Panel Paper: Spatial Analysis As a Policy Evaluation Tool: The Case of Droughts Adaptation Technologies in the Brazilian Semiarid

Saturday, March 10, 2018
Room 16 (Burkle Family Building at Claremont Graduate University)

*Names in bold indicate Presenter

Marcelo S.O. Goncalves, University of California, San Diego

In Semiarid regions around the world, millions of people suffer water shortages due to successive droughts. In many cases, water scarcity couples with poverty, causing episodes of social calamities that trigger unwanted migrations flows and, no rarely, cost the life of thousands of people.

The Brazilian Semiarid region has a long history of recurring droughts. A rough estimation indicates that, since the 16th century, 124 droughts were registered in the region. In the 1990s, a social movement, headed by the Semiarid Articulation (ASA), proposed a program to build 1 million rain harvest units (cisterns) to attend millions of the poorest people in the Brazilian Semiarid. Its demand got stronger when, in 2000, another severe drought ravaged part of the Northeastern Region, causing not only the usual food security problems, but also a water shortage that reached some of the main coastal cities and the hydroelectric reservoirs.

In 2003, the Federal Government decided to fund the program. The focus of the program is to build low-cost tanks for low-income households in rural areas of the Semiarid Region. In 2016, the program was present in almost 1400 cities and had built more than 900 thousand units, considering only the “water for consumption” modality. In 2017, this policy was recognized with the Future Policy Silver Award 2017, awarded by the World Future Council in partnership with the UNCCD.

This study uses two types of spatial regressions (OLS and geographically weighted regressions) to assess whether the program has been placing the cisterns in the localities that needed them the most. For that, it regresses the program coverage (number of cisterns per capita) on indicators associated with the qualifying dimensions of the program: exposure to droughts (a climate component), poverty prevalence, and rural population proportion. The data used is aggregated at the county level.

This study’s identifying assumption can be stated as follows: if the program is well-target, the variables associated to the qualifying dimension should present relatively high marginal effects at significant statistical levels (90%, 95% and 99% confidence intervals). It means that relatively low marginal effects or statistically insignificant coefficients indicate poor targeting or that some unobservable has higher impact on placement decision than the stated qualifying indicators.

The results suggest that the program is not as well target as the government claims. Taken as a group, the three variables produce estimates that are not statistically significant; individually, the climate variable is the only that seems relevant – it systematically presents sizable and statistically significant marginal effects. The other two variables of interest return very low marginal effects that are not statistically significant in many cases. More troublesome is the fact that rural population concentration seems to be negatively associated to program coverage, what suggests that there is a “penalty” for county with a higher proportion of its population living in rural areas. Thus, although placed in the regions more exposed to droughts, the cisterns are not concentrated in the municipalities where the concentrations of poor people and rural population are the highest.