We use multiple measures of agricultural total factor productivity (TFP) change to examine the relationship between agricultural productivity and poverty in developing countries. We employ a stochastic frontier analysis to estimate agricultural TFP changes for 113 countries using output distance function in a multi input multi output framework. We then make alternative groupings of countries to allow for the possibility of different production frontiers for countries with different income level, and we examine the effect of these various measurements of agricultural TFP on poverty reduction. Results from the TFP analysis show that TFP change estimates by income groups differ from those estimated using all countries in a pooled model. This indicates that agricultural technology and production frontiers may differ across countries based on income levels. Preliminary results show that TFP change from the pooled model has significant impact on poverty reduction. However, TFP estimates from different income groups didn’t indicate significant impact on poverty. The relationship between TFP change and poverty is therefore sensitive to the method used to estimate agricultural productivity.