/* Test that perfect foresight simulation with expectation errors gives correct results with marginal linearization. Homotopy is not really triggered, but the use of homotopy_max_completion_share is enough to test the computations. This test is similar to pfwee_homotopy_marginal_linearization.mod, except that the numerical values are different (the shock is smaller to facilitate computations). */ var Consumption, Capital, LoggedProductivity; varexo LoggedProductivityInnovation; parameters beta, alpha, delta, rho; beta = .985; alpha = 1/3; delta = alpha/10; rho = .4; model; 1/Consumption = beta/Consumption(1)*(alpha*exp(LoggedProductivity(1))*Capital^(alpha-1)+1-delta); Capital = exp(LoggedProductivity)*Capital(-1)^alpha+(1-delta)*Capital(-1)-Consumption; LoggedProductivity = rho*LoggedProductivity(-1)+LoggedProductivityInnovation; end; initval; LoggedProductivityInnovation = 0.1; end; steady; endval; LoggedProductivityInnovation = 0.3; end; shocks(learnt_in = 5); var LoggedProductivityInnovation; periods 7; values -0.5; end; endval(learnt_in = 10); LoggedProductivityInnovation = 0.6; end; // Save initial steady state (it will be modified by pfwee) orig_steady_state = oo_.steady_state; orig_exo_steady_state = oo_.exo_steady_state; perfect_foresight_with_expectation_errors_setup(periods = 50); perfect_foresight_with_expectation_errors_solver(homotopy_max_completion_share = 0.8, homotopy_marginal_linearization_fallback, steady_solve_algo = 13); verbatim; if ~oo_.deterministic_simulation.status error('Perfect foresight simulation failed') end pfwee_endo_simul = oo_.endo_simul; pfwee_exo_simul = oo_.exo_simul; oo_.steady_state = orig_steady_state; oo_.exo_steady_state = orig_exo_steady_state; oo_ = make_ex_(M_,options_,oo_); oo_ = make_y_(M_,options_,oo_); options_.simul.homotopy_marginal_linearization_fallback = 0; options_.simul.homotopy_max_completion_share = 1; options_.solve_algo = 13; %% Information arriving in period 1 (permanent shock now) % Simulation at 80% disp('*** info_period=1, share=80%') oo_.exo_steady_state = 0.1+0.2*0.8; oo_.steady_state = evaluate_steady_state(oo_.steady_state, oo_.exo_steady_state, M_, options_, true); oo_.endo_simul(:,end) = oo_.steady_state; oo_.exo_simul(2:end,:) = oo_.exo_steady_state; sim1_1 = perfect_foresight_solver(M_, options_, oo_); % Simulation at 79% disp('*** info_period=1, share=79%') oo_.exo_steady_state = 0.1+0.2*0.79; oo_.steady_state = evaluate_steady_state(oo_.steady_state, oo_.exo_steady_state, M_, options_, true); oo_.endo_simul(:,end) = oo_.steady_state; oo_.exo_simul(2:end,:) = oo_.exo_steady_state; sim2_1 = perfect_foresight_solver(M_, options_, oo_); %% Information arriving in period 5 (temporary shock in period 7) options_.periods = 46; % Simulation at 80% disp('*** info_period=5, share=80%') oo_.endo_simul = sim1_1.endo_simul(:,5:end); oo_.exo_simul = sim1_1.exo_simul(5:end,:); oo_.steady_state = sim1_1.steady_state; oo_.exo_steady_state = sim1_1.exo_steady_state; oo_.exo_simul(4,1) = 0.1-0.6*0.8; sim1_5 = perfect_foresight_solver(M_, options_, oo_); % Simulation at 79% disp('*** info_period=5, share=79%') oo_.endo_simul = sim2_1.endo_simul(:,5:end); oo_.exo_simul = sim2_1.exo_simul(5:end,:); oo_.steady_state = sim2_1.steady_state; oo_.exo_steady_state = sim2_1.exo_steady_state; oo_.exo_simul(4,1) = 0.1-0.6*0.79; sim2_5 = perfect_foresight_solver(M_, options_, oo_); %% Information arriving in period 10 (permanent shock now) options_.periods = 41; % Simulation at 80% disp('*** info_period=10, share=80%') oo_.endo_simul = sim1_5.endo_simul(:,6:end); oo_.exo_simul = sim1_5.exo_simul(6:end,:); oo_.exo_steady_state = 0.1+0.5*0.8; oo_.steady_state = evaluate_steady_state(sim1_5.steady_state, oo_.exo_steady_state, M_, options_, true); oo_.endo_simul(:,end) = oo_.steady_state; oo_.exo_simul(2:end,:) = oo_.exo_steady_state; sim1_10 = perfect_foresight_solver(M_, options_, oo_); % Simulation at 79% disp('*** info_period=10, share=79%') oo_.endo_simul = sim2_5.endo_simul(:,6:end); oo_.exo_simul = sim2_5.exo_simul(6:end,:); oo_.exo_steady_state = 0.1+0.5*0.79; oo_.steady_state = evaluate_steady_state(sim2_5.steady_state, oo_.exo_steady_state, M_, options_, true); oo_.endo_simul(:,end) = oo_.steady_state; oo_.exo_simul(2:end,:) = oo_.exo_steady_state; sim2_10 = perfect_foresight_solver(M_, options_, oo_); %% Reconstruct the two simulations sim1_endo_simul = [ sim1_1.endo_simul(:,1:5), sim1_5.endo_simul(:,2:6), sim1_10.endo_simul(:,2:end);]; sim2_endo_simul = [ sim2_1.endo_simul(:,1:5), sim2_5.endo_simul(:,2:6), sim2_10.endo_simul(:,2:end);]; sim1_exo_simul = [ sim1_1.exo_simul(1:5,:); sim1_5.exo_simul(2:6,:); sim1_10.exo_simul(2:end,:);]; sim2_exo_simul = [ sim2_1.exo_simul(1:5,:); sim2_5.exo_simul(2:6,:); sim2_10.exo_simul(2:end,:);]; %% Compute marginal linearization endo_simul = sim1_endo_simul + 20*(sim1_endo_simul - sim2_endo_simul); exo_simul = sim1_exo_simul + 20*(sim1_exo_simul - sim2_exo_simul); endo_abs_diff = max(max(abs(pfwee_endo_simul-endo_simul))); endo_rel_diff = max(max(abs(endo_simul./pfwee_endo_simul-1))); fprintf('Max absolute difference on endogenous: %g\nMax relative difference on endogenous: %g%%\n', endo_abs_diff, 100*endo_rel_diff) exo_abs_diff = max(max(abs(pfwee_exo_simul-exo_simul))); exo_rel_diff = max(max(abs(exo_simul./pfwee_exo_simul-1))); fprintf('Max absolute difference on exogenous: %g\nMax relative difference on exogenous: %g%%\n', exo_abs_diff, 100*exo_rel_diff) if endo_abs_diff > 1e-3 || endo_rel_diff > 3e-3 || exo_abs_diff > 1e-15 || exo_rel_diff > 1e-13 error('perfect foresight with expectation errors + marginal linearization gives wrong result') end end; // verbatim