| 1 |
/* |
| 2 |
* This file replicates the estimation of the cash in advance model (termed M1 |
| 3 |
* in the paper) described in Frank Schorfheide (2000): "Loss function-based |
| 4 |
* evaluation of DSGE models", Journal of Applied Econometrics, 15(6), 645-670. |
| 5 |
* |
| 6 |
* The data are in file "fsdat_simul.m", and have been artificially generated. |
| 7 |
* They are therefore different from the original dataset used by Schorfheide. |
| 8 |
* |
| 9 |
* The prior distribution follows the one originally specified in Schorfheide's |
| 10 |
* paper, except for parameter rho. In the paper, the elicited beta prior for rho |
| 11 |
* implies an asymptote and corresponding prior mode at 0. It is generally |
| 12 |
* recommended to avoid this extreme type of prior. Some optimizers, for instance |
| 13 |
* mode_compute=12 (Mathworks' particleswarm algorithm) may find a posterior mode |
| 14 |
* with rho equal to zero. We lowered the value of the prior standard deviation |
| 15 |
* (changing .223 to .100) to remove the asymptote. |
| 16 |
* |
| 17 |
* The equations are taken from J. Nason and T. Cogley (1994): "Testing the |
| 18 |
* implications of long-run neutrality for monetary business cycle models", |
| 19 |
* Journal of Applied Econometrics, 9, S37-S70. |
| 20 |
* Note that there is an initial minus sign missing in equation (A1), p. S63. |
| 21 |
* |
| 22 |
* This implementation was originally written by Michel Juillard. Please note that the |
| 23 |
* following copyright notice only applies to this Dynare implementation of the |
| 24 |
* model. |
| 25 |
*/ |
| 26 |
|
| 27 |
/* |
| 28 |
* Copyright © 2004-2017 Dynare Team |
| 29 |
* |
| 30 |
* This file is part of Dynare. |
| 31 |
* |
| 32 |
* Dynare is free software: you can redistribute it and/or modify |
| 33 |
* it under the terms of the GNU General Public License as published by |
| 34 |
* the Free Software Foundation, either version 3 of the License, or |
| 35 |
* (at your option) any later version. |
| 36 |
* |
| 37 |
* Dynare is distributed in the hope that it will be useful, |
| 38 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 39 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 40 |
* GNU General Public License for more details. |
| 41 |
* |
| 42 |
* You should have received a copy of the GNU General Public License |
| 43 |
* along with Dynare. If not, see <https://www.gnu.org/licenses/>. |
| 44 |
*/ |
| 45 |
|
| 46 |
var m P c e W R k d n l gy_obs gp_obs y dA; |
| 47 |
varexo e_a e_m; |
| 48 |
|
| 49 |
parameters alp bet gam mst rho psi del; |
| 50 |
|
| 51 |
alp = 0.33; |
| 52 |
bet = 0.99; |
| 53 |
gam = 0.003; |
| 54 |
mst = 1.011; |
| 55 |
rho = 0.7; |
| 56 |
psi = 0.787; |
| 57 |
del = 0.02; |
| 58 |
|
| 59 |
model; |
| 60 |
dA = exp(gam+e_a); |
| 61 |
log(m) = (1-rho)*log(mst) + rho*log(m(-1))+e_m; |
| 62 |
-P/(c(+1)*P(+1)*m)+bet*P(+1)*(alp*exp(-alp*(gam+log(e(+1))))*k^(alp-1)*n(+1)^(1-alp)+(1-del)*exp(-(gam+log(e(+1)))))/(c(+2)*P(+2)*m(+1))=0; |
| 63 |
W = l/n; |
| 64 |
-(psi/(1-psi))*(c*P/(1-n))+l/n = 0; |
| 65 |
R = P*(1-alp)*exp(-alp*(gam+e_a))*k(-1)^alp*n^(-alp)/W; |
| 66 |
1/(c*P)-bet*P*(1-alp)*exp(-alp*(gam+e_a))*k(-1)^alp*n^(1-alp)/(m*l*c(+1)*P(+1)) = 0; |
| 67 |
c+k = exp(-alp*(gam+e_a))*k(-1)^alp*n^(1-alp)+(1-del)*exp(-(gam+e_a))*k(-1); |
| 68 |
P*c = m; |
| 69 |
m-1+d = l; |
| 70 |
e = exp(e_a); |
| 71 |
y = k(-1)^alp*n^(1-alp)*exp(-alp*(gam+e_a)); |
| 72 |
gy_obs = dA*y/y(-1); |
| 73 |
gp_obs = (P/P(-1))*m(-1)/dA; |
| 74 |
end; |
| 75 |
|
| 76 |
shocks; |
| 77 |
var e_a; stderr 0.014; |
| 78 |
var e_m; stderr 0.005; |
| 79 |
end; |
| 80 |
|
| 81 |
steady_state_model; |
| 82 |
dA = exp(gam); |
| 83 |
gst = 1/dA; |
| 84 |
m = mst; |
| 85 |
khst = ( (1-gst*bet*(1-del)) / (alp*gst^alp*bet) )^(1/(alp-1)); |
| 86 |
xist = ( ((khst*gst)^alp - (1-gst*(1-del))*khst)/mst )^(-1); |
| 87 |
nust = psi*mst^2/( (1-alp)*(1-psi)*bet*gst^alp*khst^alp ); |
| 88 |
n = xist/(nust+xist); |
| 89 |
P = xist + nust; |
| 90 |
k = khst*n; |
| 91 |
|
| 92 |
l = psi*mst*n/( (1-psi)*(1-n) ); |
| 93 |
c = mst/P; |
| 94 |
d = l - mst + 1; |
| 95 |
y = k^alp*n^(1-alp)*gst^alp; |
| 96 |
R = mst/bet; |
| 97 |
W = l/n; |
| 98 |
ist = y-c; |
| 99 |
q = 1 - d; |
| 100 |
|
| 101 |
e = 1; |
| 102 |
|
| 103 |
gp_obs = m/dA; |
| 104 |
gy_obs = dA; |
| 105 |
end; |
| 106 |
|
| 107 |
steady; |
| 108 |
|
| 109 |
check; |
| 110 |
|
| 111 |
estimated_params; |
| 112 |
alp, beta_pdf, 0.356, 0.02; |
| 113 |
bet, beta_pdf, 0.993, 0.002; |
| 114 |
gam, normal_pdf, 0.0085, 0.003; |
| 115 |
mst, normal_pdf, 1.0002, 0.007; |
| 116 |
rho, beta_pdf, 0.129, 0.100; |
| 117 |
psi, beta_pdf, 0.65, 0.05; |
| 118 |
del, beta_pdf, 0.01, 0.005; |
| 119 |
stderr e_a, inv_gamma_pdf, 0.035449, inf; |
| 120 |
stderr e_m, inv_gamma_pdf, 0.008862, inf; |
| 121 |
end; |
| 122 |
|
| 123 |
varobs gp_obs gy_obs; |
| 124 |
|
| 125 |
estimation(order=1,mode_compute=5,silent_optimizer, datafile='../fs2000/fsdat_simul.m', nobs=192, loglinear, mh_replic=20, mh_nblocks=1, mh_jscale=0.8,moments_varendo, |
| 126 |
conditional_variance_decomposition=[2,2000],consider_all_endogenous,sub_draws=2); |
| 127 |
|
| 128 |
stoch_simul(order=1,conditional_variance_decomposition=[2,2000],noprint,nograph); |
| 129 |
par=load([M_.fname filesep 'metropolis' filesep M_.fname '_posterior_draws1']); |
| 130 |
|
| 131 |
for par_iter=1:size(par.pdraws,1) |
| 132 |
M_=set_parameters_locally(M_,par.pdraws{par_iter,1}); |
| 133 |
[info, oo_, options_, M_]=stoch_simul(M_, options_, oo_, var_list_); |
| 134 |
correlation(:,:,par_iter)=cell2mat(oo_.autocorr); |
| 135 |
covariance(:,:,par_iter)=oo_.var; |
| 136 |
conditional_variance_decomposition(:,:,:,par_iter)=oo_.conditional_variance_decomposition; |
| 137 |
variance_decomposition(:,:,par_iter)=oo_.variance_decomposition; |
| 138 |
end |
| 139 |
|
| 140 |
correlation=mean(correlation,3); |
| 141 |
nvars=M_.orig_endo_nbr; |
| 142 |
for var_iter_1=1:nvars |
| 143 |
for var_iter_2=1:nvars |
| 144 |
if max(abs(correlation(var_iter_1,var_iter_2:nvars:end)'-oo_.PosteriorTheoreticalMoments.dsge.correlation.Mean.(M_.endo_names{var_iter_1}).(M_.endo_names{var_iter_2})))>1e-8 |
| 145 |
error('Correlations do not match') |
| 146 |
end |
| 147 |
end |
| 148 |
end |
| 149 |
|
| 150 |
covariance=mean(covariance,3); |
| 151 |
nvars=M_.orig_endo_nbr; |
| 152 |
for var_iter_1=1:nvars |
| 153 |
for var_iter_2=var_iter_1:nvars |
| 154 |
if max(abs(covariance(var_iter_1,var_iter_2)-oo_.PosteriorTheoreticalMoments.dsge.covariance.Mean.(M_.endo_names{var_iter_1}).(M_.endo_names{var_iter_2})))>1e-8 |
| 155 |
error('Covariances do not match') |
| 156 |
end |
| 157 |
end |
| 158 |
end |
| 159 |
|
| 160 |
variance_decomposition=mean(variance_decomposition,3); |
| 161 |
nvars=M_.orig_endo_nbr; |
| 162 |
for var_iter_1=1:nvars |
| 163 |
for shock_iter=1:M_.exo_nbr |
| 164 |
if max(abs(variance_decomposition(var_iter_1,shock_iter)/100-oo_.PosteriorTheoreticalMoments.dsge.VarianceDecomposition.Mean.(M_.endo_names{var_iter_1}).(M_.exo_names{shock_iter})))>1e-8 |
| 165 |
error('Variance decomposition does not match') |
| 166 |
end |
| 167 |
end |
| 168 |
end |
| 169 |
|
| 170 |
conditional_variance_decomposition=mean(conditional_variance_decomposition,4); |
| 171 |
nvars=M_.orig_endo_nbr; |
| 172 |
horizon_size=size(conditional_variance_decomposition,2); |
| 173 |
for var_iter_1=1:nvars |
| 174 |
for shock_iter=1:M_.exo_nbr |
| 175 |
for horizon_iter=1:horizon_size |
| 176 |
if max(abs(conditional_variance_decomposition(var_iter_1,horizon_iter,shock_iter)-oo_.PosteriorTheoreticalMoments.dsge.ConditionalVarianceDecomposition.Mean.(M_.endo_names{var_iter_1}).(M_.exo_names{shock_iter})(horizon_iter)))>1e-8 |
| 177 |
error('Conditional Variance decomposition does not match') |
| 178 |
end |
| 179 |
end |
| 180 |
end |
| 181 |
end |
| 182 |
|
| 183 |
// case with measurement error |
| 184 |
estimated_params(overwrite); |
| 185 |
alp, beta_pdf, 0.356, 0.02; |
| 186 |
bet, beta_pdf, 0.993, 0.002; |
| 187 |
gam, normal_pdf, 0.0085, 0.003; |
| 188 |
mst, normal_pdf, 1.0002, 0.007; |
| 189 |
rho, beta_pdf, 0.129, 0.100; |
| 190 |
psi, beta_pdf, 0.65, 0.05; |
| 191 |
del, beta_pdf, 0.01, 0.005; |
| 192 |
stderr e_a, inv_gamma_pdf, 0.035449, inf; |
| 193 |
stderr e_m, inv_gamma_pdf, 0.008862, inf; |
| 194 |
stderr gp_obs, inv_gamma_pdf, 0.003, inf; |
| 195 |
end; |
| 196 |
|
| 197 |
estimation(order=1,mode_compute=5,silent_optimizer, datafile='../fs2000/fsdat_simul.m', nobs=192, loglinear, mh_replic=20, mh_nblocks=1, mh_jscale=0.8,moments_varendo, |
| 198 |
conditional_variance_decomposition=[2,2000],consider_all_endogenous,sub_draws=2); |
| 199 |
|
| 200 |
stoch_simul(order=1,conditional_variance_decomposition=[2,2000],noprint,nograph); |
| 201 |
par=load([M_.fname filesep 'metropolis' filesep M_.fname '_posterior_draws1']); |
| 202 |
|
| 203 |
for par_iter=1:size(par.pdraws,1) |
| 204 |
M_=set_parameters_locally(M_,par.pdraws{par_iter,1}); |
| 205 |
[info, oo_, options_, M_]=stoch_simul(M_, options_, oo_, var_list_); |
| 206 |
correlation(:,:,par_iter)=cell2mat(oo_.autocorr); |
| 207 |
covariance(:,:,par_iter)=oo_.var; |
| 208 |
conditional_variance_decomposition(:,:,:,par_iter)=oo_.conditional_variance_decomposition; |
| 209 |
conditional_variance_decomposition_ME(:,:,:,par_iter)=oo_.conditional_variance_decomposition_ME; |
| 210 |
variance_decomposition(:,:,par_iter)=oo_.variance_decomposition; |
| 211 |
variance_decomposition_ME(:,:,par_iter)=oo_.variance_decomposition_ME; |
| 212 |
[~,obs_order]=sort(options_.varobs_id); |
| 213 |
end |
| 214 |
|
| 215 |
correlation=mean(correlation,3); |
| 216 |
nvars=M_.orig_endo_nbr; |
| 217 |
for var_iter_1=1:nvars |
| 218 |
for var_iter_2=1:nvars |
| 219 |
if max(abs(correlation(var_iter_1,var_iter_2:nvars:end)'-oo_.PosteriorTheoreticalMoments.dsge.correlation.Mean.(M_.endo_names{var_iter_1}).(M_.endo_names{var_iter_2})))>1e-8 |
| 220 |
error('Correlations do not match') |
| 221 |
end |
| 222 |
end |
| 223 |
end |
| 224 |
|
| 225 |
covariance=mean(covariance,3); |
| 226 |
nvars=M_.orig_endo_nbr; |
| 227 |
for var_iter_1=1:nvars |
| 228 |
for var_iter_2=var_iter_1:nvars |
| 229 |
if max(abs(covariance(var_iter_1,var_iter_2)-oo_.PosteriorTheoreticalMoments.dsge.covariance.Mean.(M_.endo_names{var_iter_1}).(M_.endo_names{var_iter_2})))>1e-8 |
| 230 |
error('Covariances do not match') |
| 231 |
end |
| 232 |
end |
| 233 |
end |
| 234 |
|
| 235 |
variance_decomposition=mean(variance_decomposition,3); |
| 236 |
nvars=M_.orig_endo_nbr; |
| 237 |
for var_iter_1=1:nvars |
| 238 |
for shock_iter=1:M_.exo_nbr |
| 239 |
if max(abs(variance_decomposition(var_iter_1,shock_iter)/100-oo_.PosteriorTheoreticalMoments.dsge.VarianceDecomposition.Mean.(M_.endo_names{var_iter_1}).(M_.exo_names{shock_iter})))>1e-8 |
| 240 |
error('Variance decomposition does not match') |
| 241 |
end |
| 242 |
end |
| 243 |
end |
| 244 |
|
| 245 |
if isoctave |
| 246 |
tol = 2.2e-3; |
| 247 |
else |
| 248 |
tol = 1e-8; |
| 249 |
end |
| 250 |
|
| 251 |
variance_decomposition_ME=mean(variance_decomposition_ME,3); |
| 252 |
nvars=length(options_.varobs); |
| 253 |
for var_iter_1=1:nvars |
| 254 |
for shock_iter=1:M_.exo_nbr |
| 255 |
if max(abs(variance_decomposition_ME(obs_order(var_iter_1),shock_iter)/100-oo_.PosteriorTheoreticalMoments.dsge.VarianceDecompositionME.Mean.(options_.varobs{var_iter_1}).(M_.exo_names{shock_iter}))) > tol |
| 256 |
error('Variance decomposition does not match') |
| 257 |
end |
| 258 |
end |
| 259 |
end |
| 260 |
|
| 261 |
conditional_variance_decomposition=mean(conditional_variance_decomposition,4); |
| 262 |
nvars=M_.orig_endo_nbr; |
| 263 |
horizon_size=size(conditional_variance_decomposition,2); |
| 264 |
for var_iter_1=1:nvars |
| 265 |
for shock_iter=1:M_.exo_nbr |
| 266 |
for horizon_iter=1:horizon_size |
| 267 |
if max(abs(conditional_variance_decomposition(var_iter_1,horizon_iter,shock_iter)-oo_.PosteriorTheoreticalMoments.dsge.ConditionalVarianceDecomposition.Mean.(M_.endo_names{var_iter_1}).(M_.exo_names{shock_iter})(horizon_iter))) > tol |
| 268 |
error('Conditional Variance decomposition does not match') |
| 269 |
end |
| 270 |
end |
| 271 |
end |
| 272 |
end |
| 273 |
|
| 274 |
conditional_variance_decomposition_ME=mean(conditional_variance_decomposition_ME,4); |
| 275 |
exo_names=[M_.exo_names;'ME']; |
| 276 |
nvars=length(options_.varobs); |
| 277 |
horizon_size=size(conditional_variance_decomposition_ME,2); |
| 278 |
for var_iter_1=1:nvars |
| 279 |
for shock_iter=1:M_.exo_nbr+1 |
| 280 |
for horizon_iter=1:horizon_size |
| 281 |
if max(abs(conditional_variance_decomposition_ME(obs_order(var_iter_1),horizon_iter,shock_iter)-oo_.PosteriorTheoreticalMoments.dsge.ConditionalVarianceDecompositionME.Mean.(options_.varobs{var_iter_1}).(exo_names{shock_iter})(horizon_iter))) > tol |
| 282 |
error('Conditional Variance decomposition does not match') |
| 283 |
end |
| 284 |
end |
| 285 |
end |
| 286 |
end |
| 287 |
|
| 288 |
|
| 289 |
options_.prior_mc=100; |
| 290 |
options_.qz_criterium = 1+1e-6; |
| 291 |
oo_ = compute_moments_varendo('prior',options_,M_,oo_,estim_params_,M_.endo_names(1:M_.orig_endo_nbr)); |
| 292 |
|
| 293 |
|
| 294 |
/* |
| 295 |
* The following lines were used to generate the data file. If you want to |
| 296 |
* generate another random data file, comment the "estimation" line and uncomment |
| 297 |
* the following lines. |
| 298 |
*/ |
| 299 |
|
| 300 |
//stoch_simul(periods=200, order=1); |
| 301 |
//datatomfile('fsdat_simul', char('gy_obs', 'gp_obs')); |