Hi all,
I am dealing with sample selection bias in a paper I am writing. I have selected my exclusion restrictions that I believe affect selection and use the heckman twostep command to deal with the issue of potential sample selection bias. I am however concerned that the exclusion restrictions I have used are weak and wish to address this. From my understanding Woodbridge (2010) sets out 3 assumptions that exclusion restrictions must meet:
1. First, the independent variable of interest (exclusion restriction) is available for the broader population, while the dependent variable is only available in the selected sample.
2. Second, an omitted variable creates a correlation between the two error terms (i.e., e and u) in the selection and structural equations. This is verified by the significance of lamda.
3. The Independent variable (exclusion restriction) is not correlated with either e or u, and is thus exogenous in both stages.
Now my exclusion restrictions meet the first two assumptions, however I am concerned that the exclusion restrictions I use are not exogenous in both stages. My question is, how would I address this and verify that all the exclusion restrictions I use are exogenous and are in fact "strong" variables, i.e is their a command I can use?
Any help is much appreciated,
Many Thanks,
Amit
I am dealing with sample selection bias in a paper I am writing. I have selected my exclusion restrictions that I believe affect selection and use the heckman twostep command to deal with the issue of potential sample selection bias. I am however concerned that the exclusion restrictions I have used are weak and wish to address this. From my understanding Woodbridge (2010) sets out 3 assumptions that exclusion restrictions must meet:
1. First, the independent variable of interest (exclusion restriction) is available for the broader population, while the dependent variable is only available in the selected sample.
2. Second, an omitted variable creates a correlation between the two error terms (i.e., e and u) in the selection and structural equations. This is verified by the significance of lamda.
3. The Independent variable (exclusion restriction) is not correlated with either e or u, and is thus exogenous in both stages.
Now my exclusion restrictions meet the first two assumptions, however I am concerned that the exclusion restrictions I use are not exogenous in both stages. My question is, how would I address this and verify that all the exclusion restrictions I use are exogenous and are in fact "strong" variables, i.e is their a command I can use?
Any help is much appreciated,
Many Thanks,
Amit
