Code should execute sequentially if run in a Jupyter notebook

- See the set up page to install Jupyter, Julia (0.6+) and all necessary libraries
- Please direct feedback to contact@quantecon.org or the discourse forum

# Consumption and Tax Smoothing with Complete and Incomplete Markets¶

Contents

## Overview¶

This lecture describes two types of consumption-smoothing model

- one is in the
**complete markets**tradition of Lucas and Stokey [LS83] - the other is in the
**incomplete markets**tradition of Hall [Hal78] and Barro [Bar79]

*Complete markets* allow a consumer or government to buy or sell claims contingent on all possible states of the world

*Incomplete markets* allow a consumer or government to buy or sell only a limited set of securities, often only a single risk-free security

Hall [Hal78] and Barro [Bar79] both assumed that the only asset that can be traded is a risk-free one period bond

Hall assumed an exogenous stochastic process of nonfinancial income and an exogenous gross interest rate on one period risk-free debt that equals \(\beta^{-1}\), where \(\beta \in (0,1)\) is also a consumer’s intertemporal discount factor

Barro [Bar79] made an analogous assumption about the risk-free interest rate in a tax-smoothing model that we regard as isomorphic to Hall’s consumption-smoothing model

We maintain Hall and Barro’s assumption about the interest rate when we describe an incomplete markets version of our model

In addition, we extend their assumption about the interest rate to an appropriate counterpart that we use in a “complete markets” model in the style of Lucas and Stokey [LS83]

While we are equally interested in consumption-smoothing and tax-smoothing models, for the most part we focus explicitly on consumption smoothing versions of these models

But for each version of the consumption-smoothing model there is a natural tax-smoothing counterpart obtained simply by

- relabeling consumption as tax collections and nonfinancial income as government expenditures
- relabeling the consumer’s debt as the government’s
*assets*

We’ll consider two closely related assumptions about the consumer’s exogenous non-financial income process (or in the tax-smoothing interpretation, the government’s exogenous expenditure process):

- that it obeys a finite \(N\) state Markov chain (setting \(N=2\) most of the time)
- that it is described by a linear state space model with a continuous state vector in \({\mathbb R}^n\) driven by a Gaussian vector iid shock process

We’ll spend most of this lecture studying the finite-state Markov specification, but will briefly treat the linear state space specification before concluding

### Relationship to other lectures¶

This lecture can be viewed as a followup to The Permanent Income Model: II and a warm up for a model of tax smoothing described in Optimal Taxation with State-Contingent Debt

Linear-quadratic versions of the Lucas-Stokey tax-smoothing model are described in Optimal Taxation in an LQ Economy

The key difference between those lectures and this one is

- Here the decision maker takes all prices as exogenous, meaning that his decisions do not affect them
- In Optimal Taxation in an LQ Economy and Optimal Taxation with State-Contingent Debt, the decision maker – the government in the case of these lectures – recognizes that his decisions affect prices

So these later lectures are partly about how the government should manipulate prices of government debt

## Background¶

Outcomes in consumption-smoothing (or tax-smoothing) models emerge from two sources:

- a decision maker – a consumer in the consumption-smoothing model or
a government in the tax-smoothing model – who wants to maximize an
intertemporal objective function that expresses its preference for
paths of consumption (or tax collections) that are
*smooth*in the sense of not varying across time and Markov states - a set of trading opportunities that allow the optimizer to transform a possibly erratic nonfinancial income (or government expenditure) process into a smoother consumption (or tax collections) process by purchasing or selling financial securities

In the complete markets version of the model, each period the consumer can buy or sell one-period ahead state-contingent securities whose payoffs depend on next period’s realization of the Markov state

In the two-state Markov chain case, there are two such securities each period

In an \(N\) state Markov state version of the model, there are \(N\) such securities traded each period

These state-contingent securities are commonly called Arrow securities, after Kenneth Arrow who first theorized about them

In the incomplete markets version of the model, the consumer can buy and sell only one security each period, a risk-free bond with gross return \(\beta^{-1}\)

### Finite state Markov income process¶

In each version of the consumption-smoothing model, nonfinancial income is governed by a two-state Markov chain (it’s easy to generalize this to an \(N\) state chain)

In particular, the *state of the world* is given by \(s_t\) that follows
a Markov chain with transition probability matrix

Nonfinancial income \(\{y_t\}\) obeys

A consumer wishes to maximize

### Market Structure¶

The two models differ in how effectively the market structure allows the consumer to transfer resources across time and Markov states, there being more transfer opportunities in the complete markets setting than in the incomplete markets setting

Watch how these differences in opportunities affect

- how smooth consumption is across time and Markov states
- how the consumer chooses to make his levels of indebtedness behave over time and across Markov states

## Model 1 (Complete Markets)¶

At each date \(t \geq 0\), the consumer trades **one-period ahead
Arrow securities**

We assume that prices of these securities are exogenous to the consumer (or in the tax-smoothing version of the model, to the government)

*Exogenous* means that the decision maker can take them as given
and unaffected by his decisions

In Markov state \(s_t\) at time \(t\), one unit of consumption in state \(s_{t+1}\) at time \(t+1\) costs \(q(s_{t+1} \,|\, s_t)\) units of the time \(t\) consumption good

At time \(t=0\), the consumer starts with an inherited level of debt due at time \(0\) of \(b_0\) units of time \(0\) consumption goods

The consumer’s budget constraint at \(t \geq 0\) in Markov state \(s_t\) is

where \(b_{t+1}(\bar s_j\,|\, s_t)\) are the consumer’s time \(t\) sales of the time \(t+1\) consumption good in Markov state \(\bar s_j\)

An analogue of Hall’s assumption that the one-period risk-free gross interest rate is \(\beta^{-1}\) is

To understand this, observe that in state \(\bar s_i\) it costs \(\sum_j q(\bar s_j \,|\, \bar s_i)\) to purchase one unit of consumption next period *for sure*, i.e., meaning no matter what state of the world occurs at \(t+1\)

Hence the implied price of a risk-free claim on one unit of consumption next period is

This confirms that (2) is a natural analogue of Hall’s assumption about the risk-free one-period interest rate

First-order necessary conditions for maximizing the consumer’s expected utility are

or, under our assumption (2) on Arrow security prices,

Thus, our consumer sets \(c_t = \bar c\) for all \(t \geq 0\) for some value \(\bar c\) that it is our job now to determine

We’ll make the plausible guess that

so that the amount borrowed today turns out to depend only on *tomorrow’s* Markov state. (Why is this is a plausible guess?)

To determine \(\bar c\), we shall pursue the implications of our guess about the consumer’s debt level choices and the consumer’s budget constraints in each Markov state today

For \(t \geq 1\), we write these as

or

These are \(2\) equations in the \(3\) unknowns, \(\bar c, b(\bar s_1), b(\bar s_2)\)

To get a third equation, we assume that at time \(t=0\), \(b_0\) is the debt due; and we assume that at time \(t=0\), the Markov state is \(\bar s_1\)

Then the budget constraint at time \(t=0\) is

If we substitute (5) into the first equation of (4) and rearrange, we discover that

We can then use the second equation of (4) to deduce the restriction

an equation in the unknown \(b(\bar s_2)\)

Knowing \(b(\bar s_1)\) and \(b(\bar s_2)\), we can solve equation (5) for the constant level of consumption \(\bar c\)

### Key outcomes¶

The preceding calculations indicate that in the complete markets version of our model, we obtain the following striking results:

- The consumer chooses to make consumption perfectly constant across time and Markov states

We computed the constant level of consumption \(\bar c\) and indicated how that level depends on the underlying specifications of preferences, Arrow securities prices, the stochastic process of exogenous nonfinancial income, and the initial debt level \(b_0\)

- The consumer’s debt neither accumulates, nor decumulates, nor drifts. Instead the debt level each period is an exact function of the Markov state, so in the two-state Markov case, it switches between two values

We computed how one of those debt levels depends entirely on initial debt – it equals it – and how the other value depends on virtually all remaining parameters of the model

### Code¶

Here’s some code that, among other things, contains a function called consumption_complete()

This function computes \(b(\bar s_1), b(\bar s_2), \bar c\) as outcomes given a set of parameters, under the assumption of complete markets

```
using QuantEcon
"""
The data for a consumption problem, including some default values.
"""
struct ConsumptionProblem{TF<:AbstractFloat}
beta::TF
y::Vector{TF}
b0::TF
P::Matrix{TF}
end
"""
Parameters
----------
beta : discount factor
P : 2x2 transition matrix
y : Array containing the two income levels
b0 : debt in period 0 (= state_1 debt level)
"""
function ConsumptionProblem(;
beta = 0.96,
y = [2.0, 1.5],
b0 = 3.0,
P = [0.8 0.2;
0.4 0.6])
ConsumptionProblem(beta, y, b0, P)
end
"""
Computes endogenous values for the complete market case.
Parameters
----------
cp : instance of ConsumptionProblem
Returns
-------
c_bar : constant consumption
b1 : rolled over b0
b2 : debt in state_2
associated with the price system
Q = beta * P
"""
function consumption_complete(cp::ConsumptionProblem)
beta, P, y, b0 = cp.beta, cp.P, cp.y, cp.b0 # Unpack
y1, y2 = y # extract income levels
b1 = b0 # b1 is known to be equal to b0
Q = beta * P # assumed price system
# Using equation (7) calculate b2
b2 = (y2 - y1 - (Q[1, 1] - Q[2, 1] - 1) * b1)/(Q[1, 2] + 1 - Q[2, 2])
# Using equation (5) calculae c_bar
c_bar = y1 - b0 + ([b1 b2] * Q[1, :] )[1]
return c_bar, b1, b2
end
"""
Computes endogenous values for the incomplete market case.
Parameters
----------
cp : instance of ConsumptionProblem
N_simul : Integer
"""
function consumption_incomplete(cp::ConsumptionProblem;
N_simul::Integer=150)
beta, P, y, b0 = cp.beta, cp.P, cp.y, cp.b0 # Unpack
# For the simulation define a quantecon MC class
mc = MarkovChain(P)
# Useful variables
y = y''
v = inv(eye(2) - beta * P) * y
# Simulat state path
s_path = simulate(mc, N_simul, init = 1)
# Store consumption and debt path
b_path, c_path = ones(N_simul + 1), ones(N_simul)
b_path[1] = b0
# Optimal decisions from (12) and (13)
db = ((1 - beta) * v - y) / beta
for (i, s) in enumerate(s_path)
c_path[i] = (1 - beta) * (v - b_path[i] * ones(2, 1))[s, 1]
b_path[i + 1] = b_path[i] + db[s, 1]
end
return c_path, b_path[1:end-1], y[s_path], s_path
end
```

Let’s test by checking that \(\bar c\) and \(b_2\) satisfy the budget constraint

```
cp = ConsumptionProblem()
c_bar, b1, b2 = consumption_complete(cp)
debt_complete = [b1, b2]
isapprox((c_bar + b2 - cp.y[2] - debt_complete' * (cp.beta * cp.P)[2, :])[1], 0)
```

```
true
```

Below, we’ll take the outcomes produced by this code – in particular the implied consumption and debt paths – and compare them with outcomes from an incomplete markets model in the spirit of Hall [Hal78] and Barro [Bar79] (and Gallatin (1807) [Gal37])

## Model 2 (One-Period Risk Free Debt Only)¶

This is a version of the original models of Hall (1978) and Barro (1979) in which the decision maker’s ability to substitute intertemporally is constrained by his ability to buy or sell only one security, a risk-free one-period bond bearing a constant gross interest rate of \(\beta^{-1}\)

Given an initial debt \(b_0\) at time \(0\), the consumer faces a sequence of budget constraints

where \(\beta\) is the price at time \(t\) of a risk-free claim on one unit of time consumption at time \(t+1\)

First-order conditions for the consumer’s problem are

For our assumed quadratic utility function this implies

which is Hall’s (1978) conclusion that consumption follows a random walk

As we saw in our first lecture on the permanent income model, this leads to

and

which expresses \(c_t\) as a net interest rate factor \(1 - \beta\) times the sum of the expected present value of non-financial income \(\mathbb E_t \sum_{j=0}^\infty \beta^j y_{t+j}\) and financial wealth \(-b_t\)

Substituting (10) into the one-period budget constraint and rearranging leads to

Now let’s do a useful calculation that will yield a convenient expression for the key term \(\mathbb E_t \sum_{j=0}^\infty\beta^j y_{t+j}\) in our Markov setting

Define

In our Markov setting, \(v_t = v(1)\) when \(s_t= \bar s_1\) and \(v_t = v(2)\) when \(s_t=\bar s_2\)

Therefore, we can write

or

where \(\vec v = \begin{bmatrix} v(1) \cr v(2) \end{bmatrix}\) and \(\vec y = \begin{bmatrix} y(1) \cr y(2) \end{bmatrix}\)

We can also write the last expression as

In our Markov setting, from expression (10), consumption at date \(t\) when debt is \(b_t\) and the Markov state today is \(s_t = i\) is evidently

and the increment in debt is

### Summary of outcomes¶

In contrast to outcomes in the complete markets model, in the incomplete markets model

- consumption drifts over time as a random walk; the level of consumption at time \(t\) depends on the level of debt that the consumer brings into the period as well as the expected discounted present value of nonfinancial income at \(t\)
- the consumer’s debt drifts upward over time in response to low realizations of nonfinancial income and drifts downward over time in response to high realizations of nonfinancial income
- the drift over time in the consumer’s debt and the dependence of current consumption on today’s debt level account for the drift over time in consumption

### Code for the incomplete markets model¶

The code above also contains a function called consumption_incomplete() that uses (12) and (13) to

- simulate paths of \(y_t, c_t, b_{t+1}\)
- plot these against values of of \(\bar c, b(s_1), b(s_2)\) found in a corresponding complete markets economy

Let’s try this, using the same parameters in both economies

```
using PyPlot
srand(1)
N_simul = 150
cp = ConsumptionProblem()
c_bar, b1, b2 = consumption_complete(cp)
debt_complete = [b1, b2]
c_path, debt_path, y_path, s_path = consumption_incomplete(cp, N_simul=N_simul)
fig, ax = subplots(1, 2, figsize = (15, 5))
ax[1][:set_title]("Consumption paths", fontsize = 17)
ax[1][:plot](1:N_simul, c_path, label = "incomplete market", lw = 3)
ax[1][:plot](1:N_simul, c_bar * ones(N_simul),
label = "complete market", lw = 3)
ax[1][:plot](1:N_simul, y_path, label = "income",
lw = 2, alpha = .6, linestyle = "--")
ax[1][:legend](loc = "best", fontsize = 15)
ax[1][:set_xlabel]("Periods", fontsize = 13)
ax[1][:set_ylim]([1.4, 2.1])
ax[2][:set_title]("Debt paths", fontsize = 17)
ax[2][:plot](1:N_simul, debt_path, label = "incomplete market", lw = 2)
ax[2][:plot](1:N_simul, debt_complete[s_path], label = "complete market", lw = 2)
ax[2][:plot](1:N_simul, y_path, label = "income", lw = 2,
alpha = .6, linestyle = "--")
ax[2][:legend](loc = "best", fontsize = 15)
ax[2][:axhline](0, color = "k", lw = 1)
ax[2][:set_xlabel]("Periods", fontsize = 13)
```

In the graph on the left, for the same sample path of nonfinancial income \(y_t\), notice that

- consumption is constant when there are complete markets, but it varies as a random walk in the incomplete markets version of the model
- the consumer’s debt oscillates between two values that are functions of the Markov state in the complete markets model, while the consumer’s debt drifts in a “unit root” fashion in the incomplete markets economy

## Example: Tax Smoothing with Complete Markets¶

It is useful to focus on a simple tax-smoothing example with complete markets.

This example will illustrate how, in a complete markets model like that of Lucas and Stokey [LS83], the government purchases insurance from the private sector.

- Purchasing insurance protects the government against the need to raise taxes too high or issue too much debt in the high-expenditure war event.

We assume that government expenditures move between two values \(G_1 < G_2\), where Markov state \(1\) means “peace” and Markov state \(2\) means “war”

The government budget constraint in Markov state \(i\) is

where

is the price of one unit of output next period in state \(j\) when
today’s Markov state is \(i\) and \(b_i\) is the government’s
level of *assets* in Markov state \(i\)

As above, we’ll assume that the initial Markov state is state \(1\)

In addition, to simplify our example, we’ll set the government’s initial asset level to \(b_0= 0\), so that \(b_1 =0\)

Here’s our Python code to compute a quantitative example with zero debt in peace time:

```
# Parameters
beta = .96
y = [1.0, 2.0]
b0 = 0.0
P = [0.8 0.2;
0.4 0.6]
cp = ConsumptionProblem(beta, y, b0, P)
Q = beta*P
N_simul = 150
c_bar, b1, b2 = consumption_complete(cp)
debt_complete = [b1, b2]
println("P = ", P)
println("Q= ", Q, "\n")
println("Govt expenditures in peace and war =", y)
println("Constant tax collections = ", c_bar)
println("Govt assets in two states = ", debt_complete)
msg = """
Now let's check the government's budget constraint in peace and war.
Our assumptions imply that the government always purchases 0 units of the
Arrow peace security.
"""
println(msg)
AS1 = Q[1,2] * b2
println("Spending on Arrow war security in peace = ", AS1)
AS2 = Q[2,2]*b2
println("Spending on Arrow war security in war = ", AS2)
println("\n")
println("Government tax collections plus asset levels in peace and war")
TB1=c_bar+b1
println("T+b in peace = ",TB1 )
TB2 = c_bar + b2
println("T+b in war = ", TB2)
println("\n")
println("Total government spending in peace and war")
G1= y[1] + AS1
G2 = y[2] + AS2
println("total govt spending in peace = ", G1)
println("total govt spending in war = ", G2)
println("\n")
println("Let's see ex post and ex ante returns on Arrow securities")
Pi= 1./Q#reciprocal(Q)
exret= Pi
println("Ex post returns to purchase of Arrow securities = $exret")
exant = Pi.*P
println("Ex ante returns to purchase of Arrow securities = $exant")
```

Here’s the output it produces

```
P = [0.8 0.2; 0.4 0.6]
Q= [0.768 0.192; 0.384 0.576]
Govt expenditures in peace and war =[1,2]
Constant tax collections = 1.3116883116883118
Govt assets in two states = [0.0,1.62338]
Now let's check the government's budget constraint in peace and war.
Our assumptions imply that the government always purchases 0 units of the
Arrow peace security.
Spending on Arrow war security in peace = 0.3116883116883117
Spending on Arrow war security in war = 0.9350649350649349
Government tax collections plus asset levels in peace and war
T+b in peace = 1.3116883116883118
T+b in war = 2.9350649350649354
Total government spending in peace and war
total govt spending in peace = 1.3116883116883118
total govt spending in war = 2.935064935064935
Let's see ex post and ex ante returns on Arrow securities
Ex post returns to purchase of Arrow securities = [1.30208 5.20833; 2.60417 1.73611]
Ex ante returns to purchase of Arrow securities = [1.04167 1.04167; 1.04167 1.04167]
```

### Explanation¶

In this example, the government always purchase \(0\) units of the Arrow security that pays off in peace time (Markov state \(1\))

But it purchases a positive amount of the security that pays off in war time (Markov state \(2\))

We recommend plugging the quantities computed above into the government budget constraints in the two Markov states and staring

This is an example in which the government purchases *insurance* against
the possibility that war breaks out or continues

- the insurance does not pay off so long as peace continues
- the insurance pays off when there is war

*Exercise:* try changing the Markov transition matrix so that

Also, start the system in Markov state \(2\) (war) with initial government assets \(b_0 = - 10\), so that the government starts the war in debt

## Linear state space version of complete markets model¶

Now we’ll use a setting like that in first lecture on the permanent income model

In that model, there were

- incomplete markets: the consumer could trade only a single risk-free one-period bond bearing gross one-period risk-free interest rate equal to \(\beta^{-1}\)
- the consumer’s exogenous nonfinancial income was governed by a linear state space model driven by Gaussian shocks, the kind of model studied in an earlier lecture about linear state space models

We’ll write down a complete markets counterpart of that model

So now we’ll suppose that nonfinancial income is governed by the state space system

where \(x_t\) is an \(n \times 1\) vector and \(w_{t+1} \sim {\cal N}(0,I)\) is IID over time

Again, as a counterpart of the Hall-Barro assumption that the risk-free gross interest rate is \(\beta^{-1}\), we assume the scaled prices of one-period ahead Arrow securities are

where \(\phi(\cdot \,|\, \mu, \Sigma)\) is a multivariate Gaussian distribution with mean vector \(\mu\) and covariance matrix \(\Sigma\)

Let \(b(x_{t+1})\) be a vector of state-contingent debt due at \(t+1\) as a function of the \(t+1\) state \(x_{t+1}\).

Using the pricing function defined in (14), the value at \(t\) of \(b(x_{t+1})\) is

In the complete markets setting, the consumer faces a sequence of budget constraints

We can solve the time \(t\) budget constraint forward to obtain

We assume as before that the consumer cares about the expected value of

In the incomplete markets version of the model, we shall assume that \(u(c_t) = - (c_t -\gamma)^2\), so that the above utility functional becomes

But in the complete markets version, we can assume a more general form of utility function that satisfies \(u' > 0\) and \(u'' < 0\)

The first-order condition for the consumer’s problem with complete markets and our assumption about Arrow securities prices is

which again implies \(c_t = \bar c\) for some \(\bar c\)

So it follows that

or

where the value of \(\bar c\) satisfies

where \(b_0\) is an initial level of the consumer’s debt, specified as a parameter of the problem

Thus, in the complete markets version of the consumption-smoothing model, \(c_t = \bar c, \forall t \geq 0\) is determined by (16) and the consumer’s debt is a fixed function of the state \(x_t\) described by (15)

Here’s an example that shows how in this setting the availability of insurance against fluctuating nonfinancial income allows the consumer completely to smooth consumption across time and across states of the world.

```
"""
Computes the path of consumption and debt for the previously described
complete markets model where exogenous income follows a linear
state space
"""
function complete_ss(beta::AbstractFloat,
b0::Union{AbstractFloat, Array},
x0::Union{AbstractFloat, Array},
A::Union{AbstractFloat, Array},
C::Union{AbstractFloat, Array},
S_y::Union{AbstractFloat, Array},
T::Integer=12)
# Create a linear state space for simulation purposes
# This adds "b" as a state to the linear state space system
# so that setting the seed places shocks in same place for
# both the complete and incomplete markets economy
# Atilde = vcat(hcat(A, zeros(size(A,1), 1)),
# zeros(1, size(A,2) + 1))
# Ctilde = vcat(C, zeros(1, 1))
# S_ytilde = hcat(S_y, zeros(1, 1))
lss = LSS(A, C, S_y, mu_0=x0)
# Add extra state to initial condition
# x0 = hcat(x0, 0)
# Compute the (I - beta*A)^{-1}
rm = inv(eye(size(A,1)) - beta*A)
# Constant level of consumption
cbar = (1-beta) * (S_y * rm * x0 - b0)
c_hist = ones(T)*cbar[1]
# Debt
x_hist, y_hist = simulate(lss, T)
b_hist = (S_y * rm * x_hist - cbar[1]/(1.0-beta))
return c_hist, vec(b_hist), vec(y_hist), x_hist
end
using PyPlot
N_simul = 150
# Define parameters
alpha, rho1, rho2 = 10.0, 0.9, 0.0
sigma = 1.0
# N_simul = 1
# T = N_simul
A = [1.0 0.0 0.0;
alpha rho1 rho2;
0.0 1.0 0.0]
C = [0.0, sigma, 0.0]
S_y = [1.0 1.0 0.0]
beta, b0 = 0.95, -10.0
x0 = [1.0, alpha/(1-rho1), alpha/(1-rho1)]
# Do simulation for complete markets
s = rand(1:10000)
srand(s) # Seeds get set the same for both economies
out = complete_ss(beta, b0, x0, A, C, S_y, 150)
c_hist_com, b_hist_com, y_hist_com, x_hist_com = out
fig, ax = subplots(1, 2, figsize = (15, 5))
# Consumption plots
ax[1][:set_title]("Cons and income", fontsize = 17)
ax[1][:plot](1:N_simul, c_hist_com, label = "consumption", lw = 3)
ax[1][:plot](1:N_simul, y_hist_com, label = "income",
lw = 2, alpha = 0.6, linestyle = "--")
ax[1][:legend](loc = "best", fontsize = 15)
ax[1][:set_xlabel]("Periods", fontsize = 13)
ax[1][:set_ylim]([-5.0, 110])
# Debt plots
ax[2][:set_title]("Debt and income", fontsize = 17)
ax[2][:plot](1:N_simul, b_hist_com, label = "debt", lw = 2)
ax[2][:plot](1:N_simul, y_hist_com, label = "Income",
lw = 2, alpha = 0.6, linestyle = "--")
ax[2][:legend](loc = "best", fontsize = 15)
ax[2][:axhline](0, color = "k", lw = 1)
ax[2][:set_xlabel]("Periods", fontsize = 13)
```

Here’s the resulting figure

### Interpretation of graph¶

In the above graph, please note that:

- non financial income fluctuates in a stationary manner
- consumption is completely constant
- the consumer’s debt fluctuates in a stationary manner; in fact, in this case because non-financial income is a first-order autoregressive process, the consumer’s debt is an exact affine function (meaning linear plus a constant) of the consumer’s non-financial income

### Incomplete markets version¶

The incomplete markets version of the model with nonfinancial income being governed by a linear state space system is described in the first lecture on the permanent income model and the followup lecture on the permanent income model

In that version, consumption follows a random walk and the consumer’s debt follows a process with a unit root

We leave it to the reader to apply the usual isomorphism to deduce the corresponding implications for a tax-smoothing model like Barro’s [Bar79]

### Government manipulation of Arrow securities prices¶

In a linear quadratic Ramsey problem and recursive optimal taxation, we study **complete-markets**
models in which the government recognizes that it can manipulate Arrow securities prices

In optimal taxation with incomplete markets, we study an **incomplete-markets** model in which the government manipulates asset prices