Back

Using Machine Learning to Target Treatment: The Case of Household Energy Use

Christopher R. Knittel and Samuel Stolper

January 2020

We use causal forests to evaluate the heterogeneous treatment effects (TEs) of repeated behavioral nudges towards household energy conservation. The average response is a monthly electricity reduction of 9 kilowatt-hours (kWh), but the full distribution of responses ranges from -30 to +10 kWh. Selective targeting of treatment using the forest raises social net benefits by 12-120 percent, depending on the year and welfare function. Pre-treatment consumption and home value are the strongest predictors of treatment effect. We find suggestive evidence of a “boomerang effect”: households with lower consumption than similar neighbors are the ones with positive TE estimates.

Keywords: machine learning, program evaluation, targeting, energy efficiency.
JEL Codes: C53; Q40; D90


For Associates Only

As a benefit to our Associates, the latest Working Papers are embargoed for a period of up to six months before becoming accessible to the public. If you are interested in becoming an Associate or learning more about the benefits of sponsorship, please click here, or email us at ceepr@mit.edu

If you are a CEEPR Associate or CEEPR staff member, please visit the login page here: http://ceepr.mit.edu/associates