Assessing changes in poverty levels over time is bedeviled by problems in questionnaire design, the choice of the poverty line, the exact timing of the survey and uncertainty about the appropriate cost-of-living deflators. In this paper, we focus on testing the robustness of measured changes in poverty to these common problems, using household panel data collected in rural Ethiopia in 1989, 1994 and 1995: in particular, we implement a simple graphical technique for assessing the impact of uncertainty in measured inflation rates. We find that poverty declined between 1989 and 1994, but remained virtually unchanged between 1994 and 1995. However, the last result disguises substantial seasonal fluctuations in 1994. We also find that households with substantial human and physical capital, and better access to roads and towns have both lower poverty levels and are more likely to get better off over time. Human capital and access to roads and towns also reduce the fluctuations in poverty across the seasons.