Quartile in DID analysis.

[Pranshu Tripathi]

I am working on panel data having 1410 firms, six variables, and eight years.
I have a law as the treatment, which was passed in 2016. So, 2013-2016 is the pre-treatment period, whereas 2017-2020 is the post-treatment period.
Also, I have categorized the firm into two groups.

I run a regression with three-way interactions where a continuous dependent variable interacts with the other two dummy variables. Now I want to divide my data into four quartiles based on the deviation of the dependent and the industry-based median. (observed value-industrial median).

Now the problem is that I have grouped the firms based on the average of the first four years (average of the pre-treatment period), so it is constant over time. If I categorize the deviations based on the quartile, the observed values of the same firm will fall into different quartiles in different years. This would affect the causality. I am thinking of taking out the average/median of the deviation of the particular firm (first four years) and then checking in which quartile it falls. Is it the right way to do so?

Thanks and Regards

[George Ford]

My advice is to do a basic DD analysis first and get a handle on what it's telling you. Soup it up from there.

[Pranshu Tripathi]

Dear George
The third-order interaction is insignificant. Now, I want to check whether the different level of deviations has different results.

[George Ford]

You have two types of firms, I presume. Are all firms treated, or just some firms (of which there are two types)?

[Pranshu Tripathi]

I have two groups. One is treated and other is kept as control.