Test significant difference of individuals' coefficients

[Julia Martinsen]

Hi all,

I have a panel data with countries and different macro variables per each country, over time. The analysis I am running consists in regressing one macro variable on another for each country and what I am interested in is the coefficient (beta). I.e.

regress macro_a macro_b if country_id == 1

I'm using a loop and I get the coefficient for each country - let's call it beta_b. What I want to do now is to test whether these beta_b are different between countries. I tried using "test", with the following code, but it doesn't work:

foreach y of global c_all {
  regress macro_a macro_b if country=="`y'"
  mat beta=e(b)
  svmat double beta, names("`y'")
}
test country_1 == country_2 == coutnry_3

What would you suggest using?

The error I get is:

country_1 not found

when country_1 appears in my variables list.

[Clyde Schechter]

I don't think you're on the right track with this approach. If I were going to do this, I would do it as follows:

regress macro_a .i.country_id##c.macro_b
testparm i.country_id#c.macro_b

The idea is to simultaneously regress in all countries, and have a separate slope for each country by using country#macro_b interaction. Then you can test whether they are all equal at the end with the -testparm- command.

-regress- is OK here even though you have panel data, because it also includes the indicators for the countries, so that is equivalent to -xtreg, fe-.

If you are not familiar with the i., c. and ## notation, read -help fvvarlist- and the associated chapter in the manuals.

[Julia Martinsen]

Thank you Clyde!

I use -regress- cause I'm kind of running a time series when I write -if country_id==1- and I control for time series issues using HAC s.e. (I didn't include them for simplicity).

[Clyde Schechter]

Richard Williams has pointed out that there is a typo in the code in #2: I mistakenly put an extra . in front of i.country_id. It should read:

regress macro_a i.country_id##c.macro_b

Thank you, Richard. And sorry, Julia, if that mistake threw you off track.

[Julia Martinsen]

I saw that but imagined it was a typo - so no prob! Thanks