Do consumption choices explain and help project energy demand and CO2 emissions? We develop and estimate the parameters of a general-equilibrium model with nonhomothetic preferences, that is, with income-dependent consumption baskets, for a large set of countries and sectors. We find that direct household energy consumption increases less than proportionally to income in rich countries and is more income elastic in developing countries. Income effects for indirect embodied energy consumption, traced through intermediate use and trade linkages, are weaker but significant: high-income-elasticity goods and services have on average the lowest CO2 intensity, medium-income-elasticity goods the highest. Income effects in consumption patterns thus partially explain the cross-country inverted-U relationship between per capita GDP and emissions intensity (“environmental Kuznets curve”). Simulations suggest that further economic growth, through changing consumption choices, would lower emissions intensity in middle- and high-income countries and increase it at low incomes. Reductions in aggregate world emissions would therefore be modest.