34. Optimal Growth I: The Stochastic Optimal Growth Model#
Contents
34.1. Overview#
In this lecture we’re going to study a simple optimal growth model with one agent.
The model is a version of the standard one sector infinite horizon growth model studied in
The technique we use to solve the model is dynamic programming.
Our treatment of dynamic programming follows on from earlier treatments in our lectures on shortest paths and job search.
We’ll discuss some of the technical details of dynamic programming as we go along.
34.2. The Model#
Consider an agent who owns an amount \(y_t \in \mathbb R_+ := [0, \infty)\) of a consumption good at time \(t\).
This output can either be consumed or invested.
When the good is invested it is transformed one-for-one into capital.
The resulting capital stock, denoted here by \(k_{t+1}\), will then be used for production.
Production is stochastic, in that it also depends on a shock \(\xi_{t+1}\) realized at the end of the current period.
Next period output is
where \(f \colon \mathbb{R}_+ \to \mathbb{R}_+\) is called the production function.
The resource constraint is
and all variables are required to be nonnegative.
34.2.1. Assumptions and Comments#
In what follows,
The sequence \(\{\xi_t\}\) is assumed to be IID.
The common distribution of each \(\xi_t\) will be denoted \(\phi\).
The production function \(f\) is assumed to be increasing and continuous.
Depreciation of capital is not made explicit but can be incorporated into the production function.
While many other treatments of the stochastic growth model use \(k_t\) as the state variable, we will use \(y_t\).
This will allow us to treat a stochastic model while maintaining only one state variable.
We consider alternative states and timing specifications in some of our other lectures.
34.2.2. Optimization#
Taking \(y_0\) as given, the agent wishes to maximize
subject to
where
\(u\) is a bounded, continuous and strictly increasing utility function and
\(\beta \in (0, 1)\) is a discount factor
In (34.3) we are assuming that the resource constraint (34.1) holds with equality — which is reasonable because \(u\) is strictly increasing and no output will be wasted at the optimum.
In summary, the agent’s aim is to select a path \(c_0, c_1, c_2, \ldots\) for consumption that is
nonnegative,
feasible in the sense of (34.1),
optimal, in the sense that it maximizes (34.2) relative to all other feasible consumption sequences, and
adapted, in the sense that the action \(c_t\) depends only on observable outcomes, not future outcomes such as \(\xi_{t+1}\)
In the present context
\(y_t\) is called the state variable — it summarizes the “state of the world” at the start of each period.
\(c_t\) is called the control variable — a value chosen by the agent each period after observing the state.
34.2.3. The Policy Function Approach#
One way to think about solving this problem is to look for the best policy function.
A policy function is a map from past and present observables into current action.
We’ll be particularly interested in Markov policies, which are maps from the current state \(y_t\) into a current action \(c_t\).
For dynamic programming problems such as this one (in fact for any Markov decision process), the optimal policy is always a Markov policy.
In other words, the current state \(y_t\) provides a sufficient statistic for the history in terms of making an optimal decision today.
This is quite intuitive but if you wish you can find proofs in texts such as [SLP89] (section 4.1).
Hereafter we focus on finding the best Markov policy.
In our context, a Markov policy is a function \(\sigma \colon \mathbb R_+ \to \mathbb R_+\), with the understanding that states are mapped to actions via
In what follows, we will call \(\sigma\) a feasible consumption policy if it satisfies
In other words, a feasible consumption policy is a Markov policy that respects the resource constraint.
The set of all feasible consumption policies will be denoted by \(\Sigma\).
Each \(\sigma \in \Sigma\) determines a continuous state Markov process \(\{y_t\}\) for output via
This is the time path for output when we choose and stick with the policy \(\sigma\).
We insert this process into the objective function to get
This is the total expected present value of following policy \(\sigma\) forever, given initial income \(y_0\).
The aim is to select a policy that makes this number as large as possible.
The next section covers these ideas more formally.
34.2.4. Optimality#
The policy value function \(v_{\sigma}\) associated with a given policy \(\sigma\) is the mapping defined by
when \(\{y_t\}\) is given by (34.5) with \(y_0 = y\).
In other words, it is the lifetime value of following policy \(\sigma\) starting at initial condition \(y\).
The value function is then defined as
The value function gives the maximal value that can be obtained from state \(y\), after considering all feasible policies.
A policy \(\sigma \in \Sigma\) is called optimal if it attains the supremum in (34.8) for all \(y \in \mathbb R_+\).
34.2.5. The Bellman Equation#
With our assumptions on utility and production function, the value function as defined in (34.8) also satisfies a Bellman equation.
For this problem, the Bellman equation takes the form
This is a functional equation in \(w\).
The term \(\int w(f(y - c) z) \phi(dz)\) can be understood as the expected next period value when
\(w\) is used to measure value
the state is \(y\)
consumption is set to \(c\)
As shown in EDTC, theorem 10.1.11 and a range of other texts.
The value function \(v^*\) satisfies the Bellman equation
In other words, (34.9) holds when \(w=v^*\).
The intuition is that maximal value from a given state can be obtained by optimally trading off
current reward from a given action, vs
expected discounted future value of the state resulting from that action
The Bellman equation is important because it gives us more information about the value function.
It also suggests a way of computing the value function, which we discuss below.
34.2.6. Greedy policies#
The primary importance of the value function is that we can use it to compute optimal policies.
The details are as follows.
Given a continuous function \(w\) on \(\mathbb R_+\), we say that \(\sigma \in \Sigma\) is \(w\)-greedy if \(\sigma(y)\) is a solution to
for every \(y \in \mathbb R_+\).
In other words, \(\sigma \in \Sigma\) is \(w\)-greedy if it optimally trades off current and future rewards when \(w\) is taken to be the value function.
In our setting, we have the following key result
A feasible consumption policy is optimal if and only if it is \(v^*\)-greedy
The intuition is similar to the intuition for the Bellman equation, which was provided after (34.9).
See, for example, theorem 10.1.11 of EDTC.
Hence, once we have a good approximation to \(v^*\), we can compute the (approximately) optimal policy by computing the corresponding greedy policy.
The advantage is that we are now solving a much lower dimensional optimization problem.
34.2.7. The Bellman Operator#
How, then, should we compute the value function?
One way is to use the so-called Bellman operator.
(An operator is a map that sends functions into functions)
The Bellman operator is denoted by \(T\) and defined by
In other words, \(T\) sends the function \(w\) into the new function \(Tw\) defined (34.11).
By construction, the set of solutions to the Bellman equation (34.9) exactly coincides with the set of fixed points of \(T\).
For example, if \(Tw = w\), then, for any \(y \geq 0\),
which says precisely that \(w\) is a solution to the Bellman equation.
It follows that \(v^*\) is a fixed point of \(T\).
34.2.8. Review of Theoretical Results#
One can also show that \(T\) is a contraction mapping on the set of continuous bounded functions on \(\mathbb R_+\) under the supremum distance
See EDTC, lemma 10.1.18.
Hence it has exactly one fixed point in this set, which we know is equal to the value function.
It follows that
The value function \(v^*\) is bounded and continuous.
Starting from any bounded and continuous \(w\), the sequence \(w, Tw, T^2 w, \ldots\) generated by iteratively applying \(T\) converges uniformly to \(v^*\).
This iterative method is called value function iteration.
We also know that a feasible policy is optimal if and only if it is \(v^*\)-greedy.
It’s not too hard to show that a \(v^*\)-greedy policy exists (see EDTC, theorem 10.1.11 if you get stuck).
Hence at least one optimal policy exists.
Our problem now is how to compute it.
34.2.9. Unbounded Utility#
The results stated above assume that the utility function is bounded.
In practice economists often work with unbounded utility functions — and so will we.
In the unbounded setting, various optimality theories exist.
Unfortunately, they tend to be case specific, as opposed to valid for a large range of applications.
Nevertheless, their main conclusions are usually in line with those stated for the bounded case just above (as long as we drop the word “bounded”).
Consult, for example, section 12.2 of EDTC, [Kam12] or [MdRV10].
34.3. Computation#
Let’s now look at computing the value function and the optimal policy.
34.3.1. Fitted Value Iteration#
The first step is to compute the value function by value function iteration.
In theory, the algorithm is as follows
Begin with a function \(w\) — an initial condition.
Solving (34.11), obtain the function \(T w\).
Unless some stopping condition is satisfied, set \(w = Tw\) and go to step 2.
This generates the sequence \(w, Tw, T^2 w, \ldots\).
However, there is a problem we must confront before we implement this procedure: The iterates can neither be calculated exactly nor stored on a computer.
To see the issue, consider (34.11).
Even if \(w\) is a known function, unless \(Tw\) can be shown to have some special structure, the only way to store it is to record the value \(Tw(y)\) for every \(y \in \mathbb R_+\).
Clearly this is impossible.
What we will do instead is use fitted value function iteration.
The procedure is to record the value of the function \(Tw\) at only finitely many “grid” points \(y_1 < y_2 < \cdots < y_I\) and reconstruct it from this information when required.
More precisely, the algorithm will be
Begin with an array of values \(\{ w_1, \ldots, w_I \}\) representing the values of some initial function \(w\) on the grid points \(\{ y_1, \ldots, y_I \}\).
Build a function \(\hat w\) on the state space \(\mathbb R_+\) by interpolation or approximation, based on these data points.
Obtain and record the value \(T \hat w(y_i)\) on each grid point \(y_i\) by repeatedly solving (34.11).
Unless some stopping condition is satisfied, set \(\{ w_1, \ldots, w_I \} = \{ T \hat w(y_1), \ldots, T \hat w(y_I) \}\) and go to step 2.
How should we go about step 2?
This is a problem of function approximation, and there are many ways to approach it.
What’s important here is that the function approximation scheme must not only produce a good approximation to \(Tw\), but also combine well with the broader iteration algorithm described above.
One good choice from both respects is continuous piecewise linear interpolation (see this paper for further discussion).
The next figure illustrates piecewise linear interpolation of an arbitrary function on grid points \(0, 0.2, 0.4, 0.6, 0.8, 1\).
using LinearAlgebra, Statistics
using LaTeXStrings, Plots, Interpolations, NLsolve, Optim, Random
using Optim: maximum, maximizer
f(x) = 2 .* cos.(6x) .+ sin.(14x) .+ 2.5
c_grid = 0:0.2:1
f_grid = range(0, 1, length = 150)
Af = LinearInterpolation(c_grid, f(c_grid))
plt = plot(xlim = (0, 1), ylim = (0, 6))
plot!(plt, f, f_grid, color = :blue, lw = 2, alpha = 0.8, label = "true function")
plot!(plt, f_grid, Af.(f_grid), color = :green, lw = 2, alpha = 0.8,
label = "linear approximation")
plot!(plt, f, c_grid, seriestype = :sticks, linestyle = :dash, linewidth = 2,
alpha = 0.5,
label = "")
plot!(plt, legend = :top)
Another advantage of piecewise linear interpolation is that it preserves useful shape properties such as monotonicity and concavity / convexity.
34.3.2. The Bellman Operator#
Here’s a function that implements the Bellman operator using linear interpolation
function T(w; p, tol = 1e-10)
(; beta, u, f, Xi, y) = p # unpack parameters
w_func = LinearInterpolation(y, w)
Tw = similar(w)
sigma = similar(w)
for (i, y_val) in enumerate(y)
# solve maximization for each point in y, using y itself as initial condition.
results = maximize(c -> u(c; p) +
beta * mean(w_func.(f(y_val - c; p) .* Xi)),
tol, y_val)
Tw[i] = maximum(results)
sigma[i] = maximizer(results)
end
return (; w = Tw, sigma) # returns named tuple of results
end
T (generic function with 1 method)
Notice that the expectation in (34.11) is computed via Monte Carlo, using the approximation
where \(\{\xi_i\}_{i=1}^n\) are IID draws from \(\phi\).
Monte Carlo is not always the most efficient way to compute integrals numerically but it does have some theoretical advantages in the present setting.
(For example, it preserves the contraction mapping property of the Bellman operator — see, e.g., [PalS13])
34.3.3. An Example#
Let’s test out our operator when
\(f(k) = k^{\alpha}\)
\(u(c) = \ln c\)
\(\phi\) is the distribution of \(\exp(\mu + \sigma \zeta)\) when \(\zeta\) is standard normal
As is well-known (see [LS18], section 3.1.2), for this particular problem an exact analytical solution is available, with
The optimal consumption policy is
Let’s code this up now so we can test against it below
In addition to the model parameters, we need a grid and some shock draws for Monte Carlo integration.
Random.seed!(42) # for reproducible results
u(c; p) = log(c) # utility
f(k; p) = k^p.alpha # deterministic part of production function
function OptimalGrowthModel(; alpha = 0.4, beta = 0.96, mu = 0.0, s = 0.1,
u = u, f = f, # defaults defined above
y = range(1e-5, 4.0, length = 200), # grid on y
Xi = exp.(mu .+ s * randn(250)))
return (; alpha, beta, mu, s, u, f, y, Xi)
end # named tuples defaults
# True value and policy function
function v_star(y; p)
(; alpha, mu, beta) = p
c1 = log(1 - alpha * beta) / (1 - beta)
c2 = (mu + alpha * log(alpha * beta)) / (1 - alpha)
c3 = 1 / (1 - beta)
c4 = 1 / (1 - alpha * beta)
return c1 + c2 * (c3 - c4) + c4 * log(y)
end
c_star(y; p) = (1 - p.alpha * p.beta) * y
c_star (generic function with 1 method)
34.3.4. A First Test#
To test our code, we want to see if we can replicate the analytical solution numerically, using fitted value function iteration.
Now let’s do some tests.
As one preliminary test, let’s see what happens when we apply our Bellman operator to the exact solution \(v^*\).
In theory, the resulting function should again be \(v^*\).
In practice we expect some small numerical error.
p = OptimalGrowthModel() # use all default parameters from named tuple
w_star = v_star.(p.y; p) # evaluate closed form value along grid
w = T(w_star; p).w # evaluate operator, access Tw results
plt = plot(ylim = (-35, -24))
plot!(plt, p.y, w, linewidth = 2, alpha = 0.6, label = L"T(v^*)")
plot!(plt, p.y, w_star, linewidth = 2, alpha = 0.6, label = L"v^*")
plot!(plt, legend = :bottomright)
The two functions are essentially indistinguishable, so we are off to a good start.
Now let’s have a look at iterating with the Bellman operator, starting off from an arbitrary initial condition.
The initial condition we’ll start with is \(w(y) = 5 \ln (y)\)
w = 5 * log.(p.y) # An initial condition -- fairly arbitrary
n = 35
plot(xlim = (extrema(p.y)), ylim = (-50, 10))
lb = "initial condition"
plt = plot(p.y, w, color = :black, linewidth = 2, alpha = 0.8, label = lb)
for i in 1:n
w = T(w; p).w
plot!(p.y, w, color = RGBA(i / n, 0, 1 - i / n, 0.8), linewidth = 2,
alpha = 0.6,
label = "")
end
lb = "true value function"
plot!(plt, p.y, v_star.(p.y; p), color = :black, linewidth = 2, alpha = 0.8,
label = lb)
plot!(plt, legend = :bottomright)
The figure shows
the first 36 functions generated by the fitted value function iteration algorithm, with hotter colors given to higher iterates
the true value function \(v^*\) drawn in black
The sequence of iterates converges towards \(v^*\).
We are clearly getting closer.
We can write a function that computes the exact fixed point
function solve_optgrowth(initial_w; p, iterations = 500, m = 3, show_trace = false)
results = fixedpoint(w -> T(w; p).w, initial_w; iterations, m, show_trace) # Anderson iteration
v_star = results.zero
sigma = T(results.zero; p).sigma
return (; v_star, sigma, results)
end
solve_optgrowth (generic function with 1 method)
We can check our result by plotting it against the true value
initial_w = 5 * log.(p.y)
sol = solve_optgrowth(initial_w; p)
v_star_approx = sol.v_star
println("Converged in $(sol.results.iterations) to an ||residuals||_∞ of $(sol.results.residual_norm)")
plt = plot(ylim = (-35, -24))
plot!(plt, p.y, v_star_approx, linewidth = 2, alpha = 0.6,
label = "approximate value function")
plot!(plt, p.y, v_star.(p.y; p), linewidth = 2, alpha = 0.6,
label = "true value function")
plot!(plt, legend = :bottomright)
Converged in 15 to an ||residuals||_∞ of 2.179099567456433e-9
The figure shows that we are pretty much on the money
Note that this converges in fewer than the 36 iterations printed above because it is using Anderson iteration - where the \(m=0\) parameter is naive fixed point iteration
34.3.5. The Policy Function#
To compute an approximate optimal policy, we take the approximate value function we just calculated and then compute the corresponding greedy policy.
The next figure compares the result to the exact solution, which, as mentioned above, is \(\sigma(y) = (1 - \alpha \beta) y\).
plt = plot(p.y, T(v_star_approx; p).sigma, lw = 2, alpha = 0.6,
label = "approximate policy function")
plot!(plt, p.y, c_star.(p.y; p), lw = 2, alpha = 0.6,
label = "true policy function")
plot!(plt, legend = :bottomright)
The figure shows that we’ve done a good job in this instance of approximating the true policy.
34.4. Exercises#
34.4.1. Exercise 1#
Once an optimal consumption policy \(\sigma\) is given, income follows (34.5).
The next figure shows a simulation of 100 elements of this sequence for three different discount factors (and hence three different policies).
In each sequence, the initial condition is \(y_0 = 0.1\).
The discount factors are discount_factors = (0.8, 0.9, 0.98).
We have also dialed down the shocks a bit.
p = OptimalGrowthModel(s = 0.05)
p.Xi
Show code cell output
250-element Vector{Float64}:
0.9819961668245721
1.0126664087134005
0.9843739746850739
0.9845578517434581
1.0416597330713688
1.0241232892126522
0.9579326855331085
0.9291692140795736
0.8996794537879703
1.0021914808932793
0.9595731512645318
1.0429095238242871
1.0219313258836822
⋮
1.0013570363566526
0.936028836092597
0.9944175678826758
0.9775640668429453
1.022229941967392
1.028686047499314
0.9341510619913468
0.9070123342340449
0.9489873846557617
1.0177798574430181
0.9897225222212319
0.9845828858353305
Otherwise, the parameters and primitives are the same as the log linear model discussed earlier in the lecture.
Notice that more patient agents typically have higher wealth.
Replicate the figure modulo randomness.
34.5. Solutions#
34.5.1. Exercise 1#
Here’s one solution (assuming as usual that you’ve executed everything above)
function simulate_og(sigma, p, y0, ts_length)
y = zeros(ts_length)
y[1] = y0
for t in 1:(ts_length - 1)
y[t + 1] = (y[t] - sigma(y[t]))^p.alpha * exp(p.mu + p.s * randn())
end
return y
end
beta_vals = [0.9 0.94 0.98]
ts_length = 100
y0 = 0.1
plt = plot()
for beta in beta_vals
p = OptimalGrowthModel(; beta) # change beta from default
initial_w = 5 * log.(p.y)
sol = solve_optgrowth(initial_w; p)
sigma_func = LinearInterpolation(p.y, sol.sigma)
y = simulate_og(sigma_func, p, y0, ts_length)
plot!(plt, 0:(ts_length - 1), y, lw = 2, alpha = 0.6, label = L"\beta = %$beta")
end
plt