Hello Stata Community,
I am estimating the impact of household's exposure to land grabbing on farm income and food security. My treatment variable (i.e. exposure to land grabbing) is dichotomous and captured as 1 if exposed to land grabbing and 0 if not. The treatment variable (i.e. exposure to land grabbing) was constructed based on households responses to three qualitative questions regarding land grabbing. These questions include whether households were (i) involved in negotiations over land deals (ii) informed (iii) compensated. A household answering no to at least one question was then classified as been exposed to land grabbing and a household answering yes to all is said to be non-exposed. however, my main problem is misclassification resulting from potential information bias which is feasible at the stage of data collection. For instance, households may answer no to compensation especially if they detect that they will be compensated for the loss or decline to report if they detect that providing such information might lead to further exposure to land grabbing. This, in turn, has the potential to bias the effect estimate.
My question is: Can I employ the Maximum likelihood estimation of endogenous switching regression model proposed by Lokshin and Sajaia (2004)?
I would be very grateful for your help.
Thank you.