I am trying to implement IV-PPML in stata and fairly new to this command. I am using the “ivpoisson” command to implement it.
I am interested in the signs of corruptions indices of the origin and destination countries (CCI_WB_d, CCI_WB_o) on trade flows. Both of these variables are considered to be endogenous and I am instrumenting them with their lagged values. I have given my commands below for your reference.
ivpoisson gmm FLOW LGDP_o LGDP_d ldist Population_o Population_d Contig comlang_off colony rta landlocked (CCI_WB_d CCI_WB_o = lagCCI_WB_d lagCCI_WB_o) if year >1994, add twostep wmatrix(cluster pair) vce (cluster pair)
I have few questions and I would really appreciate your clarifications/comments.
  1. Am I using the appropriate options to reflect the IV-PPML given that I am using ivpoisson to estimate it?
  2. If I am controlling for time effects by using year dummies, the instruments are dropped as they found to be collinear with year dummies. Is it fine to estimate the model without controlling for year fixed effects?
  3. Once I estimate the model without year fixed effects, I am finding the following message:
“convergence not achieved
The Gauss-Newton stopping criterion has been met but missing standard errors indicate some of the parameters are not identified.”
Can you please suggest me what could be done to achieve convergence?
  1. Can I control for origin and destination fixed effects? While reading the statalist, I found that we cannot control for these fixed effects. Can you please provide me some appropriate citation for this?
  2. Is there any way I can get the first stage regression outcomes similar to those in 2SLS?
Many thanks