In order to contribute to the debate on environmental damages generated by carbon emissions and smoke from dirty fuels, we investigate the determinants of fuel choice and fuel switching in Chinese rural households. To this end, we specify a non-separable farm household model, focusing on the substitution of fuels from wood/straw and coal to LNG and electricity, is estimated by using a large micro-household panel dataset. We find that the pattern of fuel use depends not only on income, fuel prices and demand-side socioeconomic factors, as would appear in standard demand models, but also on agricultural production characteristics, food prices, and a set of original household and community characteristics shaping the household responses to market failures.