5.1 Stationary test and co-integration test
Whether the panel data is stable or has the same single integer order is an important prerequisite for regression analysis of panel data. To avoid “spurious regression” in empirical analysis, it is necessary to perform unit root test on each variable in the model. In this paper, LLC test and ADF test are used to test the unit root of panel data, and the test results are shown in Table 2. All variables have passed LLC test and ADF test in the first-order difference, and they are all first-order single integer I(1) sequences.
To further test whether there is a stable long-term relationship between residents’ health risks and explanatory variables, Kao test and Pedroni test are used to conduct co-integration tests on all variables. The test results are shown in Table 3 The explained variable of the co-integration test in the upper part of Table 3 is VI, and the explained variable in the lower part is DE. The results of most panel data co-integration tests show that there is a co-integration relationship between explanatory variables and explained variables, so there is a stable long-term relationship between residents’ health risks and explanatory variables.
Table 2
Variables
|
type
|
LLC
|
type
|
ADF
|
result
|
\(lnUR\)
|
I&T
|
-34.3646***
|
I&T
|
291.3890***
|
I(1)
|
\({lnUR}^{2}\)
|
I&T
|
-33.7509***
|
I&T
|
288.0890***
|
I(1)
|
\({lnUR}^{3}\)
|
I&T
|
-33.1518***
|
I&T
|
284.4940***
|
I(1)
|
\(lnVI\)
|
N
|
-2.1875**
|
I
|
89.5838**
|
I(1)
|
\(lnDE\)
|
I&T
|
-14.1434***
|
I&T
|
233.7290***
|
I(1)
|
\(lnGDP\)
|
I&T
|
-13.5449***
|
I&T
|
159.8890***
|
I(1)
|
\(lnME\)
|
N
|
-2.9992***
|
I
|
86.9923**
|
I(1)
|
\(lnED\)
|
I&T
|
-30.6588***
|
I&T
|
384.0530***
|
I(1)
|
\(lnEN\)
|
I&T
|
-12.9084***
|
I&T
|
155.5100***
|
I(1)
|
\(lnOR\)
|
I&T
|
-15.5315***
|
I&T
|
220.0280***
|
I(1)
|
Notes: “I&T”, “I” and “N” in the type indicate unit root test with both trend term and intercept term, only intercept term and neither trend term nor intercept term. The lag order is determined by SIC criterion, and *, **, *** respectively indicate the significant level of 10%, 5% and 1%. |
Table 3
method
|
type
|
|
statistic
|
Kao test
|
I
|
ADF
|
-7.3579***
|
Perdroni test
|
I&T
|
Panel v-Statistic
|
-2.7524
|
I&T
|
Panel rho-Statistic
|
8.2356
|
I&T
|
Panel PP-Statistic
|
-3.6605***
|
I&T
|
Panel ADF-Statistic
|
-4.4587***
|
I&T
|
Group rho-Statistic
|
10.1517
|
I&T
|
Group PP-Statistic
|
-9.6385***
|
I&T
|
Group ADF-Statistic
|
-5.7230***
|
Kao test
|
I
|
ADF
|
-3.8521***
|
Perdroni test
|
I&T
|
Panel v-Statistic
|
-6.5505
|
I&T
|
Panel rho-Statistic
|
7.2888
|
I&T
|
Panel PP-Statistic
|
-12.6850***
|
I&T
|
Panel ADF-Statistic
|
-7.7167***
|
I&T
|
Group rho-Statistic
|
8.9366
|
I&T
|
Group PP-Statistic
|
-21.7562***
|
I&T
|
Group ADF-Statistic
|
-7.9497***
|
Notes: “I&T”, “I” and “N” in the type indicate unit root test with both trend term and intercept term, only intercept term and neither trend term nor intercept term. The lag order is determined by SIC criterion, and *, **, *** respectively indicate the significant level of 10%, 5% and 1%. |
5.2 Estimation results of panel model at the national level
Table 4 reports the estimated results of UR on VI and DE in 31 provinces (autonomous regions and municipalities directly under the central government) in China during 2004–2019. Hausman test results show that the estimated results are suitable to be analyzed by the fixed effect model (p = 0), regardless of the health risks measured by VI or \(\)DE.
To study whether there is a nonlinear relationship between urbanization and residents’ health risks, models (2), (3) and (5) show the estimation results of quadratic term and cubic term of UR respectively. From the estimated results of models (3) and (5), it can be seen that there is a significant N-shaped relationship between UR and VI, while as far as China’s current urbanization level is concerned, there is a U-shaped relationship between UR\(\)and DE.
Urban development will cause changes in the living environment, lifestyle and income level of residents. On the one hand, the deterioration of living environment, the rise of living pressure and the change in working habits will make people suffer more external threats, resulting in an increase in health risks; On the other hand, due to the increase in income level and the popularity of health and medical insurance knowledge, people not only pay more attention to their physical health, but also gradually pay attention to their mental health. When attacked by diseases or in poor mental condition, they will take the initiative to choose hospitals for treatment. At the same time, urbanization has also brought more abundant medical facilities and resources, and the level of medical technology has been improved. In addition, residents take the initiative to seek medical treatment, which may reduce residents’ health risks.
From the viewpoint of VI:
(1) In the early stage, the negative impact of urbanization on the environment may not be obvious. At this time, the increase in VI is mainly due to the improvement of income level, the number of medical and health institutions and the level of medical technology. When UR is low, most areas lack medical infrastructure, and the level of medical technology is low. Residents may encounter difficulties or disbelief in medical treatment so that VI is low. The increase of medical facilities and resources availability and the improvement of medical technology level gradually result in a rapid increase in VI.
(2) With the in-depth development of urbanization, although the negative impact of urbanization on the environment has increased, and pollutants such as industrial smoke dust and sulfur dioxide have harmed residents’ health. Since urbanization has enabled a large number of people to seek medical treatment in the early stage, the residents attending the doctor’s treatment at this time are mainly caused by the deterioration of their living environment, thus reducing VI overall.
(3) When urbanization develops to a higher level, on the one hand, the deterioration of living environment continues to threaten residents’ health; on the other hand, residents also pay more attention to their physical and mental health, and take more proactive medical care. As a result, VI has continued increasing.
From the viewpoint of DE:
(1) In the early stage of urbanization, the effect of environmental factors on residents’ health is not obvious. With the increase of income level, the abundance of medical resources and the improvement of medical technology, residents tend to take the initiative to seek medical treatment, so DE has reduced.
(2) In the middle period of urbanization, the environmental pollution problem is becoming more and more serious, and the living pressure is also increasing. The probability of residents suffering from illness, especially serious illness, has increased, resulting in an increase in DE.
Table 4
The impact of urbanization on residents’ health risks
Variables
|
\(lnVI\)
|
\(lnDE\)
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
\(lnUR\)
|
0.1165***
|
-2.3816***
|
40.1438***
|
-0.0787**
|
-1.3277***
|
|
(2.86)
|
(-5.60)
|
(7.00)
|
(-2.03)
|
(-3.19)
|
\({lnUR}^{2}\)
|
|
0.3449***
|
-11.2227***
|
|
0.1724***
|
|
|
(5.90)
|
(-7.34)
|
|
(3.01)
|
\({lnUR}^{3}\)
|
|
|
1.0436***
|
|
|
|
|
|
(7.70)
|
|
|
\(lnGDP\)
|
-0.0686**
|
-0.0536*
|
0.0263*
|
0.0662**
|
0.0708**
|
|
(-2.03)
|
(-1.82)
|
(1.82)
|
(2.04)
|
(2.20)
|
\(lnME\)
|
0.3264***
|
0.3338***
|
0.5047***
|
0.0418*
|
0.0455*
|
|
(12.40)
|
(13.13)
|
(24.97)
|
(1.67)
|
(1.83)
|
\(lnED\)
|
-0.0746***
|
-0.0597**
|
-0.0076
|
-0.1361***
|
-0.1287***
|
|
(-2.63)
|
(-2.17)
|
(-0.50)
|
(-5.03)
|
(-4.78)
|
\(lnEN\)
|
-0.0351***
|
-0.0325***
|
-0.0299***
|
-0.0033
|
-0.0021
|
|
(-3.57)
|
(-3.44)
|
(-4.13)
|
(-0.31)
|
(-0.18)
|
\(lnOR\)
|
-0.0408
|
-0.0402
|
0.0486*
|
0.2241***
|
0.2244***
|
|
(-1.44)
|
(-1.47)
|
(1.73)
|
(8.32)
|
(8.41)
|
\(Constant\)
|
0.5443*
|
4.7860***
|
-49.173***
|
1.8949***
|
4.0156***
|
|
(1.67)
|
(6.10)
|
(-6.85)
|
(6.11)
|
(5.23)
|
\(Observations\)
|
496
|
496
|
496
|
496
|
496
|
\(R-squared\)
|
0.953
|
0.957
|
0.940
|
0.301
|
0.315
|
\(Province FE\)
|
YES
|
YES
|
YES
|
YES
|
YES
|
\(Year FE\)
|
YES
|
YES
|
|
YES
|
YES
|
\(Hausman\)
|
P = 0.0000
|
P = 0.0000
|
P = 0.0000
|
P = 0.0000
|
P = 0.0000
|
Notes: *, **, *** respectively indicate the significant level of 10%, 5% and 1%. |
(3) Theoretically, when urbanization develops at a higher stage, urban will form a situation in which people and nature coexist in harmony. As a country pursues economic growth, governance also gradually enhances the awareness of environmental protection. Residents’ living environment will be improved and the probability of illness can be reduced. While from the data of China’s national level, the equation after adding the cubic term is not significant, indicating that the urbanization level of most provinces has not yet reached this stage. And there is still a U-shaped relationship between UR and DE, which is also confirmed by the results of regional analysis later.
Other variables have different effects on residents’ health. GDP is positively correlated with VI and DE. Areas with higher income level are often areas with serious environmental deterioration and great living pressure, and residents have higher risk of illness. ME\(\)has a positive correlation with VI and DE, but it has less influence on DE. The effect of ED on VI is not significant, but it has a significant negative relationship with DE. The improvement of education level will help residents to acquire more knowledge about health and disease prevention, increase residents’ attention to health, and effectively reduce DE. OR is positively correlated with VI and DE. As the aging problem of China’s population intensifies, more attention should be paid to the health problems of the elderly. We should actively promote the construction of old-age welfare undertakings, improve the health level of the whole people, and reduce the possible negative impact of aging problem on the construction of “Healthy China”.
To clearly show the relationship between urbanization and residents’ health risks, the fitting curves of UR to VI and DE are respectively shown in Fig. 4.
From Fig. 4, it is obvious that:
(1) UR has an N-shaped relationship with VI, with inflection points of 3.423 and 3.745 respectively. That means when UR is below about 30.66%, VI increases with the increase of UR. When UR is between 30.66% and 42.31%, VI decreases with the increase of UR. When the UR exceeds about 42.31%, the increase of UR will continue to promote the increase of VI.
(2) There is a U-shaped relationship between UR and DE, with an inflection point of 3.841. When UR exceeds about 46.57%, the increase of UR will increase DE.
5.4 Heterogeneity test
5.4.1 Heterogeneity of grouping by region
China has a vast territory, and the level of development in different regions is obviously unbalanced. The level of urbanization and medical infrastructure of different regions are quite different. It is also unreasonable to require different regions to adopt the same development mode and environmental supervision policies. To analyze the regional differences of urbanization on residents’ health risks, China is divided into three regions: eastern, central and western regions, to estimate the differences of impacts of urbanization on residents’ health risks.
When residents’ health risks are measured by VI, the STIRPAT model (3) mentioned above is used in the eastern, central and western regions. When DE is used to measure residents’ health risks, the STIRPAT model (2) mentioned above is used in the central and western regions, and the STIRPAT model (2) and model (3) mentioned above are used in the eastern region.
Table 7 reports the regional differences of the impact of urbanization on residents’ health risks. According to models (1)-(3): (1) The coefficients of core variables in the eastern, central and western regions are significant, indicating that there is an N-type relationship between UR and VI, but the impact in the eastern region is obviously stronger than that in the central and western regions. The possible reasons are: compared with the central and western regions, the eastern region has a better economic foundation and marketization degree, richer medical resources, better public health services. Residents have higher human capital and more emphasis on their physical and mental health. At the same time, as the living and working pressure of cities in the eastern region is relatively high, the impact of urbanization on residents’ health risks in the eastern region will be stronger;
(2) In central and western regions, urban medical resources are relatively scarce compared with the eastern regions, and the public infrastructure still needs to be improved, and the supply of educational and medical resources lags behind the speed of population gathering in the cities, resulting in a smaller effect of UR on VI than that in the east.
According to models (4)-(7): the impact of UR on DE in different areas is quite different. (1) After the quadratic term and cubic term of urbanization rate were added to the eastern region, it is significant, which indicates that there is an inverted N-type relationship between UR and DE in the eastern region. It also means there is an optimization turning point between UR and DE in the eastern region, and DE would be reduced when UR developed to a higher level. With the continuous development of urbanization, the industrial structure in the eastern region is gradually transformed and upgraded, and the proportion of tertiary industry in the three industries is increasing. The negative externality of environment brought by economic growth are decreasing, and the environmental problems are improving day by day. The improvement of urbanization has a positive impact on residents’ health by increasing residents’ income, medical technology and service level, and improving public facilities, thus reducing DE.
Table 7
Health impact of urbanization based on regional groups
Variables
|
\(lnVI\)
|
\(lnDE\)
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
Eastern
|
Central
|
Western
|
Eastern
|
Eastern
|
Central
|
Western
|
\(lnUR\)
|
102.2696***
|
76.4811**
|
34.6141***
|
-2.9353*
|
-73.2417***
|
1.6403
|
-0.7348
|
|
(3.10)
|
(2.52)
|
(3.86)
|
(-1.95)
|
(-2.73)
|
(1.17)
|
(-1.42)
|
\({lnUR}^{2}\)
|
-26.7411***
|
-20.3931**
|
-9.8523***
|
0.3713*
|
18.0225***
|
-0.2821
|
0.0985
|
|
(-3.24)
|
(-2.53)
|
(-4.05)
|
(1.93)
|
(2.68)
|
(-1.45)
|
(1.36)
|
\({lnUR}^{3}\)
|
2.3328***
|
1.8093**
|
0.9319***
|
|
-1.4757***
|
|
|
|
(3.38)
|
(2.55)
|
(4.24)
|
|
(-2.62)
|
|
|
\(lnGDP\)
|
0.1018***
|
-0.0228
|
0.0103
|
0.0195
|
0.0421
|
0.1060***
|
-0.0420***
|
|
(2.72)
|
(-0.74)
|
(0.55)
|
(0.65)
|
(1.38)
|
(3.34)
|
(-2.78)
|
\(lnME\)
|
0.3930***
|
0.5601***
|
0.5587***
|
0.0002
|
-0.0329
|
0.0009
|
-0.0007
|
|
(8.57)
|
(15.98)
|
(20.28)
|
(0.01)
|
(-0.88)
|
(0.03)
|
(-0.04)
|
\(lnED\)
|
0.0065
|
0.0054
|
-0.0171
|
-0.0341
|
-0.0299
|
-0.0307
|
0.0234
|
|
(0.21)
|
(0.18)
|
(-0.81)
|
(-1.37)
|
(-1.22)
|
(-1.05)
|
(1.48)
|
\(lnEN\)
|
-0.0155
|
-0.0512***
|
-0.0194
|
0.0356***
|
0.0268**
|
-0.0166
|
-0.0280***
|
|
(-1.16)
|
(-2.99)
|
(-1.57)
|
(3.35)
|
(2.43)
|
(-0.94)
|
(-2.92)
|
\(lnOR\)
|
0.0669
|
0.1635**
|
0.0554
|
0.2452***
|
0.2326***
|
0.2891***
|
0.1445***
|
|
(1.41)
|
(2.40)
|
(1.11)
|
(6.28)
|
(6.01)
|
(4.64)
|
(3.63)
|
\(Constant\)
|
-132.4125***
|
-96.9736**
|
-41.8443***
|
6.8451**
|
100.0532***
|
-1.7811
|
3.1433***
|
|
(-3.03)
|
(-2.55)
|
(-3.82)
|
(2.28)
|
(2.80)
|
(-0.66)
|
(3.26)
|
\(Observations\)
|
176
|
128
|
192
|
176
|
176
|
128
|
192
|
\(R-squared\)
|
0.926
|
0.959
|
0.955
|
0.247
|
0.279
|
0.439
|
0.209
|
\(Province FE\)
|
YES
|
YES
|
YES
|
YES
|
YES
|
YES
|
|
\(Hausman\)
|
P = 0.0000
|
P = 0.0000
|
P = 0.0000
|
P = 0.0004
|
P = 0.0270
|
P = 0.0000
|
P = 0.5836
|
Notes: *, **, *** respectively indicate the significant level of 10%, 5% and 1%. |
(2) For the central and western regions, there is no significant relationship between UR and DE. Only when urbanization develops to a certain level and can have a great impact on the living environment and infrastructure of residents, its impact on residents’ health risks will be obvious.
5.4.2 Heterogeneity according to the level of infrastructure
In the process of urbanization, the phenomenon of lagging infrastructure often occurs. The development of urban will not have a direct impact on residents’ health, the improvement of infrastructure is an important way for urbanization to affect residents’ health. If the infrastructure lags behind the development of urbanization, the impact of urbanization on residents’ health may not be obvious or even counterproductive.
Table 8
Health impact of urbanization based on infrastructure level groups
Variables
|
\(lnVI\)
|
\(lnDE\)
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
Low
|
Medium
|
High
|
Low
|
Medium
|
High
|
\(lnUR\)
|
-18.9061
|
54.3760***
|
72.3364***
|
-0.5459
|
-0.8546
|
-2.4687***
|
|
(-1.48)
|
(5.73)
|
(6.53)
|
(-0.56)
|
(-0.83)
|
(-3.25)
|
\({lnUR}^{2}\)
|
4.5268
|
-14.9825***
|
-20.1002***
|
0.0964
|
0.0995
|
0.3493***
|
|
(1.34)
|
(-5.91)
|
(-6.65)
|
(0.70)
|
(0.70)
|
(3.40)
|
\({lnUR}^{3}\)
|
-0.3498
|
1.3729***
|
1.8505***
|
|
|
|
|
(-1.17)
|
(6.10)
|
(6.77)
|
|
|
|
\(lnGDP\)
|
0.0254
|
0.1788***
|
0.0343
|
-0.0133
|
-0.0340
|
-0.0762***
|
|
(0.99)
|
(6.35)
|
(1.16)
|
(-0.56)
|
(-1.20)
|
(-2.77)
|
\(lnME\)
|
0.4551***
|
0.2767***
|
0.5371***
|
-0.0382
|
0.1150**
|
0.0375
|
|
(13.38)
|
(4.59)
|
(14.67)
|
(-1.21)
|
(2.02)
|
(1.12)
|
\(lnED\)
|
-0.0577**
|
0.0497**
|
-0.0159
|
-0.0050
|
-0.0155
|
-0.0095
|
|
(-2.37)
|
(2.56)
|
(-0.56)
|
(-0.22)
|
(-0.77)
|
(-0.37)
|
\(lnEN\)
|
0.0105
|
-0.0175
|
-0.0152
|
0.0038
|
0.0078
|
-0.0020
|
|
(0.78)
|
(-1.54)
|
(-1.10)
|
(0.30)
|
(0.70)
|
(-0.17)
|
\(lnOR\)
|
0.0928*
|
0.0018
|
0.0672
|
0.1135**
|
0.1087**
|
0.1777***
|
|
(1.71)
|
(0.04)
|
(1.39)
|
(2.23)
|
(2.14)
|
(4.09)
|
\(Constant\)
|
24.9788
|
-68.0380***
|
-88.3311***
|
2.4684
|
3.3082
|
6.2436***
|
|
(1.55)
|
(-5.73)
|
(-6.52)
|
(1.37)
|
(1.64)
|
(4.34)
|
\(Observations\)
|
166
|
165
|
165
|
166
|
165
|
165
|
\(R-squared\)
|
0.935
|
0.907
|
0.923
|
0.087
|
0.117
|
0.309
|
\(Province FE\)
|
YES
|
YES
|
YES
|
YES
|
YES
|
YES
|
\(Hausman\)
|
P = 0.0000
|
P = 0.0042
|
P = 0.0000
|
P = 0.0009
|
P = 0.0299
|
P = 0.0007
|
Notes: *, **, *** respectively indicate the significant level of 10%, 5% and 1%. |
The sample was divided equally into three groups according to the proportion of medical and health personnel in the total population, in order to study the differences in the impact of different level of infrastructure on residents’ health risks while urbanization is developing. When residents’ health risks are measured by VI, STIRPAT model (3) mentioned above is used in different levels of areas; When DE is used to measure residents’ health risks, STIRPAT model (2) is used.
In Table 8: the impact of UR on VI is significant only when the infrastructure is at a medium to high level, with an N-shaped relationship between the two, while the impact of UR on DE only significant when the infrastructure is at a high level, with a U-shaped relationship between the two.
That means urbanization will only have a significant impact on residents’ health if it achieves simultaneous improvement of infrastructure. With the deepening of urbanization, if the improvement of medical facilities is backward and the improvement of medical technology is slow, even if people have higher income levels and higher self-health awareness, the problem of difficulty in seeing and treating patients may still occur, thus threatening the residents’ health.