31. Job Search III: Search with Learning#
Contents
31.1. Overview#
In this lecture we consider an extension of the previously studied job search model of McCall [McC70].
In the McCall model, an unemployed worker decides when to accept a permanent position at a specified wage, given
his or her discount rate
the level of unemployment compensation
the distribution from which wage offers are drawn
In the version considered below, the wage distribution is unknown and must be learned.
The following is based on the presentation in [LS18], section 6.6.
31.1.1. Model features#
Infinite horizon dynamic programming with two states and one binary control.
Bayesian updating to learn the unknown distribution.
31.2. Model#
Let’s first review the basic McCall model [McC70] and then add the variation we want to consider.
31.2.1. The Basic McCall Model#
Recall that, in the baseline model, an unemployed worker is presented in each period with a permanent job offer at wage \(W_t\).
At time \(t\), our worker either
accepts the offer and works permanently at constant wage \(W_t\)
rejects the offer, receives unemployment compensation \(c\) and reconsiders next period
The wage sequence \(\{W_t\}\) is iid and generated from known density \(h\).
The worker aims to maximize the expected discounted sum of earnings \(\mathbb{E} \sum_{t=0}^{\infty} \beta^t y_t\) The function \(V\) satisfies the recursion
The optimal policy has the form \(\mathbf{1}\{w \geq \bar w\}\), where \(\bar w\) is a constant depending called the reservation wage.
31.2.2. Offer Distribution Unknown#
Now let’s extend the model by considering the variation presented in [LS18], section 6.6.
The model is as above, apart from the fact that
the density \(h\) is unknown
the worker learns about \(h\) by starting with a prior and updating based on wage offers that he/she observes
The worker knows there are two possible distributions \(F\) and \(G\) — with densities \(f\) and \(g\).
At the start of time, “nature” selects \(h\) to be either \(f\) or \(g\) — the wage distribution from which the entire sequence \(\{W_t\}\) will be drawn.
This choice is not observed by the worker, who puts prior probability \(\pi_0\) on \(f\) being chosen.
Update rule: worker’s time \(t\) estimate of the distribution is \(\pi_t f + (1 - \pi_t) g\), where \(\pi_t\) updates via
This last expression follows from Bayes’ rule, which tells us that
The fact that (31.2) is recursive allows us to progress to a recursive solution method.
Letting
we can express the value function for the unemployed worker recursively as follows
Notice that the current guess \(\pi\) is a state variable, since it affects the worker’s perception of probabilities for future rewards.
31.2.3. Parameterization#
Following section 6.6 of [LS18], our baseline parameterization will be
\(f\) is \(\operatorname{Beta}(1, 1)\) scaled (i.e., draws are multiplied by) some factor \(w_m\)
\(g\) is \(\operatorname{Beta}(3, 1.2)\) scaled (i.e., draws are multiplied by) the same factor \(w_m\)
\(\beta = 0.95\) and \(c = 0.6\)
With \(w_m = 2\), the densities \(f\) and \(g\) have the following shape
using LinearAlgebra, Statistics
using Distributions, LaTeXStrings, Plots, QuantEcon, Interpolations
w_max = 2
x = range(0, w_max, length = 200)
G = Beta(3, 1.6)
F = Beta(1, 1)
plot(x, pdf.(G, x / w_max) / w_max, label = L"g")
plot!(x, pdf.(F, x / w_max) / w_max, label = L"f")
31.2.4. Looking Forward#
What kind of optimal policy might result from (31.3) and the parameterization specified above?
Intuitively, if we accept at \(w_a\) and \(w_a \leq w_b\), then — all other things being given — we should also accept at \(w_b\).
This suggests a policy of accepting whenever \(w\) exceeds some threshold value \(\bar w\).
But \(\bar w\) should depend on \(\pi\) — in fact it should be decreasing in \(\pi\) because
\(f\) is a less attractive offer distribution than \(g\)
larger \(\pi\) means more weight on \(f\) and less on \(g\)
Thus larger \(\pi\) depresses the worker’s assessment of her future prospects, and relatively low current offers become more attractive.
Summary: We conjecture that the optimal policy is of the form \(\mathbb 1\{w \geq \bar w(\pi) \}\) for some decreasing function \(\bar w\).
31.3. Take 1: Solution by VFI#
Let’s set about solving the model and see how our results match with our intuition.
We begin by solving via value function iteration (VFI), which is natural but ultimately turns out to be second best.
The code is as follows.
# use key word argment
function SearchProblem(; beta = 0.95, c = 0.6, F_a = 1, F_b = 1,
G_a = 3, G_b = 1.2, w_max = 2.0,
w_grid_size = 40, pi_grid_size = 40)
F = Beta(F_a, F_b)
G = Beta(G_a, G_b)
# scaled pdfs
f(x) = pdf.(F, x / w_max) / w_max
g(x) = pdf.(G, x / w_max) / w_max
pi_min = 1e-3 # avoids instability
pi_max = 1 - pi_min
w_grid = range(0, w_max, length = w_grid_size)
pi_grid = range(pi_min, pi_max, length = pi_grid_size)
nodes, weights = qnwlege(21, 0.0, w_max)
return (; beta, c, F, G, f,
g, n_w = w_grid_size, w_max,
w_grid, n_pi = pi_grid_size, pi_min,
pi_max, pi_grid = pi_grid, quad_nodes = nodes,
quad_weights = weights)
end
function q(sp, w, pi_val)
new_pi = 1.0 / (1 + ((1 - pi_val) * sp.g(w)) / (pi_val * sp.f(w)))
# Return new_pi when in [pi_min, pi_max] and else end points
return clamp(new_pi, sp.pi_min, sp.pi_max)
end
function T!(sp, v, out;
ret_policy = false)
# simplify names
(; f, g, beta, c) = sp
nodes, weights = sp.quad_nodes, sp.quad_weights
vf = extrapolate(interpolate((sp.w_grid, sp.pi_grid), v,
Gridded(Linear())), Flat())
# set up quadrature nodes/weights
# q_nodes, q_weights = qnwlege(21, 0.0, sp.w_max)
for (w_i, w) in enumerate(sp.w_grid)
# calculate v1
v1 = w / (1 - beta)
for (pi_j, _pi) in enumerate(sp.pi_grid)
# calculate v2
function integrand(m)
[vf(m[i], q.(Ref(sp), m[i], _pi)) *
(_pi * f(m[i]) + (1 - _pi) * g(m[i])) for i in 1:length(m)]
end
integral = do_quad(integrand, nodes, weights)
# integral = do_quad(integrand, q_nodes, q_weights)
v2 = c + beta * integral
# return policy if asked for, otherwise return max of values
out[w_i, pi_j] = ret_policy ? v1 > v2 : max(v1, v2)
end
end
return out
end
function T(sp, v;
ret_policy = false)
out_type = ret_policy ? Bool : Float64
out = zeros(out_type, sp.n_w, sp.n_pi)
T!(sp, v, out, ret_policy = ret_policy)
end
get_greedy!(sp, v, out) = T!(sp, v, out, ret_policy = true)
get_greedy(sp, v) = T(sp, v, ret_policy = true)
function res_wage_operator!(sp, phi, out)
# simplify name
(; f, g, beta, c) = sp
# Construct interpolator over pi_grid, given phi
phi_f = LinearInterpolation(sp.pi_grid, phi, extrapolation_bc = Line())
# set up quadrature nodes/weights
q_nodes, q_weights = qnwlege(7, 0.0, sp.w_max)
for (i, _pi) in enumerate(sp.pi_grid)
function integrand(x)
max.(x, phi_f.(q.(Ref(sp), x, _pi))) .* (_pi * f(x) + (1 - _pi) * g(x))
end
integral = do_quad(integrand, q_nodes, q_weights)
out[i] = (1 - beta) * c + beta * integral
end
end
function res_wage_operator(sp, phi)
out = similar(phi)
res_wage_operator!(sp, phi, out)
return out
end
res_wage_operator (generic function with 1 method)
The type SearchProblem is used to store parameters and methods needed to compute optimal actions.
The Bellman operator is implemented as the method T(), while get_greedy()
computes an approximate optimal policy from a guess v of the value function.
We will omit a detailed discussion of the code because there is a more efficient solution method.
These ideas are implemented in the .res_wage_operator() method.
Before explaining it let’s look at solutions computed from value function iteration.
Here’s the value function:
# Set up the problem and initial guess, solve by VFI
sp = SearchProblem(; w_grid_size = 100, pi_grid_size = 100)
v_init = fill(sp.c / (1 - sp.beta), sp.n_w, sp.n_pi)
f(x) = T(sp, x)
v = compute_fixed_point(f, v_init)
policy = get_greedy(sp, v)
# Make functions for the linear interpolants of these
vf = extrapolate(interpolate((sp.w_grid, sp.pi_grid), v, Gridded(Linear())),
Flat())
pf = extrapolate(interpolate((sp.w_grid, sp.pi_grid), policy,
Gridded(Linear())), Flat())
function plot_value_function(; w_plot_grid_size = 100,
pi_plot_grid_size = 100)
pi_plot_grid = range(0.001, 0.99, length = pi_plot_grid_size)
w_plot_grid = range(0, sp.w_max, length = w_plot_grid_size)
Z = [vf(w_plot_grid[j], pi_plot_grid[i])
for j in 1:w_plot_grid_size, i in 1:pi_plot_grid_size]
p = contour(pi_plot_grid, w_plot_grid, Z, levels = 15, alpha = 0.6,
fill = true, size = (400, 400), c = :lightrainbow)
plot!(xlabel = L"\pi", ylabel = L"w", xguidefont = font(12))
return p
end
plot_value_function()
Compute iterate 10 with error 0.1980171015328338
Compute iterate 20 with error 0.007608221868114384
Compute iterate 30 with error 0.0002901698734376623
Converged in 34 steps
The optimal policy:
function plot_policy_function(; w_plot_grid_size = 100,
pi_plot_grid_size = 100)
pi_plot_grid = range(0.001, 0.99, length = pi_plot_grid_size)
w_plot_grid = range(0, sp.w_max, length = w_plot_grid_size)
Z = [pf(w_plot_grid[j], pi_plot_grid[i])
for j in 1:w_plot_grid_size, i in 1:pi_plot_grid_size]
p = contour(pi_plot_grid, w_plot_grid, Z, levels = 1, alpha = 0.6, fill = true,
size = (400, 400), c = :coolwarm)
plot!(xlabel = L"\pi", ylabel = "wage", xguidefont = font(12), cbar = false)
annotate!(0.4, 1.0, "reject")
annotate!(0.7, 1.8, "accept")
return p
end
plot_policy_function()
The code takes several minutes to run.
The results fit well with our intuition from section looking forward.
The black line in the figure above corresponds to the function \(\bar w(\pi)\) introduced there.
It is decreasing as expected.
31.4. Take 2: A More Efficient Method#
Our implementation of VFI can be optimized to some degree.
But instead of pursuing that, let’s consider another method to solve for the optimal policy.
We will use iteration with an operator that has the same contraction rate as the Bellman operator, but
one dimensional rather than two dimensional
no maximization step
As a consequence, the algorithm is orders of magnitude faster than VFI.
This section illustrates the point that when it comes to programming, a bit of mathematical analysis goes a long way.
31.4.1. Another Functional Equation#
To begin, note that when \(w = \bar w(\pi)\), the worker is indifferent between accepting and rejecting.
Hence the two choices on the right-hand side of (31.3) have equal value:
Together, (31.3) and (31.4) give
Combining (31.4) and (31.5), we obtain
Multiplying by \(1 - \beta\), substituting in \(\pi' = q(w', \pi)\) and using \(\circ\) for composition of functions yields
Equation (31.6) can be understood as a functional equation, where \(\bar w\) is the unknown function.
Let’s call it the reservation wage functional equation (RWFE).
The solution \(\bar w\) to the RWFE is the object that we wish to compute.
31.4.2. Solving the RWFE#
To solve the RWFE, we will first show that its solution is the fixed point of a contraction mapping.
To this end, let
\(b[0,1]\) be the bounded real-valued functions on \([0,1]\)
\(\| \psi \| := \sup_{x \in [0,1]} | \psi(x) |\)
Consider the operator \(Q\) mapping \(\psi \in b[0,1]\) into \(Q\psi \in b[0,1]\) via
Comparing (31.6) and (31.7), we see that the set of fixed points of \(Q\) exactly coincides with the set of solutions to the RWFE.
If \(Q \bar w = \bar w\) then \(\bar w\) solves (31.6) and vice versa.
Moreover, for any \(\psi, \phi \in b[0,1]\), basic algebra and the triangle inequality for integrals tells us that
Working case by case, it is easy to check that for real numbers \(a, b, c\) we always have
Combining (31.8) and (31.9) yields
Taking the supremum over \(\pi\) now gives us
In other words, \(Q\) is a contraction of modulus \(\beta\) on the complete metric space \((b[0,1], \| \cdot \|)\).
Hence
A unique solution \(\bar w\) to the RWFE exists in \(b[0,1]\).
\(Q^k \psi \to \bar w\) uniformly as \(k \to \infty\), for any \(\psi \in b[0,1]\).
31.4.2.1. Implementation#
These ideas are implemented in the .res_wage_operator() method from odu.jl as shown above.
The method corresponds to action of the operator \(Q\).
The following exercise asks you to exploit these facts to compute an approximation to \(\bar w\).
31.5. Exercises#
31.5.1. Exercise 1#
Use the default parameters and the .res_wage_operator() method to compute an optimal policy.
Your result should coincide closely with the figure for the optimal policy shown above.
Try experimenting with different parameters, and confirm that the change in the optimal policy coincides with your intuition.
31.6. Solutions#
31.6.1. Exercise 1#
This code solves the “Offer Distribution Unknown” model by iterating on
a guess of the reservation wage function. You should find that the run
time is much shorter than that of the value function approach in
examples/odu_vfi_plots.jl.
sp = SearchProblem(pi_grid_size = 50)
phi_init = ones(sp.n_pi)
f_ex1(x) = res_wage_operator(sp, x)
w_bar = compute_fixed_point(f_ex1, phi_init)
plot(sp.pi_grid, w_bar, linewidth = 2, color = :black,
fillrange = 0, fillalpha = 0.15, fillcolor = :blue)
plot!(sp.pi_grid, 2 * ones(length(w_bar)), linewidth = 0, fillrange = w_bar,
fillalpha = 0.12, fillcolor = :green, legend = :none)
plot!(ylims = (0, 2), annotations = [(0.42, 1.2, "reject"),
(0.7, 1.8, "accept")])
Compute iterate 10 with error 0.007194437603255555
Compute iterate 20 with error 0.00043487034178757433
Converged in 26 steps
The next piece of code is not one of the exercises from QuantEcon – it’s just a fun simulation to see what the effect of a change in the underlying distribution on the unemployment rate is.
At a point in the simulation, the distribution becomes significantly worse. It takes a while for agents to learn this, and in the meantime they are too optimistic, and turn down too many jobs. As a result, the unemployment rate spikes.
The code takes a few minutes to run.
# Determinism and random objects.
using Random
Random.seed!(42)
# Set up model and compute the function w_bar
sp = SearchProblem(pi_grid_size = 50, F_a = 1, F_b = 1)
phi_init = ones(sp.n_pi)
g(x) = res_wage_operator(sp, x)
w_bar_vals = compute_fixed_point(g, phi_init)
w_bar = extrapolate(interpolate((sp.pi_grid,), w_bar_vals,
Gridded(Linear())), Flat())
# Holds the employment state and beliefs of an individual agent.
mutable struct Agent{TF <: AbstractFloat, TI <: Integer}
_pi::TF
employed::TI
end
Agent(_pi = 1e-3) = Agent(_pi, 1)
function update!(ag, H)
if ag.employed == 0
w = rand(H) * 2 # account for scale in julia
if w >= w_bar(ag._pi)
ag.employed = 1
else
ag._pi = 1.0 ./ (1 .+ ((1 - ag._pi) .* sp.g(w)) ./ (ag._pi * sp.f(w)))
end
end
nothing
end
num_agents = 5000
separation_rate = 0.025 # Fraction of jobs that end in each period
separation_num = round(Int, num_agents * separation_rate)
agent_indices = collect(1:num_agents)
agents = [Agent() for i in 1:num_agents]
sim_length = 600
H = sp.G # Start with distribution G
change_date = 200 # Change to F after this many periods
unempl_rate = zeros(sim_length)
for i in 1:sim_length
if i % 20 == 0
println("date = $i")
end
if i == change_date
H = sp.F
end
# Randomly select separation_num agents and set employment status to 0
shuffle!(agent_indices)
separation_list = agent_indices[1:separation_num]
for agent in agents[separation_list]
agent.employed = 0
end
# update agents
for agent in agents
update!(agent, H)
end
employed = Int[agent.employed for agent in agents]
unempl_rate[i] = 1.0 - mean(employed)
end
plot(unempl_rate, linewidth = 2, label = "unemployment rate")
vline!([change_date], color = :red, label = "")
Compute iterate 10 with error 0.007194437603255555
Compute iterate 20 with error 0.00043487034178757433
Converged in 26 steps
date = 20
date = 40
date = 60
date = 80
date = 100
date = 120
date = 140
date = 160
date = 180
date = 200
date = 220
date = 240
date = 260
date = 280
date = 300
date = 320
date = 340
date = 360
date = 380
date = 400
date = 420
date = 440
date = 460
date = 480
date = 500
date = 520
date = 540
date = 560
date = 580
date = 600