Yup, this combined with the difference in income for couples vs. individuals mentioned by another commented smells to me like the program had a lot of means tes

Yup, this combined with the difference in income for couples vs. individuals mentioned by another commented smells to me like the program had a lot of means testing built in. I wouldn't be surprised if this skewed the results of the experiment in ways that reflect badly on it.