Analyses of responses to reforms in Ghana seem to indicate that current policies may be benefiting different segments of society disproportionately. Also, experience in the 1990s suggests that recurring budget deficits may adversely affect reform and poverty alleviation programmes. The aim of this paper is to carry out some experiments using variants of a stylised CGE model, to ascertain the possible effects on poverty of a range of budget-neutral redistributive income transfers. The analysis is based on a social accounting matrix (SAM) for Ghana for the year 1993, which has been substantially modified for the present application. The CGE model is a real-side, static model and therefore excludes the monetary and financial sectors and is designed in the tradition of other OECD Development Centre models. The experimental design follows one employed by Adelman and Robinson (1978) for Korea, and Chia et al. (1992) for the Côte d’Ivoire. However the experiments are designed with a view to examining the sensitivity of the results to alternative specifications, within otherwise broadly similar, SAM-based model structures. The main outcome is to show that the results are very sensitive to (long and short run) closure rules, to the financing rules in a budget-neutral setting, and to the method of computing poverty ratios (parametric and non-parametric approaches). A new decomposition method is introduced to assist in interpreting the results. A wide range of simulations demonstrates that poverty is not eradicated via redistributive income transfers, and may even increase, especially in the short run, after taking into account the secondary effects.