time | Calls | line |
---|
| | 7 | function result_map = ff_abzr_fibs_vf_vec(varargin)
|
| | 8 | %% FF_ABZR_FIBS_VF_VEC solve infinite horizon exo shock + endo asset problem
|
| | 9 | % With R shock.
|
| | 10 | %
|
| | 11 | % The model could be invoked mainly in sveral ways:
|
| | 12 | %
|
| | 13 | % # param_map('bl_default') = true; param_map('bl_bridge') = false;
|
| | 14 | % param_map('bl_rollover') = true; Given these, default is possible, bridge
|
| | 15 | % loans are not needed because rollover is allowed for formal loans (or
|
| | 16 | % informal loans)
|
| | 17 | % # we change param_map('bl_bridge') = true, that means
|
| | 18 | % rollover is still allowed, but only allowed using informal sources,
|
| | 19 | % formal loans no longer allow for roll-over. Furthermore, if both
|
| | 20 | % bl_bridge and bl_rollover are false, that means we are not allowing for
|
| | 21 | % rollover at all, so households can not borrow such that they end up with
|
| | 22 | % negative cash-on-hand.
|
| | 23 | %
|
| | 24 | % Default simulation bl_bridge = false.
|
| | 25 | %
|
| | 26 | % @param param_map container parameter container
|
| | 27 | %
|
| | 28 | % @param support_map container support container
|
| | 29 | %
|
| | 30 | % @param armt_map container container with states, choices and shocks
|
| | 31 | % grids that are inputs for grid based solution algorithm
|
| | 32 | %
|
| | 33 | % @param func_map container container with function handles for
|
| | 34 | % consumption cash-on-hand etc.
|
| | 35 | %
|
| | 36 | % @return result_map container contains policy function matrix, value
|
| | 37 | % function matrix, iteration results, and policy function, value function
|
| | 38 | % and iteration results tables.
|
| | 39 | %
|
| | 40 | % keys included in result_map:
|
| | 41 | %
|
| | 42 | % * mt_val matrix states_n by shock_n matrix of converged value function grid
|
| | 43 | % * mt_pol_a matrix states_n by shock_n matrix of converged policy function grid
|
| | 44 | % * ar_val_diff_norm array if bl_post = true it_iter_last by 1 val function
|
| | 45 | % difference between iteration
|
| | 46 | % * ar_pol_diff_norm array if bl_post = true it_iter_last by 1 policy
|
| | 47 | % function difference between iterations
|
| | 48 | % * mt_pol_perc_change matrix if bl_post = true it_iter_last by shock_n the
|
| | 49 | % proportion of grid points at which policy function changed between
|
| | 50 | % current and last iteration for each element of shock
|
| | 51 | %
|
| | 52 | % @example
|
| | 53 | %
|
| | 54 | % % Get Default Parameters
|
| | 55 | % it_param_set = 4;
|
| | 56 | % [param_map, support_map] = ffs_abzr_fibs_set_default_param(it_param_set);
|
| | 57 | % % Chnage param_map keys for borrowing
|
| | 58 | % param_map('fl_b_bd') = -20; % borrow bound
|
| | 59 | % param_map('bl_default') = false; % true if allow for default
|
| | 60 | % param_map('fl_c_min') = 0.0001; % u(c_min) when default
|
| | 61 | % % Change Keys in param_map
|
| | 62 | % param_map('it_a_n') = 75;
|
| | 63 | % param_map('fl_z_r_borr_n') = 3;
|
| | 64 | % param_map('it_z_wage_n') = 5;
|
| | 65 | % param_map('it_z_n') = param_map('it_z_wage_n') * param_map('fl_z_r_borr_n');
|
| | 66 | % param_map('fl_a_max') = 100;
|
| | 67 | % param_map('fl_w') = 1.3;
|
| | 68 | % % Change Keys support_map
|
| | 69 | % support_map('bl_display') = false;
|
| | 70 | % support_map('bl_post') = true;
|
| | 71 | % support_map('bl_display_final') = false;
|
| | 72 | % % Call Program with external parameters that override defaults.
|
| | 73 | % ff_abzr_fibs_vf_vec(param_map, support_map);
|
| | 74 | %
|
| | 75 | % @include
|
| | 76 | %
|
| | 77 | % * <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/m_abzr_paramfunc/html/ffs_abzr_fibs_set_default_param.html ffs_abzr_fibs_set_default_param>
|
| | 78 | % * <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/m_abzr_paramfunc/html/ffs_abzr_fibs_get_funcgrid.html ffs_abzr_fibs_get_funcgrid>
|
| | 79 | % * <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/paramfunc_fibs/html/ffs_fibs_min_c_cost_bridge.html ffs_fibs_min_c_cost_bridge>
|
| | 80 | % * <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/paramfunc_fibs/html/ffs_fibs_inf_bridge.html ffs_fibs_inf_bridge>
|
| | 81 | % * <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/paramfunc_fibs/html/ffs_fibs_min_c_cost.html ffs_fibs_min_c_cost>
|
| | 82 | % * <https://fanwangecon.github.io/CodeDynaAsset/m_az/solvepost/html/ff_az_vf_post.html ff_az_vf_post>
|
| | 83 | %
|
| | 84 | % @seealso
|
| | 85 | %
|
| | 86 | % * for/inf + save + borr loop: <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/m_abzr_solve/html/ff_abzr_fibs_vf.html ff_abzr_fibs_vf>
|
| | 87 | % * for/inf + borr vectorized: <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/m_abzr_solve/html/ff_abzr_fibs_vf_vec.html ff_abzr_fibs_vf_vec>
|
| | 88 | % * for/inf + borr optimized-vectorized: <https://fanwangecon.github.io/CodeDynaAsset/m_fibs/m_abzr_solve/html/ff_abzr_fibs_vf_vecsv.html ff_abzr_fibs_vf_vecsv>
|
| | 89 | %
|
| | 90 |
|
| | 91 |
|
| | 92 | %% Default
|
| | 93 | % * it_param_set = 1: quick test
|
| | 94 | % * it_param_set = 2: benchmark run
|
| | 95 | % * it_param_set = 3: benchmark profile
|
| | 96 | % * it_param_set = 4: press publish button
|
| | 97 |
|
| | 98 | it_param_set = 3;
|
| | 99 | bl_input_override = true;
|
| | 100 | [param_map, support_map] = ffs_abzr_fibs_set_default_param(it_param_set);
|
| | 101 |
|
| | 102 | % Note: param_map and support_map can be adjusted here or outside to override defaults
|
| | 103 | % To generate results as if formal informal do not matter
|
| | 104 |
|
| | 105 | % param_map('fl_r_fsv') = 0.025;
|
| | 106 | % param_map('fl_r_fbr') = 0.035;
|
| | 107 | % param_map('bl_b_is_principle') = false;
|
| | 108 | % param_map('st_forbrblk_type') = 'seg3';
|
| | 109 | % param_map('fl_forbrblk_brmost') = -19;
|
| | 110 | % param_map('fl_forbrblk_brleast') = -1;
|
| | 111 | % param_map('fl_forbrblk_gap') = -1.5;
|
| | 112 | % param_map('bl_b_is_principle') = false;
|
| | 113 | % param_map('it_a_n') = 750;
|
| | 114 | % param_map('fl_z_r_borr_n') = 5;
|
| | 115 | % param_map('it_z_wage_n') = 15;
|
| | 116 | % param_map('it_z_n') = param_map('it_z_wage_n') * param_map('fl_z_r_borr_n');
|
| | 117 |
|
| | 118 | [armt_map, func_map] = ffs_abzr_fibs_get_funcgrid(param_map, support_map); % 1 for override
|
| | 119 | default_params = {param_map support_map armt_map func_map};
|
| | 120 |
|
| | 121 | %% Parse Parameters 1
|
| | 122 |
|
| | 123 | % if varargin only has param_map and support_map,
|
| | 124 | params_len = length(varargin);
|
| | 125 | [default_params{1:params_len}] = varargin{:};
|
| | 126 | param_map = [param_map; default_params{1}];
|
| | 127 | support_map = [support_map; default_params{2}];
|
| | 128 | if params_len >= 1 && params_len <= 2
|
| | 129 | % If override param_map, re-generate armt and func if they are not
|
| | 130 | % provided
|
| | 131 | [armt_map, func_map] = ffs_abzr_fibs_get_funcgrid(param_map, support_map);
|
| | 132 | else
|
| | 133 | % Override all
|
| | 134 | armt_map = [armt_map; default_params{3}];
|
| | 135 | func_map = [func_map; default_params{4}];
|
| | 136 | end
|
| | 137 |
|
| | 138 | % append function name
|
| | 139 | st_func_name = 'ff_abzr_fibs_vf_vec';
|
| | 140 | support_map('st_profile_name_main') = [st_func_name support_map('st_profile_name_main')];
|
| | 141 | support_map('st_mat_name_main') = [st_func_name support_map('st_mat_name_main')];
|
| | 142 | support_map('st_img_name_main') = [st_func_name support_map('st_img_name_main')];
|
| | 143 |
|
| | 144 | %% Parse Parameters 2
|
| | 145 |
|
| | 146 | % armt_map
|
| | 147 | params_group = values(armt_map, {'ar_a', 'mt_z_trans', 'ar_z_r_infbr_mesh_wage', 'ar_z_wage_mesh_r_infbr'});
|
| | 148 | [ar_a, mt_z_trans, ar_z_r_infbr_mesh_wage, ar_z_wage_mesh_r_infbr] = params_group{:};
|
| | 149 |
|
| | 150 | % Formal choice Menu/Grid and Interest Rate Menu/Grid
|
| | 151 | params_group = values(armt_map, {'ar_forbrblk_r', 'ar_forbrblk'});
|
| | 152 | [ar_forbrblk_r, ar_forbrblk] = params_group{:};
|
| | 153 |
|
| | 154 | % func_map
|
| | 155 | params_group = values(func_map, {'f_util_log', 'f_util_crra', 'f_coh', 'f_cons_coh_fbis', 'f_cons_coh_save'});
|
| | 156 | [f_util_log, f_util_crra, f_coh, f_cons_coh_fbis, f_cons_coh_save] = params_group{:};
|
| | 157 |
|
| | 158 | % param_map
|
| | 159 | params_group = values(param_map, {'it_a_n', 'it_z_n', 'fl_crra', 'fl_beta', 'fl_c_min',...
|
| | 160 | 'fl_nan_replace', 'bl_default', 'bl_bridge', 'bl_rollover', 'fl_default_aprime'});
|
| | 161 | [it_a_n, it_z_n, fl_crra, fl_beta, fl_c_min, ...
|
| | 162 | fl_nan_replace, bl_default, bl_bridge, bl_rollover, fl_default_aprime] = params_group{:};
|
| | 163 | params_group = values(param_map, {'it_maxiter_val', 'fl_tol_val', 'fl_tol_pol', 'it_tol_pol_nochange'});
|
| | 164 | [it_maxiter_val, fl_tol_val, fl_tol_pol, it_tol_pol_nochange] = params_group{:};
|
| | 165 |
|
| | 166 | % param_map, Formal informal
|
| | 167 | params_group = values(param_map, {'fl_r_fsv', 'bl_b_is_principle'});
|
| | 168 | [fl_r_fsv, bl_b_is_principle] = params_group{:};
|
| | 169 |
|
| | 170 | % support_map
|
| | 171 | params_group = values(support_map, {'bl_profile', 'st_profile_path', ...
|
| | 172 | 'st_profile_prefix', 'st_profile_name_main', 'st_profile_suffix',...
|
| | 173 | 'bl_display_minccost', 'bl_display_infbridge', ...
|
| | 174 | 'bl_time', 'bl_display_defparam', 'bl_display', 'it_display_every', 'bl_post'});
|
| | 175 | [bl_profile, st_profile_path, ...
|
| | 176 | st_profile_prefix, st_profile_name_main, st_profile_suffix, ...
|
| | 177 | bl_display_minccost, bl_display_infbridge, ...
|
| | 178 | bl_time, bl_display_defparam, bl_display, it_display_every, bl_post] = params_group{:};
|
| | 179 |
|
| | 180 | %% Initialize Output Matrixes
|
| | 181 | % include mt_pol_idx which we did not have in looped code
|
| | 182 |
|
| | 183 | mt_val_cur = zeros(it_a_n,it_z_n);
|
| | 184 | mt_val = mt_val_cur - 1;
|
| | 185 | mt_pol_a = zeros(it_a_n,it_z_n);
|
| | 186 | mt_pol_a_cur = mt_pol_a - 1;
|
| | 187 | mt_pol_idx = zeros(it_a_n,it_z_n);
|
| | 188 | mt_pol_cons = zeros(it_a_n,it_z_n);
|
| | 189 |
|
| | 190 | % collect optimal borrowing formal and informal choices
|
| | 191 | mt_pol_b_bridge = zeros(it_a_n,it_z_n);
|
| | 192 | mt_pol_inf_borr_nobridge = zeros(it_a_n,it_z_n);
|
| | 193 | mt_pol_for_borr = zeros(it_a_n,it_z_n);
|
| | 194 | mt_pol_for_save = zeros(it_a_n,it_z_n);
|
| | 195 |
|
| | 196 | %% Initialize Convergence Conditions
|
| | 197 |
|
| | 198 | bl_vfi_continue = true;
|
| | 199 | it_iter = 0;
|
| | 200 | ar_val_diff_norm = zeros([it_maxiter_val, 1]);
|
| | 201 | ar_pol_diff_norm = zeros([it_maxiter_val, 1]);
|
| | 202 | mt_pol_perc_change = zeros([it_maxiter_val, it_z_n]);
|
| | 203 |
|
| | 204 | %% Iterate Value Function
|
| | 205 | % Loop solution with 4 nested loops
|
| | 206 | %
|
| | 207 | % # loop 1: over exogenous states
|
| | 208 | % # loop 2: over endogenous states
|
| | 209 | % # loop 3: over choices
|
| | 210 | % # loop 4: add future utility, integration--loop over future shocks
|
| | 211 | %
|
| | 212 |
|
| | 213 | % Start Profile
|
| | 214 | if (bl_profile)
|
| | 215 | close all;
|
| | 216 | profile off;
|
| | 217 | profile on;
|
< 0.001 | 1 | 218 | end
|
| | 219 |
|
| | 220 | % Start Timer
|
< 0.001 | 1 | 221 | if (bl_time)
|
< 0.001 | 1 | 222 | tic;
|
< 0.001 | 1 | 223 | end
|
| | 224 |
|
| | 225 | % Value Function Iteration
|
< 0.001 | 1 | 226 | while bl_vfi_continue
|
< 0.001 | 122 | 227 | it_iter = it_iter + 1;
|
| | 228 |
|
| | 229 | %% Solve Optimization Problem Current Iteration
|
| | 230 |
|
| | 231 | % loop 1: over exogenous states
|
< 0.001 | 122 | 232 | for it_z_i = 1:it_z_n
|
| | 233 |
|
| | 234 | %% Solve the Formal Informal Problem for each a' and coh: c_forinf(a')
|
| | 235 | % find the today's consumption maximizing formal and informal
|
| | 236 | % choices given a' and coh. The formal and informal choices need to
|
| | 237 | % generate exactly a', but depending on which formal and informal
|
| | 238 | % joint choice is used, the consumption cost today a' is different.
|
| | 239 | % Note here, a is principle + interests. Three areas:
|
| | 240 | %
|
| | 241 | % * *CASE A* a' > 0: savings, do not need to optimize over formal and
|
| | 242 | % informal choices
|
| | 243 | % * *CASE B* a' < 0 & coh < 0: need bridge loan to pay for unpaid debt, and
|
| | 244 | % borrowing over-all, need to first pick bridge loan to pay for
|
| | 245 | % debt, if bridge loan is insufficient, go into default. After
|
| | 246 | % bridge loan, optimize over formal+informal, borrow+save joint
|
| | 247 | % choices.
|
| | 248 | % * *CASE C* a' < 0 & coh > 0: do not need to get informal bridge loans,
|
| | 249 | % optimize over for+inf save, for+save+borr, inf+borr only, for
|
| | 250 | % borrow only.
|
| | 251 | %
|
| | 252 |
|
| | 253 | % 1. Current Shock
|
0.003 | 9150 | 254 | fl_z_r_borr = ar_z_r_infbr_mesh_wage(it_z_i);
|
0.003 | 9150 | 255 | fl_z_wage = ar_z_wage_mesh_r_infbr(it_z_i);
|
| | 256 |
|
| | 257 | % 2. cash-on-hand
|
0.113 | 9150 | 258 | ar_coh = f_coh(fl_z_wage, ar_a);
|
| | 259 |
|
| | 260 | % 3. *CASE A* initiate consumption matrix as if all save
|
14.436 | 9150 | 261 | mt_c = f_cons_coh_save(ar_coh, ar_a');
|
| | 262 |
|
| | 263 | % 3. if Bridge Loan is Needed
|
| | 264 |
|
| | 265 | % 4. *CASE B+C* get negative coh index and get borrowing choices index
|
0.022 | 9150 | 266 | ar_coh_neg_idx = (ar_coh <= 0);
|
| | 267 |
|
0.013 | 9150 | 268 | ar_a_neg_idx = (ar_a < 0);
|
0.070 | 9150 | 269 | ar_coh_neg = ar_coh(ar_coh_neg_idx);
|
0.033 | 9150 | 270 | ar_a_neg = ar_a(ar_a_neg_idx);
|
| | 271 |
|
| | 272 | % 5. if coh > 0 and ap < 0, can allow same for+inf result to all coh.
|
| | 273 | % The procedure below works regardless of how ar_coh is sorted. get
|
| | 274 | % the index of all negative coh elements as well as first
|
| | 275 | % non-negative element. We solve the formal and informal problem at
|
| | 276 | % these points, note that we only need to solve the formal and
|
| | 277 | % informal problem for positive coh level once.
|
0.087 | 9150 | 278 | ar_coh_first_pos_idx = (cumsum(ar_coh_neg_idx == 0) == 1);
|
0.010 | 9150 | 279 | ar_coh_forinfsolve_idx = (ar_coh_first_pos_idx | ar_coh_neg_idx);
|
0.036 | 9150 | 280 | ar_coh_forinfsolve_a_neg_idx = (ar_coh(ar_coh_forinfsolve_idx) <= 0);
|
| | 281 |
|
| | 282 | % 6. *CASE B + C* Negative asset choices (borrowing), 1 col Case C
|
| | 283 | % negp1: negative coh + 1, 1 meaning 1 positive coh, first positive
|
| | 284 | % coh column index element grabbed.
|
1.319 | 9150 | 285 | mt_coh_negp1_mesh_neg_aprime = zeros(size(ar_a_neg')) + ar_coh(ar_coh_forinfsolve_idx);
|
0.976 | 9150 | 286 | mt_neg_aprime_mesh_coh_negp1 = zeros(size(mt_coh_negp1_mesh_neg_aprime)) + ar_a_neg';
|
| | 287 |
|
0.004 | 9150 | 288 | if (bl_bridge)
|
| | 289 | % mt_neg_aprime_mesh_coh_1col4poscoh = zeros([length(ar_a_neg), (length(ar_coh_neg)+1)]) + ar_a_neg';
|
| | 290 | % ar_coh_neg_idx_1col4poscoh = ar_coh_neg_idx(1:(length(ar_coh_neg)+1));
|
| | 291 |
|
| | 292 | % 6. *CASE B* Solve for: if (fl_ap < 0) and if (fl_coh < 0)
|
9.109 | 9150 | 293 | [mt_aprime_nobridge_negcoh, ~, mt_c_bridge_negcoh] = ffs_fibs_inf_bridge(...
|
| 9150 | 294 | bl_b_is_principle, fl_z_r_borr, ...
|
| 9150 | 295 | mt_neg_aprime_mesh_coh_negp1(:,ar_coh_forinfsolve_a_neg_idx), ...
|
| 9150 | 296 | mt_coh_negp1_mesh_neg_aprime(:,ar_coh_forinfsolve_a_neg_idx), ...
|
| 9150 | 297 | bl_display_infbridge, bl_input_override);
|
| | 298 |
|
| | 299 | % generate mt_aprime_nobridge
|
0.342 | 9150 | 300 | mt_neg_aprime_mesh_coh_negp1(:, ar_coh_forinfsolve_a_neg_idx) = mt_aprime_nobridge_negcoh;
|
| | 301 | else
|
| | 302 | % no bridge loan needed means roll over is allowed.
|
| | 303 | mt_neg_aprime_mesh_coh_negp1 = ar_a_neg';
|
0.002 | 9150 | 304 | end
|
| | 305 |
|
| | 306 | % 7. *CASE B + C* formal and informal joint choices, 1 col Case C
|
0.001 | 9150 | 307 | bl_input_override = true;
|
31.332 | 9150 | 308 | [ar_max_c_nobridge, ~, ~, ~] = ...
|
| | 309 | ffs_fibs_min_c_cost(...
|
| 9150 | 310 | bl_b_is_principle, fl_z_r_borr, fl_r_fsv, ...
|
| 9150 | 311 | ar_forbrblk_r, ar_forbrblk, ...
|
| 9150 | 312 | mt_neg_aprime_mesh_coh_negp1(:), ...
|
| 9150 | 313 | bl_display_minccost, bl_input_override);
|
| | 314 |
|
| | 315 | %% Update Consumption Matrix *CASE A + B + C* Consumptions
|
| | 316 | % Current mt_c is assuming all to be case A
|
| | 317 | %
|
| | 318 | % * Update Columns for case B (negative coh)
|
| | 319 | % * Update Columns for case C (1 column): ar_coh_first_pos_idx,
|
| | 320 | % included in ar_coh_forinfsolve_idx
|
| | 321 | % * Update Columns for all case C: ~ar_coh_neg_idx using 1 column
|
| | 322 | % result
|
| | 323 | %
|
| | 324 |
|
| | 325 | % 1. Initalize all Neg Aprime consumption cost of aprime inputs
|
| | 326 | % Initialize
|
4.820 | 9150 | 327 | mt_max_c_nobridge_a_neg = zeros([length(ar_a_neg), length(ar_coh)]) + 0;
|
4.526 | 9150 | 328 | mt_c_bridge_coh_a_neg = zeros(size(mt_max_c_nobridge_a_neg)) + 0;
|
| | 329 |
|
| | 330 | % 2. Fill in *Case B* and *Case C* (one column) Other C-cost
|
0.511 | 9150 | 331 | mt_max_c_nobridge_negcohp1 = reshape(ar_max_c_nobridge, [size(mt_neg_aprime_mesh_coh_negp1)]);
|
| | 332 |
|
0.003 | 9150 | 333 | if (bl_bridge)
|
| | 334 | % 2. Fill in *Case B* Bridge C-cost
|
0.512 | 9150 | 335 | mt_c_bridge_coh_a_neg(:, ar_coh_neg_idx) = mt_c_bridge_negcoh;
|
| | 336 |
|
| | 337 | % 2. Fill in *Case B* and *Case C* (one column) Other C-cost
|
0.433 | 9150 | 338 | mt_max_c_nobridge_a_neg(:, ar_coh_forinfsolve_idx) = mt_max_c_nobridge_negcohp1;
|
7.503 | 9150 | 339 | mt_max_c_nobridge_a_neg(:, ~ar_coh_forinfsolve_idx) = ...
|
| 9150 | 340 | zeros(size(mt_c(ar_a_neg_idx, ~ar_coh_forinfsolve_idx))) ...
|
| 9150 | 341 | + mt_max_c_nobridge_negcohp1(:, ~ar_coh_forinfsolve_a_neg_idx);
|
| | 342 | else
|
| | 343 | mt_max_c_nobridge_a_neg = zeros([length(ar_a_neg), length(ar_coh)]) + mt_max_c_nobridge_negcohp1;
|
0.002 | 9150 | 344 | end
|
| | 345 |
|
| | 346 | % 3. Consumption for B + C Cases
|
| | 347 | % note, the c cost of aprime is the same for all coh > 0, but mt_c
|
| | 348 | % is different still for each coh and aprime.
|
9.290 | 9150 | 349 | mt_c_forinfsolve = f_cons_coh_fbis(ar_coh, mt_c_bridge_coh_a_neg + mt_max_c_nobridge_a_neg);
|
| | 350 |
|
| | 351 | % 4. Update with Case B and C
|
1.944 | 9150 | 352 | mt_c(ar_a_neg_idx, :) = mt_c_forinfsolve;
|
| | 353 |
|
| | 354 | %% Solve Optimization Problem: max_{a'} (u(c_forinf(a')) + EV(a',z'))
|
| | 355 | % 1. EVAL current utility: N by N, f_util defined earlier
|
0.003 | 9150 | 356 | if (fl_crra == 1)
|
| | 357 | mt_utility = f_util_log(mt_c);
|
| | 358 | fl_u_cmin = f_util_log(fl_c_min);
|
0.001 | 9150 | 359 | else
|
111.327 | 9150 | 360 | mt_utility = f_util_crra(mt_c);
|
0.176 | 9150 | 361 | fl_u_cmin = f_util_crra(fl_c_min);
|
0.002 | 9150 | 362 | end
|
| | 363 |
|
| | 364 | % 2. f(z'|z)
|
0.039 | 9150 | 365 | ar_z_trans_condi = mt_z_trans(it_z_i,:);
|
| | 366 |
|
| | 367 | % 3. EVAL EV((A',K'),Z'|Z) = V((A',K'),Z') x p(z'|z)', (N by Z) x (Z by 1) = N by 1
|
| | 368 | % Note: transpose ar_z_trans_condi from 1 by Z to Z by 1
|
| | 369 | % Note: matrix multiply not dot multiply
|
1.001 | 9150 | 370 | mt_evzp_condi_z = mt_val_cur * ar_z_trans_condi';
|
| | 371 |
|
| | 372 | % 4. EVAL add on future utility, N by N + N by 1, broadcast again
|
1.847 | 9150 | 373 | mt_utility = mt_utility + fl_beta*mt_evzp_condi_z;
|
| | 374 |
|
0.004 | 9150 | 375 | if (bl_default)
|
| | 376 | % if default: only today u(cmin), transition out next period, debt wiped out
|
47.286 | 9150 | 377 | mt_utility(mt_c <= fl_c_min) = fl_u_cmin + fl_beta*mt_evzp_condi_z(ar_a == fl_default_aprime);
|
| | 378 | else
|
| | 379 | % if default is not allowed: v = fl_nan_replace
|
| | 380 | mt_utility(mt_c <= fl_c_min) = fl_nan_replace;
|
0.003 | 9150 | 381 | end
|
| | 382 |
|
| | 383 | % Set below threshold c to c_min
|
18.899 | 9150 | 384 | mt_c(mt_c < fl_c_min) = fl_c_min;
|
| | 385 |
|
| | 386 | % 5. no bridge and no rollover allowed
|
0.003 | 9150 | 387 | if( ~bl_rollover && ~bl_bridge)
|
| | 388 | if (bl_default)
|
| | 389 | % if default: only today u(cmin), transition out next period, debt wiped out
|
| | 390 | mt_utility(:, ar_coh_neg_idx) = fl_u_cmin + fl_beta*mt_evzp_condi_z(ar_a == fl_default_aprime);
|
| | 391 | else
|
| | 392 | % if default is not allowed: v = fl_nan_replace
|
| | 393 | mt_utility(:, ar_coh_neg_idx) = fl_nan_replace;
|
| | 394 | end
|
| | 395 | end
|
| | 396 |
|
| | 397 | % 5. Optimization: remember matlab is column major, rows must be
|
| | 398 | % choices, columns must be states
|
| | 399 | % <https://en.wikipedia.org/wiki/Row-_and_column-major_order COLUMN-MAJOR>
|
| | 400 | % mt_utility is N by N, rows are choices, cols are states.
|
3.256 | 9150 | 401 | [ar_opti_val_z, ar_opti_idx_z] = max(mt_utility);
|
0.075 | 9150 | 402 | [it_choies_n, it_states_n] = size(mt_utility);
|
0.261 | 9150 | 403 | ar_add_grid = linspace(0, it_choies_n*(it_states_n-1), it_states_n);
|
0.018 | 9150 | 404 | ar_opti_linear_idx_z = ar_opti_idx_z + ar_add_grid;
|
0.089 | 9150 | 405 | ar_opti_aprime_z = ar_a(ar_opti_idx_z);
|
0.170 | 9150 | 406 | ar_opti_c_z = mt_c(ar_opti_linear_idx_z);
|
| | 407 |
|
| | 408 | % 6. Handle Default is optimal or not
|
0.003 | 9150 | 409 | if (bl_default)
|
| | 410 | % if defaulting is optimal choice, at these states, not required
|
| | 411 | % to default, non-default possible, but default could be optimal
|
0.054 | 9150 | 412 | ar_opti_aprime_z(ar_opti_c_z <= fl_c_min) = fl_default_aprime;
|
0.072 | 9150 | 413 | ar_opti_idx_z(ar_opti_c_z <= fl_c_min) = find(ar_a == fl_default_aprime);
|
| | 414 | else
|
| | 415 | % if default is not allowed, then next period same state as now
|
| | 416 | % this is absorbing state, this is the limiting case, single
|
| | 417 | % state space point, lowest a and lowest shock has this.
|
| | 418 | ar_opti_aprime_z(ar_opti_c_z <= fl_c_min) = min(ar_a);
|
0.001 | 9150 | 419 | end
|
| | 420 |
|
| | 421 | % 6. no bridge and no rollover allowed
|
0.002 | 9150 | 422 | if( ~bl_rollover && ~bl_bridge)
|
| | 423 | if (bl_default)
|
| | 424 | % if default: only today u(cmin), transition out next period, debt wiped out
|
| | 425 | ar_opti_aprime_z(ar_coh_neg_idx) = fl_default_aprime;
|
| | 426 | else
|
| | 427 | % if default is not allowed: v = fl_nan_replace
|
| | 428 | ar_opti_aprime_z(ar_coh_neg_idx) = ar_a(fl_nan_replace);
|
| | 429 | end
|
| | 430 | end
|
| | 431 |
|
| | 432 | %% Store Optimal Choices Current Iteration
|
0.046 | 9150 | 433 | mt_val(:,it_z_i) = ar_opti_val_z;
|
0.028 | 9150 | 434 | mt_pol_a(:,it_z_i) = ar_opti_aprime_z;
|
0.018 | 9150 | 435 | mt_pol_cons(:,it_z_i) = ar_opti_c_z;
|
| | 436 |
|
0.003 | 9150 | 437 | if (it_iter == (it_maxiter_val + 1))
|
< 0.001 | 75 | 438 | mt_pol_idx(:,it_z_i) = ar_opti_idx_z;
|
< 0.001 | 75 | 439 | end
|
0.005 | 9150 | 440 | end
|
| | 441 |
|
| | 442 | %% Check Tolerance and Continuation
|
| | 443 |
|
| | 444 | % Difference across iterations
|
0.142 | 122 | 445 | ar_val_diff_norm(it_iter) = norm(mt_val - mt_val_cur);
|
0.133 | 122 | 446 | ar_pol_diff_norm(it_iter) = norm(mt_pol_a - mt_pol_a_cur);
|
0.013 | 122 | 447 | mt_pol_perc_change(it_iter, :) = sum((mt_pol_a ~= mt_pol_a_cur))/(it_a_n);
|
| | 448 |
|
| | 449 | % Update
|
0.006 | 122 | 450 | mt_val_cur = mt_val;
|
0.004 | 122 | 451 | mt_pol_a_cur = mt_pol_a;
|
| | 452 |
|
| | 453 | % Print Iteration Results
|
< 0.001 | 122 | 454 | if (bl_display && (rem(it_iter, it_display_every)==0))
|
| | 455 | fprintf('VAL it_iter:%d, fl_diff:%d, fl_diff_pol:%d\n', ...
|
| | 456 | it_iter, ar_val_diff_norm(it_iter), ar_pol_diff_norm(it_iter));
|
| | 457 | tb_valpol_iter = array2table([mean(mt_val_cur,1); mean(mt_pol_a_cur,1); ...
|
| | 458 | mt_val_cur(it_a_n,:); mt_pol_a_cur(it_a_n,:)]);
|
| | 459 | tb_valpol_iter.Properties.VariableNames = strcat('z', string((1:size(mt_val_cur,2))));
|
| | 460 | tb_valpol_iter.Properties.RowNames = {'mval', 'map', 'Hval', 'Hap'};
|
| | 461 | disp('mval = mean(mt_val_cur,1), average value over a')
|
| | 462 | disp('map = mean(mt_pol_a_cur,1), average choice over a')
|
| | 463 | disp('Hval = mt_val_cur(it_a_n,:), highest a state val')
|
| | 464 | disp('Hap = mt_pol_a_cur(it_a_n,:), highest a state choice')
|
| | 465 | disp(tb_valpol_iter);
|
| | 466 | end
|
| | 467 |
|
| | 468 | % Continuation Conditions:
|
| | 469 | % 1. if value function convergence criteria reached
|
| | 470 | % 2. if policy function variation over iterations is less than
|
| | 471 | % threshold
|
< 0.001 | 122 | 472 | if (it_iter == (it_maxiter_val + 1))
|
< 0.001 | 1 | 473 | bl_vfi_continue = false;
|
0.002 | 121 | 474 | elseif ((it_iter == it_maxiter_val) || ...
|
| 121 | 475 | (ar_val_diff_norm(it_iter) < fl_tol_val) || ...
|
| 121 | 476 | (sum(ar_pol_diff_norm(max(1, it_iter-it_tol_pol_nochange):it_iter)) < fl_tol_pol))
|
| | 477 | % Fix to max, run again to save results if needed
|
< 0.001 | 1 | 478 | it_iter_last = it_iter;
|
< 0.001 | 1 | 479 | it_iter = it_maxiter_val;
|
< 0.001 | 1 | 480 | end
|
| | 481 |
|
< 0.001 | 122 | 482 | end
|
| | 483 |
|
| | 484 | % End Timer
|
< 0.001 | 1 | 485 | if (bl_time)
|
< 0.001 | 1 | 486 | toc;
|
< 0.001 | 1 | 487 | end
|
| | 488 |
|
| | 489 | % End Profile
|
< 0.001 | 1 | 490 | if (bl_profile)
|
0.004 | 1 | 491 | profile off
|
| | 492 | profile viewer
|
| | 493 | st_file_name = [st_profile_prefix st_profile_name_main st_profile_suffix];
|
| | 494 | profsave(profile('info'), strcat(st_profile_path, st_file_name));
|
| | 495 | end
|
| | 496 |
|
| | 497 | %% Process Optimal Choices
|
| | 498 |
|
| | 499 | result_map = containers.Map('KeyType','char', 'ValueType','any');
|
| | 500 | result_map('mt_val') = mt_val;
|
| | 501 | result_map('mt_pol_idx') = mt_pol_idx;
|
| | 502 |
|
| | 503 | % Find optimal Formal Informal Choices. Could have saved earlier, but was
|
| | 504 | % wasteful of resources
|
| | 505 | for it_z_i = 1:it_z_n
|
| | 506 | for it_a_j = 1:it_a_n
|
| | 507 | fl_z_r_borr = ar_z_r_infbr_mesh_wage(it_z_i);
|
| | 508 | fl_z_wage = ar_z_wage_mesh_r_infbr(it_z_i);
|
| | 509 |
|
| | 510 | fl_a = ar_a(it_a_j);
|
| | 511 | fl_coh = f_coh(fl_z_wage, fl_a);
|
| | 512 | fl_a_opti = mt_pol_a(it_a_j, it_z_i);
|
| | 513 |
|
| | 514 | param_map('fl_r_inf') = fl_z_r_borr;
|
| | 515 |
|
| | 516 | % call formal and informal function.
|
| | 517 | [~, fl_opti_b_bridge, fl_opti_inf_borr_nobridge, fl_opti_for_borr, fl_opti_for_save] = ...
|
| | 518 | ffs_fibs_min_c_cost_bridge(fl_a_opti, fl_coh, ...
|
| | 519 | param_map, support_map, armt_map, func_map, bl_input_override);
|
| | 520 |
|
| | 521 | % store savings and borrowing formal and inf optimal choices
|
| | 522 | mt_pol_b_bridge(it_a_j,it_z_i) = fl_opti_b_bridge;
|
| | 523 | mt_pol_inf_borr_nobridge(it_a_j,it_z_i) = fl_opti_inf_borr_nobridge;
|
| | 524 | mt_pol_for_borr(it_a_j,it_z_i) = fl_opti_for_borr;
|
| | 525 | mt_pol_for_save(it_a_j,it_z_i) = fl_opti_for_save;
|
| | 526 |
|
| | 527 | end
|
| | 528 | end
|
| | 529 |
|
| | 530 | result_map('cl_mt_pol_a') = {mt_pol_a, zeros(1)};
|
| | 531 | result_map('cl_mt_coh') = {f_coh(ar_z_r_infbr_mesh_wage, ar_a'), zeros(1)};
|
| | 532 |
|
| | 533 | result_map('cl_mt_pol_c') = {mt_pol_cons, zeros(1)};
|
| | 534 | result_map('cl_mt_pol_b_bridge') = {mt_pol_b_bridge, zeros(1)};
|
| | 535 | result_map('cl_mt_pol_inf_borr_nobridge') = {mt_pol_inf_borr_nobridge, zeros(1)};
|
| | 536 | result_map('cl_mt_pol_for_borr') = {mt_pol_for_borr, zeros(1)};
|
| | 537 | result_map('cl_mt_pol_for_save') = {mt_pol_for_save, zeros(1)};
|
| | 538 |
|
| | 539 | result_map('ar_st_pol_names') = ["cl_mt_pol_a", "cl_mt_pol_coh", "cl_mt_pol_c", ...
|
| | 540 | "cl_mt_pol_b_bridge", "cl_mt_pol_inf_borr_nobridge", "cl_mt_pol_for_borr", "cl_mt_pol_for_save"];
|
| | 541 |
|
| | 542 | % Get Discrete Choice Outcomes
|
| | 543 | result_map = ffs_fibs_identify_discrete(result_map, bl_input_override);
|
| | 544 |
|
| | 545 | %% Post Solution Graph and Table Generation
|
| | 546 | % Note in comparison with *abzr*, results here, even when using identical
|
| | 547 | % parameters would differ because in *abzr* solved where choices are
|
| | 548 | % principle. Here choices are principle + interests in order to facilitate
|
| | 549 | % using the informal choice functions.
|
| | 550 | %
|
| | 551 | % Note that this means two things are
|
| | 552 | % different, on the one hand, the value of asset for to coh is different
|
| | 553 | % based on the grid of assets. If the asset grid is negative, now per grid
|
| | 554 | % point, there is more coh because that grid point of asset no longer has
|
| | 555 | % interest rates. On the other hand, if one has positive asset grid point
|
| | 556 | % on arrival, that is worth less to coh. Additionally, when making choices
|
| | 557 | % for the next period, now choices aprime includes interests. What these
|
| | 558 | % mean is that the a grid no longer has the same meaning. We should expect
|
| | 559 | % at higher savings levels, for the same grid points, if optimal grid
|
| | 560 | % choices are the same as before, consumption should be lower when b
|
| | 561 | % includes interest rates and principle. This is however, not true when
|
| | 562 | % arriving in a period with negative a levels, for the same negative a
|
| | 563 | % level and same a prime negative choice, could have higher consumption
|
| | 564 | % here becasue have to pay less interests on debt. This tends to happen for
|
| | 565 | % smaller levels of borrowing choices.
|
| | 566 | %
|
| | 567 | % Graphically, when using interest + principle, big difference in
|
| | 568 | % consumption as a fraction of (coh - aprime) figure. In those figures,
|
| | 569 | % when counting in principles only, the gap in coh and aprime is
|
| | 570 | % consumption, but now, as more is borrowed only a small fraction of coh
|
| | 571 | % and aprime gap is consumption, becuase aprime/(1+r) is put into
|
| | 572 | % consumption.
|
| | 573 |
|
| | 574 | if (bl_post)
|
| | 575 | result_map('ar_val_diff_norm') = ar_val_diff_norm(1:it_iter_last);
|
| | 576 | result_map('ar_pol_diff_norm') = ar_pol_diff_norm(1:it_iter_last);
|
| | 577 | result_map('mt_pol_perc_change') = mt_pol_perc_change(1:it_iter_last, :);
|
| | 578 |
|
| | 579 | % Standard AZ graphs
|
| | 580 | result_map = ff_az_vf_post(param_map, support_map, armt_map, func_map, result_map);
|
| | 581 |
|
| | 582 | % Graphs for results_map with FIBS contents
|
| | 583 | result_map = ff_az_fibs_vf_post(param_map, support_map, armt_map, func_map, result_map);
|
| | 584 | end
|
| | 585 |
|
| | 586 | %% Display Various Containers
|
| | 587 | if (bl_display_defparam)
|
| | 588 |
|
| | 589 | %% Display 1 support_map
|
| | 590 | fft_container_map_display(support_map);
|
| | 591 |
|
| | 592 | %% Display 2 armt_map
|
| | 593 | fft_container_map_display(armt_map);
|
| | 594 |
|
| | 595 | %% Display 3 param_map
|
| | 596 | fft_container_map_display(param_map);
|
| | 597 |
|
| | 598 | %% Display 4 func_map
|
| | 599 | fft_container_map_display(func_map);
|
| | 600 |
|
| | 601 | %% Display 5 result_map
|
| | 602 | fft_container_map_display(result_map);
|
| | 603 |
|
| | 604 | end
|
| | 605 |
|
| | 606 | end
|
Other subfunctions in this file are not included in this listing.