Filippo Ferroni, Stefano Grassi, and Miguel Leon-Ledesma, "Selecting Structural Innovations in DSGE models", Journal of Applied Econometrics, Vol. 34, No. 2, 2019, pp. 205-220. All files are ASCII files in DOS format. They are zipped in the file fgl-files.zip. Unix/Linux users should use "unzip -a". DATA The file usmodel_newdata.csv contains the data used in estimation in section 4 of the paper. The data have been downloaded from FRED database of the St. Louis FED. List of series: - Real Gross Domestic Product, 3 Decimal (GDPC96) - GDP Deflator (GDPDEF) - Personal Consumption Expenditures (PCEC) - Fixed Private Investment (FPI) - Civilian Employment (CE16OV) - Effective Federal Funds Rate (FEDFUNDS) - Civilian Noninstitutional Population (CNP16OV) - Nonfarm Business Sector: Average Weekly Hours (PRS85006023) - Average Weekly Hours Of Production And Nonsupervisory Employees: Total private (AWHNONAG) For the estimation of the models, we use the data as in Smets and Wouters (SW), 2007, Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach, American Economic Review, Vol. 97, No. 3, June 2007 and follow their transformations. The dynare mod file of the SW model can be found at https://www.aeaweb.org/articles?id=10.1257/aer.97.3.586 MATLAB CODES section_3_a: contains the codes to replicate the figure 1 of section 3 of the paper. section_3_b: contains the codes to replicate the table 1 of section 3 of the paper and of table 1 and figure 1 of the OnlineAppendix. Note that some of these codes use a modified version of dynare. There are two modifications: 1. In order to decompose the marginal likelihood in bits, make the following change at line 737 of dynare_estimation_1.m [likelihood,DLIK] = feval(objective_function,xparam1,dataset_,options_,M_,estim_params_,bayestopt_,oo_); oo_.MarginalDensity.LaplaceApproximation = .5*estim_params_nbr*log(2*pi) + .5*log_det_invhess - likelihood; oo_.MarginalDensity.LaplaceConstant = .5*estim_params_nbr*log(2*pi); oo_.MarginalDensity.LaplaceHessian = .5*log_det_invhess; oo_.MarginalDensity.LaplaceKernel = - likelihood; oo_.MarginalDensity.LaplacePrior = - DLIK(1); oo_.MarginalDensity.LaplaceLikelihood = - (likelihood - DLIK(1)); 2. To accomodate the random sign switch in the MH algo when using the normal prior on SDT, make the following change in random_walk_metropolis_hastings_core.m Add after lines 164 vSwitch = -1 + 2 * round(rand(1, M_.exo_nbr)); tmp(1:M_.exo_nbr) = vSwitch.*tmp(1:M_.exo_nbr); This change is not necessary if only the posterior mode is computed. Additional Material and Results are reported in the online appendix.