In light of a slow buildup in CO2 emissions since the recovery, this paper revisits the relationship between CO2 emissions and the US economy using a nonlinear autoregressive distributed lag model, in which the determinants are identified through an expanded real business cycle model. We find convincing evidence that CO2 emissions decline more rapidly during recessions than increase during expansions over the long run. Of all determinants considered, long-run asymmetry is fostered once vehicle miles traveled is controlled. This calls for a greater attention to public transportation development and vehicle miles traveled tax for slowing down stock buildup of CO2 emissions during good times.