Shedding light on aid: Using night lights to analyse the effectiveness of geocoded aid in Malawi

Abstract

This thesis uses two unique microeconomic datasets to analyse the effectiveness of foreign aid on economic activity, where economic activity is measured using night light intensity. First, I create an innovative night lights panel dataset for Malawi from 2005 to 2010. This dataset extracts individual pixel values from stable night light images captured from the Defense Meteorological Satellite Programme - Operational Linescan System (DMSP-OLS). The individual pixel values are then summed together in each town or village to represent a proxy for economic activity. Second, I use a dataset created by AidData that includes foreign aid projects in Malawi. This second dataset is geocoded, which means that longitude and latitude coordinates are attached to each aid project. Using ArcGIS 10, I overlay the geocoded aid projects on the stable night light images. This allows for analysis to be conducted at the town or village level.

I use both a random effects panel model and a spatial dynamic panel model, with a GMM estimator for the spatial dynamic panel model. I find that completed aid projects have a significant impact on the night light intensity in a town or village. The results with respect to the effectiveness of aid on night light intensity are sensitive to the type of econometric technique employed. In addition, I find a negative precipitation coefficient in a random effects panel model, which suggests that towns or villages with more cloud cover are associated with lower night light intensity levels. Finally, I find a negative spatial correlation in night light intensity between neighbouring towns or villages. This suggests that there is some form of mobility present with regards to commerce or resources.