I'm getting a different error variance than regress...

Hi All, I'm confused about how the error variance is computed in regress. Why is STATS(4) sometimes larger than var(Y)? Why isn't it always equal to var(Resid)?

clear all;clc;
X=rand(100,3);
Y=rand(100,1);
[B,BINT,R,RINT,STATS] = regress(Y,[ones(size(X,1),1),X]);
varY=var(Y);
varResid=var(Y'-B'*[ones(size(X,1),1),X]');
disp(['var(Y) = ' num2str(varY)]);
disp(['var(Resid) = ' num2str(varResid)]);
disp(['STATS(4) = ' num2str(STATS(4))]);

thanks!