Bangladesh has made remarkable progress in reduction of poverty headcounts, poverty gap, and squared poverty gap since 2000. While scores of studies rigorously looked into drivers of poverty reduction from the perspective of sources, both the cross country and single country literature on poverty convergence is scant. Even most of these convergence estimates are biased arising out of omitted variable due to ignorance of unobserved individual heterogeneity and endogeneity of at least a subset of regressors. This paper exploits a unique set of data to provide credible evidence of convergence in poverty across the districts of Bangladesh during 2000-2016 using the Arellano-Bond system dynamic panel estimator and panel generalized method of moments estimator. We find that poverty convergence is present during this period both through direct estimation and decomposition with relevant indirect estimates. Our results are robust to the alternative frequency of data (cross-section vis-à-vis panel) and the consequent estimation techniques, sources of data (direct estimates from the Household Income and Expenditure Surveys vis-à-vis small area estimates) and the alternative transformation of the dependent variables. Both growth-accounted poverty convergence effect and strong growth effect dominate the adverse effect of initial poverty on growth effectiveness to ensure strong overall poverty convergence found across the districts.