Hi! Not exactly sure why you disagree, but some context might help. We found 5 studies of hshold spillovers for CTs. Taken at face value, the imply other members of hshold get approx 100% of SWB value as recipient.
Weirdly,we could only find 1 study of intra-hshold therapy effects; that was non RCT and had N<200. It found hshold spillover was 77% size of that to recipient. Taking this at face value too, and assuming hshold size is about 4, therapy goes from 12x to 10x better.
I think the right thing to do here (besides further research) would be to give less weight to the psychotherapy spillover estimate and adjust the effect downwards (or at least more downwards than CTs’ spillover, which has evidence from multiple studies, presumably some better designed and with larger sample sizes), based on a skeptical prior.
PT=psychotherapy, CT=cash transfers
PT= 12 x CT without spillovers
CT’ = 4 x CT with spillovers
PT’ = (1+3s) x PT with spillovers, where s=spillover effect for psychotherapy
PT’= ((1+3s)/4) * 12 x CT’ = 3(1+3s) x CT’.
The worst case, s=0, has PT’=3 x CT’. With s=0.25=25%, PT’=5.25 x CT’, and with s=0.5=50%, PT’=7.5 x CT’.
On Twitter, Michael Plant wrote:
I think the right thing to do here (besides further research) would be to give less weight to the psychotherapy spillover estimate and adjust the effect downwards (or at least more downwards than CTs’ spillover, which has evidence from multiple studies, presumably some better designed and with larger sample sizes), based on a skeptical prior.
PT=psychotherapy, CT=cash transfers
PT= 12 x CT without spillovers
CT’ = 4 x CT with spillovers
PT’ = (1+3s) x PT with spillovers, where s=spillover effect for psychotherapy
PT’= ((1+3s)/4) * 12 x CT’ = 3(1+3s) x CT’.
The worst case, s=0, has PT’=3 x CT’. With s=0.25=25%, PT’=5.25 x CT’, and with s=0.5=50%, PT’=7.5 x CT’.