You are on page 1of 39

Annual Review of Economics

Computing Economic
Equilibria Using
Projection Methods
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Alena Miftakhova,1 Karl Schmedders,2


and Malte Schumacher3
1
Department of Management, Technology, and Economics, Eidgenössische Technische
Hochschule (ETH) Zurich, 8092 Zurich, Switzerland; email: alenam@ethz.ch
2
International Institute for Management Development (IMD), 1003 Lausanne, Switzerland;
email: karl.schmedders@imd.org
3
Department of Business Administration, University of Zurich, 8044 Zurich, Switzerland;
email: malte.schumacher@business.uzh.ch

Annu. Rev. Econ. 2020. 12:317–53 Keywords


First published as a Review in Advance on
dynamic models, equilibria, nonlinearities, projection methods
May 29, 2020

The Annual Review of Economics is online at Abstract


economics.annualreviews.org
The analysis of dynamic economic models routinely leads to the mathemati-
https://doi.org/10.1146/annurev-economics-
cal problem of determining an unknown function for which no closed-form
080218-025711
solution exists. Economists must then resort to methods of numerical ap-
Copyright © 2020 by Annual Reviews.
proximation when analyzing such models. Among the computational meth-
All rights reserved
ods that have been successfully applied in economics and finance, one set of
JEL codes: C61, C63, D58, G12, J64, Q54
techniques stands out due to its flexibility and robustness: projection meth-
ods. In this article, we describe the basic steps of these methods for sev-
eral different applications, surveying many successful applications of pro-
jection methods to dynamic economic models. Importantly, we emphasize
that the ever-increasing complexity and dimensionality of dynamic models
have made the previously used simpler methods obsolete and the applica-
tions of projection methods all but mandatory. We closely examine the most
recent endeavors in the literature on solving economic models with projec-
tion methods.

317
1. INTRODUCTION
Modern research in economics and finance is dominated by the analysis of dynamic models. In dy-
namic economics, key economic objects change over time and economic agents react to changes
in economic conditions. In many applications, the analysis of dynamic economic models leads
to the mathematical problem of determining an unknown function. Examples of such problems
arising in economic and financial modeling include solving differential equations in continuous-
time finance and economics, computing value functions that solve the Bellman equation in a dy-
namic programming problem, computing price and policy functions from Euler equations and
market-clearing conditions in dynamic equilibrium models, determining price/dividend ratios in
consumption-based asset-pricing models, and many other problems. All too often, the solutions
to these problems have no closed form. Economists must then resort to approximation techniques
in the analysis of dynamic models.
Traditionally, economists solving difficult models often ignored standard numerical methods
Access provided by Arizona State University on 10/21/21. For personal use only.

in favor of idiosyncratic and ad hoc methods with poorly understood properties that produced
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

approximate solutions of unknown quality. Judd (1998, p. 19) described the state of the art in
the late 1990s as follows: “Many ‘innovations’ in computational economics are straightforward
applications of standard numerical methods, inferior reinventions of earlier work, or ad hoc pro-
cedures with little mathematical foundation.” While Judd’s criticism has remained valid well into
the twenty-first century, the innovations of the information age during the first two decades of
the new century have begun to engulf economics and finance. Economists increasingly use so-
phisticated numerical methods as well as the latest available hardware and software in the analysis
of their models. This rapid growth in sophistication has also been apparent in the function ap-
proximations necessary in dynamic economics. Among several classes of useful methods applied
by economists, one particular class of methods stands out due to its flexibility and robustness:
projection methods.
Projection methods were developed by applied mathematicians seeking numerical solutions
to differential equations during the 1960s and 1970s (see Reddien 1980 and the many citations
therein). Although such methods had also been applied in some economic work in an ad hoc fash-
ion (see, for example, Taylor & Uhlig 1990), Judd (1992) formally introduced projection methods
to economics. Importantly, Judd observed that these methods are so flexible that they can also
be applied to many other economic problems, such as solving discrete-time dynamic economic
problems in which equilibria are characterized by functions on a continuous domain.
Applying projection methods requires that we formulate economic equilibrium conditions as a
functional equation N ( f ) = 0, where the operator N is a mapping between function spaces. The
equation N ( f ) = 0 could, for example, be a differential equation, a Bellman equation of a dynamic
programming problem, or a system of necessary first-order conditions and market-clearing equa-
tions in a recursive equilibrium problem. The solution function f is then the solution function to
the differential equation, the value function of the dynamic programming problem, or the vector
of equilibrium policy and price functions defined on an appropriate state space, respectively. For
most dynamic economic models, the functional equation does not have a closed-form solution.
Because it is impossible to represent an exact solution on a continuous domain on a computer,
all we can hope for is to find a finite representation of an approximate solution fˆ that satisfies
N ( fˆ ) ≈ 0. For this purpose, we restrict the search to a set of functions spanned by a finite set of

basis functions, {φ j }nj=0 . Any candidate solution fˆ = nj=0 a j φ j is then a linear combination of the
basis functions. Naturally, we want to choose the weights {a j }nj=0 such that some chosen norm of

the residuals N ( fˆ ) = N ( n a j φ j ) over the domain of interest for the function f is sufficiently
j=0
small in a chosen sense—for example, a norm. The minimization of some error norm can often be

318 Miftakhova • Schmedders • Schumacher


rewritten as solving a system of projection conditions, which gives the computational technique
its name. An important feature of projection methods is that they incorporate information con-
cerning N ( f ) not only at a single point, but at several points over the domain of f. Consequently,
economists often describe projection approaches as global methods.
In this article, we provide a survey of many successful applications of projection methods to
dynamic economic models. In particular, we point to some models in which the application of
projection methods has resulted not only in more accurate quantitative results but also in new
qualitative insights; in some cases, these insights substantially alter the models’ economic im-
plications. Therefore, we strongly encourage economists to include projection methods in their
computational toolboxes, particularly if they want reliable solutions to modern dynamic economic
models.
A second set of approximation methods—as an alternative to projection methods—has en-
joyed widespread application during the last two decades, particularly in macroeconomic model-
ing. Judd & Guu (1993, 1997) and Judd (1996) popularized perturbation methods in economics.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Macroeconomists enthusiastically adopted perturbation methods because they are a natural ex-
tension of linearizations, which have a long history of applications in macroeconomics since the
emergence of work by Hall (1971) and Magill (1977). Perturbation methods construct an approx-
imation of an unknown function around a particular point in the domain at which the function
value is known exactly. For example, a stochastic economic model may have a deterministic steady
state for the degenerate case of no uncertainty. Perturbation methods then build an approximation
around this point relying on implicit function theorems and several orders of derivatives at this
point. In contrast to the global projection approach, economists refer to perturbation techniques
as local methods.
Naturally, the question arises of which of the two sets of techniques works better on dynamic
economic models. In their pioneering paper, Gaspar & Judd (1997) compare these methods across
a variety of dynamic economic problems with respect to computational speed and accuracy. They
conclude that perturbation methods are much faster than projection methods (on the comput-
ers available at the time of their research) and are capable of producing good approximations if
the ergodic distribution of the dynamic model has a small support. In the presence of substantial
nonlinearities, however, they conclude that projection methods should be used. About a decade
later, Aruoba et al. (2006) compare the performance of perturbation and projection methods in
the case of the neoclassical growth model. These authors strongly argue that economists should
abandon linear and log-linear approximations whenever possible in favor of higher-order pertur-
bation methods and projection methods. They prefer perturbation methods, considering them
a good compromise between accuracy, speed, and programming burden, although they do ac-
knowledge the excellent performance of projection methods with finite elements or Chebyshev
polynomials. Caldara et al. (2012) compare computational methods for solving dynamic stochas-
tic general equilibrium (DSGE) models with recursive utility and stochastic volatility and find
that perturbation methods are competitive in terms of accuracy with Chebyshev polynomials but
offer significantly faster speed. They also conclude that when numerical errors are the key con-
cern, projection with Chebyshev polynomials dominate. Fernández-Villaverde & Levintal (2018)
compare computational methods for solving DSGE models with rare disasters and find that even
higher-order perturbation methods sometimes produce unacceptably large numerical errors.
The general conclusion of comparison studies in the economic literature is clear. For compar-
atively simple models, perturbation methods can produce approximations of sufficient accuracy
faster than projection methods. For more complex models, in most cases we must abandon local
methods and apply projection methods. Such cases include, first of all, models in which economic
constraints (for example, borrowing or collateral constraints, or zero-lower-bound constraints)

www.annualreviews.org • Computing Economic Equilibria 319


lead to nondifferentiable solution functions. Another such class includes models in which large
shocks (such as rare disasters or large persistence of shocks) result in large domains for the equi-
librium functions. The third such category are models in which these functions are differentiable
but exhibit large curvature on their domain (for example, as a result of utility functions with habit
or large risk aversion, or for some types of constraints).
Since many economic applications in recent years have contained one or several of the de-
scribed difficulties, we encourage the use of projection methods when solving the latest generation
of dynamic economic models. We believe that considerations regarding computational speed are
of increasingly smaller concern in the age of the digital revolution and that our focus should turn
instead to computational accuracy. Therefore, we exclusively focus on projection methods below.
The remainder of this article is organized as follows: In Section 2, we present the basic ideas
of projection methods in a context that is familiar to economists. In Section 3, we provide a brief
description of general projection methods and then stress the advantages of Chebyshev bases.
In Sections 4 and 5, we survey successful applications of projection methods in macroeconomics
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

and finance, respectively, and in Section 6, we describe applications in other economic fields such
as microeconomics, econometrics, and climate-change economics. In addition, we describe some
exciting new endeavors on the current research frontier. We conclude in Section 7.

2. FIRST EXAMPLE
For a first look at projection methods, we apply two such methods to a simple ordinary differential
equation (ODE). We choose this particular example because the ODE has a simple analytical
solution, which helps us illustrate the analysis of numerical errors in a projection method. We
choose these two particular methods because they are rather straightforward and intuitive.

2.1. Problem Setup


Consider the problem of finding the function f : [0, T ] → R satisfying the ODE

∂f
− r f = 0, 1.
∂t

with the boundary condition

f (T ) = 1, 2.

for positive constants r and T. This ODE has the analytical solution

f ∗ (t ) = er(t−T ) . 3.

Our task is now to solve the ODE in Equation 1 numerically, subject to the boundary condition
in Equation 2. Following Judd (1992), the first step of a projection method requires us to choose
a set of basis functions with which we want to approximate the (unknown) function f ∗ . A simple
set of basis functions is a set of ordinary polynomials, {1, t, t2 , . . . , tn }, for some positive integer n.
The approximation fˆ is then defined as


n
fˆ (t; a) = a jt j , 4.
j=0

320 Miftakhova • Schmedders • Schumacher


with the coefficients aj for j ࢠ {0, 1, . . . , n}. Observe that the choice of a finite value for n reduces
the infinite-dimensional problem of finding a general function f to the finite-dimensional problem
of finding a vector of coefficients a ∈ Rn+1 .
If we replace the unknown function f by the approximation fˆ in the ODE in Equation 1, we
obtain the residual function

∂ fˆ
R(t; a) = − r fˆ ,
∂t

which leads to


n 
n
R(t; a) = a j jt j−1 − r a jt j . 5.
j=1 j=0
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Similarly, replacing f by fˆ in the boundary condition of Equation 2, the following obtains:


n
fˆ (T ; a) = a j T j = 1. 6.
j=0

Ideally, we should like to find a coefficient vector a∗ so that the residuals satisfy R(t; a∗ ) ≡ 0 on
the domain [0, T ], but since the true function is not polynomial, such coefficients do not exist.
Instead, the approximation will typically lead to errors |R(t; a)| > 0. Therefore, in the next step
of a projection method, we need to choose the projection conditions—that is, conditions on the
residual function R that a satisfactory approximation fˆ must satisfy.

2.2. Least-Squares Projection


Economists are most familiar with least-squares methods due to their common training in econo-
metrics. Therefore, a natural starting point for finding a suitable approximation fˆ is to determine
the coefficient vector a ∈ Rn+1 by minimizing a sum of squared residuals under the constraint in
Equation 6. For this purpose, we choose a finite number, m + 1, of points on the domain [0, T];
for example, we can choose the equally spaced grid
  
T 
[0,T ] =
Gm i  i = 0, 1, 2, . . . m . 7.
m

Therefore we can consider the following constrained least-squares problem,


⎛ ⎞2
 n 
n 
n
min ⎝ a j jt j−1
−r a jt ⎠ ,
j
subject to a j T j = 1. 8.
a∈Rn+1
t∈Gm
[0,T ]
j=1 j=0 j=0

Next, we must solve the constrained minimization problem in Equation 8 in the unknown
coefficients a ∈ Rn+1 . In our specific instance, the residual function is linear in the elements of a.
Consequently, the problem in Equation 8 is a straightforward quadratic programming problem,
which even allows for a closed-form solution. For our purposes here, a numerical solution, denoted
by â, suffices.

www.annualreviews.org • Computing Economic Equilibria 321


a Relative errors b Residuals
1 × 10–6
5× 10–7

0 5 × 10–7
f̂(t;â) – f*(t) –5 × 10–7
R(t;â) 0
f*(t)
–1 × 10–6
–5 × 10–7
–1.5 × 10–6
–1 × 10–6
2 4 6 8 10 2 4 6 8 10
t t
Figure 1
fˆ (t;â)− f ∗ (t )
(a) A plot of the relative errors, f ∗ (t )
, in the approximated ODE solution function on the entire
domain [0, T]. (b) The residuals, R(t; â). Abbreviation: ODE, ordinary differential equation.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

We can now compute an approximate solution for given constants r, T > 0 and parameters m
[0,10] = {0, 1, 2, . . . , 10} and obtain
and n. For r = 0.05, T = 10, m = 10, and n = 4, we have G10

fˆ (t; â) = 0.606530 + 0.0303258t + 7.59394 · 10−4 t 2 + 1.22365 · 10−5 t 3 + 2.03659 · 10−7 t 4 .

In the final step of a projection method, we want to verify that our solution yields a reason-
able approximation. Since we have an exact solution in Equation 3 for the model, we can easily
determine the relative approximation errors,

fˆ (t; â) − f ∗ (t )
∀ t ∈ [0, T ]. 9.
f ∗ (t )

Figure 1a shows a plot of these relative errors. Due to the constraint fˆ (T ; a) = 1, the relative
error in the solution function fˆ (t; â) is zero at t = T = 10. The maximal relative error is less than
2 × 10−6 . In typical applications of projection methods, however, we do not know an exact solution
f ∗ , so we cannot determine the relative errors of Equation 9 in the approximation of the unknown
function. Instead, all we can do is to compute the residuals R(t; â). Figure 1b shows a plot of the
residuals. We observe that the maximal absolute residual is less than 10−6 . If we consider this error
to be too large, we could now try to increase the degree n of the approximating polynomial in the
hope of reducing the numerical errors.

2.3. Uniform Collocation


Another intuitive approach to the approximation of the unknown function f ∗ is uniform colloca-
tion. This approach attempts to set the residual errors on a grid of points to zero.
We start again with an approximating polynomial fˆ as in Equation 4; and so, obviously, the re-
sulting residual function in Equation 5 and the boundary condition in Equation 6 remain the same
as in the least-squares approach. For each grid point ti ∈ Gm
[0,T ] , we obtain an equation R(ti , a) = 0.
In addition to these (m + 1) linear equations in the unknown coefficient vector a ∈ Rn+1 , we con-
tinue to have the boundary condition in Equation 6, which is also linear in a. Thus, we have (m + 2)
linear equations in (n + 1) unknowns. We choose the grid size (m + 1) and the degree n of the

322 Miftakhova • Schmedders • Schumacher


a Relative errors b Residuals
0
8 × 10–10
–5 × 10–10 6 × 10–10
f̂(t;â) – f*(t) –1 × 10–9 R(t;â)
4 × 10–10
f*(t) –1.5 × 10–9 2 × 10–10
0
–2 × 10–9
–2 × 10–10
2 4 6 8 10 2 4 6 8 10
t t
Figure 2
fˆ (t;â)− f ∗ (t )
(a) A plot of the relative errors, f ∗ (t )
, in the approximated ODE solution function on the entire
domain [0, T]. (b) The residuals, R(t; â). Abbreviation: ODE, ordinary differential equation.

approximating polynomial such that m + 1 = n in order to obtain a square system:


Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org


n
j−1

n
j
a j jti −r a j ti = 0 ∀ ti ∈ Gn−1
[0,T ] , 10.
j=1 j=0


n
a j T j = 1. 11.
j=0

The system is solved for given positive constants r and T and parameters m + 1 = n. For r =
0.05, T = 10, and m = 5 (so n = 6), the linear system has seven equations and unknowns. We have
G5[0,10] = {0, 2, 4, 6, 8, 10} and obtain

fˆ (t; â) = 0.606531 + 0.0303265t + 7.58165 · 10−4 t 2 + 1.26348 · 10−5 t 3 + 1.58317 · 10−7 t 4
+ 1.52554 · 10−9 t 5 + 1.69504 · 10−11t 6 .

Figure 2a depicts a plot of the relative errors in Equation 9. Recall that without knowing the
exact solution we would not be able to determine these errors. Instead, all we can typically do is
to compute the residuals R(t; â) (see Figure 2b for a plot of the residuals). We observe that the
residuals are indeed zero at the six elements of G5[0,10] = {0, 2, 4, 6, 8, 10}. In addition, the relative
error in the solution function fˆ (t; â) is also zero at t = T = 10 due to the constraint fˆ (T ; a) = 1.

3. PROJECTION METHODS: BRIEF INTRODUCTION


3.1. Main Steps
We briefly describe the main steps of a projection method in general terms.

3.1.1. Step 1: identifying the functional equation describing the economic problem. For
an application of projection methods, we need to formulate the economic problem as finding a
function f : D ⊂ RK → RM that must satisfy a functional equation

N ( f ) = 0, 12.

where N is a mapping (i.e., an operator) from a function space B1 to a function space B2 . Note that
f ࢠ B1 and N ( f ) ∈ B2 . Equation 12 is satisfied if and only if N ( f (x)) = 0 for all x ࢠ D. The

www.annualreviews.org • Computing Economic Equilibria 323


operator N may depend on additional parameters, which then also affect the solution to Equation
12. (For ease of notation, we suppress this dependence.)
For the simple ODE in the previous section, we have K = 1, M = 2, and D = [0, T]. The
function f : D → R must be smooth and so we obtain B1 = C∞ (D). The operator N satisfies
∂ f (t )
∂t
− r f (t )
N ( f (t )) = , 13.
f (T ) − 1

with t ࢠ D and N ( f ) ∈ B2 = C ∞ R2 . Note that, as is typically the case in economic applications,
the unknown function f is an element of an infinite-dimensional vector space.

3.1.2. Step 2: choosing a representation for approximate solutions. On a computer, we can


neither represent a general function f nor can we check the general functional Equation 12 at all
points of a function f in its domain D. Instead, we can only consider finite-dimensional objects.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Therefore, we must restrict attention to functions fˆ that have a finite representation. And so, for
some finite n ∈ N, we choose an (n + 1)-dimensional vector space of functions V n that is a subset
 n
of B1 . Denote by n = φ j j=0 a basis of V n , so that we obtain
⎧ ⎫
⎨ 
n ⎬
Vn = g | g(x; a) = a j φ j (x), x ∈ D, a ∈ Rn+1 . 14.
⎩ ⎭
j=0

The projection method then restricts attention to functions fˆ ∈ V n ⊂ B1 in the search for an
approximate solution N ( fˆ ) ≈ 0.
For the ODE in Section 2, the basis n is the set {1, t, t2 , . . . , tn } of the first (n + 1) monomials,
which spans the (n + 1)-dimensional vector space of polynomials of degree less than or equal to
n. Hence, we obtain
⎧ ⎫
⎨ 
n ⎬
fˆ ∈ g | g(t; a) = a j t , t ∈ [0, T ], a ∈ R
j n+1
.
⎩ ⎭
j=0

3.1.3. Step 3: defining a residual function. Projection methods return a function fˆ ∈ V n that
yields a reasonable approximation of the true f. For the evaluation of possible approximations, we
define the residual function R : D × Rn+1 → RM by R(·; a) = N ( fˆ ) with the term
⎛ ⎞

n
R(x; a) = N ⎝ a j φ j (x)⎠. 15.
j=0

The residual R(x; a) is the error in the equation N ( fˆ (x; a)) = 0. Note that we do not approximate
the operator N but rather search for an approximation fˆ of the unknown function f. Judd (1992)
points out that in many economic applications the operator N cannot be represented exactly on
a computer, and thus it requires an approximation as well.
For the ODE in Section 2, we obtain the residual R by replacing f (t) by fˆ (t; a) in
Equation 13:
n n
j=1 ja j t j−1 − r j=0 a jt j
R(t; a) = n . 16.
j=0 ajT j − 1

324 Miftakhova • Schmedders • Schumacher


3.1.4. Step 4: determining an approximate solution. We choose conditions that the approx-
imation fˆ ∈ V n must satisfy. The perhaps simplest conditions require the residuals to be zero on
a finite subset G of the domain D, that is,

R(x; a) = 0 ∀ x ∈ G ⊂ D. 17.

Typically, we choose the number of elements in G to be (n + 1) and the number of unknown


coefficients to be a ∈ Rn+1 . In that case we obtain a square system of (n + 1) equations in (n + 1)
unknowns, which we can solve with a nonlinear (or even linear) equation solver.
Popular alternative sets of conditions require a function F : RM → RM of the residuals to be
zero on a finite subset G of the domain D, which obtains

F (R(x; a)) = 0 ∀ x ∈ G ⊂ D. 18.

Yet another alternative approach to determining the unknown coefficients a ∈ Rn+1 is to min-
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

imize the L2 norm,


·
2G , of the residuals on a subset G ⊆ D, so that

min
R(x; a)
2G such that a ∈ Rn+1 . 19.

Solving a system of equations or an optimization problem yields a solution â and a function



term fˆ (x; â) = nj=0 â j φ j (x) approximating the true solution f ∗ (x).
The least-squares approach in Section 2.2 shows an example of the optimization approach in
Equation 19 in the context of the ODE example (see the problem in Equation 8). The uniform
collocation approach in Section 2.3 is an example of the general equation approach in Equation 17
(see Equations 10 and 11).

3.1.5. Step 5: evaluating the approximate solution. We evaluate the quality of the approxi-
mate solution by determining a close upper bound on the residuals R(x; â) on the domain D. If we
are assured that

max |R(x; â)| ≤  20.


x∈D

for some predefined error tolerance  (for example,  = 10−8 ), then we accept the computed solu-
tion fˆ (x; â) as an acceptable approximation for the true function term f ∗ (x). If, however, the max-
imal residuals exceed the error tolerance, then we change the function representation for fˆ (x; a),
for example, by increasing the number of basis functions. We return to Step 4 and compute a new
approximation. We repeat these steps until the condition in Equation 20 is satisfied.
The least-squares approach in Section 2.2 delivered an upper bound on the residuals of 10−6 .
This bound exceeds  = 10−8 . So, if we require the maximal residual to fall below this bound, we
may want to increase n, repeat the optimization, and check whether we can reduce the error. The
uniform collocation approach in Section 2.3 delivers an upper bound on the residuals below 10−9
and thus satisfies our error requirement.

3.2. What’s in a Name?


The description in Step 4 of the different methods to determine suitable coefficient vectors a ∈
Rn+1 is quite tedious. The mathematics of projection—which gives the methods their general
name—allows us to describe all the methods much more succinctly in a uniform fashion. This
convenient mathematical description somewhat hides the numerical solution methods.

www.annualreviews.org • Computing Economic Equilibria 325


Consider inner products, ·,·, of the vector space B2 of the form

 
f (x), g(x) ≡ f (x)g(x)w(x)dx
D

for some weighting function w(x). Using inner products, we can now formally state all methods
in Step 4 in a standardized fashion. All methods determine the unknown coefficients a such that

R(x; a), p j (x) = 0 21.

for some specified set of functions pj ࢠ B2 , j = 0, 1, . . . , n. The conditions in Equation 21 require


that the residual term R(x; a) be orthogonal to the chosen functions pj ࢠ B2 . Put differently, the
projection of the residual function on each of the “directions” pj ࢠ B2 must be zero.
Different sets of basis functions pj ࢠ B2 lead to different projection methods. In the least-
squares projection method, the functions pj ࢠ B2 derive from the first-order conditions of the
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

problem, and the inner products in Equation 21 become


 
∂R(x; a)
R(x; a), = 0,
∂a j

with the weighting function w(x) = 1 for all x ࢠ D. In econometrics these conditions are often
formulated in discrete form on a finite set of data points G ⊆ D, so that we have

 ∂R(xi ; a)
R(xi ; a) = 0, i ∈ {0, 1, . . . , |G| − 1}.
i
∂a j

In the uniform-collocation projection method, we can write an equation R(xi ; a) = 0 at a grid


point xi ࢠ D as
 
R(x; a), δ(x − xi ) = 0,

with δ denoting the Dirac delta function [and the weighting function w(x) = 1 for all x ࢠ D].
In sum, although the notation via projections allows for mathematically elegant statements,
these are not directly helpful to an explanation or in the implementation of projection methods.

3.3. Projection Methods with Chebyshev Bases for D = [L, U]


Steps 1, 3, and 5 are very similar across projection methods and require few if any decisions by the
user. In many applications of projection methods in economics, V n in Step 2 is a vector space of
functions with a polynomial basis. The methods differ only in the choice of basis functions. The
strongest point of differentiation among projection methods are the conditions on the residuals
in Step 4, which determine an approximation fˆ of f ∗ .
In the following discussion, we restrict ourselves to finding an unknown function f with a do-
main D = [L, U]. Obviously, many economic problems lead to functional equations N ( f ) = 0
with f having a multidimensional domain D ⊂ RK and K > 1. We cite many papers with multidi-
mensional domains in the discussion below.
In the simple ODE example in Section 2, we applied first a least-squares approach with or-
dinary polynomials and a uniform grid and then uniform collocation. Even though these two

326 Miftakhova • Schmedders • Schumacher


methods may come natural to many economists and are, thus, a good starting point to learn about
projection methods, we advise against using them for serious economic applications. For the basis
in Step 2 we prefer to use Chebyshev polynomials instead of ordinary polynomials—that is, the

approximation is of the form nj=0 a j Tj (x), where Tj (x) denotes the degree-j Chebyshev polyno-
mial (of the first kind). Trefethen (2013) provides a beautiful introduction to the approximation
of one-dimensional functions in theory and practice. He strongly advocates interpolation meth-
ods relying on Chebyshev polynomials for the practical approximation of functions on intervals,
calling such methods “unbeatable.”
If we want to use a collocation method in Step 4 (see Equation 17), then we must first choose
a grid G. An excellent grid to determine the coefficient vector a ∈ Rn+1 in Equation 15 is given
by the zeros of the degree (n + 1) Chebyshev polynomial. With these grid points, the approach
is called the Chebyshev collocation method. Since the Chebyshev zeros lie in the interval [−1,
1], we need a bijective linear transformation from this interval to [L, U]. For x ࢠ [L, U] and y ࢠ
[−1, 1], we use
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

x−L U −L U +L U −L
y=2 − 1 and x= y+ = (y + 1) + L. 22.
U −L 2 2 2
The resulting grid for the problem is then given by
⎧  ⎫
⎨U − L    ⎬
U + L  2i − 1
G= y+  y = cos π , i = 1, 2, . . . , n + 1 , 23.
⎩ 2 2  2(n + 1) ⎭

 
where cos 2i−1
2(n+1)
π is a convenient representation of the Chebyshev zeros. The approximation
to f (x) is


n  
x−L
fˆ (x; a) = a j Tj 2 −1 , 24.
j=0
U −L

where we retransform the grid G to ensure the evaluation of Chebyshev polynomials on [−1, 1].
After replacing the unknown term f (x) by fˆ (x; a) in the residual R, we must then solve the resulting
square system of residual equations.
An alternative approach, which closely follows the language of projection, is Galerkin’s method.
The conditions to pin down the coefficients are orthogonality conditions on the residual function
and the basis functions. On the domain [−1, 1], these conditions are
 1
1
R(y; a)Tj (y)  dy = 0 for j = 0, 1, . . . , n.
−1 1 − y2

Since we need these conditions typically on a different domain [L, U], we apply again the linear
change of variables in Equation 22. For a general residual R(x; a) on [L, U], the Galerkin conditions
are
 U  
x−L 1
R(x; a)Tj 2 −1 2 dx = 0 for j = 0, 1, . . . , n. 25.
U − L
L
1 − 2 Ux−L
−L
−1

Before we can solve the system in Equation 25 we must evaluate the integral in each equation. Due
to the inclusion of the particular weight function, a great choice for the numerical evaluation of

www.annualreviews.org • Computing Economic Equilibria 327


the integrals is Gauss–Chebyshev quadrature (see Judd 1998). The general approximation formula
is

π 
1 k
g(z)
√ dz = g(zi ),
−1 1 − z2 k i=1

with quadrature nodes


 
2i − 1
zi = cos π for i = 1, 2, . . . , k.
2k

U −L
Applying the transformation in Equation 22 including dx = 2
dz and this quadrature rule to the
integral in Equation 25, we obtain the system of equations
Access provided by Arizona State University on 10/21/21. For personal use only.

 
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

U −Lπ 
k
U −L
R (zi + 1) + L; a Tj (zi ) = 0 for j = 0, 1, . . . , n. 26.
2 k i=1 2

We must now solve this square system of equations in the unknown coefficient vector a to obtain
the desired approximation fˆ (x; â) of f (x).
Observe that both the Chebyshev collocation method and Galerkin’s method require us to
solve a square system of equations. Even though in some rare cases the system may be linear,
we typically face a nonlinear system of equations, and such a nonlinear system may have several
solutions.

3.4. Beyond Chebyshev Polynomials in One Dimension


To simplify our introduction to projection methods, we focus on the approximation of an unknown
function f in a single variable. However, these methods would not have been so successful in eco-
nomic applications had they been limited to one-dimensional problems. Our survey of successful
applications in Sections 4–6 contains many dynamic models that require the approximation of
functions with several variables. Historically, economists have been wary of projection methods in
three or more dimensions because of their exponentially increasing computationally burden (see
Gaspar & Judd 1997, Aruoba et al. 2006, Caldara et al. 2012). Huge advances in hardware as well
as algorithms and their software implementations mean that economists can now solve models of
increasingly large dimensions with projection methods (see our discussion in Sections 4.2 and 6).
Also, for reasons of simplicity, in this review we apply only Chebyshev bases for the numerical
solutions of economic applications. We justify this focus with our experience that these bases de-
liver approximations of sufficient accuracy for many applications. Nevertheless, we also acknowl-
edge that Chebyshev bases do not deliver the best approximations for all models. For example,
Judd et al. (2003) choose cubic splines for their collocation method to solve heterogeneous-agent
incomplete-market models because the price and policy functions in these models exhibit re-
peated changes in curvature, so that Chebyshev collocation methods do not deliver acceptably
small numerical errors. More generally, certainly models with high-dimensional domains or non-
differentiable solution functions may require different bases for accurate solutions. Versatile basis
functions are low-order polynomials that are defined only on a subset of the domain, such as piece-
wise linear functions or other splines. (For introductions to splines we recommend de Boor 1978
and Schumaker 2007.) These methods are examples of finite-element methods (see Silvester &
Ferrari 1996). We describe great recent advances for economic models with large-dimensional
state spaces and nondifferentiable solutions in Section 6.

328 Miftakhova • Schmedders • Schumacher


4. MACROECONOMICS
Once formally introduced, projection methods found their immediate and extensive use in
macroeconomics, where the modeled phenomena are complex, dynamic, and ever-changing in
their nature. Originally, Judd (1992) demonstrated their application on optimal growth problems
and suggested their potential use for many more types of typical dynamic economic models.
As the modelers’ ambition grew, so did the dimensionality and complexity of the constructed
problems—incorporating heterogeneity, aggregate uncertainty, and (nontrivial) constraints. Even
though alternative, often cheaper, methods have also received attention in this literature, the many
examples of inefficient or obscure—not to mention incorrect—solutions have led projection
methods to gradually become the dominant toolkit for accurate and reliable solutions.

4.1. In Search of the Best Method


Modeling macroeconomic fluctuations observed in the real world requires nontrivial features—
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

such as heterogeneity across countries, aggregate shocks, and frictions—embedded in simple


Walrasian models. These models quickly grow out of simple and even any feasible analytical so-
lutions; a variety of computational solution methods have therefore emerged in the last decades,
together with a natural need to understand how these methods compare.
This need gave rise to large computational exercises, which applied available techniques of
perturbation, projection, and stochastic simulations to nontrivial DSGE models. In the first such
experiment, the economy is characterized by incomplete financial markets, aggregate productivity
shocks, and a continuum of agents who face idiosyncratic income shocks and a borrowing con-
straint (Den Haan et al. 2010). A particularly challenging feature is the fact that the distribution
of income and wealth levels across the agents is a high-dimensional part of the state space. Some
methods therefore rely—to a different extent—on simulations to infer this cross-sectional dis-
tribution. The simulation-based methods used by Maliar et al. (2010) and Young (2010) rely on
projection methods to solve for the individual policies and on simulations to derive the aggregate
law of motion. Two methods—backward induction (Reiter 2010) and parameterized distribution
(Algan et al. 2010)—use simulations to learn about the shape of the cross-sectional distribution and
then proceed with the projection routines to actually solve for individual policies and the aggre-
gate law of motion. The algorithm of Den Haan & Rendahl (2010) relies on projection techniques
only. This projection approach turns out to be both extremely fast and accurate, though it com-
petes with the backward induction in accuracy. The key components of the successful solution
methods are the smart choices of numerical procedures (such as time iteration instead of fixed-
point iteration) and of the grid over the state space with higher density of nodes in critical areas
(for example, where the constraint nearly binds).
Another computational experiment collates the methods of perturbation, projection, and
stochastic simulations and applies them to a DSGE model with 10 countries (20 continuous state
variables), heterogeneous preferences and technologies, and both country-specific and worldwide
productivity shocks (Den Haan et al. 2011). In a summary of the results by Kollmann et al. (2011),
unsurprisingly, projection methods clearly dominate the other methods in terms of accuracy.
Curiously, the efficiency of each projection method varies greatly with the choices and combi-
nations of the individual components—integration rule, approximation technique and precision,
type of grid over the state space, and approach to the problem itself. In all three projection
methods that participate in the experiment by Kollmann et al. (2011), the classical formulation by
Judd (1992) is modified in a certain way to handle the size of the problem. The primary focus of
the Galerkin method implemented by Pichler (2011) is the high dimensionality at the integration
steps of the algorithm. This model therefore substitutes the conventional, computationally costly

www.annualreviews.org • Computing Economic Equilibria 329


Gaussian methods with much cheaper, nonproduct monomial rules for computing both condi-
tional expectations in the model’s Euler equations and weighted residuals. These nonproduct
integration rules operate on a sparse grid of points in the state space; as a result, computational
costs increase only polynomially, and not exponentially, in the dimension of the problem, while
maintaining the accuracy at acceptable levels. The key feature of the Smolyak collocation
method implemented by Malin et al. (2011) is a special sparse grid of points for approximation,
which utilizes the extrema of Chebyshev polynomials as a basis. The same polynomial family
is subsequently used to interpolate the solution; a time-iteration algorithm efficiently finds the
coefficients for such approximation. Here, too, monomial rules for integration are applied to
ease the computation of conditional expectations. Finally, both the cluster-grid algorithm (CGA)
( Judd et al. 2010) and the stochastic simulation algorithm (SSA) ( Judd et al. 2011)1 employed
by Maliar et al. (2011) limit their attention to the ergodic set of points—that is, the constructed
grid covers only the area of the state space that is likely to be visited in equilibrium and spares
the effort of solving the model elsewhere. SSA uses a set of points simulated on this subspace as
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

a grid, while CGA further clusters these simulated points to represent each such cluster with one
central grid point. Besides, based on the very structure of the model, the optimal policy is split
into the optimal consumption and labor (static, intratemporal policy) and the optimal rule for
capital (dynamic, intertemporal policy). Solving the intratemporal problem in isolation—that is,
conditioned on the choice for capital—greatly improves the accuracy of the overall solution to
the model. It is therefore a smart choice of the approximation, integration, and grid construction
techniques—together with an informed approach to the problem itself—that delivers the best
resolution of the trade-off between computational costs and the accuracy of the obtained solution.

4.2. Overcoming Classical Limitations


The long record of successful applications of projection methods in macroeconomics would push
this section way beyond its page limits; we nevertheless name a few interesting examples, starting
with advances in classical models.
Since the establishment of the neoclassical Ramsey framework, many models have sought
to overcome its many limitations. One class of such models explores the role of international
trade in capturing the observed macroeconomic dynamics. Using the Galerkin method, Cuñat &
Maffezzoli (2004) show how in the Ramsey setting a mere difference in countries’ initial endow-
ments with production factors can lead to a disparity in the steady-state levels of income per capita.
The introduction of international trade leads to a solution in which the convergence of the growth
rates of the poor and the rich regions—typical for the neoclassical growth framework—is not ac-
companied by a convergence in the relative income levels. In a later work, Cuñat & Maffezzoli
(2007) attempt to explain the puzzle of unprecedented growth of global trade by introducing a
trade friction to the model. The solution—obtained with orthogonal collocation—shows how a
slight relaxation in trade barriers can, through specialization, lead to disproportionally large in-
creases in international trade volumes in the long run.
Another contradiction of the neoclassical growth framework becomes apparent under unequal
time preferences of households: The assumed price-taking behavior contradicts the model’s steady
state with a monopolistic household (Becker 1980). The studies that try to achieve more realistic
long-run wealth distributions develop the so-called strategic Ramsey–Cass–Koopmans models.
Pichler & Sorger (2009), for example, introduce the households’ awareness of their own market
power. The least-squares projection method applied to this model finds a steady state in which

1 Section 6.2 presents a discussion of stochastic simulation–based methods.

330 Miftakhova • Schmedders • Schumacher


even less patient households keep a positive share of wealth; this share happens to depend on the
individual intertemporal elasticity of substitution.
Endogenous growth is a fundamental component often omitted from conventional models. To
showcase its relevance, Jones et al. (2005) equip a real business cycle (RBC) model with an endoge-
nous growth mechanism including asymmetric depreciation rates for physical and human capital.
The extended model solved with projection methods provides a better fit to the distributions of
the observed data: It captures the cyclical fluctuations in the labor supply and thereby partially
matches the observed autocorrelation in output growth rate.
Large overlapping generations (OLG) models with aggregate uncertainty are traditionally con-
sidered particularly difficult to solve. Here projection methods are infeasible in their classical form,
as the state space expands with the number of periods that each generation lives. To overcome
the curse of dimensionality, Krueger & Kubler (2004) develop a collocation-based algorithm that
combines the use of complete polynomials with a Smolyak algorithm to generate a sparse grid for
approximation. The former reduces the number of coefficients to be estimated, while the latter
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

keeps the number of points where the model is evaluated moderately (polynomially) growing. The
challenging feature of their model is that the production shocks change the wealth distribution;
therefore, the future wealth distributions are part of the state space of an agent’s problem. The
algorithm can nevertheless handle the model with up to 30 agents.
The method pioneered by Krueger & Kubler (2004) enables the assessment of the welfare
effects of social security systems in stochastic OLG models. Sánchez-Marcos & Sánchez-Martín
(2006) use the method to solve an OLG model with demographic shocks and a pay-as-you-go
pension system; this system appears to have a limited ability of smoothing the welfare effect of
the fluctuations in cohort sizes. The introduction of the pension system to the economy has an
insurance effect of intergenerational risk sharing. The consequent welfare gains, however, do not
outweigh the reduction in per-capita income in comparison to the crowding out of private savings
in the absence of pensions. Gottardi & Kubler (2011) subsequently show that the introduction of a
social security system can in fact be Pareto-improving, given a few characteristics of the economy.
Their model with production shocks features ex-ante welfare evaluation and an infinitely lived
asset (land). These two conditions appear crucial for a pay-as-you-go pension system to prove
beneficial: The former carries the potential of a Pareto-improvement even when the markets are
complete, while the latter can compensate for the welfare loss from decreasing capital stock.
Heterogeneous agent models in general share the challenge of a state space expanded by the
cross-sectional distributions of state variables. In addition to the aforementioned tools, this chal-
lenge can be tackled with hybrid approaches that combine the best features of projection and
perturbation methods. Reiter (2009) uses such a hybrid approach to solve a heterogeneous agents
model with both idiosyncratic and aggregate shocks and liquidity constraints. The projection part
of the algorithm parameterizes the stationary solution of the model with idiosyncratic shocks only,
while the perturbation part consequently linearizes the solution to the model with added aggre-
gate shocks. Reiter (2015) solves a large OLG model with an inverse technique—it starts with a
solution linearized very close to the steady state and gradually builds a global, projection-based
solution as the aggregate shocks grow.

4.3. Winding Paths to Equilibria


One primary objective in macroeconomics is finding economies’ balanced, optimal states that are
consistent with the observed wealth distribution across countries. An equally relevant question
is whether the post-shock paths toward these optimal states—as prescribed by macroeconomic
models—can resemble the stochastic dynamics of real economies. Moving away from steady states

www.annualreviews.org • Computing Economic Equilibria 331


requires accurate numerical methods that can deal with substantial nonlinearities, which makes
conventional first-order approximation unsuitable for this task (Atolia et al. 2010, 2011). In this
regard, not only can projection methods solve these models reliably even far away from equi-
libria, but they also promote analyses of the transition paths toward the equilibria. For exam-
ple, Papageorgiou & Perez-Sebastian (2004) use Chebyshev collocation to solve a research and
development (R&D)-based endogenous growth model and challenge the conventional notion of
country-specific steady states. In contrast, they find that countries’ transition dynamics on the way
to a single, common balanced growth path can to a large extent explain the observed cross-country
income differences. Papageorgiou & Perez-Sebastian (2007) question the very informativeness of
the asymptotic speed of convergence as a proxy for the transitional growth path of an economy.
Based on this proxy, the presence of endogenous human capital accumulation does not enhance
the model, whereas in fact it does improve the model’s ability to capture other components of the
growth path. In the monetary policy context, the same method allows van Zandweghe & Wolman
(2019) to study the transition dynamics of the economy after unexpected policy maker’s commit-
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

ments and discretion shocks—revealing non-negligible differences in the inflation dynamics of


the two settings.

4.4. Facing Zero Lower Bound


The introduction of constraints to DSGE models significantly complicates the solution proce-
dures. In many cases, however, such complications are necessary, as the very object of study is the
optimal policy function under a specific constraint. One example particularly important in prac-
tice is the zero lower bound (ZLB) on the nominal interest rate, which effectively deprives central
banks of the nominal interest rate as a policy instrument. Recent dynamics in some economies
(e.g., Japan, the United States, and the Eurozone) marked by the ZLB or near-ZLB nominal in-
terest rate evoked strong interest in this phenomenon. Incorporation of ZLB explicitly requires
introducing a kink-generating constraint into a DSGE model and focusing on the cases in which
this constraint binds—which is a curse for conventional local methods. Judd et al. (2010) show
that global methods are essential tools to solve such models. Perturbation methods even of higher
order cannot treat the economy near to (let alone at) the ZLB appropriately; their application is
characterized by considerable approximation errors and underestimated length of the ZLB events.
The accuracy of the newly introduced CGA, on the contrary, is high and stays insensitive to ZLB.
Maliar & Maliar (2015) come to a similar conclusion: With high volatility of shocks or binding
ZLB, perturbation methods cannot achieve a satisfactory level of accuracy, while projection meth-
ods combined with a smartly constructed sparse grid provide very accurate solutions in reasonable
computational time.2
Some of the models’ dynamics are simply impossible to detect with linear methods even when
accurate solutions are attainable. Ngo (2014) uses a linear method to calculate the optimal discre-
tionary monetary policy, provided that the distortions to the steady state are minimal. However,
already in this setting linear methods cannot tell how the relative price dispersion affects the nom-
inal interest rate—a dynamics revealed by spline collocation. In work by Fernández-Villaverde
et al. (2015) even more critical features—nonlinearities in consumption, inflation, and, somewhat
surprisingly, in the cumulative effects of shocks—are discovered by a projection method (Smolyak
collocation) but not by linearization or perturbation methods. Once the steady state is disturbed
enough and the ZLB binds, nonlinear methods become the only feasible option.

2 Sections 6.2 and 6.3 provide more details on efficient grid construction.

332 Miftakhova • Schmedders • Schumacher


A notable example of a further application is provided by Aruoba et al. (2018), who use a CGA-
based algorithm to compute a sunspot equilibrium in a ZLB-constrained economy with both a
rich set of fundamental shocks to the economy and an exogenous sunspot shock.

4.5. To Central Banks and Policy Makers


Apart from the ZLB case, projection methods proved extremely useful in general for optimal mon-
etary and fiscal policy problems. For example, Chari et al. (1994) use the Galerkin method to study
the optimal alignment of fiscal policy with the business cycle in a neoclassical growth model. They
find that state-contingent instruments such as return on debt and tax on capital income—and not
labor tax—are the optimal shock absorbers in the economy. Devereux & Siu (2007) trace non-
linear and asymmetric responses of the economy to monetary policy shocks; by the nature of the
problem, they have to resort to nonlinear, global methods. Using orthogonal collocation, Azacis &
Gillman (2010) model the transition of Baltic countries to flat tax rate systems. Song et al. (2012)
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

apply Chebyshev collocation to resolve a trade-off between debt and taxes in an economy with
conflicting views of different generations on the provision of public goods. In this conflict, only a
strong concern with public goods can induce fiscal discipline and avoid the otherwise inevitable
debt accumulation and deterioration of public goods. Niemann et al. (2013) use the collocation
method to analyze the interplay between fiscal and monetary policy when, in contrast to the cen-
tral bank, the fiscal authority is impatient. The central bank’s choice between money growth rate
and nominal interest rate as a policy instrument turns out to determine the equilibrium allocation.

4.6. Deceptive Similarities


Known for their accuracy, projection methods in fact help macroeconomists differentiate between
solutions to seemingly analogous models by uncovering the models’ distinct features overlooked
by coarse methods. New Keynesian analysis emerged in the search of a way to model the ef-
fect of monetary policies on real prices and quantities. This framework requires some form of
nominal rigidity, and it typically implements one of the two forms of price stickiness, through
contracts that temporarily prohibit price adjustment (Calvo 1983) or explicit adjustment costs
(Rotemberg 1982). These two approaches were often deemed interchangeable because, assuming
an efficient zero-inflation steady state, their policy implications are first-order equivalent. A pro-
jection method allowed Leith & Liu (2016) to clearly differentiate between the two setups; they
find significant differences in the economic dynamics—including inflationary bias, consumption,
and labor supply—implied by the two models.
Global methods can fully exploit the nonlinearities obscured by linear approximations and
thereby produce inference that is more consistent with the observed macroeconomic phenomena.
Anderson et al. (2010), for example, find that—in comparison to a nonlinear solution obtained via
Chebyshev collocation—linear-quadratic approximation heavily underestimates the inflation bias
implied by Calvo-type price stickiness.
It is sometimes tempting for modelers to make what might appear to be an innocuous sim-
plifying assumption, such as assuming that a model constraint is always satisfied because it allows
for faster and easier computations. Werner (2016) presents a successful application of projection
methods that shows the danger of such a simplification. In the context of an RBC model, he shows
that the technical treatment of a financial constraint has a significant influence on the economic
predictions of the model; the simplification of enforcing the constraint throughout the state space
leads to substantial distortions in risky steady states, impulse responses, and higher-order mo-
ments. Werner (2016) obtains accurate solutions for both the original and the simplified model
formulation using cubic spline collocation and a multi-grid procedure.

www.annualreviews.org • Computing Economic Equilibria 333


4.7. A Successful Application
As we just explained, the mere application of a suitable solution method can yield new insights or
even change the qualitative implication of a model. Petrosky-Nadeau & Zhang (2017) present a
great study that delivers new insights and represents a huge recent success of projection methods
over local solution methods. This paper applies a projection algorithm to the stylized Diamond–
Mortensen–Pissarides (DMP) model of search and matching in labor markets as presented by
Hagedorn & Manovskii (2008) (see Diamond 1982, Mortensen 1982, Pissarides 1985). The sim-
ple analysis of a projection solution for these types of models yields significant differences in both
qualitative and quantitative implications compared to previous solutions obtained by other meth-
ods. Historically, the economic framework is used to model dynamics in labor markets. It is in-
herent to the model that the tightness of the labor market (described by the ratio of job postings
to unemployed workers) strongly influences the job-finding rate (or matching). This relationship
is highly nonlinear and cannot be accurately captured by local approximations usually deployed
Access provided by Arizona State University on 10/21/21. For personal use only.

in this strand of literature (see Lan 2018 for a review of accuracies for different calibrations of
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

the DMP model). Petrosky-Nadeau et al. (2018) show that—embedding the stylized labor market
into a general equilibrium economy and accounting for nonlinearities introduced by the match-
ing process—a series of negative disturbances can depress the economy into disasters. Particularly,
Petrosky-Nadeau & Zhang (2017) show that, for the correct solution, recent calibrations of the
DMP model as presented by Hagedorn & Manovskii (2008) understate unemployment volatility
by a factor of two and strongly overstate negative correlation of vacancies and unemployment.
In the following, we lay out the key equations of the DMP model and demonstrate how we can
solve them with a projection algorithm (for more details see Petrosky-Nadeau & Zhang 2017).

4.7.1. Solving the Diamond–Mortensen–Pissarides model. The framework assumes an


economy populated by a representative household and a single firm. The firm is owned by the
household and uses labor as its only production input in a production function of the form

Yt = Xt Nt , 27.

where Nt is the level of employment and Xt denotes marginal productivity. Logarithmic produc-
tivity, xt , follows the AR(1) autoregressive process

xt+1 = ρxt + σ εt+1 , 28.

with εt + 1 being a random shock with standard normal distribution, σ being its volatility, and
0 ≤ ρ < 1 being the persistence of log productivity. Family members face the risk of being either
employed or unemployed, for which they grant each other insurance. Employment dynamics in
the economy are described by

Nt+1 = (1 − s)Nt + q(θt )Vt , 29.

with s denoting the job separation rate. The amount of vacancies posted is given by Vt , and the
job-filling rate q(θ t ) is defined as

G(Vt , Ut ) 1
q(θt ) = = 1/ι 30.
Vt 1 + θtι

with the elasticity parameter ι. Here G(Vt , Ut ) is a constant-elasticity-of-substitution function,


which describes the process whereby matches between job openings and unemployed workers are

334 Miftakhova • Schmedders • Schumacher


formed. Labor market tightness θ t is defined as Vt /Ut —i.e., vacancies divided by unemployment—
and fully explains how many vacancies will result in a match.
Due to full risk sharing in the household, we can price all cash flows in the economy with a
single pricing kernel. The household is risk neutral, such that the stochastic discount factor reduces
to the time-preference parameter β. Following Petrosky-Nadeau & Zhang (2017), we derive the
Euler equation determining how many new jobs will be posted by the firm,
!   "
κt κt
− λt = Et β (Xt+1 − Wt+1 ) + (1 − s) − λt+1 , 31.
q (θt ) q (θt+1 )

where κ t is the cost of posting one vacancy. The term on the left-hand side of Equation 31 is the
difference between the marginal cost of one hire and the Lagrange multiplier λt ; the multiplier
denotes the marginal cost of not being able to post negative vacancies—that is, of not actively
firing employees. The accompanying Kuhn–Tucker conditions are
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

q (θt ) Vt ≤ 0, λt ≥ 0, λt q (θt ) Vt = 0.

The right-hand side of Equation 31 depends on two terms. First, the cash-flow term, Xt + 1 −
Wt + 1 , denotes the difference between a worker’s marginal output, Xt + 1 , and their wage, Wt + 1 .
The wage follows from Nash bargaining between the firm and the worker,

Wt = η (Xt + κt θt ) + (1 − η) b,

where b denotes the cash flow in case of unemployment. The second term of the right-hand side
of Equation 31 accounts for the continuation value of having one worker employed, discounted by
the probability of separation. The firm’s inability to actively fire a worker also lowers the continu-
ation value, just as on the left-hand side of the equation. Due to the linear production technology,
the level of employment does not alter the marginal productivity of labor, and so employment
does not show up in the Euler Equation 31. Solving for an equilibrium now requires us to solve
this Euler equation by approximating unknown functions for θ and λ.

4.7.2. The projection solution. To apply projection methods, we view Equation 31 not as a
sequence of conditions in infinite discrete time, t = 0, 1, 2, . . . , but instead as a functional equation.
The exogenous stochastic process xt has as support the real line, R. The two unknown functions
θ, λ : R → R++ must satisfy the functional equation
!      "
κ (x) κ (x ) 

− λ(x) = E β ex − W (x ) + (1 − s) − λ(x 
) x , 32.
q (θ (x)) q (θ (x )) 

with

x = ρx + ε, ε ∼ N (0, σ 2 ).

In the formulation of the functional equation, we use the fact that the variables W, κ, and q are
explicitly given functions of x.
Compared to the introductory example in Section 2, this equation entails a nontrivial complica-
tion. The functional equation has to be solved subject to the nonnegativity constraint on vacancies.
We make sure that our solution satisfies this condition using the Kuhn–Tucker conditions below.

www.annualreviews.org • Computing Economic Equilibria 335


As a result, we must expect that both θ and λ are possibly kinked and continuous but nondifferen-
tiable functions. Namely, we expect θ to be increasing in x and to be zero below a certain level of
productivity, once the marginal value of one worker is so low that the firm prefers to decrease the
number of workers. The opposite should be true for the Lagrange multiplier. When productivity
is high, the constraint is not binding and λ = 0. For sufficiently low productivity, firms would like
to choose θ < 0, and therefore λ is positive. Using a polynomial basis approximating such kinked
functions can be problematic. Petrosky-Nadeau & Zhang (2017) point out that instead of approx-
imating the above variables directly, it is advisable to first approximate the conditional expectation
on the right-hand side of Equation 31 and subsequently infer θ and λ.
Furthermore, as another complication compared to the introductory example, we face uncer-
tainty, captured by the integral over the normal distribution. The conditional expectation of x
given x is an expected value based on the normal distribution of the shock ε. So we can write the
right-hand side of Equations 31 as
Access provided by Arizona State University on 10/21/21. For personal use only.

   
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

∞ ρx+ε κ (ρx + ε) 1 − ε
2
β e − W (ρx + ε) + (1 − s) − λ (ρx + ε) √ e 2σ 2 dε,
−∞ q (θ (ρx + ε)) σ 2π

where we replace the expectations operator from Equation 31 with an integral including the nor-
mal density. We must approximate this integral. For this purpose we apply Gauss–Hermite quadra-
ture (see Judd 1998). This approximation replaces the integral by a finite sum. And so we obtain
an expression that includes values of the unknown functions at a few points in the domain,
#

n  √ √ 
wiGH β eρx+ 2σ εi
− W (ρx + 2σ εi )
i=1 ⎛ √ ⎞⎞ 33.
κ (ρx + 2σ ε )  √ 
+ (1 − s) ⎝    − λ ρx + 2σ εi ⎠⎠ ,
i

q θ ρx + 2σ εi

with the Gauss–Hermite quadrature nodes ε i and weights wiGH . Note that here we replace the
right-hand side in the exact functional Equation 32, involving an integral, by the approximation
in Equation 33, involving a finite sum. This completes Step 1, the identification of the functional
equation. However, before moving to Step 2, we need to take the time to specify how we move
from this approximation to obtaining θ and λ. Denote the term in Equation 33 by E (x). We need
to differentiate between two cases—namely, whether the constraint is binding or not.
Case 1 (not binding). To obtain θ, we begin by assuming that the constraint on vacancies
is not binding. Therefore, λ = 0 and Equation 32 simplifies to

κ (x)
q∗ (θ (x)) = , 34.
E (x)

with the job filling probability (see Equation 30) being weakly between 0 and 1.

Case 2 (binding). Once κ (x)/E (x) > 1, which would mean that the probability of filling
a vacancy exceeds 1, the algorithm ought to set q∗ (θ(x)) = 1 and calculate the Lagrange
multiplier as
κ (x)
λ (x) = − E (x). 35.
1

336 Miftakhova • Schmedders • Schumacher


Note that in both cases θ follows from inverting Equation 30, such that we have θ =
q−1 (q∗ (θ (x))). We now know how a value of E translates into values for θ and λ. This brings us
to the part where we need to approximate E.
In Step 2, we use Chebyshev polynomials (of the first kind) as a basis of V n in Equation 14. So,
the polynomial approximation term Ê (x) is of the form


m  
x−L
Ê (x; a) = a j Tj 2 −1 ,
j=0
U −L

where we need to specify the interval [L, U] of interest for our application. In Step 3, we replace
the function terms E (x) in Equation 33 by the polynomial approximation Ê (x; a) and then obtain
the residual R(x; a) for this model. Observe that here we approximate only one function with a
Access provided by Arizona State University on 10/21/21. For personal use only.

polynomial, namely Ê (x; a). The other two variables, λ̂(x, â) and θ̂ (x, â), follow from the case-by-
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

case analysis and are, once we have approximated E, given in closed form.
For Step 4 we choose the Galerkin projection method with a Chebyshev basis. Therefore, we
need to solve the system of equations

k  
U −L
R (zi + 1) + L; a Tj (zi ) = 0 for j = 0, 1, . . . , n
i=1
2

(see Equation 26 in Section 3.3).


Figure 3a,b shows plots of the Lagrange multiplier λ̂(x, â) and labor market tightness θ̂ (x, â)
for a particular model parameterization (see the caption of the figure for the values of the model
parameters and for the parameters used in the numerical solution). As expected, λ̂(x, â) is positive
and strictly decreasing in the lower range of productivity values. The labor market tightness is close
to being linear for large levels of productivity. The lower two graphs show plots of the marginal
value of the worker, Ê (x, â), and log10 Euler equation errors. Note how the errors are relatively
larger around the kink of the Lagrange multiplier in the region where the curvature of θ̂ (x, â) is
large. Section 6.3 discusses new advances that help increase the accuracy of model solutions; for
example, endogenous-grid methods use spline bases and place more grid points where curvature
is high.

5. MACRO-FINANCE
Numerical methods in the macroeconomics literature are of use to approximate policy functions
describing agents’ optimal reactions to changes in their economic environments. These policy
functions can, for instance, determine the amount of labor deployed, the utilization of production
facilities, or the investments in the capital stock. A projection—that is to say, a global solution—is
well suited when accounting for extreme curvature, changes of curvature, and even kinks intro-
duced by, for example, borrowing or nonnegativity constraints on investments. Contrary to this
focus on the real economy, the finance literature prevailingly focuses on the impact of macroe-
conomic quantities on financial markets. Here, projection methods are used to approximate vari-
ables such as the risk-free rate, the price/dividend ratio, or the conditional expected return of an
asset.

www.annualreviews.org • Computing Economic Equilibria 337


a Lagrange multiplier b Labor market tightness
0.4 2

0.3 1.5

λ̂(x;â) 0.2 θ̂ (x;â) 1

0.1 0.5

0 0
–0.05 0 0.05 –0.05 0 0.05
x x

c Marginal value of worker d log10 Euler equation error


5 –2
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

4
–4

Ê (x;â) 3 |Ê (x;â) – E (x)|


–6
E (x)
2
–8
1

0 –10
–0.05 0 0.05 –0.05 0 0.05
x x
Figure 3
(a) The approximated Lagrange multiplier, λ̂(x; â). (b) Labor market tightness, θ̂ (x; â). (c) The marginal value
of a worker, Ê (x; â). (d) The log10 of the relative Euler equation error, |Ê (x;â)−E
E (x)
(x)|
. All four graphs depict
functions on the domain [−0.0852, 0.0852]. The economic parameters underlying the graphs are the
time-preference parameter, β = 0.991/12 ; the persistence of log productivity, ρ = 0.9895; the volatility
parameter, σ = 0.0034; the workers’ bargaining weight, η = 0.052; the workers’ flow value of unemployment
activities, b = 0.955; the job separation rate, s = 0.0081; and the elasticity of the matching function, ι =
0.407. The graphs show numerical errors using Gauss–Hermite quadrature with 15 nodes and weights and
Chebyshev polynomials of degree 15. The approximations are obtained for the Galerkin method using 25
points. Productivity is represented by x. The polynomial approximation, Ê, of the marginal value of a worker,
E, has as arguments the productivity x and the coefficient vector â.

5.1. Financial Frictions


In macro-finance—the research area at the interface of macroeconomics and finance—the analysis
does not stop at approximating asset prices as a function of the underlying economy, but often the
authors are further interested in possible feedback loops from the financial markets to the real
economy. Here we cannot do justice to this huge literature and only cite a few recent landmark
contributions. An excellent review of this literature together with an assessment of the impact of
different solution methods on model implications is provided by Dou et al. (2020).
Jermann & Quadrini (2012) build an economic model that distinguishes between debt and
equity financing. They use this framework to study the effects of financial shocks on financial vari-
ables and on the real economy. They thus present a mechanism describing how financial shocks
tightened firms’ financing conditions and as a consequence contributed to the Great Recession
of 2008–2009. They solve their model on a four-dimensional grid using linear interpolation be-
tween grid points. They use Gauss–Hermite quadrature to approximate expectations. They show,
however, that a closed-form solution yields results indistinguishable from the projection solution
that they then use to estimate their model.

338 Miftakhova • Schmedders • Schumacher


Gourio (2012) incorporates a model of rare disasters into an otherwise standard RBC economy
to analyze the effects of time-varying disaster probability on real variables such as investment and
on financial variables such as risk spreads. The model is solved on a grid with linear interpolation
between grid points. Gourio (2013) deploys a similar model of disaster risk to link the bond market
with the real economy and explain the level, volatility, and cyclicality of credit spreads as well as
the variation in bond risk premia in a general equilibrium setting. Due to significant nonlinearities
and the need to approximate risk premia, the latter model is solved using Chebyshev polynomial
approximation. Bocola (2016) builds on the model by Gertler & Kiyotaki (2010) and Gertler &
Karadi (2011) to analyze the spillover of sovereign risks to the real economy. Due to a funding
constraint on investment, the model is solved globally. The author takes the model to the data
by estimating it with a particle filter. The solution to the model is obtained with sparse colloca-
tion using the Smolyak algorithm. Schumacher & Żochowski (2017) use a similar constraint on
leverage and combine it with an endogenous growth mechanism as presented by Kung & Schmid
(2015). They present an increase in risk premia driven by the leverage constraint as a potential
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

explanation for the extended period of low growth after the financial crisis of 2007. Again, due to
the constraint on leverage, the model is solved using a projection algorithm.
Another class of market frictions that has received much attention after the 2007 financial cri-
sis is borrowing constraints resulting from collateral requirements. In many circumstances, bor-
rowers must put up substantial amounts of collateral assets to receive loans from lenders. Brumm
et al. (2015a) investigate such collateral constraints and their impact on asset returns in an infinite-
horizon general equilibrium model with heterogeneous agents. They show that borrowing against
collateral substantially increases the return volatility of long-lived assets such as stocks. To solve
their model, they apply the projection method of Brumm & Grill (2014), which is explicitly de-
signed for occasionally binding constraints and uses piecewise linear functions as bases. Brumm
et al. (2015b) use the same approach to explain why the active management of margin require-
ments on US stock markets from 1947 until 1974 had little if any measurable effect on market
volatility.

5.2. Consumption-Based Asset Pricing


One of the major challenges in finance is to reconcile the quantities observed in financial markets
with agents’ preferences and observed macroeconomic variables such as consumption growth,
investment, and GDP. A model that is among the first to align dynamics in consumption growth
with a high equity risk premium and a low risk-free rate is proposed by Campbell & Cochrane
(1999). They propose an economy in which agents form their consumption habit level relative to
others. They are exposed to time-varying consumption growth. When consumption growth is low,
marginal utility is high, such that the equity risk premium is countercyclical and predictable by the
price/dividend ratio. As discussed by Chen et al. (2008), the above results are strongly driven by
events with consumption growth below −25%. Not allowing for those events makes it impossible
to replicate financial moments as observed in the data in this model.

5.3. A Projection Solution of an Asset-Pricing Model


Before we discuss some exciting asset-pricing applications of projection methods in Section 5.4,
we show how to apply such methods in a consumption-based asset-pricing model by example of
an early asset-pricing model in partial equilibrium. Due to space limitations, the presentation of
the application here must be very brief. We refer interested readers to the textbooks by Cochrane
(2001) and Campbell (2018) for in-depth treatments of consumption-based asset-pricing models.

www.annualreviews.org • Computing Economic Equilibria 339


5.3.1. Solving an asset-pricing model. Burnside (1998) considers a standard version of the
consumption-based asset-pricing model with a representative agent having power utility. The
agent’s first-order condition yields the basic pricing equation for this model. For an asset with
unit price Pt and dividend Dt in period t, the agent’s Euler equation states

   
Ct+1 −γ
Pt = Et β (Pt+1 + Dt+1 ) for t = 0, 1, 2, . . . .
Ct

Here, β ࢠ (0, 1) denotes the agent’s discount factor and γ > 0 denotes the agent’s coefficient of
relative risk aversion. Under the assumption that dividends are the only source of income, Ct =
Dt , we can rewrite the Euler equation as
   $
Pt Ct+1 1−γ Pt+1
= Et β +1 .
Access provided by Arizona State University on 10/21/21. For personal use only.

Ct Ct Ct+1
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Following Burnside (1998), we define the price/consumption ratio v t = Pt /Ct and obtain
  $
Ct+1 1−γ
vt = Et β (vt+1 + 1) . 36.
Ct

The evolution of the stochastic consumption


 process,
 Ct , is exogenously given. Specifically, loga-
Ct+1
rithmic consumption growth ct+1 = log Ct
is an AR(1) process,

ct+1 ≡ xt+1 = (1 − ρ )μ + ρxt + t+1 , t+1 ∼ N (0, σ 2 ),

with the persistence parameter ρ ࢠ (−1, 1). Since the exogenous process provides log consump-
tion growth xt instead of actual consumption or consumption growth, it is helpful to rewrite the
equation for the price/consumption ratio as
  $
1−γ
vt = Et β ext+1 (vt+1 + 1) . 37.

Burnside (1998) derives a necessary and sufficient condition that a finite solution v t of Equation
37 exists given a value for the exogenous process xt . The price/consumption ratio Burnside (1998,
equation 19) is



vt = β i eai +bi (xt −μ) , 38.
i=1

with the parameter sequences ai and bi given by

! "
1 σ2 ρ 21−ρ
2i
ai = (1 − γ )iμ + (1 − γ )2 i − 2 (1 − ρ i
) + ρ
2 (1 − ρ )2 1−ρ 1 − ρ2

and

ρ
bi = (1 − γ ) (1 − ρ i ),
1−ρ

340 Miftakhova • Schmedders • Schumacher


respectively. The sum in Equation 38 converges if and only if

(1−γ )μ+ 12 (1−γ )2 σ2


βe (1−ρ )2 <1

(see Burnside 1998, theorem 1, condition 20).

5.3.2. Projection solutions. To apply projection methods, we view Equation 37 not as a se-
quence of conditions in infinite discrete time, t = 0, 1, 2, . . . , but instead as a functional equa-
tion. The exogenous stochastic process xt has as support the real line, R. The unknown function
v : R → R++ must satisfy the functional equation
    
 1−γ 
v(x) = E β ex (v(x ) + 1) x 39.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

with

x = (1 − ρ )μ + ρx + ,  ∼ N (0, σ 2 ).

The conditional expectation of x given x is just an expected value based on the normal distribution
of the shock . So, we can simply write the functional equation as
 ∞ 1−γ   1 2
− 
v(x) = β e(1−ρ )μ+ρx+ v((1 − ρ )μ + ρx + ) + 1 √ e 2σ 2 d. 40.
−∞ σ 2π
To obtain a numerical approximation of the integral on the right-hand side, we apply Gauss–
Hermite quadrature (see Judd 1998). This approximation replaces the integral by a finite sum.
And so we obtain an expression that includes values of the unknown function v at a few points in
the domain, namely

1  GH  (1−ρ )μ+ρx+√2σ i 1−γ   √  


n
v(x) − β √ wi e v (1 − ρ )μ + ρx + 2σ i + 1 = 0, 41.
π i=1

with the Gauss–Hermite quadrature nodes  i and weights wiGH . Note that here we replace the
exact functional Equation 40, involving an integral, by an approximation in Equation 41, involving
a finite sum. This completes Step 1, the identification of the functional equation.
In the following we explain how we can approximate the unknown price-consumption func-
tion v by a sum of Chebyshev polynomials. We employ our two favorite projection methods,
Chebyshev collocation and Galerkin projection with Chebyshev polynomials, which we described
in Section 3.3.
In Step 2, we use Chebyshev polynomials (of the first kind) as a basis of V n . So, the polynomial
approximation term v̂(x) is of the form


m  
x−L
v̂(x; a) = a j Tj 2 −1 ,
j=0
U −L

where we need to specify the interval [L, U] of interest for our application. In Step 3, we replace
the function terms v(x) in Equation 41 by the polynomial approximation v̂(x; a). The resulting
left-hand side is the residual R(x; a) for this model.

www.annualreviews.org • Computing Economic Equilibria 341


a Relative errors b Residuals

–5 × 10–6 6 × 10–7
–0.000010 4 × 10–7
–0.000015 2 × 10–7
v̂ (x;â) – v*(x)
–0.000020 R(x;â) 0
v*(x)
–0.000025 –2 × 10–7
–0.000030 –4 × 10–7
0 –6 × 10–7
–0.05 0 0.05 0.10 –0.05 0 0.05 0.10
x x
Figure 4

(a) A plot of the relative errors, v̂(x;â)−v
v ∗ (x)
(x)
, in the approximated solution function on the domain [−0.06, 0.1]. (b) The residuals, R(x; â).
The economic parameters underlying the two graphs are β = 0.95, γ = 5, ρ = 0.8, μ = 0.02, and σ = 0.0205. The graphs show
numerical errors using Gauss–Hermite quadrature with 15 nodes and weights and Chebyshev polynomials of degrees 0 to 8. The state
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

variable, log consumption growth, is denoted by x. The polynomial approximation, v̂, of the price/consumption ratio, v ∗ , has as
arguments the variable x and the coefficient vector â.

In Step 4 we choose Chebyshev collocation and use the m + 1 zeros of the Chebyshev polyno-
mial of degree m + 1 as collocation nodes. We obtain a linear square system of m + 1 equations—
one residual equation R(xj ; a) = 0 for each grid point xj ࢠ G, j = 0, 1, . . . , m. The collocation
points are the zeros of the (m + 1)-degree Chebyshev polynomial transformed to the interval
[L, U] = [−0.06, 0.1] for the state variable x. We sum the first 1,000 elements of the infinite sum
in Equation 38 to obtain (an excellent approximation of ) the exact solution to the model.
Figure 4a shows a plot of the relative errors in Equation 9. Recall that without knowing the
exact solution we would not be able to determine these errors. Instead, all we can typically do is
to compute the residuals R(x; â) (see Figure 4b for a plot of these residuals on [−0.06, 0.1]). We
observe that the absolute relative errors are smaller than 4 × 10−5 , while the absolute residuals
are smaller than 7 × 10−7 . The residuals closely resemble the equi-oscillation property, which we
expect from a good approximation via Chebyshev polynomials.

5.4. Recursive Utility and Priced State Variables


Another class of models aiming at explaining moments in financial markets taking dynamics in
consumption growth as given is the class of long-run risk models. These models disentangle the
agents’ intertemporal elasticities of substitution from their risk aversion by means of recursive
preferences, as discussed by Weil (1989) and Epstein & Zin (1991). Bansal & Yaron (2004) pro-
vide an equilibrium explanation for several asset-pricing quantities. They model consumption and
dividend growth dynamics to depend on two latent state variables, a predictable long-run growth
rate and a process driving conditional volatilities in the economy. When pairing this framework
with recursive preferences, agents care about these latent state variables. Beyond generating a low
risk-free rate together with a large and volatile return on the dividend claim, fluctuating economic
uncertainty makes the equity premium countercyclical and predictable. The authors use a linear
return approximation as deployed by Campbell & Shiller (1988) to obtain closed-form solutions
for valuation ratios. They test the accuracy of their model solution with standard polynomial-
based projection methods. However, Pohl et al. (2018) show that the generation of asset-pricing
models following the seminal paper by Bansal & Yaron (2004) exhibit economically significant
nonlinearities. These make a log-linear model solution unsuitable as it leads to a large numeri-
cal approximation error. The authors point out that these nonlinearities are largely driven by an

342 Miftakhova • Schmedders • Schumacher


increased persistence of state variables and by the fact that these processes spend more time far
away from the point of expansion of the linear solution. These extreme values have a significant
impact on asset-price dynamics. More recently, the literature has incorporated even more intricate
dynamics to consumption growth to explain a larger host of quantitative facts characterizing asset
markets, such as the implied volatility curve or surface. For instance, Benzoni et al. (2011) explain
the steepening of the volatility curve after the 1987 stock market crash by adding rare jumps to
consumption growth and economic uncertainty. Even though they use a log-linear return approxi-
mation, the model solution is not available in closed form anymore, so they rely on approximations
with Chebyshev polynomials of order 20 via least-squares minimization. Dergunov et al. (2019)
use a similar model setup and introduce inflation dynamics to their economy to price nominal
assets. Again—using the same solution method—they first use a log-linear return approximation
and in a second step solve their model numerically with a fourth-order Chebyshev approximation.
The long-run risk asset-pricing literature has put forth numerous models with jumps in economic
fundamentals to reproduce the properties and the time-variation of the variance risk premium.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Lorenz et al. (2020) compare the results of the workhorse model of this literature presented
by Drechsler & Yaron (2010) with an accurate solution obtained by a projection algorithm using
complete polynomials. They show that the additions to state variable dynamics severely worsen
the inaccuracies introduced by log-linearization of returns, up to the point that their qualitative
and quantitative results are largely driven by numerical errors.
Another possible extension to long-run risk models is to modify the agent’s preferences beyond
the recursive time aggregation of utility. Gallant et al. (2019) use a Bayesian method to estimate
consumption-based asset-pricing models featuring smooth-ambiguity preferences. All their esti-
mations rely on a model solution obtained with collocation projection using Chebyshev polyno-
mials. They show in a model comparison that models with ambiguity, learning, and time-varying
volatility are preferred to the standard long-run risk model. Lorenz & Schumacher (2018) study
the role of generalized disappointment aversion (cf. Routledge & Zin 2010) in explaining asset-
pricing quantities, using insights from behavioral economics. The type of preferences introduces
large nonlinearities, which the authors account for by using a Galerkin projection approach to
solve the model.
While all of the above build on a single-agent economy, Pohl et al. (2020) add a second agent
to the model to establish trade. The agents disagree on the persistence of the long-run growth
rate of consumption. Adding a second agent introduces an endogenous state variable—the wealth
share—complicating the solution procedure. The wealth share evolves over time and influences
the optimal consumption decision of agents. The value functions of the two agents are a function
of both exogenous and endogenous state variables. The value functions are approximated with a
two-dimension cubic spline. On the boundaries, with the wealth share being 0 and 1, the solution
is replaced by the solution of an otherwise equal representative-agent economy. Pohl et al. (2020)
use a discrete-time framework and a full nonlinear solution. Branger et al. (2016) use a continuous-
time setup, featuring disagreement on jump intensities and a solution method inspired by Benzoni
et al. (2011) and outlined above. The final solution is obtained with a partial differential equation
(PDE) solver from the Numerical Algorithms Group (NAG) (http://nag.com). Again, for the lim-
iting case and a boundary condition of the PDE, the authors use a representative-agent economy
solution.

5.5. Asset Pricing with Production and Investment


As already pointed out by Cochrane (1991), production-based asset pricing is analogous to the
standard consumption-based asset-pricing model. While consumption-based asset pricing ties as-
set returns to the marginal rate of substitution of agents, production-based asset pricing ties asset
www.annualreviews.org • Computing Economic Equilibria 343
returns to the return on investment—that is, the marginal rate of transformation. Asset pricing
with production and investment then tries to explain both in one unified model. Solving such mod-
els poses the challenge of devising numerical solution methods that can find good approximate
solutions when there are smooth frictions in investment and high curvatures in agents’ utilities.
Jermann (1998) studies the implications of habit-formation preferences and capital adjust-
ment costs on asset prices in an otherwise standard RBC model. The model aligns the equity risk
premium and the return of the risk-free asset with the business cycle. He uses log-normal solu-
tion methods if applicable; when no closed-form solution is available, the model is solved with
a Galerkin projection algorithm. Building on the insight that a persistent component on con-
sumption growth combined with agents of Epstein–Zin type increases the price of risk in the
economy, Kaltenbrunner & Lochstoer (2010) study how this pattern can arise endogenously in
a standard production economy model. Their economy has two state variables, the capital stock
and detrended labor productivity. The authors approximate the value function of the representa-
tive agent with a tenth-order Chebyshev polynomial. Aldrich & Kung (2017) point out that for
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

DSGE models with asset-pricing implications, higher-order perturbation methods are ill suited.
They conduct an analysis for economies with transitory and permanent shocks to total factor pro-
ductivity. With calibrations featuring high volatility and risk aversion, asset-pricing quantities tend
to not be accurately matched by the perturbation method. They explain this by the fact that a local
method does very poorly at capturing the curvature of the value function on the whole domain of
the state variables.
Gräber & Schumacher (2019) extend this analysis to the effects of persistent growth risks
on the solution accuracy of DSGE models as introduced by Croce (2014).3 They show that the
presence of persistent growth risks severely worsens the performance of a perturbation-based so-
lution. In a high-volatility calibration, the risk-free rate is understated by more than 100% and the
wealth/consumption ratio is approximately 65% below the true value. Also, macroeconomic values
such as the volatility of investment growth are wrongly approximated by almost 10%. They fur-
ther show that even a second-order collocation projection performs significantly better than any
solution obtained by perturbation. Boldrin et al. (2001) add habit formation and a two-sector setup
to a standard RBC model to improve asset-pricing performance. The model’s equilibrium condi-
tions are solved utilizing a projection algorithm. The authors display impulse response functions
to analyze differences in reactions to exogenous disturbances in both sectors. Petrosky-Nadeau
et al. (2013) use early insights of Petrosky-Nadeau et al. (2018) and study the asset-pricing impli-
cations of matching frictions in labor markets. The solution method to this framework is discussed
in detail in Sections 4.7.1 and 4.7.2. Their economy featuring a DMP-matching process generates
a sizable equity risk premium together with high stock market volatility while also producing a
low risk-free rate. Excess returns are countercyclical and predictable.

5.6. Asset Pricing with Heterogeneity


Favilukis & Lin (2015) show the effects of wage stickiness on equity valuations in a heterogeneous
agents framework in the fashion of the models discussed in Section 4.1. A very impressive
example of heterogeneous agents models in asset pricing is offered by Tuzel (2010). She studies
the impact of capital adjustment cost on the cross-section of asset returns and finds—in line
3 Ai et al. (2012, 2018) and Colacito et al. (2018) use a similar setup of long-run risks in productivity growth to

explain joint phenomena of the real economy and asset prices. They all use perturbation-based solutions for
their analyses. Favilukis & Lin (2013) also use a setup featuring shocks to the growth rate of productivity but
solve their model with a value function algorithm. They use this framework to explain several salient properties
of investment in a production economy.

344 Miftakhova • Schmedders • Schumacher


with empirical data—that firms with a rigid capital stock earn a higher risk premium compared
to firms with investments in assets that can be adjusted more easily. The model is solved for
a large number of firms, which requires her to keep track of the distribution of capital stocks.
She uses the parameterized expectations algorithm—a simulation-based solution method—to
reduce the computational burden of obtaining a model solution. İmrohoroğlu & Tüzel (2014)
use a similar solution algorithm to analyze the impact of firm-level total factor productivity
on asset returns. Gomes & Michaelides (2007) use a heterogeneous agents framework to study
asset-pricing implication together with stock market participation. The agents in their framework
are borrowing-constrained and risk sharing is incomplete. Their model shows how an agent’s
stock market participation arises endogenously. However, they find it hard to reconcile agents’
portfolio decisions with a set of asset-pricing implications.

6. OTHER APPROACHES AND NEW ENDEAVORS


Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Macroeconomics and finance are certainly the two fields in economic modeling that have seen
the most applications of projection methods. However, it would be an omission not to mention
at least briefly a few applications in microeconomics and climate-change modeling, which we do
next. Moreover, we survey other numerical methods that involve projection techniques. Finally,
we briefly describe recent progress on solving models with large-dimensional state spaces and on
taking advantage of parallel computing on high-performance computing architectures.

6.1. Omni-Applicable: From Microeconomics to Climate Economics


Advanced dynamic microeconomic models have also witnessed some prominent applications of
projection methods. For example, Doraszelski (2003) challenges the memoryless property of con-
ventional R&D models and, instead, allows knowledge to accumulate and to affect firms’ strate-
gies. The solution to this model obtained via Chebyshev collocation implies a behavior consistent
with observations: The firm that lags behind acts to catch up with the leader. In general, given a
large-enough stock of knowledge, the follower eventually invests more than the leader.
The optimal stopping problem analyzed by Doraszelski (2004) aims to explain the observed
phenomenon of delayed implementation of newly discovered technologies. The model, which dis-
tinguishes between innovations and improvements that follow the innovations, is characterized by
high nonlinearity, also addressed with Chebyshev collocation. The ultimate feature of the solution
is that firms might indeed choose to wait until a new technology is improved sufficiently and only
then adopt it.
Projection methods also proved powerful in solving asymmetric auctions. Bajari (2001) con-
trasts competitive and collusive bidding in an asymmetric procurement auction. The algorithm
he uses to infer the inverse bid functions is essentially collocation based. The inference based on
this algorithm suggests that a firm bids more aggressively the more numerous and more efficient
its competitors are. Hubbard & Paarsch (2009) model procurement auctions with so-called bid-
preference programs: In such auctions, at the evaluation stage the bids of some preferred firms are
scaled down. This preference-induced asymmetry creates unusual bidding behavior. To solve the
model, they modify the method proposed by Bajari (2001). Even though firms’ entry decisions are
not strongly affected by the bid-preference policy, the solution reveals two other notable effects,
with the preferred firms inflating their bids and the nonpreferred firms bidding more aggressively.
A relatively young branch of economics, climate-change economics, has very quickly grown be-
yond analytically solvable formulations. In climate-economic models, economic activity induces
an adverse effect on the climate system, which in turn damages the economy. Powerful compu-
tational methods are especially helpful here, as challenging macroeconomic models are merged

www.annualreviews.org • Computing Economic Equilibria 345


with models that emulate the climate system. The first applications of projection-based meth-
ods in this literature are to models in which decision makers face vast uncertainties. In one such
model, Lontzek & Narita (2011) study how the degree of risk aversion affects the optimal climate
policy. Given only minor departures from the steady state, the perturbation-based solution for
their model shows that higher risk aversion implies more stringent mitigation; projection meth-
ods (namely, Chebyshev collocation), in contrast, can track the dynamics further away from the
steady state.
In the model of Lemoine & Traeger (2016), the changes to the climate induce both gradual
damage and irreversible climate events (so-called tipping points) that permanently shift climate
sensitivity, the persistence of CO2 concentrations, or the impact on the economy. They use col-
location methods to solve for the optimal climate policy that would take into account both the
possible occurrence and the interaction of these tipping points. These irreversibilities increase
the optimal carbon tax in the model almost twofold and significantly raise the cost of delaying
climate action.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Bansal et al. (2016) further associate a climate-economic model and the long-run risk frame-
work of finance. In their joint model, solved by a projection-based algorithm, the damage induced
by rising temperatures to economic growth transfers into a positive risk premium on equity mar-
kets. The information from these markets is thereby used to estimate the costs of optimal climate
policy; in particular, it leads to a high optimal carbon tax.
The dynamic stochastic integrated model of climate and economy (DSICE) model used by Cai
et al. (2017) is an impressive example of both advanced computational methods and high-power
computing applied to a climate-economic problem. The model has nine state variables, is enriched
by both climate and economic uncertainty, and spans centuries with annual time steps. Solving a
problem of this size is only feasible on a supercomputer, yet the algorithm builds on projection
techniques in that it approximates the value function with multivariate orthogonal polynomials
and solves the Bellman equation with value function iteration. The ultimate finding is that the
climate and economic risks might well raise the optimal carbon tax tenfold.

6.2. Hybrid Approaches


As ever-developing models include more and more of the nontrivial features of the real world,
the solution methods evolve along. We witness projection approaches being modified and inter-
laced with other approaches; modern implementations go well beyond the classical projection
paradigm. The parameterized expectation approach (PEA), for example, is tailored to solve non-
linear stochastic dynamic models. The key idea of the method is to approximate the expecta-
tion function in the Euler equation with some basis functions by fitting simulated time series.
To find the coefficients of such an approximation, early works suggest nonlinear regression (see,
for example, Den Haan & Marcet 1990). Compared to linearization and log-linearization, PEA
performs much better on the accuracy test proposed by Den Haan & Marcet (1994) and ap-
plied to a cash-in-advance model. Once PEA is viewed as a projection-type method, it appears
natural that Christiano & Fisher (2000) consider—along with the conventional PEA procedure—
Galerkin PEA and collocation PEA. The advantageous features of such blends include the use of
Chebyshev polynomials as basis functions, of Chebyshev zeros as a base for the grid, and of linear
(instead of nonlinear) regression to find the coefficients. When applied to a one-sector optimal
growth model with productivity shocks and constrained gross investments, these hybrid algo-
rithms achieve higher accuracy while at the same time drastically reducing computational burden.
Because the conventional application of the PEA does not guarantee global convergence, Maliar
& Maliar (2003) suggest to impose moving bounds on a model’s endogenous state variables; these

346 Miftakhova • Schmedders • Schumacher


bounds—tight initially and wider as the algorithm converges—can prevent these variables from
exploding or imploding.
Ergodic-set algorithms operate on a subdomain comprised by all points that are likely to be
visited in a model’s equilibrium. In combination with projection methods, such algorithms nar-
row the focus to only a small fraction of the original hypercube grid—maximizing the accuracy of
approximation over this relevant subdomain. An extensive analysis of the combinations of approx-
imation and integration techniques by Judd et al. (2011) yields the generalized stochastic simula-
tions algorithm (GSSA). GSSA indeed uses stochastic simulations to restrict the solution domain
to the endogenous ergodic set of a model, but the solutions are subsequently computed by de-
terministic integration methods. The former renders the method applicable to high-dimensional
models, while the latter, in combination with robust approximation methods, ensures the numer-
ical stability.
Finally, Maliar & Maliar (2015) focus on the construction of an efficient grid for projection
from the simulated points. The first of the three proposed algorithms clusters the simulated points
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

and selects the points that represent these clusters. The accuracy of this clustering method corre-
sponds to the density of the points, with better fit in high-probability areas. The second method
selects only a subset of points based on their distance from each other. This distance can be further
flexibly adjusted to provide a more dense grid in the areas that are harder to fit.

6.3. Too Large to Handle?


Most of the recent developments for projection methods aim at relieving the computational bur-
den associated with their implementation; specifically, studies continue to search for better ways
to deal with the curse of dimensionality. Heer & Maussner (2018) compare the ability of three dif-
ferent projection methods—Galerkin, collocation, and least squares—to handle high-dimensional
state space problems. Monomial formulas improve this ability substantially for least-squares and
Galerkin projection methods; the latter is shown to be especially accurate and fast. Levintal (2018)
proposes a hybrid of projection and Taylor approximations—Taylor projection—which is less sen-
sitive to the curse of dimensionality. The method is bound to a point of a problem’s state space
and sets both the residual function and its derivatives to zero at this point—to obtain an approx-
imation for the solution there. In contrast to perturbation, Taylor projection is applicable to any
point in the state space and, under certain regularity conditions, converges to the true solution in
the neighborhood of this point.
An inefficient grid is a potential bottleneck of any (even otherwise optimal) solution algorithm.
The literature has long since moved away from classical tensor-product grids and toward more
economical alternatives, which can overcome the curse of dimensionality (Bungartz & Griebel
2004). In contrast to predetermined sparse grids, flexible grids determine the grid points at every
iteration and can thereby take advantage of the properties of the very problem in question. For
example, the endogenous grid introduced by Carroll (2006) sets specific grid points at any period
t derived from the state variables of the already computed period t + 1. The efficiency of the
procedure lies in the absence of a need to compute expectations at the unused grid points.
For some problems, however, iterative grid adaptation might complicate the calculations, e.g.,
by introducing discontinuities. Reich & Wilms (2015) adapt interpolation grids continuously,
without any disruption to a problem’s structure. The grid nodes move to the optimal locations
dictated by the optimality conditions of the problem itself. The applications show that the algo-
rithm ultimately requires fewer nodes to attain a given level of accuracy, compared to a uniform
grid and even to Chebyshev nodes.
Brumm & Scheidegger (2017) offer a remarkable application of an efficient adaptive sparse grid
algorithm to supersized economic problems. The chosen adaptive grid algorithm refines the grid
www.annualreviews.org • Computing Economic Equilibria 347
only in the regions where approximation is especially difficult (e.g., regions with nondifferentiabil-
ities or high curvature). Implemented with the use of high-performance computing technology,
the algorithm solves an RBC model with capital-adjustment costs and irreversible investments
with up to 20 dimensions (and up to 100 dimensions in the absence of irreversibility).
Finally, sampling-based, grid-free methods present a new opportunity to break free from the
curse of dimensionality. Machine learning methods, such as Gaussian process regression (GPR),
can be used to approximate functions with irregular local features. The algorithm learns from the
sampled points (also called design points) and therefore avoids explicit grid construction. In work
by Renner & Scheidegger (2018) and Scheidegger & Bilionis (2019), GPR combined with high
parallelization successfully solves large-scale models with irregularities. Duffy & McNelis (2001)
suggest a neural network specification for approximating the expectation function in the PEA
procedure. Applied to a stochastic growth model, the algorithm performs with accuracy similar to
that of polynomial approximations. Azinović et al. (2019) use deep neural networks as an approx-
imation for the equilibrium functions of a model. The fact that the algorithm learns the shape of
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

a model’s ergodic set from simulations makes it grid-free and flexible enough to handle irregular
shapes. The power of the method is demonstrated on a large OLG model à la Krueger & Kubler
(2004)—expanded further with borrowing constraints in addition to aggregate stochastic shocks
and agents living up to 60 periods. The structure overall leaves the model with around 120 dimen-
sions, successfully processed by the neural networks. Seemingly unorthodox, the algorithm can be
generally classified as a projection method in which neural networks approximate the equilibrium
conditions and the loss function essentially minimizes mean squared errors in these conditions.

7. CONCLUSION
In this review we have discussed the impact of the use of projection methods on different strands
of economic research. We demonstrated that these methods are an excellent choice of numer-
ical techniques to address the nonlinearities arising from constraints, other potentially smooth
frictions, or extreme dynamics of state variables. Not only can these methods deliver accurate ap-
proximations of nonlinear solution functions, but—because they incorporate information from
several points in the domain—they can also do so on large domains. In recent years, projection
methods have delivered new insights into prominent models in economics and finance, which less
accurate solutions, obtained in older studies using simpler techniques, had missed.
As dynamic economic models continue to evolve, so must the numerical methods to solve them.
In light of the recent advances in projection methods, such as new adaptive-grid methods and
the implementation of such methods on high-performance computer architectures, we strongly
believe that projection methods will be the methods of choice for many years to come.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings that
might be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS
We are grateful to an anonymous referee, Ken Judd, and Harry Paarsch for detailed comments on
an earlier draft. We thank Thomas Cosimano, Uli Doraszelski, Gregor Reich, Simon Scheidegger,
and Ole Wilms for helpful suggestions.

348 Miftakhova • Schmedders • Schumacher


LITERATURE CITED
Ai H, Croce MM, Diercks AM, Li K. 2018. News shocks and the production-based term structure of equity
returns. Rev. Financ. Stud. 31:2423–67
Ai H, Croce MM, Li K. 2012. Toward a quantitative general equilibrium asset pricing model with intangible
capital. Rev. Financ. Stud. 26:491–530
Aldrich EM, Kung H. 2017. Computational methods for production-based asset pricing models with recursive utility.
Work. Pap., Duke Univ., Durham, NC
Algan Y, Allais O, Den Haan WJ. 2010. Solving the incomplete markets model with aggregate uncertainty
using parameterized cross-sectional distributions. J. Econ. Dyn. Control 34:59–68
Anderson GS, Kim J, Yun T. 2010. Using a projection method to analyze inflation bias in a micro-founded
model. J. Econ. Dyn. Control 34:1572–81
Aruoba SB, Cuba-Borda P, Schorfheide F. 2018. Macroeconomic dynamics near the ZLB: a tale of two coun-
tries. Rev. Econ. Stud. 85:87–118
Aruoba SB, Fernández-Villaverde J, Rubio-Ramírez JF. 2006. Comparing solution methods for dynamic equi-
Access provided by Arizona State University on 10/21/21. For personal use only.

librium economies. J. Econ. Dyn. Control 30:2477–508


Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Atolia M, Awad B, Marquis M. 2011. Linearization and higher-order approximations: How good are they?
Comput. Econ. 38:1–31
Atolia M, Chatterjee S, Turnovsky SJ. 2010. How misleading is linearization? Evaluating the dynamics of the
neoclassical growth model. J. Econ. Dyn. Control 34:1550–71
Azacis H, Gillman M. 2010. Flat tax reform: the Baltics 2000–2007. J. Macroecon. 32:692–708
Azinović M, Gaegauf L, Scheidegger S. 2019. Deep equilibrium nets. Work. Pap., Univ. Zurich, Zurich, Switz.
Bajari P. 2001. Comparing competition and collusion: a numerical approach. Econ. Theory 18:187–205
Bansal R, Kiku D, Ochoa M. 2016. Price of long-run temperature shifts in capital markets. NBER Work. Pap.
22529
Bansal R, Yaron A. 2004. Risks for the long run: a potential resolution of asset pricing puzzles. J. Finance
59:1481–509
Becker RA. 1980. On the long-run steady state in a simple dynamic model of equilibrium with heterogeneous
households. Q. J. Econ. 95:375–82
Benzoni L, Collin-Dufresne P, Goldstein RS. 2011. Explaining asset pricing puzzles associated with the 1987
market crash. J. Financ. Econ. 101:552–73
Bocola L. 2016. The pass-through of sovereign risk. J. Political Econ. 124:879–926
Boldrin M, Christiano LJ, Fisher JD. 2001. Habit persistence, asset returns, and the business cycle. Am. Econ.
Rev. 91:149–66
Branger N, Konermann P, Schlag C. 2016. Optimists, pessimists, and the stock market: the role of preferences and
market (in)completeness. Work. Pap., Münster Univ., Münster, Ger.
Brumm J, Grill M. 2014. Computing equilibria in dynamic models with occasionally binding constraints.
J. Econ. Dyn. Control 38:142–60
Brumm J, Grill M, Kubler F, Schmedders K. 2015a. Collateral requirements and asset prices. Int. Econ. Rev.
56:1–25
Brumm J, Grill M, Kubler F, Schmedders K. 2015b. Margin regulation and volatility. J. Monet. Econ. 75:54–68
Brumm J, Scheidegger S. 2017. Using adaptive sparse grids to solve high-dimensional dynamic models.
Econometrica 85:1575–612
Bungartz HJ, Griebel M. 2004. Sparse grids. Acta Numer. 13:147–269
Burnside C. 1998. Solving asset pricing models with Gaussian shocks. J. Econ. Dyn. Control 22:329–40
Cai Y, Judd KL, Lontzek TS. 2017. The social cost of carbon with economic and climate risks. Work. Pap. 18113,
Hoover Inst. Econ., Stanford Univ., Stanford, CA
Caldara D, Fernández-Villaverde J, Rubio-Ramírez JF, Yao W. 2012. Computing DSGE models with recursive
preferences and stochastic volatility. Rev. Econ. Dyn. 15:188–206
Calvo GA. 1983. Staggered prices in a utility-maximizing framework. J. Monet. Econ. 12:383–98
Campbell JY. 2018. Financial Decisions and Markets: A Course in Asset Pricing. Princeton, NJ: Princeton Univ.
Press

www.annualreviews.org • Computing Economic Equilibria 349


Campbell JY, Cochrane JH. 1999. By force of habit: a consumption-based explanation of aggregate stock
market behavior. J. Political Econ. 107:205–51
Campbell JY, Shiller RJ. 1988. The dividend-price ratio and expectations of future dividends and discount
factors. Rev. Financ. Stud. 1:195–228
Carroll CD. 2006. The method of endogenous gridpoints for solving dynamic stochastic optimization prob-
lems. Econ. Lett. 91:312–20
Chari VV, Christiano LJ, Kehoe PJ. 1994. Optimal fiscal policy in a business cycle model. J. Political Econ.
102:617–52
Chen Y, Cosimano TF, Himonas AA. 2008. Analytic solving of asset pricing models: the by force of habit case.
J. Econ. Dyn. Control 32:3631–60
Christiano LJ, Fisher JD. 2000. Algorithms for solving dynamic models with occasionally binding constraints.
J. Econ. Dyn. Control 24:1179–232
Cochrane JH. 1991. Production-based asset pricing and the link between stock returns and economic fluctu-
ations. J. Finance 46:209–37
Cochrane JH. 2001. Asset Pricing. Princeton, NJ: Princeton Univ. Press
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Colacito R, Croce M, Ho S, Howard P. 2018. BKK the EZ way: international long-run growth news and
capital flows. Am. Econ. Rev. 108:3416–49
Croce MM. 2014. Long-run productivity risk: a new hope for production-based asset pricing? J. Monet. Econ.
66:13–31
Cuñat A, Maffezzoli M. 2004. Neoclassical growth and commodity trade. Rev. Econ. Dyn. 7:707–36
Cuñat A, Maffezzoli M. 2007. Can comparative advantage explain the growth of US trade? Econ. J. 117:583–
602
de Boor C. 1978. A Practical Guide to Splines. New York: Springer Verlag
Den Haan WJ, Judd KL, Juillard M. 2010. Computational suite of models with heterogeneous agents: incom-
plete markets and aggregate uncertainty. J. Econ. Dyn. Control 34:1–3
Den Haan WJ, Judd KL, Juillard M. 2011. Computational suite of models with heterogeneous agents II:
multi-country real business cycle models. J. Econ. Dyn. Control 35:175–77
Den Haan WJ, Marcet A. 1990. Solving the stochastic growth model by parameterizing expectations. J. Bus.
Econ. Stat. 8:31–34
Den Haan WJ, Marcet A. 1994. Accuracy in simulations. Rev. Econ. Stud. 61:3–17
Den Haan WJ, Rendahl P. 2010. Solving the incomplete markets model with aggregate uncertainty using
explicit aggregation. J. Econ. Dyn. Control 34:69–78
Dergunov I, Meinerding C, Schlag C. 2019. Extreme inflation and time-varying consumption growth. Discuss.
Pap. 16/2019, Deutsche Bundesbank, Frankfurt
Devereux MB, Siu HE. 2007. State dependent pricing and business cycle asymmetries. Int. Econ. Rev. 48:281–
310
Diamond PA. 1982. Aggregate demand management in search equilibrium. J. Political Econ. 90:881–94
Doraszelski U. 2003. An R&D race with knowledge accumulation. RAND J. Econ. 34:20–42
Doraszelski U. 2004. Innovations, improvements, and the optimal adoption of new technologies. J. Econ. Dyn.
Control 28:1461–80
Dou W, Fang X, Lo AW, Uhlig H. 2020. Macro-finance models with nonlinear dynamics. Work. Pap., Becker
Friedman Inst. Econ., Univ. Chicago, Chicago
Drechsler I, Yaron A. 2010. What’s vol got to do with it. Rev. Financ. Stud. 24:1–45
Duffy J, McNelis PD. 2001. Approximating and simulating the stochastic growth model: parameterized ex-
pectations, neural networks, and the genetic algorithm. J. Econ. Dyn. Control 25:1273–303
Epstein LG, Zin SE. 1991. Substitution, risk aversion, and the temporal behavior of consumption and asset
returns: an empirical analysis. J. Political Econ. 99:263–86
Favilukis J, Lin X. 2013. Long run productivity risk and aggregate investment. J. Monet. Econ. 60:737–51
Favilukis J, Lin X. 2015. Wage rigidity: a quantitative solution to several asset pricing puzzles. Rev. Financ.
Stud. 29:148–92
Fernández-Villaverde J, Gordon G, Guerrón-Quintana P, Rubio-Ramirez JF. 2015. Nonlinear adventures at
the zero lower bound. J. Econ. Dyn. Control 57:182–204

350 Miftakhova • Schmedders • Schumacher


Fernández-Villaverde J, Levintal O. 2018. Solution methods for models with rare disasters. Quant. Econ. 9:903–
44
Gallant AR, Jahan-Parvar MR, Liu H. 2019. Does smooth ambiguity matter for asset pricing? Rev. Financ.
Stud. 32:3617–66
Gaspar J, Judd KL. 1997. Solving large-scale rational-expectations models. Macroecon. Dyn. 1:45–75
Gertler M, Karadi P. 2011. A model of unconventional monetary policy. J. Monet. Econ. 58:17–34
Gertler M, Kiyotaki N. 2010. Financial intermediation and credit policy in business cycle analysis. In Handbook
of Monetary Economics, Vol. 3, ed. BM Friedman, M Woodford, pp. 547–99. Amsterdam: Elsevier
Gomes F, Michaelides A. 2007. Asset pricing with limited risk sharing and heterogeneous agents. Rev. Financ.
Stud. 21:415–48
Gottardi P, Kubler F. 2011. Social security and risk sharing. J. Econ. Theory 146:1078–106
Gourio F. 2012. Disaster risk and business cycles. Am. Econ. Rev. 102:2734–66
Gourio F. 2013. Credit risk and disaster risk. Am. Econ. J. Macroecon. 5:1–34
Gräber N, Schumacher M. 2019. Solving DSGE models—when local approximations fail. Work. Pap., Univ.
Münster, Münster, Ger.
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Hagedorn M, Manovskii I. 2008. The cyclical behavior of equilibrium unemployment and vacancies revisited.
Am. Econ. Rev. 98:1692–706
Hall RE. 1971. The dynamic effects of fiscal policy in an economy with foresight. Rev. Econ. Stud. 38:229–44
Heer B, Maussner A. 2018. Projection methods and the curse of dimensionality. J. Math. Finance 8:317–34
Hubbard TP, Paarsch HJ. 2009. Investigating bid preferences at low-price, sealed-bid auctions with endoge-
nous participation. Int. J. Ind. Organ. 27:1–14
İmrohoroğlu A, Tüzel Ş. 2014. Firm-level productivity, risk, and return. Manag. Sci. 60:2073–90
Jermann UJ. 1998. Asset pricing in production economies. J. Monet. Econ. 41:257–75
Jermann UJ, Quadrini V. 2012. Macroeconomic effects of financial shocks. Am. Econ. Rev. 102:238–71
Jones LE, Manuelli RE, Siu HE. 2005. Fluctuations in convex models of endogenous growth, II: business cycle
properties. Rev. Econ. Dyn. 8:805–28
Judd KL. 1992. Projection methods for solving aggregate growth models. J. Econ. Theory 58:410–52
Judd KL. 1996. Approximation, perturbation, and projection methods in economic analysis. In Handbook of
Computational Economics, Vol. 1, ed. K Schmedders, KL Judd, pp. 509–85. Amsterdam: Elsevier
Judd KL. 1998. Numerical Methods in Economics. Cambridge, MA: MIT Press
Judd KL, Guu SM. 1993. Perturbation solution methods for economic growth models. In Economic and Finan-
cial Modeling with Mathematica® , ed. HR Varian, pp. 80–103. New York: Springer
Judd KL, Guu SM. 1997. Asymptotic methods for aggregate growth models. J. Econ. Dyn. Control 21:1025–42
Judd KL, Kubler F, Schmedders K. 2003. Computational methods for dynamic equilibria with heteroge-
neous agents. In Advances in Economics and Econometrics: Theory and Applications, Eight World Congress, ed.
M Dewatripont, LP Hansen, SJ Turnovsky, pp. 243–90. Cambridge, UK: Cambridge Univ. Press
Judd KL, Maliar L, Maliar S. 2010. A cluster-grid projection method: solving problems with high dimensionality.
NBER Work. Pap. 15965
Judd KL, Maliar L, Maliar S. 2011. Numerically stable and accurate stochastic simulation approaches for
solving dynamic economic models. Quant. Econ. 2:173–210
Kaltenbrunner G, Lochstoer LA. 2010. Long-run risk through consumption smoothing. Rev. Financ. Stud.
23:3190–224
Kollmann R, Maliar S, Malin BA, Pichler P. 2011. Comparison of solutions to the multi-country Real Business
Cycle model. J. Econ. Dyn. Control 35:186–202
Krueger D, Kubler F. 2004. Computing equilibrium in OLG models with stochastic production. J. Econ. Dyn.
Control 28:1411–36
Kung H, Schmid L. 2015. Innovation, growth, and asset prices. J. Finance 70:1001–37
Lan H. 2018. Comparing solution methods for DSGE models with labor market search. Comput. Econ. 51:1–34
Leith C, Liu D. 2016. The inflation bias under Calvo and Rotemberg pricing. J. Econ. Dyn. Control 73:283–97
Lemoine D, Traeger CP. 2016. Economics of tipping the climate dominoes. Nat. Climate Change 6:514–19
Levintal O. 2018. Taylor projection: a new solution method for dynamic general equilibrium models. Int. Econ.
Rev. 59:1345–73

www.annualreviews.org • Computing Economic Equilibria 351


Lontzek TS, Narita D. 2011. Risk-averse mitigation decisions in an unpredictable climate system. Scand. J.
Econ. 113:937–58
Lorenz F, Schmedders K, Schumacher M. 2020. Nonlinear dynamics in conditional volatility. Work. Pap., Univ.
Münster, Münster, Ger.
Lorenz F, Schumacher M. 2018. Downside risks and the price of variance uncertainty. Work. Pap., Univ. Münster,
Münster, Ger.
Magill MJ. 1977. A local analysis of N-sector capital accumulation under uncertainty. J. Econ. Theory 15:211–19
Maliar L, Maliar S. 2003. Parameterized expectations algorithm and the moving bounds. J. Bus. Econ. Stat.
21:88–92
Maliar L, Maliar S. 2015. Merging simulation and projection approaches to solve high-dimensional problems
with an application to a new Keynesian model. Quant. Econ. 6:1–47
Maliar L, Maliar S, Valli F. 2010. Solving the incomplete markets model with aggregate uncertainty using the
Krusell–Smith algorithm. J. Econ. Dyn. Control 34:42–49
Maliar S, Maliar L, Judd K. 2011. Solving the multi-country real business cycle model using ergodic set meth-
ods. J. Econ. Dyn. Control 35:207–28
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Malin BA, Krueger D, Kubler F. 2011. Solving the multi-country real business cycle model using a Smolyak-
collocation method. J. Econ. Dyn. Control 35:229–39
Mortensen DT. 1982. Property rights and efficiency in mating, racing, and related games. Am. Econ. Rev.
72:968–79
Ngo PV. 2014. Optimal discretionary monetary policy in a micro-founded model with a zero lower bound on
nominal interest rate. J. Econ. Dyn. Control 45:44–65
Niemann S, Pichler P, Sorger G. 2013. Central bank independence and the monetary instrument problem.
Int. Econ. Rev. 54:1031–55
Papageorgiou C, Perez-Sebastian F. 2004. Can transition dynamics explain the international output data?
Macroecon. Dyn. 8:466–92
Papageorgiou C, Perez-Sebastian F. 2007. Is the asymptotic speed of convergence a good proxy for the tran-
sitional growth path? J. Money Credit Bank. 39:1–24
Petrosky-Nadeau N, Zhang L. 2017. Solving the Diamond–Mortensen–Pissarides model accurately. Quant.
Econ. 8:611–50
Petrosky-Nadeau N, Zhang L, Kuehn LA. 2013. Endogenous disasters and asset prices. Work. Pap., Carnegie
Mellon Univ., Pittsburgh, PA
Petrosky-Nadeau N, Zhang L, Kuehn LA. 2018. Endogenous disasters. Am. Econ. Rev. 108:2212–45
Pichler P. 2011. Solving the multi-country Real Business Cycle model using a monomial rule Galerkin method.
J. Econ. Dyn. Control 35:240–51
Pichler P, Sorger G. 2009. Wealth distribution and aggregate time-preference: Markov-perfect equilibria in
a Ramsey economy. J. Econ. Dyn. Control 33:1–14
Pissarides CA. 1985. Short-run equilibrium dynamics of unemployment vacancies, and real wages. Am. Econ.
Rev. 75:676–90
Pohl W, Schmedders K, Wilms O. 2018. Higher order effects in asset pricing models with long-run risks.
J. Finance 73:1061–111
Pohl W, Schmedders K, Wilms O. 2020. Asset pricing with heterogeneous agents and long-run risk. J. Financ.
Econ. In press
Reddien GW. 1980. Projection methods for two-point boundary value problems. SIAM Rev. 22:156–71
Reich G, Wilms O. 2015. Adaptive grids for the estimation of dynamic models. Work. Pap., Univ. Zurich, Zurich,
Switz.
Reiter M. 2009. Solving heterogeneous-agent models by projection and perturbation. J. Econ. Dyn. Control
33:649–65
Reiter M. 2010. Solving the incomplete markets model with aggregate uncertainty by backward induction.
J. Econ. Dyn. Control 34:28–35
Reiter M. 2015. Solving OLG models with many cohorts, asset choice and large shocks. Work. Pap., Inst. Adv. Stud.,
Vienna, Austria
Renner P, Scheidegger S. 2018. Machine learning for dynamic incentive problems. Work. Pap., Univ. Lancaster,
Lancaster, UK

352 Miftakhova • Schmedders • Schumacher


Rotemberg JJ. 1982. Sticky prices in the United States. J. Political Econ. 90:1187–211
Routledge BR, Zin SE. 2010. Generalized disappointment aversion and asset prices. J. Finance 65:1303–32
Sánchez-Marcos V, Sánchez-Martín AR. 2006. Can social security be welfare improving when there is demo-
graphic uncertainty? J. Econ. Dyn. Control 30:1615–46
Scheidegger S, Bilionis I. 2019. Machine learning for high-dimensional dynamic stochastic economies. J. Com-
put. Sci. 33:68–82
Schumacher M, Żochowski D. 2017. The risk premium channel and long-term growth. Work. Pap. 2114, Eur.
Cent. Bank, Frankfurt
Schumaker L. 2007. Spline Functions: Basic Theory. Cambridge, UK: Cambridge Univ. Press. 3rd ed.
Silvester PP, Ferrari RL. 1996. Finite Elements for Electrical Engineers. Cambridge, UK: Cambridge Univ. Press.
3rd ed.
Song Z, Storesletten K, Zilibotti F. 2012. Rotten parents and disciplined children: a politico-economic theory
of public expenditure and debt. Econometrica 80:2785–803
Taylor JB, Uhlig H. 1990. Solving nonlinear stochastic growth models: a comparison of alternative solution
methods. J. Bus. Econ. Stat. 8:1–17
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Trefethen LN. 2013. Approximation Theory and Approximation Practice. Philadelphia: SIAM
Tuzel S. 2010. Corporate real estate holdings and the cross-section of stock returns. Rev. Financ. Stud. 23:2268–
302
van Zandweghe W, Wolman AL. 2019. Discretionary monetary policy in the Calvo model. Quant. Econ.
10:387–418
Weil P. 1989. The equity premium puzzle and the risk-free rate puzzle. J. Monet. Econ. 24:401–21
Werner M. 2016. Occasionally binding liquidity constraints and macroeconomic dynamics. Work. Pap., Univ. Zurich,
Zurich, Switz.
Young ER. 2010. Solving the incomplete markets model with aggregate uncertainty using the Krusell–Smith
algorithm and non-stochastic simulations. J. Econ. Dyn. Control 34:36–41

www.annualreviews.org • Computing Economic Equilibria 353


EC12_FrontMatter ARI 12 June 2020 12:50

Annual Review of
Economics

Contents Volume 12, 2020

Economics with a Moral Compass? Welfare Economics: Past, Present,


and Future
Amartya Sen, Angus Deaton, and Timothy Besley p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Trade Policy in American Economic History
Access provided by Arizona State University on 10/21/21. For personal use only.
Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Douglas A. Irwin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p23


An Econometric Perspective on Algorithmic Subsampling
Sokbae Lee and Serena Ng p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p45
Behavioral Implications of Causal Misperceptions
Ran Spiegler p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p81
Poverty and the Labor Market: Today and Yesterday
Robert C. Allen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 107
The Econometrics of Static Games
Andrés Aradillas-López p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 135
On Measuring Global Poverty
Martin Ravallion p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 167
Taxation and the Superrich
Florian Scheuer and Joel Slemrod p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 189
How Distortions Alter the Impacts of International Trade in Developing
Countries
David Atkin and Amit K. Khandelwal p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 213
Robust Decision Theory and Econometrics
Gary Chamberlain p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 239
Cities in the Developing World
Gharad Bryan, Edward Glaeser, and Nick Tsivanidis p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 273
New Developments in Revealed Preference Theory: Decisions Under
Risk, Uncertainty, and Intertemporal Choice
Federico Echenique p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 299
Computing Economic Equilibria Using Projection Methods
Alena Miftakhova, Karl Schmedders, and Malte Schumacher p p p p p p p p p p p p p p p p p p p p p p p p p p p p 317

v
EC12_FrontMatter ARI 12 June 2020 12:50

Social Identity and Economic Policy


Moses Shayo p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 355
Empirical Models of Lobbying
Matilde Bombardini and Francesco Trebbi p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 391
Political Effects of the Internet and Social Media
Ekaterina Zhuravskaya, Maria Petrova, and Ruben Enikolopov p p p p p p p p p p p p p p p p p p p p p p p p p 415
Nash Equilibrium in Discontinuous Games
Philip J. Reny p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 439
Revealed Preference Analysis of School Choice Models
Nikhil Agarwal and Paulo Somaini p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 471
Access provided by Arizona State University on 10/21/21. For personal use only.

Social Networks and Migration


Annu. Rev. Econ. 2020.12:317-353. Downloaded from www.annualreviews.org

Kaivan Munshi p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 503


Informality: Causes and Consequences for Development
Gabriel Ulyssea p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 525
The Theory and Empirics of the Marriage Market
Pierre-André Chiappori p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 547
Modeling Imprecision in Perception, Valuation, and Choice
Michael Woodford p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 579
Peer Effects in Networks: A Survey
Yann Bramoullé, Habiba Djebbari, and Bernard Fortin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 603
Alternative Work Arrangements
Alexandre Mas and Amanda Pallais p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 631
Shotgun Wedding: Fiscal and Monetary Policy
Marco Bassetto and Thomas J. Sargent p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 659
Social Identity, Group Behavior, and Teams
Gary Charness and Yan Chen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 691
Aspirations and Economic Behavior
Garance Genicot and Debraj Ray p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 715
The Search Theory of Over-the-Counter Markets
Pierre-Olivier Weill p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 747
Econometric Models of Network Formation
Áureo de Paula p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 775
Dynamic Taxation
Stefanie Stantcheva p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 801
Capital Flows and Leverage
Şebnem Kalemli-Özcan and Jun Hee Kwak p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 833

vi Contents

You might also like