[notahacker]

I think the short answer is you can't. This study is probably less likely to be susceptible to another major problem with philanthropy experiments, which is recipients reporting what they think the survey designers want to hear because they believe they will be further rewarded as a consequence.

(which I think is present and very difficult to avoid in RCT trials in developing countries; a notable example was people reporting that they were less likely to have recently been subjected to domestic violence after their neighbours had received money.)