Open Access

Explaining provincial government health expenditures in China: evidence from panel data 2007–2013

China Finance and Economic Review20175:9

DOI: 10.1186/s40589-017-0054-2

Received: 1 May 2017

Accepted: 22 June 2017

Published: 3 July 2017

Abstract

Background

Since the mid-2000s, the Chinese government has increased government health expenditures (GHE) significantly to address widespread complaints about health delivery. This study examines the real per capita provincial GHE over the period 2007–2013 to identify the determinants of provincial GHE during the most recent round of health reforms.

Methods

A range of theoretically grounded socioeconomic indicators were collected from the China Statistical Yearbooks and then factored to reduce the number of highly correlated indicators. Maps were drawn to visualise the spatial patterns of key variables and fixed-effects regressions were run to test relationships between the real per capita provincial GHE and various variables. GMM estimators were used to address endogeneity problems.

Results

Key determinants of provincial GHE in China include the real per capita budgetary deficits, economy, and industrial structure (two factors composed from an exploratory factor analysis). Increasing 1000 yuan real per capita budgetary deficits was expected to increase the real per capita GHE by 34 yuan. A one-unit increase in the economy was associated with a 249 yuan higher real per capita GHE, while a one-unit increase in the industrial structure was expected to decrease the real per capita GHE by 33 yuan.

Conclusions

The findings of this study reveal a worrisome picture: potential inefficiencies of the central government’s funding efforts and the overwhelming importance of economic development for GHE.

Keywords

Government health expenditures Panel data analysis China

Background

In contrast to its rapid economic growth, the equality of health care in China deteriorated significantly in the first two decades of economic reform. The gaps in health status and access to health services were widened both between urban and rural areas and across provinces (Liu et al. 1999; Zhang and Kanbur 2005). In 2000, China ranked 188th among 191 countries for fairness in health finance in the World Health Report (World Health Organization 2000), illustrating the severity of the inequality problem in China.

As a response to widespread complaints of growing inequality-related health problems, the Chinese government has initiated and implemented a series of policy changes since the mid-2000s. During the Sixth Plenum of the 16th Congress held in 2006, the unaffordability and inaccessibility of health services were formally conceptualized as “disharmonious features,” and the establishment of a harmonious society by 2020 became the new chief task for China (Chinese Communist Party 2006; Woo 2009). Furthermore, in early April 2009, the central government announced a massive expansion of its health provisions with the aim of providing basic health services to the whole population by 2020 (Chinese Communist Party and State Council 2009). This is an important milestone signifying the Chinese state’s determination to reverse the trend of deteriorating equity throughout the 1980s and 1990s. A range of efforts have been made to expand the coverage of health insurance schemes and improve the accessibility of health services.

Following the change in national policy direction, government health expenditures (GHE) jumped from 196 billion yuan in 2007 to 1195 billion yuan in 2015, a 510% increase within 8 years (National Statistical Bureau of China 2008; 2016). In comparison, the increase in gross domestic product (GDP) was only 152% during the same time period (National Statistical Bureau of China 2016). This signifies the government’s determination to assert its role in the health sector.

Although the most recent round of health reforms has been documented and evaluated by a number of recent studies (for example, Chen 2009; Yip and Hsiao 2009; Zhao and Huang 2010; Yip et al. 2012), there has been a lack of research devoted to the pattern of provincial GHE. Studying provincial GHE is necessary because it shows how government resources have been allocated across the country and whether the pattern has been aligned with the prevailing target of improving health equality. The objective of this research is twofold. The first is to identify the pattern of provincial GHE during the most recent round of health reforms. The second is to explain this pattern by examining the impacts of three sets of key potential explanatory variables including economic development, globalization, and fiscal institutional structure.

Literature review

The link between inequality in health outcomes and access to health services in China has been well documented (Zhang and Kanbur 2005; Li and Wei 2010; Fang et al. 2010; Uchimura and Jutting 2009). However, the understanding of determinants of why certain provinces have more or less GHE is more limited. For this reason, this review draws heavily from the literature on the determinants of welfare and public expenditure more broadly to compensate. In general, scholarly discussions of the determinants of GHE and welfare/public expenditure can be grouped into three strands. The first strand of literature highlights the importance of economic development. The German economist Adolph Wagner was among the earliest to predict that the development of an industrial economy would be accompanied by an increased share of public expenditure, later known as “Wagner’s Law.” Welfare state theorists, adopting the “logic of industrialism” approach, echoed this proposition and further argued that welfare state developments could also be attributed to economic growth (Wilensky and Lebeaux 1965; Wilensky 1975). They explained that the increase in government expenditure on welfare was driven by a growing demand originating from the industrialization process.

Empirical tests of Wagner’s law, or the industrialization thesis, in the context of China have obtained mixed results. While some researchers identified a positive relationship between economic development and the expansion in public expenditure (Tobin 2005), others challenged this observation and cast doubt on the existence of Wagner’s law in China (Lin and Song 2002; Huang 2006; Wu and Lin 2012). Leaving aside the contentiousness of the impact of economic development on the total public expenditure, studies examining the relationship between economic development and GHE exclusively in China have been scarce. Among the limited number of studies, Pan and Liu (2012) used panel data regression analyses for panel data from Chinese provinces over the period of 2002–2006 and concluded that provincial GHE is indeed affected by the local economy, a relationship further explored in this study.

The second strand of literature focuses on the impacts of globalization based on two competing debates. On the one hand, the race-to-bottom thesis claims that increases in trade and capital openness pressure governments to race to the bottom in social spending and labor standards to increase their competitiveness (Clayton and Pontusson 1998; Mishra 1999; Scharpf and Schmidt 2000; Swank 2010). On the other hand, the compensation thesis argues that economies that are more exposed to external risks expand the scale of their welfare states to compensate for these risks (Cameron 1978; Katzenstein 1985; Rodrik 1997, 1998). Apart from these two major theses, a large volume of work has been generated to question the impacts of globalization. The skeptics de-emphasize the effects of globalization and contend that globalization alone has limited influence on welfare state development (Swank 2002).

The race-to-bottom thesis has some support among researchers studying China. For example, Walker and Wong (2005) and Chau and Yu (2005) reasoned that the Chinese government intentionally keeps social welfare provisions limited to provide a preferred investment environment to please foreign investors out of fear that capitalists would withdraw their investments if welfare and labor costs were raised. However, this theory has yet to be tested with empirical data. Wu and Lin (2012) showed some evidence that openness to trade and foreign direct investment curtailed government expansion, but they did not differentiate between the different types of public expenditure. Overall, there is a lack of literature on the relationship between globalization and GHE in China.

Researchers in the third school of thought emphasize the role of institutional settings. Although focusing on different aspects of institutional structures, these institutionalists all point to the importance of institutional structures in shaping welfare state development (Immergut 1992; Pierson 1994, 2001; Obinger et al. 2005). In the case of China, among various institutional structures, it is the local-central relationship, particularly fiscal decentralization that has received the most scholarly attention. In their widely cited paper, West and Wong (1995) suggested that fiscal decentralization in China contributed to large and growing interregional inequalities in the provision of public services, including health. When higher level governments struggle to balance their own budgets, which has been the case since the 1994 tax reform, they tend to devolve expenditure responsibilities downward. However, the lower level governments are likely to have even more difficulties in balancing their budgets, therefore resulting in an inefficient provision of public services. As noted by Wong (2007), the intergovernmental system was undermined by the piecemeal reforms applied to the fiscal system during the 1980s and especially the 1990s, which eroded the ability of local governments to perform many of their assigned functions.

In terms of GHE, Pan and Liu (2012) revealed that a 10% increase in transfers from the central government (which are used to compensate for budget deficits) increase GHE by 2.27%. In a case study of a poor rural county, Tang and Bloom (2000) found little evidence that lower level governments mobilized additional financial resources, thereby confirming the argument that fiscal decentralization is detrimental to public health services in China. Two relevant studies used the total health expenditure as the dependent variable. Chou (2007) performed the panel LM unit roots tests using a sample of data from 28 provinces in China, covering the period 1978–2004. They demonstrated that government budget deficits have a significant impact, with every 10 million yuan increase in budget deficits associated with an approximate 26.3% decrease in health expenditure in the long term. Based on the same dataset, Chou and Wang (2009) later carried out cross-section regressions and a cluster analysis to prove that provincial government budget deficits are useful in explaining the disparity in health expenditures. Their findings indicate that budget deficits decrease the total health expenditure, which may also extend to GHE.

Apart from the three major strands of literature discussed above, some empirical studies also included socio-economic indicators in their analysis and showed certain patterns. For example, Brixi et al. (2013) found GHE to be regressive (negatively correlated with the population’s basic health needs and financial barriers in accessing health care) across and within provinces. Pan and Liu (2012) highlighted that the proportion of the population under age 15, medical insurance coverage, and urbanization are also key determinants of real per capita provincial GHE in China.

A review of the existing literature revealed two major gaps. First, empirical examinations of the determinants of GHE in China have been rare. Although the studies of public expenditure or health expenditure shed some light on determinants of GHE, their results could not be easily applied to GHE. The limited number of studies devoted to investigate determinants of GHE have focused on the time period between 1978 and 2006. However, the Chinese government’s approach to health has changed significantly since the mid-2000s, indicating a need to reassess these relationships. Furthermore, the increases in government spending and the changes in health policies may also mean new determinants of GHE. Second, among the three strands of literature, economic growth and fiscal decentralization received relatively more scholarly attention in China, but as the theoretical literature implies, globalization could also have an impact on GHE. A statistical test of this relationship is warranted.

To fill these gaps in the literature, this paper tests the relationship between GHE and all three potential explanatory factors identified: economic growth, globalization, and budgetary deficits. It covers the most recent period to examine whether there have been significant changes in determinants of GHE since the mid-2000s.

An overview of government health expenditure in China

Importance of government health expenditure

The importance of GHE can be reflected in its substantial contribution to the total health expenditure (THE) in China. The government’s growing input has increased GHE contributions during the study period: while GHE made up 22% of THE in 2007, the percentage jumped to 30% in 2013 (National Statistical Bureau of China 2014). Since the government expenditure is usually expected to have the strongest equalizing effect among the key components of THE,1 it is essential to look at the allocative pattern of GHE to investigate whether it has indeed played an equalizing role.

As Table 1 shows, GHE in China consists of four major components. The most significant one is medical insurance. In particular, two social health insurance schemes, the New Rural Cooperative Medical Scheme (NRCMS) and the Urban Residents’ Basic Medical Insurance (URBMI)—the former for rural residents and the latter for urban residents who are not formally employed—are heavily subsidized by the government and have absorbed the majority of GHE on medical insurance. The remaining GHE is almost evenly divided among subsidies to public hospitals, primary care facilities, and public health. In other words, examining the allocation of GHE is particularly helpful for us to understand the aggregate allocative patterns of subsidies to medical insurance, public hospitals, primary care facilities, and public health.
Table 1

Key components of government health expenditure, 2010–2013

 

2010 (%)

2011 (%)

2012 (%)

2013 (%)

Administration

3

2

2

2

Public hospitals

18

15

14

14

Primary care facilities

9

9

12

11

Public health

16

17

15

15

 Basic Public Health Services

4

5

5

5

Medical insurance

46

51

50

52

 NRCMS

22

27

28

29

 URBMI

4

6

6

7

Data source: Author’s calculation based on statistics compiled from the national final accounts released on the official website of the Ministry of Finance (http://yss.mof.gov.cn/zhengwuxinxi/caizhengshuju/), various years

Origin of local variations

By making national policies, the central government determines the broad outline of budgeting and expenditure assignments. Since the mid-2000s, there has been a tendency for the central government to expand the list of expenditure assignments under the banner of “harmonious society” (Wong 2016). In the health sector, for example, a series of notifications have been issued to set government subsidy standards for the NRCMS and URBMI. Since 2009, the central government also clarified the government subsidy standard for providing basic public health services.

Although the central government has the authority to assign expenditure responsibilities, large variations exist across localities. This is possible for several reasons. First, national policies in China usually maintain some flexibility and allow local governments to make adjustments according to local circumstances. Second, the budget law establishes the legal foundation for local governments’ autonomy in budgeting.2 Governments at all five levels, including the central, provincial, prefectural, county, and township levels, need to have an independent budget that must be approved by the People’s Congress at that level.3 Third, the complex institutional arrangement makes China’s fiscal system rather decentralized. The structure is a nested hierarchy: the central government deals directly only with the provincial administrations, which in turn only deal with their respective prefectures, and so on. This arrangement creates a large number of de facto decision makers in budgetary processes, making GHE at different levels results of complicated negotiations.

In addition, China’s fiscal system is further characterized by two salient features: heavy expenditure responsibilities for sub-national governments and a high level of reliance on transfers (World Bank 2002). In aggregate, 99% of GHE is spent at local levels, and approximately 30% of the spending is financed by transfers from the central government (Table 2). The total scale of transfers should be larger because, similar to the central government, sub-national governments also make transfers to lower level governments. The consequence is that GHE at each level of government is financed by a combination of local revenues and transfers from higher level governments.4 Given that the availability of a local government’s revenues is usually dependent upon local economic development, the industrial structure, and other factors, while the amount of transfers can be attributed to higher level governments’ own resources, preferences and negotiations between different levels of governments, GHE is a result of many factors. To identify the key determinants among the long list of potential factors is the central task of this study.
Table 2

Central-local division of government health expenditure, 2010–2013

 

2010 (%)

2011 (%)

2012 (%)

2013 (%)

The central government

1

1

1

1

Local governments

99

99

99

99

 Transfers from the central government

29

26

27

30

Data source: Author’s calculation based on statistics compiled from the national final accounts released on the official website of the Ministry of Finance (http://yss.mof.gov.cn/zhengwuxinxi/caizhengshuju/), various years

Methods

Data

Data for 31 provinces in China (excluding Hong Kong, Macao, and Taiwan) for the period 2007–2013 were collected from China Statistical Yearbooks (20082014), the most comprehensive source of official data in China. Only post-2007 data were applied because 2007 is the year when classification methods for budgetary items (including GHE) changed significantly, which makes the data since then incomparable to those before 2007.

The dependent variable for this study is GHE, measured by per capita budgetary expenditures on health in each province. As identified in previous research, three key explanatory factors (independent variables) for GHE need to be tested: economic development, globalization, and budgetary deficits. After a selection based on the relevance, importance, and availability of data, indicators for the three independent variables were limited to one indicator for budgetary deficits (per capita budgetary deficits) and 15 indicators for economy and globalization. Three indicators of population characteristics were also identified as potential control variables for later analysis. Since the dataset covers the years between 2007 and 2013, price-related items need to be converted to real values to subtract the influence of inflation and in this way to better facilitate comparison between years. For this purpose, the consumer price indices (CPI) reported by the Yearbooks were used to convert all price-related indicators to values at the 2010 price level. Descriptive statistics for each indicator are summarized and presented in Table 3.
Table 3

List of indicators and descriptive statistics

Category

Indicators (unit)

Obs.

Mean

SD

Min

Max

GHE

Real per capita GHE (yuan)

217

429

226

101

1159

Budgetary deficits

Real per capita budgetary deficits (yuan)

217

3532

3632

334

26179

Economic development and openness

Real per capita gross regional product (GRP) (yuan)

217

33120

17865

7967

87915

Real per capita disposable income in urban areas (yuan)

217

18356

5560

11257

39633

Real per capita net income in rural areas (yuan)

217

6516

2962

2595

17710

Real per capita salary in urban enterprises (yuan)

217

36004

11331

19696

82546

Real per capita consumption (yuan)

217

11057

6054

3523

35450

Percentage of GRP from the first industry

217

0.11

0.06

0.01

0.30

Percentage of GRP from the secondary industry

217

0.48

0.08

0.22

0.62

Percentage of GRP from the tertiary industry

217

0.41

0.09

0.29

0.77

Unemployment rate

215a

0.04

0.01

0.01

0.05

Percentage of population employed in urban enterprises

217

0.11

0.06

0.06

0.35

Percentage of population who are self-employed or work in private enterprises

217

0.12

0.06

0.04

0.32

Funds from Hong Kong, Macao, and Taiwan/total investment in fixed assets

216b

0.03

0.03

0.00

0.15

FDI/total investment in fixed assets

217

0.03

0.03

0.00

0.12

Trade (by location of importers or exporters)/GRP

217

0.29

0.39

0.01

1.80

Trade (by place of destination or origin)/GRP

217

0.28

0.35

0.00

1.71

Population characteristics

Percentage of urban population

217

0.51

0.15

0.23

0.90

Adolescent dependency rate

217

0.22

0.07

0.09

0.42

Dependency rate of the aged

217

0.12

0.03

0.05

0.19

Data sources: Author’s calculation based on statistics compiled from National Statistical Bureau of China (2008–2014) and Ministry of Health (2008)

aThe two missing values are the values for Tibet in 2007 and 2008

bThe missing value is the value for Tibet in 2009

Visualization

To visualize the indicators, the provincial average values for the years between 2007 and 2013 were calculated5 and presented as maps. The average values were used instead of values in any single year to show the overall pattern across the study period. Maps were produced by using ArcMap (ArcGIS 10.3). China’s shapefile (province map) was downloaded from the website StatSilk. “Quantile” was used as the method to classify data for clearer visualization, and this classification scheme was followed in all maps in this paper for consistency.

Exploratory factor analysis

One problem in running regression with the identified indicators is that many of them are highly correlated. This is intuitive because these 15 indicators all measure either the economy or openness, which are correlated with each other. If all these variables were entered into the same regression, the estimates of their separate effects would be hampered by multicollinearity, thereby making the coefficient estimates of the multiple regression unreliable across samples.

To address this issue, an exploratory factor analysis was performed to identify a smaller number of less correlated factors to capture the common variance among the original indicators. An exploratory instead of confirmatory factor analysis is more appropriate in this study because there is no theory-based hypothesis of the meaning and number of factors (Bartholomew et al. 2008). The factors were estimated for the general linear factor model (p observed indicators and q factors):
$$ {X}_i = {\alpha}_{\mathrm{i}0} + {\displaystyle \sum_{j=1}^q}{\alpha}_{i j}{f}_j + {\varepsilon}_{\mathrm{i}}\ \left( i=1,\dots,\ p\right) $$
where X i are the indicators, f j are the common factors, ε i are residuals, and α ij are the factor loadings. The factor loadings indicate the degree of correlation between the indicators and the factors. The higher the load is, the more relevant it is in defining the factor’s dimensionality.
The principal component approach to factor analysis was adopted to extract factors. The number of retained factors was determined by a combination of the Kaiser criterion and observation of the proportion of variance explained by each factor and the scree plot. The judgment of whether one indicator should be included for a factor analysis was based on observation of its factor loadings and uniqueness. When the factor loadings were low and uniqueness high, indicating that they were not well accounted for by factors, the indicators were removed from the factor analysis. To fine tune the model, promax rotation with Kaiser normalization was applied. Promax, an oblique rotation, was chosen because it allows us to relax the assumption of the linear factor model that the factors be independent. Given the indicators listed in Table 3, it is more reasonable to expect that factors are correlated. The analysis was conducted by using Stata 13.0. For later analyses, the factor scores, the provincial scores for the factors, were calculated and obtained as a linear combination of the indicators following the formula:
$$ {F}_j = {\displaystyle \sum_{i=1}^p}{C}_{i j}{X}_i\ \left( j=1,\dots,\ q\right) $$
where F i are the factor scores, X j are the indicators, and C ij are the factor score coefficients.

Panel data regression analysis

The relationships between GHE and various independent variables were tested by fixed-effects regression analyses to account for the impact of changes in our predictors on our outcome measure, GHE. The independent variables included the common factors obtained from factor analysis. Two control variables included one variable that was removed from the factor model due to its high level of uniqueness and the other one describing population characteristics. Scatterplots were first drawn to identify outliers and visualize the relationships between the dependent variables and four key independent variables.

To account for panel effects, two models are available: fixed effects and random effects. The major advantage of the fixed-effects regression model is that it eliminates the omitted variable bias arising both from unobserved variables that are constant over time and from unobserved variables that are constant across provinces (Stock and Watson 2012). This is a desirable property for data in this study because unobserved province-specific and time-specific effects are expected. For example, one unobserved province-specific effect could be the government administrative capacity in a given province. A stronger capacity may be associated with higher GHE while also contributing to the local economy and openness. If not controlled, this type of effect would lead to an omitted variable bias, which makes the regression results unreliable. On the other hand, time-specific effects are also relevant because national policies in a given year could affect both the dependent and independent variables in the regression. The fixed effects regression model, by focusing on the changes in the same province over years and variations across entities in the same year, can control for unobserved variables that are constant either over time or across provinces. The regression model (n independent variables) including both the province and time fixed effects could be written as follows:
$$ {Y}_{i t} = {\alpha}_i + {\lambda}_t + {\displaystyle \sum_{j=1}^n}{\beta}_j{X}_{j it}+{u}_{i t} $$
where Y it is the real per capita GHE for province i at year t, X jit are independent/control variables for province i at year t, u it are error terms, α i are the province fixed effects, λ t are the time fixed effects, and β j are unknown coefficients.

The other option, the random-effects model, assumes the variation across provinces to be random and uncorrelated with the predictor or independent variables included in the model. The model looks very similar to the fixed-effects model, with the only difference being the addition of the term ε it at the end of the formula to account for within-entity errors. Since the choice between the fixed-effects and random-effects models depends on whether the province-specific effects are correlated with the independent variables, a Hausman test was conducted to determine which model is more appropriate. The results indicated that the fixed effects model should be adopted.

One problem with the regression results is potential endogeneity due to reversed causality. GHE can potentially enlarge budgetary deficits and promote the economy through the enhancement of human capital. In other words, real per capita budgetary deficits and economy are potentially endogenous. To test for this, endogeneity tests were conducted. This study then followed Wu and Lin (2012)’s and Checherita and Rother (2010)’s approach by adopting the generalized method of moments estimators and using lag regressors as instruments to mitigate the possibility of reversed causation.6 A series of standard tests, including those for under-identification, weak identification, and over-identification, were carried out to test the appropriateness of models.

Results

Spatial patterns

To capture variation in GHE, Fig. 1 presents a graphical overview of real per capita GHE between 2007 and 2013. Interestingly, the fifth quintile group (with the highest real per capita GHE) includes Beijing, Tianjin, Shanghai, Tibet, Qinghai, and Ningxia, the former three of which are municipalities, while the latter three are poor provinces in the western region. Given that the western region is the most economically underdeveloped and the people there are more likely to find health services inaccessible or unaffordable, the overall higher level of GHE in the west appears to reflect that the central government’s efforts to improve equality across provinces have succeeded to some extent. Nevertheless, it should be noted that provinces in the central region, which is also a relatively poor region, had a low level of GHE in general.
Fig. 1

Real per capita GHE, 2007–2013 (yuan). The figure shows the spatial pattern of real per capita government health expenditure. For each province, per capita government health expenditure was calculated by dividing the government appropriation for health by the local population, which was adjusted to the 2010 price level by the corresponding consumer price index. The values from 2007 to 2013 were then averaged

From the review of the literature and China’s GHE, the central-local fiscal dynamics were identified as an important determinant of GHE. Figure 2 shows the spatial pattern of real per capita budgetary deficits. While western provinces had larger budgetary deficits, provinces in the coastal region had healthier balance sheets. The two figures together provide some descriptive evidence that GHE appears to be positively correlated with budgetary deficits, a relationship further explored in the “Regression results” section.
Fig. 2

Real per capita budgetary deficits, 2007–2013 (yuan). The figure shows the spatial pattern of the real per capita budgetary deficits. For each province, the per capita budgetary deficit was calculated by subtracting the government budgetary revenue from the government budgetary expenditure and then dividing the difference by the local population. The values in different years (from 2007 to 2013) were all adjusted to the 2010 price level by the corresponding consumer price index and were then averaged

Common factors for indicators

In the first round of factor analysis, all 15 indicators for economy and openness were included. The uniqueness of only one variable, the unemployment rate, was particularly high, with its uniqueness as high as 62.51%, meaning that the majority of its variance could not be explained by the three common factors. The variable was therefore removed, and the 14 other variables were included for a second round of factor analysis. The data for the second round of factor analysis had a Kaiser-Meyer-Olkin (KMO) index of 0.843, meaning that the dataset was suitable for factor analysis. The Bartlett test of sphericity also confirmed the suitability of the factor analysis with the p value at 0.00. Three factors were identified, together explaining 94.22% (68.56% by the first factor, 13.42% by the second factor, and 12.24% by the third factor) of the common variance of the 14 indicators. After a promax rotation with Kaiser normalization, the results are shown in Table 4.
Table 4

Factor loadings of the 14 economic and globalization indicators

Indicators

F1

F2

F3

Uniqueness

Real per capita gross regional product (GRP) (yuan)

0.9441

  

0.0875

Real per capita disposable income in urban areas (yuan)

0.9288

  

0.1132

Real per capita net income in rural areas (yuan)

0.9191

  

0.0965

Real per capita salary in urban enterprises (yuan)

0.9539

  

0.1278

Real per capita consumption (yuan)

0.8816

  

0.0592

Percentage of GRP from the first industry

−0.9048

 

0.3696

0.2118

Percentage of GRP from the secondary industry

  

−1.0326

0.0026

Percentage of GRP from the tertiary industry

0.4321

 

0.7191

0.0661

Percentage of population employed in urban enterprises

0.7789

  

0.2089

Percentage of population who are self-employed or work in private enterprises

0.6000

0.3157

 

0.2876

Funds from Hong Kong, Macao, and Taiwan/total investment in fixed assets

 

0.8297

 

0.4048

FDI/total investment in fixed assets

 

0.8446

 

0.2306

Trade (by location of importers or exporters)/GRP

 

0.7565

 

0.1191

Trade (by place of destination or origin)/GRP

 

0.9191

 

0.1026

Notes: The total number of observations is 216 because the value for the FDI variable of Tibet in 2009 is missing. Only factor loadings higher than 0.3 are displayed. For the sensitivity test, the data for the years between 2009 and 2013 were also factored, and the results were similar and therefore are not reported separately here

The first factor loaded heavily (|factor loadings| > 0.90) on per capita salary in urban enterprises, per capita GRP, per capita disposable income in urban areas, per capita net income in rural areas, and first industry (negative). The factor was named “economy” because all the variables listed here mainly describe the level of economic development. The second factor was named “openness” because it loaded heavily (|factor loadings| > 0.80) on the four items measuring the percentage of funds from non-domestic sources in the total fixed-asset investment and the percentage of trade in GRP. The last factor was named “industrial structure” because it mainly loaded on the three indicators describing the industrial structure.

In this way, the 14 indicators for economic development and globalization were captured by three factor variables. The first factor and third factor both describe the economy, the first independent variable to be tested in this study. The former concerns the level of economic development, while the latter describes the industrial structure. The second factor describes openness, the second independent variable to be investigated.

The average provincial factor scores for the years between 2007 and 2013 are presented in Figs. 3, 4, and 5. For economy, the northern and coastal provinces tended to have higher factor scores compared to their counterparts in the western and middle regions. The degree of openness was also higher in coastal areas than those in the western and middle regions. The greatest difference is that provinces in the northern area no longer belong to the high-performer group and the southern areas tended to be more open. For industrial structure, the provinces in the south-western regions had higher scores.
Fig. 3

Economy, 2007–2013. The figure shows the spatial pattern of scores for the first factor from the factor analysis. The factor was named “economy” because it loaded heavily on the per capita salary in urban enterprises, per capita GRP, per capita disposable income in urban areas, per capita net income in rural areas, and first industry (negative)

Fig. 4

Openness, 2007–2013. The figure shows the spatial pattern of scores for the second factor from the factor analysis. The second factor was named “openness” because it loaded heavily on the four items measuring the percentage of funds from non-domestic sources in total fixed-asset investment and the percentage of trade in GRP

Fig. 5

Industrial structure, 2007–2013. The figure shows the spatial pattern of scores for the third factor from the factor analysis. The last factor was named “industrial structure” because it mainly loaded on the three indicators describing the industrial structure

As discussed in the “Exploratory factor analysis” section, the reason the factor analysis was carried out in the first place was that the original indicators were highly inter-correlated (the correlation coefficients between many indicators were higher than 0.80 or even 0.90), therefore posing problems for the regression analysis. Table 5 shows the correlation coefficients between the new indicators generated from the factor analysis. According to the table, the largest coefficient is the one between economy and openness, with a value of 0.5617. Given that this figure is much lower than the correlation coefficients between many original indicators and indicates a moderate (instead of strong) relationship, it was considered acceptable for the following regression analyses.
Table 5

Correlation coefficients between factors

 

Economy

Openness

Industrial structure

Economy

1.0000

  

Openness

0.5617

1.0000

 

Industrial structure

0.3039

0.2080

1.0000

Regression results

After the factor analysis was performed, three key indicators were generated to serve as proxies for the first two independent variables explored in this study: economy and industrial structure, which together measured the first key independent variable in this study, economic development; and openness captured globalization, the second key independent variable. The per capita budgetary gap quantified the fiscal institutional structure between central and local governments, the third independent variable as identified in the literature review.

The indicator of the unemployment rate, although capturing some aspects of the economy, was excluded from the factor analysis due to its high uniqueness. It was added back to the regression to serve as a control variable due to its potential correlation with GHE (because unemployed people usually have a greater demand for health care services and are in particular need of government support). Among the three population characteristics identified at the beginning, the percentage of the urban population, the adolescent dependency rate, and the dependency rate of the aged, only the last was entered into the regression because the first two were found to be strongly correlated with economy. This is intuitive because more urbanized areas tend to have higher levels of economic development, and it is much more expensive to raise children in these areas, therefore lowering the number of children. In total, four indicators plus two control variables were included in the regression analyses.

The data of the dependent variable and six independent/control variables for the years between 2007 and 2013 had a p value of 0.000 for the Breusch and Pagan Lagrangian multiplier test for random effects, indicating that the null hypothesis that variance across entities is zero was rejected at the 1% significance level. In other words, OLS regression is inappropriate and panel effects should be considered.

The solution to address the problem with the panel effects was to examine the changes in various variables across years instead of their values in different years. Scatterplots for the average annual difference in real per capita GHE and various independent variables were drawn. In all figures, Qinghai is clearly an outlier due to its large increase in real per capita GHE. Tibet is also an obvious outlier, with its large increase in real per capita budgetary deficits. It appears that Qinghai and Tibet may not fit into the same model with the other provinces. Therefore, regressions were run both with and without these two provinces for sensitivity test.

Two time periods, the years between 2007 and 2013 and the years between 2009 and 2013, were also analyzed separately to see whether there were significant changes since the initiation of the most recent round of heath reforms. The results for the regression analysis are presented in Table 6. Year effects were tested for all models, and the null hypothesis that the coefficients for all years are jointly equal to zero was always rejected, with a p value of 0.000. Therefore, control of time effects is essential. Two additional model specification issues with the models are heteroskedasticity and autocorrelation. The results for the modified Wald test for group wise heteroskedasticity and the autocorrelation test suggested the necessity of correcting both. For this reason, heteroskedasticity and autocorrelation were controlled for in all regression models presented in this paper.
Table 6

Fixed-effects (FE) models

 

Model 1

Model 2

Model 3

Model 4

Number of provinces

31

29 (Tibet and Qinghai excluded)

31

29 (Tibet and Qinghai excluded)

Years

2007–2013

2007–2013

2009–2013

2009–2013

Real per capita budgetary deficits

0.019** (0.009)

0.042*** (0.006)

0.009 (0.012)

0.031*** (0.009)

Economy

84.042** (40.975)

124.566*** (31.532)

85.828 (64.939)

145.280** (55.846)

Openness

0.539 (8.048)

−3.593 (7.674)

−12.485 (9.183)

−13.167 (9.569)

Industrial structure

−17.628 (11.124)

−18.217** (8.769)

−34.009 (21.513)

−28.022 (18.270)

Unemployment rate

−816.269 (2848.755)

−2101.171 (1383.514)

0.471 (2550.578)

−1789.106 (1473.1)

Dependency rate of the aged

961.461*** (208.640)

769.547*** (161.244)

1136.961*** (246.679)

865.509*** (213.074)

Factors = 0

0.100

0.001

0.214

0.030

Year dummies = 0

0.000

0.000

0.005

0.010

Panel effects = 0

p value = 0.000

p value = 0.000

p value = 0.000

p value = 0.000

Hausman test

p value = 0.000

p value = 0.000

p value = 0.000

p value = 0.000

Heteroskedasticity test (Modified Wald)

p value = 0.000

p value = 0.000

p value = 0.000

p value = 0.000

Autocorrelation test

p value = 0.000

p value = 0.004

p value = 0.000

p value = 0.001

N

214

203

154

145

R-squared (within)

0.9580

0.9732

0.9133

0.9416

Notes: Standard errors are given in parentheses after the coefficients. The individual coefficient is statistically significant at the *10%, **5%, or ***1% significance level. The statistics are robust to heteroscedasticity and autocorrelation. The null hypothesis for the Hausman test is that the preferred model is random effects. A significantly low p value indicates that fixed effects should be used

According to Table 6, in both time periods, the results for the fixed effects are sensitive to whether Tibet and Qinghai are included, confirming the earlier observation that both are outliers in the scatterplots. Given the particularities of these two provinces, it makes more sense to rely on the regression models where they are excluded. Furthermore, the results for the fixed-effects and random-effects models were different, demonstrating the impacts of whether the variance across provinces was assumed to be random. Hausman tests were conducted for all fixed-effects models presented here and their corresponding random-effects models. The p value remained at 0.000, which means that the null hypothesis that the unique errors are uncorrelated with the regressions was rejected. In other words, the variance across provinces is unlikely to be random, and fixed-effects models are more appropriate. For these reasons, only fixed-effects models with Tibet and Qinghai excluded are discussed below.

For the period between 2007 and 2013, coefficients for the real per capita budgetary deficits, economy, and dependency rate of the aged are all significant at the 1% significance level. The coefficient for real per capita budgetary deficits is 0.042, meaning that increasing 1000 yuan for real per capita budgetary deficits would lead to 42 yuan more real per capita GHE. The coefficient for economy is 125. This indicates that increasing one unit of economy would contribute to 125 yuan more real per capita GHE. The coefficient for the dependency rate of the aged is 770, which suggests that an increase of one percentage point of the dependency rate of the aged is associated with 8 yuan more real per capita GHE. Compared to the model for the years between 2007 and 2013, the corresponding model for the period between 2009 and 2013 yielded similar results. Still, the coefficients for the real per capita budgetary deficits, economy, and dependency rate of the aged are the three variables that are significant, the difference being that the coefficient for economy is no longer significant at the 1% significance level but instead at the 5% significance level. The coefficient for real per capita budgetary deficits dropped from 0.042 to 0.031, indicating that the influence of budgetary deficits has decreased since 2009. On the other hand, the magnitude of the coefficients for the economy and dependency rate of the aged both improved. A one-unit increase of economy would this time lead to 145 yuan more real per capita GHE, and increasing one percentage point of the dependency rate of the aged would mean 9 yuan more real per capita GHE.

Correcting for endogeneity

To test and control for endogeneity, lag one to lag two real per capita budgetary deficits were used as instruments in model 5 and lag one to lag two economy were used as instruments in model 6. The results are summarized in Table 7. Both models passed the under-identification, weak identification, and over-identification tests (statistics shown in the table), suggesting the appropriateness of the instruments.
Table 7

General method of moments (GMM) models

 

Model 5

Model 6

Instrumented variable

Real per capita budgetary deficits

Economy

Instruments

L(1/2). real per capita budgetary deficits

L(1/2).economy

Endogeneity test

p value = 0.6055

p value = 0.0161

Real per capita budgetary deficits

0.037*** (0.014)

0.034*** (0.008)

Economy

167.859*** (54.612)

248.886*** (44.731)

Openness

−12.046 (9.079)

−9.525 (9.356)

Industrial Structure

−26.148 (17.685)

−32.773** (14.302)

Unemployment rate

−1844.576

(1324.935)

−2493.487**

(1064.039)

Dependency rate of the aged

804.165*** (198.447)

1018.315*** (191.3554)

Province dummies

Included (29, Qinghai and Tibet excluded from data)

Included (29, Qinghai and Tibet excluded from data)

Time dummies

Included

Included

Under identification

p value = 0.0094

p value = 0.0045

Weak identification (Cragg-Donald Wald F statistic)

53.349

71.133

Over identification (Hansen J statistic)

p value = 0.4606

p value = 0.1140

N

145

145

R-squared

0.9412

0.9374

Notes: Standard errors are given in parentheses after the coefficients. The individual coefficient is statistically significant at the *10%, **5%, or ***1% significance level. The GMM estimates reported are all two-step results. The statistics are robust to heteroscedasticity and autocorrelation. The null hypothesis for the endogeneity test is that the specified endogenous regressor can be treated as exogenous. A significantly low p value suggests endogeneity. The null hypotheses for the under-identification test are that the model is under identified. For the weak identification test, the null hypothesis is that the model is weakly identified. Stock-Yogo weak ID test critical values are 19.93 for 10% maximal IV size, 11.59 for 15% maximal IV size; and 8.75 for 20% maximal IV size. The Cragg-Donald Wald F statistics for both model 5 and model 6 are significantly higher than the critical values, therefore rejecting the null hypothesis. The null hypothesis for the over-identification test is that the model is identified

The null hypothesis for the endogeneity test is that the instrumented variable is exogenous. As the p values indicate, the null hypothesis cannot be rejected in model 5 but can be rejected in model 6 at a 5% level of significance. In other words, the variable of real per capita budgetary deficits is likely to be exogeneous while the economy is likely to be endogenous. Additionally, the other two factors, openness and industrial structure, were also instrumented to test for endogeneity. Both tests failed to reject the null hypothesis.

Given that economy was identified as an endogenous variable, the results of model 6 are more reliable compared to the previous models because it corrected for endogeneity. Most results generated from fixed-effects models over the same period of time (model 2) still hold: real per capita budgetary deficits, economy, industrial structure, and dependency rate of the aged are key predictors of GHE; openness does not matter much.

There were several changes after the mitigation of endogeneity: first, the magnitude of coefficients for economy, industrial structure, and dependency rate of the aged increased. A one-unit increase in economy would be expected to increase the real per capita GHE by 249 yuan, in contrast to 125 yuan in model 2; a one-unit increase in industrial structure would lead to a decrease of 33 yuan in real per capita GHE, compared to 9 yuan in model 2. The coefficient for the dependency rate of the aged also increased from 770 to 1018. Second, the coefficient for the real per capita budgetary deficits decreased, from 0.042 to 0.034. Third, the coefficient for the unemployment rate became statistically significant. An increase of one percentage point of the unemployment rate would be associated with 25 yuan less real per capita GHE.

Discussion

Budgetary deficit increases GHE

This study discovered that increasing 1000 yuan real per capita budgetary deficits would lead to 34 yuan more real per capita GHE for the period between 2007 and 2013. This means that the less capable a province is in financing its expenditures from its own revenue, the more it spends on health, which is possible because these deficits should have been filled by central transfers to compensate for any gap. In other words, provinces with larger budgetary deficits receive more money from the central government, which is then spent on province-level GHE. In this, the central government subsidizes high deficit province health expenditures.

Nevertheless, there are several clues from the results to cast doubt on the effectiveness of this mechanism. To begin with, the magnitude of the coefficient for budgetary deficits decreased significantly from the period of 2007–2013 to 2009–2013. This probably shows that the marginal effect of the central government’s spending has diminished over time. In addition, when the endogeneity of economy was mitigated, the coefficient for budgetary deficits also decreased, meaning that the effect of budgetary deficits tends to be overestimated due to the noises posed by economic factors. These results appear to confirm some criticisms with central transfers in China, which argue that transfers were not properly targeted.

Economic development increases GHE even more

Given that economy and budgetary deficits are both key determinants of provincial GHE, it is interesting to make a comparison of the magnitude of the two coefficients. For economy, the average annual increase ranged from 0.11 to 0.29 across different provinces (Qinghai and Tibet excluded). The coefficient of 249 indicates that economy could explain 27- to 72-yuan average annual difference in real per capita GHE. In comparison, the average annual increase in real per capita budgetary deficits ranged from 118 to 192 yuan (Qinghai and Tibet excluded). The coefficient of 0.034 means that real per capita budgetary deficits could potentially explain 4- to 7-yuan increase in average annual difference in real per capita GHE. The influence of economic development is much more significant than that of budgetary deficits. This means that despite the central government’s efforts, GHE is still largely determined by economic development: richer provinces can enjoy a higher level of GHE. Moreover, a comparison of the coefficients for the overall period and the post-2009 period shows a worrisome picture. The magnitude of the coefficient for economy became even larger over time, suggesting that the impact of economic development was intensified.

Industrial structure, the other factor used to describe economic development, was also found to be statistically significant. Given that the industrial structure loaded most heavily on the percentage of GRP from the secondary industry (negative) and the sign of the coefficient for this factor is negative, the result means that provinces with heavier reliance on the secondary industry would have higher GHE. This is in accordance with the industrialization thesis discussed in the literature review. However, the magnitude of the coefficient for this factor is much smaller than that for economy, suggesting that the economy itself, rather than the structure of the economy, matters more for provincial GHE in China.

Openness is irrelevant

Across all regression models in this paper, none of the coefficients for openness is significant. In other words, openness fails to be a helpful predictor of GHE. This finding provides some empirical evidence that contradicts globalization arguments and justifies the exclusion of openness variables in previous studies of provincial GHE (Pan and Liu 2012). This result is also consistent with a study of health status and resources (Li and Wei 2010). In their paper, although the two authors found a positive relationship between FDI and health resources, they argued that it was because FDI largely determines and reflects local economic growth. In this study, openness was separated from the economy and composed as an individual factor, and the coefficient was indeed insignificant.

Conclusions

Since the mid-2000s, the Chinese government has asserted its role in the health sector, as manifested by the significant increase in GHE. One major goal of the reforms has been to curtail inequality across provinces, between urban and rural areas, and between different groups of people. This study mainly considers the first type of inequality, which is across provinces. As expected, the western region has benefited from the health reforms, with a high level of per capita GHE in many provinces between 2007 and 2013. On the other hand, provinces in the central region, which is also relatively economically underdeveloped, generally had a low level of GHE. This means that people in the central region may have experienced more financial difficulties when seeking medical services because they had neither a high level of government subsidy as in the western region nor a strong economy as in the coastal area.

Based on a panel dataset for Chinese provinces between 2007 and 2013, this study examined the key explanatory variables of provincial GHE. Different from previous research, a range of socio-economic indicators were factored before any regression analysis was performed to address the high correlation between variables, thereby ameliorating the measurement. The panel data regression analysis shows that real per capita budgetary deficits and economic development are the most important determinants of real per capita provincial GHE. More specifically, it was found that increasing 1000 yuan real per capita budgetary deficits would lead to 34 yuan more real per capita GHE; a one-unit increase in the economy was associated with 249 yuan higher real per capita GHE; and a one-unit increase in the industrial structure was expected to decrease the real per capita GHE by 33 yuan. A comparison of these coefficients showed that the influence of economic development is much larger than that of budgetary deficits. Furthermore, the comparison between the periods of 2007–2013 and 2009–2013 revealed an even influence of economic development and a diminishing effect of budgetary deficits, indicating that the central government’s funding efforts have probably diminished over time.

This study contributes to the existing literature by adding empirical evidence to three major theoretical debates in the field of welfare state development and public expenditure. It confirms the importance of economic development and the institutional structure but rejects the relevance of globalization in China’s GHE. The results also help us better understand the complicated budgeting process in China with a simple and straightforward message: economic development is the key to the availability of provincial GHE. Although central transfers also play a role, the impact is much smaller than that of economic development.

Footnotes
1

In China as well as other countries, the three key components of THE include GHE, social health insurance, and out-of-pocket payments. China’s classification is slightly different from the international practice. While the latter classifies subsidies to social health insurance under the category of social health insurance, China classifies the subsidies under GHE. Nevertheless, the difference in classification does not change the conclusion that GHE is expected to have the strongest equalizing effects among the three key components of THE.

 
2

The budget law was implemented between 1995 and 2014. A revised version was approved in 2014, and the new budget law was implemented in 2015. The provision for the local government autonomy remains the same. The new budget law introduced new procedures for budget preparation and approval, and budget reporting to the National People’s Congress was strengthened.

 
3

In other words, provincial GHE, the subject of this study, is actually an aggregate of health budgets of the provincial government itself plus all prefectural, county, and township governments in a given province.

 
4

The system of transfers is fragmented. Governments at lower levels can receive transfers from multiple higher level governments. For example, a county government can receive transfers from central, provincial, and prefectural governments.

 
5

\( \frac{1}{7}\times {\displaystyle {\sum}_{\mathrm{t}=2007}^{2013}}{X}_t \)

 
6

For this purpose, function xtivreg2 in Stata was used; xtivreg2 implements IV/GMM estimation of the fixed-effects and the first-differences panel data models with possibly endogenous regressors. It is essentially a wrapper for ivreg2, which was developed by Baum, Schaffer, and Stillman (2007).

 

Declarations

Acknowledgements

I would like to thank Dr. Leah Ruppanner at the University of Melbourne for proofreading the initial draft. All errors and omissions are my own.

Funding

This study is not supported by any funding.

Competing interests

The author declares that there are no competing interests.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

(1)
Asia Institute, The University of Melbourne

References

  1. Bartholomew DJ, Steele F, Moustaki I, Galbraith J (2008) Analysis of multivariate social science data, 2nd edn. CRC press, Boca RatonGoogle Scholar
  2. Baum CF, Shaffer ME, Stillman S (2007) Enhanced routines for instrumental variables/generalized method of moments estimation and testing. Stata J 7(4):465–506Google Scholar
  3. Brixi H, Mu Y, Targa B, Hipgrave D (2013) Engaging sub-national governments in addressing health equities: challenges and opportunities in China’s health system reform. Health Policy Plan 28:809–824View ArticleGoogle Scholar
  4. Cameron DR (1978) The expansion of the public economy: a comparative analysis. Am Polit Sci Rev 72(4):1243–1261View ArticleGoogle Scholar
  5. Checherita C, Rother P (2010) The impact of high and growing government debt on economic growth, European Central Bank Working Paper Series, 1237Google Scholar
  6. Party CC (2006) The resolutions of the CCP Central Committee on major issues regarding the building of a harmonious socialist society. CCP Document, BeijingGoogle Scholar
  7. Chinese Communist Party & State Council (2009) Opinions of the CPC Central Committee and the State Council on deepening the health care system reform. CCP Document, BeijingGoogle Scholar
  8. Chau R, Yu WK (2005) Is welfare unAsian? In: Walker A, Wong CK (eds) East Asian welfare regimes in transition: from confucianism to globalisation. Policy Press, Bristol, pp 21–45Google Scholar
  9. Chen Z (2009) Launch of the health-care reform plan in China. Lancet 373:1322–1324View ArticleGoogle Scholar
  10. Chou WL (2007) Explaining China’s regional health expenditures using LM-type unit root tests. J Health Econ 26:682–698View ArticleGoogle Scholar
  11. Chou WL, Wang Z (2009) Regional inequality in China’s health care expenditures. Health Econ 18:S137–146View ArticleGoogle Scholar
  12. Clayton R, Pontusson J (1998) Welfare-state retrenchment revisited: entitlement cuts, public sector restructuring, and inegalitarian trends in advanced capitalist societies. World Politics 51(1):67–98View ArticleGoogle Scholar
  13. Fang P, Dong S, Xiao J, Liu C, Feng X, Wang Y (2010) Regional inequality in health and its determinants: evidence from China. Health Policy 94:14–25View ArticleGoogle Scholar
  14. Huang C (2006) Government expenditures in China and Taiwan: do they follow Wagner’s law? J Econ Dev 31(2):139–148Google Scholar
  15. Immergut EM (1992) Health politics: interests and institutions in Western Europe. Cambridge University Press, New YorkGoogle Scholar
  16. Katzenstein PJ (1985) Small states in world markets: industrial policy in Europe. Cornell University Press, New YorkGoogle Scholar
  17. Li Y, Wei YHD (2010) A spatial-temporal analysis of health care and mortality inequalities in China. Eurasian Geography and Economics 51(6):767–787View ArticleGoogle Scholar
  18. Lin S, Song S (2002) Urban economic growth in China: theory and evidence. Urban Stud 39(12):2251–2266View ArticleGoogle Scholar
  19. Liu YL, Hsiao WC, Eggleston K (1999) Equity in health and health care: the Chinese experience. Soc Sci Med 49(10):1349–1356View ArticleGoogle Scholar
  20. Ministry of Health (2008) China public health statistical yearbook 2008. China Union Medical University Press, BeijingGoogle Scholar
  21. Mishra R (1999) Globalization and the welfare state. Edward Elgar, CheltenhamGoogle Scholar
  22. National Statistical Bureau of China (2008) China statistical yearbook 2008. China Statistics Press, BeijingGoogle Scholar
  23. National Statistical Bureau of China (2009) China statistical yearbook 2009. China Statistics Press, BeijingGoogle Scholar
  24. National Statistical Bureau of China (2010) China statistical yearbook 2010. China Statistics Press, BeijingGoogle Scholar
  25. National Statistical Bureau of China (2011) China statistical yearbook 2011. China Statistics Press, BeijingGoogle Scholar
  26. National Statistical Bureau of China (2012) China statistical yearbook 2012. China Statistics Press, BeijingGoogle Scholar
  27. National Statistical Bureau of China (2013) China statistical yearbook 2013. China Statistics Press, BeijingGoogle Scholar
  28. National Statistical Bureau of China (2014) China statistical yearbook 2014. China Statistics Press, BeijingGoogle Scholar
  29. National Statistical Bureau of China (2016) China statistical yearbook 2016. China Statistics Press, BeijingGoogle Scholar
  30. Obinger H, Leibfried S, Castles FG (eds) (2005) Federalism and the welfare state: New world and European experiences. Cambridge University Press, CambridgeGoogle Scholar
  31. Pan J, Liu GG (2012) The determinants of Chinese provincial government health expenditures: evidence from 2002-2006 data. Health Econ 21:757–777View ArticleGoogle Scholar
  32. Pierson P (1994) Dismantling the welfare state? Reagan, Thatcher, and the politics of retrenchment. Cambridge University Press, CambridgeView ArticleGoogle Scholar
  33. Pierson P (ed) (2001) The new politics of the welfare state. Oxford University Press, OxfordGoogle Scholar
  34. Rodrik D (1997) Has globalization gone too far? Institute for International Economics, WashingtonGoogle Scholar
  35. Rodrik D (1998) Why do more open economies have bigger governments? J Polit Econ 106(5):997–1032View ArticleGoogle Scholar
  36. Scharpf FW, Schmidt VA (eds) (2000) Welfare and work in the open economy: from vulnerability to competitiveness, vol Vol.1. Oxford University Press, OxfordGoogle Scholar
  37. Stock JH, Watson MM (2012) Introduction to econometrics, 3rd edn., Pearson EducationGoogle Scholar
  38. Swank D (2002) Global capital, political institutions, and policy change in developed welfare states. Cambridge University Press, CambridgeView ArticleGoogle Scholar
  39. Swank D (2010) Globalization. In: Castles FG, Leibfried S, Lewis J, Obinger H, Pierson C (eds) The Oxford Handbook of the Welfare State. Oxford University Press, OxfordGoogle Scholar
  40. Tang S, Bloom G (2000) Decentralizing rural health services: a case study in China. Int J Health Plann Manag 15(3):189–200View ArticleGoogle Scholar
  41. Tobin D (2005) Economic liberalization, the changing role of the state and “Wagner’s law”: China’s development experience since 1978. World Dev 33(5):729–743View ArticleGoogle Scholar
  42. Uchimura H, Jutting JP (2009) Fiscal decentralization, Chinese style: good for health outcomes? World Dev 37(12):1926–2934View ArticleGoogle Scholar
  43. Walker A, Wong CK (2005) Conclusion: from confucianism to globalisation. In: Walker A, Wong CK (eds) East Asian welfare regimes in transition: From Confucianism to Globalisation. Policy Press, Bristol, pp 213–224Google Scholar
  44. West LA, Wong C (1995) Fiscal decentralization and growing regional disparities in rural China: some evidence in the provision of social services. Oxf Rev Econ Policy 11(4):70–84View ArticleGoogle Scholar
  45. Wilensky HL, Lebeaux CN (1965) Industrial society and social welfare. Russell Sage, New YorkGoogle Scholar
  46. Wilensky HL (1975) The welfare state and equality. University of California Press, CaliforniaGoogle Scholar
  47. Wong C (2007) Can the retreat from equality be reversed? In: Shue V, Wong C (eds) Paying for progress in China: Public finance, human welfare and changing patterns of inequality. Routledge, New York, pp 12–28Google Scholar
  48. Wong C (2016) Budget reform in China: progress and prospects in the Xi Jinping era. OECD J Budg 15(3):27–36View ArticleGoogle Scholar
  49. Woo WT (2009) Assessing China’s capability to manage the high-probability risks to economic growth: fiscal, governance and ecological problems. In: Lee K, Kim J, Woo WT (eds) Power and the sustainability of the Chinese state. Routledge, London, pp 75–99Google Scholar
  50. World Bank (2002) China national development and sub-national finance: a review of provincial expenditures. The World Bank, WashingtonGoogle Scholar
  51. World Health Organization (2000) The world health report 2000: health systems, improving performance. World Health Organization, GenevaGoogle Scholar
  52. Wu AM, Lin M (2012) Determinants of government size: evidence from China. Public Choice 151(1-2):255–270View ArticleGoogle Scholar
  53. Yip W, Hsiao W (2009) China’s healthcare reform: a tentative assessment. China Econ Rev 20:613–9View ArticleGoogle Scholar
  54. Yip W, Hsiao W, Chen W, Hu S, Ma J, Maynard A (2012) Early appraisal of China’s huge and complex health care reforms. Lancet 379:833–42View ArticleGoogle Scholar
  55. Zhang X, Kanbur R (2005) Spatial inequality in education and health care in China. China Econ Rev 16(2):189–204View ArticleGoogle Scholar
  56. Zhao L, Huang Y (2010) China’s blueprint for health care reform. East Asian Policy 2(1):51–9Google Scholar

Copyright

© The Author(s). 2017

Advertisement