def main():
# Set targets for K/Y and the Lorenz curve based on the data
lorenz_target = getLorenzShares(Params.SCF_wealth,weights=Params.SCF_weights,percentiles=Params.percentiles_to_match)
lorenz_long_data = np.hstack((np.array(0.0),getLorenzShares(Params.SCF_wealth,weights=Params.SCF_weights,percentiles=np.arange(0.01,1.0,0.01).tolist()),np.array(1.0)))
KY_target = 10.26
# Make AgentTypes for estimation
InfiniteType = cstwMPCagent(**Params.init_infinite)
InfiniteType.AgeDstn = np.array(1.0)
EstimationAgentList = []
total_types = 1
EstimationAgentList.append(deepcopy(InfiniteType))
# Make an economy for the consumers to live in
EstimationEconomy = cstwMPCmarket(**Params.init_market)
EstimationEconomy.agents = EstimationAgentList
EstimationEconomy.KYratioTarget = KY_target
EstimationEconomy.LorenzTarget = lorenz_target
EstimationEconomy.LorenzData = lorenz_long_data
EstimationEconomy.PopGroFac = 1.0
EstimationEconomy.TypeWeight = [1.0]
EstimationEconomy.act_T = Params.T_sim_PY
EstimationEconomy.ignore_periods = Params.ignore_periods_PY
# Choose the bounding region for the parameter search
spec_name = 'BetaDistPY'
param_name = 'DiscFac'
dist_type = 'uniform'
if param_name == 'CRRA':
param_range = [0.2,70.0]
elif param_name == 'DiscFac':
param_range = [0.95,0.99]
else:
print('Parameter range for ' + Params.param_name + ' has not been defined!')
# Run the param-point estimation only
paramPointObjective = lambda center : getKYratioDifference(Economy = EstimationEconomy,
param_name = param_name,
param_count = total_types,
center = center,
spread = 0.0,
dist_type = dist_type)
t_start = clock()
center_estimate = brentq(paramPointObjective,param_range[0],param_range[1],xtol=1e-6)
spread_estimate = 0.0
t_end = clock()
# Display statistics about the estimated model
EstimationEconomy.LorenzBool = True
EstimationEconomy.ManyStatsBool = True
EstimationEconomy.distributeParams(param_name,total_types,center_estimate,spread_estimate,dist_type)
EstimationEconomy.solve()
EstimationEconomy.calcLorenzDistance()
print('Estimate is center=' + str(center_estimate) + ', spread=' + str(spread_estimate) + ', took ' + str(t_end-t_start) + ' seconds.')
EstimationEconomy.center_estimate = center_estimate
EstimationEconomy.spread_estimate = spread_estimate
EstimationEconomy.showManyStats(spec_name)
def calcLorenzDistance(SomeTypes):
'''
Calculates the Euclidean distance between the simulated and actual (from SCF data) Lorenz curves at the
20th, 40th, 60th, and 80th percentiles.
Parameters
----------
SomeTypes : [AgentType]
List of AgentTypes that have been solved and simulated. Current levels of individual assets should
be stored in the attribute aLvlNow.
Returns
-------
lorenz_distance : float
Euclidean distance (square root of sum of squared differences) between simulated and actual Lorenz curves.
'''
# Define empirical Lorenz curve points
lorenz_SCF = np.array([-0.00183091, 0.0104425, 0.0552605, 0.1751907])
# Extract asset holdings from all consumer types
aLvl_sim = np.concatenate([ThisType.aLvlNow for ThisType in MyTypes])
# Calculate simulated Lorenz curve points
lorenz_sim = getLorenzShares(aLvl_sim, percentiles=[0.2, 0.4, 0.6, 0.8])
# Calculate the Euclidean distance between the simulated and actual Lorenz curves
lorenz_distance = np.sqrt(np.sum((lorenz_SCF - lorenz_sim)**2))
# Return the Lorenz distance
return lorenz_distance
def calculateLorenzDifference(sim_wealth, weights, percentiles, target_levels):
'''
Calculates the sum of squared differences between the simulatedLorenz curve
at the specified percentile levels and the target Lorenz levels.
Parameters
----------
sim_wealth : numpy.array
Array with simulated wealth values.
weights : numpy.array
List of weights for each row of sim_wealth.
percentiles : [float]
Points in the distribution of wealth to match.
target_levels : np.array
Actual U.S. Lorenz curve levels at the specified percentiles.
Returns
-------
distance : float
Sum of squared distances between simulated and target Lorenz curves.
'''
sim_lorenz = getLorenzShares(sim_wealth,
weights=weights,
percentiles=percentiles)
distance = sum((100 * sim_lorenz - 100 * target_levels)**2)
return distance
def makeLorenzFig(real_wealth, real_weights, sim_wealth, sim_weights):
'''
Produces a Lorenz curve for the distribution of wealth, comparing simulated
to actual data. A sub-function of makeCSTWresults().
Parameters
----------
real_wealth : np.array
Data on household wealth.
real_weights : np.array
Weighting array of the same size as real_wealth.
sim_wealth : np.array
Simulated wealth holdings of many households.
sim_weights :np.array
Weighting array of the same size as sim_wealth.
Returns
-------
these_percents : np.array
An array of percentiles of households, by wealth.
real_lorenz : np.array
Lorenz shares for real_wealth corresponding to these_percents.
sim_lorenz : np.array
Lorenz shares for sim_wealth corresponding to these_percents.
'''
these_percents = np.linspace(0.0001, 0.9999, 201)
real_lorenz = getLorenzShares(real_wealth,
weights=real_weights,
percentiles=these_percents)
sim_lorenz = getLorenzShares(sim_wealth,
weights=sim_weights,
percentiles=these_percents)
plt.plot(100 * these_percents, real_lorenz, '-k', linewidth=1.5)
plt.plot(100 * these_percents, sim_lorenz, '--k', linewidth=1.5)
plt.xlabel('Wealth percentile', fontsize=14)
plt.ylabel('Cumulative wealth ownership', fontsize=14)
plt.title('Simulated vs Actual Lorenz Curves', fontsize=16)
plt.legend(('Actual', 'Simulated'), loc=2, fontsize=12)
plt.ylim(-0.01, 1)
plt.show()
return (these_percents, real_lorenz, sim_lorenz)
def calcCSTWmpcStats(Agents):
'''
Calculate and print to screen overall and education-specific aggregate
wealth to income ratios, as well as the 20th, 40th, 60th, and 80th percentile
points of the Lorenz curve for (liquid) wealth.
Parameters
----------
Agents : [AgentType]
List of AgentTypes in the economy.
Returns
-------
None
'''
yLvlAll = np.concatenate([ThisType.lLvlNow for ThisType in Agents])
aLvlAll = np.concatenate([ThisType.aLvlNow for ThisType in Agents])
AgeAll = np.concatenate([ThisType.t_age for ThisType in Agents])
EducAll = np.concatenate([
ThisType.EducType * np.ones(ThisType.AgentCount) for ThisType in Agents
])
WeightAll = 1.01**(-0.25 * AgeAll)
yAgg = np.dot(yLvlAll, WeightAll)
aAgg = np.dot(aLvlAll, WeightAll)
yAggD = np.dot(yLvlAll, WeightAll * (EducAll == 0))
yAggH = np.dot(yLvlAll, WeightAll * (EducAll == 1))
yAggC = np.dot(yLvlAll, WeightAll * (EducAll == 2))
aAggD = np.dot(aLvlAll, WeightAll * (EducAll == 0))
aAggH = np.dot(aLvlAll, WeightAll * (EducAll == 1))
aAggC = np.dot(aLvlAll, WeightAll * (EducAll == 2))
LorenzPts = getLorenzShares(aLvlAll,
weights=WeightAll,
percentiles=[0.2, 0.4, 0.6, 0.8])
print('Overall aggregate wealth to income ratio is ' + mystr(aAgg / yAgg) +
' (target 6.60).')
print('Aggregate wealth to income ratio for dropouts is ' +
mystr(aAggD / yAggD) + ' (target 1.60).')
print('Aggregate wealth to income ratio for high school grads is ' +
mystr(aAggH / yAggH) + ' (target 3.78).')
print('Aggregate wealth to income ratio for college grads is ' +
mystr(aAggC / yAggC) + ' (target 8.84).')
print('Share of liquid wealth of the bottom 20% is ' +
mystr(100 * LorenzPts[0]) + '% (target 0.0%).')
print('Share of liquid wealth of the bottom 40% is ' +
mystr(100 * LorenzPts[1]) + '% (target 0.4%).')
print('Share of liquid wealth of the bottom 60% is ' +
mystr(100 * LorenzPts[2]) + '% (target 2.5%).')
print('Share of liquid wealth of the bottom 80% is ' +
mystr(100 * LorenzPts[3]) + '% (target 11.7%).')
def objectiveFuncWealth(center,spread):
'''
Objective function of the beta-dist estimation, similar to cstwMPC.
Minimizes the distance between simulated and actual 20-40-60-80 Lorenz
curve points and average wealth to income ratio.
Parameters
----------
center : float
Mean of distribution of discount factor.
spread : float
Half width of span of discount factor.
Returns
-------
distance : float
Distance between simulated and actual moments.
'''
DiscFacSet = approxUniform(N=TypeCount,bot=center-spread,top=center+spread)[1]
for j in range(TypeCount):
Agents[j](DiscFac = DiscFacSet[j])
multiThreadCommands(Agents,['solve()','initializeSim()','simulate()'])
aLvl_sim = np.concatenate([agent.aLvlNow for agent in Agents])
aNrm_sim = np.concatenate([agent.aNrmNow for agent in Agents])
aNrmMean_sim = np.mean(aNrm_sim)
Lorenz_sim = list(getLorenzShares(aLvl_sim,percentiles=percentile_targets))
moments_sim = np.array([aNrmMean_sim] + Lorenz_sim)
moments_diff = moments_sim - moments_data
moments_diff[1:] *= 1 # Rescale Lorenz shares
distance = np.sqrt(np.dot(moments_diff,moments_diff))
print('Tried center=' + str(center) + ', spread=' + str(spread) + ', got distance=' + str(distance))
print(moments_sim)
return distance
print("The mean of individual wealth is "+ str(sim_wealth.mean()) + ";\n the standard deviation is "
+ str(sim_wealth.std())+";\n the median is " + str(np.median(sim_wealth)) +".")
# %% {"code_folding": []}
# Get some tools for plotting simulated vs actual wealth distributions
from HARK.utilities import getLorenzShares, getPercentiles
# The cstwMPC model conveniently has data on the wealth distribution
# from the U.S. Survey of Consumer Finances
from HARK.cstwMPC.SetupParamsCSTW import SCF_wealth, SCF_weights
# %% {"code_folding": []}
# Construct the Lorenz curves and plot them
pctiles = np.linspace(0.001,0.999,15)
SCF_Lorenz_points = getLorenzShares(SCF_wealth,weights=SCF_weights,percentiles=pctiles)
sim_Lorenz_points = getLorenzShares(sim_wealth,percentiles=pctiles)
# Plot
plt.figure(figsize=(5,5))
plt.title('Wealth Distribution')
plt.plot(pctiles,SCF_Lorenz_points,'--k',label='SCF')
plt.plot(pctiles,sim_Lorenz_points,'-b',label='Benchmark KS')
plt.plot(pctiles,pctiles,'g-.',label='45 Degree')
plt.xlabel('Percentile of net worth')
plt.ylabel('Cumulative share of wealth')
plt.legend(loc=2)
plt.ylim([0,1])
make_figs('wealth_distribution_1', True, False)
# remark.show('')
def runRoszypalSchlaffmanExperiment(CorrAct, CorrPcvd, DiscFac_center, DiscFac_spread, numTypes, simPeriods):
'''
Solve and simulate a consumer type who misperceives the extent of serial correlation
of persistent shocks to income.
Parameters
----------
CorrAct : float
Serial correlation coefficient for *actual* persistent income.
CorrPcvd : float
List or array of *perceived* persistent income serial correlation
DiscFac_center : float
A measure of centrality for the distribution of the beta parameter, DiscFac.
DiscFac_spread : float
A measure of spread or diffusion for the distribution of the beta parameter.
numTypes: int
Number of different types of agents (distributed using DiscFac_center and DiscFac_spread)
simPeriods: int
Number of periods to simulate before calculating distributions
Returns
-------
AggWealthRatio: float
Ratio of Aggregate wealth to income.
Lorenz: numpy.array
A list of two 1D array representing the Lorenz curve for assets in the most recent simulated period.
Gini: float
Gini coefficient for assets in the most recent simulated period.
Avg_MPC: numpy.array
Average marginal propensity to consume by income quintile in the latest simulated period.
'''
# Make a dictionary to construct our consumer type
ThisDict = copy(BaselineDict)
ThisDict['PrstIncCorr'] = CorrAct
# Make a N=numTypes point approximation to a uniform distribution of DiscFac
DiscFac_list = approxUniform(N=numTypes,bot=DiscFac_center-DiscFac_spread,top=DiscFac_center+DiscFac_spread)[1]
type_list = []
# Make a PersistentShockConsumerTypeX for each value of beta saved in DiscFac_list
for i in range(len(DiscFac_list)):
ThisDict['DiscFac'] = DiscFac_list[i]
ThisType = PersistentShockConsumerTypeX(**ThisDict)
# Make the consumer *believe* he will face a different level of persistence
ThisType.PrstIncCorr = CorrPcvd
ThisType.updatepLvlNextFunc() # *thinks* E[p_{t+1}] as a function of p_t is different than it is
# Solve the consumer's problem with *perceived* persistence
ThisType.solve()
# Make the consumer type experience the true level of persistence during simulation
ThisType.PrstIncCorr = CorrAct
ThisType.updatepLvlNextFunc()
# Simulate the agents for many periods
ThisType.T_sim = simPeriods
#ThisType.track_vars = ['cLvlNow','aLvlNow','pLvlNow','MPCnow']
ThisType.initializeSim()
ThisType.simulate()
type_list.append(ThisType)
# Get the most recent simulated values of X = cLvlNow, MPCnow, aLvlNow, pLvlNow for all types
cLvl_all = np.concatenate([ThisType.cLvlNow for ThisType in type_list])
aLvl_all = np.concatenate([ThisType.aLvlNow for ThisType in type_list])
MPC_all = np.concatenate([ThisType.MPCnow for ThisType in type_list])
pLvl_all = np.concatenate([ThisType.pLvlNow for ThisType in type_list])
# The ratio of aggregate assets over the income
AggWealthRatio = np.mean(aLvl_all) / np.mean(pLvl_all)
# first 1D array: Create points in the range (0,1)
wealth_percentile = np.linspace(0.001,0.999,201)
# second 1D array: Compute Lorenz shares for the created points
Lorenz_init = getLorenzShares(aLvl_all, percentiles=wealth_percentile)
# Stick 0 and 1 at the boundaries of both arrays to make it inclusive on the range [0,1]
Lorenz_init = np.concatenate([[0],Lorenz_init,[1]])
wealth_percentile = np.concatenate([[0],wealth_percentile,[1]])
# Create a list of wealth_percentile 1D array and Lorenz Shares 1D array
Lorenz = np.stack((wealth_percentile, Lorenz_init))
# Compute the Gini coefficient
Gini = 1.0 - 2.0*np.mean(Lorenz_init[1])
# Compute the average MPC by income quintile in the latest simulated period
Avg_MPC = calcSubpopAvg(MPC_all, pLvl_all, cutoffs=[(0.0,0.2), (0.2,0.4), (0.4,0.6), (0.6,0.8), (0.8,1.0)])
return AggWealthRatio, Lorenz, Gini, Avg_MPC
def calcStats(self, aLvlNow, pLvlNow, MPCnow, lIncomeLvl, EmpNow, t_age,
LorenzBool, ManyStatsBool):
'''
Calculate various statistics about the current population in the economy.
Parameters
----------
aLvlNow : [np.array]
Arrays with end-of-period assets, listed by each ConsumerType in self.agents.
pLvlNow : [np.array]
Arrays with permanent income levels, listed by each ConsumerType in self.agents.
MPCnow : [np.array]
Arrays with marginal propensity to consume, listed by each ConsumerType in self.agents.
lIncomeLvl : [np.array]
Arrays with labor income levels, listed by each ConsumerType in self.agents.
EmpNow : [np.array]
Arrays with employment states: True if employed, False otherwise.
t_age : [np.array]
Arrays with periods elapsed since model entry, listed by each ConsumerType in self.agents.
LorenzBool: bool
Indicator for whether the Lorenz target points should be calculated. Usually False,
only True when DiscFac has been identified for a particular nabla.
ManyStatsBool: bool
Indicator for whether a lot of statistics for tables should be calculated. Usually False,
only True when parameters have been estimated and we want values for tables.
Returns
-------
None
'''
# Combine inputs into single arrays
aLvl = np.hstack(aLvlNow)
pLvl = np.hstack(pLvlNow)
age = np.hstack(t_age)
IncLvl = np.hstack(lIncomeLvl)
Emp = np.hstack(EmpNow)
# Calculate the capital to income ratio in the economy
CohortWeight = self.PopGroFac**(-age)
CapAgg = np.sum(aLvl * CohortWeight)
IncAgg = np.sum(IncLvl * CohortWeight)
KtoYnow = CapAgg / IncAgg
self.KtoYnow = KtoYnow
# Store Lorenz data if requested
self.LorenzLong = np.nan
if LorenzBool:
order = np.argsort(aLvl)
aLvl = aLvl[order]
CohortWeight = CohortWeight[order]
wealth_shares = getLorenzShares(aLvl,
weights=CohortWeight,
percentiles=self.LorenzPercentiles,
presorted=True)
self.Lorenz = wealth_shares
if ManyStatsBool:
self.LorenzLong = getLorenzShares(aLvl,
weights=CohortWeight,
percentiles=np.arange(
0.01, 1.0, 0.01),
presorted=True)
else:
self.Lorenz = np.nan # Store nothing if we don't want Lorenz data
# Calculate a whole bunch of statistics if requested
if ManyStatsBool:
# Reshape other inputs
MPC = np.hstack(MPCnow)
# Sort other data items if aLvl and CohortWeight were sorted
if LorenzBool:
pLvl = pLvl[order]
MPC = MPC[order]
IncLvl = IncLvl[order]
age = age[order]
Emp = Emp[order]
aNrm = aLvl / pLvl # Normalized assets (wealth ratio)
# Calculate overall population MPC and by subpopulations
# MPC_cf_BPP is the MPC that is comparable with the empirical estimation method
MPC_cf_BPP = 1.0 - 0.25 * ((1.0 - MPC) + (1.0 - MPC)**2 +
(1.0 - MPC)**3 + (1.0 - MPC)**4)
self.MPCall = np.sum(
MPC_cf_BPP * CohortWeight) / np.sum(CohortWeight)
employed = Emp
unemployed = np.logical_not(employed)
self.MPCbyWealthRatio = calcSubpopAvg(MPC_cf_BPP, aNrm,
self.cutoffs, CohortWeight)
self.MPCbyIncome = calcSubpopAvg(MPC_cf_BPP, IncLvl, self.cutoffs,
CohortWeight)
# Calculate the wealth quintile distribution of "hand to mouth" consumers
quintile_cuts = getPercentiles(aLvl,
weights=CohortWeight,
percentiles=[0.2, 0.4, 0.6, 0.8])
wealth_quintiles = np.ones(aLvl.size, dtype=int)
wealth_quintiles[aLvl > quintile_cuts[0]] = 2
wealth_quintiles[aLvl > quintile_cuts[1]] = 3
wealth_quintiles[aLvl > quintile_cuts[2]] = 4
wealth_quintiles[aLvl > quintile_cuts[3]] = 5
MPC_cutoff = getPercentiles(
MPC_cf_BPP, weights=CohortWeight, percentiles=[
2.0 / 3.0
]) # Looking at consumers with MPCs in the top 1/3
these = MPC_cf_BPP > MPC_cutoff
in_top_third_MPC = wealth_quintiles[these]
temp_weights = CohortWeight[these]
hand_to_mouth_total = np.sum(temp_weights)
hand_to_mouth_pct = []
for q in range(1, 6):
hand_to_mouth_pct.append(
np.sum(temp_weights[in_top_third_MPC == q]) /
hand_to_mouth_total)
self.HandToMouthPct = np.array(hand_to_mouth_pct)
else: # If we don't want these stats, just put empty values in history
self.MPCall = np.nan
self.MPCunemployed = np.nan
self.MPCemployed = np.nan
self.MPCretired = np.nan
self.MPCbyWealthRatio = np.nan
self.MPCbyIncome = np.nan
self.HandToMouthPct = np.nan
def main():
# Set targets for K/Y and the Lorenz curve based on the data
lorenz_target = getLorenzShares(Params.SCF_wealth,weights=Params.SCF_weights,percentiles=Params.percentiles_to_match)
lorenz_long_data = np.hstack((np.array(0.0),getLorenzShares(Params.SCF_wealth,weights=Params.SCF_weights,percentiles=np.arange(0.01,1.0,0.01).tolist()),np.array(1.0)))
KY_target = 10.26
# Make AgentTypes for estimation
InfiniteType = cstwMPCagent(**Params.init_infinite)
InfiniteType.AgeDstn = np.array(1.0)
EstimationAgentList = []
#Rsave_list = [1.015,1.016,1.017,1.018,1.018,1.019,1.019,1.02,1.021,1.021,1.022,1.022,1.023,1.024,1.026] #20percent truncation
Rsave_list = [1.013,1.015,1.016,1.017,1.018,1.019,1.019,1.02,1.021,1.022,1.023,1.023,1.024,1.026,1.029] #10percent truncation
pref_type_count = 3 # Number of discrete beta types in beta-dist
r_type_count = len(Rsave_list) # declare the number of types we want
total_types = r_type_count * pref_type_count
for n in range(total_types):
EstimationAgentList.append(deepcopy(InfiniteType))
assignRdistribution(EstimationAgentList,Rsave_list)
# Make an economy for the consumers to live in
EstimationEconomy = cstwMPCmarket(**Params.init_market)
EstimationEconomy.agents = EstimationAgentList
EstimationEconomy.KYratioTarget = KY_target
EstimationEconomy.LorenzTarget = lorenz_target
EstimationEconomy.LorenzData = lorenz_long_data
EstimationEconomy.PopGroFac = 1.0
EstimationEconomy.TypeWeight = [1.0]
EstimationEconomy.act_T = Params.T_sim_PY
EstimationEconomy.ignore_periods = Params.ignore_periods_PY
# Choose the bounding region for the parameter search
spec_name = 'BetaDistPY'
param_name = 'DiscFac' # Which parameter to introduce heterogeneity in
dist_type = 'uniform' # Which type of distribution to use
if param_name == 'CRRA':
param_range = [0.2,70.0]
spread_range = [0.00001,1.0]
elif param_name == 'DiscFac':
param_range = [0.95,0.99]
spread_range = [0,0.02]
else:
print('Parameter range for ' + param_name + ' has not been defined!')
# Run the param-dist estimation
paramDistObjective = lambda spread : findLorenzDistanceAtTargetKY(
Economy = EstimationEconomy,
param_name = param_name,
param_count = total_types,
center_range = param_range,
spread = spread,
dist_type = dist_type)
t_start = clock()
spread_estimate = golden(paramDistObjective,brack=spread_range,tol=1e-6)
center_estimate = EstimationEconomy.center_save
t_end = clock()
# Display statistics about the estimated model
EstimationEconomy.LorenzBool = True
EstimationEconomy.ManyStatsBool = True
EstimationEconomy.distributeParams(param_name,total_types,center_estimate,spread_estimate,dist_type)
EstimationEconomy.solve()
EstimationEconomy.calcLorenzDistance()
print('Estimate is center=' + str(center_estimate) + ', spread=' + str(spread_estimate) + ', took ' + str(t_end-t_start) + ' seconds.')
EstimationEconomy.center_estimate = center_estimate
EstimationEconomy.spread_estimate = spread_estimate
EstimationEconomy.showManyStats(spec_name)
def main():
# Set targets for K/Y and the Lorenz curve based on the data
if Params.do_liquid:
lorenz_target = np.array([0.0, 0.004, 0.025,0.117])
KY_target = 6.60
else: # This is hacky until I can find the liquid wealth data and import it
lorenz_target = getLorenzShares(Params.SCF_wealth,weights=Params.SCF_weights,percentiles=Params.percentiles_to_match)
lorenz_long_data = np.hstack((np.array(0.0),getLorenzShares(Params.SCF_wealth,weights=Params.SCF_weights,percentiles=np.arange(0.01,1.0,0.01).tolist()),np.array(1.0)))
#lorenz_target = np.array([-0.002, 0.01, 0.053,0.171])
KY_target = 10.26
# Make AgentTypes for estimation
if Params.do_lifecycle:
DropoutType = cstwMPCagent(**Params.init_dropout)
DropoutType.AgeDstn = calcStationaryAgeDstn(DropoutType.LivPrb,True)
HighschoolType = deepcopy(DropoutType)
HighschoolType(**Params.adj_highschool)
HighschoolType.AgeDstn = calcStationaryAgeDstn(HighschoolType.LivPrb,True)
CollegeType = deepcopy(DropoutType)
CollegeType(**Params.adj_college)
CollegeType.AgeDstn = calcStationaryAgeDstn(CollegeType.LivPrb,True)
DropoutType.update()
HighschoolType.update()
CollegeType.update()
EstimationAgentList = []
for n in range(Params.pref_type_count):
EstimationAgentList.append(deepcopy(DropoutType))
EstimationAgentList.append(deepcopy(HighschoolType))
EstimationAgentList.append(deepcopy(CollegeType))
else:
if Params.do_agg_shocks:
PerpetualYouthType = cstwMPCagent(**Params.init_agg_shocks)
else:
PerpetualYouthType = cstwMPCagent(**Params.init_infinite)
PerpetualYouthType.AgeDstn = np.array(1.0)
EstimationAgentList = []
for n in range(Params.pref_type_count):
EstimationAgentList.append(deepcopy(PerpetualYouthType))
# Give all the AgentTypes different seeds
for j in range(len(EstimationAgentList)):
EstimationAgentList[j].seed = j
# Make an economy for the consumers to live in
EstimationEconomy = cstwMPCmarket(**Params.init_market)
EstimationEconomy.agents = EstimationAgentList
EstimationEconomy.KYratioTarget = KY_target
EstimationEconomy.LorenzTarget = lorenz_target
EstimationEconomy.LorenzData = lorenz_long_data
if Params.do_lifecycle:
EstimationEconomy.PopGroFac = Params.PopGroFac
EstimationEconomy.TypeWeight = Params.TypeWeight_lifecycle
EstimationEconomy.T_retire = Params.working_T-1
EstimationEconomy.act_T = Params.T_sim_LC
EstimationEconomy.ignore_periods = Params.ignore_periods_LC
else:
EstimationEconomy.PopGroFac = 1.0
EstimationEconomy.TypeWeight = [1.0]
EstimationEconomy.act_T = Params.T_sim_PY
EstimationEconomy.ignore_periods = Params.ignore_periods_PY
if Params.do_agg_shocks:
EstimationEconomy(**Params.aggregate_params)
EstimationEconomy.update()
EstimationEconomy.makeAggShkHist()
# Estimate the model as requested
if Params.run_estimation:
# Choose the bounding region for the parameter search
if Params.param_name == 'CRRA':
param_range = [0.2,70.0]
spread_range = [0.00001,1.0]
elif Params.param_name == 'DiscFac':
param_range = [0.95,0.995]
spread_range = [0.006,0.008]
else:
print('Parameter range for ' + Params.param_name + ' has not been defined!')
if Params.do_param_dist:
# Run the param-dist estimation
paramDistObjective = lambda spread : findLorenzDistanceAtTargetKY(
Economy = EstimationEconomy,
param_name = Params.param_name,
param_count = Params.pref_type_count,
center_range = param_range,
spread = spread,
dist_type = Params.dist_type)
t_start = clock()
spread_estimate = golden(paramDistObjective,brack=spread_range,tol=1e-4)
center_estimate = EstimationEconomy.center_save
t_end = clock()
else:
# Run the param-point estimation only
paramPointObjective = lambda center : getKYratioDifference(Economy = EstimationEconomy,
param_name = Params.param_name,
param_count = Params.pref_type_count,
center = center,
spread = 0.0,
dist_type = Params.dist_type)
t_start = clock()
center_estimate = brentq(paramPointObjective,param_range[0],param_range[1],xtol=1e-6)
spread_estimate = 0.0
t_end = clock()
# Display statistics about the estimated model
#center_estimate = 0.986609223266
#spread_estimate = 0.00853886395698
EstimationEconomy.LorenzBool = True
EstimationEconomy.ManyStatsBool = True
EstimationEconomy.distributeParams(Params.param_name,Params.pref_type_count,center_estimate,spread_estimate,Params.dist_type)
EstimationEconomy.solve()
EstimationEconomy.calcLorenzDistance()
print('Estimate is center=' + str(center_estimate) + ', spread=' + str(spread_estimate) + ', took ' + str(t_end-t_start) + ' seconds.')
EstimationEconomy.center_estimate = center_estimate
EstimationEconomy.spread_estimate = spread_estimate
EstimationEconomy.showManyStats(Params.spec_name)
j += 1
# Only run below this line if module is run rather than imported:
if __name__ == "__main__":
# =================================================================
# ====== Make the list of consumer types for estimation ===========
#==================================================================
# Set target Lorenz points and K/Y ratio (MOVE THIS TO SetupParams)
if Params.do_liquid:
lorenz_target = np.array([0.0, 0.004, 0.025, 0.117])
KY_target = 6.60
else: # This is hacky until I can find the liquid wealth data and import it
lorenz_target = getLorenzShares(
Params.SCF_wealth,
weights=Params.SCF_weights,
percentiles=Params.percentiles_to_match)
#lorenz_target = np.array([-0.002, 0.01, 0.053,0.171])
KY_target = 10.26
# Make a vector of initial wealth-to-permanent income ratios
a_init = drawDiscrete(N=Params.sim_pop_size,
P=Params.a0_probs,
X=Params.a0_values,
seed=Params.a0_seed)
# Make the list of types for this run, whether infinite or lifecycle
if Params.do_lifecycle:
# Make cohort scaling array
cohort_scale = Params.TFP_growth**(-np.arange(Params.total_T + 1))
cohort_scale_array = np.tile(
return AgeDstn
###############################################################################
### ACTUAL WORK BEGINS BELOW THIS LINE #######################################
###############################################################################
if __name__ == '__main__':
# Set targets for K/Y and the Lorenz curve based on the data
if do_liquid:
lorenz_target = np.array([0.0, 0.004, 0.025, 0.117])
KY_target = 6.60
else: # This is hacky until I can find the liquid wealth data and import it
lorenz_target = getLorenzShares(
Params.SCF_wealth,
weights=Params.SCF_weights,
percentiles=Params.percentiles_to_match)
lorenz_long_data = np.hstack(
(np.array(0.0),
getLorenzShares(Params.SCF_wealth,
weights=Params.SCF_weights,
percentiles=np.arange(0.01, 1.0, 0.01).tolist()),
np.array(1.0)))
#lorenz_target = np.array([-0.002, 0.01, 0.053,0.171])
KY_target = 10.26
# Set total number of simulated agents in the population
if do_param_dist:
if do_agg_shocks:
Population = Params.pop_sim_agg_dist
else:
