def __init__(self, n_households):
self.n_households = n_households
super().__init__()
# Chose a schedule
self.schedule = RandomActivation(self)
self.production = 10
self.kapital_households = 0 # To change in a list of the sum
self.kapital_households_speculators = 0 # # To change in a list of the sum
self.sum_loans_households = self.n_households * loan_households
self.travail = 0
self.alpha = 0
self.beta = 0
# Create a bank, a firm and n household
# self.bank = Bank(1, self)
# self.schedule.add(bank)
# firm = Firm(2, self)
# self.schedule.add(firm)
x = risk_lovers_rate * self.n_households
for i in range(self.n_households):
if i <= x:
h = Household(i + 2, 1, self)
self.schedule.add(h)
else:
hp = Household(i + 2, -1, self)
self.schedule.add(hp)
def init_house(self, l_0, theta, gamma, phi, omega_p=None, f=None, r=None):
"""
Initialize a household object as instance of economy class. Refer to household class.
:param l_0: baseline work offer,
:param theta: preferency factors,
:param gamma: aversion to work parameter,
:param phi: Frisch index,
:param omega_p: confidence parameter
:param f: fraction of budget to save,
:param r: savings growth rate.
:return: Initializes household class with given parameters.
"""
self.house = Household(l_0, theta, gamma, phi, omega_p, f, r)
def _form_single_hh(self, t, ind):
"""
Move specified individual, with no partner, into their own household.
:param t: the current time step.
:type t: int
:param i_id: the first partner in the couple.
:type i_id: int
:returns: the ID of the new household.
"""
assert not ind.partner
# assert len(self.housemates(i_id)) > 0
# remove any guardians
self._remove_from_guardians(t, ind, \
'c-', "Lost dependent %d (leaving)" % ind.ID, ind.ID)
# remove from old household
self._remove_individual_from_hh(t, ind, 'l', "Individual left")
# add to newly created household
new_hh = self.add_group('household', [ind])
ind.with_parents = False
self.households[new_hh] = Household(t)
if self.logging:
ind.add_log(t, 'l1', "Leaving household (single)")
self.households[new_hh].add_log(t, 'f1', "Household formed", 1)
return new_hh
def duplicate_household(self, t, hh_id):
"""
Create a duplicate of household hh_id, consisting of
new individuals with same household age composition.
Return id of newly created household.
:param t: The current time step.
:type t: int
:param hh_id: The household to duplicate
:type hh_id: int
:returns the id of the newly created household.
"""
new_hh = []
for ind in self.groups['household'][hh_id]:
new_ind = self.add_individual(ind.age,
ind.sex,
logging=self.logging)
if self.logging:
new_ind.add_log(t, 'i', "Individual (immigrated)")
new_hh.append(new_ind)
new_hh_id = self.add_group('household', new_hh)
self.households[new_hh_id] = Household(t)
self._init_household(new_hh_id)
if self.logging:
self.households[new_hh_id].add_log(
t, 'i', "Household (immigrated)",
len(self.groups['household'][new_hh_id]))
return new_hh_id
def __init__(self, n_agents):
"""
Basic constructor with mostly default values
Parameters
n_agents : dict containing the number of each
type of agents that we want our model to include
"""
# "Private parameters"
# These parameters will be used to communicate to the agents that
# a new day or a new month started
#self.daypassed = False
self.monthpassed = False
# Attributes related to time
self.start_datetime = datetime(2005, 1, 1, 0, 0, 0, tzinfo=None)
self.current_datetime = self.start_datetime
self.step_interval = "month"
self.n_banks = n_agents['banks']
self.n_firms = n_agents['firms']
self.n_households = n_agents['households']
# Chose a scheduler
self.scheduler = RandomActivation(self)
# Create the required number of each agent
for i in range(self.n_banks):
a = Bank(i, self)
a.liquidity = 1000
self.scheduler.add(a)
for i in range(self.n_firms):
a = Firm(i + self.n_banks, self)
self.scheduler.add(a)
for i in range(self.n_households):
a = Household(i + self.n_banks + self.n_firms, self)
self.scheduler.add(a)
# If any deposits is to be given to banks it should be given now
# Init all firms with their employees
self.init_all_firms()
# Create the data collector
self.datacollector = DataCollector(
model_reporters = {"Household liquidity": agent2_liquidity,
"Household deposit": agent2_deposit,
"Bank liquidity": agent0_liquidity,
"Networth of household": agent2_netWorth,
"Networth of bank": agent0_netWorth,
"Firm net worth": agent1_netWorth})
self.running = True
def init_households(self):
employer_idx = 0
for hh in range(self.hh_param.get("num_hh")):
# first, hhs are distributed such that each firm has one employee
# afterwards, hhs are randomly assigned to an employer
# employers need to be informed who their employees are
employer = self.firm_list[employer_idx] if employer_idx < len(self.firm_list) else random.choice(self.firm_list)
new_household = Household(self, employer)
employer.list_employees.append(new_household)
self.hh_list.append(new_household)
employer_idx += 1
def gen_hh_age_structured_pop(self, pop_size, hh_probs, age_probs_i,
cutoffs, rng):
"""
Generate a population of individuals with age structure and household
composition.
Household composition here is approximated by the number of
individuals who are:
- pre-school age (0--4)
- school age (5--17)
- adult (18+)
This is a bit ad hoc, but serves current purposes.
:param pop_size: The size of the population to be generated.
:type pop_size: int
:param hh_probs: A table of household size probabilities.
:type hh_probs: list
:param age_probs_i: A table of age probabilities.
:type age_probs_i: list
:param rng: The random number generator to use.
:type rng: :class:`random.Random`
"""
age_probs = [[x, int(y[0])] for x, y in age_probs_i]
split_probs = split_age_probs(age_probs, cutoffs)
split_probs.reverse()
i = 0
while i < pop_size:
# get list of [adults, school age, preschool age]
hh_type = [int(x) for x in sample_table(hh_probs, rng)]
# hh_type = [int(x) for x in sample_uniform(hh_probs, rng)]
cur_hh = []
for cur_hh_type, cur_prob in zip(hh_type, split_probs):
for _ in itertools.repeat(None, cur_hh_type):
sex = rng.randint(0, 1)
cur_age = sample_table(cur_prob, rng)
cur_ind = self.add_individual(cur_age,
sex,
adam=True,
logging=self.logging)
if self.logging:
cur_ind.add_log(0, 'f', "Individual (bootstrap)")
cur_hh.append(cur_ind)
i += 1
hh_id = self.add_group('household', cur_hh)
self.households[hh_id] = Household(0, adam=True)
if self.logging:
self.households[hh_id].add_log(
0, 'f', "Household (bootstrap)",
len(self.groups['household'][hh_id]))
def _form_couple_hh(self, t, inds):
"""
Make specified single individuals into a couple and move them into a
new household. Any dependents of either individual accompany them to
the new household.
:param t: the current time step.
:type t: int
:param i_ids: the individuals to move into a couple household.
:type i_ids: list
:returns: the ID of the new household.
"""
ind_a = inds[0]
ind_b = inds[1]
inds[0].with_parents = False
inds[1].with_parents = False
# move dependents along with guardians
new_inds = list(inds)
combined_deps = ind_a.deps + ind_b.deps
ind_a.deps = combined_deps[:]
ind_b.deps = combined_deps[:]
new_inds.extend(combined_deps)
# remove any guardians of new couple
self._remove_from_guardians(t, inds[0], \
'c-', "Lost dependent %d (leaving)" % inds[0].ID, inds[0].ID)
self._remove_from_guardians(t, inds[1], \
'c-', "Lost dependent %d (leaving)" % inds[1].ID, inds[1].ID)
# remove individuals from prior households and create new household
for ind in new_inds:
self._remove_individual_from_hh(t, ind, 'l', "Individual left")
hh_id = self.add_group('household', new_inds)
self.households[hh_id] = Household(t)
if self.logging:
ind_a.add_log(t, 'l2',
"Leaving household - couple with %d" % ind_b.ID)
ind_b.add_log(t, 'l2',
"Leaving household - couple with %d" % ind_a.ID)
self.households[hh_id].add_log(t, 'f2', \
"Household formed (%d individuals)" % \
len(new_inds), len(new_inds))
return hh_id
def gen_single_hh_pop(self, hh_size, rng):
"""
Generate a dummy population consisting of a single household.
"""
cur_hh = []
for i in xrange(hh_size):
sex = rng.randint(0, 1)
cur_ind = self.add_individual(20,
sex,
adam=True,
logging=self.logging)
if self.logging:
cur_ind.add_log(0, 'f', "Individual (bootstrap)")
cur_hh.append(cur_ind)
hh_id = self.add_group('household', cur_hh)
self.households[hh_id] = Household(0, adam=True)
if self.logging:
self.households[hh_id].add_log(
0, 'f', "Household (bootstrap)",
len(self.groups['household'][hh_id]))
def agents_update(self):
'''
Update household(and person)'s demographic status; do everything except for marriage and child birth.
'''
temp_hh_list = list()
# Annual household status update
for HID in self.hh_dict:
temp_hh_list.append(Household.household_demographic_update(self.hh_dict[HID], self.current_year, self.model_parameters_dict))
# Reset the households dict
self.hh_dict = dict()
# Refresh hh_dict
for h in temp_hh_list:
self.hh_dict[h.HID] = h
if h.is_exist == 0: # If the household does not exist...
if h.is_dissolved_this_year == True: # ...and it becomes non-exist this year
# Dissolve non-existing households
self.dissolve_household(h.HID)
def agents_update(self):
'''
Update household(and person)'s demographic status; do everything except for marriage and child birth.
'''
temp_hh_list = list()
# Annual household status update
for HID in self.hh_dict:
temp_hh_list.append(Household.household_demographic_update(self.hh_dict[HID], self.current_year, self.model_parameters_dict))
# Reset the households dict
self.hh_dict = dict()
# Refresh hh_dict
for h in temp_hh_list:
self.hh_dict[h.HID] = h
if h.is_exist == 0: # If the household does not exist...
if h.is_dissolved_this_year == True: # ...and it becomes non-exist this year
# Dissolve non-existing households
self.dissolve_household(h.HID)
def setup_households(self):
if self.adaptive:
for i in range(self.init_households):
new_household = Household(model=self,
coords=self.random_coords(),
energy=self.init_energy,
move_threshold=random(),
move_dist=randint(1, 5),
farm_dist=randint(1, 20),
fission_rate=uniform(1, 2),
min_fertility=random())
new_household.claim_land()
else:
for i in range(self.init_households):
new_household = Household(model=self,
coords=self.random_coords(),
energy=self.init_energy,
move_threshold=(self.init_energy *
self.move_rate),
move_dist=randint(1, 5),
farm_dist=self.swidden_radius,
fission_rate=self.fission_energy,
min_fertility=0.8)
new_household.claim_land()
class Simulation():
def __init__(self, runParameters, firmParameters, householdParameters):
self.runParameters = runParameters
self.firmParameters = firmParameters
self.householdParameters = householdParameters
self.duration = runParameters['duration']
self.money = runParameters['money']
self.stock = runParameters['stock']
self.verbose = runParameters['verbose']
self.expectation = runParameters['expectation']
self.demandKnown = runParameters['demandKnown']
def equilibrium(self, money, householdParameters, firmParameters):
Lmax = householdParameters['Lmax']
alpha = householdParameters['alpha']
beta = householdParameters['beta']
gamma = firmParameters['gamma']
zeta_0 = firmParameters['zeta_0']
L_star = beta / (alpha + beta) * Lmax
p_star = L_star**(1 - gamma)
S_star = L_star**gamma
z_star = p_star * S_star**zeta_0
print(
'Equilibrium values are: p = {:.4f}, S = {:.4f}, L = {:.4f}, z = {:.4f}'
.format(p_star, S_star, L_star, z_star))
return (p_star, L_star, S_star, z_star)
def saveSimResults(self):
self.firmSimResults = self.firmSimResults.append(self.firm.__dict__,
ignore_index=True)
self.householdSimResults = self.householdSimResults.append(
self.household.__dict__, ignore_index=True)
self.marketSimResults = self.marketSimResults.append(
self.market.__dict__, ignore_index=True)
def advanceAttributes(self):
self.market.SM_t, self.market.LM_t = self.market.SM_tp1, self.market.LM_tp1
self.firm.LD_t, self.firm.SSP_t, self.firm.SS_t = self.firm.LD_tp1, self.firm.SSP_tp1, self.firm.SS_tp1
self.firm.p_t = self.firm.p_tp1
self.firm.z_t, self.firm.zeta_t = self.firm.z_tp1, self.firm.zeta_tp1
if self.stock: self.firm.stock_t = self.firm.stock_tp1
if self.money: self.firm.mF_t = self.firm.mF_tp1
self.household.LS_t, self.household.SD_t = self.household.LS_tp1, self.household.SD_tp1
if self.money: self.household.mH_t = self.household.mH_tp1
def Run(self):
self.p_star, self.L_star, self.S_star, self.z_star = self.equilibrium(
self.money, self.householdParameters, self.firmParameters)
self.firm = Firm(self.money, self.firmParameters, self.p_star,
self.S_star, self.L_star, self.z_star,
self.expectation)
self.household = Household(self.householdParameters, self.p_star,
self.S_star, self.z_star, self.firm.zeta_0)
self.market = Market(self.S_star, self.L_star)
self.firmSimResults = pd.DataFrame()
self.householdSimResults = pd.DataFrame()
self.marketSimResults = pd.DataFrame()
for t in range(self.duration):
if self.verbose: print('\nstep ', str(t + 1))
#add time step as instance attribute
self.market.step = t + 1
self.firm.step = t + 1
self.household.step = t + 1
#firms plan sugar production
self.firm.decideProduction(self.verbose, self.money,
self.market.SM_t, self.household.Lmax)
#households decide labor supply
self.household.decideLabor(self.verbose)
#market decides labor supply
self.market.laborTransaction(self.verbose, self.firm.LD_tp1,
self.household.LS_tp1)
#firms produce sugar
self.firm.produce(self.verbose, self.market.LM_tp1)
#households decide sugar consumption
self.household.decideConsumption(self.verbose, self.money,
self.market.LM_tp1,
self.firm.p_tp1)
#market decides demand for sugar
self.market.stuffTransaction(self.verbose, self.firm.SS_tp1,
self.household.SD_tp1)
#update firm stock
if self.stock:
self.firm.updateStock(self.verbose, self.market.SM_tp1)
#update firm ledgers
if self.money:
self.firm.updateLedger(self.verbose, self.market.SM_tp1,
self.market.LM_tp1)
#no labor curve updating first round
if t == 0: self.market.LM_t = self.market.LM_tp1
#update sugar demand an labor supply function parameters
if self.demandKnown:
self.firm.updateExpectations(self.verbose, self.money,
self.household.SD_t,
self.household.SD_tp1,
self.market.LM_tp1,
self.market.LM_t,
self.market.LM_tp1)
else:
self.firm.updateExpectations(self.verbose, self.money,
self.market.SM_t,
self.market.SM_tp1,
self.market.LM_tp1,
self.market.LM_t,
self.market.LM_tp1)
#update household ledgers
if self.money:
self.household.updateLedger(self.verbose, self.firm.p_tp1,
self.market.SM_tp1,
self.market.LM_tp1)
#save step results
self.saveSimResults()
# update timed variables
self.advanceAttributes()
#print('step', t + 1, 'complete')
print(
'Final simulation values are: p = {:.4f}, S = {:.4f}, L = {:.4f}'.
format(self.firm.p_tp1, self.market.SM_t, self.market.LM_t))
return self.firmSimResults, self.householdSimResults, self.marketSimResults
class Economy:
def __init__(self, n, d, netstring, directed, j0, a0, q, b):
# Network initialization
self.n = n
self.j = create_net(netstring, directed, n, d)
self.j0 = j0
self.j_a = None
self.a0 = a0
a = np.multiply(np.random.uniform(0, 1, (n, n)), self.j)
self.a = np.array([(1 - a0[i]) * a[i] / np.sum(a[i])
for i in range(n)])
self.a_a = None
self.q = q
self.zeta = 1 / (q + 1)
self.b = b
# Auxiliary network variables
self.lamb = None
self.a_a = None
self.j_a = None
self.lamb_a = None
self.m_cal = None
self.v = None
self.kappa = None
self.zeros_j_a = None
# Firms and household sub-classes
self.firms = None
self.house = None
# Equilibrium quantities
self.p_eq = None
self.g_eq = None
self.mu_eq = None
self.labour_eq = None
self.cons_eq = None
self.b_eq = None
self.utility_eq = None
def init_house(self, l_0, theta, gamma, phi, omega_p=None, f=None, r=None):
"""
Initialize a household object as instance of economy class. Refer to household class.
:param l_0: baseline work offer,
:param theta: preferency factors,
:param gamma: aversion to work parameter,
:param phi: Frisch index,
:param omega_p: confidence parameter
:param f: fraction of budget to save,
:param r: savings growth rate.
:return: Initializes household class with given parameters.
"""
self.house = Household(l_0, theta, gamma, phi, omega_p, f, r)
def init_firms(self, z, sigma, alpha, alpha_p, beta, beta_p, omega):
"""
Initialize a firms object as instance of economy class. Refer to firms class.
:param z: Productivity factors,
:param sigma: Depreciation of stocks parameters,
:param alpha: Log-elasticity of prices' growth rates against surplus,
:param alpha_p: Log-elasticity of prices' growth rates against profits,
:param beta: Log-elasticity of productions' growth rates against profits,
:param beta_p: Log-elasticity of productions' growth rates against surplus,
:param omega: Log-elasticity of wages' growth rates against labor-market tensions.
:return: Initializes firms class with given parameters.
"""
self.firms = Firms(z, sigma, alpha, alpha_p, beta, beta_p, omega)
# Setters for class instances
def set_house(self, house):
"""
Sets a household object as instance of economy class.
:param house:
:return:
"""
self.house = house
def set_firms(self, firms):
"""
Sets a firms object as instance of economy class.
:param firms:
:return:
"""
self.firms = firms
# Update methods for firms and household
def update_firms_z(self, z):
self.firms.update_z(z)
self.set_quantities()
self.compute_eq()
def update_firms_sigma(self, sigma):
self.firms.update_sigma(sigma)
def update_firms_alpha(self, alpha):
self.firms.update_alpha(alpha)
def update_firms_alpha_p(self, alpha_p):
self.firms.update_alpha_p(alpha_p)
def update_firms_beta(self, beta):
self.firms.update_beta(beta)
def update_firms_beta_p(self, beta_p):
self.firms.update_beta_p(beta_p)
def update_firms_w(self, omega):
self.firms.update_w(omega)
def update_house_labour(self, labour):
self.house.update_labour(labour)
self.compute_eq()
def update_house_theta(self, theta):
self.house.update_theta(theta)
self.compute_eq()
def update_house_gamma(self, gamma):
self.house.update_gamma(gamma)
self.compute_eq()
def update_house_phi(self, phi):
self.house.update_phi(phi)
self.compute_eq()
def update_house_w_p(self, omega_p):
self.house.update_w_p(omega_p)
def update_house_f(self, f):
self.house.update_f(f)
def update_house_r(self, r):
self.house.update_r(r)
# Setters for the networks and subsequent instances
def set_j(self, j):
"""
Sets a particular input-output network.
:param j: a n by n matrix.
:return: side effect
"""
if j.shape != (self.n, self.n):
raise ValueError('Input-output network must be of size (%d, %d)' %
(self.n, self.n))
self.j = j
self.set_quantities()
self.compute_eq()
def set_a(self, a):
"""
Sets a particular input-output network.
:param a: a n by n matrix.
:return: side effect
"""
if a.shape != (self.n, self.n):
raise ValueError('Substitution network must be of size (%d, %d)' %
(self.n, self.n))
self.a = a
self.set_quantities()
self.compute_eq()
def set_quantities(self):
"""
Sets redundant economy quantities as class instances.
:return: side effect
"""
if self.q == 0:
self.lamb = self.j
self.a_a = np.hstack((np.array([self.a0]).T, self.a))
self.j_a = np.hstack((np.array([self.j0]).T, self.j))
self.lamb_a = self.j_a
self.m_cal = np.diag(self.firms.z) - self.lamb
self.v = np.array(self.lamb_a[:, 0])
elif self.q == np.inf:
self.lamb = self.a
self.a_a = np.hstack((np.array([self.a0]).T, self.a))
self.j_a = np.hstack((np.array([self.j0]).T, self.j))
self.lamb_a = self.a_a
self.m_cal = np.eye(self.n) - self.lamb
self.v = np.array(self.lamb_a[:, 0])
else:
self.lamb = np.multiply(np.power(self.a, self.q * self.zeta),
np.power(self.j, self.zeta))
self.a_a = np.hstack((np.array([self.a0]).T, self.a))
self.j_a = np.hstack((np.array([self.j0]).T, self.j))
self.lamb_a = np.multiply(np.power(self.a_a, self.q * self.zeta),
np.power(self.j_a, self.zeta))
self.m_cal = np.diag(np.power(self.firms.z, self.zeta)) - self.lamb
self.v = np.array(self.lamb_a[:, 0])
self.mu_eq = np.power(
np.power(self.house.gamma, 1. / self.house.phi) *
np.sum(self.house.theta) * (1 - (1 - self.house.f) *
(1 + self.house.r)) /
(self.house.f * np.power(self.house.l_0, 1 + 1. / self.house.phi)),
self.house.phi / (1 + self.house.phi))
self.kappa = self.house.theta / self.mu_eq
self.zeros_j_a = self.j_a != 0
def get_eps_cal(self):
"""
Computes the smallest eigenvalue of the economy matrix
:return: smallest eigenvalue
"""
return np.min(np.real(np.linalg.eigvals(self.m_cal)))
def set_eps_cal(self, eps):
"""
Modifies firms instance to set smallest eigenvalue of economy matrix to given epsilon.
:param eps: a real number,
:return: side effect.
"""
min_eig = self.get_eps_cal()
z_n = self.firms.z * np.power(
1 +
(eps - min_eig) / np.power(self.firms.z, self.zeta), self.q + 1)
sigma = self.firms.sigma
alpha = self.firms.alpha
alpha_p = self.firms.alpha_p
beta = self.firms.beta
beta_p = self.firms.beta_p
omega = self.firms.omega
self.init_firms(z_n, sigma, alpha, alpha_p, beta, beta_p, omega)
self.set_quantities()
self.compute_eq()
def update_b(self, b):
"""
Sets return to scale parameter
:param b: return to scale
:return: side effect
"""
self.b = b
self.compute_eq()
def update_q(self, q):
self.q = q
self.zeta = 1 / (q + 1)
self.set_quantities()
self.compute_eq()
def update_network(self, netstring, directed, d, n):
self.j = create_net(netstring, directed, n, d)
a = np.multiply(np.random.uniform(0, 1, (n, n)), self.j)
self.a = np.array([(1 - self.a0[i]) * a[i] / np.sum(a[i])
for i in range(n)])
self.set_quantities()
self.compute_eq()
def update_a0(self, a0):
self.a0 = a0
a = np.multiply(np.random.uniform(0, 1, (self.n, self.n)), self.j)
self.a = np.array([(1 - self.a0[i]) * a[i] / np.sum(a[i])
for i in range(self.n)])
self.set_quantities()
self.compute_eq()
def update_j0(self, j0):
self.j0 = j0
self.set_quantities()
self.compute_eq()
def production_function(self, q_available):
"""
CES production function.
:param q_available: matrix of available labour and goods for production,
:return: production levels of the firms.
"""
if self.q == 0:
return np.power(
np.nanmin(np.divide(q_available, self.j_a), axis=1), self.b)
elif self.q == np.inf:
return np.power(
np.nanprod(np.power(np.divide(q_available, self.j_a),
self.a_a),
axis=1), self.b)
else:
return np.power(
np.nansum(self.a_a * np.power(self.j_a, 1. / self.q) /
np.power(q_available, 1. / self.q),
axis=1), -self.b * self.q)
def compute_eq(self):
"""
Computes the competitive equilibrium of the economy. We use least-squares to compute solutions of linear
systems Ax=b for memory and computational efficiency. The non-linear equations for non-constant return to scale
parameters are solved using generalized least-squares with initial guesses taken to be the solution of the b=1
linear equation. For a high number of firms, high heterogeneity of close to 0 epsilon, this function might
can output erroneous results or errors.
:return: side effect.
"""
if self.q == np.inf:
h = np.sum(
self.a_a *
np.log(np.ma.masked_invalid(np.divide(self.j_a, self.a_a))),
axis=1)
v = lstsq(np.eye(self.n) - self.a.T, self.kappa, rcond=10e-7)[0]
log_p = lstsq(np.eye(self.n) / self.b - self.a,
-np.log(self.firms.z) / self.b +
(1 - self.b) * np.log(v) / self.b + h,
rcond=10e-7)[0]
log_g = -np.log(self.firms.z) - log_p + np.log(v)
self.p_eq, self.g_eq = np.exp(log_p), np.exp(log_g)
else:
if self.b != 1:
if self.q == 0:
init_guess_peq = lstsq(self.m_cal, self.v, rcond=10e-7)[0]
init_guess_geq = lstsq(self.m_cal.T,
np.divide(self.kappa,
init_guess_peq),
rcond=10e-7)[0]
par = (self.firms.z, self.v, self.m_cal, self.b - 1,
self.kappa)
pert_peq = lstsq(self.m_cal,
self.firms.z * init_guess_peq *
np.log(init_guess_geq),
rcond=10e-7)[0]
pert_geq = lstsq(
np.transpose(self.m_cal),
-np.divide(self.kappa, np.power(init_guess_peq, 2)) *
pert_peq +
self.firms.z * init_guess_geq * np.log(init_guess_geq),
rcond=10e-7)[0]
pg = leastsq(
lambda x: self.non_linear_eq_qzero(x, *par),
np.array(
np.concatenate(
(init_guess_peq + (1 - self.b) * pert_peq,
np.power(
init_guess_geq + (1 - self.b) *
(pert_geq -
init_guess_geq * np.log(init_guess_geq)),
1 / self.b))).reshape(2 * self.n)))[0]
# pylint: disable=unbalanced-tuple-unpacking
self.p_eq, g = np.split(pg, 2)
self.g_eq = np.power(g, self.b)
else:
# The numerical solving is done for variables u = p_eq ^ zeta and
# w = z ^ (q * zeta) * u ^ q * g_eq ^ (zeta * (bq+1) / b)
init_guess_u = lstsq(self.m_cal, self.v, rcond=None)[0]
init_guess_w = lstsq(self.m_cal.T,
np.divide(self.kappa, init_guess_u),
rcond=None)[0]
par = (np.power(self.firms.z,
self.zeta), self.v, self.m_cal, self.q,
(self.b - 1) / (self.b * self.q + 1), self.kappa)
uw = leastsq(
lambda x: self.non_linear_eq_qnonzero(x, *par),
np.concatenate((init_guess_u, init_guess_w)),
)[0]
# pylint: disable=unbalanced-tuple-unpacking
u, w = np.split(uw, 2)
self.p_eq = np.power(u, 1. / self.zeta)
self.g_eq = np.power(
np.divide(
w,
np.power(self.firms.z, self.q * self.zeta) *
np.power(u, self.q)),
self.b / (self.zeta * (self.b * self.q + 1)))
else:
if self.q == 0:
self.p_eq = lstsq(self.m_cal, self.v, rcond=10e-7)[0]
self.g_eq = lstsq(self.m_cal.T,
np.divide(self.kappa, self.p_eq),
rcond=10e-7)[0]
else:
# The numerical solving is done for variables u = p_eq ^ zeta and
# w = z ^ (q * zeta) * u ^ q * g_eq
u = lstsq(self.m_cal, self.v, rcond=None)[0]
self.p_eq = np.power(u, 1. / self.zeta)
w = lstsq(self.m_cal.T,
np.divide(self.kappa, u),
rcond=None)[0]
self.g_eq = np.divide(
w,
np.power(self.firms.z, self.q * self.zeta) *
np.power(u, self.q))
self.labour_eq = np.power(self.mu_eq * self.house.f,
1. / self.house.phi) / self.house.v_phi
self.cons_eq = self.kappa / self.p_eq
self.b_eq = np.sum(self.house.theta) / self.mu_eq
self.utility_eq = np.dot(self.house.theta, np.log(
self.cons_eq)) - self.house.gamma * np.power(
self.labour_eq / self.house.l_0,
self.house.phi + 1) / (self.house.phi + 1)
def save_eco(self, name):
"""
Saves the economy as multi-indexed data-frame in hdf format along with networks in npy format.
:param name: name of file,
"""
first_index = np.concatenate(
(np.repeat('Firms', 11), np.repeat('Household', 11)))
second_index = np.concatenate(([
'q', 'b', 'z', 'sigma', 'alpha', 'alpha_p', 'beta', 'beta_p', 'w',
'p_eq', 'g_eq'
], ['l', 'theta', 'gamma', 'phi']))
multi_index = [first_index, second_index]
values = np.vstack((
self.q * np.ones(self.n),
self.b * np.ones(self.n),
self.firms.z,
self.firms.sigma,
self.firms.alpha * np.ones(self.n),
self.firms.alpha_p * np.ones(self.n),
self.firms.beta * np.ones(self.n),
self.firms.beta_p * np.ones(self.n),
self.firms.w * np.ones(self.n),
self.p_eq,
self.g_eq,
self.house.l_0 * np.ones(self.n),
self.house.theta,
self.house.gamma * np.ones(self.n),
self.house.phi * np.ones(self.n),
))
df_eco = pd.DataFrame(values,
index=multi_index,
columns=[np.arange(1, self.n + 1)])
df_eco.to_hdf(name + '/eco.h5', key='df', mode='w')
np.save(name + '/network.npy', self.j_a)
if self.q != 0:
np.save(name + '/sub_network.npy', self.a_a)
# Fixed point equations for equilibrium computation
@staticmethod
def non_linear_eq_qnonzero(x, *p):
"""
Function used for computation of equilibrium with non constant return to scale
and general CES production function.
:param x: guess for equilibrium
:param p: tuple z, z_zeta, v, m_cal, zeta, q, theta, theta_bar, power
:return: function's value at x
"""
if len(x) % 2 != 0:
raise ValueError("x must be of even length")
# pylint: disable=unbalanced-tuple-unpacking
u, w = np.split(x, 2)
z_zeta, v, m_cal, q, exponent, kappa = p
w_over_uq_p = np.power(
np.divide(w,
np.power(z_zeta, q) * np.power(u, q)), exponent)
v1 = np.multiply(z_zeta, np.multiply(u, 1 - w_over_uq_p))
m1 = np.dot(m_cal, u)
m2 = u * np.dot(m_cal.T, w) - w * m1
return np.concatenate((m1 - v1 - v, m2 + w * v - kappa))
@staticmethod
def non_linear_eq_qzero(x, *par):
"""
Function used for computation of equilibrium with non constant return to scale
and Leontieff production function.
:param x: guess for equilibrium
:param par: tuple z_zeta, v, m_cal, q power
:return: function's value at x
"""
if len(x) % 2 != 0:
raise ValueError('x must be of even length')
# pylint: disable=unbalanced-tuple-unpacking
p, g = np.split(x, 2)
z, v, m_cal, exponent, kappa = par
v1 = np.multiply(z, np.multiply(p, 1 - np.power(g, exponent)))
m1 = np.dot(m_cal, p)
m2 = g * m1 - p * np.dot(m_cal.T, g)
return np.concatenate((m1 - v1 - v, m2 - g * v + kappa))
def __init__(self, db, model_table_name, model_table, hh_table_name, hh_table, pp_table_name, pp_table,
land_table_name, land_table, business_sector_table_name, business_sector_table,
policy_table_name, policy_table, stat_table_name, stat_table, start_year):
'''
Initialize the society class;
'''
# Set the start year and current time stamps
self.start_year = start_year
self.current_year = start_year
# Create a dictionary to store model parameters, indexed by Variable_Name, and contents are Variable_Value
self.model_parameters_dict = dict()
# Fill in the model parameters dictionary from the model table (fetched from DB)
for record in model_table:
self.model_parameters_dict[record.Variable_Name] = record.Variable_Value
# Get the variable lists for household, person, land, business sector, policy, and statistics classes
self.hh_var_list = DataAccess.get_var_list(db, hh_table_name)
self.pp_var_list = DataAccess.get_var_list(db, pp_table_name)
self.land_var_list = DataAccess.get_var_list(db, land_table_name)
self.business_sector_var_list = DataAccess.get_var_list(db, business_sector_table_name)
self.policy_var_list = DataAccess.get_var_list(db, policy_table_name)
self.stat_var_list = DataAccess.get_var_list(db, stat_table_name)
# Initialize the land instances, and create a land dictionary to store all the land parcels
# Note that the land parcels here do not necessarily belong to any household. i.e. land_parcel.HID could be "None".
self.land_dict = dict()
for land in land_table:
land_parcel = Land(land, self.land_var_list, self.current_year)
self.land_dict[land_parcel.ParcelID] = land_parcel # Indexed by ParcelID
# Initialize the household instances (household capital property and land class instances are initialized at the initialization of household class);
# And create a dictionary to store them, indexed by HID.
self.hh_dict = dict()
# Add household instances to hh_dict
for hh in hh_table:
hh_temp = Household(hh, self.hh_var_list, self.current_year, db, pp_table_name, pp_table,
self.land_dict, self.model_parameters_dict)
self.hh_dict[hh_temp.HID] = hh_temp # Indexed by HID
# Initialize the business sector instances;
# And create a dictionary to store them, indexed by sector name.
self.business_sector_dict = dict()
for sector in business_sector_table:
sector_temp = BusinessSector(sector, self.business_sector_var_list)
self.business_sector_dict[sector_temp.SectorName] = sector_temp # Indexed by SectorName
# Initialize the policy program instances;
# And create a dictionary to store them, indexed by policy program type.
self.policy_dict = dict()
for program in policy_table:
program_temp = Policy(program, self.policy_var_list)
self.policy_dict[program_temp.PolicyType] = program_temp # Indexed by PolicyType
# Create a statistics dictionary; indexed by Variable Names.To be filled later in Statistics Class.
self.stat_dict = dict()
# Create some variables to record the confiscated (due to ownerlessness) money and land
self.ownerless_land = list()
self.ownerless_money = 0
def editHousehold(self, hhid, newhhid, householdname, dateofcollection):
household = Household( self.pid, hhid )
household.setData( householdname, dateofcollection, newhhid )
return household
def Run(self):
self.p_star, self.L_star, self.S_star, self.z_star = self.equilibrium(
self.money, self.householdParameters, self.firmParameters)
self.firm = Firm(self.money, self.firmParameters, self.p_star,
self.S_star, self.L_star, self.z_star,
self.expectation)
self.household = Household(self.householdParameters, self.p_star,
self.S_star, self.z_star, self.firm.zeta_0)
self.market = Market(self.S_star, self.L_star)
self.firmSimResults = pd.DataFrame()
self.householdSimResults = pd.DataFrame()
self.marketSimResults = pd.DataFrame()
for t in range(self.duration):
if self.verbose: print('\nstep ', str(t + 1))
#add time step as instance attribute
self.market.step = t + 1
self.firm.step = t + 1
self.household.step = t + 1
#firms plan sugar production
self.firm.decideProduction(self.verbose, self.money,
self.market.SM_t, self.household.Lmax)
#households decide labor supply
self.household.decideLabor(self.verbose)
#market decides labor supply
self.market.laborTransaction(self.verbose, self.firm.LD_tp1,
self.household.LS_tp1)
#firms produce sugar
self.firm.produce(self.verbose, self.market.LM_tp1)
#households decide sugar consumption
self.household.decideConsumption(self.verbose, self.money,
self.market.LM_tp1,
self.firm.p_tp1)
#market decides demand for sugar
self.market.stuffTransaction(self.verbose, self.firm.SS_tp1,
self.household.SD_tp1)
#update firm stock
if self.stock:
self.firm.updateStock(self.verbose, self.market.SM_tp1)
#update firm ledgers
if self.money:
self.firm.updateLedger(self.verbose, self.market.SM_tp1,
self.market.LM_tp1)
#no labor curve updating first round
if t == 0: self.market.LM_t = self.market.LM_tp1
#update sugar demand an labor supply function parameters
if self.demandKnown:
self.firm.updateExpectations(self.verbose, self.money,
self.household.SD_t,
self.household.SD_tp1,
self.market.LM_tp1,
self.market.LM_t,
self.market.LM_tp1)
else:
self.firm.updateExpectations(self.verbose, self.money,
self.market.SM_t,
self.market.SM_tp1,
self.market.LM_tp1,
self.market.LM_t,
self.market.LM_tp1)
#update household ledgers
if self.money:
self.household.updateLedger(self.verbose, self.firm.p_tp1,
self.market.SM_tp1,
self.market.LM_tp1)
#save step results
self.saveSimResults()
# update timed variables
self.advanceAttributes()
#print('step', t + 1, 'complete')
print(
'Final simulation values are: p = {:.4f}, S = {:.4f}, L = {:.4f}'.
format(self.firm.p_tp1, self.market.SM_t, self.market.LM_t))
return self.firmSimResults, self.householdSimResults, self.marketSimResults
def create_households(self):
"""Creates X number of households and adds them to list of households."""
for i, _ in enumerate(range(self.NEIGHBORHOOD_APARTMENTS), 1):
household = Household(i)
self.households.append(household)
def __init__(self, n_households, df3):
self.n_households = n_households
self.df3 = pd.read_excel(df3)
super().__init__()
# Chose a schedule
self.schedule = RandomActivation(self)
self.production = 0
self.kapital_households = []
self.kapital_households_speculators = []
self.kapital_global_btcModel = []
self.kapital_global = 7.5 * n_households
self.kh_high = 0
self.kh_medium = 0
self.kh_low = 0
self.kh = 0
self.sum_speculator_portfolio = 0
self.sum_wages_households = 0
self.sum_consumption_households = 0
self.sum_loans_households = 0
self.travail = 1000
self.alpha = 0.5
self.beta = 0.5
self.techno = 1.3 # technology factor
self.gamma = 0.67 # coefficient of production applie to salaries 0.67
self.n = 12
self.start_datetime = datetime(2017, 1, 1, tzinfo=None)
self.current_datetime = self.start_datetime
self.datacollector = DataCollector(model_reporters={
"Production": production,
"KapitalG": evolution_kapital_global,
"KapitalH": kapital_new,
"Kapital_households_risk_averse":
graph_households_risk_averse_kapital,
"Kapital_households_speculator":
graph_households_speculator_kapital,
"WagesH": graph_households_wage,
"sum_wage": increase_wages_households,
"Conso": graph_households_conso,
"KapitalF": graph_kapital_firm,
"KapitalB": graph_kapital_bank,
"diff": diff
},
agent_reporters={"Wage": "wage"})
self.running = True
# Create a bank, a firm and n household
self.firm = Firm(2, self)
self.bank = Bank(1, self)
xh = risk_high_rate * self.n_households
xm = xh + risk_medium_rate * self.n_households
xl = xm + risk_low_rate * self.n_households
for i in range(self.n_households):
if i < xh:
h = Household(i + 2, 2, P2, self)
self.schedule.add(h)
elif i >= xh and i < xm:
h = Household(i + 2, 1, P1, self)
self.schedule.add(h)
elif i >= xm and i < xl:
h = Household(i + 2, 0, P0, self)
self.schedule.add(h)
else:
h = Household(i + 2, -1, 0, self)
self.schedule.add(h)
#Init production with kapital initialisation of agents and give loans
production(self)
self.bank.give_loan()