Several studies have shown conceptually that assets form a more robust basis for identifying the poor than do flow variables like expenditures or income. Asset-based poverty classification can be used to distinguish structural from stochastic poverty and can enable projection of poverty dynamics through time. Nonetheless, little work has empirically compared poverty measurement based on assets and expenditures to indicate the practical implications of the choice of poverty measure. This article uses panel data (between 1994 and 2004) from Ethiopia to generate asset- and expenditures-based poverty classifications. Asset dynamics are then explored to test for the existence of multiple asset index equilibria that could constitute evidence of poverty traps. Results provide evidence of multiple equilibria and show that the asset-based poverty classifications predict future asset and expenditures poverty status more accurately than expenditures-based measures. The findings confirm that the asset-based measure could be used to more carefully target poverty interventions.