Modelling poverty transitions in Spain: Do attrition and initial conditions really matter?

The availability of panel data has allowed a comprehensive description of poverty exits and entries in Spain. However, most of the literature, so far, has ignored or not explicitly modelled the process of sample attrition and/or the initial conditions problem we face when studying poverty dynamics with survey data. The main objective of this work is to assess whether attrition and poverty status in the base year are endogenous processes to poverty transitions through unobserved heterogeneity. Our estimation follows the model recently proposed by Cappellari and Jenkins (2004a) and uses Maximum Likelihood Simulation techniques. Data is from the European Community Household Panel and refers to poverty transitions that take place between 1994 and 2000 in Spain. Results show that unobservables affecting initial conditions and sample retention are exogenous to those related with poverty transience. Yet, results prove to be sensitive to the choice of the poverty line.