A Bayesian factor-augmented interacted vector autoregression framework purified of expectations is employed to analyze how government spending shocks have impacted CO2 emissions in the United States from the 1980s to the pre-pandemic period. Consumption-generated emissions are found to have generally risen following fiscal expansions, although their elasticity to government spending has declined substantially over time—with the five-year elasticity dropping from about 0.5 in the early 1980s to 0.1 by 2019. In contrast, positive government spending shocks increased production-generated emissions in the early 1980s—with a five-year elasticity near 0.4—but reversed course by the 1990s, eventually reaching an elasticity of –0.5 by the end of the sample. Examination of time-varying interaction variables suggests that environmental regulation, tertiarization, and a larger share of spending on public goods can mitigate—or even reverse—the emissions growth associated with economic expansions driven by government spending. Furthermore, government consumption, rather than investment, is chiefly responsible for these shifts in emissions elasticities.