Hello,
I’m trying to replicate the NK model featuring credit frictions by Curdia and Woodford (2015). I’m referring to the “Complete simplified system of log-linear equations” at page 53 of Technical Appendix. Also, I’m following the calibration “Exogenous, takes resources” and implementing the Taylor rule stated at page 62 (Equation 7.1), with persistence of monetary shocks equal to 0.6.
I’ve double checked the model equations and the auxiliary parameters definitions many times, but when I try to run a stochastic simulation featuring a 1% mon. policy shock, my IRFs are different from the one at page 63 (Figure H.1).
In particular, b (private credit) persistently goes above the SS level instead of going below it, output falls too much, the interest rate goes back to 0 later than when it should according to the Figure H.1.
I’d be extremely grateful if you could give a look at my .mod file and tell me if you spot any errors! Thank you!
CW_15_LINv2.mod (6.4 KB)
UPDATE:
I’ve rewritten the code referring to the “Full system of log-linearised equations” at page 48. This way I get rid of many of the auxiliary parameters definitions that may contain errors.
I’ve also slightly modified the optimal credit condition (3.9) because it was not clear to me what the parameters zeta_Xi and zeta_Chi were. I derived the equation by log-linearising myself the non-linear condition and following the authors’ definition for the hatted variables.
Now the shape of the IRFs are correct, private credit persistently goes below SS! However, the magnitude of b and Y is still incorrect wrt Figure H.1.
Any idea why?
Many thanks!
CW_15_LINv3.mod (6.8 KB)
Are those the only variables where the size does not match? And did you ask the authors for their codes?
Thanks for replying!
I can only compare the 5 variables whose IRFs they include in the Technical Appendix (spread, private credit, output, inflation and policy/deposit rate). Among these, yes, output and private credit are the only ones whose magnitude is different (both Y and b should fall less than they do), while the response of inflation and the spread seems like correct.
No, I didn’t ask the authors for their code.
Another inconsistency: combining the two Steady State Euler equations for borrowers and savers, one should be able to pin down the discount factor BETA for given calibration of steady state policy/deposit rate, spread, probability of changing type DELTA, and probability of borrowers PI_b. Indeed, at page 35 of the Appendix the authors state that they calibrate BETA in such way.
However, plugging into equation (2.16) the calibrated values for DELTA, PI_b, OMEGA and R_d I obtain BETA = 0.5, while in the calibration table at the bottom of the appendix they assume BETA = 0.9874, which is inconsistent with the steady state relationship 2.16… Also, I’ve tried to derive a relationship similar to 2.16 myself analytically combining the 2 Steady State Euler equations to compute BETA, and I obtain the same result for given calibration of other parameters, BETA = 0.5.
My experience is that such inconsistencies can only be resolved with access to the original codes.
Thanks for the advice, I’ll try that and let you know if I sort this out. Thanks again for your help.