Predictions of future temperature increases depend critically on the projections of future greenhouse gas emissions. Yet there is a vigorous debate about how these projections should be undertaken. This paper explores a range of methodological issues surrounding projecting greenhouse emissions over the next century. It points out that understanding future emissions requires a framework that deals with the sources of economic growth and allows for endogenous structural change. It also explores the role of convergence assumptions and the "Castles and Henderson Critique" of the Special Report on Emission Scenarios (SRES) regarding use of Market Exchange Rates (MERs) rather than Purchasing Power Parity exchange rates (PPPs) to benchmark income differentials in the world economy. In the G-Cubed multi-country model, we show that emission projections based on convergence assumptions defined in MER terms, are 40% higher by 2100 than emissions generated using a PPP comparison of income differentials between economies. We support the argument presented by Castles and Henderson, that the use of MERs in the SRES represents a serious analytical error. It is not clear what this means for the SRES projections because the SRES is not transparent in its assumptions. In the G-Cubed model, the error leads to considerably higher emissions projections.