Dynare forums

Hi everybody,

I'm estimating a new-keynesian model with a production sector of monopolistically competitive firms that combine raw materials, capital, and labor to produce output. I also have importers of consumption and investment goods, and importers of raw materials, all of which have market power. At the same time, households offer differentiated labour in a monopolistically competitive labour market, and thus choose the nominal wage. I have paremeters of calvo that indicate the percent of firms/households that can't adjust their prices/wages at each moment t.
When I try to estimate these parameters, I get the following posteriors (see attached file). Epsq and epsw, the parameters of price stickiness associated to output prices and wages, do not have a posterior distribution, but only a value, which coincides with the initial value I give to them in the estimated_params block. I have beta_pdf distributions for these parameters, but when I try to use an uniform between 0 and 1, the estimation fails. (I'm using the Metropolis-Hastings algortithm)

Any ideas on how to procede or what might be deriving these results?

Thank you very much,

Rafael

If your model implementation is correct, this indicates problems with the jumping distribution of the MCMC. Try a different mode_compute. Also try using the option

Code: Select all

mcmc_jumping_covariance=prior_variance

or

Code: Select all

mcmc_jumping_covariance=identity_matrix

which is available in Dynare 4.4.

Hi Prof. Pfeifer,

Thank you very much for your reply. I tried both the prior_variance and the identity_matrix for the MCMC, and neither worked out. I also tried with different mode_computes and the estimation is even worse. I tried using the option mode_check and this is what I get (see attached file) . I am a bit confused with the red points, does anybody know what do they mean? Is this a signal of something wrong in my model or identification problems in the parameters? I would really appreciate if someone could guide me on the next step to tackle these problems.

Thanks again,

Rafael

I need the mod-file.