mortality trends and variation in china€¦ · mortality trends: rural/urban gap (unadjusted e 0...
TRANSCRIPT
Mortality Trends and Variation in China
Yong Cai [email protected]
University of North Carolina, Chapel Hill
May 15, 2019
Harnack Haus, Berlin
Satellite meeting
Mortality data and methodological approaches in
estimating mortality in developing countries
Sponsored by Max Planck Institute for Demographic Research
Background
Data sources & challenges
Mortality trends and variation
Overview
Economic growth
Changes in nutrition and consumption
Social transformation
Individual awareness and life style
Establishment of “new” social welfare system
Healthcare and delivery
Public health investment
Socioeconomic background
Main sources
Population census and surveys
Health surveillance systems
Death registration
Household registration (Hukou)
Social pension system
Main challenges
Data quality
Underreporting
Migration
Government evaluation
Data availability
Segmentation
Undocumented adjustment
Data sources and challenges
Chinese data: good and bad
Mortality trends: e0t
70
72
74
76
78
80
82
1995 2000 2005 2010 2015 2020
Unadjusted Survey
NBS Official
WPP2017
Mortality trends: qxt
0.0001
0.001
0.01
0.1
1
0 20 40 60 80
q2000 q2005 q2006
q2007 q2008 q2009
q2010 q2011 q2012
q2013 q2014 q2015
q2016 q2017
Mortality trends: lgt(lxt) compared to 2000
-2
-1
0
1
2
3
4
5
6
-3 -2 -1 0 1 2 3 4
2005 2006 2007 2008 2009 2010 2011
2012 2013 2014 2015 2016 2017
Mortality trends: lgt(lxt) compared to 2000
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0
0.2
0.4
0.6
0.8
1
1.2
1.4
2004 2006 2008 2010 2012 2014 2016 2018
BET
A
beta alpha Linear (beta) Linear (alpha )
Mortality trends: gender gap e0f - e0
m
0
1
2
3
4
5
6
7
1995 2000 2005 2010 2015 2020
Unadjusted Survey
NBS Official
WPP2017
Morality trends: Shanghai e0
70
73
76
79
82
85
88
1980 1985 1990 1995 2000 2005 2010 2015
HK e0f HK e0m
Morality trends: Shanghai e0
70
73
76
79
82
85
88
1980 1985 1990 1995 2000 2005 2010 2015
HK e0f HK e0m
NBS e0f NBS e0m
Survey e0m Survey e0f
Mortality trends: infant mortality rate
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
Surveillance Adjusted Population Survey Surveillance Raw
Rising life expectancy fits well with the general pattern of mortality decline as it was driven by multiple factors happening in China: socioeconomic development, investment in public health, change of life style…
Mortality underreporting is a constant problem, and will remain to be.
While overall mortality decline can be collaborated with other data sources, ascertain of mortality level requires careful cross-examination with other population data.
Mortality trends: basic observations
Mortality trends by rural/urban (unadjusted e0t)
70.0
72.0
74.0
76.0
78.0
80.0
82.0
84.0
86.0
88.0
2000 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
cities towns villages
Mortality trends: rural/urban gap (unadjusted e0t)
comparing to e0t for villages
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
towns
cities
Regional trends and variation, e0t
65
70
75
80
85Xizang
YunnanGuizhou
Qinghai
Xinjinag
Gansu
Jiangxi
Neime…
Shaanxi
Ningxia
Hunan
Hubei
Sichuan
Guangxi
ChinaHenan
ShanxiChong…
Anhui
Helong…
Hebei
Fujian
Hainan
Jilin
Guang…
Liaoning
Jiangsu
Shang…
Zhejiang
Tianjin
BeijingShang… NBS2000
Census2000
Census2005
Census2010
Census2015
Mortality variation: e0 (2000, unadjusted)
Percentile
Variable Mean S.D. Min 5 Med 95 Max
e0m 69.9 3.9 46.4 62.9 70.5 75.3 86.9
e0f 73.8 4.7 45.3 65.4 74.5 80.0 106.7
1q0m 25.7 23.4 0.0 4.6 22.9 91.1 376.7
1q0f 32.8 32.0 0.0 4.9 18.9 68.6 307.3
Mortality variation: county level e0 (2000, raw)
Brass (1971) relational model: linear relationship between lx’s of an observed population and of a standard population at logit scale
Consistency of mortality across different age groups
Choice of standard life table: “must be some kind of average Brass (1971)”. We use provincial average.
)()(s
xx llogitllogit
Brass Logit Model Smoothing
Empirical Bayes Estimates
Empirical Bayes Estimates (Efron and Morris 1973, 1975; Clayton and Kaldor 1987; Marshall 1991): pool information across areas to produce more stable and robust estimates using the Empirical Bayes methods.
Information from neighboring units serves as priori for the local estimate.
)(ˆ2
2
ni
i
nn
n
ni mm
n
m
n
mS
n
mS
m
Empirical Bayes Estimates
14.44
6.46
14.61
7.99
15.65
14.696.90
11.02
14.693.50 10.06
11.16
11.33
9.96
12.77
14.6314.82
13.89
15.65
4.30
16.42
6.84
22.62
15.61 6.3111.12
15.412.70 9.83
11.16
10.32
7.33
14.21
15.0715.22
14.68
Percentile
Variable Mean S.D. Min 5 Med 95 Max
e0m 69.8 3.7 46.6 62.9 70.5 74.5 79.3
e0f 73.4 4.2 45.5 65.3 74.3 78.6 84.0
1q0m 26.0 22.1 0.9 6.1 23.8 93.0 355.2
1q0f 33.6 31.3 1.2 6.6 19.2 66.9 267.7
Mortality variation: county level e0 (2000 smoothed)
Mortality variation: county level e0t (2000)
Legend46 - 6060 - 65
65 - 67.567.5 - 7070 - 72.572.5 - 7575 - 80.8
Mortality variation: county level e0f clusters (2000)
cluster, sig
low, <.10low, <.05low, <.01high, <.01high, <.05high, <.10not sig.
PredictorMale Model Female Model Difference
coef std beta coef std beta (Male - Female)
(Intercept) 61.526 1.9371 *** 64.099 2.3035 *** -2.573
GDP 0.6292 0.1119 0.1205*** 1.0133 0.1331 0.1676*** -0.384*
Urbanization 0.0046 0.0042 0.0296 -0.0014 0.0050 -0.0080 0.006
Employment 0.0179 0.0089 0.0435* 0.0312 0.0106 0.0654** -0.013
Health Care -0.6532 0.2782 -0.0518* -0.4184 0.3308 -0.0286 -0.235
Water 0.0044 0.0026 0.0337 0.0040 0.0031 0.0265 0.000
Lavatory 0.0058 0.0022 0.0448** 0.0069 0.0026 0.0458** -0.001
Bath Facility 0.0121 0.0042 0.0659** 0.0162 0.0050 0.0761** -0.004
Education 0.9515 0.0982 0.3524*** 1.0719 0.1167 0.3428*** -0.120
Discrimination 0.0465 0.1629 0.0060 0.0353 0.1937 0.0039 0.011
Ethnicity -0.0253 0.0029 -0.2152*** -0.0151 0.0035 -0.1107*** -0.010*
Marriage -0.0417 0.0133 -0.0677** -0.0705 0.0158 -0.0987*** 0.029
Family Structure 0.0117 0.0095 0.0229 0.0106 0.0113 0.0180 0.001
Latitude 0.0324 0.0177 0.0580 -0.0023 0.0210 -0.0035 0.035
Longititude 0.0064 0.0153 0.0167 0.0168 0.0182 0.0379 -0.010
Elevation -0.0003 0.0001 -0.0780 -0.0002 0.0002 -0.0415 0.000
Terrain -0.0009 0.0003 -0.0544** -0.0011 0.0004 -0.0548** 0.000
*** denotes p<.001; ** denotes p< .01; * denotes p< .05
Mortality variation: e0 2000 regression
Mortality variation: county level e0f (2000)
cluster, sig
low, <.10low, <.05low, <.01high, <.01high, <.05high, <.10not sig.
Huge regional variation: China is a world of its own!
Regional variation is relatively stable
Regional differences are largely driven by developmental factors
Mortality variation: observations
A model/system simultaneously estimating fertility, mortality, population age structure using all available survey/census data to consistently estimate demographic parameters.
Subnational evaluation
HMD standards/protocols
Next Steps
Thank you!