Steps to compute HHI in SAS

1. At minimum, you need SICH, SALE, and FYEAR from COMPUSTAT.
2. SICH is four digits. As an example, if you want to compute HHI by three-digit SIC, do the following:
   char_sic=put(sich,4.0);*This converts SICH from numeric to character; three_char_sic=substr(char_sic,1,3);*This takes the first three components of the string; sicthree=input(three_char_sic,best3.);*This converts the character back;
3. We first need to compute the total size of the market for a given industry-year:
   proc summary data=sic; by sicthree fyear; var sale; output out=hhidenom (drop=_type_ _freq_) sum=; run;
4. Now, we can compute the HHI:

   *Sort before merge; proc sort data=sic; by sicthree fyear; run; proc sort data=hhidenom; by sicthree fyear; run;

   *Merge denominator; proc sql;*http://sbaleone.bus.miami.edu/PERLCOURSE/SASFILES/SQL_EXAMPLES.sas; create table compstatdenom as select a.*, b.sale as denom from sic a left join hhidenom b on (a.sicthree=b.sicthree) and (a.fyear=b.fyear); quit;

   *Create market share; data compstatshare; set compstatdenom; share=sale/denom; sharesq=share**2; run;

   *Create HHI; *http://www.talkstats.com/showthread.php/10892-Sum-up-values-by-groups; proc summary data=compstatshare; by sicthree fyear; var sharesq; output out=sumhhi (drop=_type_ _freq_) sum=; run;

   *Merge HHI; proc sql;*http://sbaleone.bus.miami.edu/PERLCOURSE/SASFILES/SQL_EXAMPLES.sas; create table compstathhi as select a.*, b.sharesq as hhi from compstatshare a left join sumhhi b on (a.sicthree=b.sicthree) and (a.fyear=b.fyear); quit;

Dates in SAS

Suppose you imported data from a spreadsheet with dates that look like 20120409. How can you convert them to the familiar date9 format in SAS?

If they are already in character format, do the following:

datadate2=input(datadate,mmddyy8.);
format datadate2 date9.;

Take care to replace the mmddyy8. depending on how the date is initially written.

If they are in numeric format, then do the following:

date13dchar=put(date13d,8.);
date13d2=input(date13dchar,yymmdd8.);
format date13d2 date9.;

That is, you first convert numeric to character and then character to date. Be sure to keep in mind that this process is vastly made easier if you take care to input the dates in a "nice" format to begin with in the spreadsheet.

How do I identify runs of consecutive observations in panel data?

The STATA mailing list has a way to identify runs of consecutive observations. With some googling, SAS can do the same thing. Here's how.

Suppose you want to figure out how many observations you have per GVKEY:

data merge2; set merge; by gvkey; cnt+1; if first.gvkey then cnt=1; run;
proc sort data=merge2; by gvkey descending cnt; run;
data merge3; set merge2; by gvkey; retain totcnt; if first.gvkey then totcnt=cnt; output; run;

Merging with CIK

char_cik=put(cik,z10.)

CIKs are 10-digit characters, but if you manually code them from EDGAR, they won't have leading zeroes. So use "z10."

Averages across years

PROC MEANS nolabel DATA=[dataset] ;
CLASS [year];
VAR [variable];
OUTPUT OUT=[dataset] MEAN= ;
RUN;

MEAN can be replaced by MEDIAN.

Source

Esttab and multiple dummy indicators

Suppose your empirical specification has both unit and time fixed effects. You don't the table to be cluttered with n or m variables, do you?

Including indicate("Time fixed effects = " "Unit fixed effects = ") after esttab will do the trick. Note the STATA output to determine what to put after the equal signs.

Ado Path

Emerald at UNC has STATA 11.2 but is not cooperative when it comes to downloading *.ado files (e.g. estout). So I downloaded estout in STATA 9.2 through Latte at Fuqua and then copied the folder to my Emerald account. Do note that you must use "scp" and not "sftp" due to the latter's restriction on recursive copying (i.e. folders).

The help for adopath suggested that the folder be copied to the "/netscr/[username]" folder. After that, esttab worked just fine.