Profiling Poverty with Multivariate Adaptive Regression Splines

Using data from the 2003 Family Income and Expenditure Survey and 2005 community-based monitoring system for a city, Multivariate Adaptive Regression Splines (MARS) is used in identifying household poverty correlates in the Philippines. Models produced by MARS are more parsimonious, yet contain theoretically and empirically sound set of household poverty correlates and have high accuracy in identifying a poor household. MARS provides a better alternative to logistic regression for a more efficient and effective implementation of a proxy means test in the identification of potential beneficiaries of poverty alleviation programs.