Remarkably little is known about the long-term impacts of project aid to lagging poor areas (Chen, Mu et al. 2006, 2008). This paper contributes to the debate about the role of rural transport infrastructure development in explaining the long-term rural development. In line with Grimm and Klasen (2008) we agree that there is value-added to consider this debate at the micro level within a country as particularly questions of parameter heterogeneity and unobserved heterogeneity are likely to be smaller than between countries. Moreover, at the micro level it is possible to identify more precise transmission mechanisms from rural transport infrastructure to socio-economic development outcomes. This is done empirically by analyzing a UNDP&UNCDF; financed rural development project in Zambia's Eastern Province running from 1997-2002. The secondary datasets consist of respectively a series of repeated cross-sectional living conditions monitoring surveys (LCMSs). The LCMSs were collected in 1998 (baseline) and 2004 (follow-up), that is both prior, during and after the project implementation. Our aim is to assess the ability of the parametric and semi-parametric models as well as using a time- series of cross-sections to provide an adequate description of the logarithm of per adult equivalent consumption of rural household conditional on few covariates, including an infrastructure treatment dummy variable. Although, the mean cotton sales share of household income has more than doubled despite the fact that the mean distance to the input market remained unchanged from 1998 to 2004, the parametric and semi-parametric estimation results are only small and statistically insignificant in terms of gains to mean consumption emerged in the longer-term. The main results are robust to corrections for various sources of selection bias