I am using Stata 14, and the command ivmediate.
This command is first used in:
Dippel, C., Gold, R., Heblich, S., & Pinto, R. (2022). The effect of trade on workers and voters. The Economic Journal, 132(641), 199-217.
And it was first elaborated in:
Dippel, C., Gold, R., Heblich, S., & Pinto, R. (2017). Instrumental variables and causal mechanisms: Unpacking the effect of trade on workers and voters (No. w23209). National Bureau of Economic Research.
This is what I am running:
ivmediate Y X1 X2 X3 X4, mediator(M) treatment(T) instrument(Z) vce(cluster X5) full
One of the outputs (and perhaps the most important one) is the % of the total effect of T on Y that passes through M.
I have two problems, and I cannot find their solution or at least their explanation.
- regardless of the mediator that I use, the result is always 0.38%
- this percentage is far from what I compute by “hand” (as explained in Section 3.2 of Dippel et al. (2022), the estimated indirect effect and the effect calculated effect should be the same, besides from rounding)
Is anybody facing the same issue?
Can anybody explain the reason for my two issues?
Any error in my code? (I am replacing the actual name of the values, besides that, the code is the same)