Thanks for this insightful discussion. I hope you don't mind if I add one clarification about the results of my paper with Aprajit Mahajan, Brian Blackburn et al. on ITNs sold on credit. The lack of improvements in malaria prevalence in treatment (that is more ITNs) areas of course has not changed, but in the new, revised version of the paper we also show that self-reported six-month malaria INCIDENCE did decrease considerably in treatment areas. Recall that prevalence tells you the fraction of a population who suffer from a condition, while incidence tells you the number of cases in the population over a longer period of time. Of course self-reported information is not as reliable as blood tests, but it does have the benefit of concentrating on episodes severe enough to be symptomatic and hence recognized by the households (although it is again not clear that such cases were indeed malaria!). In the paper we discuss at length why such discrepancy between prevalence and incidence may have taken place. In short, we go back to the idea (highlighted by Shanta) that the low coverage with ITNs is likely at least partly responsible, and we discuss how this is consistent with a number of earlier studies on the effectiveness of ITNs. Let me also add (and sorry for being pedantic!) that a threshold for externalities/community effects of ITNs to start kicking in has not to the best of our knowledge been conclusively and experimentally established, and indeed a number of models suggest that there isn't a precise threshold, with even some ITNs providing some benefits to non-users. The new version of the paper (that can be found here http://www.econ.upf.edu/~tarozzi/TarozziEtAl2013RCT.pdf) provided lots of additional details and citations in case anyone want to know more.