisfaction among rural residents is persistent and gradually increases within at least 4 years. This finding is consistent with the evidence from the randomized expansion of Medicaid in Oregon as analysed by Finkelstein et al. (2012), who reveal that Medicaid enrolment is associated with about a 32% increase in self-reported overall happiness 1 year later. Indeed, that study found that self-reported health status of low-income US adults improved significantly (about two-thirds of the later value) right after lottery selection, when there was not yet any impact on utilization, suggesting that the coverage itself led to an improved sense of well-being (Finkelstein et al., 2012). Similarly for China’s rural residents, we find that expansion of coverage through subsidized urban–rural resident insurance contributed to a general improvement in the sense of well-being. Moreover, this effect did not dissipate over time, but rather strengthened as more residents became familiar with and benefitted from the insurance integration policy.

We also find that the improvement in life satisfaction is concentrated among rural residents with lower economic status, with 10.3 and 7.5 percentage point improvement in life satisfaction among the lowest and middle tertiles of household non-medical expenditures, respectively. Integration has no significant effect on the highest tertile economic group in rural areas. This finding is consistent with recent studies finding that integration reduces pro-rich inequity in outpatient benefits and significantly improves financial risk protection for low-income rural residents (Ren et al., 2022; Zhou et al., 2022). As noted, it is also consistent with the Oregon Medicaid programme evidence for low-income adults in the USA (Finkelstein et al., 2012).

However, we observed a positive but minor effect of integration on life satisfaction among people with chronic diseases. This finding indicates that the level and range of medical security after integration still need to be improved for residents with chronic diseases, although outpatient care utilization has increased after the integration due to better service packages and coverage of outpatient services. Other studies have shown that rural residents can benefit from primary care chronic disease management programmes and reduce avoidable hospitalizations (Ding et al., 2021). Similar and complementary policies should be tailored according to the characteristics and health service demands of patients with chronic diseases, assuring that they benefit from the URRBMI.

We found no evidence of any significant effect of integration on the SWB of urban residents, as expected given that their benefit package essentially did not change. Fortunately, we do not find any evidence of negative effects such as worsened crowding at urban tertiary hospitals due to increased utilization by rural residents. However, since urban residents also include vulnerable populations (e.g. low-income elderly and chronically ill), understanding why the integration policy has not delivered tangible benefits to them warrants policy attention as China continues to strengthen UHC.

Regarding potential mechanisms, we find that integration significantly reduced depressive symptoms and improved the self-assessed health status of rural residents, although with a lagged effect as more rural residents had reason to seek care or otherwise experience the benefits of the better insurance coverage. This improvement in mental health is significant, given that like many countries around the world, China is facing the challenge of increasing prevalence of mental disorders accompanied by the fast and accelerated urbanization by decade (Phillips et al., 2009; Yang et al., 2013; Chen et al., 2014). The 12-month adjusted prevalence of any mental disorder is 9.3% in China, and the lifetime incidence is 16.6% (Huang et al., 2019). Treatment rates, however, were very low, and few people received adequate treatment. Only 9.5% with 12-month depressive disorders were treated at all, and only 0.5% were treated adequately (Lu et al., 2021). Meanwhile, mental illness is a strong risk factor for suicide in China; approximately half of the suicide victims had one or more mental illness (Zhang et al., 2010).⁷ Hence, any interventions that help to improve mental health are important.

The empirical results also show that higher reimbursement rates after integration led to greater utilization, especially of outpatient care; total expenditures for outpatient services increased, although the likelihood of an outpatient visit did not increase after integration. Furthermore, rural residents were more likely to be satisfied with their health after integration. Some rural residents’ health behaviours (e.g. physical activity such as walking; social activities such as taking part in a community-related organization) also improved after integration and its associated decrease in depressive symptoms, and their willingness to choose inpatient services at local government-owned facilities (i.e. public county/district hospitals) increased. These ancillary impacts of the improved coverage for rural residents hint at gradual improvement along several dimensions, reducing inequality in healthcare access and health outcomes without negative spillovers on their urban counterparts.

This study has several limitations. First, because the CHARLS data focus on middle-aged and older adults, additional studies are required to understand the impacts of integration on younger age groups. Second, in addition to the mechanisms we have tested in this study, other mechanisms, such as stress-related factors, availability of human resources and their empathetical behaviour in hospital and other healthcare settings, might also be correlated with personal satisfaction and SWB. More research is warranted to further collect related data and test potential mechanisms. Furthermore, while our study design based on the staggered implementation of the policy and sensitivity analysis using individual fixed effects provide robust evidence, residents were not randomly assigned insurance and thus our estimates fall short of the gold-standard randomized controlled trial. Additional studies of this natural experiment impacting over one billion people would be valuable.

In summary, this study provides evidence that enhanced health insurance for low-income populations can improve SWB and reduce depressive symptoms. Our empirical analyses find that China’s urban–rural health insurance integration policy significantly improved the SWB of rural residents, with the positive impact gradually increasing in the few years after integration. Elderly and lower-income rural residents benefitted the most. The positive impact of integration on well-being appears to stem from improvement of health status (especially mental health), reduced co-payments and increased outpatient care utilization. There was no discernible impact of integration on SWB among urban residents and rural people with other chronic diseases. Further policies should be designed to deepen and expand the beneficiary population. This study contributes to a better understanding of the impacts of urban–rural health insurance integration reform in China and provides evidence-based guidance for policy improvement. As China has the largest population covered by social health insurance in the world and urban–rural health insurance integration reform impacts over two-thirds of China’s population, such evidence might have important implications for other countries pursuing UHC.

Acknowledgements

We are grateful to the National Social Science Fund of China (grant number: 17CGL051) for financial support. The authors would like to acknowledge the China Health and Retirement Longitudinal Survey (CHARLS) for providing data. The data underlying this article are available at http://charls.pku.edu.cn/. The authors are responsible for all remaining errors.

Author contributions

Q.Z. contributed to conception or design of the work, data collection, data analysis and interpretation, drafting the article and final approval of the version to be submitted; K.E. contributed to conception or design of the work, critical revision of the article and final approval of the version to be submitted; G.G.L. contributed to conception or design of the work and final approval of the version to be submitted.

Reflexivity statement

The authors include two females and one male and span multiple levels of seniority. Q.Z. mainly studied in the fields of health economics, labour economics and public health and management. She has rich experience in using quantitative research methods to study China’s health insurance system reform. K.E. specializes in research on government and market roles in the health sector and Asia health policy, especially in China, India, Japan and Korea; healthcare productivity; and the economics of the demographic transition. G.G.L. research focuses on health and development economics, China’s healthcare system reform and medical economics. All three authors have extensive experience of understanding China’s health insurance system reform.

Ethical approval.

Our data come from the CHARLS. The investigation of the survey asked about respondents’ and their spouses’ social-demographic characteristics, health status, self-reported health outcomes, healthcare utilization and health insurance status and other socio-demographic information. The CHARLS data are de-identified, publicly available data and thus exempt from Institutional Review Board (IRB) review.

Conflict of interest:

The authors have no conflicts of interest to report.

Footnotes

1.

According to data from the CHARLS, 23 counties/cities implemented the integration policy before 2016, accounting for 18% of total surveyed counties/cities.

2.

The CHARLS also collected the 2020 round of survey data, although many variables used in our method are not available for the 2020 wave (e.g. medical expenditures for outpatient care and inpatient care, chronic disease and disability status); hence, we report our main results using 2011–18 data, and provide evidence adding the latest 2020 data as a robust check. Reassuringly, the latter results show that our main findings are robust.

3.

We exclude the sample of participating in NRCMS, URBMI and URRMI in registered residence and other provinces, accounts for 4% (2208 observations) of total sample.

4.

For example, the variable of the integration treatment (‘Policy_jt’ in the equation) is defined as 1 if the jurisdiction j implemented the integration policy in 2017, and 0 if the jurisdiction implemented the scheme after 2017.

5.

The relationship between education and happiness among the Chinese population deserves further research.

6.

The integration may reduce participants’ worries about future medical expenses, and the generous policy may also promote their utilization of mental health services, all of which contribute to higher life satisfaction. However, it should be noted that depressive symptoms and life satisfaction are highly correlated. This study has not yet extracted the causal effect of depressive symptoms on life satisfaction, which requires further research.

7.

Mental disorders also have significant effects on healthcare costs, with about 7% of total personal medical expenditure attributable to depression and depressive symptoms (Hsieh and Qin, 2018).

8.

Rural residents are defined as those holding agricultural household registration (agricultural hukou) and living in rural areas at the time of the survey. Urban residents are defined as those holding non-agricultural household registration (non-agricultural hukou) and living in urban areas.

9.

CHARLS asked the respondents about physical activities that they did for at least 10 min at a time in a usual week. Vigorous activities make one breathe much harder than normal and may include heavy lifting, digging, ploughing, aerobics, fast bicycling and cycling with a heavy load. Moderate physical activities make one breathe somewhat harder than normal and may include carrying light loads, bicycling at a regular pace or mopping the floor. The survey also asked the respondents about the time they spend walking in a usual week. This time includes walking at work and at home, walking to travel from place to place and any other walking that might be done solely for recreation, sport, exercise or leisure.

A. Data

The CHARLS is a nationally representative survey. The target population of CHARLS comprises adults over the age of 45 years and their spouses in both urban and rural households in China. The survey was based on a design similar to that of the Health and Retirement Study in the USA, the English Longitudinal Study of Ageing in England and the Survey of Health, Ageing and Retirement in Europe in 11 Continental European countries. In the CHARLS baseline survey of 2011, a multi-stage cluster sampling design was employed with three levels of sampling frames, including county/city/prefecture, village/community and households, and a total sample size of about 10 000 households was randomly selected from 150 jurisdictions (e.g. municipality, county, prefecture) in 30 provincial-level administrative units in mainland China (Tibet is excluded from the survey) with the probability proportional to size method (Zhao et al., 2013; 2020). We used baseline survey data from 2011 and follow-up surveys in 2013, 2015 and 2018, and focused on rural and urban respondents aged 40–89 years.⁸

Before 2011, three jurisdictions (i.e. Tianjin, Chengdu and Chongqing) voluntarily administered the urban–rural residents’ health insurance integration pilot. As no specific guidance document is available at the national level, the corresponding integration policy is localized and distinguished from each other in each of the pilot areas. We excluded samples from these three cities from our study. Thus, all jurisdictions included in our analytic sample had not yet implemented the URRBMI programme in the baseline year of this study. Finally, we excluded observations with missing values and outliers, leaving 35 974 rural person-years and 14 887 urban person-years for this study.

For potential mechanisms, we mainly investigated seven areas, including healthcare utilization, financial risk protection, selection of healthcare providers, health status, whether they were satisfied with their health, physical activities and family economic benefits. First, healthcare utilization is measured by four variables: outpatient visits in the past month (yes = 1, no = 0), total expenditures for outpatient care in the past month, inpatient admission in the past year (yes = 1, no = 0) and total expenditure for inpatient care. Second, financial risk protection is measured by the ratio of out-of-pocket expenditures to family expenditures and the reimbursement ratio at the municipal level, which refers to the ratio of reimbursement amount to total healthcare expenditures of the sample population in the same municipality. The reimbursement ratio was also calculated separately for inpatient and outpatient care.

Third, selection of healthcare providers included whether a patient chose a public or private facility for outpatient and inpatient care, and the facility levels for outpatient and inpatient care (i.e. county/district, regional/city, provincial/affiliated with a ministry or military). Fourth, health status was measured using self-reported health status and the survey questions regarding mental health. We analysed self-reported health as a dummy variable with 1 indicating good and very good self-assessed health status and zero otherwise. The mental health index is constructed from the 10-item version of the CES-D, which is widely used as a pre-clinical measure of depression. The CES-D score is a continuous variable ranging from 0 to 30, with higher scores indicating more severe depressive symptoms. As suggested by Andresen et al., (1994), we defined CES-D scores of 10 and above as having depressive symptoms and set a dummy variable with 1 as having depressive symptoms and zero otherwise.

Fifth, to measure satisfaction with health, the CHARLS respondents were asked the following question in the 2015 and 2018 rounds of the survey, ‘How satisfied are you with your health?’. Response options included ‘completely satisfied’, ‘very satisfied’, ‘general’, ‘not very satisfied’ and ‘not at all satisfied’. We used a dummy variable, namely, ‘satisfied with health’, with 1 denoting completely or very satisfied and zero otherwise.

Sixth, health behaviours were measured using physical and social activities. Physical activities (in a usual week) included vigorous activities, moderate physical activities and walking.⁹ We suppose that some physical activity would be good for health. Social activities (in the past month) include whether they interacted with friends, and whether they took part in a community-related organization.

Seventh, in terms of family economic benefits, we use yearly per capita expenditures, and debts owned by individuals and working units. These two indicators were logarithmically transformed.

B. Results

Table A2 provides descriptive statistics for our analytic sample from CHARLS, and Table A3 reports the subgroup analysis results.

C. Distributional effects on life satisfaction for rural residents

In addition to the main results reported in the text, we further check the distributional effects on the life satisfaction of rural residents because, according to the findings in Table 1 and Figure 2, the rural residents are significantly influenced by the integration policy. We used the continuous variable of life satisfaction as an independent variable in the quantile regressions. Figure A1 shows that individuals with different distributions of life satisfaction are positively affected by the integration policy, and individuals at the right of the distribution report a greater increase in life satisfaction after integration but with large variation.

D. Mechanisms analysis

Complementing the summary in the main text, this section of the Appendix amplifies our discussion of the possible mechanisms underlying the positive impact of integration on life satisfaction among rural residents.

Most health economics literature concludes that insured people use more healthcare when holding health insurance, given the discounted price at point of use (Zeckhauser, 1970; Finkelstein et al., 2012). Some studies provide evidence that insured people have better financial risk protection with lower out-of-pocket medical expenditures and medical debt, which may also positively affect their health status by reducing worry over the financial risk caused by illness (Finkelstein et al., 2012; Zhou et al., 2022). Some studies also find that insurance affects enrollees’ health behaviours, although both positive and negative impacts may occur and findings are somewhat mixed (e.g. Newhouse, 1993; Zweifel and Manning, 2000; Zhou et al., 2019). Furthermore, health insurance might reduce the financial need for coping mechanisms (such as borrowing against assets) during other stresses on family income, yielding consumption-smoothing benefits. The above impacts of insurance coverage may ultimately affect individuals’ SWB. Accordingly, we examine the potential mechanisms underlying integration’s impact on well-being from the perspective of healthcare utilization, financial risk protection, selection of healthcare providers, health, health behaviours (physical activities) and family economic benefits (family consumption and debts) in Table A4.

After the integration, the URRMI programme provides a relatively more comprehensive health service benefit package and lower patient cost-sharing compared with NRCMS, which may improve access to healthcare utilization and financial risk protection for rural residents previously under NRCMS. Our empirical analyses support this hypothesis. Panels A, B and C in Table A4 present the time-varying DID estimation results for outpatient and inpatient care utilization and financial risk protection. The findings indicate that integration significantly increases outpatient care utilization, with higher medical expenditures for outpatient care. For inpatient care, the actual reimbursable rate for rural residents (i.e. the share of total healthcare expenditures covered by insurance and reimbursed to the patient) increased by 7.8% after the integration, a magnitude that is both statistically significant (at the 1% level) and economically meaningful. Thus, these improvements may provide direct benefits to rural residents.

The findings in Panel D, E and F indicate that among rural residents that were hospitalized, the facility of choice is more likely to be a public hospital after the integration. They are also more likely to go to county-/district-level hospitals for outpatient and inpatient care, instead of regional/city, provincial/affiliated to a ministry and military hospitals. These findings indicate that URRBMI programme has more attraction to rural residents for inpatient care at local public healthcare providers, and that they may use county hospitals more often than township-level facilities or higher-level urban hospitals. This result also indicates that the integration did not bring the problem of more congestion in higher-level hospitals to cause unnecessary medical expenditures and worsened crowding of local providers.

Good health increases individuals’ utility (i.e. happiness). The findings in Table A4 Panel G validate that the self-assessed health status of rural residents significantly improved after integration, although with lagged health impacts. The improved insurance coverage after integration significantly reduced the depressive symptoms of rural residents. Additionally, we found that rural residents are more likely to be satisfied with their health after integration (Panel H), which is consistent with the results of better mental health after integration.

The results in Panels I and J show that integration significantly increases the probability of walking activity and taking part in a community-related organization for rural enrollees, providing evidence of the mechanisms of lifestyle and health behaviour. For family economic benefits, Panel K shows that integration has significant effect on reducing family debts owning to individuals and working unit.

The above results suggest that improved well-being arose in part through the mechanism of better mental health, probably because insurance can reduce residents’ worry about how to cope with catastrophic medical expenditures and other adverse economic impacts caused by illness. The results of the subgroup analysis of the integration impacts among people with different health statuses (Figure 3) also support this mechanism.

E. Robustness checks

The estimates of the dynamic impact of the integration policy on satisfaction in Figure 2 provide evidence of the validity of the parallel trends assumption of the DID estimation. We also use the interaction between the treatment indicator and the year dummy to check the assumption of parallel trends in the pre-treatment periods. If the assumption of parallel trends holds, the estimates of the interactions should be insignificant. The results in Panel A of Table A5 do not reject the null hypothesis of parallel trends.

Moreover, we used a placebo test of the DID model by falsely setting the integration treatment time 2 years ahead to check the soundness of our findings. Panel B shows the outcome of the placebo test, where the coefficient of the interaction term ‘(Treat *Post)_false’ is close to zero and not significant at the 10% level, suggesting that our findings about the impact of the integration on life satisfaction are robust.

We also apply a robustness check with individual fixed effects, given the longitudinal structure of CHARLS, and the results of Panel C are consistent with the time-varying DID estimates. We also check whether the main results are robust to using an ordered probit model; the results are robust, as shown in Panel D.

Considering that there might be differences in the integration impacts before and after 2016, we conduct a DID design on data before 2016 (2011, 2013 and 2015) and after (2015 and 2018), respectively. Panels E and F signify that the integration impacts on rural residents’ SWB before and after 2016 are both significant at the 1% level and have no significant difference.

The CHARLS also collected 2020 round survey data, although many variables used in our method were either not collected or not released for the 2020 wave, such as medical expenditures for outpatient care and inpatient care, access to public or private facility, facility level for outpatient care and inpatient care, chronic disease and disability; all of these variables are critical for our analyses either as main control variables or as variables capturing potential mechanisms of integration impact. Hence, we report our main results using 2011–18 data, and provide evidence adding the latest 2020 data as a robust check. The results in Panel G show that our findings are robust if the post pandemic data are taken into consideration.

Table A1.

Basic health insurance before and after the integration of urban–rural health insurance in Beijing

	NRCMS (in year 2017)	URBMI (in year 2017)	URRBMI (in year 2018)
Financing	Personal contribution: 160 yuan; Government subsidy: 1040 yuan	Personal contribution for elderly people, children and unemployed people: 360, 160 and 660 yuan, respectively; Government subsidy: 1000 yuan	Personal contribution for elderly people, children and unemployed people: 180, 180 and 300 yuan, respectively; Government subsidy: 1430 yuan
Fund revenue	2.58 billion yuan	2.56 billion yuan	8.05 billion yuan
Fund expenditure	2.88 billion yuan	2.21 billion yuan	7.92 billion yuan
Number of insured	1.869 million	2.022 million	3.908 million
Number of users of the insurance	24.21 million person-time		26.37 million person-time

Note: Data are from the website of the Beijing Human Resources and Social Security Bureau, and Reports on the Development of Social Insurance in Beijing in 2017 and 2018. One USD was equivalent to about 6.4–6.9 yuan in year 2017 and 2018. Beijing launched its integration policy on 1 January 2018. After the integration, the financing of benefits (i.e. the premium level) increased significantly for both rural and urban residents, mainly due to the government subsidy increasing by nearly 30%. The number of URRBMI insured in 2018 was approximately 3.908 million, which was slightly higher than that before the integration. The total number of visits covered by URRBMI was 26.37 million cases, which increased by 8.9% compared with the previous year before the integration. The revenue for the health insurance fund was 8.05 billion and the expenditure was 7.92 billion, both of which were significantly higher than the level before integration. The above statistics imply that the financing level, number of beneficiaries, reimbursement level and funding management of the URRBMI were significantly improved after integration in Beijing, so that rural residents in Beijing’s outlying rural areas enjoyed benefits as comprehensive as those enjoyed by Beijing’s urban residents not covered by UEBMI.

Figure A1.

Distributional effects on the life satisfaction of rural residents

Note: The figure plots the DIDs estimates at each quantile of the distribution of life satisfaction, along with their 95% CIs. All estimations were adjusted for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status and local- and year-fixed effects.

Table A2.

Descriptive statistics

	Full sample			Treatment group		Control group
Variable	Observations	Mean	SD	Mean	SD	Mean	SD
Self-perceived life satisfaction	44 744	3.22	0.79	3.22	0.78	3.22	0.80
Not at all satisfied		0.03	0.16	0.03	0.16	0.03	0.17
Not very satisfied		0.10	0.30	0.10	0.30	0.10	0.30
Average/general		0.55	0.50	0.56	0.50	0.54	0.50
Very satisfied		0.28	0.45	0.28	0.45	0.29	0.45
Completely satisfied		0.04	0.21	0.04	0.20	0.05	0.21
Satisfied with life (yes = 1)	44 744	0.33	0.47	0.32	0.47	0.33	0.47
Male (yes = 1)	48 745	0.46	0.50	0.46	0.50	0.46	0.50
Age	48 745	60.13	9.89	59.97	9.80	60.36	10.02
Married (yes = 1)	48 745	0.87	0.34	0.87	0.34	0.86	0.34
Educational level
Illiterate	48 745	0.28	0.45	0.27	0.45	0.30	0.46
Primary school	48 745	0.44	0.50	0.45	0.50	0.43	0.50
Middle school	48 745	0.20	0.40	0.20	0.40	0.19	0.40
Senior high school and above	48 745	0.07	0.26	0.07	0.26	0.07	0.26
Working status:
Unemployed	48 745	0.26	0.44	0.26	0.44	0.26	0.44
Farmers	48 745	0.45	0.50	0.44	0.50	0.47	0.50
Retired	48 745	0.05	0.21	0.05	0.22	0.04	0.20
Employed	48 745	0.18	0.38	0.18	0.38	0.17	0.37
Self-employed	48 745	0.06	0.24	0.06	0.24	0.06	0.23
Childhood health status:
Poor	48 745	0.13	0.34	0.14	0.34	0.12	0.33
Average/general	48 745	0.52	0.50	0.53	0.50	0.51	0.50
Good	48 745	0.35	0.48	0.33	0.47	0.37	0.48
Physical disability (yes = 1)	48 745	0.32	0.47	0.32	0.47	0.32	0.47
Pension insurance (yes = 1)	48 745	0.70	0.46	0.70	0.46	0.70	0.46
Family annual non-medical per capita expenditures (yuan)	48 745	10 068	23 221	10 056	22 688	10 085	23 976
Self-reported good health (yes = 1)	47 625	0.23	0.42	0.23	0.42	0.24	0.43
CES-D scores	48 745	7.97	6.47	7.96	6.47	7.99	6.47
Outpatient visit (yes = 1)	48 662	0.20	0.40	0.20	0.40	0.19	0.39
Total medical expenditures for outpatient care	48 425	166	4090	150	1714	190	6067
Hospital admission (yes = 1)	48 727	0.13	0.34	0.13	0.34	0.13	0.34
Total medical expenditures for hospitalization	48 411	1106	7162	1147	7941	1045	5845
Reimbursement rate for outpatient care	48 233	0.13	0.12	0.14	0.11	0.12	0.12
Reimbursement rate of hospitalization	48 323	0.37	0.14	0.37	0.14	0.38	0.15
Ratio of out-of-pocket expenditures to family expenditures	45 049	0.05	0.14	0.05	0.14	0.05	0.14
Access to public facility for outpatient care (yes = 1)	8645	0.72	0.45	0.73	0.45	0.71	0.45
Utilized a public facility for hospitalization (yes = 1)	5320	0.92	0.27	0.92	0.28	0.93	0.26
Facility levels for outpatient care
County/district	3253	0.81	0.39	0.81	0.39	0.82	0.38
Regional/city	3253	0.14	0.35	0.15	0.36	0.13	0.34
Provincial/affiliated to a ministry	3253	0.04	0.19	0.04	0.19	0.04	0.19
Military	3253	0.01	0.09	0.01	0.07	0.01	0.11
Facility levels for inpatient care
County/district	4074	0.79	0.41	0.77	0.42	0.81	0.39
Regional/city	4074	0.16	0.36	0.17	0.38	0.14	0.35
Provincial/affiliated to a ministry	4074	0.05	0.21	0.05	0.22	0.04	0.19
Military	4074	0.01	0.10	0.01	0.09	0.01	0.12
Satisfied with health (yes = 1)	25 774	0.26	0.44	0.26	0.44	0.27	0.44
Physical activity:
Vigorous activity	28 345	0.37	0.48	0.37	0.48	0.37	0.48
Moderate activity	28 333	0.53	0.50	0.54	0.50	0.52	0.50
Walking activity	28 321	0.80	0.40	0.80	0.40	0.81	0.40
Social activities
Interacted with friends	39 475	0.42	0.49	0.42	0.49	0.42	0.49
Took part in a community-related organization	30 990	0.26	0.44	0.01	0.10	0.01	0.10
Yearly per capita expenditures (yuan)	48 745	11 816	24 358	11 851	23 957	11 767	24 931
Debts owning to individuals and working unit (excluding mortgage loans)	47 649	8161	39 555	7731	39 964	8788	38 944

Note: Mean for continuous variables and percentage for discrete variables. CHARLS added the questionnaire on satisfaction with health in the 2015 and 2018 surveys. The treatment group comprises all respondents residing in the jurisdictions where urban–rural health insurance had been integrated before 2017. The control group comprises all respondents residing jurisdictions where the integration reform has not yet been implemented by 2017. ‘Illiterates’ are those who had no formal education. ‘Primary school’ includes three categories: primary school, ‘sishu’/home school and not finishing primary school but were capable of reading and/or writing.

Table A3.

Subgroup analysis

	Rural	Observations	Urban	Observations
Panel A: full sample	0.046***	31 769	0.009	12 975
	(0.010)		(0.015)
Panel B: by family economic status
The lowest 33%	0.103***	10 060	0.039	4186
	(0.020)		(0.031)
The middle 33%	0.075***	10 780	0.005	4383
	(0.018)		(0.026)
The highest 33%	−0.012	10 929	−0.025	4402
	(0.017)		(0.026)
Panel C: by age
40–59	0.025*	15 851	0.017	6829
	(0.014)		(0.020)
60–89	0.064***	15 918	0.003	6144
	(0.014)		(0.023)
Panel D: by gender
Male	0.048***	14 842	−0.012	5623
	(0.014)		(0.023)
Female	0.045***	16 927	0.023	7351
	(0.013)		(0.020)
Panel E: by chronic disease
Chronic disease (yes = 1)	0.023*	16 004	0.001	6725
	(0.013)		(0.021)
Chronic disease (yes = 0)	0.082***	15 759	0.040	6245
	(0.018)		(0.025)
The mean of the dependent variable	0.326		0.325

Note: All models control for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status and year and local fixed effects.

***, ** and * denote statistical significance at 1%, 5% and 10% level, respectively.

Table A4.

Potential mechanisms (rural sample)

Dependent variables	Rural	Observations
Panel A: outpatient care utilization in the past month
Outpatient visit (yes = 1)	0.001	34 563
	(0.008)
Log(total medical expenditures for outpatient care)	0.314**	34 391
	(0.137)
Panel B: inpatient care utilization in the past year
Hospital admission (yes = 1)	0.008	34 613
	(0.007)
Log(total medical expenditures for hospitalization)	−0.189*	34 396
	(0.099)
Panel C: financial risk protection
Ratio of out-of-pocket expenditures to family expenditures	0.002	32 031
	(0.003)
Reimbursement rate for outpatient care (at the municipal level)	0.014	317
	(0.020)
Reimbursement rate of hospitalization (at the municipal level)	0.084***	320
	(0.019)
Panel D: access to public or private facility
Access to public facility for outpatient care (yes = 1)	−0.034	6148
	(0.023)
Access to public facility for hospitalization (yes = 1)	0.041**	3686
	(0.018)
Panel E: facility level for outpatient care		2170
County/district	0.052*
	(0.029)
Regional/city	−0.032*
	(0.018)
Provincial/affiliated to a ministry	−0.013*
	(0.007)
Military	−0.004*
	(0.002)
Panel F: facility level for hospitalization		2728
County/district	0.074***
	(0.028)
Regional/city	−0.044***
	(0.017)
Provincial/affiliated to a ministry	−0.022***
	(0.008)
Military	−0.007**
	(0.003)
Panel G: health status
Self-reported good health (lag-1)	0.009	33 875
	(0.008)
Self-reported good health (lag-2)	0.028***	34 532
	(0.010)
CES-D scores	−0.395***	34 627
	(0.118)
CES-D ≥ 10 (yes = 1)	−0.032***	34 627
	(0.009)
Panel H: satisfaction with health (yes = 1)	0.029***	20 077
	(0.011)
Panel I: physical activity
Vigorous activity (yes = 1)	−0.001	20 720
	(0.012)
Moderate activity (yes = 1)	−0.025**	20 101
	(0.013)
Walking activity (yes = 1)	0.019*	20 088
	(0.011)
Panel J: social activities
Interacted with friends	0.007	28 421
	(0.010)
Took part in a community-related organization	0.003*	34 627
	(0.001)
Panel K: family economic benefits
Log(yearly per capita expenditure)	0.030	35 974
	(0.024)
Log(debts owning to individuals and working unit) (excluding mortgage loans)	−0.137*	33 878
	(0.076)
The mean of the dependent variable	0.326

Note: All models control for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status, year and local fixed effects. We use method of Heckman selection model for estimation in Panels 1 and 2.

***, ** and *denote statistical significance at 1%, 5% and 10% level, respectively.

Table A5.

Robustness checks

	Rural	Observations	Urban	Observations
Panel A: parallel trend test using interactions		25 654		9989
Treat* 2013	0.016		0.034
	(0.014)		(0.023)
Treat* 2015	−0.014		0.010
	(0.014)		(0.024)
Panel B: placebo test		31 172		12 374
(Treat *Post)_false	0.006		0.003
	(0.010)		(0.016)
Panel C: individual fixed effects	0.049***	31 769	0.005	12 975
	(0.010)		(0.015)
Panel D: average marginal effects of ordered probit regression		31 769		12 975
Not at all satisfied	−0.006***		−0.000
	(0.002)		(0.002)
Not very satisfied	−0.013***		−0.001
	(0.003)		(0.005)
General	−0.013***		−0.001
	(0.003)		(0.006)
Very satisfied	0.023***		0.001
	(0.006)		(0.009)
Completely satisfied	0.008***		0.001
	(0.002)		(0.003)
Panel E: using three-round data in 2011, 2013 and 2015	0.068***	22 609	−0.025	9091
	(0.023)		(0.029)
Panel F: using 2015 and 2018 round survey data	0.058***	16 922	0.026	6819
	(0.013)		(0.021)
Panel G: using five-round data in 2011–20	0.037***	39 477	0.010	16 022
	(0.009)		(0.013)

Note: All models control for gender, age, marital status, educational level, working status, childhood health status, physical disability, pension insurance, family economic status and year and local fixed effects. Since disability information was not collected in 2020 data, we use this information in 2018 data instead.

***, ** and * denote statistical significance at 1%, 5% and 10% level, respectively.