Author's Accepted Manuscript

A multilevel analysis of the role of the familyand the state in self-rated health of elderlyChinese

Zhixin Feng, Wenfei Winnie Wang, KelvynJones

PII: S1353-8292(13)00088-9DOI: http://dx.doi.org/10.1016/j.healthplace.2013.07.001Reference: JHAP1334

To appear in: Health & Place

Received date: 25 November 2012Revised date: 28 February 2013Accepted date: 3 July 2013

Cite this article as: Zhixin Feng, Wenfei Winnie Wang, Kelvyn Jones, Amultilevel analysis of the role of the family and the state in self-rated health ofelderly Chinese, Health & Place, http://dx.doi.org/10.1016/j.healthplace.2013.07.001

This is a PDF file of an unedited manuscript that has been accepted forpublication. As a service to our customers we are providing this early version ofthe manuscript. The manuscript will undergo copyediting, typesetting, andreview of the resulting galley proof before it is published in its final citable form.Please note that during the production process errors may be discovered whichcould affect the content, and all legal disclaimers that apply to the journalpertain.

www.elsevier.com/locate/healthplace

1��

A multilevel analysis of the role of the family and the state in self-rated health of elderly Chinese

First Author: Zhixin Feng

First Author’s Institution: University of Bristol, UK

Second Author: Wenfei Winnie Wang

Second Author’s Institution: University of Bristol, UK

Third Author: Kelvyn Jones

Third Author’s Institution: University of Bristol, UK

Corresponding Author: Zhixin Feng

Corresponding Author’s Institution: University of Bristol

Address: School of Geographical Sciences

University of Bristol

University Road

Bristol

UK

BS8 1SS

Email address: frankfengs@gmail.com

Abstract:

This study examines the geographical variations of self-rated health of the elderly based on the

2008 Chinese Longitudinal Healthy Longevity Survey. Multilevel logistic models are employed

to estimate how individual, family, and institutional factors affect the health of the elderly at both

individual and province levels. Results show that while individual characteristics help to explain

2��

self-rated health, the family remains an important determinant. Those with nobody to care for

them, those in poverty and those who have to rely on medical insurance report the worst health.

The role of the state is relatively limited in contributing to the health of the elderly. There are

substantial between province differences.

Key words: China, health, family, social security, multilevel modelling

Words (including abstract, references and tables): 6,283

Introduction

China has the largest and most rapidly growing elderly population in the world. In 2009, there

were approximately 110 million aged over 65, some 8.3% of the total (NBS, 2010). This is

forecast to grow to 16% in 2030 and to 23% in 2050 (United Nations, 2001). Traditionally, the

family has played a vital role in caring for the elderly, providing emotional, physical and

financial support (Zhang and Liu, 2007). However, the family’s capability for caring has been

eroded due to reduced family size caused by urbanization and lower fertility rates, and increased

mobility of young adults especially from the rural areas to cities. Consequently there are growing

geographical differences in such familial support. To address this demographic challenge, China

has been engaging in social security reforms, including pension, insurance, financial and medical

subsidies to provide institutional support for the elderly (Wang, 2006). While China has achieved

remarkable recent economic success, its social security provision for the elderly remains

extremely underfunded, underdeveloped and geographically unevenly distributed (Liu and

Zhang, 2004; Yip and Hsiao, 2009). This study focuses on the role of the family and the State in

3��

the health of the elderly to shed light on China’s demographic processes and to provide

implications for policy makers.

Support for the elderly in the Chinese context

Family Support

Filial piety is a notable Chinese cultural trait and has long been considered a major social force

holding together China’s family care system for the elderly (Zhan et al., 2008). Filial piety and

dutiful repayment is normally expressed in co-residence; the elderly living with their married

adult children and receiving emotional, physical and material help from family members

(Zimmer, 2005). Indeed, around 50.3% of elderly are living with their family members in 2010

(National Committee on Ageing, 2012).

Previous studies have illustrated the multifaceted relationships between living arrangements and

the health outcomes (e.g. Li et al., 2009; Sereny, 2011). There is some debate, however, about

the benefits of such arrangements. Although co-residence may provide the elderly with a sense

of pride as well as support to maintain health, it can also lead to dependence and acceleration of

loss of physical ability (Li et al., 2009). Some studies find that the elderly living alone are less

limited functionally because they have to undertake their own daily activities (e.g. Gu & Xu,

2007). A similar conclusion is drawn by Beydoun and Popkin (2005) who found that elderly

living independently were healthier than those living with non-spousal family members.

4��

However, other studies conclude that the family provides vital social support and interaction

resulting in the elderly living with their family exhibiting better health (Li et al., 2009).

Formal and professional personal social services for the elderly are extremely underdeveloped

in China and the primary financial responsibility for elderly care, irrespective of living

arrangements, traditionally falls on the family (Zhang and Liu, 2007). Recent studies have

reported that some 55% of the elderly in urban and 48% in rural areas want improved social

services (National Committee on Ageing, 2012). It is known that elderly living in households

with higher income have better health including lower risk of dysfunctional status, better

emotional health, and lower risk of mortality (Beydoun and Popkin, 2005).

Both traditional family living arrangements and financial support have been weakened in recent

decades due to reduced family size and increased mobility of young adults. These changes have

an important geographical dimension as there has been considerable urbanization with young

adults moving to urban areas without their elderly relatives (Kim, 2007). Moreover, the rural

elderly are usually not covered by state pension and health care plans. Consequently, the rural

elderly are more dependent on support from their families than the urban elderly when they lose

the physical ability to work (Cai et al., 2012). Indeed, some 87% of rural elderly above 80 rely

on family support compared to 29% of those aged between 60 and 64 (Cai et al., 2012).

.

State Support

5��

Institutional support for the elderly is provided in the form of social security which includes

pension, health insurance and social services. The provision varies greatly between places and

especially between rural and urban areas. In rural areas, the collective-based social security

collapsed in 1978 and was replaced by the Household Responsibility System whereby farming

families are responsible for their own welfare protection including old-age support (Zhang,

2011). The state only provides the “five guarantees” (food, clothing, medical care, fuel and

burial) for the elderly who are childless, disabled and without any support (Smart and Smart,

2001). A basic rural old-age insurance program was launched in 1992 with funding sources

primarily coming from individuals supplemented with collective-sponsored contributions and

governmental support. This scheme failed due to poor management, inconsistent policies, and

suspicions related to the financial sustainability of the promised returns. Only 1.5% of rural

elderly received social old age insurance in 2000 and only 11% participated in 2004 (Wang,

2006). A new rural old-age insurance program was introduced in 2009 targeted at covering all

rural areas by 2012. It has achieved a higher participation rate since its implementation, but its

impact on the elderly population is largely un-researched.

In urban areas prior to 1978, state-owned enterprises and the public sector provided social

welfare including retirement pension and medical care (Lin, 1992). After 1978, a multi-pillared

system, combining a basic social pillar with supplemental enterprise-sponsored pensions and

individual savings was introduced. In this system, pension financing is shared by the

government, enterprises and individuals. In 1997, a basic unified old-age insurance system using

a social-pool-plus-personal-accounts scheme was implemented and extended in 1999 to cover all

urban employees (Government White Papers, 2004). By the end of 2005, 175 million people,

6��

including 43.67 million retirees participated in the basic old-age insurance scheme (Government

White Papers, 2006). Even though the urban elderly enjoy better social security coverage, 40%

of non-working urban elderly still relied on family support in 2000 (Wang, 2006).

In terms of state health care provision, most of the urban residents continue to be covered by

medical insurance systems funded by the government and their work units after the economic

reform. In contrast, the old Cooperative Medical System which covered 90 percent of the rural

residents became insolvent after 1978 leaving the majority of the rural residents uninsured until

2003 (Yip and Hsiao, 2009). Since then the government established a new Rural Cooperative

Medical System and increased its financial support. The coverage for the rural elderly increased

from 14% in 2003 to 94% in 2008 according to China’s 2008 National Health Services Survey.

However, the increased coverage has not led to better access to formal medical care or improved

health statuses of rural residents due to high deductibles and limited budget at the county level

(Lei and Lin, 2009). The coverage for urban elderly also increased from 66% in 2003 to 86% in

2008. Moreover, the urban elderly have better access to higher quality medical services as health

spending per capita is increasingly skewed towards urban areas (Brixi et al., 2010). Since the

public resources disproportionately benefit higher-income households, people from poor

households, especially those in rural areas, face financial obstacles in accessing cheap effective

care (Liu et al., 2008).

Professional care is extremely limited and only 1.5% of the elderly stayed at nursing homes in

2000 (Leung, 1997). In recent years, this provision has been extended to people who could afford

7��

to pay. Community-based social services for single elderly in cities have also been promoted

actively by the government (Chu et al., 2008). However, the provision of the services is highly

localized and varies greatly between places.

Geographical Variations in Elderly Support and Health

A growing body of research contends that where people live makes a difference to health

outcomes (Jones and Moon, 1993; Mohan et al., 2005; Skapinakis et al., 2005). Health outcomes

are not only affected by individual characteristics but also by the surrounding environment where

individuals live and work (Duncan et al., 1998; Jones et al., 2000).

China is highly spatially differentiated in terms of its physical and cultural landscapes as well as

economic development. The quality and availability of the health resources vary greatly between

coastal and inland areas as a result of their different levels of economic development (Liu et al.,

2008; Brixi et al., 2010). Even though the main institutions of social security provision are

ostensibly implemented through central government, large variations exist in how they are

performed locally. Indeed, since the mid-2000s, each province implements its own social

security policy. Geographical variations also exist in family support with a U-shaped

relationship; the weakest family support in mid-developed urban regions and the highest in the

most and least developed provinces (Du and Wu, 1998; Lin, 2002).

8��

There have been few national scale studies of area variations on the health of the elderly in

China. Yin and Lu (2007) examined individual and community factors associated with activities

of daily living among the oldest-old and found unexplained geographical variations. The uneven

economic development, varied social security provision as well as different family support at

regional levels may all contribute to the geographical variations on the health outcomes of the

elderly. The objective of this study is to characterise the nature of province level differences in

the health of the elderly and to examine how family support and social security provision

contribute to these differences. We do not undertake an aggregate analysis at the province level

but a multilevel analysis where individual and province differences are considered

simultaneously.

Data and methodology

The empirical analysis of this study mainly uses the 2008 wave of the Chinese Longitudinal

Healthy Longevity Survey (CLHLS). The survey was conducted in a randomly selected half of

the total number of counties and cities in 23 provinces1. It includes the oldest-old (aged 80 and

older) as well as younger elders (aged 60-79). The survey areas covered 1.1 billion people, 85

percent of the total population in China. The 2008 wave has the largest dataset among all waves

and includes 12,591 interviewees with complete data. The dataset has a natural hierarchal

structure with individuals nested within provinces.

������������������������������������������������������������1 Eight provinces in northwest China are excluded from the study due to lack of reliable information on self-reported age; they are: Xinjiang, Inner Mongolia, Gansu, Qinghai, Ningxia, Tibet, Yunnan and Guizhou.

9��

The dependent variable is Self-Rated Health (SRH); previous research has indicated that this is

a sensitive and reliable indicator of an individual’s current health status (Wu and Schimmele,

2006). Although the focus of this research is on the effects of the state and family support, it is

important to control for other relevant individual factors but not to over control so that the main

effects are underestimated. Consequently, age, gender and educational status are included as

individual variables. The remaining variables are grouped into three categories: care support,

financial support and medical support; they are chosen to represent the characteristics of

contemporary China as discussed above. Details and summary statistics are given in Table 1

where the importance of the support of the family can clearly be seen. Some 49% live with their

family (not including living with their spouse) and the main source of finance is the family for

two thirds of the respondents. At the province level we include total health institutional beds per

1000 population to reflect health resources; this includes both hospital and health centre beds.

(Table 1 about here)

Methods

Given the ordinal nature of SRH and the hierarchical structure of individuals in provinces, the

appropriate model is the multilevel multinomial logistic (Duncan et al., 1998). This model has

three parts: the response, the fixed part and the random part. The observed response is whether

the individuals rate themselves into one of the three categories. This, for good statistical reasons

(Snijders and Bosker, 2011) is modelled as the log odds of being in fair or good health as

10��

compared to the base category of poor health. The fixed part consists of the predictor variables at

both the individual and province level. The associated estimated effects are on the logit scale but

can readily transformed to relative odds for easier comparison. The random part consists of two

elements: the un-accounted for individual differences which are assumed to come from

multinomial distribution and the residual province differentials on a logit scale. The variance of

the latter summarises the between-province differences conditional on what variables are

specified in the fixed part. This can be transformed to a Median Odds Ratio (Merlo et al., 2005)

to make comparisons with the size of the effects in the fixed part of the model. This is the

median value of the odds ratio between provinces of higher and lower propensity when randomly

picking two provinces. It can be conceptualised as the increased propensity of better health that

would result on average from moving to a province with a higher propensity.

All the models are estimated with the MLwiN software (Rasbash et al., 2009). Due to the

discrete nature of the outcome and the rather small number of higher level units, Bayesian

MCMC estimation is used because of its good properties in this situation (Browne and Draper,

2006).

Results

The models were estimated as a sequence, adding variables from one category at a time and

finally including all variables (Table 2). Model 1 is null model which only includes a constant in

the fixed part. Models 2 and 3 include variables for individual factors (Age, Gender and

Education). From Model 4a to Model 7, each category is added to model 3; Model 8 includes all

11��

the variables. The Deviance Information Criterion (DIC) is used to measure the badness-of-fit for

each model (Spiegelhalter et al., 2002). The DIC is penalized for complexity which is estimated

as the degrees of freedom (dof) consumed in the fit. Any reduction in the DIC is an improvement

in the fit of the model, and a larger reduction in the DIC shows a better fitting model. The change

in DIC between each model and Model 3 is also shown in Table 3. The category with greatest

reduction in the DIC is Financial Support with the bulk of the improvement in fit coming from

the sub-category of Family Income; the Care support category is important with most of the

improvement coming from the Source of Care. The Medical Support category plays a lesser but

still important role. Province health services have little explanatory power. In summary, the

major single determinant is of elderly health is family equivalised income.

(Table 2 about here)

(Table 3 about here)

(Figure 1 about here)

Table 3 presents the results of null model and the full model in the form of logits; it also gives

the residual variance between provinces, and the MOR statistic. All estimates also have their

95% Bayesian credible intervals. In order to appreciate the nature and the relative size of the

effects, the logits of the fixed effects in Model 8 are converted into relative odds ratios with the

base category; the lowest likelihood to report good health, being set to one. The results are

graphed in Figure 1 with a common vertical scale for comparison.

12��

Taking each variable and category in turn we can characterise the nature of the relationships.

For individual factors, the odds of having good health have a U-shaped relationship with the

lowest odds being reported by the 80 year olds with those younger and older reporting higher

odds. Elderly men are 10 percent more likely to have good health than elderly women. Elderly

with formal education are 7% more likely to be in good health than those without but even with a

large sample size the credible intervals include the estimate of no effect. The same can be said

for living arrangement and community social services; there are no substantial differences

according to whom they live with, nor if they receive social services, despite a fifth of the survey

being in receipt of such support. In contrast, the source of care is important. The key difference

in reporting good health is between those who have nobody and those receiving family care

when they are sick (odds of 1.8). The wide credible interval for social services and private live-in

carer as the main sources of care reflects how few have access to such care.

A sizeable effect is found for family income with the lowest odds of being good health being

found for those in the lower quartile with fewer than 360 US dollars. Those in the top quartile of

income (above 840 US dollars) have odds of 1.61 and those in middle quartiles all report higher

odds for good health than those in poverty. Family income matters a lot. The source of this

finance play a role too, with the main difference being the low odds of good health for those

relying on local government as compared to other sources (family, work, and pension). Who

bears the medical expense also has something of a role with the main contrast between those

relying on insurance (with the lowest odds) and the other categories (family, public care and

13��

self). The wide credible interval for not being able to pay reflects the very small numbers in this

category in the survey. There are no sizeable effects for community medical services or for the

province provision of hospital beds.

Table 3 also shows there for the level-2 between-province variance; the unexplained variance in

null model is 0.154, and it declines to 0.129 in the full model, which suggests that these variables

vary systematically in their mean proportion between provinces (Jones and Subramanian, 2012).

The MOR statistic for the full model 8 is 1.41, and the lower and upper CI’s are 1.27 and 1.61,

implying that are still unexplained differences between provinces. These odds are directly

comparable with the ORs of fixed effects variables. On average (in the median), conceptually

moving a person to a province with a higher probability of having good health, the propensity of

reporting good health will increase 1.41 times, which is greater than all the main individual

effects.

To illustrate the differences between provinces, Figure 2 plots the provincial differentials on an

odds scale for Model 2 (age and sex) and Model 8 (all variables) with the lowest odds set to 1.

The graph also provides the 95% comparative confidence intervals. The plot depicts substantial

differences with the best province (Zhejiang) 5.6 times higher than Hainan (the poorest

provinces) in model 2 and still 4.6 times higher in the full model. Figure 3 maps these

differential odds. Darker colour shows higher odds of reporting good health. It is clear that the

north part of China (excluding Liaoning) have better health than the middle and southern China.

Elderly in Hubei, Guangxi and Hainan are the lowest odds of reporting good health. Zhejiang is

14��

the province with best health for the elderly. These differential odds are large compared to all the

variables that we have included in the model; there are substantial (residual) differences between

provinces.

(Figure 2 about here)

(Figure 3 about here)

Discussion and conclusion

In terms of age, previous studies (Smith and Kington 1997; Zeng et al., 2002) found that the

physical and cognitive functions of the elderly declined rapidly as they became older. Our study

shows similar results for most of the elderly, but also exhibits a positive association between

health status and advanced age: elderly who are over 100 have a higher odds of reporting good

health than those who are octogenarians. Similar results were also identified by Liu and Zhang

(2004). This may be explained by Chinese cultural values as well as institutional effects. Chinese

families and societies show great respect for those of advanced age and the local government

also provides financial assistance to the oldest of the old. Such financial subsidy is substantially

higher for people who are over 100 than those who are in their 80s. Selective survivorship and

cohort differences may also play a role in this pattern. In terms of gender, the better health for

men is also found by Liu and Zhang (2004), and Sereny (2011). For education attainment,

Sereny (2011) found that educated elderly are more likely to self-rated health as poor; while, Liu

and Zhang (2004) found that the higher the educational attainment, the better health status the

15��

elderly reported. Li et al., (2009) did not find any sign of effects of education among SRH,

mortality and ADL disability using a two waves study. In conclusion the effect does indeed

appear to be small when other variables are taken into account.

The results suggest that elderly who are more independent, such as those who rely on

themselves for financial support or medical expenses are more likely to report good health.

Previous research showed that elderly who achieved financial and living independence report

better self-rated health based on a survey in Beijing (Meng and Zhang, 2010). A study on the

oldest of the old in China also suggested that elderly who are financially independent express

greater life satisfaction and exhibit better mental status (Li et al., 2007). These results may also

explain the high working rate among the elderly, notably the rural elderly (over 50 percent,

Xinhuanet.com, 2012) as they receive little or no income if they stop working even after

retirement age. With these results it is a possibility that there is reverse causality such that those

who are in poor health are less able to live and support themselves independently.

This independence of the elderly, however, does not indicate lack of need for care support;

those who receive no care are the most vulnerable. Moreover, those who rely on the state

economically are least likely to report good health. In China, only elderly with no children, no

work ability and no income (Three No’s) are eligible to receive state financial state assistance

(Zhang and Liu, 2007) and the coverage has only been recently extended to the oldest old. It is

not surprising, therefore, to find this group of people reporting the poorest health. It reflects the

inadequacy and underdevelopment of the state financial provision for the elderly; as well the

16��

assumption that the state’s responsibility for coverage of the elderly only extends to those who

are the most vulnerable (Ikels, 1992). However, adequate coverage from public medical care

does contribute more to the health of the elderly than those who have to rely on insurance for

medical expenses or out-of-pocket payment.

The results confirm the essential role the family plays in the care system for the elderly, and

those who receive family care are more likely to report good health than those who rely upon

other sources. Social services and professional personal care for the elderly in China is extremely

underdeveloped, the former is provided by the Street Office and primarily targets single elderly

people who are in the most urgent need (Leung, 1997); the latter is accessed by the few elderly

who can afford the services after privatization. For the majority, the family becomes the de facto

carer, particularly after the elderly become older and sick. A positive relationship exists between

family income and health, with those on the poorest income reporting much lower odds of being

in good health. As discussed earlier, high-income households benefit disproportionally from

public resources such as medical resources (Brixiet al., 2010). With access to higher quality

medical treatment, facilities and services, it is not surprising that elderly from these households

report better health. Breakdowns between age category and family support (finance and medical

expense) indicate that the reliance of the elderly on their family increases as they grow older.

Evidence suggests that sizeable geographical inequality exists in the health outcomes of the

elderly. Elderly from Zhejiang are nearly six times more likely to report good health than those

from Hainan. Elderly from Northern provinces on average report better health than those from

17��

southern provinces with some exceptions. Elderly from the old northeast industrial provinces

Liaoning and Jilin report poorer health. Industrial pollution may have a role in this (Zeng and

Vaupel, 2002). Zhejiang, one of the wealthiest provinces, reports the best health result for the

elderly. However, other more developed areas such as Shanghai, Guangdong and Beijing display

fairly poor average health which suggests that factors other than economic level may be

important. Results from a previous study suggest that the health of the elderly is affected by

provincial income inequality level (Feng et al., 2012). Elderly from Hainan are least likely to

report good health from our results and Hainan also experiences the highest income inequality

level (Feng et al., 2012). The provincial level variable--- Total health institutions beds per 1000

population --- is not significant which indicates that the availability of health facilities in a

province does not result in a higher health status of the elderly. The heavy medical expenses

might be the reason for the lack of relationship. Neither province GDP nor province mean

income was important variables in accounting for province differentials. The large residual

province differentials may relate to more intangible cultural differences. In models not presented

here, there were no important differences between urban and rural residents and between the Han

Chinese and minority Chinese ethnic groups; the province differentials did not reduce when

these variables were included.

As China’s society ages, individuals, families and the state are all facing daunting challenges.

Our findings are important for understanding that challenge. We have found that those in poverty

who are missing out on the economic transformation, those who have no one to care for them,

and those who have to rely on insurance are the fractions of the elderly with the poorest health.

All these groups can be expected to grow with increasing economic inequality, the urbanization

18��

of the young, and the further development of the empty nest problem as the elderly are

physically dislocated from their family. Moreover, province differentials which are already large

can be expected to grow alongside greater economic inequality. Although the family can be

expected to retain its vital welfare role, its capacity will be diminished due to reduced family size

and increased lifespan of the old population. Policies need to be implemented to transfer part of

the families’ responsibility, mainly financial support and personal care to other agents. Programs

that help enable the economic independence of the elderly especially rural elderly, such as the

new old-age insurance need to be strengthened. The government needs not only to pursue wider

coverage of these programs, but also extend the benefits that the elderly actually receive. This

future development needs to be targeted more towards less developed areas and underprivileged

groups in order to reduce health inequality amongst the elderly. Our findings reveal that it is not

just the scale of unprecedented demographic that is challenging but also the nature of this

change.

19��

References

Beydoun, M.A., & Popkin, B.M. (2005). The impact of socio-economic factors on functional status decline among community-dwelling older adults in China. Social Science & Medicine, 60, 2045-2057.

Brixi, H., Mu, Y., Targa, B., & Hipgrave, D. (2010). Equity and public governance in health system reform: Challenges and opportunities for China. World Bank Policy Research Working Paper.

Browne, W.J., & Draper, D. (2006). A comparison of Bayesian and likelihood-based methods for fitting multilevel models. Bayesian Analysis, 1, 473-513.

Cai, F., Giles, J., O'Keefe, P., & Wang, D. (2012). The Elderly and Old Age Support in Rural China : Challenges and prospects. In W. Bank (Ed.).

Chu, L.W., & Chi, I. (2008). Nursing homes in China. Journal of the American Medical Directors Association, 9, 239-245.

Du, P., & Wu, C. (1998). The analysis of primary source of economic support among the elderly in China. Population Research, 51-57.

Duncan, C., Jones, K., & Moon, G. (1998). Context, composition and heterogeneity: Using multilevel models in health research. Social Science & Medicine, 46, 97-117.

Feng, Z.X., Wang, W.F.W.N., Jones, K., & Li, Y.Q. (2012). An exploratory multilevel analysis of income, income inequality and self-rated health of the elderly in China. Social Science & Medicine, 75, 2481-2492.

Government White Papers, 2004. http://www.china.org.cn/e-white/20040907/1.htm retrieved on 15 July 2012.

Government White Papers, 2006. http://www.china.org.cn/english/aged/192046.htm retrieved on 15 July 2012.

Gu, D., & Xu, Q. (2007). Sociodemographic effects on dynamics of task-specific ADL functioning at the oldest-old ages. Journal of Cross-cultural Gerontology, 22, 61-81.

Ikels, C. (1992). Family caregiving and the elderly in China. In D. Biegel & A.Blum (Eds.), Aging and caregiving: Theory and practice (pp. 270-284). NewburyPark, CA: Sage Publications.

Jones, K., Gould, M.I., & Duncan, C. (2000). Death and deprivation: an exploratory analysis of deaths in the Health and Lifestyle survey. Social Science & Medicine, 50, 1059-1079.

20��

Jones, K., & Moon, G. (1993). Medical Geography - Taking Space Seriously. Progress in Human Geography, 17, 515-524.

Jones, K., & Subramanian, V.S. (2012). Developing multilevel models for analysing contextuality, heterogeneity and change. Bristol: Centre for Multilevel Modelling.

Kim, I.K. (2007). Social support for older persons: the role of family, community and state in selected Asian countires. Global review of Madrid Plan of action on Aging. Bangkok.

Lei, X., & Lin, W. (2009). The New Cooperative Medical Scheme in rural China: does more coverage mean more service and better health? Health Economics, 18, S25-S46.

Leung, J.C.B. (1997). Family Support for the Elderly in China. Journal of Aging & Social Policy, 9, 87-101.

Li, D.-m., Chen, Y.-Y., & Zhen-yun, W. (2007). Mental status of the oldest-old in China and its related factors. Chinese Mental Health Journal, 2, 614-617.

Li, L.W., Zhang, J., & Liang, J. (2009). Health among the oldest-old in China: Which living arrangements make a difference? Social Science & Medicine, 68, 220-227.

Li, M. 2012. The average monthly pension of rural elderly in China only 74 yuan, they still need to farm work. Xinhuanet.

Lin, G. (2002). Regional variation in family support for the elderly in China: a geodevelopmental perspective. Environment and Planning A, 34, 1617-1633.

Lin, J.Y.F. (1992). Rural Reforms and Agricultural Growth in China. American Economic Review, 82, 34-51.

Liu, G.P., & Zhang, Z. (2004). Sociodemographic differentials of the self-rated health of the oldest-old Chinese. Population Research and Policy Review, 23, 117-133.

Liu, Y., Rao, K., Wu, J., & Gakidou, E. (2008). China’s health system performance. Lancet, 372.

Meng, Q., & Zhang, T. (2010). Social support, self-care status and self-rated health in elderly. Journal of Medical Research, 39, 86-89.

Merlo, J., Chaix, B., Ohlsson, H., Beckman, A., Johnell, K., Hjerpe, P., et al. (2006). A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. Journalof Epidemiology and Community Health, 60, 290-297.

Mohan, J., Twigg, L., Barnard, S., & Jones, K. (2005). Social capital, geography and health: a small-area analysis for England. Social Science & Medicine, 60, 1267-1283.

21��

National Bureau of Statistics (2010), Beijing, China: Chinese Statistics Press.

National Committee on Ageing (2012), China’s urban and rural elderly population status tracking survey report, Beijing, China.

Rasbash, J., Steele, F., Browne, W.J., & Goldstein, H. (2009). A user's guide to MLwiN, version 2.10. Bristol: University of Bristol.

Sereny, M. (2011). Living Arrangements of Older Adults in China: The Interplay Among Preferences, Realities, and Health. Research on Aging, 33, 172-204.

Skapinakis, P., Lewis, G., Araya, R., Jones, K., & Williams, G. (2005). Mental health inequalities in Wales, UK: multi-level investigation of the effect of area deprivation. British Journal of Psychiatry, 186, 417-422.

Smart, A., & Smart, J. (2001). Local citizenship: welfare reform urban/rural status, and exclusion in China. Environment and Planning A, 33, 1853-1869.

Smith, J.P., & Kington, R. (1997). Demographic and economic correlates of health in old age. Demography, 34, 159-170.

Snijders, T.A.B., & Bosker, R.J. (2011). Multilevel Analysis: An Introduction to basic and advanced multilevel modeling. London: Sage.

Spiegelhalter, D.J., Best, N.G., Carlin, B.R., & van der Linde, A. (2002). Bayesian measures of model complexity and fit. Journal of the Royal Statistical Society Series B-Statistical Methodology, 64, 583-616.

United Nations. 2001. World population prospects. New York.

Wang, D.W. (2006). China's urban and rural old age security system: Challenges and options. China & World Economy, 14, 102-116.

Wu, Z., & Schimmele, C.M. (2006). Psychological disposition and self-reported health among the 'oldest-old' in China. Ageing & Society, 26, 135-151.

Yin, D., & Lu, J. (2007). Individual and community factors associated with activities of daily living among the oldest-old: An application of the HLM method. Population Research and Policy Review, 60-70.

Yip, W., & Hsiao, W. (2009). China's health care reform: A tentative assessment. Chinaeconomic review, 20, 613-619.

Zeng, Y., & Vaupel, J.W. (2002). Functional capacity and self-evaluation of health and life of the oldest old in China. Journal of Social Issues, 58, 733-748.

22��

Zeng, Y., Vaupel, J.W., Xiao, Z., Zhang, C., & Liu, Y. (2002). Sociodemographic and health profiles of oldest old in China. Population and Development Review, 28.

Zhan, H.J., Feng, X.T., & Luo, B.Z. (2008). Placing elderly parents in institutions in urban China - A reinterpretation of filial piety. Research on Aging, 30, 543-571.

Zhang, C. (2011). Children, support in old age and social insurance in rural China. Munich University Library of Munich, Germany.

Zhang, W., & Liu, G. (2007). Childlessness, psychological well-being, and life satisfaction among the elderly in China. Journal of Cross-cultural Gerontology, 22, 185-203.

Zimmer, Z. (2005). Health and living arrangement transitions among China's oldest-old. Research on Aging, 27, 526-555.

Figure captions

Figure 1 The odds of good health: results from Model 8

Figure 2 Differential relative odds for provinces derived from models 2 and 8

Figure 3 Maps of odds of good health for provinces derived from model 2 and model 8�

Table 1 Descriptive univariate information on the 2008 CLHLS sample

Dependent variable

Self-rated health Poor (17.5%), Fair (35.2%), Good (47.3%),

Individual factors (n = 12,591)

Age60-69(11.1%), 70-79(19.2%), 80-89(27.3%), 90-99(25.9%), 100+ (16.4%)

Gender Male (44.4%), Female (55.6%)

Education No Schooling (59.5%); Schooling (40.5%)

Care Support

Co-residence Live with spouse (33.5%), Live with family member(s) 49.1%, Alone (15.8%), In a nursing house (1.6%)

23��

Main source of care Family (94.2%), Social service (2.1%), Live-in care giver (1.5%), Nobody (2.2%)

Community social services No (78.8%), Yes (21.2%)

Financial Support

Family equivalised income0~14 (unit: 1,000 US dollars); Quartiles: 0.36, 0.84 and 2.00*

Main source of finance Local government (5.6%), Pension (19.1%), Family (66%), Work (9.4%)

Medical Support

Main medical expense bearer**

Family (62.7%), Public medical care (9.6%), Insurance (6.4%), Self (21%), Cannot pay (0.3%)

Community medical services No (81.4%), Yes (18.6%)

Province-level (n=23)

Health resources Total health institutional beds per 1000 population (2.17~5.17)

*In 2008, 1 US dollar is approximately equal to 7 Chinese Yuan

** This variable is not a true reflection of the public medical care coverage but reflects the extent of the adequate coverage. In addition to those without medical coverage, people with poor medical care coverage may also have to pay a large amount of out-of-pocket medical expenses.

Table 2 Model comparisons

Model Terms dof DIC Change in DIC compared with Model 3

1 Null 25480.5

Individual

2 1 + Age and Sex 27 25417.2

24��

3 2 + Education 28 25408.1

Care support

4a 3+ Co-residence 32 25404.7 3.4

4b 3+ Source of Care 31 25378.6 29.5

4c 3+ Community Social Services 29 25404.2 3.9

4d 3 + Care Support ( All three) 35 25372.4 35.7

Financial Support

5a 3 + Family Income 31 25306.4 101.7

5b 3 + Source of Finance 32 25375.0 33.1

5c 3+ Financial Support (Both) 34 25276.2 131.9

Medical Support

6a 3 + Main medical support 33 25396.7 11.4

6b 3+ Community Medical services 30 25400.1 8.0

6c 3 + Medical Support (Both ) 34 25388.7 19.5

Province level

7 3 +Province health resources 29 25408.6 -0.5

Full Model

8 3+ Care Support + Financial + Medical + Province health resources 46 25228.6 179.6

25��

Table 3 Multilevel multinomial logit estimates

Null Model8

Fixed Part Estimate CI Estimate CI

Intercept.(<=Good) -0.068 (-0.192,0.099) -1.673 (-2.037: -1.305)

Intercept(<=Fair) 1.632 (1.506, 1.804) 0.06 (-0.302: 0.428)

Age ref:80-89

60-69 0.165 ( 0.031: 0.300)

70-79 -0.006 (-0.111: 0.099)

90-99 0.114 (0.020: 0.207)

100+ 0.331 (0.219: 0.445)

Male(ref: Female) 0.101 (0.020: 0.181)

Schooling(ref: Non-schooling) 0.077 (-0.008: 0.161)

Care Support

Co-residence ref: Live with spouse

Live with family member (s) 0.121 (0.027: 0.215)

Alone 0.080 (-0.031: 0.193)

In a nursing house 0.151 (-0.167: 0.472)

Main source of care ref: Nobody

Family 0.590 (0.355: 0.825)

Social service 0.249 (-0.091: 0.595)

Live-in care giver 0.181 (-0.187: 0.547)

Community social services ref: No 0.058 (-0.039: 0.154)

Financial Support

Family equivalised income ref: First quartile

Second quartile 0.309 (0.213: 0.405)

26��

Third quartile 0.477 (0.376: 0.580)

Fourth quartile 0.479 (0.370: 0.589)

Main source of finance ref: Local government

Pension 0.209 (0.030: 0.388)

Family 0.225 (0.062: 0.386)

Work 0.502 (0.308: 0.698)

Medical and health care support

Main medical expense bearer ref: Insurance

Public medical care 0.248 (0.060 : 0.435)

Self 0.346 (0.162: 0.530)

Family 0.119 (-0.067: 0.308)

Cannot pay 0.321 (-0.304: 0.960)

Community medical services ref: No 0.113 (0.012: 0.215)

Total health institutional beds per1000 0.141 (-0.053: 0.336)

Random Part: between provinces

Variance 0.154 (0.079: 0.29) 0.129 (0.064: 0.248)

MOR (odds) 1.45 (1.31: 1.67) 1.41 (1.27: 1.61)

�

Figu

re 1

Figu

re(s

)

Figu

re 2

Figure 3