options nocenter ls=80;

libname temp 'd:\empiricalmethods\jagwangcapm\';
filename tempdir 'd:\empiricalmethods\jagwangcapm\';

*Read Returns on the size portfolios (pid identifies different portfolios);
data rets;
  infile tempdir(dat.ret) missover firstobs=5;
  input month ret;
  if month=. then delete;
  if _n_=1 then pid=1;
  else if month=1 then pid=pid+1;
  retain pid;
proc sort;
  by month;
proc print data=one (obs=100);
proc means;
run;

*Read Value Weighted Index Returns;
data index;
  infile tempdir(dat.vwd) missover firstobs=2;
  input month vw;
proc means;

*Read Factors Prem and Labor;
data facs;
  infile tempdir(dat.fc1) missover firstobs=2;
  input month prem labor;
proc means;
run;

data rets;
  merge rets (in=A) index (in=B) facs (in=C);
  by month;
  if A and B and C;
  premvw=prem*vw;
  proc sort;
    by pid month;
proc means;
run;

* get vwbeta;
proc reg data=rets noprint outest=betavw;
  model ret=vw;
  by pid;
run;

data betavw (keep=pid betavw);
  set betavw;
  betavw=vw;
proc print data=betavw;
run;

* get prembeta;
proc reg data=rets noprint outest=betapr;
  model ret=prem;
  by pid;
run;

data betapr (keep=pid betapr);
  set betapr;
  betapr=prem;
proc print data=betapr;
run;

*Get mean returns over the sample period for each portfolio;
proc means data=rets noprint;
  var ret;
  by pid;
  output out=mret mean=mret;
run;

*Merge data and estimate regression of sample mean returns.;
*This gives R-Squared from the paper;
data all1;
  merge mret (in=A) betavw (in=B) betapr (in=C);
  by pid;
  if A and B and C;
run;

proc reg data=all1 outest=estr edf;
  model mret=betavw betapr;
run;
proc print data=estr;
 var _rsq_;
run;

*Merge betas into monthly returns and estimate FM regressions to get t-stats in the paper;
data all;
 merge rets (in=A) betavw (in=B) betapr (in=C);
 by pid;
 if A and B and C;
 betatv=betaa+betab*prem;
 proc sort;
   by month;
run;

proc reg data=all outest=est1 noprint;
  model ret=betavw;
  by month;
run;

proc means data=est1 n mean t prt;
run;

proc reg data=all outest=est1 noprint;
  model ret=betavw betapr;
  by month;
run;

proc means data=est1 n mean t prt;
run;