Dear Ben:

Thanks for your email sent on Monday.

Yes, there are quite a few statistical packages that can let you use instrumental variables (IVs) in multiple regressions. I recommend SPSS as it is very easy to use. But, you can also conduct IV analysis by using other software like LISREL, SAS, or Stata. For example, in SAS, use SAS/ETS Pro Syslin with the 2SLS option. In Stata, use the ivreg command. You may want to use LISREL if your next step of modeling work is to build some structural equations.

They are basically two approaches that you can use instrumental variables in these packages. The first approach is that you perform two regression analyses in two steps. That is, you use instrumental variables for a regression at step one and use the predicted values saved from first step for your second regression at step two. Specifically, at the first step, regress your independent variables needing instrumental variables on the instrumental variables and other independent variables not needing instrumental variables. Then, save the predicted values to form some new variables. At the second step, regress your dependent variable on these new variables and other independent variables not needing instrumental variables. For example, if you have an equation Y = aX + bZ + u and have a set of Is as the instrumental variables for X needing them. In the first step, Regress X on Is and Z, AND save the predicted values as variable P. Then, in the second step, regress Y on Z and the new variable P.

The second approach is to let your statistical packages like SPSS or LISREL or SAS or STATA to perform these two steps for you all together. In other words, the 2SLS procedure in these computer programs can execute the above-mentioned two steps in one single step. In STATA, this can be done by using  ivreg Y Z (X=I). In SAS, use SAS/ETS Pro Syslin with the 2SLS option.

To implement this second approach in SPSS and in LISREL, use the followings instructions.

(1)   in SPSS

Step 1: click on File, then Read Text Data to read in your data file

Step 2: click on Analyze, then Regression, then 2-Stage Least Squares…

            A 2-Stage Least Squares box will open that you should (1) move your dependent variable to the box with Dependent: above it, then (2) move your instrumental variables AND your other independent variables not needing instrumental variables to the box with Instrumental: on the top, and (3) move all your independent variables (not IVs) to the box with Explanatory: on the top.

Click on OK to get your results.

(2)   in LISREL

Step 1:  click on File, then on  Import External Data in Other Formats to import your data into LISREL

Step 2:  click on Statistics, then on Two Stage Least-Squares. A Two Stage Least-Squares box will open for you to input your Y Variables, X Variables AND your Instrumental Variables.

Step 3: Select your output options by click on Output Options. Then, click on RUN to get your results.

For your convenience, I just created the following web page containing one example of using instrumental variables in SPSS that you may want to take a look.

You may prefer the first approach as it gives you more control over the process and allow you to conduct some diagnostics if necessary. Actually, we do need to take a look at the first step results to ensure we have strong instrumental variables. That is, we need to ensure the effects of our IVs are significant. The problem of weak instruments can make our 2SLS useless. Actually, most software packages can allow us to check the first step results when implementing the second approach. If in STATA, the software has the option first allowing us to check the first stage results. Usually, the results produced by the second approach are preferred as their st errors are more accurately estimated than that estimated by the first approach.

I think the above description is long enough already, and hope it covers all the necessary issues on this topic. If you need any more assistance or have some follow-up questions, please do not hesitate to contact me.

Sincerely,

RM Publications

RM Programs

RM Platforms