Impact of Income Inequality on Health: Does Environment Quality Matter?

Similar documents
Volume 30, Issue 4. Who likes circus animals?

International Trade and California Employment: Some Statistical Tests

1 Basic concepts for quantitative policy analysis

An Empirical Study about the Marketization Degree of Labor Market from the Perspective of Wage Determination Mechanism

emissions in the Indonesian manufacturing sector Rislima F. Sitompul and Anthony D. Owen

Evaluating the statistical power of goodness-of-fit tests for health and medicine survey data

Bulletin of Energy Economics.

Extended Abstract for WISE 2005: Workshop on Information Systems and Economics

Appendix 6.1 The least-cost theorem and pollution control

Biomass Energy Use, Price Changes and Imperfect Labor Market in Rural China: An Agricultural Household Model-Based Analysis.

WISE 2004 Extended Abstract

Willingness to Pay for the Quality of Drinking Water

LIFE CYCLE ENVIRONMENTAL IMPACTS ASSESSMENT FOR RESIDENTIAL BUILDINGS IN CHINA

Labour Demand Elasticities in Manufacturing Sector in Kenya

Driving Factors of SO 2 Emissions in 13 Cities, Jiangsu, China

The Impact of CO 2 Emission Cuts on Income

Relative income and the WTP for public goods

Market Dynamics and Productivity in Japanese Retail Industry in the late 1990s

FDI and Intra-industrial Technology Spillovers ---Empirical Study on China s Manufacturing Industries

Do not turn over until you are told to do so by the Invigilator.

The Effect of Outsourcing on the Change of Wage Share

A Longer Tail?: Estimating The Shape of Amazon s Sales Distribution Curve in Erik Brynjolfsson, Yu (Jeffrey) Hu, Michael D.

Calculation and Prediction of Energy Consumption for Highway Transportation

Innovation in Portugal:

Development trajectory of energy consumption and carbon emissions in developing countries

Experiments with Protocols for Service Negotiation

Firm Performance and Foreign Direct Investment: Evidence from Transition Economies. Abstract. Department of Economic, University of Texas at Arlington

A Two-Echelon Inventory Model for Single-Vender and Multi-Buyer System Through Common Replenishment Epochs

The Substitutability of Labor of Selected Ethnic Groups in the US Labor Market

Spatial difference of regional carbon emissions in China

The Role of Price Floor in a Differentiated Product Retail Market

NBER WORKING PAPER SERIES UNMASKING THE POLLUTION HAVEN EFFECT. Arik Levinson M. Scott Taylor. Working Paper

Uniform Standard or Emissions Trading? Efficient coverage of installations in a GHG Emission Trading Scheme

Public goods and the value of product quality regulations: the case of food safety

Analyses Based on Combining Similar Information from Multiple Surveys

Volume 29, Issue 2. How do firms interpret a job loss? Evidence from the National Longitudinal Survey of Youth

IS INCOME INEQUALITY ENDOGENOUS IN REGIONAL GROWTH? Yohannes G. Hailu 1 Mulugeta S. Kahsai 2 Tesfa G. Gebremedhin 3 Randall W.

Introducing income distribution to the Linder hypothesis

CONSUMER PRICE INDEX METHODOLOGY (Updated February 2018)

A Neoclassical Growth Model for Environmental Expenditure of S-shaped Curve and Pollution of Inverted U-shaped Curve

Do Remittances Alter Labor Market Participation? A Study of Albania

Evaluation Method for Enterprises EPR Project Risks

Economic incentives and the quality of domestic waste: counterproductive effects through waste leakage 1

Econometric Methods for Estimating ENERGY STAR Impacts in the Commercial Building Sector

Overeducation in Cyprus

Volume-3, Issue-8, January-2017 ISSN No:

Study on the Coupling Development between Urbanization and Ecosystem-- The Comparative Analysis Based on Guizhou, Yunnan, Hunan and Zhejiang Province

Field Burning of Crop Residues

Selected Economic Aspects of Water Quality Trading

Demand for U.S. Lamb and Mutton by Country of Origin: A Two-Stage Differential Approach

The Allocation of Time and Goods: Three Essays on American Household Shopping Behavior

PRODUCTIVE PUBLIC EXPENDITURE AND IMPERFECT COMPETITION WITH ENDOGENOUS PRICE MARKUP: COMMENT

STRATEGIC OUTPUT AND GREEN TECHNOLOGY RIVALRY IN A GLOBALIZED WORLD: THEORY, EMPIRICS AND POLICY IMPLICATION. A Thesis

The link between immigration and trade in Spain

The Impact of Carbon Tax on Economic Growth in China

Household Budget and Calorie Consume of Livestock Products: Evidence from Indonesia SUMMARY

An Analysis of the Impact of ICT Investment on Productivity in Developing Countries: Evidence from Cameroon

Welfare Gains under Tradable CO 2 Permits * Larry Karp and Xuemei Liu

Building Energy Consumption and CO 2 Emissions in China

Chinese Economic Reform and Labor Market Efficiency

Supplementary Appendix to. Rich Communication, Social Preferences, and Coordinated Resistance against Divide-and-

The Impact of Agricultural Extension on Farmer Nutrient Management Behavior in Chinese Rice Production: A Household-Level Analysis

NBER WORKING PAPER SERIES INNOVATION AND PRODUCTIVITY ACROSS FOUR EUROPEAN COUNTRIES. Rachel Griffith Elena Huergo Jacques Mairesse Bettina Peters

Regulating monopoly price discrimination

Impact of Internet Technology on Economic Growth in South Asia with Special Reference to Pakistan

A Group Decision Making Method for Determining the Importance of Customer Needs Based on Customer- Oriented Approach

Gender Wage Differences in the Czech Public Sector: A Micro-level Case

Research on Evaluation Index System for Automobile Enterprise Brand Value

Consideration of the nexus between Institutions and Outreach. -The study of the caste discrimination in India -

Regression model for heat consumption monitoring and forecasting

Modelling and analysis of international recycling between developed and developing countries

A Reevaluation of the Effect of Human Capital Accumulation on Economic Growth: Using Natural Disasters as an Instrument

Saving Investment Correlation in South Asia- A Panel Approach

Relationship Between the Uncompensated Price-Elasticity and the Income- Elasticity of Demand Under Conditions of Additive Preferences

Impact of public research on industrial innovation

Absenteeism and Peer Interaction Effects: Evidence from an Italian Public Institute

Lecture 5: Applications of Consumer Theory

Sources of information

Short- and Long-Run Structural International Tourism Demand Modeling Based on the Dynamic AIDS Model: Empirical Research in Japan

Product Innovation Risk Management based on Bayesian Decision Theory

6.4 PASSIVE TRACER DISPERSION OVER A REGULAR ARRAY OF CUBES USING CFD SIMULATIONS

More Information Isn t Always Better: The Case of Voluntary Provision of Environmental Quality. Ann L. Owen. Julio Videras.

research paper series

Are the Chinese Really More Risk Averse? A Cross-Cultural Comparison of the Risk-Return Relationship

Applying Emission Tax and Emission Permit Schemes Controlling Greenhouse Gas Emissions in the Thai-Cement Industry: A Social Welfare Analysis

Numerical Analysis about Urban Climate Change by Urbanization in Shanghai

The Labor Market Impacts of. Adult Education and Training in Canada

Energy 45 (2012) 867e873. Contents lists available at SciVerse ScienceDirect. Energy. journal homepage:

Consumption capability analysis for Micro-blog users based on data mining

Coupon Redemption and Its Effect on Household Cheese Purchases

The Greenness of Cities: Carbon Dioxide Emissions and Urban Development

Analysis on the Regional Disparity in China and the Influential Factors

The Spatial Equilibrium Monopoly Models of the Steamcoal Market

Maternal Education, Linkages and Child Nutrition in the Long and Shortrun: Evidence from the Ethiopia Demographic and Health Surveys

RELATIONSHIP BETWEEN BUSINESS STRATEGIES FOLLOWED BY SERVICE ORGANIZATIONS AND THEIR PERFORMANCE MEASUREMENT APPROACH

Gender differentials in agricultural productivity: evidence from Nepalese household data

Evaluation on external economies of renewable energy resource utilization: Taken wind power engineering project as example

THE IMPACT OF HIGHER EDUCATION INSTITUTION-FIRM KNOWLEDGE LINKS ON FIRM-LEVEL PRODUCTIVITY IN BRITAIN. RICHARD HARRIS, QIAN CHER LI AND JOHN MOFFAT.

Firm Productivity in Jamaica: An Investigation of Total Factor Productivity with Micro-Level Firm Data.

Transcription:

Document de traval de la sére Etudes et Documents E 2010.06 Impact of Income Inequalty on Health: Does Envronment Qualty Matter? Alassane DRABO 1 CERDI - Unversty of Auvergne - France Janver 2010 1 Tel: +33 (0)6 27 60 49 64. Fax: +33 (0)4 73 17 74 28. E-mal address : alassanedrabo@hotmal.com 1

Abstract : Ths paper examnes the lnk between health ndcators, envronmental varables and ncome nequaltes. Theoretcally, all the mechansms developed n the lterature underlne a negatve mpact of ncome nequalty on health status. However, emprcal studes fnd dfferent results and the conclusons are far from a consensus. In ths paper we nvestgate how envronment degradaton could be consdered as a channel through whch ncome dstrbuton affects populaton health. We frst develop a smple theoretcal model based on Magnan (2000), n whch relatve ncome affects health status through the level of polluton abatement expendtures. Our econometrc analyss shows that ncome nequaltes negatvely affect envronmental qualty and envronment degradaton worsens populaton s health. Ths negatve effect of ncome nequaltes on envronment s mtgated by good nsttutons. We also show that ncome nequaltes negatvely affect health status. Another nterestng result s that when envronmental varables are taken nto account, the level and the statstcal sgnfcance of the coeffcent of ncome nequalty varable vansh. Ths confrms that envronment qualty s an mportant channel through whch ncome nequaltes affect populaton health. These results hold for ar polluton ndcators (CO2 and SO2) and water polluton ndcator (BOD). It s also robust for rch and developng countres. Countres wth hgh ncome nequaltes may mplement dstrbutve polces n order to avod ts negatve mpact on health. Keywords: health status, ncome nequalty, envronmental qualty, nstrumental varables method JL classfcaton: C13, D63, I1; Q5 2

1. Introducton Populaton health s an mportant economc concern for many developng countres. It plays a crucal role n development process, snce t consttutes a component of nvestment n human captal and workforce s the most abundant producton factor n these countres. It consttutes also a major preoccupaton for the nternatonal communty, especally when t s consdered as a publc good. The mportance gven to health status could be llustrated through ts relatvely hgh weght among the Mllennum Development Goals (MDGs), of whch three are related to health preoccupatons. It s therefore mportant to know the factors that nfluence populaton health n order to undertake sutable economc polcy. Rodgers (1979) s one of the frst economsts to consder ncome dstrbuton as a determnant of health outcomes. He shows that ncome nequalty nfluences health status not only n developed countres, but also n developng countres, openng the debate about the assocaton between ncome dstrbuton and health. Wlknson (1992) reopens the debate by showng through eleven ndustralzed countres that ncome nequalty s an essental determnant of health status. Even though major part of the studes on ths topc confrm the negatve effect of nequalty on health, some authors reject ths hypothess and show that hgh nequalty may be ndfferent to health status or mprove t (Pampel et Pella 1986 ; Mellor et Mylo, 2001; Deaton, 2003). All the mechansms through whch ncome dstrbuton mpacts health status developed n the lterature show that an ncrease n nequalty worsens populaton health. These mechansms rely on the absolute and relatve ncome hypothess, psychosocal hypothess and neomateralsm hypothess as well. In ths paper we add the envronment as another mechansm through whch ncome dstrbuton could affect health status. Durng the past ffteen years, wth the emergence of envronmental concerns, many studes examne the assocaton between ncome nequalty and natural envronment qualty. But they found dfferent results. On the one hand, some show that more nequalty may mprove envronment qualty (Scruggs, 1998; Ravallon et al., 2000). On the other hand, other studes underlne the negatve mpact of nequalty on envronmental qualty (Boyce, 1994; Torras & Boyce, 1998). If envronmental qualty s degraded by an ncrease n nequalty, t may be a channel that renforces the negatve effect of the other mechansms. But f t s mproved by an ncrease n nequalty, t maybe a mechansm that mtgates or cancels the negatve effect predcted by the other mechansms and justfy the dscrepances between the fndngs. Our results show theoretcally and emprcally that an ncrease n ncome nequalty s assocated to envronmental degradaton and envronment qualty s an mportant determnant of health status. Ths negatve effect of ncome nequalty on envronment qualty s mtgated by good nsttutons. When the effect of envronment qualty on health s taken nto account, the effect of ncome dstrbuton on health decreases and become less sgnfcant statstcally. That s, an ncrease n nequalty worsens populaton s health va envronmental degradaton. The rest of ths paper s organzed n four sectons. Secton 2 revews the lterature on the assocaton between ncome dstrbuton, envronmental degradaton and populaton s health. In ths secton we explan why and how ncome nequalty affects health before ntroducng the arguments that defend the assocaton between ncome dstrbuton and envronmental qualty. Secton 3 develops a theoretcal model n whch ncome dstrbuton affects health through envronmental degradaton. In secton 4, we nvestgate emprcally the effects of ncome dstrbuton on health va envronment qualty. The last secton concludes. 3

2. Lterature revew The relatonshp between ncome nequalty and populaton health has been nvestgated by many macroeconomc studes durng the past 15 years. Scholars examne how and why ncome nequalty affects health theoretcally and emprcally wthn and between natons. We wll frst revew the tradtonal mechansms, namely the ways ncome dstrbuton affects populaton s health already developed n the lterature. Then, we wll explan how ncome nequalty mpacts health through envronmental degradaton. 2.1. Tradtonal effects of ncome nequalty on health Theoretcally, four mechansms are underlned, through whch ncome nequalty can harm drectly populaton health (Mayer & Sarn, 2005). The frst mechansm s the absolute ncome hypothess. In fact, ncome may be an mportant determnant of populaton health, snce t allows them to buy better nutrton or medcal care or reduces ther stress. If the relatonshp between an ndvdual ncome level and ts health status s lnear, an extra unt of ncome wll have the same effect on health regardless of whether t goes to the rch or to the poor. In ths case takng a unt of ncome from the rch and gvng t to the poor wll lower health status among the rch and rase t among the poor by exactly equal amounts, leavng the global health unchanged. The realty s that standard economc models predct that the health gans from an extra unt of ncome should dmnsh as ncome rses (Preston, 1975; Laporte, 2002; Deaton, 2003; Backlund et al., 1996; Babones, 2008), n other words, health should be a concave functon of ncome. That s, a transfer of a unt of ncome from the rch to the poor mght mprove aggregate populaton s health status. The second mechansm developed n the lterature s the relatve ncome hypothess. The effect of economc nequalty s lkely to depend to some extent on the geographc proxmty of the rch to the poor (Mayer & Sarn, 2005). In fact, f people assess ther ncome by comparng themselves to ther neghbours, the ncome of others can affect ther health. The chronc stress provoked by ths comparson may lower resstance to some dseases and cause premature death. For Wlknson (1997), f ndvduals evaluate ther well-beng by comparng themselves to others wth more ncome than themselves, ncreases n economc nequalty wll engender low control, nsecurty, and loss of self esteem. The thrd way developed n the lterature through whch ncome nequalty may worsen populaton health s psychosocal hypothess. Inequalty can mpact health through socal comparsons by reducng socal captal, trust and effcacy (Kawach & Kennedy, 1997; Marmot & Bobak, 2000). Accordng to Wlknson (1996), ncome nequalty worsens health because a low rankng n the socal herarchy produces negatve emotons such as shame and dstrust that lead to worse health va neuro-endocrne mechansms and stress-nduced behavors such as smokng, excessve drnkng, takng dangerous drugs, and other rsky actvtes (Mayer & Sarn, 2005). Lynch et al. (2001) found weak assocatons between a varety of measures of the psychosocal envronment, (dstrust, belongng to organzatons, volunteerng, and effcacy), and nfant mortalty, but they found that economc nequalty s strongly related to nfant deaths. Neo-materalsm hypothess s the fourth mechansm through whch ncome nequalty may harm health status. Accordng to some authors defendng ths dea, ncome nequalty affects health manly through ts effect on the level and the dstrbuton of materal resources (Coburn, 2000 and Lynch, 2000). Ths argument suggests that bad health could be the consequence of an ncrease n ncome nequalty that reduces state spendng on medcal care, goods and servces for the poor. 4

If theoretcally, all the arguments found n the lterature ndcate a negatve mpact of ncome nequalty on health status, emprcal fndngs are far from a consensus. Lynch et al. (2004) revew 98 aggregate and multlevel studes to examne the assocatons between ncome nequalty and health. They conclude that overall, there seems to be lttle support for the dea that ncome nequalty s a major, generalzable determnant of populaton health dfferences wthn or between rch countres. Income nequalty may, however, drectly nfluence some health outcomes, such as homcde n some contexts. Mayer & Sarn (2005) revew ten studes that use cross-sectonal data to estmates the assocaton between economc nequalty and nfant mortalty. Eght (8) of these ten (10) use cross-natonal data and produce eleven (11) estmates. Nne (9) fnd that more unequal countres have hgher nfant mortalty rates, and two (2) (Pampel & Pella, 1986; Mellor& Mlyo, 2001) fnd that more unequal countres have lower nfant mortalty rates than countres wth less nequalty. Wlknson & Pckett (2006) compled one hundred sxty eght (168) analyses n one hundred ffty fve (155) papers reportng research fndngs on the assocaton between ncome dstrbuton and populaton health, and classfed them accordng to how far ther fndngs supported the hypothess that greater ncome dfferences are assocated wth lower standards of populaton health. They fnd that for eghty seven (87) of these studes the coeffcent of ncome nequalty s always statstcally sgnfcant wth the correct sgn. Forty four (44) present mxed results and thrty seven (37) no sgnfcant coeffcent. They explan the dvergence of emprcal fndng by the sze of area, choce of control varables and don t fnd any explanaton for some nternatonal studes. We argue here that n addton to the tradtonal mechansms through whch ncome nequalty degrades populaton s health, found n the lterature, there exsts at least another channel through whch ncome nequalty may affect health, namely envronmental qualty. 2.2. Income nequalty and envronment A large body of research has reported strong assocatons between ncome nequalty and envronmental degradaton: most theoretcal arguments explan how ncome nequalty may mprove envronmental qualty. Frst, ncome nequalty can ncrease envronment protecton through ndvdual preference toward envronmental qualty. In fact, for a gven level of average ncome, greater nequalty means not only hgher ncomes for the rch, but also lower ncomes for the poor. Assumng that the ncome elastcty of demand for envronmental qualty s postve 2, and takng a unt of ncome from the poor and gvng t to the rch ncreases the demand for envronmental qualty of the rch, but at the same tme t decreases the demand of the poor. The net effect on envronmental qualty depends on whether the demand-ncome relaton s lnear, concave or convex (Scruggs, 1998; Boyce, 2003). If ths relaton s lnear, the transfer wll not have any effect on envronmental qualty snce an extra unt of ncome wll have the same effect on envronmental demand regardless of whether t goes to the rch or to the poor. If the envronmental demand s lnked to ncome by a convex (concave) relaton, the transfer of ncome from the poor to the rch wll ncrease (decrease) envronmental demand. It s more convncng to assume that the wealthest prefer more envronmental qualty than the poor for many reasons. Frst, economc theores suggest that the rch prefer less envronmental degradaton than the poor. Ths may be due to the fact that envronmental qualty s a superor good and demand ncreases faster than ncome (Baumol and Oates, 1988). Ths s one of the explanatons behnd the envronmental Kuznets Curve (EKC) hypothess (Grossman & Krueger, 1995). As argued by Scruggs (1998), greater demand for envronmental protecton 2 Ths supposes that envronmental qualty s a normal good 5

among the wealthest s also expected to result n a greater wllngness and ablty to pay for more envronmental protecton. In addton, wealth ncreases ndvduals concern for the future, maybe because they expect hgher lfe expectances than the poorest or because t ncreases ther concern for ther chldren n the future. Another reason to explan why rch prefer more envronmental qualty s that envronmental protests are usually composed of mddle and upper classes, not the poor (Dalton, 1994). Income nequalty can also reduce envronmental degradaton through the margnal propensty to emt (MPE) as argue by Ravallon et al. (2000). Accordng to these authors, each ndvdual has an mplct demand functon for carbon emssons snce the consumpton of almost every good mples some emssons ether drectly va consumpton or ndrectly va ts own producton. They call margnal propensty to emt (MPE) the dervatve of ths mplct demand functon wth respect to ncome. If poor people have a hgher (lower) MPE than rch ones, a redstrbuton polcy that reduces nequaltes wll ncrease (decrease) carbon emssons. One can assume that the poorests have hgher MPE than wealthests, frst because less emsson goods need hgh technology and are thus generally expensve. Therefore, the poorest cannot afford t. In addton, poor tend to use energy less effcently than the rch, whch entals a hgher MPE (Ravallon et al., 2000). If these arguments predct an mprovement of envronment qualty channelled by ncome nequalty, t s also largely argued by some authors that nequalty may degrade envronment rather than mprovng t. Boyce (1994) s the frst author to examne how ncome nequaltes affect envronmental degradaton. He supports the hypothess that greater nequalty may ncrease envronmental degradaton and ths for two reasons. Frst, he argues that a greater nequalty ncreases the rate of envronmental tme preference for both poor and rch. In fact, when nequalty ncreases, the poor tend to overexplot natural captal, because they perceve t as the only resource they have and the only source of ncome that can help them secure ther survval. In addton, economc nequalty often provokes poltcal nstablty and rsks of revolts. Ths leads rch people to prefer a polcy that conssts n explotng the envronment and nvestng the returns abroad rather than nvestng n the protecton of local natural resources. Therefore, for Boyce an ncrease n nequalty nduces both rch and poor to degrade more ther own envronment. The second argument put forward concerns the power of the rch. Boyce (1994) argues that n a socety wth greater nequalty, rch people are lkely to have large poltcal power and can heavly nfluence decsons on envronmentally damagng projects. Such decsons are based on the competton between those who beneft from the envronmentally degradng acton and those who bear the costs of t. Boyce (1994) argues that rch people are generally the wnners, whle poor people tend to be the losers of the nvestments that have an ecologcal mpact. Therefore, economc nequalty favours the mplementaton of envronmentally damagng projects and nvestments snce t renforces the power of the rch to mpose envronmental costs on the poor (Ravallon et al., 2000, p.6). Scruggs (1998) has crtczed the hypotheses supported by Boyce. He states that the nfluence va cost-beneft analyss s based on two wrong assumptons. Frst, accordng to Scruggs, evdence ndcates that better off members of socety tend to have hgher envronmental concern than those wth lower ncome (Scruggs, 1998, p.260). Moreover Boyce (1994) assumes that a democratc socal choce crteron leads to hgher envronmental protecton than a non-democratc decson process (.e. a power-weghted socal decson rule), whle evdence suggests that ths s not necessarly true. Another theoretcal argument to explan why more nequalty leads to more degradaton s developed by Borghes (2000). He argues that much of the theoretcal envronmental lterature has stressed the need of cooperatve solutons to envronmental problems. In an unequal socety ths s more dffcult to acheve than n an equal socety snce there are 6

generally more conflcts among the poltcal agents (government, trade unons, lobbes etc...) on many socal ssues. In ths sense, greater nequalty can contrbute to ncrease envronmental degradaton (Borghes, 2000). In addton to these arguments, some theoretcal model supports the envronmental degradng effect of ncome nequalty. It s the case of Magnan (2000) who examnes the mpact of ncome dstrbuton on publc research and development expendtures for envronmental protecton. Through a model n whch socal decsons are determned by the preferences of the medan voter, she hypotheszes that ncome nequalty reduces pro-envronmental publc spendng due to a relatve ncome effect, and hgher nequalty shfts the preferences of those wth below-average ncome n favour of greater consumpton of prvate goods and lower expendture on envronmental publc goods. Marslan and Renström (2000) have also recently nvestgated how ncome dstrbuton affects poltcal decsons on envronmental protecton. Through an overlappng-generatons model, they show that the hgher the level of nequalty n terms of medan-mean dstance, the lower the polluton tax set by a majorty elected representatve. Therefore, nequalty nduces redstrbuton polces that dstort economc decsons and lower producton. Inequalty may be negatvely correlated wth envronmental protecton as t leads to less strngent envronmental polces. It s a pror dffcult to predct the effect of ncome dstrbuton on envronment qualty theoretcally even though degradng effect seems n our vewpont more convncng. Let us see emprcal fndngs. Many authors have emprcally studed the relaton between ncome dstrbuton and envronment qualty and ther conclusons are qute not consensual. In appendx 1, we report nne (9) mportant papers and thrty one (31) studes on the assocaton between ncome dstrbuton and envronment qualty. Among these studes, ten (10) conclude that nequalty mproves envronment qualty, nne (9) fnd the opposte concluson and twelve (12) don t fnd any sgnfcant assocaton. Let explore some of them. Scruggs (1998) performs two cross-country emprcal analyses to assess the effect of ncome nequalty on the envronment through pooled models. In the frst one, four dfferent pollutants (sulphur doxde, partculate matter, fecal colform and dssolved oxygen) are used as dependant varable n a panel of 22 up to 29 countres. The second nvestgaton examnes the mpact of several varables on a composte ndex of envronmental qualty n a panel of 17 OECD countres. Ths ndex s constructed by combnng fve polluton ndcators. In the frst case, he fnds conflctng results: greater nequalty mproves envronmental qualty for one envronmental ndcator (partculates), whereas the opposte holds for the other ndcator (dssolved oxygen). For the other ndcators (sulphur doxde, fecal colform), the coeffcents are not statstcally sgnfcant. In the second analyss, ncome nequalty decreases envronmental degradaton. Through a panel of 42 countres n the perod 1975-92, Ravallon et al. (2000) frst estmate CO2 emssons as a cubc functon of average per capta ncome and of populaton and tme trend. They estmate ther equaton wth fxed effect model and smple pooled model usng ordnary least squares. They conclude that hgher nequalty wthn countres reduces carbon emssons. However, the mpact of ncome dstrbuton on the envronment decreases at hgher average ncomes. Borghes (2000) performs an emprcal analyss smlar to that of Ravallon et al. (2000). He uses CO2 per capta as envronmental varable and Gn from Dennger and Squre as ncome nequalty ndcator wth a panel of 37 countres from 1988-1995. In the pooled OLS model, an ncrease n nequalty lowers CO2 emssons, whereas t does not have a sgnfcant mpact on CO2 emssons accordng to the FE model. 7

Magnan (2000) assessed the mpact of nequalty on R&D expendtures for the envronment taken as proxy for the ntensty of publc engagement n envronmental problems through pooled ordnary least squares and random effects estmatons. Usng a panel of 19 OECD countres n the perod 1980-1991, he fnds that hgher nequalty reduces envronmental care, however, the effect s statstcally sgnfcant at 5% level n the pooled ordnary least squares model only. Usng the prncpal components analyss, Boyce et al. (1999) estmate statstcally a measure of nter-state varatons n power dstrbuton based on voter partcpaton, tax farness, Medcad accessblty, and educatonal attanment levels. They fnd that ncome nequalty, per capta ncome, race, and ethncty affect power dstrbuton n the expected drectons. Inequalty n power dstrbuton s assocated wth lower envronmental polces, and these n turn are assocated wth hgher envronmental stress. Both envronmental stress and power nequalty are assocated wth adverse publc health outcomes. Torras and Boyce (1998) examne the effect of ncome dstrbuton on a set of water and ar polluton varables usng the Global Envronment Montorng System (GEMS) data, Gn ndex, adult lteracy rates and an aggregate of poltcal rghts and cvl lbertes. Wth a OLS estmaton, they obtan mxed results on the envronmental mpact of ncome nequalty. The Gn coeffcent s postve for some envronmental ndcators and negatve for others. It s also possble that more envronmental degradaton ncreases ncome nequalty. In fact, envronmental degradaton n many ways affects the lvelhood of the poor. The poorest are vulnerable to envronmental degradaton snce they depend heavly on natural resources and have less alternatve resource. They are also exposed to envronment hazards and are less capable of copng to envronmental rsks (Dagusta and Mäler, 1994; World Bank, DFID, EC, UNDP, 2002). Furthermore, the rch are more capable of lookng after themselves from envronmental dseases than the poorest. 3 Ths revew explans the complexty of the relaton between ncome dstrbuton and envronment. Fgure 1 summarzes the relaton lnkng ncome nequalty and populaton s health. Fgure 1: Relatons between Income level, Income Inequalty, Ecologcal Degradaton and Health Income level Income nequalty Health status Envronmental degradaton Source : author 3 Ths s not the object of the present study. 8

3. The model The purpose of ths model s to assess theoretcally how ncome nequalty affects health status through the level of polluton abatement expendtures. It conssts n the ntroducton of health varable n Magnan s model 4. Let us assume an addtvely separable utlty functon for ndvdual : U = c + α h ( Q) (3.1) Where c s the level of consumpton of a prvate good and h s the health status of ndvdual. We consder health not merely as absence of llness or nfrmty, but also as a state of complete physcal, mental and socal well-beng. h s postvely lnked to envronment qualty Q (a pure publc good) and the effect of envronment on health s the same for every ndvdual ( ( h ) ( Q) = k ). α s the contrbuton of health to s utlty. It expresses also the preference for envronment qualty as n Magnan s model because f the contrbuton of health n ndvdual s utlty s hgh, he wll prefer a better envronment qualty n order to mprove hs health. Furthermore, n ths model, health s wdely defned. The publc good nature of Q mples that envronmental polcy E s necessary to solve market falure, that s Q = Q( E), where E s publc expendtures for envronmental care, and Q '(.) > 0. Envronmental care s fnanced through taxaton by a fracton τ y of ndvdual 2 ncome y and we have: E = Y ( τ τ 2), where τ s the envronmental tax rate ( τ (0,1) ) and Y s the average ncome 5. In ths economy, ndvduals dffer by personal ncome levels and ncome s dstrbuted accordng to a unmodal functon f ( y ) where y (0, yh ) and y H s the maxmum level of personal ncome. Income nequalty mples that the majorty of the populaton has ncome below the average and ( ym Y ) < 1, where y m s the medan ncome of the dstrbuton f ( y ). We assume that α, the preference for envronment qualty and the contrbuton of health to utlty s postvely correlated wth the ndvdual relatve ncome R = ( y Y ). Ths assumpton s crucal for our analyss. That s, α = α ( R ) and α ' (.) > 0. The margnal rate of substtuton between c and h depends on ndvdual relatve ncome. Ths assumpton s supported by some theoretcal and emprcal studes (Ng and Wang 1993, Konrad 1996 and Magnan 2000). The ndrect utlty functon for the ndvdual can be wrtten as: 2 V = (1 τ ) y + αh Y ( τ τ 2) (3.2) The optmal tax rate for ndvdual s obtaned by dfferentaton of (3.2) wth respect to τ * and we have: τ = 1 (1 α k) R. The margnal effect of relatve ncome of ndvdual on hs deal tax rate s: * ' 2 ( τ ) ( R ) ( α αr ) ( kα ) = + = (1 kα )[ 1 ( α R )( R α )] +. Ths * effect s postve ( ( τ ) ( R ) > 0 ) f the relatve ncome elastcty of the preference for 4 Magnan, E., Ecologcal Economcs, 32 (2000) 431-443 5 The functonal form for publc envronmental protecton s qute general and expresses envronmental cost of publc funds (Magnan 2000). 9

envronmental care ε s more than 1, or ( α R )( R α ) > 1. For ε <1, the optmal tax rate for ndvdual s a decreasng functon of relatve ncome. If we are n a democracy wth majorty votng system, the poltcan wll maxmze the ndrect utlty functon of the medan voter accordng to the medan voter theorem. The optmal tax rate chosen by the economy wll be that of the medan voter and we have: * τ = 1 (1 α k) R (3.3) m m Where m s the ndex for the medan voter. Ths equaton (3.3) shows that the equlbrum level of envronmental abatement expendture s functon of ncome dstrbuton. E = E ( Y, y Y ) = Y ( τ ( τ ) 2) (3.4) * * * * 2 m And the margnal effect of ncome dstrbuton on the optmal taxaton rate s gven by: τ = α + α α = (1 kα )[ 1 + ( α R )( R α )] (3.5) * ' 2 ( ) ( Rm ) ( m mrm ) ( k m) m m m m m ' Where α = ( α ) ( R ) s by assumpton postve. m m m The margnal effect of ncome nequalty * E s gven by: R m on the optmal envronmental publc expendture * * * ( E ) ( Rm ) = Y ( τ ) ( Rm ) 1 τ (3.6) * τ * * (0,1), therefore 1 τ > 0. The sgn of ( E ) ( R m ) only depends on the sgn of * ( τ ) ( ). Envronmental publc expendture s an ncreasng functon of ncome equalty m R m R f (1 kα )[ 1 ( α R )( R α )] + >0 and ths condton holds f the relatve ncome m m m m m elastcty of the preference for envronment care of the medan voter s greater than one ( ( α R )( R α ) > 1). m m m m Ths result shows that ncome nequalty affects negatvely envronmental publc expendture and therefore populaton s health. 4. Emprcal analyss 4.1. Estmatons The analyss s subdvded nto three steps. We examne, frst, the mpact of ncome nequalty on envronmental qualty. Then, we study the assocaton between envronment qualty and health status. Fnally, we compare the effect of ncome dstrbuton on populaton s heath n presence and n absence of envronmental varables. The econometrc relaton between nequalty and envronment can be wrtten as: envronment = λ + β EHII + δ X + ε (4.1) t t k kt t 10

Where envronment and EHII represent respectvely the logarthm of envronment qualty and ncome nequalty measure. X s the matrx of the control varables. The country fxed k effects are represented by λ and ε t s the error term. Ths equaton could be estmated by the Ordnary Least Squares (OLS), but t s very lkely that envronmental degradaton ncreases ncome nequalty as explaned n secton 2. Ths potental smultanety can be a source of endogenety. Another source of endogenety could arse from the measurement error of our nequalty ndcator. In order to solve ths problem, we defne as nstrumental varable the dependency rato and we estmate equaton (4.1) wth the Two Step Least Square (2SLS) method. As a proxy for demographc varable, age dependency rato s an mportant determnant of ncome nequalty because of ts dstrbutve effect and t s less convncng to ague that t affects drectly envronment qualty. To control for the effect of ncome nequalty dependng on development level and nsttuton qualty, we progressvely, add to equaton (4.1), the nteracton of ncome nequalty wth development level dummy and nsttuton qualty. In the second model, health status s expressed as a functon of envronment qualty and other explanatory varables. Health = η + γ envronment + θ Z + ω (4.2) t t k kt t Where health represents health status measure and Z t s the matrx of the control varables. η represents the country fxed effects and ω t s the error term. Equaton (4.2) s estmated wth standard fxed effects estmaton. The thrd model expresses health status as a functon of ncome nequalty wth and wthout consderaton of envronmental varables. The coeffcent of EHII must decrease wth the addton of envronmental varables f ts effect s n part channelled by these varables. Health = φ + ψ EHII + ρenvronment + σ Z + τ (4.3) t t t k kt t Ths equaton could be estmated by the Ordnary Least Squares (OLS), but t s very lkely that populaton s health affects ncome nequalty through productvty, educaton and other factors. Ths potental smultanety can be a source of endogenety. To solve for ths problem, we estmate equaton (4.3) wth the Generalzed Method of Moments (GMM system). 4.2. Data and varables The data used n ths paper cover the perod 1970-2000 subdvded nto 6 perods of 5 years and we retan for the basc regresson 90 developed and developng countres (because of data avalablty). As health varable we use the logt of under fve survval rate (LOGIT SURVIVAL). The under-fve survval ndcator s lmted asymptotcally, and an ncrease n ths ndcator does not represent the same performance when ts ntal level s weak or hgh. The best functonal form to examne that s where the varable s expressed nto a logt form, as Grgorou (2005) underlned. survval log t survval= ln( ). 1 survval Data on under fve mortalty rates are from the World Health Organzaton (WHO). The envronmental qualty s represented by three varables: the carbon doxde emsson per GDP (CO2), the bologcal oxygen demand (BOD) both taken from WDI 2007 and the sulphur doxde emsson per GDP (SO2) from Stern (2005). For these varables, a hgher 11

value ndcates more envronmental degradaton. CO2 and SO2 are ar polluton ndcators and BOD n a water qualty ndcator. Income nequalty s measured by the Gn coeffcent taken from the database created by Galbrath and assocates and known as the Unversty of Texas Inequalty Project (UTIP) database. It contans two dfferent types of data on nequalty: the UTIP-UNIDO and the EHII ndexes. The EHII (that we use here) s an ndex (rangng from 0, low nequalty to 1, hgh nequalty) of Estmated Household Income Inequalty and s bult combnng the nformaton n the Dennger and Squre (D&S) data wth the nformaton n the UTIP-UNIDO data. The other varables used are gross domestc product per capta (GDPCAP), populaton densty (POPDENS), fertlzer use (FERTILIZER), foregn drect nvestment (FDI), dependency rato (DEPENDENCY) and trade openness (OPEN), all taken from WDI 2007 and prmary school enrolment (SCHOOL) from Barro and Lee (2000). Appendx 2 summarzes the characterstcs of the mportant varables. Ths table shows the mean, the mnmum, the maxmum, the standard devaton and the coeffcent of varaton of each varable. These statstcs are completed by appendx 3 whch presents the correlaton between mportant varables. These statstcs are confrmed by appendx 7, whch dsplays the statstcal relaton between EHII and envronmental varables. These relatons are just a smple correlaton and don t take nto account the nfluence of other varables. The econometrcal secton wll solve for ths. 4.3. Results 4.3.1. Income nequalty and envronment The results obtaned from equaton (4.1) for the whole sample of developed and developng countres (the relaton between nequalty and envronment qualty), are reported n table 1. The column 1 of ths table shows the results when the logarthm of carbon doxde emsson per GDP (CO2) s used as envronmental varable. The envronmental Kuznets Curve (EKC) hypothess s verfed, snce the coeffcent of the logarthm of GDP per capta (GDPCAP) s postve and statstcally sgnfcant, and the coeffcent of ts square (GDPCAPSQ) s negatve and also sgnfcant. In ths column, the coeffcent of nequalty varable (EHII) s postve and statstcally sgnfcant at 5%, showng that an ncrease n ncome nequalty worsens envronmental qualty. Columns 2 and 3 summarze the results when sulphur doxde emsson per GDP (SO2) and the bologcal demand (BOD) are respectvely used as envronmental varables. The mportant results reman unchanged, namely, ncome nequalty s an mportant cause of envronment degradaton, except for SO2 where the coeffcent of nequalty s not statstcally sgnfcant. We estmated agan equaton (1) by addng as addtonal varable, the nteracton between ncome nequalty and economc development level dummy to assess the dfferental effect of ncome nequalty dependng on development level. The results obtaned are summarzed n the frst three columns (1, 2 and 3) of appendx 4. The relatonshp between ncome nequalty and envronment s confrmed for CO2 n the frst column. In ths column, the coeffcent of the nteracton term s negatve and statstcally sgnfcant. Ths result shows that ncome nequalty ncreases CO2 emsson but the effect s hgher n developed countres. For SO2 emsson n column (2), only the coeffcent of the nteracton term s statstcally sgnfcant and postve showng that ncome nequalty ncreases SO2 emsson only n developng countres. Fnally for BOD n column (3), we have not any effect. To take nto account the role played by nsttutons qualty on the nequalty effect, we add as addtonal varable, the nteracton between nsttuton and nequalty. The results are presented n the last three columns (4, 5 and 6) of appendx 4. These results show that good 12

nsttutons mtgate the negatve effect of ncome nequalty on envronment qualty, but ths effect s only sgnfcant statstcally for SO2 emsson n column (5). Table 1: Impact of ncome nequalty on envronment qualty: 2SLS FIXED EFFECTS ESTIMATIONS DEPENDENT VARIABLES INDEPENDENT (1) (2) (3) VARIABLES CO2 SO2 BOD EHII 4.405** 2.819 9.580* (2.387) (0.673) (1.736) GDPCAP 0.969*** 3.479*** 1.893** (2.653) (4.218) (2.298) GDPCAPSQ -0.0723*** -0.253*** -0.164*** (-3.220) (-4.977) (-3.192) POPDENS -0.130-0.870** -1.366*** (-0.859) (-2.522) (-2.888) SCHOOL 0.108 1.574*** 0.125 (0.643) (4.120) (0.325) FERTILIZER 4.52e-05* 0.000177*** 0.000166** (1.786) (3.033) (2.205) INSTITUTION -0.00104-0.00976*** -0.00491 (-0.801) (-3.290) (-1.587) FDI -0.317-0.878-0.327 (-0.708) (-1.019) (-0.370) OPEN -0.0778-0.150-0.165 (-0.896) (-0.764) (-0.814) CONSTANT -3.895** -28.95*** -16.40*** (-2.411) (-7.976) (-4.550) Observatons 367 367 365 NB countres 86 86 88 ***sgnfcant at 1%, **sgnfcant at 5%, *sgnfcant at 10%. t-statstcs enter parenthess. Income nequalty (EHII) s nstrumented by dependency rato. The frst step estmaton results are presented n appendx 5. 4.3.2. Envronment and health The effect of envronmental qualty on health status (equaton 4.2) s estmated wth standard fxed effects model and the results are reported n table 2. Column 1 presents the results when envronment qualty s measured by CO2 emsson. All the explanatory varables have expected sgn and are statstcally sgnfcant, except the prmary school enrollment lagged (SCHOOL(1)) whch s not statstcally sgnfcant. GDP per capta lagged (GDPCAP(1)) and mmunzaton rate (IMDPT) mprove the survval rate whle fertlty rate (FERT) and envronment qualty (BOD) degrades t. The negatve and sgnfcant coeffcent of CO2 shows that ar polluton worsens health status as expected n the lterature revew. Columns 2 and 3 shows the results when SO2 and BOD are respectvely used as envronmental ndcators. All these columns underlne the negatve effect of ar and water polluton on populaton s health. 13

Table 2: Impact of envronment qualty on health OLS FIXED EFFETS ESTIMATION INDEPENDENT Dependent varable: logt of under fve survval rate VARIABLE (1) (2) (3) GDPCAP(-1) 0.396*** 0.290*** 0.282*** (6.223) (3.640) (3.883) IMDPT 0.502*** 0.474*** 0.532*** (5.710) (5.195) (5.632) SCHOOL(-1) -0.310-0.206-0.441 (-1.206) (-0.779) (-1.532) FERT -0.202*** -0.178*** -0.153*** (-5.933) (-4.835) (-4.343) CO2-0.223* (-1.949) SO2-0.209*** (-8.060) BOD -0.237*** (-4.711) CONSTANT 0.340-3.056*** -2.073*** (0.582) (-4.711) (-3.088) Observatons 434 429 373 NB countres 97 96 93 ***sgnfcant at 1%, **sgnfcant at 5%, *sgnfcant at 10%. t-statstcs enter parenthess. 4.3.3. Income nequalty, envronment and health The effects of ncome nequalty on health status wth and wthout consderaton of envronment varables (equaton 4.3) are summarzed n table 3. Column (1) of ths table presents the results wthout consderaton of envronment qualty. Each varable has the expected sgn. Income nequalty affects negatvely and sgnfcantly populaton health. In the other columns (2, 3 and 4) of ths table, we ntroduce envronment qualty n the model. All the envronmental varables affect negatvely health status. In addton, the ntroducton of envronmental varables decreases the level and the statstcal sgnfcance of the coeffcent of ncome nequalty varable n each column. Ths confrms the channel role played by envronmental qualty concernng the effect of ncome dstrbuton on populaton health. 14

Table 3: Impact of ncome nequalty and envronment qualty on health GMM System estmaton results Dependent varable: logt of under fve survval rate Independent varables (1) (2) (3) (4) GDPCAP 0.799*** 0.774*** 0.766*** 0.495** (11.98) (11.14) (11.68) (2.455) IMDPT 0.547*** 0.550*** 0.585*** 0.500*** (4.678) (4.646) (5.112) (4.090) SCHOOL 0.180 0.482 0.230 0.264 (0.650) (1.579) (0.856) (1.469) FERT -0.125*** -0.147*** -0.119*** -0.226*** (-3.535) (-4.147) (-3.591) (-3.527) EHII -1.400** -1.200* -1.302** -1.103 (-2.144) (-1.709) (-2.067) (-1.133) CO2-0.217** (-2.050) SO2-0.0498** (-2.175) BOD -0.224* (-1.746) CONSTANT -3.109*** -2.916*** -3.901*** -3.243*** (-3.600) (-3.248) (-4.675) (-2.614) Observatons 360 359 357 354 NB countres 90 89 88 90 Sargan OID test (p.value) 0.12 0.34 0.10 0.25 AR(2) 0.58 0.63 0.69 0.56 ***sgnfcant at 1%, **sgnfcant at 5%, *sgnfcant at 10%. t-statstcs enter parenthess. 5. Concluson The purpose of ths paper was to nvestgate the effect of ncome dstrbuton on health whch passes through envronmental qualty. Theoretcally, we show that envronment degradaton could be consder as a channel through whch ncome nequalty affects populaton health n addton to the drect mechansms found n the lterature. Ths effect could renforce the negatve effect of ncome nequalty on health. Emprcally, we show through an econometrc analyss that ncome nequalty affects negatvely envronmental qualty and envronment degradaton worsens populaton s health. Ths negatve effect of ncome nequalty on envronment qualty s mtgated by good nsttutons. Another nterestng result s that ncome nequalty affects negatvely health status and n presence of envronmental varable, the level and the statstcal sgnfcance of the coeffcent of ncome nequalty varable decrease. Ths confrms that envronment qualty s an mportant channel through whch ncome nequalty affects populaton health. These results hold for ar polluton ndcators (CO2 and SO2) and water polluton ndcator (BOD). It s also robust for rch and developng countres. As polcy mplcaton, our results mean that ncome nequalty s bad for health and envronment, and countres wth hgh ncome nequalty may mplement dstrbutve polcy n order to avod ts negatve mpact on health. 15

Next studes could extend our fndng s takng t agan at ndvdual level (mcroeconomcs). Another way to extend ths artcle s to verfy t conclusons for other envronmental and nequalty varables. Bblography: Anand, S. & Ravallon, M., 1993. Human Development n Poor Countres: On the Role of Prvate Incomes and Publc Servces, Journal of Economc Perspectves 7, 133-50. Babones, S.J., 2008, Income nequalty and populaton health: Correlaton and causalty, Socal Scence & Medcne 66, 1614e1626 Backlund, E., Sorle P. and Johnson N., 1996. The shape of the relatonshp between ncome and mortalty n the Unted States: evdence from the natonal longtudnal mortalty survey. Annals of Epdemology, 6 pp.12-20 Baumol, W. & Oates, W., 1988. The Theory of Envronmental Polcy. Cambrdge Unversty Press, Cambrdge, UK. Bobak, M., Pkhart, H., Rose, R., Hertzman, C., and Marmot, M. (2000). Socoeconomc factors, materal nequaltes, and perceved control n self-rated health: cross-sectonal data from seven post-communst countres. Socal Scence & Medcne, 51, 1343-1350. Borghes, S., 2000, Income Inequalty and the Envronmental Kuznets Curve, NOTA DI LAVORO 83.2000 Boyce, J. K., Andrew R., Klemer, Paul H., Templet, and Cleve E. W., 1999, Power Dstrbuton, the Envronment, and Publc Health: A State-level Analyss, Ecologcal Economcs 29: 127-140. Boyce, J.K. 2003, Inequalty and Envronmental Protecton, Poltcal Economy Research Insttute Workng Paper, n 52. Boyce, R., 1994. Inequalty as a cause of envronmental degradaton, Ecologcal Economcs 11, 169 178. Clément, M. & Meune, A., 2008, Economc Growth, nequalty and envronment qualty: An emprcal analyss appled to developng and transton countres, Cahers du GREThA n 2008-13 Coburn, D., 2000, Income nequalty, socal coheson and the health status of populatons: the role of neo-lberalsm, Socal Scence and Medecne, 51 pp.135-146 Dasgupta, P., & Maler, K, 1994. Poverty, nsttutons, and the envronmental-resource base. Washngton, D.C.: World Bank. Dalton, R.,1994. The Green Ranbow: Envronmental Groups n Western Europe. Yale Unversty Press, New Haven, CT. Deaton, A., 2003. Health, nequalty, and economc development, Journal of Economc Lterature, 41 pp. 113-158 Glazer, A. and Konrad, K.A., 1996. A sgnallng explanaton for charty. Am. Econ. Rev. 86, 1019 1029. Grgorou, C., 2005, Essas sur la vulnérablté des enfants dans les pays en développement: l mpact de la poltque économque, Thèse pour le doctorat ès scences économques, Unversté d Auvergne, Centre d Etudes et de Recherches sur le Développement Internatonal Grossman, G. et Krueger A.B., 1995, Polluton Growth and the envronment, Quaterly Journal of Economcs, 110, 353-377 Herrng, N., Mulatu A. and, Bulte E., 2001, Income nequalty and the envronment: aggregaton bas n envronmental Kuznets curves, Ecologcal Economcs 38, 359 367 16

Kawach, I. & Kennedy B. P. 1997, Socal Captal, ncome nequalty, and mortalty, Amercan Journal of Publc Health, 87(9), pp. 1491-1498 Knowles, S. & Owen P.D., 1994, Health Captal and Cross-country varaton n ncome per capta n the Mankw-Romer-Wel model, Economcs Letter, vol. 48(1), 99-106 Laporte, A. 2002. A note on the use of a sngle nequalty ndex n testng the effect of ncome dstrbuton on mortalty, Socal Scence and Medecne, 55 pp. 1561-1570 Lynch, H. W., Smth, G, Kaplan G. and House J. S. 2000. Income nequalty and mortalty: Importance to health of ndvdual ncome, psychosocal envronment and materal condton, Brtsh Medcal Journal, 320, pp. 1200-4. Lynch, H. W., Smth, G. D., Hllemeer, M., Shaw, M., Raghunathan, T., and Kaplan, G., A. 2001. Income Inequalty, the psychosocal envronment, and health: comparson of wealthy natons. The Lancet, 358, 194-200. Lynch, J., Smth, G. D., Harper, S., Hllemeer, M., Ross, N., and Kaplan, G. A., 2004, Is ncome nequalty a determnant of populaton health? Part 1. A systematc revew. Mlbank Quarterly, 82, 5-99. Magnan, E., 2000, The envronmental Kuznets curve, envronmental protecton polcy and ncome dstrbuton, Ecologcal Economcs, vol.32, pp.431-443. Marslan, L. & Renström T.I., 2000, Inequalty, envronmental protecton and growth, CentER workng paper n.2000-34, Tlburg Unversty, The Netherlands. Marslan, L. & Renström T.I., 2000, Inequalty, envronmental protecton and growth, CentER workng paper n.2000-34, Tlburg Unversty, The Netherlands. Mayer, S. E. & Sarn A., 2005. Some mechansms lnkng economc nequalty and nfant mortalty. Socal Scence and Medecne, 60, pp.439-455. Mellor, J & Mlyo J., 2001, Reexamnng the evdence of an ecologcal assocaton between ncome nequalty and health. Journal of Health Poltcs, Polcy and Law, 26, (3), pp. 487-522. Ng, Y.-K. and Wang, J., 1993. Relatve ncome, aspraton, envronmental qualty, ndvdual and poltcal myopa. Math. Socal Sc. 26, 3 23. Pampel, F. & Pella, V., 1986, Patterns and determnants of nfant mortalty n develeloped natons, Demography, 23(4), pp. 525-541. Preston, S. H., 1975, The Changng Relaton Between Mortalty and Level of Economc Development, Populaton Studes, 29(2), 231-248. Ravallon, M., Hel M. and Jalan J., 2000, Carbon emssons and ncome nequalty, Oxford Economc Papers, 52:651-669 Rodgers, G. B., 1979, Income and Inequalty as determnants of mortalty: An nternatonal cross-secton analyss. Populaton studes, 32 pp.343-351 Scruggs, Lyle A., (1998), Poltcal and economc nequalty and the envronment, Ecologcal Economcs, 26, 259-275 Torras, M. & Boyce, J.K., 1998. Income, nequalty, and envronmental qualty: An nternatonal cross-sectonal analyss. Ecologcal Economcs 25, 147 160. Wlknson, R. G. 1996, Unhealthy Socetes: the Afflctons of Inequalty, London, England: Routledge. Wlknson, R. G., 1997. Income nequalty and socal coheson, Amercan Journal of Publc Health, (87) pp. 104-106 Wlknson, R. G., & Pckett, K. E., 2006. Income nequalty and populaton health: a revew and explanaton of the evdence. Socal Scence & Medcne, 62, 1768-1784. Wlknson, R.G., 1992, Income Dstrbuton and Lfe Expectancy, Brtsh Medcal Journal, 304, p. 165-168 World Bank, Department for Internatonal Development (DFID) UK, European Commsson (EC) & Unted Natons Development Program (UNDP), 2002, Lnkng Poverty 17

Appendces: study Clément and Meune 2008 year nequalty varable Herrng, N., Mulatu 2001 gn ndex A. and, Bulte E. Borghes 2000 Marslan and Renström 2000 Magnan 2000 Appendx 1: lterature revew envronment measure effect of nequalty effect sg. Level SO2 emsson mpr. 10% gn WIDER BOD emsson degr. 1% Gn (Dennge r and Squre) rato of househol ds ranked at top 90th percentle to the medan househol d quntles 1 / quntles 4 gn access to safe water, access to santaton, and deforestaton carbon doxde emssons, ntrogen depleton, and phosphorus depleton degr. 1% mpr. 1% sulfur doxde and mpr. NO partculate concentratons CO2 per capta sulfur, Ntrogen oxdes and carbon doxde Publc R&D expendture for envronmental protecton mpr. 1% degr. NO degr. mpr. degr. 1% 10% NO data 83 developng and transton countres n 1988-2003 16- country sample of sub- Saharan Afrcan countres panel of 37 countres from 1988-1995 two panels of 7 and 10 ndustral zed countres over 1978-1997 17 developed countres estmator revew OLS pooled OLS pooled model fxed effects smple OLS ML fxed effects fxed effects & random effects Cahers du GRETh A n 2008-13 Ecolog cal Econo mcs 38, 359 367 NOTA DI LAVOR O 83.2000 CentER workn g paper n.2000-34 Ecologc al Econom cs 32 (2000) 440 431 443 other covarates GDP, GDP², GDP 3 GDP, GDP² GDP, GDP², GDP 3, Populaton densty, ndustry share. GDP GDP, GDP², Tme trend 18

study Ravallon M., Hel M., Jalan Boyce et al. 1999 Scruggs L.A. Torras and Boyce year nequalty varable 2000 gn ndex 1998 1998 power nequalty Gn (Dennge r and Squre) gn (low ncome) gn (hgh ncome) envronment measure CO2 per capta emsson envronment polcy effect of nequalty effect sg. Level mpr. 5% degr. 1% sulfur doxde mpr. 1% partculate mpr. NO matter fecal colform degr. NO dssolved oxygen degr. 1% Sulfur doxde degr. 1% Smoke degr. 1% Heavy partcles mpr. 1% Dssolved mpr. 1% oxygen Fecal colform mpr. NO Safe water (%) degr. 1% Santaton (%) degr. NO Sulfur doxde mpr. 1% Smoke mpr. NO Heavy partcles degr. NO Dssolved degr. NO oxygen Fecal colform mpr. 1% Safe water (%) degr. NO Santaton (%) degr. NO data panel of 42 countres n the perod 1975-92 50 US states n 1990's 25 29 countres for 3 perods: 1979 1982, 1983 1986 and 1987 1990 287 statons n 58 countres estmator revew fxed effects & pooled OLS OLS OLS pooled model OLS Oxford Econo mc Papers, 52:651-669 Ecologc al Econom cs 29 (1999) 127 140 Ecologc al Econom cs 26 (1998) 259 275 Ecologc al Econom cs 25 (1998) 147 160 other covarates GDP, GDP², Populaton manufacturn g share, urbanzaton and populaton densty Democracy, Income, Industralze ste, perode GDP, GDP², GDP 3, lteracy rate, rght 19

Appendx 2: descrptve statstcs MEAN MINIMUM MAXIMUM COEF. VAR. STAND. DEV. NB. OBS. LOGIT SURVIVAL 2.988 0.672 5.293 0.406 1.214 478 CO2 0.448 0.020 2.255 0.747 0.335 436 BOD 2.34e-06 2.29e-07 0.00002 1.034 2.42e-06 369 SO2 8.18e-09 5.64e-12 2.99e-07 3.320 2.72e-08 485 EHII 0.417 0.266 0.642 0.147 0.061 485 GDPCAP 6280 122.6 36160 1.261 7922 485 SCHOOL 0.304 0 0.93 0.889 0.271 485 IMDPT 0.710 0.012 0.99 0.350 0.249 351 FERT 3.997 1.18 8.494 0.492 1.968 485 POPDENS 98.713 1.567 951.97 1.265 124.89 485 FERTILIZER 1681.06 0.896 37358 2.201 3700.6 485 Appendx 3: correlatons between mportant varables LOGIT SURVIVAL CO2 BOD SO2 EHII GDPCAP SCHOOL IMDPT FERT POPDENS LOGIT SURVIVAL 0.94* LIFE EXPECT 0.30* 1.00 CO2-0.45* 0.01 1.00 BOD -0.19* 0.06 0.20* 1.00 SO2-0.62* -0.17* 0.13* 0.11* 1.00 EHII 0.81* 0.17* -0.47* -0.14* -0.61* 1.00 GDPCAP -0.86* -0.29* 0.33* 0.12* 0.52* -0.63* 1.00 SCHOOL 0.64* 0.17* -0.20* -0.03* -0.30* 0.44* -0.59* 1.00 FERT -0.90* -0.30* 0.32* 0.22* 0.57* -0.68* 0.84* -0.61* 1.00 POPDENS 0.17* -0.01 0.12* -0.11* -0.11* 0.11* -0.12* 0.05-0.25* 1.00 FERTILIZER 0.40* 0.02-0.11* -0.08* -0.27* 0.41* -0.31* 0.25* -0.32* 0.12* *sgnfcant at 10%. 20

Appendx 4: Development level and nsttuton condtonal mpact of nequalty on envronment DEPENDENT VARIABLES (2SLS FIXED EFFECTS ESTIMATIONS) DEVELOPMENT LEVEL INSTITUTION QUALITY INDEPENDENT (1) (2) (3) (4) (5) (6) VARIABLES CO2 SO2 BOD CO2 SO2 BOD EHII 12.14*** -7.513 5.685 6.481** 15.10 14.17* (3.576) (-0.892) (0.652) (2.410) (1.405) (1.955) (EHII)x(DEV_LEVEL) -9.984*** 13.31* 5.757 (-3.149) (1.698) (0.723) (EHII)x(INSTITUTION) -0.129-0.778* -0.247 (-1.290) (-1.929) (-1.169) GDPCAP 1.606*** 2.623** 1.523 0.930** 3.276** 1.775** (3.649) (2.412) (1.427) (2.344) (2.059) (2.097) GDPCAPSQ -0.118*** -0.192*** -0.138** -0.0686*** -0.230** -0.154*** (-4.178) (-2.734) (-1.993) (-2.807) (-2.343) (-2.909) POPDENS -0.0352-0.995*** -1.477*** -0.162-1.016-1.484*** (-0.240) (-2.735) (-2.931) (-0.963) (-1.500) (-2.927) SCHOOL 0.161 1.506*** 0.0934 0.285 2.570*** 0.442 (0.958) (3.616) (0.225) (1.239) (2.832) (0.916) FERTILIZER 5.00e-05** 0.000171*** 0.000173** 6.27e-05** 0.000254** 0.000199** (1.971) (2.689) (2.168) (1.996) (2.087) (2.350) INSTITUTION -0.00313** -0.00700* -0.00386 0.0556 0.334* 0.104 (-2.097) (-1.890) (-1.036) (1.267) (1.875) (1.118) FDI -0.295-0.846-0.350-0.0841-0.995-0.407 (-0.666) (-0.908) (-0.369) (-0.164) (-0.597) (-0.446) OPEN -0.0770-0.154-0.175-0.133-0.376-0.242 (-0.895) (-0.724) (-0.804) (-1.275) (-0.940) (-1.098) CONSTANT -6.442*** -25.52*** -14.92*** -4.806** -34.81*** -18.03*** (-3.367) (-5.415) (-3.241) (-2.481) (-4.449) (-4.447) Observatons 367 367 365 367 367 365 NB countres 86 86 88 86 86 88 ***sgnfcant at 1%, **sgnfcant at 5%, *sgnfcant at 10%. t-statstcs enter parenthess. Income nequalty (EHII) s nstrumented by dependency rato; (EHII)x(DEV_LEVEL) s nstrumented by the nteracton between dependency rato and development level dummy and EHII INSTITUTION s nstrumented by the nteracton between dependency rato and nsttuton varable. The frst step estmaton results are presented n appendx 5. 21