Proc Panel HAC yields no standard errors

shenflow — Thu, 23 Jul 2020 12:17:01 GMT

Estimating the following:

proc panel data = input;
	id gvkey rankyear;
	model pers_a_diff = readj_quant readj_avg 
		 				/ fixtwo noint;
	ods output FitStatistics = fit;
run;
quit;

yields:

However, I need HAC standard errors. Inserting the HAC option in the statement above as follows:

proc panel data = input;
	id gvkey rankyear;
	model pers_a_diff = readj_quant readj_avg 
		 				/ fixtwo noint hac;
	ods output FitStatistics = fit;
run;
quit;

yields the following results:

That is, I do not get any standard errors at all. Why is that?

EDIT: What strikes me as even more surprising is that NEWEYWEST in fact yields standard errors (and NEWEYWEST are a special case of HAC standard errors as far as I am concerned).

Re: Proc Panel HAC yields no standard errors

SteveDenham — Fri, 24 Jul 2020 18:17:04 GMT

Here is something buried in the documentation about HAC:

This option is not available for between models and cannot be combined with the HCCME option.

I suspect that your model falls into the first category.

SteveDenham

topic Re: Proc Panel HAC yields no standard errors in Statistical Procedures

Proc Panel HAC yields no standard errors

Re: Proc Panel HAC yields no standard errors