Forecasting using SAS Forecast Server, SAS/ETS, and more

2-stage Heckman

Posts: 23

2-stage Heckman

I have a question regarding the output of my heckman model.  I am testing how institutional investor presence affects firm investments.  However, it has been argued that institutions choose to invest in firms based on the firm investment strategies, performance, etc.  


I am proposing using a 2SLS Heckman approach to correct for this sample bias.  However, I also have a simultaneity issue because what I'm really interested in is how institutional investors (dumii) affect investments (capx).


endogeneity problem:   dumii = capx q lnat salesgrowth lev;

equation of interest:      capx = dumii liq salesgrowth vol rec;



Would using Heckman be appropriate in this situation?  If I estimate equation (1) and substitute in invese mills ratio into (2), I cannot also include "dumii" as an independent variable because it is a linear combination of thee inverse mills ratio.  In this case, would the coefficient on the inverse mills ratio act as an instrumental variable for dumii?  Ie: will the coefficient on lambda give me a clue as to how institutional investors affect investments?


The code I run is:


Proc qlm data=data heckit;

Model dumii = adjcapxyq q lnat salesgrowth lev / discrete;                                 *selection equation;

Model adjcapxyq = liq salesgrowth vol rec / select(dumii=1);                              *equation of interest;




Thank you for your help in advance!!!

Ask a Question
Discussion stats
  • 0 replies
  • 1 in conversation