Forecasts of Chinese carbon dioxide (CO2) emissions are critical to any global agreement on mitigating possible global climate change. We provide such forecasts through 2050 using a reduced form model selected using a general to simple search strategy. These estimates are the first based upon provincial-level data (1985-2000). The model chosen by the information criterion is one that melds the standard approach taken in the science and engineering literature with the environmental Kuznets curve approach popular in the economics literature whereby per capita emissions can first rise and then fall with increases in income. Other aspects of the model allow for the possibility that the rate of technological change varies across provinces and the possibility of population density effects. We find statistical support for the presence of an inverted U shaped environmental Kuznets relationship with the projected turning point being not too far above Shanghai's current income level. Our model suggests lower estimates of CO2 emissions given similar GDP and population growth assumptions than those based on aggregate national level data such as the quasi-official Intergovernmental Panel on Climate Change (IPCC) estimates. However, in contrast to conventional wisdom, uncertainty over demographic changes is likely to dominate uncertainty over changes in per capita GDP. It also predicts that province specific per capita emissions are likely to follow very different income/pollution trajectories. This in turn suggests that province specific policies to reduce CO2 emission levels may be desirable.