[Minhaj uddin]

Thank you so much.

But if you consider the GFC to have ended in, say, 2015, and the Covid era doesn't begin until 2020 (or maybe 2019 if your data extends to the parts of the world, mostly China, where Covid activity was recognized in 2019), then I think the two would be separate enough to avoid this problem.

Exactly this is the case. So If I understand you correctly the two separate models won't have any issues?

[Clyde Schechter]

Well, you are selectively quoting what I wrote in #15. Based on your response in #16 then I think we can stop worrying about GFC and Covid eras overlapping, or nearly so. But you still have to consider the other objection I raised in #15 to doing two separate models:

Even if you defined PostGFC so that it returned to 0 before the Covid era started, using the separate model for Covid would only be appropriate if it is reasonable to believe that the real-world effects of the other variables returned to their pre-GFC values once the PostGFC era ended. Is that a reasonable assumption? I have no idea--but you should. If you are unsure, it would be safer to assume that they do not return to their pre-GFC values and allow the data to speak to this issue themselves.

It is a very strong assumption to make that the effects of all those variables returned to their pre-GFC values in the interval between the end of GFC and the start of Covid. Maybe it's true, and if you have evidence for that, then fine. But if you don't really know it to be true, I'd say it's a bit aggressive to build it into a model, as I would think it has only a very low probability of being true. I would still put GFC and Covid together in a single model unless you can counter this concern.

[Minhaj uddin]

Thank you Prof. for the response.

One issue in implementing both GFC and COVID together in a single model is that, in my model, there are around 8 other independent variables, and I am using GFC and COVID as a moderating variable, which means I will interact the two dummy variables with all other independent variables; this makes the total RHS variables count to 26 (including level terms). It seems a bit overwhelming, and I'm not sure if it's appropriate to have this many variables in one single model.

[Clyde Schechter]

Your concern about having two many variables is a reasonable one. Of course, it depends on your sample size. There are various rules of thumb about ratios of observations to variables, but even using the most permissive ones, you would need a sample size close to 1,000 observations to support 26 variables, and some would say you need more than twice that. If you do have a sample with several thousand observations, then I wouldn't worry bout having 26 variables, but with smaller samples you are right to be concerned.

One thing you might do to deal with the concern I raised in #17 is to first run the model without the Covid variable, but to use, instead of a GFC dichotomy, a three-level GFC variable that distinguishes pre-GFC, during GFC, and post-GFC. Then look to see if the post-GFC effects of your other variables are reasonably close to their pre-GFC effects. If so, that would say that things really do tend to revert (at least approximately) to the pre-GFC situation when the GFC ends. And in that case, running a separate model for the Covid effects analysis would be reasonable.

If, however, you find that the post-GFC effects of other variables are materially different from their pre-GFC effects, then your best bet might be to keep the Covid and GFC analyses separate, but for the Covid model, only include the post-GFC years.