Abstract
Reducing domestic energy demand is central to climate mitigation and fuelpoverty strategies, yet the impact of energy efficiency interventions is highlyheterogeneous. Using a causal machine learning model trained on nationallyrepresentative data of the English housing stock, we estimate average andconditional treatment effects of wall insulation on gas consumption, focusingon distributional effects across energy burden subgroups. While interventionsreduce gas demand on average (by as much as 19 percent), low energy burdengroups achieve substantial savings, whereas those experiencing high energyburdens see little to no reduction. This pattern reflects abehaviourally-driven mechanism: households constrained by high costs-to-incomeratios (e.g. more than 0.1) reallocate savings toward improved thermal comfortrather than lowering consumption. Far from wasteful, such responses representrational adjustments in contexts of prior deprivation, with potentialco-benefits for health and well-being. These findings call for a broaderevaluation framework that accounts for both climate impacts and the equityimplications of domestic energy policy.