This working paper presents an improved analysis of the potential of biofuels for climate change mitigation. It starts from an existing estimate from Havlík et al. (2011) and: Adds the emissions from changes in dead wood, litter and soil organic carbon (DWLSOC); Includes a sensitivity analysis; Investigates different options for carbon accounting; and Investigates the timing of emissions from biofuels. It is clear from Havlík et al. that the greenhouse gas emissions from biofuels are dominated by the future technology. Nevertheless, including DWLSOC in the estimate is important. The significance is directly related to the amount of deforestation predicted. We found that including DWLSOC increased greenhouse gas emissions by 21 per cent. These results are nonetheless sensitive to poorly studied assumptions about DWLSOC dynamics. The greenhouse gas accounting approach that is currently accepted internationally allocates emissions from bioenergy in the land use sector if a loss of carbon stocks occurs, rather than in the energy sector. As a result, this approach: Underestimates the true emissions from biofuels, if few countries participate in a greenhouse gas target; and Poorly allocates emissions to the region in which they occur. We present accounting alternatives which perform these two tasks better. Finally, greenhouse gas emissions from biofuels can amount to more than from the fossil fuels they displace, even if bioenergy is carbon neutral over the long term. This occurs because every addition of biofuel causes an associated emission with a ‘payback' period. If biofuels use grows quickly then the users may continually be emitting more than they are paying back.