Difference in the result of regressing when using Stata and SAS ?

[Phil_NZ]

Hi all SAS Users,

I am in a critical issue that when I use Stata and SAS, I have different results

What I want is that I want to run a Difference-in-Difference regression by using OLS with fixed effects (firms (TYPE) and years(yr)).

My code in STATA is:

encode TYPE, generate(TYPE2)


areg wROE pt wTOT_ASS_TUR wFIRM_SIZE LNGDP UNEMPLOYMENT INFLATION wLEVERAGE pri_ove_ern i.yr, a(TYPE2)

My result by running in STATA is:

As can be seen from the picture above, the coefficient of pt is 0.0026, positive.

However, I replicate this regression in SAS by using the code below:

proc glm data=merge_treat_con;
    ABSORB TYPE;
    class yr;
    model wROE= pt TOT_ASS_TUR wFIRM_SIZE LNGDP UNEMPLOYMENT INFLATION 
		 wLEVERAGE pri_ove_ear yr/solution ss3;
run;
quit;

It generates a WARNING but I hope it is not a critical mistake

WARNING: Formatted ABSORB values not sorted in ascending sequence.

And the result is as below:

Boom, the coefficient of pt now is -0.016, negative.

I am wondering which cause these difference and what is the way to get out of this situation?

I highly appreciate any comment and suggestion. Thank you for spending time with me in advance.

Warm regards.

Phil.

[PaigeMiller · Accepted solution]

YR1991 is in the SAS output, but not in the STATA output. That may be a different parameterization of the model, which of course would result in different coefficients. Or maybe it's something different.

The F-test in SAS has 44781 degrees of freedom in the numerator, but STATA shows only 33 degrees of freedom in the numerator of the F-test.

So, these are not fitting the same model.