This paper investigates the dynamic impacts of rural road improvements on farm productivity and crop choices in Zambia's Eastern Province. There are several channels through which the feeder road improvements impact on farmers. Our aim is to estimate whether the differential outcomes in the five treatment districts and three control districts generated by the expansion of market agricultural activities among small to medium scale farmers could be explained by rural road improvements that took place after the new Chiluba MMD government in 1995 had completed an IMF rights accumulation programme bringing the principal marketing agent system to an end. Our district-level empirical analysis is an extension to the Brambilla and Porto(2005, 2007) crossprovincial level approach which proposes a dynamic approach accounting for entry and exit into the agricultural cotton sector to avoid biases in the estimates of aggregate productivity, when measuring productivity in agriculture applied to a repeated cross-sections of farm-level data from the Zambian post-harvest survey (PHS). Despite the limitations of the PHS data covering the period from 1996/1997 to 2001/2002 when the Eastern Province Feeder Road Project (EPFRP) was being implemented. The identification strategy relies on differences-in-differences of outcomes (i.e., cotton productivity) approach across two phases (pre-treatment and post-treatment). We use maize productivity to difference out unobserved household and aggregate agricultural year effects. Through our descriptive analysis we do find that changes in land allocation and in yields to Eastern Province's most important cash crop - cotton did occur at the district level. However, it is difficult to conclude that these changes are linked directly to the improved accessibility obtained from the implementation of the EPFRP based on our differences-in-differences estimator or our Tobit model.