Fama and MacBeth two-step procedure with logit regressions

[Jovana Cadenovic]

Dear members,

Is there a possibility to run a Fama and MacBeth two-step procedure in STATA where logit regressions are run for each time period and then the average of coefficients are obtained with t stats showing their significance?

Thank you in advance!
Best
Jovana

[Joro Kolev]

You can do whatever you want, the computer will not break.

But I have never seen anybody doing it.

Why would you do something like this? What is the purpose here?

[Jovana Cadenovic]

Dear Joro,

I was hoping if there is a possibility to do it using xtfmb command somehow?
I have a panel data with a dummy as a dependent variable, the logit model fits.

[Erik Rothi]

Dear Jovana,

have you found a solution for this?

I am trying to do the same thing: Running logit regression for each time period and average the coefficients according to the Fama and MacBeth procedure. I suppose it would be an extremely strenuous thing to do by running it manually, storing the results etc.

Also happy if anyone else has any suggestions on this or a workaround.

Thanks and kind regards
Erik

[Jovana Cadenovic]

Dear Erik Rothi

I have just seen your post. Yes, I made a code that still requires some manual work.


statsby _b _se e(r2_p), by(year): logit Y X1 X2 , vce (cluster ID)
export excel year Y_b_X1 Y_b_X2 Y_b_cons Y_se_X1 Y_se_X2 Y_se_cons _eq2_stat_1 using "by_year", firstrow(variables) nolabel keepcellfmt

global x Y_b_X1 Y_b_X2 Y_b_cons
gen sqrt_N= sqrt(number of years)
foreach var of global x{
egen `var'_mean=mean(`var')
egen `var'_sd= sd(`var')
gen `var'_tratio=sqrt_N*`var'_mean/`var'_sd
}
export excel Y_b_X1_mean Y_b_X1_sd Y_b_X1_tratio Y_b_X2_mean Y_b_X2_sd Y_b_X2_tratio Y_b_cons_mean Y_b_cons_sd Y_b_cons_tratio using "mean_values", firstrow(variables) nolabel keepcellfmt

It does require some code writing; I am pretty new in STATA so I did step by step like this.

Hope it helps.

Best
Jovana