The ecoENERGY program was open to everyone, but the homeowners had to make the investments.
In our study, we wanted to avoid the problem of landlord-tenant split incentives.
As a result, we excluded from our study apartments and multi‑family residential buildings, such as townhouses. We really focused on the single‑family housing sector. We found that middle‑income households had the highest participation rate. However, this segment of the population's performance in terms of reducing energy consumption was nearly the lowest compared to other participants.
That implied to us that there was some combination of unobserved selection and of the characteristics of the energy performances of the houses that was leading the households that participated a lot to have lower savings. We concluded that because this was not a study in which we completed a randomized control trial, there was some unobserved selection. That's a typical problem you're going to end up with if you don't deploy a randomized control trial beforehand.
We can't test our hypothesis because this was ex post data, but it's likely that folks were planning a renovation anyway, like replacing a furnace or getting an energy-efficient AC, and thought they might as well get a rebate. That causes an issue for identifying which households should be adopting these investments and getting the rebates—and not because they would have done it anyway. We don't want to be identifying free riders. We also want to make sure that households that have the lowest incomes are getting targeted rebates.