Health effects of indoor and outdoor air pollution Jie-Sheng Tan-Soo Subhrendu K. Pattanayak Duke University Camp Resources 2012.

Health effects of indoor and outdoor air pollution

Jie-Sheng Tan-SooSubhrendu K. Pattanayak

Duke UniversityCamp Resources 2012

2

Research questions

• What are the joint impacts of indoor and outdoor air pollution on children’s health?– Are impacts different for self-reported and

objective health measurements?

3

Introduction• Respiratory illnesses is top disease burden for developing

countries (WHO, 2008)• Environmental causes: indoor and outdoor air pollution• Indoor air pollution from ‘dirty’ household fuel (PM2.5 air

particulates, carbon monoxide)• Outdoor air pollution from emissions and forest fires (PM10 air

particulates, SO2, NOX, CO, O3)• Climate change to result in worsening outdoor air pollution

(IPCC, 2007)• Need better understanding on the concurrent effects of indoor

and outdoor air pollution• Need to consider endogenous exposure to indoor air pollution

4

Why conduct another study?

• Epidemiological reason: – Most existing studies are conducted in places

where exposure is low compared to developing countries

• Econometric reason / omitted variable bias: – 1st: Indoor air pollution highly correlated with

outdoor air pollution (see review by Yocom, 1982)– 2nd: Exposure to air pollution may be endogenous

(e.g. Mullahy and Portney, 1990)

5

Study Approach

• Panel individual-level dataset from Indonesia• Include measurements for both indoor air

pollution (IAP) and outdoor air pollution (OAP)• Use household fixed effects to control time-

invariant heterogeneity• Use a wide range of self-reported and

objective health measures

6

Literature Review

• Use 1997 Indonesian forest fires as exogenous shock to OAP (Frankenberg et al., 2005 ; Jayachandran, 2008)

• Use fuel price as instrument for stove usage (Hanley, 2006)

• Include both IAP & OAP in India (Ghosh & Mukherji, 2011)

7

Model• Health production framework (Rosenzwig & Schultz, 1983)

– Maximize household utility subject to health production function and budget constraint

• X: consumption good• R: child’s respiratory health• I: indoor air pollution• O: outdoor air pollution• : prices• Y: income• : unobserved family factors such as health preferences or attitudes

• Reduced-formed child’s health equation

• ‘Hybrid’ health equation are typically estimated because of data constraint

8

Model

• Bias 1: IAP is correlated with OAP where

Marginal effect of IAP is If IAP is not included, then estimates of will be

biased as • Bias 2: Exposure to IAP is endogenous and

correlated with other determinants of health Marginal effect of IAP is

9

Empirical strategy

• Estimating equation

• i: child; j: household; k: district• Measures for both IAP and OAP– Dirty stove indicator for IAP– Aerosol index for OAP

• Inclusion of household fixed effects to control for time-invariant household heterogeneity– Neglect cross-sectional variation– Identification of the IAP parameter is from households that

switched stoves type

10

Data

• Indonesia Family and Life Survey– Main type of stove used & individual health measurements– Four waves (1993, 1997, 2000, 2007)– Panel individual and household survey– High re-contact rate/ Low attrition– ~20,000 children (under age of 12) over three waves

• NASA Total Ozone Mapping Spectrometer– Aerosol index– From 1993 onwards– Satellite data at district level

11

Source: http://www.rand.org/labor/FLS/IFLS.html

12

Sep 24, 1997

Sep 18, 2000

Sep 21, 2005

13

Data

• Self-reported health measurements– Cough in last 4 weeks?– Breathing difficulties in last 4 weeks?– Number of days ill in last 4 weeks?– Number of days in bed due to poor health in last weeks?– Assessment of overall health (1-4 with 1 being very unhealthy)

• Objective health measures:– Lung capacity– Body-mass index– Height-for-age z-score– Weight for age z-score

14

Results – self-reported health

  (1) (2) (3) (4) (5)

VARIABLESCough Breathing difficulty Days ill Days miss activity Negative health

self-assessment

Aerosol index (AI) -0.00140** 0.000144 -0.00473*** -0.00603* 2.26e-06(p-value)  (0.0168) (0.510) (0.00442) (0.0695) (0.997)

Dirty stove indicator 0.00559 -0.00559 -0.0149 -0.0117 -0.0421**

 (0.757) (0.375) (0.796) (0.907) (0.0129)

Observations 20,863 20,842 20,852 20,851 20,866

R-squared 0.021 0.003 0.006 0.022 0.013

Number of hh_id 7,009 7,008 7,009 7,010 7,011

Controls: Age, Gender, Household assets, Year FEExpected signs: +ve for column (1) to (5)

15

Results – objective health

  (1) (2) (3) (4)

VARIABLES Body-mass index Lung capacity Height-for-age Weight-for-age

Aerosol index (AI) -0.00467 -0.293** -0.00396 -0.00681**(p-value)  (0.147) (0.0181) (0.398) (0.0411)

Dirty stove indicator 0.0348 -6.296 0.0861 -0.0759

 (0.705) (0.105) (0.527) (0.378)

Observations 20,916 6,685 7,797 7,797

R-squared 0.004 0.213 0.032 0.049

Number of hh_id 7,023 3,974 4,933 4,933

Controls: Age, Gender, Household assets, Year FEsExpected signs: -ve for column (1) to (4)

16

Robustness checks

• Use of fixed effects result in loss of cross-sectional variation

• Test of FE vs. RE (Hausman test)– Household fixed effects uncorrelated with regressors

VARIABLES (p-value) VARIABLES (p-value)

Cough (0.00) Body-mass index (0.0395)

Breathing difficulty (0.453) Lung capacity (0.0001)

Days ill NA Height-for-age (0.0051)

Days miss activity (0.1140) Weight-for-age (0.0207)

Health self-assessment (0.0036)

17

Robustness checks• Hausman-Taylor method (1981) to estimate

the impact of time-invariant regressors

VARIABLES

Cough Breathing difficulty Days ill

Days miss

activity

Health self-assessment Body-mass

indexLung

capacityHeight-for-

ageWeight-for-age

Dirty stove indicator -0.208*** -0.01 -0.015 -0.38 0.0003 -0.46* -46.7*** -1.21*** -1.02***

Aerosol index -0.0007 0.0004*** -0.004*** -0.0046* 0.0008* -0.0074*** -0.36*** -0.002 -0.005**

18

Robustness checks• Hausman-Taylor method (1981) to estimate

the impact of time-invariant regressors

VARIABLES

Cough Breathing difficulty Days ill

Days miss

activity

Health self-assessment Body-mass

indexLung

capacityHeight-for-

ageWeight-for-age

Dirty stove indicator -0.208*** -0.01 -0.015 -0.38 0.0003 -0.46* -46.7*** -1.21*** -1.02***

Dirty stove indicator 0.0056 -0.006 -0.015 -0.012 0.04** 0.035 -6.3 0.09 -0.076

Aerosol index -0.0007 0.0004*** -0.004*** -0.0046* 0.0008* -0.0074*** -0.36*** -0.002 -0.005**

Aerosol index -0.001** 0.0001 -0.005*** -0.006* -2.26e-06 -0.0047 -0.29** -0.004 -0.007**

19

Summary

• Self-reported health likely unreliable• Only OAP has adverse impact on health– Results are driven by HHs that switch stove

• Hausman-Taylor estimator however shows that both OAP & IAP are damaging to health– But results are driven by non-switcher HHs

• In all, results suggest differences between switchers & non-switchers– Long term exposure vs. short term?– Switchers and non-switchers fundamentally different?

20

Key points

• Effects of air pollution on health is a very important and complex problem– Respiratory illnesses is a major burden– Close relationship between indoor and outdoor

pollution– Endogenous exposure to pollution

• Indoor and outdoor air pollution harm health– Switchers vs. non-switchers– Self-reported vs. objective health measures

• Need structural modeling