Measuring poverty with administrative data in data deprived contexts: The case of Nicaragua

Abstract

Microsimulations are routinely used to estimate poverty in contexts of data deprivation. This paper explores how microsimulations can be enhanced by adding widely available macroeconomic and administrative data. In concrete terms, we analyze the effects of including unemployment rates and affiliation to social security in microsimulations of poverty headcounts in Nicaragua. The recent political crisis in this Central American country has interrupted data collection efforts, making it impossible to monitor poverty or quantify the effect of the crisis. We consider several methods, including incorporating unemployment in our simulations; using alternative poverty lines; and comparing with a counterfactual of no crisis. Including readily available administrative data may have significant effects on the estimated poverty headcount, with this effect yielding between 0.1 and 4.6 percentage points in difference in Nicaragua. More generally, while worthwhile efforts to utilize machine learning and cross-survey imputation to estimate poverty in data deprived contexts continue, inexpensive and comparatively straightforward microsimulations can still provide substantive insights on poverty dynamics.