-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathMainLearningFile.m
218 lines (144 loc) · 12.1 KB
/
MainLearningFile.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
% in this file we test CG learning
tic
times = 200;
% define parameter values
ParameterValues = defineParameters();
% solve New Keynesian model steady state
[ SteadyStateValuesNK ] = solveNK_SteadyState( ParameterValues );
% calculate parameter values denoted in the learning tex file
[ ParameterValuesLearning ] = calculateLearningParams( SteadyStateValuesNK, ParameterValues );
% create containers to store firms and households learning data
% actual law of motion, that is, how variables evolve according to the
% model. Include all the variables
ActualLawOfMotion = struct('capital',zeros(1,times),'wage',zeros(1,times),'inflation',zeros(1,times), 'interestRate',zeros(1,times),'markup',zeros(1,times),'consumption',zeros(1,times),'labour',zeros(1,times),'capitalReturn',zeros(1,times),'A',zeros(1,times));
% Household_PLM - households perceived law of motion
% HouseholdParameters - households regression (learning) parameters
Household_PLM = struct('capital',zeros(1,times),'wage',zeros(1,times),'inflation',zeros(1,times), 'interestRate',zeros(1,times),'markup',zeros(1,times),'A',zeros(1,times));
HouseholdParameters = struct('capital_Param',zeros(3,times),'wage_Param',zeros(3,times),'inflation_Param',zeros(3,times), 'interestRate_Param',zeros(3,times),'markup_Param',zeros(3,times));
% Firms_PLM - firms perceived law of motion
% FirmsParameters - firms regression (learning) parameters
Firms_PLM = struct('capital',zeros(1,times),'inflation',zeros(1,times),'markup',zeros(1,times),'A',zeros(1,times));
FirmsParameters = struct('capital_Param',zeros(3,times),'inflation_Param',zeros(3,times),'markup_Param',zeros(3,times));
% generate shock
ActualLawOfMotion.A(1,1) = 0; % initial value of shock
ActualLawOfMotion.A(1,2) = 0.01*randn;
for i = 3:times
ActualLawOfMotion.A(1,i) = ParameterValues.rho * ActualLawOfMotion.A(1,i-1);
end
% Rational Expectations Part
% load transition matrix from dynare (RE solution)
load REmatrix_A
% Households initial parameters in RE: capital, wage, interest rate,
% inflation and markup
HouseholdParameters.capital_Param(:,1) = [ SteadyStateValuesNK.k*(1-REmatrix_A(1,1)) REmatrix_A(1,1) SteadyStateValuesNK.k*REmatrix_A(2,1) ]';
HouseholdParameters.wage_Param(:,1) = [ SteadyStateValuesNK.w*(1-REmatrix_A(1,2)) REmatrix_A(1,2)*SteadyStateValuesNK.w/SteadyStateValuesNK.k SteadyStateValuesNK.w*REmatrix_A(2,2) ]';
HouseholdParameters.interestRate_Param(:,1) = [ SteadyStateValuesNK.R*(1-REmatrix_A(1,3)) REmatrix_A(1,3)*SteadyStateValuesNK.R/SteadyStateValuesNK.k SteadyStateValuesNK.R*REmatrix_A(2,3) ]';
HouseholdParameters.inflation_Param(:,1) = [ 1-REmatrix_A(1,4) REmatrix_A(1,4)/SteadyStateValuesNK.k REmatrix_A(2,4) ]';
HouseholdParameters.markup_Param(:,1) = [ SteadyStateValuesNK.X*(1-REmatrix_A(1,5)) REmatrix_A(1,5)*SteadyStateValuesNK.X/SteadyStateValuesNK.k SteadyStateValuesNK.X*REmatrix_A(2,5) ]';
% Firms initial parameters in RE: capital, inflation and markup
FirmsParameters.capital_Param(:,1) = [ SteadyStateValuesNK.k*(1-REmatrix_A(1,1)) REmatrix_A(1,1) SteadyStateValuesNK.k*REmatrix_A(2,1) ]';
FirmsParameters.inflation_Param(:,1) = [ 1-REmatrix_A(1,4) REmatrix_A(1,4)/SteadyStateValuesNK.k REmatrix_A(2,4) ]';
FirmsParameters.markup_Param(:,1) = [ SteadyStateValuesNK.X*(1-REmatrix_A(1,5)) REmatrix_A(1,5)*SteadyStateValuesNK.X/SteadyStateValuesNK.k SteadyStateValuesNK.X*REmatrix_A(2,5) ]';
% initialize variables in steady state at time 1.
ActualLawOfMotion.capital(1,1) = SteadyStateValuesNK.k;
ActualLawOfMotion.wage(1,1) = SteadyStateValuesNK.w;
ActualLawOfMotion.inflation(1,1) = 1;
ActualLawOfMotion.interestRate(1,1) = SteadyStateValuesNK.R;
ActualLawOfMotion.markup(1,1) = SteadyStateValuesNK.X;
ActualLawOfMotion.consumption(1,1) = SteadyStateValuesNK.c;
ActualLawOfMotion.labour(1,1) = SteadyStateValuesNK.L;
ActualLawOfMotion.capitalReturn(1,1) = SteadyStateValuesNK.rk;
% Initialize Perceived Law of Motion (PLM) at steady state values
% Households
Household_PLM.capital(1,1) = SteadyStateValuesNK.k;
Household_PLM.wage(1,1) = SteadyStateValuesNK.w;
Household_PLM.interestRate(1,1) = SteadyStateValuesNK.R;
Household_PLM.inflation(1,1) = 1;
Household_PLM.markup(1,1) = SteadyStateValuesNK.X;
% Firms
Firms_PLM.capital(1,1) = SteadyStateValuesNK.k;
Firms_PLM.inflation(1,1) = 1;
Firms_PLM.markup(1,1) = SteadyStateValuesNK.X;
forecastPeriod = 100; % # of periods to compute forecast for Households and Firms
% load initial second moment matrix from RE (obtained from dynare)
load REvariance
initial_D_Matrix = [ 1 SteadyStateValuesNK.k 0; SteadyStateValuesNK.k SteadyStateValuesNK.k^2+REvariance(1,1) REvariance(1,9);0 REvariance(9,1) REvariance(9,9) ];
% initialize learning as a class. At this point we can change learning
% algorithm
% households
HH_Capital_Learning = CG_Learning(0.1,HouseholdParameters.capital_Param(:,1), initial_D_Matrix, [ 1 Household_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.capital(1,1) );
HH_Wage_Learning = CG_Learning(0.1,HouseholdParameters.wage_Param(:,1), initial_D_Matrix, [ 1 Household_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.wage(1,1) );
HH_Inflation_Learning = CG_Learning(0.1,HouseholdParameters.inflation_Param(:,1), initial_D_Matrix, [ 1 Household_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.inflation(1,1) );
HH_Interest_Learning = CG_Learning(0.1,HouseholdParameters.interestRate_Param(:,1), initial_D_Matrix, [ 1 Household_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.interestRate(1,1) );
HH_Markup_Learning = CG_Learning(0.1,HouseholdParameters.markup_Param(:,1), initial_D_Matrix, [ 1 Household_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.markup(1,1) );
% firms
FF_Capital_Learning = CG_Learning(0.1,FirmsParameters.capital_Param(:,1), initial_D_Matrix, [ 1 Firms_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.capital(1,1) );
FF_Inflation_Learning = CG_Learning(0.1,FirmsParameters.inflation_Param(:,1), initial_D_Matrix, [ 1 Firms_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.inflation(1,1) );
FF_Markup_Learning = CG_Learning(0.1,FirmsParameters.markup_Param(:,1), initial_D_Matrix, [ 1 Firms_PLM.capital(1,1) ActualLawOfMotion.A(1,1) ]',ActualLawOfMotion.markup(1,1) );
% main learning loop
for t = 2:times
% update the state variable (capital)
ActualLawOfMotion.capital(1,t) = exp(ActualLawOfMotion.A(1,t))*ActualLawOfMotion.capital(1,t-1)^ParameterValues.alpha*ActualLawOfMotion.labour(1,t-1)^(1-ParameterValues.alpha)-ActualLawOfMotion.consumption(1,t-1)+(1-ParameterValues.delta)*ActualLawOfMotion.capital(1,t-1);
% update capital in learning algorithms
HH_Capital_Learning.variable = ActualLawOfMotion.capital(1,t);
FF_Capital_Learning.variable = ActualLawOfMotion.capital(1,t);
% define zMatrix for households and firms with updated capital
zMat = [1 ActualLawOfMotion.capital(1,t) ActualLawOfMotion.A(1,t)]';
% update D matrix and parameters for both firms and households
% households
[ HouseholdParameters.capital_Param(:,t), HH_D_Out_Capital ] = HH_Capital_Learning.do_CG_Learning();
[ HouseholdParameters.wage_Param(:,t), HH_D_Out_Wage ] = HH_Wage_Learning.do_CG_Learning();
[ HouseholdParameters.inflation_Param(:,t), HH_D_Out_Inflation ] = HH_Inflation_Learning.do_CG_Learning();
[ HouseholdParameters.interestRate_Param(:,t), HH_D_Out_Interest ] = HH_Interest_Learning.do_CG_Learning();
[ HouseholdParameters.markup_Param(:,t), HH_D_Out_Markup ] = HH_Markup_Learning.do_CG_Learning();
% firms
[ FirmsParameters.capital_Param(:,t), FF_D_Out_Capital ] = FF_Capital_Learning.do_CG_Learning();
[ FirmsParameters.inflation_Param(:,t), FF_D_Out_Inflation ] = FF_Inflation_Learning.do_CG_Learning();
[ FirmsParameters.markup_Param(:,t), FF_D_Out_Markup ] = FF_Markup_Learning.do_CG_Learning();
% compute one step ahead forecast / PLM using updated parameters
% households
B_Households = [ 1 0 0; HouseholdParameters.capital_Param(:,t)'; 0 0 ParameterValues.rho ];
tmpZ_HH = B_Households*zMat;
Household_PLM.capital(1,t) = tmpZ_HH(2,1);
Household_PLM.A(1,t) = tmpZ_HH(3,1);
Household_PLM.wage(1,t) = HouseholdParameters.wage_Param(:,t)'*tmpZ_HH;
Household_PLM.interestRate(1,t) = HouseholdParameters.interestRate_Param(:,t)'*tmpZ_HH;
Household_PLM.inflation(1,t) = HouseholdParameters.inflation_Param(:,t)'*tmpZ_HH;
Household_PLM.markup(1,t) = HouseholdParameters.markup_Param(:,t)'*tmpZ_HH;
B_Firms = [1 0 0; FirmsParameters.capital_Param(:,t)';0 0 ParameterValues.rho];
tmpZ_FF = B_Firms*zMat;
Firms_PLM.capital(1,t) = tmpZ_FF(2,1);
Firms_PLM.A(1,t) = tmpZ_FF(3,1);
Firms_PLM.inflation(1,t) = FirmsParameters.inflation_Param(:,t)'*tmpZ_FF;
Firms_PLM.markup(1,t) = FirmsParameters.markup_Param(:,t)'*tmpZ_FF;
% now solve for actual law of motion (RE)
[ E_K, E_S ] = Solve_Expecations( t, SteadyStateValuesNK, ParameterValuesLearning, ParameterValues, HouseholdParameters, FirmsParameters, ActualLawOfMotion );
ActualLawOfMotion.inflation(1,t) = SolveInflation2( ParameterValues, SteadyStateValuesNK, ParameterValuesLearning, t, ActualLawOfMotion, E_K, E_S );
ActualLawOfMotion.interestRate(1,t) = SteadyStateValuesNK.R*(1-ParameterValues.r_R)*ParameterValues.r_Infl*ActualLawOfMotion.inflation(1,t)-SteadyStateValuesNK.R*(1-ParameterValues.r_R)*ParameterValues.r_Infl+SteadyStateValuesNK.R;
ActualLawOfMotion.markup(1,t) = -ParameterValues.theta*SteadyStateValuesNK.X/(1-ParameterValues.theta)/(1-ParameterValues.theta*ParameterValues.beta)*ActualLawOfMotion.inflation(1,t)+ParameterValues.theta*SteadyStateValuesNK.X...
/(1-ParameterValues.theta)/(1-ParameterValues.theta*ParameterValues.beta)+SteadyStateValuesNK.X/(1-ParameterValues.theta*ParameterValues.beta)*E_K+SteadyStateValuesNK.X;
ActualLawOfMotion.capitalReturn(1,t) = ActualLawOfMotion.interestRate(1,t)/ActualLawOfMotion.inflation(1,t)-1+ParameterValues.delta;
ActualLawOfMotion.labour(1,t) = ActualLawOfMotion.capital(1,t)*(ActualLawOfMotion.markup(1,t)*ActualLawOfMotion.capitalReturn(1,t)/ParameterValues.alpha/exp(ActualLawOfMotion.A(1,t)))^(1/(1-ParameterValues.alpha));
ActualLawOfMotion.wage(1,t) = (1-ParameterValues.alpha)*exp(ActualLawOfMotion.A(1,t))*(ActualLawOfMotion.capital(1,t)/ActualLawOfMotion.labour(1,t))^ParameterValues.alpha/ActualLawOfMotion.markup(1,t);
ActualLawOfMotion.consumption(1,t) = ActualLawOfMotion.wage(1,t)*ActualLawOfMotion.labour(1,t)^(1-ParameterValues.eta);
% Update learning algorithms for next iteration
% households
HH_Capital_Learning = CG_Learning(0.1,HouseholdParameters.capital_Param(:,t), HH_D_Out_Capital, zMat,ActualLawOfMotion.capital(1,t) );
HH_Wage_Learning = CG_Learning(0.1,HouseholdParameters.wage_Param(:,t), HH_D_Out_Wage, zMat,ActualLawOfMotion.wage(1,t) );
HH_Inflation_Learning = CG_Learning(0.1,HouseholdParameters.inflation_Param(:,t), HH_D_Out_Inflation, zMat,ActualLawOfMotion.inflation(1,t) );
HH_Interest_Learning = CG_Learning(0.1,HouseholdParameters.interestRate_Param(:,t), HH_D_Out_Interest, zMat,ActualLawOfMotion.interestRate(1,t) );
HH_Markup_Learning = CG_Learning(0.1,HouseholdParameters.markup_Param(:,t), HH_D_Out_Markup, zMat,ActualLawOfMotion.markup(1,t) );
% firms
FF_Capital_Learning = CG_Learning(0.1,FirmsParameters.capital_Param(:,t), FF_D_Out_Capital, zMat,ActualLawOfMotion.capital(1,t) );
FF_Inflation_Learning = CG_Learning(0.1,FirmsParameters.inflation_Param(:,t), FF_D_Out_Inflation, zMat,ActualLawOfMotion.inflation(1,t) );
FF_Markup_Learning = CG_Learning(0.1,FirmsParameters.markup_Param(:,t), FF_D_Out_Markup, zMat,ActualLawOfMotion.markup(1,t) );
end
toc
figure
plot(ActualLawOfMotion.capital)
hold
plot(Household_PLM.capital,'r')
figure
plot(HouseholdParameters.capital_Param')
figure
plot(Household_PLM.inflation)