Furthermore, even if what you care about is the effect of treatment with perfect compliance, the effect of intention to treat is still probably the first statistic you should look at. The most important point of intention to treat is that it is likely to measure the sign of the effect correctly, whereas a naive “as treated” measurement often won’t due to selection bias. (And, cynically, non-naive “as treated” measurement will tell whatever story the authors want it to tell)
Furthermore, even if what you care about is the effect of treatment with perfect compliance, the effect of intention to treat is still probably the first statistic you should look at. The most important point of intention to treat is that it is likely to measure the sign of the effect correctly, whereas a naive “as treated” measurement often won’t due to selection bias. (And, cynically, non-naive “as treated” measurement will tell whatever story the authors want it to tell)