| Abstract | Energy consumption is one of the main source of air pollution, greenhouse gas emission, and global warming. Reducing household energy consumption is a key goal of policymakers. We develop statistical models using socio-economics and demographics, building characteristics, location, temperature, and energy prices to estimate household energy expenditure in the U.S. We use household energy expenditure for more than 560,000 households in the U.S. from 2010 to 2012. We first employ multivariate regression models to investigate and identify the impacts of the explanatory variables on household energy expenditure. Next, we use Principal Component Analysis (PCA) to convert correlated co-linear explanatory variables into orthogonal components and estimate household energy expenditure by principal component regression. We find newer attached buildings to be effective in decreasing household energy expenditure, particularly among educated people in metropolitan areas. With sufficient data availability, our model could be used by state, regional, or even city-level policymakers and planners to optimize their infrastructural investments. |
