Global poverty: A first estimation of its uncertainty

Abstract

The dollar-a-day method, applied in monitoring the UN’s development goals against poverty, provides no confidence interval for the official figures of global poverty reduction, a practice that does not allow statistical testing. Using Monte Carlo micro-simulations we construct confidence intervals that reflect the error introduced by the process of determining the International Poverty Line, as well as the uncertainty of the involved Purchasing Power Parity exchange rates. These estimates identify a reduction of 5.19% between 1990 and 2015 at 95% confidence level, in stark contrast with the remarkable 73% reduction of global poverty reported in the World Bank official statistics published on September 18, 2018. At the same time, MDG1 obtains with a 80% confidence level. The cost-of-basic-needs method paints a more promising picture identifying a 35.71% reduction at 95% confidence level, while the confidence level at which poverty in 2015 was half of 1990 stands at 46%. We conclude that the derivation method of the international poverty line introduces high levels of uncertainty in the estimates.