Global CO2 emission forecasts span such a wide range as to yield little guidance for climate policy and analysis. But global per capita emissions appear to be better constrained than total emissions, which we argue has an economic justification. Trend stationarity of per capita emissions may provide a means of characterizing the relative likelihood of global forecasts. On data spanning 1950 to 2009 a unit root hypothesis allowing for endogenous structural breaks is rejected, but adding in the 2010 observation pushes the p-value slightly over 0.1. Since structural breaks cannot be detected at the end of sample this may simply indicate a power problem. Using Monte Carlo simulations we conclude that the lower half of a well-known suite of IPCC emission scenarios is more likely to occur than the upper half, and the top quartile is particularly difficult to justify. Copyright © 2013 John Wiley & Sons, Ltd.