Hi all,
I'm hoping to ask a question regarding the use of standardized coefficients under a 2SLS model. I have an econometrics project to do and my supervisor has recommended that rather than looking at the Beta's of my coefficients for interpretation, using the standardized coefficient would be useful to see which of my explanatory variables have the largest effect on my dependent variable.
I understand how to do this in that I look at the coefficient of my explanatory variable and then multiply it by the standard deviation of that said explanatory variable, which can the be divided by the standard deviation of my dependent variable.
BUT, my question is regarding explanatory variables that have been instrumented for. In my case, I am considering a just-identified 2SLS model where I instrument income with one instrumental variable, I will call that instrumental variable K in this thread for simplicity.
When interpreting the coefficient of my income explanatory variable in my 2SLS model, in adopting the standardized coefficient approach, should I multiply that coefficient by the standard deviation of my income explanatory variable or by the standard deviation of K?
My intuition is to choose the former of above, because if I deduced that if I had a case where I had an over-identified model using two instruments for income, I'm not sure how you would use standardized coefficients to interpret it.
However I am unsure and so would be grateful for any advice on this issue.
Thanks a lot,
Mike Dev
I'm hoping to ask a question regarding the use of standardized coefficients under a 2SLS model. I have an econometrics project to do and my supervisor has recommended that rather than looking at the Beta's of my coefficients for interpretation, using the standardized coefficient would be useful to see which of my explanatory variables have the largest effect on my dependent variable.
I understand how to do this in that I look at the coefficient of my explanatory variable and then multiply it by the standard deviation of that said explanatory variable, which can the be divided by the standard deviation of my dependent variable.
BUT, my question is regarding explanatory variables that have been instrumented for. In my case, I am considering a just-identified 2SLS model where I instrument income with one instrumental variable, I will call that instrumental variable K in this thread for simplicity.
When interpreting the coefficient of my income explanatory variable in my 2SLS model, in adopting the standardized coefficient approach, should I multiply that coefficient by the standard deviation of my income explanatory variable or by the standard deviation of K?
My intuition is to choose the former of above, because if I deduced that if I had a case where I had an over-identified model using two instruments for income, I'm not sure how you would use standardized coefficients to interpret it.
However I am unsure and so would be grateful for any advice on this issue.
Thanks a lot,
Mike Dev
