Re: Difference-in-Difference using SPSS

Posted by Nils Braakmann-2 on Apr 27, 2021; 2:42pm
URL: http://spssx-discussion.165.s1.nabble.com/Difference-in-Difference-using-SPSS-tp5740485p5740489.html

Depends on how sophisticated your DiD is supposed to be. The simple 2*2 design is basically just a regression with a treatment group effect, a post effect and their interaction. Designs with more units and time periods but where everyone who is treated is treated at the same time are straightforward extensions of that principle, i.e., more group fixed effects and/or more time effects and more interactions. If you get into differential treatment timing, it becomes more complicated. This (http://goodman-bacon.com/pdfs/ddtiming.pdf) paper by Andrew Goodman-Bacon (forthcoming in the Journal of Econometrics) is a good description of the problem. There are some new specialised estimators that deal with DiD designs in that context, e.g., https://www.aeaweb.org/articles?id=10.1257/aer.20181169, https://www.sciencedirect.com/science/article/pii/S030440762030378X or https://www.sciencedirect.com/science/article/pii/S0304407620303948. Most of these have code in R, some in Stata, probably none off the shelf in SPSS.

On Tue, Apr 27, 2021 at 9:10 AM 3J LEMA <[hidden email]> wrote:

I wonder if we can perform Difference-in-Difference analysis in SPSS.  Appreciate your comments.

Thank you. 

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

--
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD