def __init__(self,
*,
base,
targ_mode,
compare=None,
setup=None,
base_name='Model',
compare_name='Data',
graphs_title_add=None,
moments_aux=None,
noplot=False):
if setup is None: self.setup = ModelSetup()
if type(base) is str:
self.moments = filer(base, 0, 0, repeat=False)
else:
self.moments = base
self.targ_mode = targ_mode
self.base_name = base_name
self.compare_name = compare_name
if graphs_title_add:
self.graphs_title_add = '\n ' + graphs_title_add
else:
self.graphs_title_add = ''
if type(compare) is str:
targ_load = filer(compare, 0, 0, repeat=False)
# emulating targets
self.targets = {key: (targ_load[key], 0.0) for key in targ_load}
else:
self.targets = all_targets(targ_mode)
if not noplot:
try:
self.print_things()
except:
print('failed to print')
self.moments_aux = moments_aux
if not noplot: self.plot()
def __init__(self,Mlist,*args,use_pickles=False,**kwargs):
if type(Mlist) is not list:
Mlist = [Mlist]
if not use_pickles:
Agents.__init__(self,Mlist,*args,nosim=True,**kwargs)
else:
try:
pmeet_exo = filer('pmeet_exo.pkl',0,0,repeat=False)
ppreg_exo = filer('ppreg_exo.pkl',0,0,repeat=False)
Agents.__init__(self,Mlist,*args,pmeet_exo=pmeet_exo,\
ppreg_exo=ppreg_exo,nosim=True,**kwargs)
except:
print('no available pickles found!')
mdl = Mlist
assert type(mdl) is list
pmeet_exo = np.array(mdl[0].setup.pars['pmeet_t'][:mdl[0].setup.pars['Tmeet']])
ppreg_exo = np.array([mdl[0].setup.upp_precomputed_fem[t][3] for t in range(mdl[0].setup.pars['Tmeet'])])
Agents.__init__(self,Mlist,*args,pmeet_exo=pmeet_exo,\
ppreg_exo=ppreg_exo,nosim=True,**kwargs)
self.define_targets()
self.run_sim()
self.get_shares()
if use_pickles:
pmeet_exo = self.pmeet_exo.copy()
ppreg_exo = self.ppreg_exo.copy()
filer('pmeet_exo.pkl',pmeet_exo,True)
filer('ppreg_exo.pkl',ppreg_exo,True)
print('saved pickles')
def get_point(high_e,read_wisdom=False):
if high_e:
x = {'sigma_psi': 0.02351795998155682, 'sigma_psi_init': 0.06319749482803798, 'pmeet_21': 0.36231065032307264, 'pmeet_30': 1.2185173549427066, 'pmeet_40': 1.0, 'preg_21': 0.006683178914886262, 'preg_28': 0.030446109898106186, 'preg_35': 0.026257274180916695, 'u_shift_mar': 0.24749413577501758, 'util_alp': 0.5739290298121097, 'util_kap': 8.484242183675391, 'util_qbar': 5.561954628246849, 'disutil_marry_sm_mal': 4.408415907989258, 'disutil_shotgun': 0.4715857724582857, 'abortion_costs': 1.8092065614536414, 'p_abortion_access': 0.9512267376733684, 'u_lost_divorce': 0.8892578980806901, 'mu_psi_init': -0.15876965206098093,
'high education': True}
if read_wisdom:
try:
print('read wisdom from file!')
o = filer('wisdom.pkl',0,0,repeat=False)[0]
x = calibration_params()[-1](o[1])
print(x)
print('saved distance is {}'.format(o[0]))
x.update({'high education': True})
except:
print('failed to read from wisdom file')
targ_mode = 'high education'
else:
x = {'sigma_psi': 0.04119975516565719, 'sigma_psi_init': 0.07184509981781, 'pmeet_21': 0.7300641341551373, 'pmeet_30': 0.38552526708748397, 'pmeet_40': 1.4132304041226518, 'preg_21': 0.1029100967053943, 'preg_28': 0.11241132276639117, 'preg_35': 0.11203564468462099, 'u_shift_mar': 0.338428482678413, 'util_alp': 0.5195282434982275, 'util_kap': 7.152398760885778, 'util_qbar': 0.0, 'disutil_marry_sm_mal': 3.18966037249299, 'disutil_shotgun': 0.3647670950676456, 'abortion_costs': 0.2962878054482049, 'p_abortion_access': 0.6662167114665236, 'u_lost_divorce': 0.5275074834332285, 'mu_psi_init': -0.24342175587968384,
'high education': False}
targ_mode = 'low education'
if read_wisdom:
try:
print('read wisdom from file!')
o = filer('wisdom.pkl',0,0,repeat=False)[0]
x = calibration_params()[-1](o[1])
print(x)
print('saved distance is {}'.format(o[0]))
x.update({'high education': False})
except:
print('failed to read from wisdom file')
return x, targ_mode
def produce_cf_table(educ='col'):
names = [('BL','baseline'),
('NS','no social stigma'),
('FC','costless abortion'),
('NR','no remar penalty'),
#('SD','no skills depreciation'),
('FD','no divorce costs'),
('ND','infinite divorce costs'),
#('PG','no pay gap')
]
def file_name(nm): return '{} {}.pkl'.format(educ,nm)
entries = [('divorced in 10 years, kids-first','divorced by years after marriage if kids first, 10'),
('divorced in 10 years, marriage-first','divorced by years after marriage if marriage first, 10'),
('kids-first at 30','k then m in sample at 30'),
('unplanned pregnancies aborted','unplanned pregnancies aborted'),
('single mothers at 35','single mothers among mothers at 35')]
table_cols = list()
for ename, nm in names:
fname = file_name(nm)
mom = filer(fname,0,0)
table_col = [r'\textbf{' + ename + r'}'] + [mom[e] for _, e in entries]
table_cols.append(table_col)
table_left = [r'\textbf{Experiment}'] + [r'\textit{' + e + r'}' for e, _ in entries]
# build table rows
table_rows = list()
for irow in range(len(entries)+1):
table_row = table_left[irow]
for icol in range(len(names)):
try:
table_row += ' & ' + '{:02.1f}'.format(100*table_cols[icol][irow])
except ValueError:
table_row += ' & ' + '{}'.format(table_cols[icol][irow])
table_row += r' \\'
table_rows.append(table_row)
[print(r) for r in table_rows]
def __init__(self,nogrid=False,divorce_costs_k='Default',divorce_costs_nk='Default',**kwargs):
p = dict()
#age_begin = 21
#age_data = 23 # age at which the data start being available
#age_death = 76
#age_retire = 65
#age_fertile = 41
T = 55
Tret = 45 # first period when the agent is retired
Tfert = 20 # first peroid when infertile
Tdiv = 44 # first period when cannot divorce / renegotiate
Tmeet = 43 # first period when stop meeting partners
Tinc = 25 # first period where stop tracking income process and assume it to be fixed
p['T'] = T
p['Tret'] = Tret
p['Tfert'] = Tfert
p['Tsim'] = T
p['Tmeet'] = Tmeet
p['n_zf_t'] = [7]*Tret + [1]*(T-Tret)
p['n_zm_t'] = [5]*Tret + [1]*(T-Tret)
p['sigma_psi_init'] = 0.28
p['mu_psi_init'] = 0.0
p['sigma_psi'] = 0.11
p['R_t'] = [1/0.96]*T
p['n_psi'] = 17
p['beta_t'] = [0.96]*T
p['A'] = 1.0 # consumption in couple: c = (1/A)*[c_f^(1+rho) + c_m^(1+rho)]^(1/(1+rho))
p['crra_power'] = 1.5
p['couple_rts'] = 0.23
p['sig_partner_a'] = 0.1
p['mu_partner_a_female'] = 0.00
p['mu_partner_a_male'] = -0.00
p['dump_factor_z'] = 0.75
p['sig_partner_z'] = 0.25
p['mu_partner_z_male'] = -0.02
p['mu_partner_z_female'] = 0.02
p['m_bargaining_weight'] = 0.5
p['pmeet_21'] = 0.1
p['pmeet_30'] = 0.2
p['pmeet_40'] = 0.1
p['pmeet_pre25'] = None
p['ppreg_pre25'] = None
p['pmeet_exo'] = None
p['ppreg_exo'] = None
p['m_zf'] = 1.0
p['m_zf0'] = 1.0
p['no kids at meeting'] = True
p['high education'] = True # what trend to pick
p['any kids'] = True
p['z_drift'] = -0.09 if p['high education'] else -0.06
p['wret'] = 0.8
p['uls'] = 0.2
p['pls'] = 1.0
p['income_sd_mult'] = 1.0
p['pay_gap'] = True
p['preg_mult'] = 1.0
p['u_shift_mar'] = 0.0
p['u_shift_coh'] = 0.0
p['sm_shift'] = 0.0
p['disutil_marry_sm_fem'] = 0.0
p['disutil_marry_sm_mal'] = 10.0
p['disutil_shotgun'] = 2.0
p['pmeet_multiplier_fem'] = 1.0
p['p_to_meet_sm_if_mal'] = 0.1
p['taste_shock_mult'] = 1.0
p['p_abortion_access'] = 0.5
p['abortion_costs'] = 10.0
p['u_lost_divorce'] = 0.0
p['child_a_cost'] = 0.0
p['child_support_share'] = 0.2
p['child_support_awarded_nm'] = 0.284
p['child_support_awarded_div'] = 0.461
p['util_lam'] = 0.7
p['util_alp'] = 0.5
p['util_xi'] = 1.5
p['util_kap'] = 0.5
p['util_qbar'] = 0.0
p['util_out_lf'] = 0.0
p['ppreg_sim_mult'] = 1.0
p['tax_childless_couples'] = True
p['tax_couples_woth_children'] = True
p['tax_single_mothers'] = True
p['preg_21'] = 0.01
p['preg_28'] = 0.5
p['preg_35'] = 0.3
for key, value in kwargs.items():
assert (key in p), 'wrong name?'
p[key] = value
if p['high education']:
p['sig_zm'] = p['income_sd_mult']*0.16138593
p['sig_zm_0'] = p['income_sd_mult']*0.41966813
p['sig_zf'] = p['income_sd_mult']*p['m_zf']*0.19571624
p['sig_zf_0'] = p['income_sd_mult']*p['m_zf0']*0.43351219
else:
p['sig_zm'] = p['income_sd_mult']*0.17195085
p['sig_zm_0'] = p['income_sd_mult']*0.2268650
p['sig_zf'] = p['income_sd_mult']*p['m_zf']*0.1762148
p['sig_zf_0'] = p['income_sd_mult']*p['m_zf0']*0.1762148
p['sm_init'] = (0.02 if p['high education'] else 0.25) if p['any kids'] else 0.0 # initial share of single moms
# college
if p['high education']:
m_trend_data = [3.3899509,3.5031169,3.6395543,3.7025571,3.7646729,3.8168057,3.8706297,3.9168266,3.9628897,3.999365,4.0401083,4.0584735,4.114779,4.1344154,4.1570761,4.1703323,4.189479,4.1908761,4.2096766,4.2190858,4.2284456,4.2330354,4.2364118,4.2443216,4.2449467]
f_trend_data = [3.0796507,3.1468105,3.2911468,3.4024807,3.5485041,3.6067711,3.662677,3.7131363,3.7598972,3.794152,3.8221141,3.8500988,3.8791148,3.8860543,3.9176438,3.9332115,3.9479731,3.9537252,3.9578583,3.9480597,3.9559513,3.9562479,3.9585957,3.9576563,3.9589798]
nm = len(m_trend_data)-1
nf = len(f_trend_data)-1
p['m_wage_trend'] = np.array(
[m_trend_data[min(t,nm)]
for t in range(T)]
)
p['f_wage_trend'] = np.array(
[f_trend_data[min(t,nf)]
for t in range(T)]
)
else:
# no college
m_trend_data = [3.091856,3.130104,3.2259613,3.2581962,3.2772099,3.2999002,3.3206571,3.3314928,3.3573047,3.3663062,3.3801406,3.3909449,3.4160915,3.4358479,3.4488938,3.4534575,3.4654005,3.4655065,3.4815268,3.4859583,3.4967845,3.4972438,3.5118738,3.525121,3.5271331]
f_trend_data = [2.9056071,2.9427025,2.9773922,2.999882,3.0932755,3.1129375,3.1199322,3.1352323,3.1498192,3.1606338,3.168912,3.1722982,3.1775691,3.1831384,3.2040837,3.1997439,3.2122542,3.2085543,3.2209876,3.2232882,3.2273824,3.2336534,3.233879,3.244275,3.2527455]
nm = len(m_trend_data)-1
nf = len(f_trend_data)-1
p['m_wage_trend'] = np.array(
[m_trend_data[min(t,nm)]
for t in range(T)]
)
p['f_wage_trend'] = np.array(
[f_trend_data[min(t,nf)]
for t in range(T)]
)
if not p['pay_gap']:
p['sig_zf'], p['sig_zf_0'] = p['sig_zm'], p['sig_zm_0']
p['f_wage_trend'] = p['m_wage_trend']
p['preg_az'] = 0.00
p['preg_azt'] = 0.00
#Get the probability of meeting, adjusting for year-period
p['taste_shock'] = 0.0 #*p['taste_shock_mult']*0.0#p['sigma_psi']
p['is fertile'] = [p['any kids']]*Tfert + [False]*(T-Tfert)
p['can divorce'] = [True]*Tdiv + [False]*(T-Tdiv)
#p['poutsm_t'] = [p['poutsm']]*T
p['pmeet_0'], p['pmeet_t'], p['pmeet_t2'] = prob_polyfit(
(p['pmeet_21'],0),(p['pmeet_30'],9),(p['pmeet_40'],19),
max_power=2)
p['preg_a0'], p['preg_at'], p['preg_at2'] = prob_polyfit(
(p['preg_21'],0),(p['preg_28'],7),(p['preg_35'],14),
max_power=2)
if p['pmeet_exo'] is None:
p['pmeet_t'] = [np.clip(p['pmeet_0'] + t*p['pmeet_t'] + (t**2)*p['pmeet_t2'],0.0,1.0) for t in range(20)] + \
[p['pmeet_40']]*(Tmeet - 20) + [0.0]*(T-Tmeet)
p['pmeet_t'] = np.array(p['pmeet_t'])
if p['pmeet_pre25'] is not None: p['pmeet_t'][:4] = p['pmeet_pre25']
else:
p['pmeet_t'] = [p['pmeet_exo'][min(t,p['pmeet_exo'].size-1)] for t in range(Tmeet)] + [0.0]*(T-Tmeet)
p['pmeet_t'] = np.array(p['pmeet_t'])
p['n_psi_t'] = [p['n_psi']]*T
p['psi_clip'] = 8.5*p['sigma_psi_init']
p['fert_prob_t'] = [0.86*(t<=3) + 0.78*(t>3 and t<=8) +
0.63*(t>=9 and t<=13) + 0.52*(t>=14) for
t in range(T)]
#p['fert_prob_t'] = [1.0]*T
self.pars = p
self.dtype = np.float64 # type for all floats
# relevant for integration
self.state_names = ['Female, single','Male, single','Female and child','Couple, no children','Couple and child']
# female labor supply
lmin = 0.2
lmax = 1.0
nl = 2
ls = np.array([0.2,1.0]) #np.linspace(lmin,lmax,nl,dtype=self.dtype)
ps = np.array([p['pls'],0.0])
ls_ushift = np.array([p['util_out_lf'],0.0])
self.ls_levels = dict()
self.ls_levels['Couple, no children'] = np.array([1.0],dtype=self.dtype)
self.ls_levels['Female, single'] = np.array([1.0],dtype=self.dtype)
self.ls_levels['Male, single'] = np.array([1.0],dtype=self.dtype)
self.ls_levels['Couple and child'] = ls
self.ls_levels['Female and child'] = ls
self.ls_ushift = dict()
self.ls_ushift['Couple, no children'] = np.array([0.0],dtype=self.dtype)
self.ls_ushift['Female, single'] = np.array([0.0],dtype=self.dtype)
self.ls_ushift['Male, single'] = np.array([0.0],dtype=self.dtype)
self.ls_ushift['Couple and child'] = ls_ushift
self.ls_ushift['Female and child'] = ls_ushift
#self.ls_utilities = np.array([p['uls'],0.0],dtype=self.dtype)
self.ls_pdown = dict()
self.ls_pdown['Couple, no children'] = np.array([0.0],dtype=self.dtype)
self.ls_pdown['Female, single'] = np.array([0.0],dtype=self.dtype)
self.ls_pdown['Male, single'] = np.array([0.0],dtype=self.dtype)
self.ls_pdown['Female and child'] = ps
self.ls_pdown['Couple and child'] = ps
self.nls = dict()
self.nls['Couple and child'] = len(self.ls_levels['Couple and child'])
self.nls['Couple, no children'] = len(self.ls_levels['Couple, no children'])
self.nls['Female and child'] = len(self.ls_levels['Female and child'])
self.nls['Female, single'] = len(self.ls_levels['Female, single'])
self.nls['Male, single'] = len(self.ls_levels['Male, single'])
#Cost of Divorce
if divorce_costs_k == 'Default':
# by default the costs are set in the bottom
self.divorce_costs_k = DivorceCosts(u_lost_m=self.pars['u_lost_divorce'],
u_lost_f=self.pars['u_lost_divorce'])
else:
if isinstance(divorce_costs_k,dict):
# you can feed in arguments to DivorceCosts
self.divorce_costs_k = DivorceCosts(**divorce_costs_k)
else:
# or just the output of DivorceCosts
assert isinstance(divorce_costs_k,DivorceCosts)
self.divorce_costs_k = divorce_costs_k
#Cost of Separation
if divorce_costs_nk == 'Default':
# by default the costs are set in the bottom
self.divorce_costs_nk = DivorceCosts(u_lost_m=self.pars['u_lost_divorce'],
u_lost_f=self.pars['u_lost_divorce'])
else:
if isinstance(divorce_costs_nk,dict):
# you can feed in arguments to DivorceCosts
self.divorce_costs_nk = DivorceCosts(**divorce_costs_nk)
else:
# or just the output of DivorceCosts
assert isinstance(divorce_costs_nk,DivorceCosts)
self.divorce_costs_nk = divorce_costs_nk
# exogrid should be deprecated
if not nogrid:
exogrid = dict()
# let's approximate three Markov chains
# this sets up exogenous grid
# FIXME: this uses number of points from 0th entry.
# in principle we can generalize this
exogrid['zf_t'], exogrid['zf_t_mat'] = rouw_nonst(p['T'],p['sig_zf'],p['sig_zf_0'],p['n_zf_t'][0])
exogrid['zm_t'], exogrid['zm_t_mat'] = rouw_nonst(p['T'],p['sig_zm'],p['sig_zm_0'],p['n_zm_t'][0])
for t in range(Tinc,Tret):
for key in ['zf_t','zf_t_mat','zm_t','zm_t_mat']:
exogrid[key][t] = exogrid[key][Tinc]
for t in range(Tret,T):
exogrid['zf_t'][t] = np.array([np.log(p['wret'])])
exogrid['zm_t'][t] = np.array([np.log(p['wret'])])
exogrid['zf_t_mat'][t] = np.atleast_2d(1.0)
exogrid['zm_t_mat'][t] = np.atleast_2d(1.0)
# fix transition from non-retired to retired
exogrid['zf_t_mat'][Tret-1] = np.ones((p['n_zf_t'][Tret-1],1))
exogrid['zm_t_mat'][Tret-1] = np.ones((p['n_zm_t'][Tret-1],1))
exogrid['psi_t'], exogrid['psi_t_mat'] = rouw_nonst(p['T'],p['sigma_psi'],p['sigma_psi_init'],p['n_psi_t'][0])
#exogrid['psi_t'], exogrid['psi_t_mat'] = tauchen_nonst(p['T'],p['sigma_psi'],p['sigma_psi_init'],p['n_psi_t'][0],nsd=2.5,fix_0=False)
#assert False
zfzm, zfzmmat = combine_matrices_two_lists(exogrid['zf_t'], exogrid['zm_t'], exogrid['zf_t_mat'], exogrid['zm_t_mat'])
all_t, all_t_mat = combine_matrices_two_lists(zfzm,exogrid['psi_t'],zfzmmat,exogrid['psi_t_mat'])
all_t_mat_sparse_T = [sparse.csc_matrix(D.T) if D is not None else None for D in all_t_mat]
#Create a new bad version of transition matrix p(zf_t)
zf_bad = [tauchen_drift(exogrid['zf_t'][t], exogrid['zf_t'][t+1],
1.0, p['sig_zf'], p['z_drift'])
for t in range(self.pars['T']-1) ] + [None]
#zf_bad = [cut_matrix(exogrid['zf_t_mat'][t]) if t < Tret -1
# else (exogrid['zf_t_mat'][t] if t < T - 1 else None)
# for t in range(self.pars['T'])]
zf_t_mat_down = zf_bad
zfzm, zfzmmat = combine_matrices_two_lists(exogrid['zf_t'], exogrid['zm_t'], zf_t_mat_down, exogrid['zm_t_mat'])
all_t_down, all_t_mat_down = combine_matrices_two_lists(zfzm,exogrid['psi_t'],zfzmmat,exogrid['psi_t_mat'])
all_t_mat_down_sparse_T = [sparse.csc_matrix(D.T) if D is not None else None for D in all_t_mat_down]
all_t_mat_by_l_nk = [ [(1-p)*m + p*md if m is not None else None
for m , md in zip(all_t_mat,all_t_mat_down)]
for p in self.ls_pdown['Couple, no children'] ]
all_t_mat_by_l_spt_nk = [ [(1-p)*m + p*md if m is not None else None
for m, md in zip(all_t_mat_sparse_T,all_t_mat_down_sparse_T)]
for p in self.ls_pdown['Couple, no children'] ]
all_t_mat_by_l_k = [ [(1-p)*m + p*md if m is not None else None
for m , md in zip(all_t_mat,all_t_mat_down)]
for p in self.ls_pdown['Couple and child'] ]
all_t_mat_by_l_spt_k = [ [(1-p)*m + p*md if m is not None else None
for m, md in zip(all_t_mat_sparse_T,all_t_mat_down_sparse_T)]
for p in self.ls_pdown['Couple and child'] ]
zf_t_mat_by_l_sk = [ [(1-p)*m + p*md if md is not None else None
for m , md in zip(exogrid['zf_t_mat'],zf_bad)]
for p in self.ls_pdown['Female and child'] ]
exogrid['all_t_mat_by_l_nk'] = all_t_mat_by_l_nk
exogrid['all_t_mat_by_l_spt_nk'] = all_t_mat_by_l_spt_nk
exogrid['all_t_mat_by_l_k'] = all_t_mat_by_l_k
exogrid['all_t_mat_by_l_spt_k'] = all_t_mat_by_l_spt_k
exogrid['zf_t_mat_by_l_sk'] = zf_t_mat_by_l_sk
exogrid['all_t'] = all_t
Exogrid_nt = namedtuple('Exogrid_nt',exogrid.keys())
self.exogrid = Exogrid_nt(**exogrid)
self.pars['nexo_t'] = [v.shape[0] for v in all_t]
self.compute_child_support_transitions(child_support_share=p['child_support_share'])
#assert False
#Grid Couple
self.na = 40
self.amin = 0
self.amax = 1000.0
self.agrid_c = np.linspace(self.amin**0.5,self.amax**0.5,self.na,dtype=self.dtype)**2
#tune=1.5
#self.agrid_c = np.geomspace(self.amin+tune,self.amax+tune,num=self.na)-tune
# this builds finer grid for potential savings
s_between = 7 # default numer of points between poitns on agrid
s_da_min = 0.2 # minimal step (does not create more points)
s_da_max = 10.0 # maximal step (creates more if not enough)
self.sgrid_c = build_s_grid(self.agrid_c,s_between,s_da_min,s_da_max)
self.vsgrid_c = VecOnGrid(self.agrid_c,self.sgrid_c)
#Grid Single
self.amin_s = 0
self.amax_s = self.amax/2.0
self.agrid_s = self.agrid_c/2.0
#tune_s=1.5
#self.agrid_s = np.geomspace(self.amin_s+tune_s,self.amax_s+tune_s,num=self.na)-tune_s
self.sgrid_s = build_s_grid(self.agrid_s,s_between,s_da_min,s_da_max)
self.vsgrid_s = VecOnGrid(self.agrid_s,self.sgrid_s)
# grid for theta
self.ntheta = 11
self.thetamin = 0.05
self.thetamax = 0.95
self.thetagrid = np.linspace(self.thetamin,self.thetamax,self.ntheta,dtype=self.dtype)
self.child_a_cost_single = np.minimum(self.agrid_s,self.pars['child_a_cost'])
self.child_a_cost_couple = np.minimum(self.agrid_c,self.pars['child_a_cost'])
assert self.pars['child_a_cost']<1e-3, 'not implemented'
#self.vagrid_child_single = VecOnGrid(self.agrid_s, self.agrid_s - self.child_a_cost_single)
#self.vagrid_child_couple = VecOnGrid(self.agrid_c, self.agrid_c - self.child_a_cost_couple)
# construct finer grid for bargaining
ntheta_fine = 10*self.ntheta # actual number may be a bit bigger
self.thetagrid_fine = np.unique(np.concatenate( (self.thetagrid,np.linspace(self.thetamin,self.thetamax,ntheta_fine,dtype=self.dtype)) ))
self.ntheta_fine = self.thetagrid_fine.size
i_orig = list()
for theta in self.thetagrid:
assert theta in self.thetagrid_fine
i_orig.append(np.where(self.thetagrid_fine==theta)[0])
assert len(i_orig) == self.thetagrid.size
# allows to recover original gird points on the fine grid
self.theta_orig_on_fine = np.array(i_orig).flatten()
self.v_thetagrid_fine = VecOnGrid(self.thetagrid,self.thetagrid_fine)
# precomputed object for interpolation
self.exo_grids = {'Female, single':exogrid['zf_t'],
'Male, single':exogrid['zm_t'],
'Female and child':exogrid['zf_t'],
'Couple and child':exogrid['all_t'],
'Couple, no children':exogrid['all_t']}
self.exo_mats = {'Female, single':exogrid['zf_t_mat'],
'Male, single':exogrid['zm_t_mat'],
'Female and child':exogrid['zf_t_mat_by_l_sk'],
'Couple and child':exogrid['all_t_mat_by_l_k'],
'Couple, no children':exogrid['all_t_mat_by_l_nk']} # sparse version?
self.utility_shifters = {'Female, single':0.0,
'Male, single':0.0,
'Female and child':p['u_shift_mar'] + p['sm_shift'],
'Couple and child':p['u_shift_mar'],
'Couple, no children':p['u_shift_coh']}
# this pre-computes transition matrices for meeting a partner
name_fem_pkl = 'az_dist_fem.pkl' if p['high education'] else 'az_dist_fem_noc.pkl'
name_mal_pkl = 'az_dist_mal.pkl' if p['high education'] else 'az_dist_mal_noc.pkl'
name_fem_csv = 'income_assets_distribution_male_col.csv' if p['high education'] else 'income_assets_distribution_male_hs.csv'
name_mal_csv = 'income_assets_distribution_female_col.csv' if p['high education'] else 'income_assets_distribution_female_hs.csv'
# this is not an error, things are switched
try:
self.partners_distribution_fem = filer(name_fem_pkl,0,0,repeat=False)
self.partners_distribution_mal = filer(name_mal_pkl,0,0,repeat=False)
except:
print('recreating estimates...')
est_fem = get_estimates(fname=name_fem_csv,
age_start=23,age_stop=42,
zlist=self.exogrid.zm_t[2:],
female=False,college=p['high education'])
filer(name_fem_pkl,est_fem,True,repeat=False)
self.partners_distribution_fem = est_fem
est_mal = get_estimates(fname=name_mal_csv,
age_start=21,age_stop=40,
zlist=self.exogrid.zf_t[0:],
female=True,college=p['high education'])
filer(name_mal_pkl,est_mal,True,repeat=False)
self.partners_distribution_mal = est_mal
self.build_matches()
# building m grid
ezfmin = min([np.min(np.exp(g+t)) for g,t in zip(exogrid['zf_t'],p['f_wage_trend'])])
ezmmin = min([np.min(np.exp(g+t)) for g,t in zip(exogrid['zm_t'],p['m_wage_trend'])])
ezfmax = max([np.max(np.exp(g+t)) for g,t in zip(exogrid['zf_t'],p['f_wage_trend'])])
ezmmax = max([np.max(np.exp(g+t)) for g,t in zip(exogrid['zm_t'],p['m_wage_trend'])])
self.money_min = 0.95*min(self.ls_levels['Female and child'])*min(ezmmin,ezfmin) # cause FLS can be up to 0
mmin = self.money_min
mmax = ezfmax + ezmmax + np.max(self.pars['R_t'])*self.amax
mint = (ezfmax + ezmmax) # poin where more dense grid begins
ndense = 1900
nm = 3500
gsparse = np.linspace(mint,mmax,nm-ndense)
gdense = np.linspace(mmin,mint,ndense+1) # +1 as there is a common pt
self.mgrid = np.zeros(nm,dtype=self.dtype)
self.mgrid[ndense:] = gsparse
self.mgrid[:(ndense+1)] = gdense
self.mgrid_c = self.mgrid
self.mgrid_s = self.mgrid
assert np.all(np.diff(self.mgrid)>0)
self.u_precompute()
self.unplanned_pregnancy_probability()
self.compute_taxes()
self.cupyfy()
def fun(x):
assert type(x) is tuple, 'x must be a tuple!'
action = x[0]
args = x[1]
assert type(action) is str, 'x[0] should be string for action'
assert len(x) <= 2, 'too many things in x! x is (action,agrs)'
if action == 'test':
return mdl_resid()
elif action == 'compute':
return mdl_resid(args)
elif action == 'minimize':
import dfols
import pybobyqa
i, N_st, xfix = args
xl, xu, x0, keys, translator = calibration_params(xfix=xfix)
#Sort lists
def sortFirst(val):
return val[0]
#Get the starting point for local minimization
#Open File with best solution so far
param = filer('wisdom.pkl', 0, False)
param.sort(key=sortFirst)
print('f best so far is {} and x is {}'.format(param[0][0],
param[0][1]))
xm = param[0][1]
#Get right sobol sequence point
xt = filer('sobol.pkl', None, False)
#Determine the initial position
dump = min(max(0.1, ((i + 1) / N_st)**(0.5)), 0.995)
xc = dump * xm + (1 - dump) * xt[:, i]
xc = xc.squeeze()
print('The initial position is {}'.format(xc))
#Standard Way
def q(pt):
try:
ans = mdl_resid(translator(pt),
return_format=['scaled residuals'])[0]
except:
print('During optimization function evaluation failed at {}'.
format(pt))
ans = np.array([1e6])
finally:
gc.collect()
return ans
res = dfols.solve(q,
xc,
rhobeg=0.01,
rhoend=1e-4,
maxfun=100,
bounds=(xl, xu),
scaling_within_bounds=True,
objfun_has_noise=False,
print_progress=True)
#res=pybobyqa.solve(q, xc, rhobeg = 0.001, rhoend=1e-6, maxfun=80, bounds=(xl,xu),
# scaling_within_bounds=True,objfun_has_noise=False,print_progress=True)
print(res)
if res.flag == -1:
raise Exception('solver returned something creepy...')
fbest = mdl_resid(translator(
res.x))[0] # in prnciple, this can be inconsistent with
# squared sum of residuals
print('fbest is {} and res.f is {}'.format(fbest, res.f))
print('Final value is {}'.format(fbest))
param_new = filer('wisdom.pkl', None, False)
param_write = param_new + [(fbest, res.x)]
#Save Updated File
param_write.sort(key=sortFirst)
filer('wisdom.pkl', param_write, True)
return fbest
else:
raise Exception('unsupported action or format')
def mdl_resid(x=None,
targets=None,
weights=w,
save_to=None,
load_from=None,
return_format=['distance'],
store_path=None,
verbose=False,
draw=False,
graphs=False,
rel_diff=False,
cs_moments=False,
moments_repeat=5,
Tsim=30,
moments_save_name=None):
from model import Model
from setup import DivorceCosts
from simulations import Agents
from crosssection import CrossSection
from calibration_params import calibration_params
if type(x) is dict:
kwords = x
if 'targets' in x:
targets = x.pop('targets')
else:
lb, ub, xdef, keys, translator = calibration_params()
if x is None:
x = xdef
kwords = translator(x)
if verbose: print(kwords)
# this is for the default model
#dc_k = DivorceCosts(unilateral_divorce=True,assets_kept = 1.0,u_lost_m=0.00,u_lost_f=0.00,eq_split=1.0)
#dc_nk = DivorceCosts(unilateral_divorce=True,assets_kept = 1.0,u_lost_m=0.00,u_lost_f=0.00,eq_split=1.0)
def join_path(name, path):
return os.path.join(path, name)
if load_from is not None:
if type(load_from) is not list:
load_from = [load_from]
if store_path is not None:
load_from = [join_path(n, store_path) for n in load_from]
if save_to is not None:
if type(save_to) is not list:
save_to = [save_to]
if store_path is not None:
save_to = [join_path(n, store_path) for n in save_to]
if load_from is None:
mdl = Model(verbose=verbose, **kwords)
mdl_list = [mdl]
else:
mdl_list = [dill.load(open(l, 'rb+')) for l in load_from]
mdl = mdl_list[0]
if save_to is not None:
if len(save_to) > 1:
print('warning: too much stuff is save_to')
dill.dump(mdl, open(save_to[0], 'wb+'))
np.random.seed(18)
agents = Agents(mdl_list, verbose=verbose, fix_seed=False, T=Tsim)
if not cs_moments:
moments_list = [agents.compute_moments()] + [
Agents(mdl_list, verbose=False, T=Tsim,
fix_seed=False).compute_moments()
for _ in range(moments_repeat - 1)
]
else:
moments_list = [
CrossSection(mdl_list,
verbose=False,
N_total=30000,
fix_seed=False).compute_moments()
for _ in range(moments_repeat)
]
mom = {
key: np.mean([m[key] for m in moments_list], axis=0)
for key in moments_list[0].keys()
}
#mom_join = Agents( mdl_list, N=10000, T=18, female=False, verbose=False).aux_moments()
#mom_men = agents_extra.compute_moments()
#mom.update(mom_join)
if moments_save_name: # is not None can be ommited
filer('{}.pkl'.format(moments_save_name), mom, True)
if targets is None:
from targets import target_values
tar = target_values()
elif type(targets) is str:
from targets import target_values
tar = target_values(mode=targets)
else:
tar = targets
resid_all, resid_sc, dist = distance_to_targets(mom,
tar,
weights=weights,
report=verbose)
#if verbose:
# print('data moments are {}'.format(dat))
# print('simulated moments are {}'.format(sim))
tt = mdl_list[0].get_total_time()
print('Distance {}, time {}'.format(dist, tt))
out_dict = {
'distance': dist,
'all residuals': resid_all,
'scaled residuals': resid_sc,
'models': mdl_list,
'agents': agents,
'moments': mom
}
out = [out_dict[key] for key in return_format]
del (out_dict)
if 'models' not in return_format:
for m in mdl_list:
del (m)
del mdl_list
if 'agents' not in return_format:
del (agents)
if len(out) == 1: out = out[0]
return out
def run(resume=False, high_e=True):
xinit, targ_mode = get_point(high_e, read_wisdom=False)
tar = target_values(targ_mode)
if resume:
x_load = filer('wisdom_refined.pkl', 0, 0)
prob_meet_load = x_load['pmeet_exo']
prob_preg_load = x_load['ppreg_exo']
xinit = x_load
out, mdl, agents, res, mom = mdl_resid(x=xinit,
targets=tar,
return_format=[
'distance', 'models', 'agents',
'scaled residuals', 'moments'
],
verbose=False)
print('initial distance is {}'.format(out))
if resume:
prob_meet_init = prob_meet_load
prob_preg_init = prob_preg_load
else:
prob_meet_init = np.array(
mdl[0].setup.pars['pmeet_t'][:mdl[0].setup.pars['Tmeet']])
prob_preg_init = np.array([
mdl[0].setup.upp_precomputed_fem[t][3]
for t in range(mdl[0].setup.pars['Tmeet'])
])
nopt = 10
yfactor = 1.5
for iopt in range(nopt):
print('running esimation round {}'.format(iopt))
print('estimating probabilities:')
prob_meet_est = 0.0
prob_preg_est = 0.0
nrep = 4 if iopt > 0 else 1
np.random.seed(12)
for rep in range(nrep):
o = AgentsEst(mdl, T=30, verbose=False, fix_seed=False)
prob_meet_est += (1 / nrep) * o.pmeet_exo.copy()
prob_preg_est += (1 / nrep) * o.ppreg_exo.copy()
print('estimated pmeet = {}'.format(prob_meet_est))
print('estimated ppreg = {}'.format(prob_preg_est))
# this does binary search
w = 1.0
factor = 0.5
ne = prob_meet_est.size
nw = 10
print('reference value is {}'.format(out))
y_previous = out
for i in range(nw):
prob_meet_w = w * prob_meet_est + (1 - w) * prob_meet_init[:ne]
prob_preg_w = w * prob_preg_est + (1 - w) * prob_preg_init[:ne]
xsearch = xinit.copy()
xsearch.update({
'pmeet_exo': prob_meet_w,
'ppreg_exo': prob_preg_w
})
out_w = mdl_resid(x=xsearch,
targets=tar,
return_format=['distance'],
verbose=False)
print('with weight = {}, distance is {}'.format(w, out_w))
if out_w < yfactor * out:
print('found a potentially imporving weight for yfactor = {}'.
format(yfactor))
break
else:
w = factor * w
if i < nw - 1:
print('trying new weight = {}'.format(w))
else:
print('no luck...')
xfix = {
k: xinit[k]
for k in [
'pmeet_21', 'pmeet_30', 'pmeet_40', 'preg_21', 'preg_28',
'preg_35'
]
}
lb, ub, _, keys, translator = calibration_params(xfix=xfix)
def tr(x):
xx = translator(x)
xx.update({'pmeet_exo': prob_meet_w, 'ppreg_exo': prob_preg_w})
return xx
x0 = [xinit[key] for key in keys]
x0, lb, ub = np.array(x0), np.array(lb), np.array(ub)
print('starting from {}'.format(tr(x0)))
tar = target_values('high education')
def q(pt):
#print('computing at point {}'.format(translator(pt)))
try:
ans = mdl_resid(tr(pt), return_format=['scaled residuals'])
except BaseException as a:
print('During optimization function evaluation failed at {}'.
format(pt))
print(a)
ans = np.array([1e6])
finally:
gc.collect()
return ans
res = dfols.solve(q,
x0,
rhobeg=0.02,
rhoend=1e-5,
maxfun=60,
bounds=(lb, ub),
scaling_within_bounds=True,
objfun_has_noise=False,
npt=len(x0) + 5,
user_params={
'restarts.use_restarts': True,
'restarts.rhoend_scale': 0.5,
'restarts.increase_npt': True
})
print(res)
print('Result is {}'.format(tr(res.x)))
filer('wisdom_refined.pkl', tr(res.x), True)
print('wrote to the file!')
xinit = tr(res.x)
out, mdl, agents, res, mom = mdl_resid(x=xinit,
return_format=[
'distance', 'models',
'agents',
'scaled residuals',
'moments'
])
if out > y_previous:
print('no reduction in function value obtained')
yfactor = 0.5 * yfactor + 0.5
y_previous = out
'sm_shift': [
0.1 * sm, 0.25 * sm, 0.5 * sm, 0.75 * sm, sm, 1.25 * sm, 1.5 * sm,
1.75 * sm, 2 * sm
],
'preg_mult': [0.0, 0.25, 0.5, 0.75, 1.0, 1.25, 1.5, 1.75, 2.0],
'poutsm': [0.05, 0.1, 0.2, 1 / 3, 1 / 2, 2 / 3, 0.99],
'z_drift': [-0.4, -0.3, -0.2, -0.15, -0.1, -0.05, -0.025, 0.0]
}
inp = list()
change = list()
for par, vlist in request.items():
for val in vlist:
change.append((par, val))
xin = x.copy()
xin.update({par: val})
inp.append(('moments', xin))
from p_client import compute_for_values
print('generated {} points'.format(len(change)))
result = compute_for_values(inp)
from tiktak import filer
filer('sens_results.py', {
'init': x,
'input': change,
'output': result
}, True)
def run(adj_name, fix, educ_name, resume=False, noplot=False):
# at first this takes point x saved in estimates.py
# by calling get_point(high_e). x is a dict with parameter names and values
# then it applies adjustment fix (a dict) to the point x
# resulting file name is educ_name + adj_name
# adj_name -- name of adjustment
# fix -- dictionary containing parameters of setup.py to change
# educ_name -- name of education group ("col" or "hs")
# high_e -- input to get_point function, True for college, False for HS
high_e = (educ_name == 'col' or educ_name == 'college')
low_e = (educ_name == 'hs' or educ_name == 'high school')
assert (high_e or low_e), 'wrong specifier for education'
x, targ_mode = get_point(high_e, read_wisdom=False)
tar = target_values(targ_mode)
print('\n\n\n\n\n\n')
print('doing {} {}'.format(educ_name, adj_name))
x_new = x.copy()
fix = fix.copy()
if 'multiply' in fix:
mult = fix.pop('multiply')
for m in mult:
x_new[m] *= mult[m]
x_new.update(fix)
print(x_new)
name = '{} {}'.format(educ_name, adj_name)
fname = '{}.pkl'.format(name)
try:
if resume:
try:
mom = filer(fname, 0, 0, repeat=False)
skip = True
except:
skip = False
else:
skip = False
if not skip:
print("computing {}".format(fname))
out, mom = mdl_resid(x=x_new,
targets=tar,
return_format=['distance', 'moments'],
verbose=False,
draw=False,
cs_moments=False,
save_to='mdl for {}'.format(fname),
moments_save_name=name,
moments_repeat=5,
Tsim=42)
print("file {} saved".format(fname))
else:
print("file {} already exists".format(fname))
print('done, doing fit plots')
try:
if adj_name == 'baseline':
fp = FitPlots(targ_mode=targ_mode,
compare=None,
base='{} baseline.pkl'.format(educ_name),
compare_name='Data',
base_name='baseline',
moments_aux=None,
noplot=noplot)
else:
fp = FitPlots(targ_mode=targ_mode,
compare='{} baseline.pkl'.format(educ_name),
base=fname,
compare_name='baseline',
base_name=adj_name,
moments_aux=None,
noplot=noplot)
except:
print('something wrong with fit plots...')
except KeyboardInterrupt:
raise (KeyboardInterrupt)
except BaseException as a:
print("file {} {}.pkl could not be produced. Exception:".format(
name, adj_name))
print(a)
print(' ')
return mom, fp
def get_new_estimates(fname='income_assets_distribution.csv',
save='za_dist.pkl',
age_start=23,age_stop=45,weighted=True):
print('obtaining estimates from file {}'.format(fname))
df = pd.read_csv(fname)
income_for_normalization = np.median(np.exp(df[df['tage']==27]['log_income']))
ages_array = np.arange(age_start,age_stop+1)
estimates = np.zeros((ages_array.size,6))
for i, age in enumerate(ages_array):
df_age = df[df['tage']==age][['log_income','log_assets','wpfinwgt']]
if weighted:
weights = df_age['wpfinwgt'] / df_age['wpfinwgt'].sum()
else:
weights = np.ones_like(df_age['wpfinwgt'])
weights = weights / weights.sum()
mean_income = (df_age['log_income']*weights).sum()
df_age['n_log_income'] = df_age['log_income'] - mean_income
data = df_age[['n_log_income','log_assets']].copy()
has_assets = data['log_assets']>0
data['log_assets'] = sps.mstats.winsorize(data['log_assets'],(None,0.1)) - np.log(income_for_normalization)
data['n_log_income'] = sps.mstats.winsorize(data['n_log_income'],(0.05,0.05))
data_npar = pd.DataFrame({'z':data['n_log_income'].copy(),'a':data['log_assets'].copy()})
def log_l(theta):
mu_a = theta[0]
sig_a = theta[1]
sig_i = theta[2]
rho = theta[3]
beta_cons = theta[4]
beta_i = theta[5]
saa = sig_a**2
sii = sig_i**2
sia = sig_a*sig_i*rho
cov = [[sii,sia],[sia,saa]]
mu = [0,mu_a]
li_pos = sps.multivariate_normal.logpdf(data,mean=mu,cov=cov).mean()
li_zero = sps.norm.logpdf(data['n_log_income'],0.0,sig_i)
p_0 = sps.norm.cdf(beta_cons + beta_i*data['n_log_income'])
p_pos = 1-p_0
is_positive = has_assets
li = weights*( (is_positive)*(np.log(p_pos) + li_pos) + (~is_positive)*(np.log(p_0) + li_zero) )
return li.sum()
bnds_z = (None,None) if has_assets.mean() > 0 else (0.0,0.0)
print('\n mean log assets {}, std {}'.format(data['log_assets'].mean(),np.std(data['log_assets']) ))
print('max assets {}'.format(np.exp(data['log_assets']).max()) )
res = spo.minimize(lambda x : -log_l(x),[3.0,0.4,0.4,0.0,0.0,0.0],
method = 'L-BFGS-B',
bounds=[(None,None),(1e-4,None),(1e-4,None),(-0.95,0.95),(None,None),bnds_z])
#print('for age = {} result is {}, percentage of non-0 is {}'.format(age,res.x,has_assets.mean()))
estimates[i,:] = res.x
print('estimated mean log assets {}, estimated std {}\n'.format(res.x[0],res.x[1]) )
out = {'Ages':ages_array,'mu_a':estimates[:,0],
'sig_a':estimates[:,1],
'sig_i':estimates[:,2],
'rho':estimates[:,3],
'beta0':estimates[:,4],
'beta1':estimates[:,5]}
if save: filer(save,out,True)
return out
def conditional_normal_parameters(zval,mu_a,mu_z,sig_a,sig_z,rho):
mean = mu_a + (sig_a/sig_z)*rho*(zval-mu_z)
std = (1-rho**2)**(0.5) * sig_a
return mean, std
if __name__ == '__main__':
dist = get_new_estimates()
#print(dist)
dist = filer('za_dist.pkl',0,0)
#print(dist)
tval = np.array([-2,-1,0,1,2],dtype=np.float64)
x, w = np.polynomial.hermite.hermgauss(5)
x = x*np.sqrt(2)
w = w/w.sum()
print((w*(x**2)).sum())
print((w*(x**3)).sum())
print((w*(x**4)).sum())
for t in range(dist['Ages'].size):
zval = dist['sig_i'][t]*tval
inputs = (zval,dist['mu_a'][t],0.0,dist['sig_a'][t],dist['sig_i'][t],dist['rho'][t])
print('inputs are {}'.format(inputs))
mu, sig = conditional_normal_parameters(*inputs)
share_0 = sps.norm.cdf(dist['beta0'][t] + dist['beta1'][t]*zval)
def fun(x):
assert type(x) is tuple, 'x must be a tuple!'
action = x[0]
args = x[1]
assert type(action) is str, 'x[0] should be string for action'
assert len(x) <= 2, 'too many things in x! x is (action,agrs)'
if action == 'test':
return mdl_resid()
elif action == 'compute':
return mdl_resid(args)
elif action == 'moments':
agents = mdl_resid(args, return_format=['agents'])
mom = agents.compute_moments()
return mom
elif action == 'minimize':
import dfols
i, N_st, xfix = args
xl, xu, x0, keys, translator = calibration_params(xfix=xfix)
#Sort lists
def sortFirst(val):
return val[0]
#Get the starting point for local minimization
#Open File with best solution so far
param = filer('wisdom.pkl', 0, False)
param.sort(key=sortFirst)
print('f best so far is {} and x is {}'.format(param[0][0],
param[0][1]))
xm = param[0][1]
#Get right sobol sequence point
xt = filer('sobol.pkl', None, False)
#Determine the initial position
dump = min(max(0.1, ((i + 1) / N_st)**(0.5)), 0.995)
xc = dump * xm + (1 - dump) * xt[:, i]
xc = xc.squeeze()
def q(pt):
try:
ans = mdl_resid(translator(pt),
moments_repeat=3,
return_format=['scaled residuals'])
except BaseException as a:
print('During optimization function evaluation failed at {}'.
format(pt))
print(a)
ans = np.array([1e6])
finally:
gc.collect()
return ans
res = dfols.solve(
q,
xc,
rhobeg=0.15,
rhoend=1e-6,
maxfun=npt,
bounds=(xl, xu),
#npt=len(xc)+5,
scaling_within_bounds=True,
#user_params={'tr_radius.gamma_dec':0.75,'tr_radius.gamma_inc':1.5,
# 'tr_radius.alpha1':0.5,'tr_radius.alpha2':0.75,
# 'regression.momentum_extra_steps':True,
#'restarts.use_restarts':True},
objfun_has_noise=True)
print(res)
if res.flag == -1:
raise Exception('solver returned something creepy...')
fbest = mdl_resid(translator(
res.x)) # in prnciple, this can be inconsistent with
# squared sum of residuals
print('fbest is {} and res.f is {}'.format(fbest, res.f))
print('Final value is {}'.format(fbest))
param_new = filer('wisdom.pkl', None, False)
param_write = param_new + [(fbest, res.x)]
#Save Updated File
param_write.sort(key=sortFirst)
filer('wisdom.pkl', param_write, True)
return fbest
else:
raise Exception('unsupported action or format')
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Tue May 26 14:25:40 2020
@author: egorkozlov
"""
import numpy as np
from tiktak import filer
V_base = filer('v_save_base.pkl',0,0)
V_comp = filer('v_save_counterfactual.pkl',0,0)
from setup import ModelSetup
import matplotlib.pyplot as plt
d = {'high education':True}
su_def = ModelSetup(**d)
agrid = su_def.agrid_s
t = 4
a0 = 0.0
ia0 = 0#np.searchsorted(agrid,a0)
sname = 'Female, single'
oname = 'V'
zfgrid = su_def.exogrid.zf_t[t]
zftrend = su_def.pars['f_wage_trend'][t]
names = ['extra assets in base to reach 0 in compare',
def __init__(self,
nogrid=False,
divorce_costs_k='Default',
divorce_costs_nk='Default',
**kwargs):
p = dict()
T = 55
Tret = 45 # first period when the agent is retired
Tfert = 18 # first peroid when infertile
Tdiv = 44 # first period when cannot divorce / renegotiate
Tmeet = 25 # first period when stop meeting partners
Tinc = 25 # first period where stop tracking income process and assume it to be fixed
p['T'] = T
p['Tret'] = Tret
p['Tfert'] = Tfert
p['Tsim'] = T
p['n_zf_t'] = [7] * Tret + [1] * (T - Tret)
p['n_zm_t'] = [5] * Tret + [1] * (T - Tret)
p['sigma_psi_mult'] = 0.28
p['sigma_psi'] = 0.11
p['R_t'] = [1.025] * T
p['n_psi'] = 15
p['beta_t'] = [0.98] * T
p['A'] = 1.0 # consumption in couple: c = (1/A)*[c_f^(1+rho) + c_m^(1+rho)]^(1/(1+rho))
p['crra_power'] = 1.5
p['couple_rts'] = 0.23
p['sig_partner_a'] = 0.1
p['mu_partner_a_female'] = 0.00
p['mu_partner_a_male'] = -0.00
p['dump_factor_z'] = 0.75
p['sig_partner_z'] = 0.25
p['mu_partner_z_male'] = -0.02
p['mu_partner_z_female'] = 0.02
p['m_bargaining_weight'] = 0.5
p['pmeet_21'] = 0.1
p['pmeet_28'] = 0.2
p['pmeet_35'] = 0.1
p['m_zf'] = 0.9
p['m_zf0'] = 1.0
p['z_drift'] = -0.09
p['no kids at meeting'] = True
p['high education'] = True # what trend to pick
p['any kids'] = True
p['wret'] = 0.8
p['uls'] = 0.2
p['pls'] = 1.0
p['income_sd_mult'] = 1.0
p['pay_gap'] = True
p['preg_mult'] = 1.0
p['u_shift_mar'] = 0.0
p['u_shift_coh'] = 0.0
p['sm_shift'] = 0.0
p['disutil_marry_sm_fem_coef'] = 0.0
p['disutil_marry_sm_mal_coef'] = 10.0
p['disutil_shotgun_coef'] = 2.0
p['pmeet_multiplier_fem'] = 1.0
p['p_to_meet_sm_if_mal'] = 0.1
p['taste_shock_mult'] = 1.0
p['p_abortion_access'] = 0.5
p['abortion_costs_mult'] = 10.0
p['u_lost_divorce_mult'] = 0.0
p['child_a_cost'] = 0.0
p['child_support_share'] = 0.0
p['util_lam'] = 0.7
p['util_alp'] = 0.5
p['util_xi'] = 1.5
p['util_kap'] = 0.5
p['util_qbar'] = 0.0
p['util_out_lf'] = 0.0
p['preg_21'] = 0.01
p['preg_28'] = 0.5
p['preg_35'] = 0.3
for key, value in kwargs.items():
assert (key in p), 'wrong name?'
p[key] = value
if p['high education']:
p['sig_zm'] = p['income_sd_mult'] * 0.16138593
p['sig_zm_0'] = p['income_sd_mult'] * 0.41966813
# FIXME: I need guidance how to pin these down
p['sig_zf'] = p['income_sd_mult'] * p['m_zf'] * 0.19571624
p['sig_zf_0'] = p['income_sd_mult'] * p['m_zf0'] * 0.43351219
else:
p['sig_zm'] = p['income_sd_mult'] * 0.2033373
p['sig_zm_0'] = p['income_sd_mult'] * 0.40317171
p['sig_zf'] = p['income_sd_mult'] * p['m_zf'] * 0.14586778
p['sig_zf_0'] = p['income_sd_mult'] * p['m_zf0'] * 0.62761052
# college
if p['high education']:
m_trend_data = [
0.0, 0.11316599, .2496034, .31260625, .37472204, .4268548,
.48067884, .52687573, .57293878, .60941412, .65015743,
.6685226, .72482815, .74446455, .76712521, .78038137,
.79952806, .80092523, .81972567, .82913486, .83849471,
.84308452, .84646086, .85437072, .85499576
]
f_trend_data = [
0.0, 0.06715984, .21149606, .32283002, .46885336, .52712037,
.58302632, .63348555, .68024646, .71450132, .74246337,
.77044807, .79946406, .80640353, .83799304, .85356081,
.86832235, .87407447, .87820755, .86840901, .87630054,
.8765972, .87894493, .87800553, .87932908
]
nm = len(m_trend_data) - 1
nf = len(f_trend_data) - 1
t0 = 4
gap = 3.0340077 - 2.8180354 # male - female
c_female = -f_trend_data[t0]
c_male = gap - m_trend_data[t0]
p['m_wage_trend'] = np.array(
[c_male + m_trend_data[min(t, nm)] for t in range(T)])
p['f_wage_trend'] = np.array(
[c_female + f_trend_data[min(t, nf)] for t in range(T)])
else:
# no college
p['m_wage_trend'] = np.array([
-0.2424105 + 0.037659 * (min(t + 2, 30) - 5) - 0.0015337 *
((min(t + 2, 30) - 5)**2) + 0.000026 *
((min(t + 2, 30) - 5)**3) for t in range(T)
])
p['f_wage_trend'] = np.array([
-0.3668214 + 0.0264887 * (min(t, 30) - 5) - 0.0012464 *
((min(t, 30) - 5)**2) + 0.0000251 * ((min(t, 30) - 5)**3)
for t in range(T)
])
if not p['pay_gap']:
p['sig_zf'], p['sig_zf_0'] = p['sig_zm'], p['sig_zm_0']
p['f_wage_trend'] = p['m_wage_trend']
# derivative parameters
p['sigma_psi_init'] = p['sigma_psi_mult'] * p['sigma_psi']
p['disutil_marry_sm_mal'] = p['disutil_marry_sm_mal_coef'] * p[
'u_shift_mar']
p['disutil_marry_sm_fem'] = p['disutil_marry_sm_fem_coef'] * p[
'u_shift_mar']
p['disutil_shotgun'] = p['disutil_shotgun_coef'] * p['sigma_psi_init']
p['abortion_costs'] = p['abortion_costs_mult'] * p['u_shift_mar']
p['u_lost_divorce'] = p['u_lost_divorce_mult'] * p['sigma_psi_init']
p['preg_az'] = 0.00
p['preg_azt'] = 0.00
#Get the probability of meeting, adjusting for year-period
p['taste_shock'] = 0.0 * p['taste_shock_mult'] * 0.0 #p['sigma_psi']
p['is fertile'] = [p['any kids']] * Tfert + [False] * (T - Tfert)
p['can divorce'] = [True] * Tdiv + [False] * (T - Tdiv)
#p['poutsm_t'] = [p['poutsm']]*T
p['pmeet_0'], p['pmeet_t'], p['pmeet_t2'] = prob_polyfit(
(p['pmeet_21'], 0), (p['pmeet_28'], 7), (p['pmeet_35'], 14),
max_power=2)
p['preg_a0'], p['preg_at'], p['preg_at2'] = prob_polyfit(
(p['preg_21'], 0), (p['preg_28'], 7), (p['preg_35'], 14),
max_power=2)
p['pmeet_t'] = [
np.clip(p['pmeet_0'] + t * p['pmeet_t'] +
(t**2) * p['pmeet_t2'], 0.0, 1.0) for t in range(Tmeet)
] + [0.0] * (T - Tmeet)
p['n_psi_t'] = [p['n_psi']] * T
self.pars = p
self.dtype = np.float64 # type for all floats
# relevant for integration
self.state_names = [
'Female, single', 'Male, single', 'Female and child',
'Couple, no children', 'Couple and child'
]
# female labor supply
lmin = 0.2
lmax = 1.0
nl = 2
ls = np.array([0.2, 1.0]) #np.linspace(lmin,lmax,nl,dtype=self.dtype)
ps = np.array([p['pls'], 0.0])
ls_ushift = np.array([p['util_out_lf'], 0.0])
self.ls_levels = dict()
self.ls_levels['Couple, no children'] = np.array([1.0],
dtype=self.dtype)
self.ls_levels['Female, single'] = np.array([1.0], dtype=self.dtype)
self.ls_levels['Male, single'] = np.array([1.0], dtype=self.dtype)
self.ls_levels['Couple and child'] = ls
self.ls_levels['Female and child'] = ls
self.ls_ushift = dict()
self.ls_ushift['Couple, no children'] = np.array([0.0],
dtype=self.dtype)
self.ls_ushift['Female, single'] = np.array([0.0], dtype=self.dtype)
self.ls_ushift['Male, single'] = np.array([0.0], dtype=self.dtype)
self.ls_ushift['Couple and child'] = ls_ushift
self.ls_ushift['Female and child'] = ls_ushift
#self.ls_utilities = np.array([p['uls'],0.0],dtype=self.dtype)
self.ls_pdown = dict()
self.ls_pdown['Couple, no children'] = np.array([0.0],
dtype=self.dtype)
self.ls_pdown['Female, single'] = np.array([0.0], dtype=self.dtype)
self.ls_pdown['Male, single'] = np.array([0.0], dtype=self.dtype)
self.ls_pdown['Female and child'] = ps
self.ls_pdown['Couple and child'] = ps
self.nls = dict()
self.nls['Couple and child'] = len(self.ls_levels['Couple and child'])
self.nls['Couple, no children'] = len(
self.ls_levels['Couple, no children'])
self.nls['Female and child'] = len(self.ls_levels['Female and child'])
self.nls['Female, single'] = len(self.ls_levels['Female, single'])
self.nls['Male, single'] = len(self.ls_levels['Male, single'])
#Cost of Divorce
if divorce_costs_k == 'Default':
# by default the costs are set in the bottom
self.divorce_costs_k = DivorceCosts(
u_lost_m=self.pars['u_lost_divorce'],
u_lost_f=self.pars['u_lost_divorce'])
else:
if isinstance(divorce_costs_k, dict):
# you can feed in arguments to DivorceCosts
self.divorce_costs_k = DivorceCosts(**divorce_costs_k)
else:
# or just the output of DivorceCosts
assert isinstance(divorce_costs_k, DivorceCosts)
self.divorce_costs_k = divorce_costs_k
#Cost of Separation
if divorce_costs_nk == 'Default':
# by default the costs are set in the bottom
self.divorce_costs_nk = DivorceCosts(
u_lost_m=self.pars['u_lost_divorce'],
u_lost_f=self.pars['u_lost_divorce'])
else:
if isinstance(divorce_costs_nk, dict):
# you can feed in arguments to DivorceCosts
self.divorce_costs_nk = DivorceCosts(**divorce_costs_nk)
else:
# or just the output of DivorceCosts
assert isinstance(divorce_costs_nk, DivorceCosts)
self.divorce_costs_nk = divorce_costs_nk
# exogrid should be deprecated
if not nogrid:
exogrid = dict()
# let's approximate three Markov chains
# this sets up exogenous grid
# FIXME: this uses number of points from 0th entry.
# in principle we can generalize this
exogrid['zf_t'], exogrid['zf_t_mat'] = rouw_nonst(
p['T'], p['sig_zf'], p['sig_zf_0'], p['n_zf_t'][0])
exogrid['zm_t'], exogrid['zm_t_mat'] = rouw_nonst(
p['T'], p['sig_zm'], p['sig_zm_0'], p['n_zm_t'][0])
for t in range(Tinc, Tret):
for key in ['zf_t', 'zf_t_mat', 'zm_t', 'zm_t_mat']:
exogrid[key][t] = exogrid[key][Tinc]
for t in range(Tret, T):
exogrid['zf_t'][t] = np.array([np.log(p['wret'])])
exogrid['zm_t'][t] = np.array([np.log(p['wret'])])
exogrid['zf_t_mat'][t] = np.atleast_2d(1.0)
exogrid['zm_t_mat'][t] = np.atleast_2d(1.0)
# fix transition from non-retired to retired
exogrid['zf_t_mat'][Tret - 1] = np.ones((p['n_zf_t'][Tret - 1], 1))
exogrid['zm_t_mat'][Tret - 1] = np.ones((p['n_zm_t'][Tret - 1], 1))
exogrid['psi_t'], exogrid['psi_t_mat'] = rouw_nonst(
p['T'], p['sigma_psi'], p['sigma_psi_init'], p['n_psi_t'][0])
zfzm, zfzmmat = combine_matrices_two_lists(exogrid['zf_t'],
exogrid['zm_t'],
exogrid['zf_t_mat'],
exogrid['zm_t_mat'])
all_t, all_t_mat = combine_matrices_two_lists(
zfzm, exogrid['psi_t'], zfzmmat, exogrid['psi_t_mat'])
all_t_mat_sparse_T = [
sparse.csc_matrix(D.T) if D is not None else None
for D in all_t_mat
]
#Create a new bad version of transition matrix p(zf_t)
zf_bad = [
tauchen_drift(exogrid['zf_t'][t], exogrid['zf_t'][t + 1], 1.0,
p['sig_zf'], p['z_drift'])
for t in range(self.pars['T'] - 1)
] + [None]
#zf_bad = [cut_matrix(exogrid['zf_t_mat'][t]) if t < Tret -1
# else (exogrid['zf_t_mat'][t] if t < T - 1 else None)
# for t in range(self.pars['T'])]
zf_t_mat_down = zf_bad
zfzm, zfzmmat = combine_matrices_two_lists(exogrid['zf_t'],
exogrid['zm_t'],
zf_t_mat_down,
exogrid['zm_t_mat'])
all_t_down, all_t_mat_down = combine_matrices_two_lists(
zfzm, exogrid['psi_t'], zfzmmat, exogrid['psi_t_mat'])
all_t_mat_down_sparse_T = [
sparse.csc_matrix(D.T) if D is not None else None
for D in all_t_mat_down
]
all_t_mat_by_l_nk = [[
(1 - p) * m + p * md if m is not None else None
for m, md in zip(all_t_mat, all_t_mat_down)
] for p in self.ls_pdown['Couple, no children']]
all_t_mat_by_l_spt_nk = [[
(1 - p) * m + p * md if m is not None else None
for m, md in zip(all_t_mat_sparse_T, all_t_mat_down_sparse_T)
] for p in self.ls_pdown['Couple, no children']]
all_t_mat_by_l_k = [[
(1 - p) * m + p * md if m is not None else None
for m, md in zip(all_t_mat, all_t_mat_down)
] for p in self.ls_pdown['Couple and child']]
all_t_mat_by_l_spt_k = [[
(1 - p) * m + p * md if m is not None else None
for m, md in zip(all_t_mat_sparse_T, all_t_mat_down_sparse_T)
] for p in self.ls_pdown['Couple and child']]
zf_t_mat_by_l_sk = [[
(1 - p) * m + p * md if md is not None else None
for m, md in zip(exogrid['zf_t_mat'], zf_bad)
] for p in self.ls_pdown['Female and child']]
exogrid['all_t_mat_by_l_nk'] = all_t_mat_by_l_nk
exogrid['all_t_mat_by_l_spt_nk'] = all_t_mat_by_l_spt_nk
exogrid['all_t_mat_by_l_k'] = all_t_mat_by_l_k
exogrid['all_t_mat_by_l_spt_k'] = all_t_mat_by_l_spt_k
exogrid['zf_t_mat_by_l_sk'] = zf_t_mat_by_l_sk
exogrid['all_t'] = all_t
Exogrid_nt = namedtuple('Exogrid_nt', exogrid.keys())
self.exogrid = Exogrid_nt(**exogrid)
self.pars['nexo_t'] = [v.shape[0] for v in all_t]
self.compute_child_support_transitions(
child_support_share=p['child_support_share'])
#assert False
#Grid Couple
self.na = 40
self.amin = 0
self.amax = 100
self.agrid_c = np.linspace(self.amin**0.5,
self.amax**0.5,
self.na,
dtype=self.dtype)**2
#tune=1.5
#self.agrid_c = np.geomspace(self.amin+tune,self.amax+tune,num=self.na)-tune
# this builds finer grid for potential savings
s_between = 7 # default numer of points between poitns on agrid
s_da_min = 0.01 # minimal step (does not create more points)
s_da_max = 0.1 # maximal step (creates more if not enough)
self.sgrid_c = build_s_grid(self.agrid_c, s_between, s_da_min,
s_da_max)
self.vsgrid_c = VecOnGrid(self.agrid_c, self.sgrid_c)
#Grid Single
self.amin_s = 0
self.amax_s = self.amax / 2.0
self.agrid_s = self.agrid_c / 2.0
#tune_s=1.5
#self.agrid_s = np.geomspace(self.amin_s+tune_s,self.amax_s+tune_s,num=self.na)-tune_s
self.sgrid_s = build_s_grid(self.agrid_s, s_between, s_da_min,
s_da_max)
self.vsgrid_s = VecOnGrid(self.agrid_s, self.sgrid_s)
# grid for theta
self.ntheta = 11
self.thetamin = 0.01
self.thetamax = 0.99
self.thetagrid = np.linspace(self.thetamin,
self.thetamax,
self.ntheta,
dtype=self.dtype)
self.child_a_cost_single = np.minimum(self.agrid_s,
self.pars['child_a_cost'])
self.child_a_cost_couple = np.minimum(self.agrid_c,
self.pars['child_a_cost'])
self.vagrid_child_single = VecOnGrid(
self.agrid_s, self.agrid_s - self.child_a_cost_single)
self.vagrid_child_couple = VecOnGrid(
self.agrid_c, self.agrid_c - self.child_a_cost_couple)
# construct finer grid for bargaining
ntheta_fine = 10 * self.ntheta # actual number may be a bit bigger
self.thetagrid_fine = np.unique(
np.concatenate((self.thetagrid,
np.linspace(self.thetamin,
self.thetamax,
ntheta_fine,
dtype=self.dtype))))
self.ntheta_fine = self.thetagrid_fine.size
i_orig = list()
for theta in self.thetagrid:
assert theta in self.thetagrid_fine
i_orig.append(np.where(self.thetagrid_fine == theta)[0])
assert len(i_orig) == self.thetagrid.size
# allows to recover original gird points on the fine grid
self.theta_orig_on_fine = np.array(i_orig).flatten()
self.v_thetagrid_fine = VecOnGrid(self.thetagrid, self.thetagrid_fine)
# precomputed object for interpolation
self.exo_grids = {
'Female, single': exogrid['zf_t'],
'Male, single': exogrid['zm_t'],
'Female and child': exogrid['zf_t'],
'Couple and child': exogrid['all_t'],
'Couple, no children': exogrid['all_t']
}
self.exo_mats = {
'Female, single': exogrid['zf_t_mat'],
'Male, single': exogrid['zm_t_mat'],
'Female and child': exogrid['zf_t_mat_by_l_sk'],
'Couple and child': exogrid['all_t_mat_by_l_k'],
'Couple, no children': exogrid['all_t_mat_by_l_nk']
} # sparse version?
self.utility_shifters = {
'Female, single': 0.0,
'Male, single': 0.0,
'Female and child': p['u_shift_mar'] + p['sm_shift'],
'Couple and child': p['u_shift_mar'],
'Couple, no children': p['u_shift_coh']
}
# this pre-computes transition matrices for meeting a partner
try:
self.partners_distribution_fem = filer('az_dist_fem.pkl',
0,
0,
repeat=False)
self.partners_distribution_mal = filer('az_dist_mal.pkl',
0,
0,
repeat=False)
except:
print('recreating estimates...')
est_fem = get_estimates(
fname='income_assets_distribution_male.csv',
age_start=23,
age_stop=42,
zlist=self.exogrid.zm_t[2:])
filer('az_dist_fem.pkl', est_fem, True, repeat=False)
self.partners_distribution_fem = est_fem
est_mal = get_estimates(
fname='income_assets_distribution_female.csv',
age_start=21,
age_stop=40,
zlist=self.exogrid.zf_t[0:])
filer('az_dist_mal.pkl', est_mal, True, repeat=False)
self.partners_distribution_mal = est_mal
self.build_matches()
# building m grid
ezfmin = min([
np.min(np.exp(g + t))
for g, t in zip(exogrid['zf_t'], p['f_wage_trend'])
])
ezmmin = min([
np.min(np.exp(g + t))
for g, t in zip(exogrid['zm_t'], p['m_wage_trend'])
])
ezfmax = max([
np.max(np.exp(g + t))
for g, t in zip(exogrid['zf_t'], p['f_wage_trend'])
])
ezmmax = max([
np.max(np.exp(g + t))
for g, t in zip(exogrid['zm_t'], p['m_wage_trend'])
])
self.money_min = 0.95 * min(self.ls_levels['Female and child']) * min(
ezmmin, ezfmin) # cause FLS can be up to 0
mmin = self.money_min
mmax = ezfmax + ezmmax + np.max(self.pars['R_t']) * self.amax
mint = (ezfmax + ezmmax) # poin where more dense grid begins
ndense = 600
nm = 1500
gsparse = np.linspace(mint, mmax, nm - ndense)
gdense = np.linspace(mmin, mint,
ndense + 1) # +1 as there is a common pt
self.mgrid = np.zeros(nm, dtype=self.dtype)
self.mgrid[ndense:] = gsparse
self.mgrid[:(ndense + 1)] = gdense
self.mgrid_c = self.mgrid
self.mgrid_s = self.mgrid
assert np.all(np.diff(self.mgrid) > 0)
self.u_precompute()
self.unplanned_pregnancy_probability()
self.compute_taxes()
def __init__(self,*,base,targ_mode,compare=None,
setup=None,
base_name='Model',
compare_name='Data',
graphs_title_add=None,
moments_aux=None):
if setup is None: self.setup = ModelSetup()
if type(base) is str:
self.moments = filer(base,0,0)
else:
self.moments = base
self.targ_mode = targ_mode
self.base_name = base_name
self.compare_name = compare_name
if graphs_title_add:
self.graphs_title_add = '\n ' + graphs_title_add
else:
self.graphs_title_add = ''
if type(compare) is str:
targ_load = filer(compare,0,0)
# emulating targets
self.targets = {key: (targ_load[key],0.0) for key in targ_load}
else:
self.targets = all_targets(targ_mode)
try:
self.print_things()
except:
print('failed to print')
try:
self.plot_estimates()
except:
print('failed to plot estimates')
try:
self.plot_hazards()
except:
print('failed to plot hazards')
try:
self.plot_cumulative()
except:
print('failed to plot cumulative')
try:
self.plot_by_years_after_marriage()
except:
print('failed to plot by years after marriage')
try:
self.plot_kfmf()
except:
print('failed to plot kfmf')
if moments_aux is not None:
try:
self.plot_men()
except:
print('failed to plot men')
try:
self.plot_kfmf_ref()
except:
print('failed to plot ref')
(r'\% MF women in population at 30', 'm then k in population at 30'),
(r'\% single mothers among mothers at 35',
'single mothers among mothers at 35'),
(r'\% unplanned pregnancies aborted', 'unplanned pregnancies aborted'),
#(r'\% with kids 5 years after marriage','ever kids by years after marriage, 5')
]
table_cols = list()
for educ in ['col', 'hs']:
def file_name(nm):
return '{} {}.pkl'.format(educ, nm)
for ename, nm in names:
fname = file_name(nm)
mom = filer(fname, 0, 0)
table_col = [r'\textbf{' + ename + r'}'] + [mom[e] for _, e in entries]
table_cols.append(table_col)
table_left = [r'\textbf{Experiment}'
] + [r'\textit{' + e + r'}' for e, _ in entries]
# build table rows
table_rows = list()
for irow in range(len(entries) + 1):
table_row = table_left[irow]
for icol in range(2 * len(names)):
try:
table_row += ' & ' + '{:02.1f}'.format(
100 * table_cols[icol][irow])
