Vulnerability to poverty revisited: Flexible modeling and better predictive performance

• DOI: 10.1007/s10888-017-9374-6
• Publication Date: 2018-02-03T09:42:04Z
• Authors: Maike Hohberg; Katja Landau; Thomas Kneib; Stephan Klasen; Walter Zucchini
• Keywords: Cut-off; Vulnerability; Conditional probability distribution

This paper analyzes several modifications to improve a simple measure of vulnerability as expected poverty. Firstly, in order model income, we apply distributional regression relating potentially each parameter the conditional income distribution covariates. Secondly, determine cutoff endogenously instead defining household vulnerable if its probability being poor next period is larger than 0.5. For this purpose, employ receiver operating characteristic curve that able consider prerequisites according particular targeting mechanism. Using long-term panel data from Germany, build both mean and models with established 0.5 our cutoff. We find new considerably increases predictive performance. Placing into framework does not predictions further but has advantage coherent where parameters are estimated simultaneously replacing original three step estimation approach.