[Paul Leeds]

If you are merging or appending different files, it is certainly possible that bad crisis codes could overwrite, or get added to, good crisis codes. I suggest that, once you finish combining files, you delete any existing crisis variable and regenerate it from scratch. If that doesn't solve the problem, then I will go to the Notre Dame Grotto and light a candle for you.

Lets hope that I don't need the candle anyway

I'l redo all the models again with the new crisis variable together with the new interpretation for the new model with the interaction which seems different now.
The interaction term now is not significant and make the crisis coefficient insignificant too (it was significant in the model without the interaction term).

[Richard Williams]

Is the problem solved? If so you may wish to go back and also fix the intermediate files so the problem doesn't come back to haunt you later.

[Paul Leeds]

I managed luckily to find a dta copy of the datasets which retains the correct crisis variable. The issue is that I will have to go back to my results section and tell another story.
Nevertheless, I have in mind another interaction term which seems highly significant as well and that may substitutes the old term which is now insignificant.

[Richard Williams]

Great. Don't rely on luck though. Hold on to intermediate data sets and do files and document the data manipulation that you do. You want to be able to reproduce what you did 5 years from now, and if there are mistakes you want to be able to track them down. (This is in the do as I say and not as I do category -- I'm much better about such things now but I would have a hard time perfectly replicating things I did decades ago.)

[Paul Leeds]

You are 100% right and now I am being punished for my mistake of not tracking changes made to my datasets regularly.