Calculates idiosyncratic volatility using time-series monthly/daily regressions for various risk models
Macro call with an example
The following example calculates idiosyncratic volatility (standard deviation of residuals) at the individual stock level (IBM, AAPL and MSFT) using estimation windows of 22 trading days (WINDOW) rolled forward by 1 day (STEP) with at least 15 trading days of non-missing returns (MIN) based on Fama-French (FF)model. The input dataset RETS should contain PERMNO identifier, date and return info. Output dataset RESULTS will contain estimates of idiosyncratic risk (standard devaition of residuals (IDRISK_STD), number of non-missing returns in the estimation window (NUSED) as well as beginning (inclusive) and ending (exclusive) date of the estimation window window (START_DATE and END_DATE, respectively)
proc sql;
create table rets
as select distinct permno, dlycaldt as date, dlyret as ret
from crsp.Stkdlysecuritydata
where permno in (14593, 12490, 10107) and dlycaldt between '01JAN2020'd and '31DEC2020'd;
quit;
%IDVOL (inset=RETS, outset=RESulTS, datevar=DATE, retvar=RET, freq=D, window=22, step=1, min=15, model=FF)Example
/* ********************************************************************************* */
/* ******************** W R D S R E S E A R C H M A C R O S ******************** */
/* ********************************************************************************* */
/* WRDS Macro: IDVOL */
/* Summary : Calculates idiosyncratic volatility using time-series monthly/daily */
/* regressions for various risk models */
/* Author : Denys Glushkov, WRDS */
/* Date : July 07, 2009 */
/* Update : November 2024 by Freda Drechsler for CRSP CIZ data format */
/* Variables : - INSET and OUTSET are input and output datasets */
/* - DATEVAR: name of the date variable in INSET dataset */
/* - RETVAR : name of the raw return variable in INSET dataset */
/* - FREQ : frequency of returns in incoming INSET dataset, */
/* m (monthly) or d(daily) */
/* - WINDOW : the length of the rolling estimation window in */
/* months/trading days over which the idiosyncratic volatility*/
/* is estimated */
/* - STEP : number of months/trading days by which the estimation */
/* window is rolled forward */
/* - MIN : minimum number of non-missing returns in the esitmation */
/* window required for generating valid estimates of IDVOL */
/* - MODEL : risk model used in estimation of idiosyncratic volatility */
/* m (market), ff (Fama-French 3 factor), ffm (FF+Momentum) */
/* ********************************************************************************* */
%MACRO IDVOL (INSET=, OUTSET=, DATEVAR=, RETVAR=, FREQ=, WINDOW=,STEP=, MIN=, MODEL=);
%local oldoptions errors;
%let oldoptions=%sysfunc(getoption(mprint)) %sysfunc(getoption(notes))
%sysfunc(getoption(source));
%let errors=%sysfunc(getoption(errors));
options nonotes nomprint nosource errors=0;
%let model=%lowcase(&model);%let freq=%lowcase(&freq);
/*Depending on the incoming return frequency, create ancillary macro variables*/
%if &freq=m %then %do;
%let file=monthly;
%let inc=month;
%let dateff=dateff;
%end;
%else %if &freq=d %then %do;
%let file=daily;
%let inc=day;
%let dateff=date;
%end;
/*Depending on the type of the risk model, create the variable list to be extracted*/
%if &model=m %then %let vars=mktrf; %else
%if &model=ff %then %let vars=mktrf smb hml; %else
%if &model=ffm %then %let vars=mktrf smb hml umd;
%put ### CREATING TRADING CALENDAR;
proc printto log=junk;
%Trade_Date_Windows_CIZ (freq=&freq, size=&window, minsize=&min, outdsn=_caldates);
proc printto;run;
%put ### DONE!;
%put ### MERGING IN THE RISK FACTORS;
proc sql noprint; create table _vol
as select a.*, b.*, (&retvar-rf) as exret
from &inset a left join ff.factors_&file (keep=&dateff rf &vars) b
on a.&datevar=b.&dateff
order by a.permno, a.&datevar;
select distinct min(&datevar) format date9.,
max(&datevar) format date9. into :mindate,:maxdate
from _vol;
quit;
%put ### DONE!;
/*Save the beginning and ending position of the earliest */
/*and the latest dates in the trading calendar */
data _caldates; set _caldates;
n+1;
if intnx('month',beg_date,0,'e')=
intnx('month',"&mindate"d,0,'e') then call symput ('start',n);
if intnx('month',end_date,0,'e')=
intnx('month',"&maxdate"d,0,'e') then call symput ('finish',n);
run;
/*Main part. Estimate Idiosyncratic Volatility using */
/*rolling time-series regressions. Boundaries for */
/*regressions are based on start and end dates in the */
/*trading calendar */
%put ### ESTIMATING IDIOSYNCRATIC VOLATILITY;
proc printto log=junk;run;
%do j=&start %to &finish %by &step;
data _null_; set _caldates (sortedby=n where=(n=&j));
call symput ('beg',beg_date);call symput ('end',end_date);
run;
data _sample/view=_sample;
do k=1 by 1 until (last.permno);
set _vol; by permno;
where &beg<=date<&end;
if missing(ret) then continue;
mcount=sum(mcount,1);
end;
keep permno mcount;
run;
data _vvol/view=_vvol;
merge _vol (sortedby=date where=(&beg<=date<&end)) _sample;
by permno;
if mcount>=&min;
drop mcount;
run;
proc reg data=_vvol edf noprint outest=_stats;
by permno;
model exret=&vars;
quit;
data _stats; set _stats;
format start_date end_date date9.;
start_date=&beg;end_date=&end;
nused=_p_+_edf_;
label _rmse_ = " "; rename _rmse_=Idrisk_std;
keep permno start_date end_date _rmse_ nused;
run;
/*Pool all estimates of idiosyncratic risk together*/
proc append base=_idvol data=_stats force;run;
%end;
proc printto;run;
/*Merge the incoming dataset with idiosyncratic risk estimates*/
proc sql; create table &outset
as select *
from &inset a left join _idvol b
on a.permno=b.permno and a.date=b.end_date;
/* house cleaning*/
drop table _stats, _vol, _caldates;
drop view _vvol, _sample;
quit;
options &oldoptions errors=&errors;
%put ### DONE!;
%put ### OUTPUT IN THE DATASET &outset;
%MEND;
/* ********************************************************************************* */
/* ************* Material Copyright Wharton Research Data Services *************** */
/* ****************************** All Rights Reserved ****************************** */
/* ********************************************************************************* */