Multiple regression in an event study

#1
Dear all,

I am currently doing an event study about COVID-19 and the differences between the G7 and BRIC. To determine possible important factors, I am considering looking at country specific characteristics (GDP per capita, Exports, country density) and COVID-19 specific data (amount of cases, health expenditures etc).

However, most of the data that is available of the country specific characteristics (like GDP per capita or amount of exports), has the same value during 3 months up to a year. I understand that I should not expect this to change daily, but is it possible to regress abnormal returns with variables that are mostly constant during these periods? Furthermore, if I choose (-3, +3), should I regress the exports and GDP per capita over the same 7 days (so that will mean I am using two constant variables) or can I regress abnormal returns with the past 2 years of GDP per capita data?

Thanks so much in advance,

David
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
If your values for variables won't change or maintain a near steady state you can put them in your model as fixed effects (e.g., similar to say, sex when modeling something at the person-level).
 
#3
Does doing that lose explanatory power? Or is it still possible to compare (lets say Russia's abnormal returns and USA's abnormal returns) with each other using the 'fixed' export variable?
 
#4
Fixed effects is the way I would go. But if I can offer up some more advise, i'd explore more variables that are measured at a frequency that better lines up with your target e.g. CPI, # of new building permits, etc.
 
#5
Fixed effects is the way I would go. But if I can offer up some more advise, i'd explore more variables that are measured at a frequency that better lines up with your target e.g. CPI, # of new building permits, etc.
Thanks! Will try do search for some more variables that are measured at a higher frequency!