Presentation name or heading here

Presentation name or heading here

HEALTH AID, GOVERNANCE AND INFANT MORTALITY Chris (Hristos) Doucouliagos Department of Economics with Jack Hennessy and Debdulal Mallick 1. Background Health outcomes differ vastly across countries: Life expectancy is 53 years in Nigeria vs 84 years in Japan Maternal mortality is 1,360 in Sierra Leone per 100,000 vs 3 in Poland Infant mortality is 63 per 1,000 in Pakistan vs 3 in Australia 2 Defeating poverty today is the equivalent of defeating slavery in the 19th century. Tim O'Connor (UNICEF, Australia)

3 International policy objective Health the focus of: four of the eight Millennium Development Goals (United Nations, 2016) three of the Sustainable Development Goals (United Nations, 2018). 4 Policy objective of many countries Infant mortality is a key national health objective: Vietnam (Glewwe et al., 2004) Tanzania (Makani et al., 2015)

5 2. Our contribution Conceptual: aid effectiveness conditional on governance Empirical: long difference estimator new IV 6 Prior studies Mixed and fragile results (Williamson, 2008; Mishra & Newhouse, 2009; Nunnenkamp & hler, 2011; Wilson, 2011; Glassman & Temin, 2016; hler, 2011; Wilson, 2011; Glassman & Temin, 2016; Tarverdi & Rammohan, 2017). How to resolve conflicting findings? make aid effectiveness conditional on good governance and by addressing endogeneity through IV estimation. 7 Core finding

The impact of health aid on infant mortality is conditional on good governance. Health aid reduces infant mortality in countries with good governance. (government effectiveness or control of corruption) 8 3. Aid and governance conditionality Willingness and capacity to respond to preferences Corruption (can go either way depending on govt. strategic behavior) Investment in public goods 9 4. Models X: income; population; sanitation; doctors; female literacy; domestic health expenditure; clean water; initial mortality. Also, time and region dummies. 10

Measures Health Aid: (1) per capita $; (2) total $; share of GDP Governance: (1) government effectiveness; (2) control of corruption 11 Government effectiveness the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the governments commitment to such policies. 12 Long difference model Griliches & Hausman (1986) First differences not suitable for mortality, as it changes slowly (Angrist & Pischke, 2009 and Allegretto et al., 2017). 13

Mortality 45 0 nt mortality 50 55 Aid and infant mortality, all recipient nations, 1995-2015 14 -.33 -.34 ent effectiveness -.32

Average government effectiveness score, all recipients, 1995-2015 15 Decile (1) Government effectiveness score (2) Infant mortality (per 1000) (3) Total health aid ($ million) 1 2

3 4 5 6 7 8 9 10 -1.29 -0.98 -0.78 -0.61 -0.44 -0.20 0.00 0.34 0.85 2.43 73.39 60.87 50.31

47.13 40.67 37.50 31.01 19.87 19.31 8.02 59.61 48.23 46.69 59.84 64.26 50.40 52.85 28.39 7.96 1.62 16 IV IV strategy motivated by recent work instrumenting

aid: Werker et al. (2009); Nunn & Quin (2014); Ahmed (2016); Dreher & Langlotz (2017). Donor government fractionalization interacted with the probability of a country receiving health aid (Dreher & Langlotz 2017). 17 IV Government fractionalization increases the overall budget of the donor government, which in turn increases the aid budget. Fragmentation in the legislature has a similar effect ( Roubini & Sachs, 1989; Annett, 2001; Dreher & Langlotz, 2017). Political, social and ethnic fragmentation will also tend to increase the size of the governments overall budget (not always!). 18 Our identification compares infant mortality in health aid recipient countries by higher donor government fractionalization to lower donor government fractionalization.

Causal inference requires the assumption that donor government fractionalization influences infant mortality in recipient countries only through health aid (conditional on the set of control variables). 19 Zero stage regression , Similar logic to: Frankel & Romer (1999); Rajan & Subramanian (2008); Dreher & Langlotz (2017). 20 IV 5=+2( 5)+5+0 5++ 21 Baseline results (plain vanilla OLS) (1) Log aid pc (2)

Log aid pc (3) Log aid pc (4) Log aid pc (5) Log total aid (6) Aid/GDP -0.001 (-0.281) -0.005** (-2.143) -0.001 (-0.533)

-0.005** (-2.104) -0.005** (-2.056) -7.339*** (-4.333) Governance -0.068** (-2.357) -0.063* (-1.762) 0.020** (2.460) 0.020** (2.375)

Aid*Governa nce -0.005** (-2.160) -0.005* (-1.962) -0.008*** (-3.944) -4.803*** (-3.894) Variables Aid 22 Long difference (5) (1)

(2) (3) (4) Log total Log aid pc Log aid pc Log aid pc Log aid pc aid Aid Governance Aid*Governance -0.006* (-1.807) -0.004 (-0.897) -0.005* (-1.699) -0.002 (-0.519)

-0.002 (-0.561) 0.001 (-0.072) -0.002 (-0.066) -0.02 (-0.699) 0.001 0.003 -0.014*** (0.639) (1.432) (-3.005)

(6) Aid/GDP -10.081*** (-3.016) -0.003 (-0.121) -7.251** (-2.579) 23 Aid (1) Log aid pc 0.004 (0.678) Governance (2) Log aid pc

0.004 (0.774) (3) Log aid pc -0.008 (-1.342) (4) Log aid pc -0.016*** (-3.164) -0.008 (-0.927) -0.620** (-2.484) -0.369** (-2.380) -0.046**

(-2.443) -0.027** (-2.334) 0.967*** (12.82) 0.221** (2.31) 0.955*** (11.24) 0.446*** (4.05) 10.497 (0.001) 5.364 1,065 14.531 (0.000) 9.048

744 Aid * Governance Fitted aid First stage regressions 0.710*** 0.714*** (11.78) (11.80) Fitted aid interaction Kleinberg-Pap rk LM Kleinberg-Pap rk Wald F Observations 130.588 (0.000) 138.719 1,068 130.056

(0.000) 139.279 1,065 24 (1) Log total aid (2) Log total aid Aid -0.001 (-0.172) -0.017*** (-3.196) Governance

0.013 (1.208) 0.022 (1.629) Aid * Governance -0.011*** (-3.330) -0.011*** (-2.656) (3) Aid/GDP (4) Aid/GDP -5.947* (-1.788)

-10.629 (-1.579) 0.046** (1.964) 0.033** (2.278) -15.386** (-2.438) -11.249*** (-4.250) 25 ffect of health aid 0 .05 .1

The marginal effect of health aid on infant mortality conditional on government effectiveness 26 Robustness: non-health aid Aid (1) (2) Log aid per Log total capita aid -0.014** -0.014** -10.060* Governance (-2.233) -0.266**

Aid*Governance (-2.205) -0.019** Non-health aid*time dummies (-2.159) YES (-2.21) YES (-3.543) YES YES 21.283 (0.000) 19.271 707 YES

78.189 (0.000) 45.002 707 YES 11.661 (0.001) 6.422 724 Other controls Kleinberg-Pap rk LM Kleinberg-Pap rk Wald F Observations 27 (-2.16) 0.019 (3) Aid/GDP

(-1.830) 0.044*** (1.25) (2.706) -0.010** -15.490*** Robustness: control of corruption (1) Log aid pc Aid Corruption Aid*Corruption 0 (0.057) 0.004 (0.824)

(2) Log aid pc 0.125** (2.341) -0.488** (-2.378) -0.036** (-2.364) (3) Log aid pc 0.062 (1.352) -0.29 (-1.631) -0.022* (-1.665)

(4) Log total aid 0.024*** (2.847) 0.015** (2.557) -0.008*** (-4.216) (5) Log total aid 0.005 (0.503) 0.019*** (2.99) -0.006** (-2.247) (6)

Aid/GDP (7) Aid/GDP 64.687 (1.588) 18.451 (1.347) 0.016* (1.915) 0.023*** (2.908) -15.969** -9.693*** (-2.114) (-3.315) 28

The 10 worst governed countries: Somalia, South Sudan, Democratic People's Republic of Korea, Democratic Republic of the Congo, Comoros, Central African Republic, Equatorial Guinea, Myanmar, Togo, and Turkmenistan. average government effectiveness score of -1.674. At sample means, the model predicts a 3% increase in infant mortality (caused by aid) for these poorly governed nations (though p-value = 0.151) 29 For our sample, we find that only 28% of the recipient-year observations had at least a 3% reduction in infant mortality arising from health aid. 30 THANK YOU FOR YOUR PATIENCE! 31

Recently Viewed Presentations

  • Nicole Fiamingo's Presentation - Yola

    Nicole Fiamingo's Presentation - Yola

    Chapter 3. Research Design. Research Design - is a framework or blueprint for conducting the marketing research project.It details the procedures necessary for obtaining the information needed to structure or solve the marketing research problem.
  • Treatment of Dizziness with Physical Therapy - A New Drug for ...

    Treatment of Dizziness with Physical Therapy - A New Drug for ...

    Treatment of Dizziness with Physical Therapy - A New Drug for Patient Management University of Arkansas Medical School - Neurology Presenter: Brian K. Werner, PT, MPT Werner Institute for Balance and Dizziness - January 10, 2011 Overview Why Physical Therapy...
  • E-nursing Presentation - Catch the Wave

    E-nursing Presentation - Catch the Wave

    Best Evidence Resources What's fast, free and efficient Melanie Browne BHSc, MLIS Information Specialist, Mount Sinai Hospital, Toronto, Canada
  • Conducting a Close Read of Text

    Conducting a Close Read of Text

    A is for Antonym definition. An antonym definition is a word of opposite meaning. The signal words are: instead, although, but, yet, however, rather, despite, either and/or or. ... On this day we gather because we have chosen hope over...
  • The Globalization of Environmentalism - Weebly

    The Globalization of Environmentalism - Weebly

    Environmental Changes. 3 major factors accelerated the impact of humans on the environment in the 20th century: Population explosion. New ability of humankind to tap the energy potential of fossil fuels coal, oil, hydroelectricity, natural gas, nuclear power
  • Language Learners' Interaction - SEAMEO REGIONAL TRAINING CENTER

    Language Learners' Interaction - SEAMEO REGIONAL TRAINING CENTER

    1. Introduction . Thu Do_International TESOL Conference, Sept. 2011. not consider the comparison between NNS/NNS and NS/NNS interactions (except for Varonis & Gass, 1985, 1994; Pica et al., 1996)
  • Natural Gas Trends and Preferred Use Midwest Energy

    Natural Gas Trends and Preferred Use Midwest Energy

    Solar Thermal/Natural Gas Hybrid Systems. Bio-Methane Production, Clean-Up, and Use. Utilization RD&D Portfolio Examples. ... direct use of natural gas should not be viewed as a bridge fuel but rather an integral part of the energy mix and a tool...
  • Cumulative impact of multiple impairments on JT performance

    Cumulative impact of multiple impairments on JT performance

    Phase drift (of slave APs relative to master) during joint TX, results in the channel seen by STA during payload changing versus the initial estimate (from preamble) as explained in [3].