Air Quality and Infant Mortality in Mexico: Evidence from Variation in Pollution levels caused by the usage of small-scale power plants

This paper exploits the discrete change in air pollutants induced by the installation of small-scale power plants throughout Mexico to measure the causal relationship between air pollution and infant mortality, and if this relationship varies by municipality’s socio-economic conditions. The estimated elasticity for changes in infant mortality due to respiratory diseases with respect to changes in air pollution concentration ranges from 0.58 and 0.84 (more than ten times higher than the OLS estimate). The effect is significantly lower in municipalities with a high presence of primary health facilities but does not vary significantly by average income and education levels. Lower income levels and lower access to health services in low and middle-income countries are likely to contribute to greater adverse effects of pollution on health outcomes. However, the unavailability of reliable measures of pollution concentrations has resulted in a lack of credible empirical studies carefully estimating this relationship for these contexts. This paper is among the first to exploit an arguably exogenous source of variation in air pollution levels across the Mexican territory to measure the relationship between pollution and infant mortality. The empirical strategy exploits, on one hand, a unique data set that provides measures of pollution concentrations in the atmosphere at a detailed level for the whole Mexican territory from satellite imagery. On the other, it isolates changes in air quality from other confounding factors that could affect health outcomes by exploiting the discrete change in air pollutants induced by the installation of small-scale power plants by existing firms in Mexico. The estimated elasticity for changes in infant mortality due to respiratory diseases with respect to changes in air pollution concentration ranges from 0.58 and 0.84, more than ten times higher than the OLS estimate for the same context (Gutiérrez, 2010). The effect is significantly lower in municipalities with a high presence of primary health facilities, and weaker evidence suggests that it varies by average income and education levels. The paper is organized as follows. The next section reviews the existing studies measuring the relationship between air pollution and health outcomes, stressing the need of a source of exogenous variation in pollution levels in order to identify the causal relationship between these variables. Section III describes the context and the nature of the instrumental variable exploited in this paper. Section IV describes the data used. Section V describes the empirical strategy. Section VI presents the results. The final section concludes. II. Related Studies The evidence on the relationship between air pollution and health outcomes (especially infant mortality) for developed countries’ settings is robust and growing. 1 The earlier studies document the positive relationship between air pollution and health outcomes (Pope et al. 1992; Schwartz 2000; Currie and Niedell 2004). More recent literature recognizes that the correlations between pollutant levels and health, due to measurement error and/or the omission of relevant controls may represent a biased estimate of the causal relationship between them. For these reasons, Chay and Greenstone (2003) instrument variations in pollution levels in the United States with the implementation of the Clean Air Act and (in a different paper) the intensity of an economic recession to estimate the impact of air pollution on infant mortality. Their results suggest that Ordinary Least Squares estimates of the relationship between air quality and infant mortality are significantly biased towards zero. Lower access to health services and lower nutrition are likely to be more prevalent in developing countries’ settings. To the extent to which they constrain individuals’ capacity to optimally respond to health shocks, the negative impacts of air pollution on health outcomes may be higher in these contexts, for which a smaller set of empirical studies exists. The existing literature, perhaps due to data limitations, has generally focused on the relationship between air pollution and health outcomes in big cities (Saldiva et al. 1994, Bharadwaj et al. 2008, Arceo et al, 2012). Others do investigate the relationship between pollution levels and infant health outcomes over larger surfaces, although they have important limitations: some lack sources of arguably exogenous variation in air pollutant levels (Gutierrez, 2010), and others exploit sources of large variation in air quality induced by major events (Jayachandran, 2009 exploits Indonesian wildfires, Tanaka, 2010 changes in regulation in China). The extent to which the events causing these large changes in pollution concentrations may have affected other 1 See Glinianaia (2004) for a review of this literature. 2 Foster et al (2009) also estimate the impact of firm’s participation in a voluntary environmental certification program on infant mortality. outcomes which can have a direct impact on infant health raises doubts about the consistency of the estimated impacts of pollution on infant mortality. More recent studies for industrialized countries’ settings have exploited sources of variation in pollutants that are less likely to be related to economic conditions (such as the introduction of EZ-pass in highways, Currie and Walker, 2011) in order to estimate pollutants’ impact on infant mortality, finding large effects. This study also exploits a source of smaller exogenous variation in air quality, but throughout the territory of a developing country setting. The first contribution of this paper to the existing literature is then that it estimates the relationship between air pollution concentrations and health outcomes across the Mexican territory exploiting a source of exogenous variation in pollution levels not likely to be correlated with local economic conditions o other factors which could bias the estimates. In addition, it provides evidence on which municipality-level socio-economic characteristics contribute to reduce or exacerbate pollution’s adverse effects on health. The next section describes in detail the nature of the source of variation in pollution exploited in this paper. III. Context Since 1994, the Commission for Energy Regulation (Comisión Reguladora de Energía, CRE) has authorized existing firms in Mexico to install small-scale power plants in their facilities for self consumption. According to the national inventory of emissions (SEMARNAT & INE, 1999), power plants contribute with 2.89% of the total emissions of PM10 and with 6.51% of the total emissions of PM2.5 at the national level. They are responsible for a major part of SOX emissions, and they are the second source of NOX after the mobile sources. Between 2000 and 2006, 275 small-scale power plants entered in operation in 116 municipalities throughout the country. By 2009, these plants’ electricity generation for self-consumption accounted to 9.4% of the electricity generated in Mexico. In addition to their relatively high contribution to electricity generation, these small power plants are generally characterized by low efficiency and high pollutant emissions (Tanaka, 2010). IV. Data Description The data used in this paper comes from four main sources: (1) from the publicly available list of permits granted by CRE to firms in Mexico to install small-scale power plants; (2) from daily satellite images containing measures of Aerosol Optic Depth (AOD), a measure of air quality, for the whole Mexican territory, for the period from January 1, 2001 to December 31, 2006; (3) from administrative records of registered infant deaths by municipality, month of occurrence and cause of death during the same time period; and (4) municipality-level socio-economic characteristics obtained from the 2000 Mexican Census, administered by the Mexican Statistics Institute (INEGI). The details on how the dataset used in the empirical analysis was constructed are described in what follows. IV.1. Power Plants The publicly available list of permits granted to Mexican firms to install small-scale power plants was used to assign to each municipality (according to the address coded in the permit), a variable indicating the total number of power plants in operation in every period for the time frame analyzed in this paper (the permits specify the date at which plants started operating). Figure 1 shows the distribution of these plants throughout the Mexican territory over time. This study doesn’t face the geographic limitations of the existing research for developing countries, as these power plants are spread across the whole Mexican territory. Because of this, exploiting the geographic variation in other socio-economic indicators, it will be then possible to estimate differential effects of pollution on mortality by municipality-level socio-economic conditions. Figure 1. Number of small-scale power plants in operation by year