coadywing.bsky.social
@coadywing.bsky.social
Trick is to think about the person-specific first stage treatment effect of the instrument (z) on treatment take up (d). It’s Delta_i = d_i(1)-d_i(0).
If it’s impossible for people assigned to control to sneak into treatment, then Delta_i is 0 or 1 for everyone— no defiers.
If it’s impossible for people assigned to control to sneak into treatment, then Delta_i is 0 or 1 for everyone— no defiers.
October 25, 2023 at 10:16 AM
Trick is to think about the person-specific first stage treatment effect of the instrument (z) on treatment take up (d). It’s Delta_i = d_i(1)-d_i(0).
If it’s impossible for people assigned to control to sneak into treatment, then Delta_i is 0 or 1 for everyone— no defiers.
If it’s impossible for people assigned to control to sneak into treatment, then Delta_i is 0 or 1 for everyone— no defiers.
I would say that this scenario is an experiment with one-sided non-compliance due to incomplete take up in the treatment group.
But people assigned to treatment who refuse treatment (and so get “control” instead) are not likely “defiers” in the IV sense of the word.
But people assigned to treatment who refuse treatment (and so get “control” instead) are not likely “defiers” in the IV sense of the word.
October 25, 2023 at 10:04 AM
I would say that this scenario is an experiment with one-sided non-compliance due to incomplete take up in the treatment group.
But people assigned to treatment who refuse treatment (and so get “control” instead) are not likely “defiers” in the IV sense of the word.
But people assigned to treatment who refuse treatment (and so get “control” instead) are not likely “defiers” in the IV sense of the word.
Another fun one is randomly assigned prices as the instrument and consumption of the good as treatment.
If the good is inferior and for some consumers the good is a Giffen good then you’d have “defiers”, I think.
These are silly. But good for discussion maybe.
If the good is inferior and for some consumers the good is a Giffen good then you’d have “defiers”, I think.
These are silly. But good for discussion maybe.
October 25, 2023 at 2:38 AM
Another fun one is randomly assigned prices as the instrument and consumption of the good as treatment.
If the good is inferior and for some consumers the good is a Giffen good then you’d have “defiers”, I think.
These are silly. But good for discussion maybe.
If the good is inferior and for some consumers the good is a Giffen good then you’d have “defiers”, I think.
These are silly. But good for discussion maybe.
A fun defiance scenario is when the instrument is a cash incentive: people are randomly offered $100 to take the treatment or $0 to take the treatment.
Defiance happens if extrinsic motivation crowds out intrinsic motivation for share of the population.
Defiance happens if extrinsic motivation crowds out intrinsic motivation for share of the population.
October 25, 2023 at 2:31 AM
A fun defiance scenario is when the instrument is a cash incentive: people are randomly offered $100 to take the treatment or $0 to take the treatment.
Defiance happens if extrinsic motivation crowds out intrinsic motivation for share of the population.
Defiance happens if extrinsic motivation crowds out intrinsic motivation for share of the population.
You can have one-sided non-compliance — for example, an experiment where everyone assigned to control gets control, but some people assigned to treatment refuse to participate.
But that is not “defiance” in the IV assumption sense.
But that is not “defiance” in the IV assumption sense.
October 25, 2023 at 1:49 AM
You can have one-sided non-compliance — for example, an experiment where everyone assigned to control gets control, but some people assigned to treatment refuse to participate.
But that is not “defiance” in the IV assumption sense.
But that is not “defiance” in the IV assumption sense.