Abstract
OECD countries have recognized the urgent need to increase energy efficiency among low-income households. Faced by high cost of public funds, policy-makers and program managers therefore seek low-cost policy alternatives, making behavioral interventions (``nudges’’) an obvious candidate. Applying such interventions to the economically disadvantaged raises problems of transportability: Policy design risks relying on biased estimates of causal effects among the target group because of insufficient overlap between study and target populations. We illustrate the presence and scale of the transportability problem by conducting an randomized control trial on scalable, low-cost design elements to improve program take-up in one of the world’s largest energy efficiency assistance programs. Observing investment decisions of over 1,800 low-income households in the Refrigerator Replacement Program, we find that the transportability problem is real and consequential: The most effective alternative is not predicted by the literature. More worryingly, relying for policy design on evidence collected outside the target population can backfire, violating ‘do no harm’ principles of policy advice. Systematic testing remains critical for addressing transportability problem, particularly for policies targeting vulnerable groups.