Interpreting fevd estimates of an SVAR (sign restrictions) model

Topics: VAR, SVAR, FAVAR
Sep 29, 2015 at 1:16 PM
Edited Oct 8, 2015 at 8:58 AM
Hi all,

I have a query regarding the interpretation of variance decomposition estimates obtained using the fevd command for SVARS. Consider a 3 variable VAR model with "constraint1" outlining the restriction. To estimate the sign restricted model, I run the code given below.
vpint = VAR({'infl','unemp','ffrate'});
[vpint,dvpint] = estimate(vpint,d,estrange,'order=',4); 
constraint1 = ['S(1,1,1) < 0 ', '&& S(2,1,1) < 0 ', '&& S(3,1,1) < 0 '];
%constraint2 = ['S(1,1,1) < 0 ','&& S(2,1,1) < 0 ', '&& S(3,1,1) < 0 ','&& S(1,2,1) > 0 ', '&& S(2,2,1) < 0 ', '&& S(3,2,1) > 0 ',
'&& S(1,3,1) < 0 ', '&& S(2,3,1) > 0 ', '&& S(3,3,1) > 0 '];
[svpint,svpintdata,B,count] = SVAR(vpint,dvpint, ...
    'method=','householder', ...
    'test=',constraint1, ...
    'ndraw=',500, ...
    'maxIter=',Inf, ...
    'progress=',true);
select = 'S(:,1,1:40)';
[ssvpint,~,index,crit] = sort(svpint,[],select,'progress=',true);
irfmed = srf(ssvpint(1:1),1:40,'select=','res_infl','presample=',true);
[Fa,Fr,VDabs,VDrel]=fevd(ssvpint(1:1),40);
  • While I get the expected IRFs, I am not sure of how one can interpret the FEVD estimates (VDrel). I want to obtain the percentage error variance of each variable explained by the identified shock, but does the first column in "VDrel.(varname)" provide this information? If not, how can I obtain the same?
  • For an SVAR with multiple structural shocks(constraint2), can the FEVD command be used to obtain the contribution of each of the 3 (or n) identified shocks? Are these contributions reported in the respective(n) columns of the FEVD output structure (VDrel) (general case)?
    Thank you.