Syndicate content

Add new comment

Submitted by Sean on

That's really interesting, Berk. Thank you for the response!

Final question - to anyone, really, but Berk would welcome your thoughts. Of course if Jeremy or Johannes want to jump in - please do!

Berk, I was interested in your response to the chart in the CGD blog post that looked at 235 vs 217 vs 188, and its accompanying interpretation, which I think you disagree with. But ... how DO we make sense of the negative spillovers - both a) technically and b) substantively?

Technically, let me just make sure I have it right.

Treatment - 235
Within-village Control (Spillover Group) - 188
Across-village Control (Should be used used to measure ITT) - 217

Let's cross out the "across-village control."

Now left with -

Treatment - 235
Within-village Control (Spillover Group) - 188

Why do we have to assume that the negative spillovers WITHIN the within-village sample apply more broadly?

If we don't have to assume that -

a. what is the best PLAUSIBLE guess as to what happened to those households that (may have) experienced spillovers?

b. is there available data on what happened to the other households WITHIN the village, who were neither the a) treatment group, b) spillover group?

Thank you, again!