WEAI/AERE 2012 - Individual Paper Abstract
Title: The Economic and Fiscal Impacts of Extreme Weather Events: Evidence from Brazil
Author(s): Sebastian MILLER, Research Department, Inter American Development Bank, E-Stop 1009, 1300 New York Ave NW, Washington, DC 20577, USA, 202-623-3972, smiller at iadb dot org; Paulo Bastos; Matias Busso, Inter American Development Bank [Photo credit: based on Wikimedia Commons, uploaded by dl91m]
Abstract:
There is a recent extensive literature on the economic impacts of natural disasters. However scholars do not agree on the effects of natural disasters on economic activity. Most papers look at cross-national comparisons, which are inconclusive for several reasons. First, natural disasters in general have impacts at a local, not national level. Second, these comparisons do not take into account differences in institutions that may undertake ex-ante preventive measures that avoid damage after a shock occurs. This paper contributes to this field looking at the economic and fiscal effects of extreme weather events, focusing on the impacts at the local level.
This is done by using an extensive dataset for Brazilian municipalities which include economic characteristics of the municipality, including, GDP, composition of GDP, and fiscal indicators such as taxes, transfers and expenditures by area. We combine this rich dataset for a little over 5500 municipalities with declarations of natural disasters for the period 2003-2008. These declarations include all types of disasters, but we focus on droughts which account for roughly 70% of the cases. We use this data to estimate the impact of a drought declaration on economic activity (namely GDP and agricultural GDP) at the municipal level. Our OLS benchmark estimates show a negative and significant effect of droughts on both aggregate and agricultural GDP.
Drought declarations however may not be a good enough indicator of a drought. There is the possibility of politically motivated declarations and moreover, some municipalities may be better at getting declarations than others simply because they know how to play the game. In order to overcome this issue we adopt two different estimation strategies. In the first method we use a matching procedure to match treatment municipalities, ie those where a drought is declared, with a control group, where droughts were not. Our main results in this case hold.
In addition we also use an instrumental variable approach, where we create our own drought measure for each month in each municipality over the 2003-2008 period. To this end we use two sources of additional data. We obtain historical precipitation from the CRU high resolution gridded dataset. It has month-by-month variations in climate over the period 1901-2009, with a resolution of 0.5 degrees. Using a geospatial method we extrapolate the precipitation to each municipality in the country. We then use this to construct rainfall monthly distributions by municipality for each Brazil municipality using the 1960-2002 data. We finally combine this with another source of data containing climate information sorted by weather stations for the period 2003-2008 obtained from SINDA. Using a similar geospatial extrapolating procedure we obtain monthly rainfall for each municipality and compare the levels with the historical distribution and generate a new drought index.
We finally use this IV in a 2SLS regression and find that the effects of droughts remain negative and significant. Moreover our coefficients imply that the negative effects on both Agriculture GDP and GDP per se can be very large.