Using a panel dataset from Uganda spanning eight growing seasons, we analyse whether the decision of maize farmers to purchase new seed (either improved or traditional) leads to higher yields or sales per acre. Purchased improved seeds are shown to cost about twice as much per kilogram as traditional seeds, while certified improved seeds are almost three times more costly than traditional seeds. Controlling for time-invariant household and plot-level unobservables using fixed effects analysis, we find that purchasing traditional seeds leads to higher yields, but purchasing improved seeds does not. Likewise, cultivating purchased traditional seeds leads to a large increase in sales per acre and the price per kilogram obtained by farmers, while we find only limited evidence of higher sales per acre for purchased “quality declared” improved seed. We find little evidence that seed policy reforms implemented in Uganda in the past 20 years had any substantial impact on agricultural productivity or farmers’ commitment to persist in using improved technologies.