This study aims to test 1. whether two indicator-based tools for poverty assessment in rural Central Sulawesi, Indonesia, developed in 2005 are capable of predicting absolute poverty in 2007, and 2. whether the indicator composition of these poverty assessment tools remains robust over time. In 2005 and 2007 we surveyed 264 households in Central Sulawesi to obtain indicators of poverty and to derive their daily per capita consumption expenditures. Ordinary least squares and quantile regression models were fit to these data sets. Each of the models yielded a different set of 15 indicators for poverty prediction. Applying the 2005 indicators to the 2007 data set, the prediction power of the indicators from 2005 mainly was influenced by the error of over-predicting poverty in 2007. When re-estimating the regression models with the 2007 data set, the accuracy levels were found to be similar, but the indicator composition of the tools changed.