Conswnption-based pCNerty comparison in three rural villages of Region 3 (Dinki,rnYetmen, and Shumsha) show that Dinki, the worst hit village in 1984 drough~ is still poorer,rnwhile Yetmen is relatively well-off. Larger household size is unambiguously a significant causernif poverty in all villages. Interestingly, unlike agricultural technology, "formal education"rndoesn't seem to have direct impact to avoid poverty in all areas, but indirectly by enablingrnhouseholds to limit household size. Also, male-headed households are not any better positionedrnthan female-headed ones to escape poverty. It seems thot only where adoption rate of modernrnagricultural technology is high (as in Yetmen) that experience and hence age helps. Non farmrnincome is significant determinant of pCNerty particularly in Dinki where agriculture is largely arnhopeless venture, which sharply contrasts with Shwnsha where the non farm jobs are despisedrnand left only for those marginalized FalashaslJews people. Assets are particularly usefoJ wherernnon-faml income is a significant part of income as in Dinki. Land not surprisingly, has a strongrnpoverty reducing effect This is not the case with Oxen. To reduce the level of poverty in thesernareas, limiting household size (through expansion of education), expansion if non-farmrnemployment opportunities, credit supply are the major policy recommendations, though therndegree of their importance vary among the areas. Yet these econometric results need to berncombined with the PartiCipatory Poverty Assessment (PPA) method in order to obtain even arnmore reliable outcome.