Programming the statistical procedures from SAS

interpreting dummy variable in Heckman 2SLS

Posts: 23

interpreting dummy variable in Heckman 2SLS

**I APOLIGIZE FOR DOUBLE POSTING THIS QUESTION,** but I am getting no bites on the other board.  My question is regarding interpreting the output of a Heckman 2SLS.



Background information: I am exploring how institutional investor presence affects investments and firm performance during recessions.  Yet, there is a bit of a selection issue because institutions may choose to invest in firms which have certain investment strategies or performances.  And so, I employ the Heckman 2SLS model.
My code is:
proc qlim data = data heckit;
model dumhi = dumsp500 capx q lnat salesgrowth lev / discrete;    *selection equation;
model capx q liq salesgrowth vol rec / select(dumhi=1);                  *equation of interest;
In an OLS or fixed effects model, I would estimate the model:
capx = dumhi rec rec*dumhi q liq salesgrowth vol
where the coefficient on the interaction term is the coefficient of interest.
And now my question:
In the second stage of Heckman 2SLS where the inverse mills ratio of my institutional investor equation is included in the estimation of investments, I cannot figure out how to include the institutional investor-recessionary interaction term due to a linearity issue.  
I did include "rec" already -- is this basically the same thing as the interaction variable? I ask this because when we set up the second equation, we are imposing institutional investors = 1. No?
So, does "rec" in the Heckman model essentially tell me the same thing as "rec*dumhi' in OLS model since we are imposing dumhi=1 in the selection criteria? Or am I not clear in my understanding of the model???
Any help or thoughts is appreciated.
Super User
Posts: 18,580

Re: interpreting dummy variable in Heckman 2SLS

It's a statistical methodology question, you should probably post on CrossValidated ( as well.  


There are several very smart statisticians on here, but the statistical questions have a longer response time. 

I'm going to move this to statisical procedure forum though it rarely affects response time. 

Posts: 23

Re: interpreting dummy variable in Heckman 2SLS

Thank you for the direction, Reeza!!!

Ask a Question
Discussion stats
  • 2 replies
  • 2 in conversation