Studying Multidimensional Poverty Dynamics: Can Synthetic Panel Methods Help?

We extend the Dang et al. (2014) synthetic panel framework for studying poverty dynamics to multidimensional poverty measures. Using the case of a three-dimensional poverty index over two time periods we show that it is possible to produce moderate bounds on predicted transitions, based on household survey data for Tanzania from 2014 and 2020. The panel structure of the National Panel Survey (NPS) of Tanzania allows us to compare synthetic panel estimates to directly measured multidimensional poverty transitions. We assess two approaches to the estimation of multidimensional poverty transitions: one based on tracking transitions of the aggregate multidimensional poverty index; and a second building on transitions of the individual dimensions of the multidimensional poverty index. The latter approach, based on a Seemingly Unrelated Regression Estimation (SURE) framework, is shown to produce more accurate and precise transition estimates. We provide R and Stata codes, which can be incorporated into the standard synthetic panel code for the purpose of extending the analysis to multidimensional poverty.