Include instrumental variable in CMP model

[Faruque Sunny]

I am trying to employ the CMP model to estimate recommended fertilizer dosage adoption's impact on ROI. The dataset contains seven years of balanced panel data. The endogenous variable is the Adoption decision (AdoptNot), and the instrumental variables for 'AdoptNot' are ‘FarmExp’ and ‘logOffFarmEarnTk.’

Following Dr. David Roodman's work, I have tried to construct the model as:

cmp (ROI = AdoptNot# IrriMachineOwn Seekinfofertilizer ACode EduC HHnumberCat HHLabor FarmExp LandSizeDec LandTypeC Landownership SoilWaterRetain CowdungAvailability SoilFertility CreditAvail BRRIKnowledge logOffFarmEarnTk year2015_d year2016_d year2017_d year2018_d year2019_d year2020_d year2021_d) ( AdoptNot =IrriMachineOwn Seekinfofertilizer ACode EduC HHnumberCat HHLabor FarmExp LandSizeDec LandTypeC Landownership SoilWaterRetain CowdungAvailability SoilFertility EnviAwareness CreditAvail BRRIKnowledge year2015_d year2016_d year2017_d year2018_d year2019_d year2020_d year2021_d logOffFarmEarnTk FarmExp), indicators ($cmp_cont $cmp_probit)


But I am confused about the correct way of constructing this model. For instance, do I need to add the ROI variable in the second model too, and what is the correct order of placing instrumental variable/variables in the model?


Thank you very much in advance, and I look forward to hearing from you soon.

Kind regards,

Faruque

[Faruque Sunny]

Dear Dr. David Roodman, it would be a great help if you could review this post. Thank you very much in advance.

[David Roodman]

Yes, if you want AdoptNot to depend on ROI, then you need to include ROI in the AdoptNot equation. It doesn't matter where. The cmp paper and the cmp help file include examples of 2-stage/IV models.

[Faruque Sunny]

Dear Dr. David Roodman, Thank you very much for the reply. As the ROI depends on AdoptNot, I am not including ROI in the AdoptNot equation. I still have another question. Please look at the model; I used the instrumental variables in the ROI model as independent variables and in the AdoptNot model as instruments. Please suggest whether I should exclude the instrumental variables from the ROI equation or not.