VARMAX for policy analysis

niam — Tue, 08 Sep 2020 12:46:26 GMT

Hi All,

Can we combine VARMAX procedure with Difference in Differences method? Here is the context of my study:

I have 3 time series, Y1t, Y2t and Xt, in which Y1t and Y2t change over time with changes in Xt. However, I also have a binary variable, Z, which is an indicator of a policy shift in the middle of the time series that only affects Y1t. So, after a specific point in time, the values of Y1t will also be affected by the policy implementation Z while there is no policy effects on Y2t, time series Xt will continue to impact both Y1t and Y2t after the policy implementation.

One thing that I can think of is to change Z into a time series with values equal to 0 before the policy shift date and 1 after the policy shift date. Does that work?

Re: VARMAX for policy analysis

dw_sas — Wed, 09 Sep 2020 20:36:20 GMT

Hi @niam ,

Yes, your idea of including your indicator variable, Z, in the equation for Y1 along with your other input variable, X, in your Y1 and Y2 equations is a reasonable approach. For example, you can specify the following MODEL statement in PROC VARMAX, where <options> is your list of desired options:

model y1=x z , y2=x / <options> ;

For more details on the MODEL statement syntax when a different set of input variables is used in each equation, please see the following link:

https://go.documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_varmax_syntax12.htm&docsetVersion=15.1&locale=en

I hope this helps!

topic VARMAX for policy analysis in SAS Forecasting and Econometrics

VARMAX for policy analysis

Re: VARMAX for policy analysis