Here is an experiment in which you can compute and graph the small sample mean of the response of hours worked to a technology, when this is computed using a VAR with long-run identifying restrictions, and the data are generated by a specified DSGE model.

Parameter values of the DSGE model are set in paramsnew.m.
par = 1: 2-shock MLE
par = 2: 2-shock CKM
par = 3: 3-shock CKM
par = 6: 3-shock MLE

Recursive or Standard version of the model.
Comment out the line, tau=ones(3,2), if you want to work with the recursive version of the DSGE model in run.m or smallsample.m. If the instruction is left active, then all period t decisions see all period t shocks.

Frequency zero.
If you want to do standard estimation, set var=1 in paramsnew.m. If you want to try out our alternative estimator that uses the zero-frequency spectral density estimator, set var=3.

B(1)
If you'd like the estimation to use the true value of the sum of the VAR coefficients (the infinite lag VAR, from our Proposition 1), then comment out the line that says Bs=[].

To compute the small sample mean of the response of hours worked to a technology shock, when the DSGE model is the two-shock CKM model, and the standard VAR estimator is used, simply execute smallsample.m and sit back a couple of minutes.