Evaluating effects of potential drivers on household indirect carbon emission quantity and patterns using machine learning

Household indirect carbon emissions (HICE), which are influenced by many interacting factors, contribute significantly to total emissions. By evaluating the effects of potential drivers on HICE quantity and patterns, this study aims to identify the most effective policies for reducing HICE and to propose suggestions for policy differentiation. We calculate HICE quantity, define HICE patterns, and evaluate the effects of 27 factors on HICE quantity and patterns using machine learning methods based on household-level data from China. The empirical results show that non-financial assets per capita, income per capita, family size, and the age of the household head are the four most important determinants of HICE quantity. For HICE patterns, family size and the age of the household head remain key determinants, while income per capita and non-financial assets per capita rank from 8 to 26 among all factors and play a much smaller role. In general, regional factors are far less important than individual factors in determining HICE quantity, but their roles are more comparable when explaining HICE patterns. We suggest that levying a progressive property tax, encouraging low-carbon consumption among younger people, promoting low-carbon diets among older people, and providing shared lifestyle services in communities with a high proportion of single-person households may be effective in reducing HICE.