You are on page 1of 31

31 July 2017

Oral Defense
Estimating the Effect of Air Pollution on Health Insurance
Coverage using Indonesian Family Life Survey
Supervisor: Professor Koji Shimada
Sub-supervisor: Professor Kiyoto Kurokawa

Dimas Hari Santoso


Master Program of Economic Development – M2
Graduate School of Economics - Ritsumeikan University
Student ID: 52211500450
e-mail: dimashari@bps.go.id
Contents
1. Background Problems

2. Method & Findings

3. Analysis & Discussion

4. Conclusion & Recommendation

2
1. Background Problems
• Indonesia: middle income country with high and rapid growth 
industrialization
• Rapid Industrialization Effect:
• increasing production scale
• increasing energy use and industrial waste  worsen environmental quality
• unplanned rapid urbanization & increasing population density  intensified
both emission and concentration of air pollution
• Air pollution substantially affects human health
• People who exposed by air pollution tend to spent more expenditure
on medical care than those who does not
3
Air Pollution Problem-Tree Framework in Indonesia
5
Riau Province
Credits: google image

6
Central
Kalimantan
Province
Credits: google image

7
Previous Studies
Authors Regions Findings
Narayan & Narayan OECD countries Income, Sulphur oxide emissions and carbon monoxide emissions
(2007) exert a statistically significant positive effect on health
expenditures.
Jerret, et. al. (2003) 49 counties of total toxic pollution output and per capita municipal environmental
Ontario, Canada expenditures have significant associations with health expenditures
Knaul (2011) 12 countries in rural residence, lowest level of income, presence of elderly people,
Latin America and the absence of health insurance coverage have become the
and the main reasons of catastrophic health expenditure
Caribbean
Xu, et. al. (2003) 59 countries Catastrophic medical spending were highest in some countries in
structural transition, and in certain Latin American countries. Three
key preconditions for catastrophic payments were identified: the
availability of health services requiring payment, low capacity to
pay, and the lack of prepayment or health insurance.

8
Research Question & Hypotheses
• Research Question
i. How do Indonesian households respond to air pollution problem concerning their health
condition?
ii. Does the air pollution in Indonesia affect health insurance coverage among households?
iii. What do other factors might affect the coverage of health insurance among households
in Indonesia?
• Hypotheses
i. Suspended Particulate Matter, as one type of air pollution, will positively affect the
insurance coverage among Indonesian household.
ii. Beside SPM concentration, other factors such as household income, health condition of
the household, and household characteristics will also positively impact on the health
insurance coverage.

9
2. Method & Findings: IFLS
• Indonesian Family Life Survey
• Conducted by Rand Corporation
• 5 waves: 1993, 1997, 2000, 2007, 2014
• Target: household in 16 provinces  87% of Indonesian population
• Using SUSENAS 1993 sample framework  based on Population Census 1990
IFLS1 IFLS2 IFLS3 IFLS4 IFLS5
Year 1993 1997 2000 2007 2014
Total Household 7,224 8,116 11,112 15,011 18,277
Split-off Household4 - 1,374 4,548 9,011 12,713
Re-contacted Household - 6,742 6,564 6,010 5,564

10
2. Method & Findings: Air Quality Monitoring
• Data: monthly average Suspended Particulate Matter (gram/m3/24
Hours)
• Provided by Indonesian Meteorological Agency
• Only 31 monitoring station (from totally 150 stations) can be collected
• Limitation:
• Several monitoring station did not continuously operating
• Missing data on several years
• Regional data

11
2. Method & Findings: Variables
Variable Categories Units Details Sources
Health Insurance Binary 1: yes, 2: no Household covered by health insurance RAND® IFLS 1-5
Suspended Particulate Matter Continues gram/m3/24 Yearly average daily SPM concentration BMKG
hours
Household Income Continues IDR Total yearly household income1 RAND® IFLS 1-5
Higher Income Level Binary 1: yes, 2: no Household with level income of more than RAND® IFLS 1-5
IDR 10 million per month (base: low income)
Middle Income Level Binary 1: yes, 2: no Household with level income of between IDR RAND® IFLS 1-5
2.5 million-10 million per month
Household Cigarettes Continues IDR Yearly consumption level of cigarettes in the RAND® IFLS 1-5
Consumption household

1] Total Household Income consist of: Total Earning of Household Member, Total Income from Household Assets,
Nonlabor Income, Net Profit from Farm Business, Net Profit from Nonfarm Business, Total Income from Farm
12
Business Assets, Total Revenue from Nonfarm Business Assets.
2. Method & Findings: Variables
Variable Categories Units Details Sources
Household Alcohol Consumption Continues IDR Yearly consumption level of alcohol in the RAND® IFLS 1-5
household
Chronic Disease Binary 1: Ever, 2: Never Household which its member ever got a RAND® IFLS 1-5
disease
Outpatient Utilization Binary 1: yes, 2: no Household who utilized outpatient care in RAND® IFLS 1-5
the past year
Inpatient Utilization Binary 1: yes, 2: no Household who utilized inpatient care in RAND® IFLS 1-5
the past year
Residential Location (Urban/Rural) Binary 1: Urban, 2: Rural Household who utilized outpatient care in RAND® IFLS 1-5
the past year
Housing Status (Owned/Rent) Binary 1: Owned, 2: Rent Housing status of the household RAND® IFLS 1-5

13
2. Method & Findings: Sample Selection
• Limitation: data matching problems
• Different unit data: IFLS-household data, SPM-regional data
• Missing data on several years of SPM: unbalanced data
• Sample selection process:
• Select SPM data from monitoring station which located on 16 provinces that
conducted IFLS1-5 (resulting SPM data from 31 monitoring stations)
• select the target household samples that are in the same city/region as one of
the selected observation stations in the previous stage
• Estimation method:
• Logit model  dichotomous dependent variable
• Fixed Effects and Random Effects  unobserved time invariant variables
14
2. Method & Findings: Variables Statistics Descriptive
1993 1997 2000 2007 2014
Variable N=419 N=470 N=491 N=953 N=1002
Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.
Health Insurance 0.2864 0.4526 0.2106 0.4082 0.1772 0.3822 0.2949 0.4562 0.5140 0.5001
Suspended Particulate Matter 254.311 147.575 236.528 142.622 243.331 185.758 132.864 82.511 132.335 86.155
Household Income (IDR) 1.10E+06 8.16E+06 6.22E+05 9.36E+05 9.69E+05 1.28E+06 2.49E+06 5.09E+06 5.97E+06 1.00E+07
Higher Income Level 1.19E-02 1.09E-01 2.13E-03 4.61E-02 2.04E-03 4.51E-02 3.88E-02 1.93E-01 1.48E-01 3.55E-01
Middle Income Level 1.67E-02 1.28E-01 3.83E-02 1.92E-01 6.52E-02 2.47E-01 2.51E-01 4.34E-01 4.46E-01 4.97E-01
Household Cigarettes Consumption (IDR) 1.57E+04 4.88E+04 2.26E+04 3.13E+04 5.35E+04 8.33E+04 1.61E+05 5.13E+05 2.03E+05 3.06E+05
Household Alcohol Consumption (IDR) 4.44E+02 6.62E+03 2.30E+01 4.63E+02 1.31E+02 1.49E+03 1.03E+04 2.60E+05 4.14E+03 5.18E+04
Chronic Disease 0.8902 0.3130 0.9532 0.2115 0.8534 0.3541 0.8206 0.3839 0.7754 0.4175
Outpatient Utilization 0.1766 0.3818 0.2255 0.4184 0.2179 0.4133 0.1700 0.3758 0.2665 0.4423
Inpatient Utilization 0.0453 0.2083 0.0723 0.2593 0.0835 0.2769 0.0902 0.2867 0.1427 0.3500
Residential Location (Urban/Rural) 0.5442 0.4986 0.5426 0.4987 0.5601 0.4969 0.5110 0.5001 0.4661 0.4991
Housing Status (Owned/Rent) 0.8735 0.3328 0.8851 0.3192 0.9104 0.2859 0.9381 0.2411 0.9321 0.2516

15
2. Method & Findings: the econometric model

16
2. Method & Findings: the econometric model
• i=1, …, n represents households • DUMDISit is medical records dummy variable (1 if there is a
• t=1993, 1997, 2000, 2007, 2014 represents years.
• p is the conditional probability that household have been common disease occurrence in the household)
covered by health insurance denoted by p=Pr(INSit=1|xit) • DUMURLit is the residential location dummy variable (1 if
• xit is the independent variables vector urban)
• INSit is the insurance coverage binary dependent variable • DUMHOSit is housing status (1 if owned by household)
(1 if household covered by insurance) • SPMit is monthly average suspended particulate matter
• INCit is the monthly average of total household income
concentration in ambient air
• DUMHICit is dummy variable for high income level
household • DUMOPTit is outpatient care dummy variable (1 if ever
• DUMMICit is dummy variable for middle income level utilize outpatient care)
household • DUMIPTit is inpatient care dummy variable (1 if ever utilize
• CIGit is the total number of household consumption of inpatient care)
cigarettes
• ui is a vector of unobserved time-invariant specific effect
• ALCit is the total number of household consumption of
• it individual error or time-varying error bounded with the
alcohol
classical statistical properties.
17
2. Method & Findings: estimation result
• Click here for the estimation result

18
3. Discussion: Insurance & Income

19
3. Discussion: Income & Air Pollution
• The environmental Kuznets hypothesis:

• Result by using Fixed Effect estimation method:


  1993** 1997*** 2000 2007 2014*** Panel FE***
INC2 -2.84E-13*** -2.94E-12 -1.47E-12 -1.54E-14 -2.58E-14*** 1.00e-13***
  (9.52E-14) (1.98E-12) (1.04E-12) (1.42E-14) (8.44E-15) (8.61e-15)
INC 2.34E-05*** 5.20E-05*** 1.11E-05 1.64E-06 2.41E-06*** -8.59e-06***
  (7.91E-06) (1.29E-05) (1.13E-05) (9.98E-07) (6.30E-07) (6.38e-07)
Constant 2.48E+02*** 2.08E+02*** 2.36E+02*** 1.29E+02*** 1.21E+02*** 197.5912***
  (7.54E+00) (8.81E+00) (1.22E+01) (3.41E+00) (3.87E+00) (2.033225)
N 419 470 491 953 1002 3335
F-test 4.45 14.77 1.04 1.53 7.78 91.21
Pr>F 0.0123 0.0000 0.3541 0.2161 0.0004 0.0000
Adj R2 0.0162 0.0555 0.0002 0.0011 0.0134 0.0808

20
3. Discussion: Income & Air Pollution

21
Conclusion
• higher income level will boost the possibility of a household to be
covered by an insurance scheme
• the SPM pollution has a statistically significant and negative effects on
the insurance scheme  deny the 1st hypothesis
• negative relationship between SPM pollution and insurance coverage
was clarified by the household income level  EKC Hypothesis
• We can also conclude that several other factors are also affecting the
insurance coverage (2nd hypothesis):
• the total household income, inpatient and outpatient utilization as the
household health status indicators, and residential area as well as housing
status as the household living characteristics
22
Policy Recommendations
1. Build a better legal framework for the implementation of the EMA
(Environmental Management Acts) 32/2009 in the specific
subsectors
2. Formulate a special scheme that help the poor to access health
services to cure any possible threat from air pollution as an
additional feature of Indonesian National Health Insurance
3. Free air pollution medical check-up provided by the government
4. Introduce an environmental education
5. optimize the number and distribution of air quality monitoring
station
23
Limitation & Further Studies
1. Further studies should enlarge the household sample coverage to
capture more accurate information
2. More control variables may be introduced in the next further studies
• Air pollution: Sox, NOx, CO, etc.  different region may have different
problems
• Household characteristics: education, housing facilities, total expenditure
• Household awareness indicators
3. Next studies will observe how the different types of health insurance
in Indonesia will be affected by any environmental degradation 
the impact of national health insurance program implementation in
2014
24
Thank you.
Appendix 1: Indonesia in Figures
GNI per capita (current US$) GDP growth (annual %)
4000
15

10
3000
5

2000 0

-5
1000
-10

0 -15
60 63 6 6 69 72 75 78 8 1 84 87 90 93 96 99 02 05 08 1 1 14

1960
1962
1964
1966
1968
1970

1974

1980
1982
1984
1986
1988
1990
1992

2002
2004
2006
2008
2010
2012
2014
1972

1976
1978

1994
1996
1998
2000
19 19 19 19 19 19 19 19 1 9 19 19 1 9 19 19 20 20 20 20 2 0

Source: The World Bank


26
Appendix 2: Indonesia in Numbers
Population (million people) Share of GDP by Sectors
300000000 70
Tertiary Sectors
60 Secondary Sectors
250000000
Primary Sectors
50
200000000
40
150000000
30
100000000
20
50000000 10

0 0
60 63 6 6 69 72 75 78 81 84 87 90 93 96 99 02 05 08 11 14

1962
1964

1968

1972
1974
1976
1978

1982
1984

1988

1992
1994

1998
2000
2002
2004

2008

2012
2014
1960

1966

1970

1980

1986

1990

1996

2006

2010
19 19 19 19 19 19 19 19 19 1 9 19 19 19 19 20 20 2 0 20 20

Source: The World Bank


27
Appendix 3: Affected Areas of forest fires in Indonesia
Ha
50000.000
44411.36000

40000.000

30000.000

20000.000

9606.53000
10000.0007619.41000 11240.78000
3500.12000
2612.09000 4918.74000
.000
2009 2010 2011 2012 2013 2014 2015

Source: (Ministry of Environment and Forestry, 2014)


Appendix 4: Forest Fires & International Comparison
• Other Causes
• Forest fire
• Automobile (Number of Registered Vehicles, National Police)
2010 2011 2012 2013 2014 2015
Pasenger Cars 8.891.041 9.548.866 10.432.259 11.484.514 12.599.038 13.423.483
Buses 2.250.109 2.254.406 2.273.821 2.286.309 2.398.846 2.402.720
Trucks 4.687.789 4.958.738 5.286.061 5.615.494 6.235.136 6.505.595
Motorcycle 61.078.188 68.839.341 76.381.183 84.732.652 92.976.240 99.675.001
Total 76.907.127 85.601.351 94.373.324 104.118.969 114.209.260 122.006.799

• Other insight: International Comparison


• Indonesia at number 8 the most polluted country in the world (Bloomberg, 2015)
29
Environment Quality Index Average Particulate Matter (PM, gr/m3)
Source: Ministry of Environment and Forestry Source: Meteorological, Climatological, And
Geophysical Agency
100.000
Yogyakarta
Cities 2010 2011 2012 2013 2014
Palangkaray
80.000 a Yogyakarta 193,13 186,47 166,71 176,72 269,41
Pekanbaru Semarang 89,11 183,43 264,98 259,03 270,34
Jakarta Surabaya 90,84 129,96 127,95 97,46 93,29
60.000 Tangerang Palangkaraya 18,60 42,82 18,93 35,15 81,63
Denpasar Pekanbaru 171,18 176,52 219,35 367,90 221,71
Makasar Bandung 182,61 195,03 165,43 151,39 134,24
40.000 Medan Jakarta 194,37 187,67 211,09 235,85 208,70
Ambon Tangerang 232,63 238,26 800,20 224,94 192,30
Jayapura Denpasar 43,48 60,27 50,75 70,17 83,59
20.000
2010 2011 2012 2013 2014 Makasar 139,55 195,17 173,44 170,77 61,68
Medan 142,55 223,37 298,12 157,77 192,20
Criteria Range Criteria Range Ambon 15,02 14,76 17,73 48,34 na
Advance > 90 Less 58-66 Jayapura 15,25 11,81 63,50 179,68 32,52
Very Good 82-90 More Less 50-58
Good 74-82 Bad < 50
Fair 66-74 PM Ambience Level 150 gr/m3

30
Appendix 5: List of IFLS Provinces
IFLS1 IFLS2 IFLS3 IFLS4 IFLS5
Sumatera Utara Sumatera Utara Sumatera Utara Sumatera Utara Sumatera Utara
Sumatera Barat Sumatera Barat Sumatera Barat Sumatera Barat Sumatera Barat
Sumatera Selatan Sumatera Selatan Sumatera Selatan Sumatera Selatan Sumatera Selatan
Babel Babel
Lampung Lampung Lampung Lampung Lampung
DKI Jakarta DKI Jakarta DKI Jakarta DKI Jakarta DKI Jakarta
Jawa Barat Jawa Barat Jawa Barat Jawa Barat Jawa Barat
BANTEN BANTEN
Jawa Tengah Jawa Tengah Jawa Tengah Jawa Tengah
Jawa Tengah
DI Yogyakarta DI Yogyakarta DI Yogyakarta DI Yogyakarta
DI Yogyakarta
Jawa Timur Jawa Timur Jawa Timur Jawa Timur Jawa Timur
Bali Bali Bali Bali Bali
Nusa Tenggara Barat Nusa Tenggara Barat Nusa Tenggara Barat Nusa Tenggara Barat Nusa Tenggara Barat
Kalimantan Selatan Kalimantan Selatan Kalimantan Selatan Kalimantan Selatan Kalimantan Selatan
Sulawesi Selatan Sulawesi Selatan Sulawesi Selatan Sulawesi Selatan Sulawesi Selatan
Sulawesi Barat Sulawesi Barat Sulawesi Barat
Sulawesi Barat
31

You might also like