% DSGE model based on replication files of % Andreasen, Fernandez-Villaverde, Rubio-Ramirez (2018), The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications, Review of Economic Studies, 85, p. 1-49 % Original code by Martin M. Andreasen, Jan 2016 % Adapted for Dynare by Willi Mutschler (@wmutschl, willi@mutschler.eu), Jan 2021 % ========================================================================= % Copyright © 2021 Dynare Team % % This file is part of Dynare. % % Dynare is free software: you can redistribute it and/or modify % it under the terms of the GNU General Public License as published by % the Free Software Foundation, either version 3 of the License, or % (at your option) any later version. % % Dynare is distributed in the hope that it will be useful, % but WITHOUT ANY WARRANTY; without even the implied warranty of % MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the % GNU General Public License for more details. % % You should have received a copy of the GNU General Public License % along with Dynare. If not, see . % ========================================================================= %-------------------------------------------------------------------------- % Variable declaration %-------------------------------------------------------------------------- var ln_k ln_s ln_a ln_g ln_d ln_c ln_r ln_pai ln_h ln_q ln_evf ln_iv ln_x2 ln_la ln_goy ln_Esdf xhr20 xhr40 Exhr @#for i in 1:40 ln_p@{i} @#endfor Obs_Gr_C Obs_Gr_I Obs_Infl Obs_r1 Obs_r40 Obs_xhr40 Obs_GoY Obs_hours ; predetermined_variables ln_k ln_s; varobs Obs_Gr_C Obs_Gr_I Obs_Infl Obs_r1 Obs_r40 Obs_xhr40 Obs_GoY Obs_hours; %-------------------------------------------------------------------------- % Exogenous shocks %-------------------------------------------------------------------------- varexo eps_a eps_d eps_g ; %-------------------------------------------------------------------------- % Parameter declaration %-------------------------------------------------------------------------- parameters BETTA B INHABIT H PHI1 PHI2 RRA PHI4 KAPAone KAPAtwo DELTA THETA ETA ALFA CHI RHOR BETTAPAI BETTAY MYYPS MYZ RHOA %STDA RHOG %STDG RHOD %STDD CONSxhr40 BETTAxhr BETTAxhr40 CONSxhr20 PAI GAMA GoY %auxiliary PHIzero AA PHI3 negVf ; %-------------------------------------------------------------------------- % Model equations %-------------------------------------------------------------------------- % Based on DSGE_model_NegVf_yieldCurve.m and DSGE_model_PosVf_yieldCurve.m % Note that we include an auxiliary parameter negVf to distinguish whether % the steady state value function is positive (negVf=0) or negative (negVf=1). % This parameter is endogenously determined in the steady_state_model block. model; %-------------------------------------------------------------------------- % Auxiliary expressions %-------------------------------------------------------------------------- % do exp transform such that variables are logged variables @#for var in [ "k", "s", "c", "r", "a", "g", "d", "pai", "h", "q", "evf", "iv", "x2", "la", "goy", "Esdf" ] #@{var}_ba1 = exp(ln_@{var}(-1)); #@{var}_cu = exp(ln_@{var}); #@{var}_cup = exp(ln_@{var}(+1)); @#endfor @#for i in 1:40 #p@{i}_cu = exp(ln_p@{i}); #p@{i}_cup = exp(ln_p@{i}(+1)); @#endfor % these variables are not transformed #xhr20_cu = xhr20; #xhr20_cup = xhr20(+1); #xhr40_cu = xhr40; #xhr40_cup = xhr40(+1); #Exhr_cu = Exhr; #Exhr_cup = Exhr(+1); % auxiliary steady state variables #K = exp(steady_state(ln_k)); #IV = exp(steady_state(ln_iv)); #C = exp(steady_state(ln_c)); #Y = (C + IV)/(1-GoY); #R = exp(steady_state(ln_r)); #G = Y-C-IV; #removeMeanXhr = 1; % The atemporal relations if possible % No stochastic trend in investment specific shocks #myyps_cu = MYYPS; #myyps_cup = MYYPS; % No stochastic trend in non-stationary technology shocks #myz_cu = MYZ; #myz_cup = MYZ; % Defining myzstar #MYZSTAR = MYYPS^(THETA/(1-THETA))*MYZ; #myzstar_cu = myyps_cu ^(THETA/(1-THETA))*myz_cu; #myzstar_cup= myyps_cup^(THETA/(1-THETA))*myz_cup; % The expression for the value function (only valid for deterministic trends!) % Note that we make use of auxiliary parameter negVf to switch signs #mvf_cup = -negVf*(d_cup/(1-PHI2)*((c_cup-B*c_cu*MYZSTAR^-1)^(1-PHI2)-1) + d_cup*PHIzero/(1-PHI1)*(1-h_cup)^(1-PHI1) - negVf* BETTA*MYZSTAR^((1-PHI4)*(1-PHI2))*AA*evf_cup^(1/(1-PHI3))); % The growth rate in lambda #myla_cup = (la_cup/la_cu)*(AA*evf_cu^(1/(1-PHI3))/mvf_cup)^PHI3*myzstar_cup^(-PHI2*(1-PHI4)-PHI4); % The relation between the optimal price for the firms and the pris and inflation %ptil_cu = ((1-ALFA*(pai_ba1^CHI/pai_cu )^(1-ETA))/(1-ALFA))^(1/(1-ETA)); %ptil_cup = ((1-ALFA*(pai_cu ^CHI/pai_cup)^(1-ETA))/(1-ALFA))^(1/(1-ETA)); #ptil_cu = ((1-ALFA*(1/pai_cu )^(1-ETA))/(1-ALFA))^(1/(1-ETA)); #ptil_cup = ((1-ALFA*(1/pai_cup)^(1-ETA))/(1-ALFA))^(1/(1-ETA)); % From the households' FOC for labor #w_cu = d_cu*PHIzero*(1-h_cu )^(-PHI1)/la_cu; #w_cup = d_cu*PHIzero*(1-h_cup)^(-PHI1)/la_cup; % Shouldn't w_cup include d_cup? Let's stick to the original (wrong) code in the replication files as results don't change dramatically... [@wmutschl] % The firms' FOC for labor #mc_cu = w_cu /((1-THETA)*a_cu *myyps_cu ^(-THETA/(1-THETA))*myz_cu ^-THETA *k_cu ^THETA*h_cu ^(-THETA)); #mc_cup = w_cup/((1-THETA)*a_cup*myyps_cup^(-THETA/(1-THETA))*myz_cup^-THETA *k_cup^THETA*h_cup^(-THETA)); % The firms' FOC for capital #rk_cu = mc_cu *THETA* a_cu *myyps_cu *myz_cu ^(1-THETA)*k_cu ^(THETA-1)*h_cu ^(1-THETA); #rk_cup = mc_cup*THETA* a_cup*myyps_cup*myz_cup^(1-THETA)*k_cup^(THETA-1)*h_cup^(1-THETA); % The income identity #y_cu = c_cu + iv_cu + g_cu; %-------------------------------------------------------------------------- % Actual model equations %-------------------------------------------------------------------------- [name='Expected value of the value function'] 0 = -evf_cu + (mvf_cup/AA)^(1-PHI3); [name='Households FOC for capital'] 0 = -q_cu+BETTA*myla_cup/myyps_cup*(rk_cup+q_cup*(1-DELTA) -q_cup*KAPAtwo/2*(iv_cup/k_cup*myyps_cup*myzstar_cup - IV/K*MYYPS*MYZSTAR)^2 +q_cup*KAPAtwo*(iv_cup/k_cup*myyps_cup*myzstar_cup - IV/K*MYYPS*MYZSTAR)*iv_cup/k_cup*myyps_cup*myzstar_cup); [name='Households FOC for investments'] 0 = -1+q_cu*(1-KAPAone/2*(iv_cu/IV-1)^2-iv_cu/IV*KAPAone*(iv_cu/IV-1)-KAPAtwo*(iv_cu/k_cu*myyps_cu*myzstar_cu - IV/K*MYYPS*MYZSTAR)); [name='Euler equation for consumption'] 0 = -1+BETTA*r_cu*exp(CONSxhr40*xhr40_cu + CONSxhr20*xhr20_cu)*myla_cup/pai_cup; [name='Households FOC for consumption'] 0 = -la_cu + d_cu*(c_cu -B*c_ba1*myzstar_cu^-1)^(-PHI2) -INHABIT*B*BETTA*d_cup*(AA*evf_cu^(1/(1-PHI3))/mvf_cup)^PHI3*(c_cup -B*c_cu*myzstar_cup^-1)^(-PHI2)*myzstar_cup^(-PHI2*(1-PHI4)-PHI4); [name='Nonlinear pricing, relation for x1 = (ETA-1)/ETA*x2'] 0= -(ETA-1)/ETA*x2_cu+y_cu*mc_cu*ptil_cu^(-ETA-1) +ALFA*BETTA*myla_cup*(ptil_cu/ptil_cup)^(-ETA-1)*(1/pai_cup)^(-ETA)*(ETA-1)/ETA*x2_cup*myzstar_cup; [name='Nonlinear pricing, relation for x2'] 0=-x2_cu+y_cu*ptil_cu^-ETA +ALFA*BETTA*myla_cup*(ptil_cu/ptil_cup)^(-ETA)*(1/pai_cup)^(1-ETA)*x2_cup*myzstar_cup; [name='Nonlinear pricing, relation for s'] 0= -s_cup+(1-ALFA)*ptil_cu^(-ETA)+ALFA*(pai_cu/1)^ETA*s_cu; [name='Interest rate rule'] 0 = -log(r_cu/R)+RHOR*log(r_ba1/R)+(1-RHOR)*(BETTAPAI*log(pai_cu/PAI)+BETTAY*log(y_cu/Y) + BETTAxhr*(BETTAxhr40*xhr40_cu - removeMeanXhr*Exhr_cu)); [name='Production function'] 0 = -y_cu*s_cup + a_cu *(k_cu *myyps_cu ^(-1/(1-THETA))*myz_cu ^-1)^THETA*h_cu ^(1-THETA); [name='Relation for physical capital stock'] 0= -k_cup + (1-DELTA)*k_cu*(myyps_cu*myzstar_cu)^-1 + iv_cu - iv_cu*KAPAone/2*(iv_cu/IV-1)^2 - k_cu*(myyps_cu*myzstar_cu)^-1*KAPAtwo/2*(iv_cu/k_cu*myyps_cu*myzstar_cu - IV/K*MYYPS*MYZSTAR)^2; [name='Goverment spending over output'] 0=-goy_cu + g_cu/y_cu; [name='The yield curve: p1'] 0= -p1_cu + 1/r_cu; @#for i in 2:40 [name='The yield curve: p@{i}'] 0= -p@{i}_cu + BETTA*myla_cup/pai_cup*p@{i-1}_cup; @#endfor [name='Stochastic discount factor'] 0= -Esdf_cu+ BETTA*myla_cup/pai_cup; [name='Expected 5 year excess holding period return'] 0= -xhr20_cu+ log(p19_cup) - log(p20_cu) - log(r_cu); [name='Expected 10 year excess holding period return'] 0= -xhr40_cu+ log(p39_cup) - log(p40_cu) - log(r_cu); [name='Mean of expected excess holding period return in Taylor rule'] 0= -Exhr_cu + (1-GAMA)*(BETTAxhr40*xhr40_cu) + GAMA*Exhr_cup; [name='Exogenous process for productivity'] 0 = -log(a_cu)+RHOA*log(a_ba1) + eps_a; [name='Exogenous process for government spending'] 0 = -log(g_cu/G)+RHOG*log(g_ba1/G) + eps_g; [name='Exogenous process for discount factor shifter'] 0 = -log(d_cu)+RHOD*log(d_ba1) + eps_d; [name='Observable annualized consumption growth'] Obs_Gr_C = 4*( ln_c -ln_c(-1) + log(MYZSTAR)); [name='Observable annualized investment growth'] Obs_Gr_I = 4*( ln_iv - ln_iv(-1) + log(MYZSTAR)+log(MYYPS)); [name='Observable annualized inflation'] Obs_Infl = 4*( ln_pai); [name='Observable annualized one-quarter nominal yield'] Obs_r1 = 4*( ln_r); [name='Observable annualized 10-year nominal yield'] Obs_r40 = 4*( -1/40*ln_p40); [name='Observable annualized 10-year ex post excess holding period return'] Obs_xhr40 = 4*( ln_p39 - ln_p40(-1) - ln_r(-1) ); [name='Observable annualized log ratio of government spending to GDP'] Obs_GoY = 4*( 1/4*ln_goy); [name='Observable annualized log of hours'] Obs_hours = 4*( 1/100*ln_h); end; %-------------------------------------------------------------------------- % Steady State Computations %-------------------------------------------------------------------------- % Based on DSGE_model_yieldCurve_ss.m, getPHI3.m, ObjectGMM.m % Note that we include an auxiliary parameter negVf to distinguish whether % the steady state value function is positive (negVf=0) or negative (negVf=1). % This parameter is endogenously determined in the steady_state_model block. steady_state_model; % The growth rate in the firms' fixed costs MYZSTARBAR = MYYPS^(THETA/(1-THETA))*MYZ; % The growth rate for lampda MYLABAR = MYZSTARBAR^(-PHI2*(1-PHI4)-PHI4); % The relative optimal price for firms PTILBAR = ((1-ALFA*PAI^((CHI-1)*(1-ETA)))/(1-ALFA))^(1/(1-ETA)); % The state variable s for distortions between output and produktion SBAR = ((1-ALFA)*PTILBAR^(-ETA))/(1-ALFA*PAI^((1-CHI)*ETA)); % The 1-period interest rate RBAR = PAI/(BETTA*MYLABAR); % The market price of capital QBAR = 1; % The real price of renting capital RKBAR = QBAR*(MYYPS/(BETTA*MYLABAR)-(1-DELTA)); % The marginal costs in the firms MCBAR = (1-ALFA*BETTA*MYLABAR*PAI^((1-CHI)*ETA)*MYZSTARBAR)*(ETA-1)/ETA*PTILBAR/(1-ALFA*BETTA*MYLABAR*PAI^((CHI-1)*(1-ETA))*MYZSTARBAR); % The capital stock KBAR = H*(RKBAR/(MCBAR*THETA*MYYPS*MYZ^(1-THETA)))^(1/(THETA-1)); % The wage level WBAR = MCBAR*(1-THETA)*MYYPS^(-THETA/(1-THETA))*MYZ^-THETA*(KBAR/H)^THETA; % The level of investment IVBAR = KBAR - (1-DELTA)*KBAR*MYYPS^(-1/(1-THETA))*MYZ^-1; % The consumption level CBAR = ((1-GoY)*(KBAR*MYYPS^(-1/(1-THETA))*MYZ^-1)^THETA*H^(1-THETA))/SBAR-IVBAR; % The output level YBAR = (CBAR + IVBAR)/(1-GoY); % The value of lambda LABAR = (CBAR-B*CBAR*MYZSTARBAR^-1)^-PHI2 - INHABIT*B*BETTA*(CBAR-B*CBAR*MYZSTARBAR^-1)^-PHI2*MYZSTARBAR^(-PHI2*(1-PHI4)-PHI4); % The value of PHIzero PHIzero = LABAR*WBAR*(1-H)^PHI1; % The level of the value function VFBAR = 1/(1-BETTA*MYZSTARBAR^((1-PHI4)*(1-PHI2)))*(1/(1-PHI2)*((CBAR-B*CBAR*MYZSTARBAR^-1)^(1-PHI2)-1)+PHIzero/(1-PHI1)*(1-H)^(1-PHI1)); UBAR = 1/(1-PHI2)*((CBAR-B*CBAR*MYZSTARBAR^-1)^(1-PHI2)-1)+PHIzero/(1-PHI1)*(1-H)^(1-PHI1); [AA, EVFBAR, PHI3, negVf, info]= AFVRR_steady_helper(VFBAR,RBAR,IVBAR,CBAR,KBAR,LABAR,QBAR,YBAR, BETTA,B,PAI,H,PHIzero,PHI1,PHI2,THETA,MYYPS,MYZ,INHABIT,RRA,CONSxhr40); % The value of X2 X2BAR = YBAR*PTILBAR^(-ETA)/(1-BETTA*ALFA*MYLABAR*PAI^((CHI-1)*(1-ETA))*MYZSTARBAR); % Government spending GBAR = GoY*YBAR; %************************************************************************** % map into model variables ln_k = log(KBAR); ln_s = log(SBAR); ln_c_ba1 = log(CBAR); ln_r_ba1 = log(RBAR); ln_a = log(1); ln_g = log(GBAR); ln_d = log(1); ln_c = log(CBAR); ln_r = log(RBAR); ln_pai = log(PAI); ln_h = log(H); ln_q = log(QBAR); ln_evf = log(EVFBAR); ln_iv = log(IVBAR); ln_x2 = log(X2BAR); ln_la = log(LABAR); ln_goy = log(GoY); ln_Esdf = log(1/RBAR); xhr20 = 0; xhr40 = 0; Exhr = 0; % The yield curve ln_p1 = log((1/RBAR)^1); ln_p2 = log((1/RBAR)^2); ln_p3 = log((1/RBAR)^3); ln_p4 = log((1/RBAR)^4); ln_p5 = log((1/RBAR)^5); ln_p6 = log((1/RBAR)^6); ln_p7 = log((1/RBAR)^7); ln_p8 = log((1/RBAR)^8); ln_p9 = log((1/RBAR)^9); ln_p10 = log((1/RBAR)^10); ln_p11 = log((1/RBAR)^11); ln_p12 = log((1/RBAR)^12); ln_p13 = log((1/RBAR)^13); ln_p14 = log((1/RBAR)^14); ln_p15 = log((1/RBAR)^15); ln_p16 = log((1/RBAR)^16); ln_p17 = log((1/RBAR)^17); ln_p18 = log((1/RBAR)^18); ln_p19 = log((1/RBAR)^19); ln_p20 = log((1/RBAR)^20); ln_p21 = log((1/RBAR)^21); ln_p22 = log((1/RBAR)^22); ln_p23 = log((1/RBAR)^23); ln_p24 = log((1/RBAR)^24); ln_p25 = log((1/RBAR)^25); ln_p26 = log((1/RBAR)^26); ln_p27 = log((1/RBAR)^27); ln_p28 = log((1/RBAR)^28); ln_p29 = log((1/RBAR)^29); ln_p30 = log((1/RBAR)^30); ln_p31 = log((1/RBAR)^31); ln_p32 = log((1/RBAR)^32); ln_p33 = log((1/RBAR)^33); ln_p34 = log((1/RBAR)^34); ln_p35 = log((1/RBAR)^35); ln_p36 = log((1/RBAR)^36); ln_p37 = log((1/RBAR)^37); ln_p38 = log((1/RBAR)^38); ln_p39 = log((1/RBAR)^39); ln_p40 = log((1/RBAR)^40); Obs_Gr_C = 4*( log(MYZSTARBAR) ); Obs_Gr_I = 4*( log(MYZSTARBAR)+log(MYYPS) ); Obs_Infl = 4*( ln_pai ); Obs_r1 = 4*( ln_r ); Obs_r40 = 4*( -1/40*ln_p40 ); Obs_xhr40 = 4*( xhr40 ); Obs_GoY = 4*( 1/4*ln_goy ); Obs_hours = 4*( 1/100*ln_h ); end; %-------------------------------------------------------------------------- % Declare moments to use in estimation %-------------------------------------------------------------------------- % These are the moments used in the paper; corresponds to momentSet=2 in the replication files matched_moments; %first moments: all Obs_Gr_C; Obs_Gr_I; Obs_Infl; Obs_r1; Obs_r40; Obs_xhr40; Obs_GoY; Obs_hours; %second moments % (i) all variances, (2) all covariances excluding GoY and hours, (3) own first autocovariances Obs_Gr_C*Obs_Gr_C; Obs_Gr_C*Obs_Gr_I; Obs_Gr_C*Obs_Infl; Obs_Gr_C*Obs_r1; Obs_Gr_C*Obs_r40; Obs_Gr_C*Obs_xhr40; Obs_Gr_I*Obs_Gr_I; Obs_Gr_I*Obs_Infl; Obs_Gr_I*Obs_r1; Obs_Gr_I*Obs_r40; Obs_Gr_I*Obs_xhr40; Obs_Infl*Obs_Infl; Obs_Infl*Obs_r1; Obs_Infl*Obs_r40; Obs_Infl*Obs_xhr40; Obs_r1*Obs_r1; Obs_r1*Obs_r40; Obs_r1*Obs_xhr40; Obs_r40*Obs_r40; Obs_r40*Obs_xhr40; Obs_xhr40*Obs_xhr40; Obs_GoY*Obs_GoY; Obs_hours*Obs_hours; Obs_Gr_C*Obs_Gr_C(-1); Obs_Gr_I*Obs_Gr_I(-1); Obs_Infl*Obs_Infl(-1); Obs_r1*Obs_r1(-1); Obs_r40*Obs_r40(-1); Obs_xhr40*Obs_xhr40(-1); Obs_GoY*Obs_GoY(-1); Obs_hours*Obs_hours(-1); end; %-------------------------------------------------------------------------- % Create Data %-------------------------------------------------------------------------- @#ifdef CreateData verbatim; % From 1961Q3 to 2007Q4 DataUS = xlsread('Data_PruningPaper_v5.xlsx','Data_used','E3:M188'); % ANNUALIZED (except for hours and GoY) % 1 2 3 4 5 6 7 8 9 % Lables: Date Gr_C Gr_I GoY hours Infl_C r1 r40 xhr40 %label_data = {'Gr_C ', 'Gr_I ','Infl ', 'r1 ', 'r40 ', 'xhr40 ','GoY ', 'hours '}; %DataUS = [DataUS(:,2:3) DataUS(:,6:8) DataUS(:,9) log(DataUS(:,4)) 4*log(DataUS(:,5))/100]; Obs_Gr_C = DataUS(:,2); Obs_Gr_I = DataUS(:,3); Obs_Infl = DataUS(:,6); Obs_r1 = DataUS(:,7); Obs_r40 = DataUS(:,8); Obs_xhr40 = DataUS(:,9); Obs_GoY = log(DataUS(:,4)); Obs_hours = 4*log(DataUS(:,5))/100; save('AFVRR_data.mat','Obs_Gr_C','Obs_Gr_I','Obs_Infl','Obs_r1','Obs_r40','Obs_xhr40','Obs_GoY','Obs_hours'); pause(1); end; @#endif