C-Score: by-year regressions

[River Huang]

Dear All, Suppose that I have the following data, i.e., 500 firms (id) over 20 years (year).

clear
set obs 500
gen id = _n
expand 20 
bys id: gen year = _n+1999

set seed 123
gen X = rnormal()
gen R = rnormal()
gen D = (R<0)

gen Size = rnormal()
gen MB = rnormal()
gen Leverage = rnormal()

I wish to calculate the C-score (for each `id' and each `year') as described below:

Any suggestions are highly appreciated.

[River Huang]

Dear All, I tried the following code, and wish to receive comments from anyone. Note that you have to ssc install runby first.

foreach v of varlist Size MB Leverage {
    gen R_`v' = R*`v'
    gen DR_`v' = D*R*`v'
    gen D_`v' = D*`v'
}
gen DR = D*R
 
reg X D R R_Size R_MB R_Leverage DR DR_Size DR_MB DR_Leverage ///
    Size MB Leverage D_Size D_MB D_Leverage
      
cap program drop C_score
program define C_score
  reg X D R R_Size R_MB R_Leverage DR DR_Size DR_MB DR_Leverage ///
      Size MB Leverage D_Size D_MB D_Leverage
  gen C_score3 = _b[DR]+(_b[Size]/_b[DR])*Size+(_b[MB]/_b[DR])*MB+(_b[Leverage]/_b[DR])*Leverage
end

runby C_score, by(year)

[Clyde Schechter]

This code does not generate the same variable that equation (7) shown in #1 defines.

That equation uses all of the gamma coefficients, which would be the coefficients _b]DR], _b[DR_Size], _b[DR_MB], and _b[DR_Leverage] from the regression in #2. The code shown in #2, however, instead uses_b[DR], _b[Size]/_b[DR], _b[MB]/_b[DR], , and _b[Leverage]/_b[DR]. Only _b[DR] is the same in both places.

Not being in finance, I have no idea what C_score is supposed to be, so I also have no idea whether equation (7) is correct or whether the code in #2 is correct. But I am completely certain that they are computing different things.

[River Huang]

Dear Clyde, Thanks for the comments. I will go through the code again.