DID models for non-panel data

[miharbi]

Hi everyone,

I found in the EViews Help this statement: (To estimate DID models on non-panel workfiles, you will need to specify the appropriate least squares model manually using standard least squares regression.)

What does this statement mean, especially for non-panel data structures? I hope to find a practical example of this point?

[EViews Gareth]

Difference in Difference estimation is just standard OLS. EViews doesn't have a built in DiD estimator for non-panel workfiles, but since it is just OLS, you can estimate it with the OLS estimator.

[miharbi]

Thank you Gareth for your response!

I have no problem with what you said, but I want to understand the structure of non-panel data. What do you mean by it? Is there a practical example or a paper that has addressed this?

[startz]

Suppose you have an outcome variable y and two dummy variables, treated and after. (So some observations of y are treated after a certain date and others are never treated--the control group.) You could do

Code: Select all

ls y c treated after treated*after

The last coefficient is the DiD estimate.

[miharbi]

Thank you Starz,
It's very clear now.