p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4

Available online at www.sciencedirect.com

Public Health

journal homepage: www.elsevier.com/puhe

Original Research

How does social development influence life

expectancy? A geographically weighted regression
analysis in China

J. Jiang a, L. Luo a, P. Xu a,b, P. Wang a,*

a
School of Health Sciences, Wuhan University, Wuhan, China
b
Institute of Sociology, Zhongnan University of Economics and Law, Wuhan, China

article info abstract

Article history: Objectives: This study aimed to examine how social development influenced life expectancy
Received 8 February 2018 from a spatio-temporal perspective.
Received in revised form Study design: Data were taken from the China Statistical Yearbooks for 2001 and 2011,
17 May 2018 which presented results from 2000 to 2010, respectively.
Accepted 4 June 2018 Methods: A geographically weighted regression method was performed. A total of 31
provinces were treated as units of our analysis, and four social development dimensions
were extracted by exploratory factor analysis.
Keywords: Results: Generally speaking, healthcare and education development could significantly
Social development improve life expectancy. Between 2000 and 2010, the positive effects of social harmony and
Life expectancy eco-environment development were gradually increasing. Spatially speaking, the
Geographically weighted regression enhancing effect of healthcare development on life expectancy generally declined from
China south to north China; the improving effect of education development generally declined
from southwest China to northeast and northwest China. The geographical distributions of
the effects of eco-environment and social harmony development on life expectancy were
different between 2000 and 2010; the former declined from the eastern coast to west China
in 2000 and declined from northwest to south China in 2010, while the latter first declined
from north to southwest China in 2000 and then declined from southwest to northeast
China in 2010.
Conclusions: Because the effects of various social development dimensions on life expec-
tancy vary with time and space, it is necessary to improve certain aspects of social
development in specific areas and social stages to efficiently improve life expectancy and
population health.
© 2018 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.

* Corresponding author. 115 Donghu Road, Wuchang District, Wuhan 430071, China. Tel.: þ86 13886001400.
E-mail addresses: jiang0111@whu.edu.cn (J. Jiang), 13006362970@163.com (L. Luo), pengxu@zuel.edu.cn (P. Xu), wpg926@whu.edu.cn,
wpg926@163.com (P. Wang).
https://doi.org/10.1016/j.puhe.2018.06.003
0033-3506/© 2018 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
96 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4
Introduction
Life expectancy is the average survival time of a specific
population with the same age in the future based on specific
mortality level,1 and it is an important index to evaluate
objective health. Although the life expectancy was 76.34 years
in 2015 in China,2 there is still a large gap between the life
expectancy in China and those of most other developed re-
gions or countries (e.g. life expectancy in Japan was 83.4 years
and in Australia was 81.9 years in 2016).3 In modern China, the
health of citizens is considered to be very important; the
Health China 2030 project further increases the awareness
and attention paid to the health of the population.4 In addi-
tion, as one of the most important indicators for health eval-
uation, life expectancy has received considerable attention
worldwide.5 Because social determinants have been treated as
the most important factors of health, life expectancy is closely
related to social development.6,7
However, in China, different areas have different levels of life
expectancy; developed regions, such as Beijing and Shanghai,
usually have a higher level of life expectancy than developing
regions. In comparison, the US citizens do not have a signifi-
cantly high level of life expectancy, despite its high economic
development; thus, many non-economic factors are also
important in determining life expectancy.8 Therefore, simply
estimating the relationship between social development and life
expectancy globally and on average cannot uncover the under-
lying association between them and the details of the mecha-
nism by which social development affects life expectancy. This
motivates us to discuss this issue in the present study.
Social development: definition and measurement
Social development is a multidimensional concept that in-
cludes various dimensions.9 In a broad sense, it contains
economy, politics, culture, and so forth, while it only, in a
narrow sense, includes the development of education,
healthcare, environment, and so on.10 In a narrow sense, so-
cial development is closely related to, but different from,
economic development.10
Most researchers have observed the multidimensional fea-
tures of social development and have widely discussed it,
mainly in a broad sense. Internationally, the Human Develop-
ment Index has been recognized as one of the most important
evaluation indices, which includes gross domestic product per
capita (economy), schooling year per capita (education), and life
expectancy (health).11 The Index of Social Progress is also a
well-known evaluation indicator system that contains econ-
omy, education, political participation, and healthcare devel-
opment.12 In addition, the United Nations and World Bank have
also established their own evaluation systems of social devel-
opment, but the former emphasizes too much on environ-
mental carrying capacity,13 while the latter ignores the social
stage and culture differences in various nations.14 In China,
social development evaluation research has been primarily
policy oriented. For example, many researchers have estab-
lished their evaluation system on the basis of various policies,
such as well-off society,15 harmonious society,16 two-oriented
society,17 and strategy of sustainable development.18
Although different studies have differences in dimensions
and indicator selection, most of them define social develop-
ment in a broad sense, containing economy, culture, education,
healthcare, environment, and so forth. In contrast to most
previous studies, we focused on social development in a narrow
sense, which was distinct from economic development.
Single effects of social development on life expectancy
The impact of social development on life expectancy has been
widely discussed by many researchers.7,19,20 However, social
development is a multidimensional concept9,18 and may have
large spatial and temporal disparities. In addition, various
aspects of social development may have distinct effect sizes
on life expectancy in different areas and periods.
The input and development of healthcare are directly
associated with population health. Before the Chinese eco-
nomic reform in 1978, a huge development of basic healthcare
greatly contributed to the improvement of life expectancy.20
After 1978, some researchers argued that owing to the un-
equal spatial distribution of healthcare resources and services
in China, there were some disparities in the spatial distribu-
tion of life expectancy.21 Ming and Dong22 reported that for life
expectancy promotion, it was not the increase in healthcare
provisions but the increase in the usage efficiency of the
healthcare services that mattered. However, in a trans-
national study in Europe, Heuvel et al.23 argued that health-
care investments had little effects on the promotion of life
expectancy, while investments in social protection could
greatly improve life expectancy, which also suggested the
indirect importance of social protection and harmony.
The development of education is a vital dimension of social
development that may also greatly influence life expectancy.
Most previous studies indicated that education development
could contribute greatly to life expectancy, but the marginal
enhancing effect was usually smaller than that of economic
development.24,25 Some people argued that the marginal
enhancing effect of education is the most important factor for
the promotion of life expectancy.26 Other evidence suggested
that there was a non-significant effect of education develop-
ment on life expectancy and other health indicators.27
The role of environmental development in life expectancy
promotion has gradually caught researchers' attention in
recent years. Based on the comparison of life expectancy values
between some developed nations and some postsocialist
countries, Feacham28 noted that the environmental disadvan-
tages produced many detrimental impacts on life expectancy
among the residents of postsocialist countries. In addition, in a
study conducted in 156 nations, Gulis26 also reported that a
better aquatic environment, more than improvements in
economy and education development, could most significantly
increase life expectancy. What is more, an investigation con-
ducted in China suggested that owing to its severe air pollution
status, residents in northern China would, on average, have a
life expectancy that was 5 years shorter than residents in
southern China.29
 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4
Spatial variations in the effect size of social development on
life expectancy
The study of spatial variations involves horizontal compari-
sons of the influencing factors of life expectancy among
various regions. Therefore, spatial variations can be uncovered
in comparison studies from the same region and also in com-
parisons between different studies conducted in various areas.
Evidence from Brazil, a developing nation, suggested that
education development could significantly enhance life ex-
pectancy, but the detrimental effect of income inequality on
life expectancy was weak and not significant.30 However, evi-
dence from Italy, a developed country, showed that the
harmful effect of income inequality on life expectancy required
more attention.31 Some researchers have argued that health-
care and education development are more important in
developing countries because they could directly use advanced
medical technologies from wealthier counties to improve their
population health.19 In addition, wars, resulting in social
instability, could negatively influence life expectancy.32,33 In
war zones with low levels of social stability and high levels of
violence,33 healthcare services have developed slowly. With the
detrimental effects of war on physical and mental health,32
coupled with increasing levels of infant mortality and dis-
eases,33 the positive effects of other factors on health might be
restricted and could not bring many health gains.
In China, some scholars found that the enhancing effects of
healthcare and education development on life expectancy
showed regional disparities, which showed a declining trend
from eastern China to western China.34 For a single dimension,
such as healthcare services, Liang et al.35 thought that disease
control and prevention strategies should be different in different
regions; the input of infectious disease prevention was more
likely to work in developing areas in China, and the input of
chronic disease prevention was more likely to benefit richer
areas in China. However, Wang36 argued that the spatial varia-
tions of the enhancing effects of healthcare and education
development on life expectancy were not statistically significant.
Temporal variations in the effect size of social development
on life expectancy
The study of temporal variations involves longitudinal ex-
aminations on the influencing factors of life expectancy in the
same area. Evidence from the Organization for Economic Co-
operation and Development (OECD) nations indicated that
the positive effect of education development on life expec-
tancy was significant at the end of the 20th century, but this
significant effect disappeared at the beginning of the 21st
century.37 Because OECD countries are a mixture of developed
and developing nations, we cannot further compare the
changing trends of different factors in various social contexts.
Interestingly, most longitudinal studies in China focused on
the increasing and decreasing effects of economic development
and eco-environment development on life expectancy, and they
obtained similar results: the significance of eco-environment
development was rising. For example, Ebenstein et al.7 found
that the negative effect of air pollution on life expectancy was
constantly significant from 1991 to 2012, and the marginal
detrimental effect of air pollution increased gradually. Other
97
Chinese researchers also concluded that with the development
of China's socio-economy, the enhancing effect of economic
factors on life expectancy would follow a marginally decreasing
trend,38 and after developing to a certain degree, the positive
effect of eco-environment development on life expectancy
would become greater.39
In summary, by comparing related studies for different
periods and regions or nations, such as the comparison be-
tween developed areas and developing areas, we could easily
find similarities and differences among these studies, which
might be related, more or less, to specific regional characteris-
tics. We could also see that there were still some disagreements
on the effect size of social development on life expectancy and
the effect of various social development dimensions on
life expectancy in different areas and periods. However, the
development of spatial analysis techniques in recent years has
laid a good foundation for the understanding of spatial re-
lationships. Nevertheless, in the research area of life expec-
tancy, the application of spatial analyses is mostly limited to
univariate analysis, such as spatial autocorrelation or hot spot
analysis,40 and those spatial regression models are not widely
used in related fields. Thus, only a few previous studies have
focused on the analysis or control of spatial relationships when
examining the influencing factors of specific health outcomes.6
Accordingly, this study introduced a geographically weighted
regression (GWR) model to further examine how social devel-
opment influenced life expectancy.
Methods
Data source
In this study, the provincial administrative unit was used as the
basic unit of spatial analysis, including 31 provincial-level
administrative regions in mainland China (4 municipalities,
22 provinces, and 5 autonomous regions were included; Hong
Kong, Macao, and Taiwan were excluded). All the data were
from the China Statistical Yearbooks (2001 and 2011).8,41 Data in
the China Statistical Yearbooks Database are all government
data from population censuses, socio-economic censuses, or
proportional sampling investigations, and the statistical caliber
of specific indexes is constant, which make the scores of a
specific index comparable to those from different time points.
Exploratory factor analysis and indicator selection
There were two main steps in the process of indicator selection.
First, we selected some important and representative indicators
according to previous studies and professional knowledge. In
the second stage, an exploratory factor analysis (EFA) and a
reliability analysis were conducted to select useful indicators.
We used the following three selection criteria: first, the Kaiser-
Meyer-Olkin (KMO) measure of sampling adequacy should be
larger than 0.70; second, the cumulative variance contribution
rate should be larger than 0.75 when the eigenvalue was larger
than one; and finally, each extracted communality should be
larger than 0.50. Accordingly, we extracted nine indicators from
15 original indicators and displayed more details in Table 1,
including four social development dimensions and the
98 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4

dependent variable. It should be noted that education, envi-
ronment, and social harmony development were measured by
outcome indicators, as we believed this would reflect their sta-
tus of development better. However, healthcare development
was measured by three input indicators, and, as life expectancy
was an outcome indicator of healthcare development, it was
improper to select other healthcare outcomes to predict it.
EFA was used to transfer many highly correlated variables
into fewer latent variables or factors based on the concept of
dimension reduction. In regression models, treating these
factors as independent variables could mostly eliminate the
collinearity and simplify models. In this study, four factors
were extracted, and each social development factor index we
obtained through EFA was used to reflect a specific aspect
of social development more exactly and comprehensively.
We used the following equations: the cumulative variance
PK
P lk
contribution rate Kk uk ¼ PNk1 , the score of each factor Fk ¼
l
n1 n
PN  mensions extracted from nine indicators, and εi was the random
j1 Mj Xij , and the score of social development
PK

k uk Fk $Xij was the standardized indictor; Mj was the score
coefficient of the factor; uk represented the weight of each
factor rotated or variance contribution rate of each factor
rotated; and lk was the eigenvalue of matrix R. This strategy
has been widely used in latent variable measurement or
development evaluation studies.42,43
Geographically weighted regression
spatial heterogeneity, we used GWR to examine the spatial
heterogeneity and the effects of various social development di-
mensions on life expectancy, and we analyzed the spatial dis-
tribution models of the effect sizes of various development
dimensions. The regression parameters in GWR were not for
global estimation but for partial estimation in subareas, and the
estimated coefficient for specific variables would vary with
spatial location; therefore, it could avoid the spatial autocorre-
lation problem and the overlook of local features in the ordinary
least square (OLS) regression.45 Accordingly, we established our
formula as follows:
X
LEi ¼ b0 ðui ; vi Þ þ bk ðui ; vi Þxk; i þ εi
 T 1
bk ðui ; vi Þ ¼ X Wðui ; vi ÞX XT Wðui ; vi ÞY
where LEi denoted the life expectancy in province i, ðui ; vi Þ was
the space coordinates of province i, xk; i represented the inde-
pendent variables, including four social development di-
F¼
disturbance term. X and Y denoted the matrices of independent
and dependent variables, respectively, Wðui ; vi Þ was the weight
matrix of GWR in province i, and the Gaussian kernel function
was used to calculate the inverse distance weighted, which
meant that larger distances would get smaller weights.
In the present study, IBM SPSS Statistics version 21.0 was used
in OLS regression and EFA methods, and the geographic infor-
mation systems package ArcGIS9.4 was used in GWR method.

GWR is a spatial regression method that can inlay the spatial Results
location of data into a regression parameter,44 and it is used to
examine non-stationary spatial relationships. Because most Exploratory factor analysis
variables would have a heterogeneous spatial distribution or
According to EFA, we obtained nine indicators, and the KMO
equaled 0.748, the total variance contribution rate equaled
Table 1 e Description of four social development 83.423%, and the Cronbach alpha coefficient was 0.843, which
dimensions and dependent variables. suggested that it was an acceptable result. The scores of the
Dimensions and variables 2000 2010 specific factors are provided in Table 2.
Mean SD Mean SD
General linear regression analysis
Healthcare development
Z1: no. of hospital beds 2.77 0.84 3.83 1.17
With the period dummy variable being controlled, the OLS
per 1000 persons
regression suggested that healthcare development and edu-
Z2: no. of doctors per 1000 persons 2.61 1.08 3.69 1.70
Z3: healthcare expenditure 51.58 45.60 429.24 188.74 cation development could significantly improve life expec-
per person (CNY) tancy (coef. ¼ 7.996, P < 0.001; coef. ¼ 11.290, P < 0.001;
Education development respectively), and the association between eco-environment
Z4: illiteracy rate [negative] (%) 8.27 5.69 4.84 4.30 development and life expectancy was not significant, neither
Z5: schooling years per capita 7.01 1.16 8.17 1.08 did the association between social harmony development and
Eco-environment development
life expectancy (see Table 3). Based on the standardized co-
Z6: emission load of SO2 and smoke 27.15 16.24 25.98 18.16
and dust per capita [negative] (kg)
efficients, the positive contribution of education development
Z7: forest coverage rate (%) 26.18 17.21 30.03 17.63 on life expectancy promotion (beta ¼ 0.612) was the largest,
Social harmony development followed by healthcare development (beta ¼ 0.464); however,
Z8: registered urban unemployment 3.17 0.71 3.61 0.60 the positive contributions of social harmony development and
rate [negative] (%) eco-environment development on life expectancy were both
Z9: crude divorce rate 1.11 0.59 2.05 0.87
relatively small (beta ¼ 0.150, beta ¼ 0.052, respectively).
[negative] (%)
Dependent variable Changes of the effect size of social development: 2000 vs
Life expectancy (years) 71.24 3.19 74.91 2.75
2010
CNY, China Yuan (the formal currency in China); SD, standard
deviation; SO2, sulfur dioxide. Table 4 provided the OLS results for both 2000 and 2010. Ac-
Data source: China Statistical Yearbooks 2001 and 2011.
cording to the comparison of these two periods, we concluded
 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4 99

Table 2 e Scores of four social development factorsa and life expectancy in 2000 and 2010 by province.
Province 2000 2010
F1 F2 F3 F4 LE F1 F2 F3 F4 LE
Beijing 0.364 0.526 0.589 1.153 76.10 0.922 0.454 0.904 0.995 80.18
Tianjin 0.141 0.486 0.257 0.712 74.91 0.465 0.439 0.559 0.518 78.89
Hebei 0.044 0.399 0.522 0.748 72.54 0.172 0.479 0.659 0.416 74.97
Shanxi 0.030 0.505 0.198 0.891 71.65 0.207 0.502 0.357 0.587 74.92
Inner Mongolia 0.002 0.414 0.386 0.603 69.87 0.171 0.575 0.210 0.332 74.44
Liaoning 0.091 0.579 0.557 0.503 73.34 0.264 0.643 0.644 0.396 76.38
Jilin 0.028 0.563 0.745 0.482 73.10 0.273 0.627 0.800 0.293 76.18
Heilongjiang 0.033 0.561 0.771 0.544 72.37 0.237 0.654 0.824 0.182 75.98
Shanghai 0.273 0.427 0.357 0.616 78.14 0.740 0.374 0.598 0.455 80.26
Jiangsu 0.005 0.335 0.515 0.707 73.91 0.245 0.412 0.598 0.598 76.63
Zhejiang 0.088 0.506 0.953 0.524 74.70 0.240 0.539 1.034 0.496 77.73
Anhui 0.047 0.280 0.740 0.649 71.85 0.145 0.343 0.782 0.443 75.08
Fujian 0.092 0.526 1.085 0.693 72.55 0.078 0.627 1.087 0.389 75.76
Jiangxi 0.128 0.526 0.995 0.672 68.95 0.040 0.575 1.045 0.474 74.33
Shandong 0.015 0.325 0.520 0.697 73.92 0.227 0.406 0.621 0.585 76.46
Henan 0.049 0.368 0.618 0.800 71.54 0.128 0.409 0.671 0.558 74.57
Hubei 0.010 0.390 0.727 0.617 71.08 0.175 0.474 0.811 0.350 74.87
Hunan 0.100 0.506 0.822 0.530 70.66 0.105 0.576 0.902 0.316 74.70
Guangdong 0.109 0.526 0.918 0.762 73.27 0.144 0.577 0.967 0.732 76.49
Guangxi 0.159 0.535 0.732 0.624 71.29 0.045 0.573 0.931 0.397 75.11
Hainan 0.034 0.435 1.000 0.663 72.92 0.148 0.480 1.090 0.608 76.30
Chongqing 0.059 0.420 0.552 0.510 71.73 0.149 0.599 0.700 0.192 75.70
Sichuan 0.078 0.410 0.680 0.448 71.20 0.173 0.478 0.808 0.265 74.75
Guizhou 0.151 0.282 0.407 0.485 65.96 0.059 0.390 0.638 0.379 71.10
Yunnan 0.050 0.316 0.852 0.718 65.49 0.143 0.420 0.980 0.309 69.54
Tibet 0.171 0.366 0.740 0.501 64.37 0.532 0.357 0.903 0.333 68.17
Shaanxi 0.045 0.447 0.667 0.741 70.07 0.193 0.510 0.797 0.407 74.68
Gansu 0.027 0.138 0.512 0.785 67.47 0.201 0.255 0.545 0.611 72.23
Qinghai 0.148 0.064 0.474 0.820 66.03 0.364 0.151 0.461 0.430 69.96
Ningxia 0.031 0.273 0.182 0.372 70.17 0.204 0.419 0.188 0.281 73.38
Xinjiang 0.165 0.404 0.411 0.439 67.41 0.438 0.511 0.345 0.368 72.35

LE, life expectancy.

a
F1 denotes healthcare development; F2 denotes education development; F3 denotes eco-environment development; F4 denotes social har-
mony development.

that different social development dimensions had distinct significant. In 2010, education development still contributed
impacts on life expectancy in different development phases. the most to the promotion of life expectancy (beta ¼ 0.732),
In 2000, education development had the greatest contribution followed by the contribution of healthcare development
to the promotion of life expectancy (beta ¼ 0.702), while the (beta ¼ 0.500), and the enhancing effect of social harmony
second greatest contribution was made by healthcare devel- development on life expectancy became significant statisti-
opment (beta ¼ 0.354), and the effects of eco-environment and cally (coef. ¼ 4.194, P ¼ 0.05). Although the effect of eco-
social harmony development on life expectancy were not environment development was still insignificant, its effect
size was larger than that in 2000 (1.244 > 0.169). From the
change in R2adj , we could also see that the explanatory power of
Table 3 e Ordinary least square estimation of social social development increased from 2000 to 2010, thus sug-
developmenta and life expectancy gesting that the role of social development in health promo-
Variables Coef. SE Beta t P tion was becoming increasingly vital.
Intercepts 231.225 177.610 1.302 0.198
F1 7.996 1.841 0.464 4.344 <0.001 Geographically weighted regression analysis: 2000 vs 2010
F2 11.290 1.449 0.612 7.790 <0.001
F3 0.747 1.115 0.052 0.670 0.506 Although three dimensions of social development signifi-
F4 2.732 1.721 0.150 1.588 0.118
cantly influenced life expectancy in the OLS regression anal-
Period dummy 0.148 0.089 0.214 1.670 0.101
ysis, we can see from Table 2 that the association between
variable
N ¼ 62, F ¼ 26.074, R2 ¼ 0.700, R2adj ¼ 0.673 healthcare development and life expectancy was more line-
arly significant when the level of healthcare development was
Coef., coefficient; SE, standard error; Beta, standard coefficient; t, t-
higher. Areas with higher levels of healthcare were usually
test score; P, P-value of t-test.
a
F1 denotes healthcare development; F2 denotes education eastern coastal areas; thus, the geographical location might
development; F3 denotes eco-environment development; F4 de- influence this process, and there might be spatial heteroge-
notes social harmony development. neity. Therefore, we used the GWR to examine whether there
100 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4
Table 4 e Comparison analysis of social developmenta and life expectancy: 2000 vs 2010.
Variables 2000
Coef. SE
Intercept 65.696*** 2.220
F1 9.315* 4.134
F2 12.038*** 2.407
F3 0.169 1.990
F4 1.089 2.885
N ¼ 31, F ¼ 7.256, R2adj ¼ 0.455
Coef., coefficient; SE, standard error; Beta, standard coefficient; *P < 0.05，**P < 0.01，***P < 0.001.
a
F1 denotes healthcare development; F2 denotes education development; F3 denotes eco-environment development; F4 denotes social har-
mony development.
was a spatial disparity in the effect sizes of different social
development dimensions on life expectancy.
The results of GWR in Fig. 1 indicated that the spatial vari-
ations of the enhancing effects of various social development
dimensions on life expectancy varied with time. Compared
with 2000, the spatial variation of the enhancing effect of
healthcare development on life expectancy decreased in 2010.
However, in both periods, the improving effect of healthcare
development on life expectancy declined from south China to
northwest and northeast China, where Tibet and Xinjiang
Province even displayed a negative effect of healthcare devel-
opment on life expectancy in 2000. The enhancing effect of
education development on life expectancy showed a tendency
of decline from southwest China to northwest and northeast
China in both 2000 and 2010, and the three northeast provinces
of China had the weakest promoting effect. The enhancing
effect of eco-environment development declined from the
eastern coast and northeast China to west and northwest
China in 2000, while it decreased from northwest and north
China to south China in 2010. The effect of social harmony
development on life expectancy seemed to be negative in most
provinces in 2000, except Xinjiang and Inner Mongolia, and this
negative effect in southwest China was larger than that seen in
north China. However, the general negative effect of social
harmony development became a positive effect across China in
2010, and it showed a trend of decline from west and southwest
China to northeast China, which reflects the social stability of
the western area during this period.
Discussion
Two innovations have been reported in this study. First, we
used EFA to extract four social development dimensions,
which could not only diminish or eliminate multicollinearity
among dependent variables but also make the analysis more
comprehensive. Second, we used GWR to analyze geograph-
ical variations, which took the spatial associations into
consideration. Accordingly, this study examined the effects of
healthcare, education, environment, and social harmony
development on life expectancy and confirmed the temporal
and spatial variations of these effects.
The healthcare and education development had profound
positive effects on life expectancy promotion. From 1949 to 1978,
the life expectancy promotion of Chinese citizens benefitted a lot
from the development of basic education and healthcare,20
2010
Beta Coef. SE Beta
65.565*** 1.482
0.354 7.040** 1.859 0.500
0.702 10.661*** 1.733 0.732
0.013 1.244 1.288 0.111
0.054 4.194* 2.042 0.254
N ¼ 31, F ¼ 13.851, R2adj ¼ 0.631
which also suggested that the most important factor of life ex-
pectancy promotion may not be economic development in
developing areas.19 Individuals with higher education are usually
less likely to be influenced by unhealthy lifestyles and more in-
clined to adopt healthy lifestyles, learn about health, and have
better living conditions; therefore, they can effectively prevent
the risks of diseases and death more in their daily life.22 The
development of healthcare, including the construction and input
of basic medical infrastructure and Medicare, directly ensures
the reduction of disease and death risks, resulting in greatly
increased life expectancy.
In the comparative analysis, the spatial distributions of the
effect of healthcare and education development on life expec-
tancy promotion were almost the same in 2000 and 2010, and
both of their enhancing effects on life expectancy displayed a
spatial model of ‘south high and north low’ globally. Because the
regional disparities in China are similar to the differences in
comparisons between developed and developing nations, this
spatial distribution model is basically in agreement with previ-
ous studies.19 On the one hand, the enhancing effect of health-
care development on life expectancy showed a decreasing trend
in south China. Healthcare developed slowly during this decade
in south China (see Table 2), which might be one reason
resulting in the decline of its enhancing effect on health. On the
other hand, the promoting effect of education development on
life expectancy demonstrated a growth trend in Tibet, and the
peak value of the enhancing effect appeared in southwest
China. In Table 2, we can see that the increasing rate of educa-
tion development was higher in west China and lower in east
China; therefore, we can deduce that its enhancing effect on life
expectancy benefited most from the long-term support of the
central government for improved basic education and some of
the helpful additive effects of the West Development Strategy.46
The total enhancing effect of eco-environment develop-
ment on life expectancy was relatively small. However, in the
comparative analysis, this enhancing effect showed an up-
ward trend from 2000 to 2010. Recently, there have been
increasingly more pollution events in China, especially air
pollution events.47 Studies on the mechanisms or impacts of
pollution on life expectancy or various diseases have been
increasingly abundant,29,48 which reflects the fact that envi-
ronmental health issues have drawn increased attention
during the process of social transition occurring in China.
However, the result provided a negative picture of eco-
environment development during this decade. With the
rapid development of economy in China, the restrictive effect
 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4 101

Fig. 1 e Spatial variation of estimated coefficients of social development on life expectancy: 2000 vs 2010. (a) Estimated
coefficients of healthcare development; (b) estimated coefficients of education development; (c) estimated coefficients of eco-
environment development; (d) estimated coefficients of social harmony development.

of the eco-environment has increasingly become larger;
therefore, the enhancing effect of eco-environment develop-
ment on life expectancy has become stronger. However, the
positive effect of eco-environment development on life ex-
pectancy displayed a spatial model of ‘southeast high and
north low’ in 2000 but ‘north high and south low’ in 2010.
Because the eco-environment in west China was more
damaged and fragile than that in east China, it had a stronger
restrictive effect on social development in 2010 according to
the cask effect.49 That is, it was the biggest weakness (eco-
environment development) that mostly limited the promotion
of life expectancy in west China. Additionally, according to the
diminishing marginal effect, the eco-environment developed
relatively slowly in west China, so the enhancing effect of eco-
environment development on life expectancy in west China
has become stronger than that in other areas in more recent
years. Certainly, the average effect of eco-environment
development on life expectancy was not so significant in
contemporary China; therefore, these results should be
treated carefully.
Based on the comparative analysis, we also found that the
enhancing effect of social harmony development on life ex-
pectancy increased significantly from 2000 to 2010. In an
advanced social stage, people are more likely to pursue high
social development goals, such as coordination or harmony;
therefore, social harmony or social stability that reflects high-
102 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4

Fig. 1 e (continued).

level goals can only have a significant effect on the life expec-
tancy of developed areas.23,30 Although China is still a devel-
oping country, the comparisons of the development level
between its east area and west area, and between 2000 and
2010, are similar to the comparison between developed and
developing countries/regions. Therefore, the average effect was
not significant in China. In fact, the positive effect of social
harmony development only emerged to be positive in Inner
Mongolia and Xinjiang in 2000. However, in 2010, the effect of
social harmony development on life expectancy became
significantly positive across China. These results do not
completely agree with previous comparison studies, which is
mainly due to the regional differences in China. For instance,
because of the impacts of national and regional characteristics,
Xinjiang is always the focus of stability maintenance in
China,50 and the improvement of social stability can be a strong
guarantee for social development; therefore, it can have a high
enhancing effect on health promotion.
Limitations
Although we tried to comprehensively analyze how social
development influenced life expectancy, this study still had
some limitations. Because of limited data availability and other
reasons, we defined social development in a narrow sense and
could not examine the effect of more social development factors
on life expectancy in detail. Furthermore, associations between
life expectancy and some dimensions of social development may
not be unidirectional;51 therefore, the causality between social
development and life expectancy needs to be further discussed.
 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4
Finally, the study period spans only 10 years, and some di-
mensions do not experience enough variation in this time frame;
therefore, a longitudinal analysis with a longer time period is
required.
Conclusions
Healthcare development, education development, and social
harmony development could significantly improve life expec-
tancy. According to the comparison between 2000 and 2010, the
enhancing effects of social harmony and eco-environment
development were gradually becoming stronger. Spatially
speaking, the enhancing effect of healthcare development on
life expectancy generally declined from south to north China;
the improving effect of education development generally
declined from southwest China to northeast and northwest
China. The geographical distributions of the effects of eco-
environment and social harmony development on life expec-
tancy were different in 2000 and 2010; the former declined from
eastern coast to west China in 2000 and declined from north-
west to south China in 2010, while the latter first declined from
north to southwest China in 2000 and declined from southwest
to northeast China in 2010.
Applications
Our findings can provide a scientific basis for formulating rele-
vant policies for improving life expectancy and public health.
We should take effective and targeted measures to improve the
level of healthcare, education, environment, and social har-
mony. For example, certain measures should be conducted to
improve the environment and social harmony in west China.
For south China, keeping the healthcare and education devel-
opment stable is important. Also, it is necessary to reduce these
disparities from a multidimensional perspective between
eastern and western China, thereby to efficiently improve the
overall life expectancy and population health in China.
Author statements
Ethical approval
Ethical approval was not required because of the public and
anonymous characteristics of macrodata.
Funding
This study was supported by ‘the Fundamental Research
Funds for the Central Universities’ (No. Whu2016026).
Competing interests
None declared.
references
1. Antonovsky A. Social class, life expectancy and overall
mortality. Milbank Mem Fund Q 1967;45:31e73.
103
2. China NBoSotPsRo. China statistical yearbook 2016. Beijing:
China Statistics Press; 2016.
3. Anonymity. The rank list of life expectancy of various
countries in 2016. http://doc.orz520.com/a/doc/2016/1109/
2042807.html2016.
4. Li B. The planning outline of Health China 2030 Project: a reference
book. Beijing: People's Medical Publishing House; 2017.
5. Klenk J, Keil U, Jaensch A, Christiansen MC, Nagel G. Changes
in life expectancy 1950-2010: contributions from age-and
disease-specific mortality in selected countries. Popul Health
Metrics 2016;14:1e11.
6. Zhao X, Wang W, Wan W. Regional inequalities of residents'
health level in China: 2003-2013. Acta Geograph Sin
2017;72:685e98.
7. Ebenstein A, Fan M, Greenstone M, He G, Yin P, Zhou M.
Growth, pollution, and life expectancy: China from 1991-2012.
Am Econ Rev 2015;105:226e31.
8. NBS P. China statistical yearbook 2011. Beijing: China Statistics
Press; 2011.
9. Kantiray A. On the measurement of certain aspects of social
development. Soc Indicat Res 1989;21:35e92.
10. Wang S. Relationship between social development and
economic development. China Population. Resour Environ
1995;5:13e7.
11. RLdC Monteiro, Pereira V, Costa HG. A multicriteria approach
to the human development index classification. Soc Indicat Res
2018;136:417e38.
12. Estes RJ. Social development trends in Asia, 1970-1994-the
challenges of a new century. Soc Indicat Res 1996;39:119e48.
13. Li T. Research progress in sustainable development indicator
systems both at home and abroad. Ecol Environ Sci 2013:1085e92.
14. Huang S. Criticism on sustainable development. Beijing: Higher
Education Press; 2001.
15. Song L. System and evaluation of China's well-off society
standards. Nanjing J Soc Sci 2010:6e14.
16. Tang Z. Research on construction and application of the
harmonious society index. 2014 [Doctoral Dissertation]. In press.
17. Gong S, Zhu H. Researches on the comprehensive evaluation
system and method of two-oriented society. Stat Decis
2009:14e6.
18. He W, Yang M, Yuan L. Ecological, economic and social
development evaluation of Wuling Mountain area using
entropy method. In: International conference on economics,
social science, arts, education and management engineering:
advances in social science education and humanities research;
2015. 670e2.
19. Ma S. Analysis on the factors of life expectancy. Popul Res
1989;13:14e8.
20. Sen A. Development as freedom. New York: Oxford University
Press; 1999.
21. Yang F. Equalization of medical and health services and life
expectancy. China Popul Newsl 2015 August 31st:1e2.
22. Ming Y, Dong Z. Life expectancy of China's population
analysis of the impact of factors. Theor Res 2010:47e50.
23. Heuvel Vd, A WJ, Marinela O. How important are health care
expenditures for life expectancy? A comparative, European
analysis. J Am Med Dir Assoc 2017;18:276. e9-.e12.
24. Chen C, Zhou T, Chen G. Analysis on influencing factors of
life expectancy. J Math Med 1997;10:71e2.
25. Ling CH, Ahmed K, Muhamad R, Shahbaz M, Loganathan N.
Testing the social cost of rapid economic development in
Malaysia: the effect of trade on life expectancy. Soc Indicat Res
2017;130:1005e23.
26. Gulis G. Life expectancy as an indicator of environmental
health. Eur J Epidemiol 2000;16:161e5.
27. Qi L. Interrelationship between growth, environment and
population health: an empirical analysis based on China's
provincial data. China Popul Resour Environ 2008;18:169e73.
104 p u b l i c h e a l t h 1 6 3 ( 2 0 1 8 ) 9 5 e1 0 4

28. Feachem R. Health decline in Eastern Europe. Nature

1994;367:313e4. Appendix
29. Chen Y, Ebensteinb A, Greenstone M, Li H. Evidence on the
impact of sustained exposure to air pollution on life
Table A1 e Eigenvalues and variances of four social
expectancy from China's Huai River policy. Proc Natl Acad Sci
development factorsa extracted (rotated).
USA 2013;110:12936e41.
30. Messias E. Income inequality, illiteracy rate, and life Factors Eigenvalue Variance (%) Cumulative variance (%)
expectancy in Brazil. J Publ Health 2003;93:1294e6. F1 2.939 32.657 32.657
31. Vogli RD, Mistry R, Gnesotto R, Cornia GA. Has the relation F2 2.171 24.125 56.782
between income inequality and life expectancy disappeared? F3 1.453 16.141 72.922
Evidence from Italy and top industrialised countries. J F4 1.286 14.294 87.217
Epidemiol Community 2005;59:158e62. a
F1 denotes healthcare development; F2 denotes education
32. Plümper T, Neumayer E. The unequal burden of war: the
development; F3 denotes eco-environment development; F4 de-
effect of armed conflict on the gender gap in life expectancy.
notes social harmony development.
Int Organ 2006;60:723e54.
33. Lindskog EE. The effect of war on infant mortality in the
Democratic Republic of Congo. BMC Publ Health 2016;16:e1059.
34. Liu C, Jian Y. Life expectancy, education capital and growth of
endogenous economy. Contemp Finance Econ 2011:15e25. Table A2 e Rotated factor loadings of indicators in four
35. Liang J, Zhai Y, Shi X. Changes and reasons of life expectancy social development factors.a
in developed nations and its enlightenment to China. Chin J Indicatorb F1 F2 F3 F4
Prev Contr Chron Dis 2013;21:491e4.
Z1 0.899 0.313 0.161 0.068
36. Wang S. Life expectancy decomposition and it's influencing
Z2 0.868 0.303 0.110 0.252
factors analysis: based on provincial panel data. NW Popul
Z3 0.838 0.067 0.129 0.280
2014;35:37e42.
Z9 0.512 0.471 0.125 0.456
37. Hossein Z, Gaskin DJ, Gerard A. Variations in life expectancy
Z4 0.124 0.959 0.044 0.031
in organization for economic co-operation and development
Z5 0.562 0.781 0.026 0.119
countries-1985-2010. Scand J Publ Health 2015;43:786e95.
Z6 0.100 0.232 0.871 0.175
38. Li P, Hu M. Analysis on influencing factors of life expectancy:
Z7 0.267 0.405 0.784 0.067
based on a quantitative analysis on provincial panel data
Z8 0.038 0.084 0.086 0.939
with dummy variables. Stat Consult 2017:29e33.
39. Lu X, Chen X. Factor on mean lifespan of each province in Note: Corresponding indices of each factor has been marked as
China. Yunnan Geograph Environ Res 2007;19:72e8. bold in the table.
a
40. Yang D, Wang X. Provincial variation and spatial F1 denotes healthcare development; F2 denotes education
autocorrelation of life expectancy in China. Soc Sci Front development; F3 denotes eco-environment development; F4 de-
2016:172e9. notes social harmony development.
b
41. NBS P. China statistical yearbook 2001. Beijing: China Statistics Z1: no. of hospital beds per 1000 persons; Z2: no. of doctors per
Press; 2001. 1000 persons; Z3: healthcare expenditure per person; Z4: illiter-
42. Zhang Y, Jiang J. Social capital and health in China: evidence acy rate (negative); Z5: schooling years per capita; Z6: emission
from the Chinese general social survey 2010. Soc Indicat Res load of sulfur dioxide and smoke and dust per capita (negative);
2018. in press. Z7: forest coverage rate; Z8: registered urban unemployment rate
43. Wang P, Xu C. Research on the comprehensive evaluation of (negative); Z9: crude divorce rate (negative).
China's environmental health, 2003-2010. Soc Indicat Res
2015;122:709e21.
44. Fotheringham A, Charlton M, Brunsdon C. The geography of Table A3 e Score coefficient matrix of nine indicatorsa
parameter space: an investigation of spatial non-stationarity. against the four social development factors.b
Int J Geogr Inf Syst 1996;10:605e27. Indicators F1 F2 F3 F4
45. Zhang Y, Ren Z. Factors affecting population distribution in
mountainous areas: geographically weighted regression Z1 0.311 0.023 0.049 0.071
using data from Bijie. Popul Res 2012;36:53e63. Z2 0.306 0.027 0.031 0.211
46. Liu S, Wang Y, Hu A. The effect of Western Development Z3 0.398 0.239 0.201 0.221
Program and regional economic convergence in China. Econ Z4 0.168 0.531 0.005 0.004
Res J 2009:94e105. Z5 0.063 0.324 0.012 0.087
47. Song C, Wu L, Xie Y, He J, Chen X, Wang T, et al. Air pollution Z6 0.186 0.205 0.632 0.070
in China: status and spatiotemporal variations. Environ Pollut Z7 0.140 0.265 0.524 0.130
2017;22:334e47. Z8 0.015 0.019 0.019 0.732
48. Luo L, Jiang J, Zhang G, Wang L, Wang Z, Yang J, et al. Stroke Z9 0.099 0.170 0.024 0.352
mortality attributable to ambient particulate matter Note: Corresponding indices of each factor has been marked as
pollution from 1990 to 2015 in China: an age-period-cohort bold in the table.
and spatial autocorrelation analysis. Int J Environ Res Publ a
Z1: no. of hospital beds per 1000 persons; Z2: no. of doctors per
Health 2017:772. 1000 persons; Z3: healthcare expenditure per person; Z4: illiter-
49. Cheng Y, Yang L, Li H. Global environmental change and acy rate (negative); Z5: schooling years per capita; Z6: emission
human health. Prog Geogr 2006;25:46e58. load of sulfur dioxide and smoke and dust per capita (negative);
50. Jia C. The infiltration and impacts of Pan-Turkism on Xinjiang Z7: forest coverage rate; Z8: registered urban unemployment rate
in China. World Ethno-national Stud 2011:2e12. (negative); Z9: crude divorce rate (negative).
51. Haruyama T, Ashiya M, Hamori S. Demographic changes, b
F1 denotes healthcare development; F2 denotes education
fiscal policy and economic growth: theory and evidence. Kobe development; F3 denotes eco-environment development; F4 de-
Univ Econ Rev 2013;59:1e23. notes social harmony development.