I am currently working in my doctoral thesis coding. The paper deals with selective default, based on Arellano (2008). Got inspiration from @quantecon notebooks.
I'm trying to create a function in Python using Numba that iterates a bunch of variables, but, one of them is conditional. That is, depending on the probability of a state of nature, the iteration must be different. I have tried to use the if/else statement, but Python points syntax error. The code is a value function iteration for my doctoral thesis. Here what I've wrote:
@jit(nopython=True)
def solve(model, tol=1e-8, maxiter=10_000):
"""
Given an Selective_Economy type, this function computes the optimal
policy and value functions
"""
# Unpack certain parameters for simplification
β, σ, r, θ_d, θ_f = model.β, model.σ, model.r, model.θ_d, model.θ_f
b_d = np.ascontiguousarray(model.b_d)
b_f = np.ascontiguousarray(model.b_f)
A, y = np.ascontiguousarray(model.A), np.ascontiguousarray(model.y)
nb_d, nb_f, ny = b_d.size, b_f.size, y.size
# Allocate space
ib_d_star = np.zeros((ny, nb_d), int64)
ib_f_star = np.zeros((ny, nb_f), int64)
dom_default_prob = np.zeros((ny, nb_d))
ext_default_prob = np.zeros((ny, nb_f))
dom_default_states = np.zeros((ny, nb_d))
ext_default_states = np.zeros((ny, nb_f))
tot_default_states = np.zeros((ny, nb_d, nb_f))
q_f = np.ones((ny, nb_f)) * 0.95
q_d = np.ones((ny, nb_d)) * 0.95
Vfd = np.zeros(ny, nb_d)
Vdd = np.zeros(ny, nb_f)
Vtd = np.zeros(ny)
Vc, V, Vupd = np.zeros((ny, nb_d,nb_f)), np.zeros((ny, nb_d,nb_f)), np.zeros((ny, nb_d,nb_f))
it = 0
dist = 10.0
while (it < maxiter) and (dist > tol):
# Compute expectations used for this iteration
EV = A@V
EVfd = A@Vfd
EVdd = A@Vdd
EVtd = A@Vtd
for iy in range(ny):
# Update value function for default state
Vfd[iy] = model.bellman_ext_default(iy, EVfd, EV)
Vdd[iy] = model.bellman_dom_default(iy, EVdd, EV)
Vtd[iy] = model.bellman_tot_default(iy, EVtd, EV)
for ib_d, ib_f in range(nb_d), range(nb_f):
# Update value function for non-default state
ib_d_star[iy, ib_d] = model.compute_savings_policy(iy, ib_d, q_d, EV)
ib_f_star[iy, ib_f] = model.compute_savings_policy(iy, ib_f, q_d, EV)
Vc[iy, ib_d,ib_f] = model.bellman_nondefault(iy, ib_d, ib_f, q_d, q_f, EV, ib_d_star[iy, ib_d], ib_f_star[iy,ib_f])
# Once value functions are updated, can combine them to get
# the full value function
Vfd_compat = np.reshape(np.repeat(Vfd, nb_f), (ny, nb_f))
Vdd_compat = np.reshape(np.repeat(Vdd, nb_d), (ny, nb_d))
Vtd_compat = np.reshape(np.repeat(Vtd, nb_d, nb_f), (ny, nb_d, nb_f))
Vupd[:, :] = np.maximum(Vc, Vfd_compat, Vdd_compat, Vtd_compat)
# Can also compute default states and update prices
ext_default_states[:, :] = 1.0 * (Vfd_compat > max(Vc, Vdd, Vtd))
dom_default_states[:, :] = 1.0 * (Vdd_compat > max(Vc, Vfd, Vtd))
tot_default_states[:, :] = 1.0 * (Vtd_compat > max(Vc, Vfd, Vdd))
ext_default_prob[:, :] = A @ ext_default_states
dom_default_prob[:, :] = A @ dom_default_states
if dom_default_prob = 0:
q_f[:, :] = (1 - ext_default_prob) / (1 + r)
else:
q_f[:, :] = θ_f * (1 - ext_default_prob) / (1 + r) + (1 - θ_f) (1 - ext_default_prob) / (1 + r)
if ext_default_prob = 0:
q_d[:, :] = β * (1 - ext_default_prob) * (1 - dom_default_prob) * u_prime(y[iy]) + ext_default_prob * (1 - dom_default_prob) * u_prime(ext_y[iy]) / u_prime(y[iy])
else:
q_d[:, :] = β * θ_f * (1 - dom_default_prob) * u_prime(y[iy]) + (1 - θ_f) * (1 - dom_default_prob) * u_prime(ext_y[iy]) / u_prime(ext_y[iy])
# Check tolerance etc...
dist = np.max(np.abs(Vupd - V))
V[:, :] = Vupd[:, :]
it += 1
return V, Vc, Vfd, Vdd, Vtd, ib_d_star, ib_f_star, ext_default_prob, dom_default_prob, ext_default_states, dom_default_states, tot_default_states, q_f, q_d
The error is in the last part, where I receive the following message:
File "<ipython-input-3-cf2f65d62110>", line 65
if dom_default_prob = 0:
SyntaxError: invalid syntax
If anyone help, I would be greateful!
Thanks :)
