Cons Book

Consumption and Saving: Theory and Evidence
Dirk Krueger
1
Department of Economics
University of Pennsylvania
October 2007
1
I wish to thank Orazio Attanasio, Richard Blundell, Jesus Fernandez-Villaverde,
Robert Hall, Narayana Kocherlakota, Fabrizio Perri, Luigi Pistaferri, Victor Rios-
Rull and Thomas Sargent for teaching me the economics of consumption and sav-
ing. c by Dirk Krueger. All comments are welcomed, please contact the author at
dkrueger@econ.upenn.edu.
ii
Contents
1 Introduction and Overview 1
2 Some Stylized Facts and Some Puzzles 5
2.1 Aggregate Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Empirical Findings from Micro Data . . . . . . . . . . . . . . . . 8
2.2.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Main Findings . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Two Benchmark Models 15
3.1 Complete Markets . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.1 Theory: The Representative Agent . . . . . . . . . . . . . 16
3.1.2 Empirical Tests of the Complete Consumption Insurance
Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Permanent Income-Life Cycle Hypothesis (PILCH) Models . . . 34
3.2.1 A Simple 2 Period Toy Model . . . . . . . . . . . . . . . . 34
3.2.2 Many Periods and Certainty Equivalence . . . . . . . . . 37
3.2.3 Empirical Tests of the Martingale Hypothesis Using Macro-
economic Data . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2.4 Consumption Responses to Income Shocks under Certainty
Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 Partial Equilibrium Extensions of PILCH Models 49
4.1 Precautionary Savings . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.1 A Simple Model and a General Result . . . . . . . . . . . 50
4.1.2 A Parametric Example for the General Model . . . . . . . 53
4.2 Liquidity Constraints . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.1 The Euler Equation with Liquidity Constraints . . . . . . 56
4.2.2 Liquidity Constraints with Quadratic Preferences: Pre-
cautionary Saving without Prudence . . . . . . . . . . . . 57
4.3 Numerical Solutions of Models with Precautionary Saving and
Liquidity Constraints . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3.1 Implementation of Particular Algorithms . . . . . . . . . 64
4.4 Combining Prudence and Liquidity Constraints . . . . . . . . . . 64
4.4.1 Case T nite . . . . . . . . . . . . . . . . . . . . . . . . . 65
iii
iv CONTENTS
4.4.2 Case T = and j < r . . . . . . . . . . . . . . . . . . . . 65
4.4.3 Case T = and j = r . . . . . . . . . . . . . . . . . . . . 66
4.4.4 Case T = and j r . . . . . . . . . . . . . . . . . . . . 67
5 PILCH Models in General Equilibrium 75
5.1 A Model Without Aggregate Uncertainty . . . . . . . . . . . . . 76
5.1.1 The Environment . . . . . . . . . . . . . . . . . . . . . . . 76
5.1.2 Theoretical Results: Existence and Uniqueness . . . . . . 82
5.1.3 Computation of the General Equilibrium . . . . . . . . . 86
5.1.4 Qualitative Results . . . . . . . . . . . . . . . . . . . . . . 86
5.1.5 Numerical Results . . . . . . . . . . . . . . . . . . . . . . 87
5.2 Unexpected Aggregate Shocks and Transition Dynamics . . . . . 91
5.2.1 Denition of Equilibrium . . . . . . . . . . . . . . . . . . 91
5.2.2 Computation of the Transition Path . . . . . . . . . . . . 93
5.2.3 Welfare Consequences of the Policy Reform . . . . . . . . 94
5.3 Aggregate Uncertainty and Distributions as State Variables . . . 96
5.3.1 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.3.2 Computation of the Recursive Equilibrium . . . . . . . . 100
5.3.3 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.3.4 Numerical Results . . . . . . . . . . . . . . . . . . . . . . 105
5.3.5 Why Quasi-Aggregation? . . . . . . . . . . . . . . . . . . 107
5.3.6 Rich People are Not Rich Enough . . . . . . . . . . . . . 108
6 Complete Market Models with Frictions 111
6.1 Limited Enforceability of Contracts . . . . . . . . . . . . . . . . . 112
6.1.1 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.1.2 Constrained Ecient Allocations . . . . . . . . . . . . . . 114
6.1.3 Recursive Formulation of the Problem . . . . . . . . . . . 115
6.1.4 Decentralization . . . . . . . . . . . . . . . . . . . . . . . 116
6.1.5 A Simple Application: Income and Consumption Inequality118
6.1.6 A Continuum Economy . . . . . . . . . . . . . . . . . . . 121
6.2 Private Information . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2.1 Partial Equilibrium . . . . . . . . . . . . . . . . . . . . . . 127
6.2.2 Endogenous Interest Rates in General Equilibrium . . . . 134
Chapter 1
Introduction and Overview
In these notes we will rst summarize some key empirical regularities charac-
terizing individual consumption and saving behavior over time and states of the
world. We will also document some empirical puzzles that we aim at explaining
with the models to be developed below. Note that to call an empirical nd-
ing a puzzle requires to take a stand on the standard economic theory relative
to which the nding is puzzling; on of the goals of the course is to develop
extensions or, if needed, radical departures, of standard theory to explain the
empirical puzzles.
We then will go on and develop two standard benchmark models for the
study of intertemporal consumption and saving decisions: a Permanent Income-
Life Cycle type model and the Complete Markets model in the spirit of Arrow
and Debreu. The theoretical part of the course will be centered around these
models and their direct extension. Both start from the basic income uctuation
problem: a consumer with stochastic income stream in each period has to decide
how much to consume and how much to save.
The Permanent Income-Life Cycle type models, developed by Friedman and
Modigliani, respectively, assume that individuals have access to a single, one
period risk-free bond and can borrow as much as desired (i.e. dont face po-
tentially binding borrowing constraints
1
) to smooth consumption over time.
The main dierence between the Permanent Income Hypothesis of Friedman
(1957) and the Life-Cycle hypothesis of Modigliani and Brumberg (1954) is the
(implicit) assumption about the life-time horizon of the consumer and the im-
portance of life-cycle phenomena: Friedmans discussion focused on an innitely
lived consumer whereas Modigliani and Brumberg assumed nite lives of agents.
Common to both studies is the focus on a single consumer (interest rates are
exogenously given and not derived endogenously in general equilibrium) and
the absence of explicit insurance arrangements against individual uncertainty
(these are ruled out in ad hoc fashion). The basic model also makes sucient
1
One obviously needs some constraint on borrowing to rule out Ponzi schemes. Whenever
the original papers dont explicitly state them, always keep in the back of your mind that they
are implicitly there to make the consumer problem well-dened.
1
2 CHAPTER 1. INTRODUCTION AND OVERVIEW
assumptions (e.g. quadratic utility) to obtain decision rules that obey certainty
equivalence: decision rules that are identical to those obtained in the absence
of income uncertainty.
Extensions of the basic theory weaken the assumptions leading to certainty
equivalence and allow for (potentially binding) borrowing constraints, leading to
precautionary and buer-stock saving behavior (these concepts will be made pre-
cise below). Further extensions allow for endogenous labor-leisure choice (mak-
ing labor income an endogenous, instead of an exogenous stochastic process)
or/and endogenous fertility choice.
Some version of these partial equilibrium models is then incorporated into
a dynamic general equilibrium model in which interest rates and wages are de-
termined endogenously in the labor and capital market. Individual households
consumption and saving decisions are aggregated to obtain aggregate labor and
capital supply, rms decisions are aggregated to obtain aggregate labor and
capital demand, and wages and interest rates move to clear both markets. De-
pending on whether individual uncertainty averages out in the aggregate (no
aggregate uncertainty) wages and interest are constant over time or are them-
selves stochastic processes (presence of aggregate uncertainty), leading to se-
vere computational problems when computing these models. The aggregation
of individual decisions also leads to (possibly time-varying) cross-sectional con-
sumption and wealth distributions; thus these models are possibly useful for the
study of the eects of redistributive and social insurance policies such as pro-
gressive taxation, unemployment insurance, welfare or social security. If time
permits, we will look at some of these applications.
Common to all these models is the assumption of the absence of explicit
insurance arrangements in an environment in which mutual insurance is po-
tentially quite benecial. A second, newer and less developed strand of the
literature, to be discussed next, aims to explain stylized consumption and sav-
ings facts with models that depart directly from the complete markets model.
Without a priori ruling out explicit insurance contracts features such as private
information, moral hazard or limited enforceability of contracts are added to the
basic model. The complete markets model itself assumes that all agents have
access to a complete set of contingent claims, giving rise to perfect consump-
tion insurance (which has to be dened more precisely below). This implies
a sharp and analytically tractable aggregation result (the existence of a repre-
sentative consumer) and a powerful (yet empirically decient) theory of asset
prices. However, complete consumption insurance, the main empirical impli-
cation of the complete markets model, is empirically rejected numerous times.
The second part of this course will provide you with the main references.
With the introduction of private information or/and limited enforceability of
contracts imperfect consumption insurance arises endogenously as (constrained)
optimal response to these frictions. These models provide a competing explana-
tion to the PILCH models for some of the empirical regularities in the consump-
tion and saving literature. In order to make these models recursive (and hence
computable) we will follow Spear and Srivastava and Abreu, Pearce and Stac-
chetti and use promised utility as state variable, as has become common in the
3
recursive mechanism design literature. We will then evaluate the empirical suc-
cess of these complete markets general equilibrium models with informational or
enforcement frictions and compare them with the general equilibrium versions
of the PILCH models.
4 CHAPTER 1. INTRODUCTION AND OVERVIEW
Chapter 2
Some Stylized Facts and
Some Puzzles
2.1 Aggregate Data
First we take a look at aggregate data from the National Income and Product
Accounts. In Figure 2.1 we plot total consumption expenditure as a fraction of
either GDP or personal income
1
for the US between 1946 and 2000. We see that
between 60% and 70% of GDP is used for private consumption. This fraction is
fairly stable over time, although one can detect an upward trend starting in the
80s and continuing through the 90s. The rapid increase in private consumption
has lead to a severe decline in the personal savings rate :
, which is dened as
:
= 1
C
1
JI
where C
is total private consumption expenditures and 1

JI
is personal dis-
posable income, which equals personal income minus income taxes. Figure 2.2
shows the personal savings rate from 1959 onwards, demonstrating the spectac-
ular decline in private households saving in the 90s.
In Figure 2.3 we plot the levels of real consumption expenditures and real
GDP against time. Note that the y-axis has logarithmic scale. We see that con-
sumption is somewhat smoother over time than GDP, indicating that households
have (limited) ability to isolate their consumption path from income uctuations
induced by the business cycle.
Turning to the decomposition of consumption, Figure 2.4 plots the fractions
of total consumption expenditures devoted to purchases of consumer durables,
nondurables and services. Note that purchases of new homes are counted as in-
vestment and hence are not included in consumption expenditures on durables.
We see that purchases of consumer durables (primarily new cars, household
1
Personal Income is used to pay for either consumption, saving or income taxes.
5
6 CHAPTER 2. SOME STYLIZED FACTS AND SOME PUZZLES
1950 1960 1970 1980 1990 2000
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Consumption as Fraction of GDP, Personal Income, 1946-2000
Year
C
o
n
s
u
m
p
t
i
o
n

a
s

F
r
a
c
t
i
o
n

o
f

G
D
P
,

P
e
r
s
.

I
n
c
.
Cons/ GDP
Cons/Pers Inc.
Figure 2.1: Consumption as Fraction of Income
appliances and electronics) make up about 13% of total consumer expenditures,
with this fraction slightly increasing over time. The quantitative importance of
nondurables (food, alcohol, apparel and the like) has declined over time, whereas
expenditures for consumer services make up about 60% of total consumer ex-
penditures now.
The previous picture gives an accurate picture of the expenditure shares
of durables, nondurables and services. However, since they were computed
using nominal data, they may give a misleading picture about how the impor-
tance of these various categories has developed over time. In particular, relative
prices of these components may have shifted over time. In Figure 2.5 we plot
shares of real consumption expenditures, where each group of expenditure is
deated with its own deator. Unfortunately these data starts only in 1987,
but the picture nevertheless indicates that as of late consumers seem to re-
allocate consumption towards durables and away from nondurables (including
2.1. AGGREGATE DATA 7
1960 1965 1970 1975 1980 1985 1990 1995 2000
0
2
4
6
8
10
12
14
16
18
20
Personal Sav ings Rate, 1959-2000
Year
P
e
r
s
o
n
a
l

S
a
v
i
n
g
s

R
a
t
e
Figure 2.2: Personal Savings Rate
services). The two previous gures are reconciled by noting that the price of
services has increased substantially over the last decade or so, relative to the
price of other nondurables and durables. For 1959 to 1996, Attanasio (1999)
documents that real nondurable consumption (including services) grew at an
average rate of 2.3% annually, whereas real expenditures on durables increased
at an annual rate of 4.8%. Moreover, consumption growth of nondurables was
highly stable (the standard deviation of the growth rate was 1.8%), whereas for
durable expenditures the same standard deviation was 6.9%. Hence although
total consumption expenditure is smoother than GDP, one of its component,
namely durables expenditures, is highly volatile and an important contributor
to the business cycle (together with inventory investment and residential xed
investment)
1950 1960 1970 1980 1990 2000
10
3
10
4
Real Consumption and GDP, 1947-2000
Year
C
o
n
s
u
m
p
t
i
o
n

a
n
d

G
D
P
,

L
o
g
a
r
i
t
h
m
i
c

S
c
a
l
e
Consumption
GDP
Figure 2.3: Levels of Real Consumption Expenditures and GDP
2.2 Empirical Findings from Micro Data
The previous section looked at aggregate data on consumption. In our the-
oretical analysis we will start with a single household as the basic entity of
investigation. Only under special assumptions to be discussed below does the
aggregation over consumers in the model lead to consumption aggregates that
look like consumption processes of the individuals from which the aggregate
was derived. Hence direct empirical tests of our models usually require empiri-
cal data on the household level.
2.2.1 Data Sources
For the US, the only household level data set with extensive information about
a wide range of consumption expenditures is the Consumer Expenditure Survey
2.2. EMPIRICAL FINDINGS FROM MICRO DATA 9
1950 1960 1970 1980 1990 2000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Components of Consumption as Fraction of Total Consumption, 1946-2000
Year
C
o
m
p
o
n
e
n
t
s

o
f

C
o
n
s
u
m
p
t
i
o
n
Durables
Nondurables
Serv ices
Figure 2.4: Components of Consumption
(CEX) or (CES).
2
The CEX is conducted by the U.S. Bureau of the Census and
sponsored by the Bureau of Labor statistics. From 1980 onwards the survey
is carried out on a yearly basis. The CEX is a so-called rotating panel: each
household in the sample is interviewed for four consecutive quarters and then
rotated out of the survey. Hence in each quarter 20% of all households is rotated
out of the sample and replaced by new households. In each quarter about
3000 to 5000 households are in the sample, and the sample is representative
of the U.S. population. The main advantage of the CEX is that it contains
very detailed information about consumption expenditures. Information about
income and wealth is inferior to the Survey of Consumer Finances (SCF), the
Current Population survey (CPS) and the Panel Study of Income Dynamics
2
In class I will also discuss some European data sources, such as the British Family Expen-
diture Survey (FES) and the German Socio-Economic Panel (SOEP) and the Einkommens-
und Verbrauchsstichprobe.
1988 1990 1992 1994 1996 1998 2000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Components of Real Consumption as Fraction of Total Real Consumption, 1987-2000
Year
C
o
m
p
o
n
e
n
t
s

o
f

C
o
n
s
u
m
p
t
i
o
n
Durables
Nondurables
Serv ices
Figure 2.5: Components of Real Consumption Expenditures
(PSID), also the panel dimension is signicantly shorter than for the PSID (one
household is only followed for 4 quarters). For further information and the
complete data set see http://www.stats.bls.gov/csxhome.htm.
With respect to income, the PSID as well as the SCF contains data that
are supposedly of higher quality than the income data from the CEX. The
SCF is conducted in three year intervals; the four available surveys are for
the years 1989, 1992, 1995 and 1998. It is conducted by the National Opin-
ion Research center at the University of Chicago and sponsored by the Federal
Reserve system. It contains rich information about U.S. households income
and wealth. In each survey about 4,000 households are asked detailed ques-
tions about their labor earnings, income and wealth. One part of the sample
is representative of the U.S. population, to give an accurate description of the
entire population. The second part oversamples rich households, to get a more
precise idea about the precise composition of this groups income and wealth
composition. As we will see, this group accounts for the majority of total house-
hold wealth, and hence it is particularly important to have good information
about this group. The main advantage of the SCF is the level of detail of in-
formation about income and wealth. The main disadvantage is that it is not
a panel data set, i.e. households are not followed over time. Hence dynamics
of income and wealth accumulation cannot be documented on the household
level with this data set. For further information and some of the data see
http://www.federalreserve.gov/pubs/oss/oss2/98/scf98home.html.
The Panel Study of Income Dynamics (PSID) is conducted by the Survey
Research Center of the University of Michigan and mainly sponsored by the
National Science Foundation. The PSID is a panel data set that started with
a national sample of 5,000 U.S. households in 1968. The same sample individ-
uals are followed over the years, barring attrition due to death or nonresponse.
New households are added to the sample on a consistent basis, making the to-
tal sample size of the PSID about 8700 households. The income and wealth
data are not as detailed as for the SCF, but its panel dimension allows to con-
struct measures of income and wealth dynamics, since the same households
are interviewed year after year. Also the PSID contains data on consumption
expenditures, albeit only food consumption. In addition, in 1990, a represen-
tative national sample of 2,000 Latinos, dierentially sampled to provide ad-
equate numbers of Puerto Rican, Mexican-American, and Cuban-Americans,
was added to the PSID database. This provides a host of information for stud-
ies on discrimination. For further information and the complete data set see
http://www.isr.umich.edu/src/psid/index.html
The Current Population Survey (CPS) is conducted by the U.S. Bureau of
the Census and sponsored by the Bureau of Labor Statistics. In its annual
March supplement detailed information about household income is collected.
The survey started to gather information about household income in 1948, but
comprehensive information about household income and income of its members
is available only since 1970s. The main advantage of the CPS is its sample size:
in each year it contains a representative sample of 40,000 to 60,000 households.
However, no information about consumption or wealth information is collected.
Also, this survey, like the SCF is a purely cross-sectional data set without panel
dimension as it does not follow individual families over time. Fore more details
see http://www.bls.census.gov/cps.
2.2.2 Main Findings
If one follows an average household over its life cycle
3
, two main stylized facts
emerge (see Attanasio (1999) for the detailed gures). First, disposable income
follows a hump over the life cycle, with a peak around the age of 45 (the age of
the household is dened by the age of the household head). This nding is hardly
surprising, given that at young ages households tend to obtain formal education
3
How to construct such an average household in the absence of panel data sets which lack
a sucient time series dimension is a challenging problem. The solution, the so-called pseudo
panel method, will be discussed in the second half of the course.
or training on the job and labor force participation of women is low because
of child bearing and rearing. As more and more agents nish their education
and learn on the job as well as promotions occur, average wages within the
cohort increase. Average personal income at age 45 is almost 2.5 times as high
as average personal income at age 25. After the age of 45 personal income rst
slowly, then more rapidly declines as more and more people retire and labor
productivity (and thus often wages) fall. The average household at age 65 has
only 60% of the personal income that the average household at age 45 obtains.
The second main nding is the surprising nding. Not only personal income,
but also consumption follows a hump over the life cycle. In other words, con-
sumption seems to track income over the life cycle fairly closely. This is one
statement of the so-called excess sensitivity puzzle: consumption appears to
be excessively sensitive to predicted changes in income. In fact, the two stan-
dard theories of intertemporal consumption allocation we will consider in the
next section both predict that (under specic assumptions spelled out explicitly
below) consumption follows a martingale and current income does not help to
forecast future consumption. The hump-shaped consumption age prole appar-
ently seems to contradict this hypothesis. Later in the course we will investigate
in detail whether, once we control for household size (which also happens to fol-
low a hump shape), the hump-shape in consumption disappears or whether the
puzzle persists. Figure 2.6 (taken from Krueger and Fernandez-Villaverde, 2003)
documents the life cycle prole of consumption, with and without adjustment
for family size by household equivalence scales. The gure is derived using a
synthetic cohort analysis, a technique from Panel data econometrics that allows
us to construct average life cycle proles for households that we do not observe
over their entire life (we will talk about this technique in detail below). The key
observation from this gure is that consumption displays a hump over the life
cycle, and that this hump persists, even after controlling for family size. The
later observation is not entirely uncontroversial, and we will discuss below the
dierent positions on this issue.
Other empirical puzzles that we would like our extensions to the standard
models to address include the
1. Excess Smoothness Puzzle: if the stochastic income process of households
is only dierence stationary (say, it follows a random walk) then a shock to
current income translates (more than) one to one into a shock to perma-
nent income and hence should induce a large shock to consumption. With
dierence-stationary income processes consumption should be as volatile
as current income, but we saw that, at least on the aggregate level, con-
sumption is smoother than income. Is in fact consumption too smooth.
This is the excess smoothness puzzle
2. Lack of Decumulation Puzzle: Household level wealth data show that a
signicant fraction of very old households still hold a large portfolio of
nancial and real estate assets. Why dont these households decumulate
their assets and enhance their consumption, as standard life-cycle theory
predicts?
20 30 40 50 60 70 80 90
1500
2000
2500
3000
3500
4000
4500
Expenditures, Total and Adult Equivalent
Age
Total
Adul t Equi valent
Figure 2.6: Consumption over the Life Cycle
3. Drop of Consumption Puzzle: As people retire, their consumption drops
by about 15% on average.
4. Portfolio Allocation Puzzle: The median wealth US household does not
own stocks, but holds its major fraction of the wealth portfolio concen-
trated in its own home and the rest in low-return checking or savings
accounts. This is despite the fact that returns to equity and returns to
human capital (i.e. the households wage) are roughly uncorrelated for
this fraction of the population. In addition, Gross and Souleles (2000)
document that a signicant fraction of the population has simultaneously
high-interest credit card debt and liquid, low return assets such as a sig-
nicantly positive checking account balance.
5. Default Puzzle: the US legal system allows private households to le for
personal bankruptcy under Chapter 7 or Chapter 11 of the US Bankruptcy
Code. In particular, under Chapter 7 households are discharged of all their
debts, are not required to use any of their future labor income to repay
the debt and can even keep their assets (nancial or real estate) below a
state-dependent exemption level. Whereas about 1% of all households per
year le for personal bankruptcy, White (1998) computes that currently
at least 15% of all US households would nancially benet from ling for
bankruptcy.
Any given model will likely not be able to resolve all these puzzles at once,
and some models will abstract from some of the issues altogether, but the styl-
ized facts of this section should be kept in mind in order to guide extensions
of the models presented next.
Chapter 3
Two Benchmark Models
Let the economy be populated by individuals, indexed by i 1 = 1, 2, . . . .
Each individual lives for T periods, where T = is allowed. In each pe-
riod there is one nonstorable consumption good. Each individual household
has a stochastic endowment process j
I
|
of this consumption good. Let by
j
I,|
= (j
I
0
, j
I
l
, . . . j
I
|
) denote a history of endowment shocks of length t 1 for
agent i and by :
|
and :
|
= (:
0
, . . . :
|
) the event and event history of this econ-
omy, respectively. By
|
(j
I,|
) and
|
(:
|
) denote the objective probabilities of
event histories of individual and aggregate endowments, respectively. Agents
subjective probability beliefs are assumed to coincide with these objective prob-
abilities. Note that at this point we do not make any assumptions about the sto-
chastic process; in particular the individual or aggregate processes need not be
Markov processes and the individual processes need not be independent across
agents.
To avoid certain measurability issues we will assume that j
I
|
1, a nite-
dimensional set of cardinality ' and that :
|
o, a nite set of cardinality
1. Sometimes we will take the individual income processes as primitives and
dene the aggregate state simply as :
|
= (j
l
|
, . . . , j
|
), sometimes we will take
the aggregate process as primitive and dene individual income processes as
functions j
I
|
(:
|
) of aggregate event histories. Using slight abuse of notation let
by o
|
= o o . . . o denote the t 1-fold Cartesian product of o and let 1
|
be dened analogously. Then :
|
o
|
and j
I,|
1
|
for all i. Also, in order to
make some of the results a little easier to derive, take the event :
0
as given (i.e.
not random; obviously which event we start with is irrelevant for the discussion
to follow).
A consumption allocation (c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
maps aggregate event histo-
ries :
|
into consumption of agents i 1 at time t. Agents have preferences over
consumption allocations that are assumed to permit von Neumann Morgenstern
expected utility representation
n
I
(c
I
) =

s
t
S
t
|
(:
|
)l
I
|
(c
I
, :
|
) (3.1)
15
16 CHAPTER 3. TWO BENCHMARK MODELS
Unless otherwise noted, we will assume that preferences are identical across
agents, additively time-separable and that agents discount the future at common
subjective time discount factor , (0, 1), so that the utility function takes the
form
n
I
(c
I
) =
T
|=0
s
t
S
t
,
|
|
(:
|
)l
I
(c
I
|
(:
|
), :
|
) (3.2)
By j =
l
o
1 let denote the subjective time discount rate (i.e. , =
l
l
). The
presence of :
|
allows for the potential presence of preference shocks to individual
agents utility from consumption. For example, when treating fertility exoge-
nous, the arrival of a new child (which is then modeled as a stochastic process)
reduces utility from a given consumption level, provided that members care
about per capita consumption (and c
I
|
(:
|
) denotes consumption expenditures of
household i).
Denition 1 A consumption allocation (c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
is feasible if
c
I
|
(:
|
) _ 0 for all i, t, :
|
(3.3)
I=l
c
I
|
(:
|
) =

I=l
j
I
|
(:
|
) for all t, :
|
(3.4)
A consumption allocation is Pareto ecient if it is feasible and there is no other
feasible consumption allocation ( c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
such that
n
I
( c
I
) _ n
I
(c
I
) for all i 1 (3.5)
n
I
( c
I
) n
I
(c
I
) for some i 1 (3.6)
3.1 Complete Markets
3.1.1 Theory: The Representative Agent
The main distinction between the complete markets model and the PILCH type
models is the set of assets that agents can trade to hedge against individual
income uncertainty, and hence the individual budget constraints. With complete
markets we suppose that there is a complete set of contingent consumption
claims that are traded at time 0, before any uncertainty has been revealed. The
individual Arrow Debreu budget constraints take the form
T
|=0
s
t
S
t
j
|
(:
|
)c
I
|
(:
|
) _
T
|=0
s
t
S
t
j
|
(:
|
)j
I
|
(:
|
) (3.7)
where j
|
(:
|
) is the period 0 price of one unit of period t consumption, delivered
if event history :
|
has realized.
3.1. COMPLETE MARKETS 17
Arrow-Debreu Equilibrium
Denition 2 An Arrow Debreu competitive equilibrium consists of allocations
(c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
and prices j
|
(:
|
)
T
|=0,s
t
S
t
such that
1. Given j
|
(:
|
)
T
|=0,s
t
S
t
, for each i 1, c
I
|
(:
|
)
T
|=0,s
t
S
t
maximizes (8.2)
subject to (8.8) and (8.7)
2. (c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
satises (8.4) for all t, :
|
.
Now let us make the following assumption
Assumption 1: The period utility functions l
I
are twice continuously dif-
ferentiable, strictly increasing, strictly concave in its rst argument and satisfy
the Inada conditions
lim
c0
l
I
c
(c, :
|
) = (3.8)
lim
co
l
I
c
(c, :
|
) = 0 (3.9)
where l
I
c
is the derivative of l with respect to its rst argument.
It is then straightforward to prove the rst welfare theorem for this econ-
omy (in fact, for this result we only need that the utility functions are strictly
increasing). Hence any competitive equilibrium allocation is the solution to the
social planners problem of
max
](c
1
t
(s
t
))
121
]
J
t=0s
t
2:
t
I=l
c
I
n
I
(c
I
) (3.10)
subject to (8.8) and (8.4), for some Pareto weights (c
I
)
I=l
satisfying c
I
_ 0
and

I=l
c
I
= 1 (see, e.g. MasColell et. al., chapter 16; this result requires
parts of assumption 1, especially concavity). Attaching Lagrange multipliers
`(:
|
) to the resource constraint and ignoring the non-negativity constraints on
consumption we obtain as rst order necessary conditions for an optimum
c
I
,
|
|
(:
|
)l
I
c
(c
I
|
(:
|
), :
|
) = `
|
(:
|
) (3.11)
for all i 1. Hence for i, , 1
l
I
c
(c
I
|
(:
|
), :
|
)
l
c
(c
|
(:
|
), :
|
)
=
c
c
I
(3.12)
for all dates t and all states :
|
. Hence with a complete set of contingent con-
sumption claims the ratio of marginal utilities of consumption of any two agents
is constant across time and states. Also agents, ceteris paribus (i.e. if they had
the same utility function and the same preference shock), with higher relative
Pareto weights will consume more in every state of the world because the utility
function is assumed to be strictly concave.
Denition 3 A consumption allocation (c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
is said to satisfy
perfect consumption insurance if the ratio of marginal utilities of consumption
between any two agents is constant across time and states of the world.
Thus, the complete markets model exhibits perfect consumption insurance.
With additional assumptions we obtain an even stronger (and hence easier to
test empirically) implication of the complete markets model.
Assumption 2: All agents have identical CRRA utility, and either prefer-
ence shocks are absent or utility is separable in consumption, i.e.
l
I
(c, :
|
) =
c
lc
1
1 o
(:
|
) (3.13)
with o _ 0 (for o = 1 it is understood that utility is logarithmic). Here o is the
coecient of relative risk aversion.
With this assumption (8.12) becomes
c
I
|
(:
|
)
c
|
(:
|
)
=
_
c
I
c
_
1
o
(3.14)
i.e. the ratio of consumption between any two agents is constant across time
and states. This, in particular, implies that there exist weights (0
I
)
I1
with
0
I
_ 0 and

I1
0
I
= 1 such that
c
I
|
(:
|
) = 0
I
I1
j
I
|
(:
|
) = 0
I
j
|
(:
|
) = 0
I
c
|
(:
|
) (3.15)
where j
|
(:
|
) =
I1
j
I
|
(:
|
) is the aggregate income in the economy and c
|
(:
|
) =
I1
c
I
|
(:
|
) is aggregate consumption. In fact, the weights are given by
0
I
=
c
1
o
I
=l
c
1
o
_ 0 (3.16)
Note that with logarithmic preferences (o = 1) we have that 0
I
= c
I
, i.e. the
share of aggregate consumption an agent i is allocated corresponds to the Pareto
weight the planner attaches to this agent.
That is, with separable CRRA utility complete markets imply that individual
consumption at each date, in each state of the world is a constant fraction of
aggregate income (or consumption). Note that it does not imply that individual
consumption is constant across time and states of the world, because it still
varies with aggregate income (a variation against which no mutual insurance
among the agents exists). It also does not imply that consumption among
agents is equalized. From (8.14) we see that the level of consumption of agent
i will depend positively on the Pareto weight of that agent.
Remember from your rst year how to construct competitive equilibria from
solutions to the social planners problem using Negishis (1960) method, and
it will be clear that the individual income processes j
I
|
(:
|
) determine the c =
(c
I
)
I1
corresponding to a (the) competitive equilibrium. Hence the level of
consumption of agent i depends on her income process j
I
|
(:
|
).
Remark 4 In case you dont remember, or havent been taught this, this reamrk
outlines Negishis method. The one element from the equilibrium denition we
havent used yet is the household budget constraint. Thus for given c there is
no reason to believe that each household can aord the corresponding allocation
c
I
|
(:
|
) = 0
I
c
|
(:
|
) =
c
1
o
I
=l
c
1
o
c
|
(:
|
) = c
I
|
(:
|
, c).
But in order to compute how much such an allocation costs we need the ap-
propriate prices. It turns out that the Lagrange multiplies (that is, the shadow
prices) from the social planner problem work. So dene the transfer functions
as
t
I
(c) =
s
t
S
t
`
|
(:
|
, c)
_
c
I
|
(:
|
, c) j
I
|
(:
|
)
.
For our economy the Lagrange multipliers are given by
`
|
(:
|
) = c
I
,
|
|
(:
|
)l
I
c
(c
I
|
(:
|
), :
|
)
= c
I
,
|
|
(:
|
)c
I
|
(:
|
, c)
c
= c
I
,
|
|
(:
|
)(0
I
)
c
c
|
(:
|
)
c
=
_
_

=l
c
1
o
_
_
c
,
|
|
(:
|
)c
|
(:
|
)
c
= `
|
(:
|
, c)
But in a competitive equilibrium households do not receive transfers, so it is
intuitive to nd the equilibrium allocations by nding the cs that solve
t
I
(c) = 0 for all i. (3.17)
Note that since we normalized

c
I
= 1 there are only 1 unknowns. But it
is easy to show that

t
I
(c) = 0, so we only have 1 independent equations
as well. Once we have solved for c
+
from these equations, the equilibrium alloca-
tions are given by c
I
|
(:
|
, c
+
) and equilibrium prices are given by `
|
(:
|
, c
+
). Note
that solving for equilibria then amounts to solving the social planner problem for
arbitrary c and solving a system of 1 equations in 1 unknowns, which
may be substantially easier than solving for the competitive equilibrium directly.
But also note that equations (8.17) can be quite messy, and in particular, can
be highly nonlinear in the cs, so that no analytical solution is available.
After this detour, lets turn back to the characterization of ecient (and
hence equilibrium) allocations. The growth rate of consumption between any
two dates and states is given from (8.1) as
log
_
c
I
|l
(:
|l
)
c
I
|
(:
|
)
_
= log
_
c
|
(:
|l
)
c
|
(:
|
)
_
(3.18)
that is, if agents have CRRA utility that is separable in consumption and we
have complete markets, then individual consumption growth is perfectly cor-
related with and predicted by aggregate consumption growth. In particular,
individual income growth should not help to predict individual consumption
growth once aggregate consumption (income) growth is accounted for. This
idea is the basis of all empirical tests of perfect consumption insurance, see e.g.
Mace (1991), Cochrane (1991), among others.
To obtain Arrow Debreu prices associated with equilibrium allocations we
obtain from the consumer problem of maximizing (8.2) subject to (8.7) that
j
|l
(:
|l
)
j
|
(:
|
)
= ,
|l
(:
|l
)
|
(:
|
)
l
I
c
(c
I
|l
(:
|l
), :
|l
)
l
I
c
(c
I
|
(:
|
), :
|
)
(3.19)
Under assumption 2 this becomes
j
|l
(:
|l
)
j
|
(:
|
)
= ,
|l
(:
|l
)
|
(:
|
)
_
c
I
|l
(:
|l
)
c
I
|
(:
|
)
_
c
(3.20)
= ,
|l
(:
|l
)
|
(:
|
)
_
c
|l
(:
|l
)
c
|
(:
|
)
_
c
(3.21)
and hence equilibrium Arrow Debreu prices can be written as functions of ag-
gregate consumption only. We then have the following
Proposition 5 Suppose allocations (c
I
|
(:
|
))
I1
T
|=0,s
t
S
t
and prices j
|
(:
|
)
T
|=0,s
t
S
t
are an Arrow-Debreu equilibrium. Then under assumption 2 the allocation
c
|
(:
|
)
T
|=0,s
t
S
t
dened by
c
|
(:
|
) =
I1
c
I
|
(:
|
) (3.22)
and prices j
|
(:
|
)
T
|=0,s
t
S
t
is an Arrow Debreu equilibrium for the single agent
economy with i = 1 in which the representative agent has endowment
j
|
(:
|
) =
I1
j
I
|
(:
|
) (3.23)
and CRRA preferences.
1
1
We have assumed idential CRRA utility functions and time discount factors across all
agents. Koulovatianos (2005), building on the large literature on aggregation in dynamic
models, such as Chatterjee (1994) and Caselli and Ventura (2000), gives necessary and su-
cient conditions on individual utility functions to obtain a representative consumer. See his
Theorem 1 and 2.
Maintaining identical time discount factors the period utility function has to be either of
power or exponential form (with heterogeneity in the CARA, but not in the CRRA possible).
If the time discount factor is allowed to vary across agents, only exponential utility gives rise
to aggregation.
Obviously the content of this proposition lies in the pricing part. It shows
that to derive Arrow-Debreu prices (and hence all other asset prices), with
complete markets it is sucient to study the representative agent economy.
This insight is the departure of the consumption-based asset pricing literature
as developed in Lucas (1978). It should also be noted that for this proposition
assumption 2 can be considerably weakened, although the construction of the
utility function of the representative agent becomes more involved.
Asset Pricing
[Add something on pricing simple assets, dene stochastic discount factors etc.]
Sequential Equilibrium
Now let trade take place sequentially in each period (more precisely, in each
period, event-history pair). We introduce one period contingent IOUs, nan-
cial contracts bought in period t, that pay out one unit of the consumption
good in t 1 only for a particular realization of :
|l
= j
tomorrow.
2
So
let
|
(:
|
, :
|l
= j
) denote the price at period t of a contract that pays out

one unit of consumption in period t 1 if (and only if) tomorrows event is
:
|l
= j
. These contracts are often called Arrow securities, contingent claims

or one-period insurance contracts. Let a
I
|l
(:
|
, :
|l
) denote the quantities of
these Arrow securities bought (or sold) at period t by agent i.
The period t, event history :
|
budget constraint of agent i is given by
c
I
|
(:
|
)
st+1
|
(:
|
, :
|l
)a
I
|l
(:
|
, :
|l
) _ j
I
|
(:
|
) a
I
|
(:
|
) (3.24)
Note that agents purchase Arrow securities a
I
|l
(:
|
, :
|l
)
s
t+12:
for all contin-
gencies :
|l
o that can happen tomorrow, but that, once :
|l
is realized,
only the a
I
|l
(:
|l
) corresponding to the particular realization of :
|l
becomes
his asset position with which he starts the current period. We assume that
a
I
0
(:
0
) = 0 for :
0
o.
We then have the following
Denition 6 A SM equilibrium is allocations
_
c
I
|
(:
|
),
_
a
I
|l
(:
|
, :
|l
_
st+1S
_
I1
T
|=0,s
t
S
t
,
and prices for Arrow securities
|
(:
|
, :
|l
)
T
|=0,s
t
S
t
,st+1S
such that
1. For i 1 given
|
(:
|
, :
|l
)
T
|=0,s
t
S
t
,st+1S
, for all i, c
I
|
(:
|
),
_
a
I
|l
(:
|
, :
|l
_
st+1S
T
|=0,s
t
S
t
maximizes (8.2) subject to (8.24) and the constraints c
I
|
(:
|
) _ 0 and a
I
|l
(:
|
, :
|l
) _
I
(:
|l
, , c
I
)
2
A full set of one-period Arrow securities is sucient to make markets sequentially com-
plete, in the sense that any (nonnegative) consumption allocation is attainable with an appro-
priate sequence of Arrow security holdings o
i
t+1
(c
t
, c
t+1
) satisfying all sequential markets
budget constraints.
2.
1
I=l
c
I
|
(:
|
) =
1
I=l
j
I
|
(:
|
) for all t, :
|
o
|
(3.25)
1
I=l
a
I
|l
(:
|
, :
|l
) = 0 for all t, :
|
o
|
and all :
|l
o (3.26)
Note that we have a market clearing condition in the asset market for each
Arrow security being traded for period t 1. Also note that, as usual, we need
a restriction on the amount borrowed to prevent Ponzi schemes. Although we
already solved for the equilibrium in its Arrow-Debreu equilibrium form, to
contrast the complete markets model better with the PILCH model we also
want to solve for the sequential market equilibrium. It turns out that with the
appropriate no Ponzi condition the set of equilibria coincide.
Let us dene the no Ponzi scheme condition, for a given process of Arrow
securities prices
|
(:
|
, :
|l
) as follows
3
: rst dene as date 0, event :
0
(re-
member that we xed that event) present value of one unit of the consumption
good at event history :
|
0
(:
0
) = 1 (3.27)
0
(:
|l
) =
0
(:
|
)
|
(:
|
, :
|l
)
0
(:
0
, :
l
) + . . . +
|
(:
|
, :
|l
) (3.28)
As No Ponzi condition we then impose
0
(:
T
)a
I
T
(:
Tl
, :
T
) _ lim
o
inf o
(:
T
) (3.29)
where
o
(:
T
) =

|=T
s
t
]s
J
0
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
(3.30)
is the present value of future savings up to date . Note that as goes to ,
the sequence of o
(:
T
) need not converge, but if it does, then
lim
o
inf o
(:
T
) =
o
|=T
s
t
]s
J
0
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
(3.31)
is well-dened.
4
3
This discussion follows Wright (1987). For those interested in a generalization of the result
in the presence of long-lived assets, see Huang and Werner (2001).
4
Remember that for any sequence an
1
n=0
the number
a
= lim
n!1
inf an
is the smallest cluster point of the sequence, i.e.
a
= sup&[an < & for at most nite number of as

Intuitively, the no Ponzi scheme condition
a
I
|l
(:
|
, :
|l
) _
I
(:
|l
, , c
I
) (3.32)
=
liminf o
(:
|l
)
0
(:
|l
)
rules out a sequence of asset holdings and consumption for which the present
value of debt could, in some event history, the present discounted value of future
savings.
Before stating and proving the equivalence result, let us further interpret the
no Ponzi scheme condition
0
(:
T
)a
I
T
(:
Tl
, :
T
) _ lim
o
inf

|=T
s
t
]s
J
0
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
(3.33)
Suppose the sequential budget constraint holds with equality. Then
j
I
|
(:
|
) c
I
|
(:
|
) =

st+1
|
(:
|
, :
|l
)a
I
|l
(:
|
, :
|l
) a
I
|
(:
|
) (3.34)
=
|
(:
|
) a
I
|l
(:
|
) a
I
|
(:
|
)
where we dene
|
(:
|
) a
I
|l
(:
|
) =

st+1
|
(:
|
, :
|l
)a
I
|l
(:
|
, :
|l
) (3.35)
But then (8.88) becomes
0
(:
T
)a
I
T
(:
Tl
, :
T
) _ lim
o
inf

|=T
s
t
]s
J
0
(:
|
)
_
|
(:
|
) a
I
|l
(:
|
) a
I
|
(:
|
)
= lim
o
inf
_
0
(:
T
)
T
(:
T
) a
I
Tl
(:
T
)
0
(:
T
)a
I
T
(:
T
)

s
J+1
]s
J
0
(:
Tl
)
Tl
(:
Tl
) a
I
T2
(:
Tl
)
0
(:
Tl
)a
I
Tl
(:
Tl
)
. . .

s
1
]s
J
0
(:
(:
) a
I
l
(:
)
0
(:
)a
I
(:
)
_
_
_
= lim
o
inf
_
_
_
0
(:
T
)a
I
T
(:
T
)

s
1
]s
J
0
(:
(:
) a
I
l
(:
)
_
_
_
(3.36)
Note that if an converges, then
a
= lim
n!1
an
and if an is unbounded below, then
a
= o
The last equality is due to the fact that all intermediate terms cancel out. We
demonstrate this for the rst terms
0
(:
T
)
T
(:
T
) a
I
Tl
(:
T
) =
0
(:
T
)

s
J+1
T
(:
T
, :
Tl
)a
I
Tl
(:
T
, :
Tl
)
(3.37)
s
J+1
]s
J
0
(:
Tl
)a
I
Tl
(:
Tl
) =

s
J+1
]s
J
0
(:
T
)
T
(:
T
, :
Tl
)a
I
Tl
(:
Tl
)
=
0
(:
T
)

s
J+1
T
(:
T
, :
Tl
)a
I
Tl
(:
T
, :
Tl
)
(3.38)
Hence the no Ponzi scheme condition becomes
lim
o
inf
_
_
_
s
1
]s
J
0
(:
(:
) a
I
l
(:
)
_
_
_
_ 0 (3.39)
and thus rules out asset sequences for which the present value of future debt is
bounded above zero. We now have the following
Proposition 7 1. If (c, a, ) is a sequential markets equilibrium, then (c, j)
is an Arrow-Debreu equilibrium where prices are given as
j
|
(:
|
) =
0
(:
|
) (3.40)
2. If (c, j) is an Arrow-Debreu equilibrium, then there exists asset holdings a
such that (c, a, ) is a sequential markets equilibrium, with
|
(:
|
, :
|l
) =
j
|l
(:
|l
)
j
|
(:
|
)
(3.41)
Proof. It is sucient to show that the budget sets described by the Arrow
Debreu budget constraint (8.7) and the sequential markets budget constraint
(8.24) plus the no Ponzi condition (8.20) (or alternatively (8.80)) contain the
same possible consumption choices.
(1) Suppose (c
I
, a
I
) satises (8.24) and (8.20). Then
0
(:
T
)a
I
T
(:
Tl
, :
T
) _ lim
o
inf

|=T
s
t
]s
J
0
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
(3.42)
Using the denition of Arrow Debreu prices and setting T = 0 yields
j
0
(:
0
)a
I
0
(:
0
) _ lim
o
inf

|=0
s
t
]s0
j
|
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
(3.43)
Noting that a
I
0
(:
0
) = 0 for all :
0
we nd that
0 _ lim
o
inf

|=T
s
t
]s0
j
|
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
= lim
o
inf

|=0
s
t
j
|
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
=
o
|=0
s
t
j
|
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
_
(3.44)
and thus c
I
satises the Arrow-Debreu budget constraint.
5
(2) Suppose c
I
satises the Arrow Debreu budget constraint (8.7). We want
to construct asset holdings a such that (c
I
, a) satisfy (8.24) and (8.20). Fix
and dene
a
(:
) = 0 for all :
(3.45)
and recursively
a
(:
|
) = c
I
|
(:
|
) j
I
|
(:
|
)
st+1
|
(:
|
, :
|l
)a
(:
|
, :
|l
) (3.46)
Multiplying by
0
(:
|
) yields
0
(:
|
)a
(:
|
) =
0
(:
|
)
_
c
I
|
(:
|
) j
I
|
(:
|
)
st+1
0
(:
|
)
|
(:
|
, :
|l
)a
(:
|
, :
|l
)
=
0
(:
|
)
_
c
I
|
(:
|
) j
I
|
(:
|
)

s
t+1
]s
t
0
(:
|l
)a
(:
|
, :
|l
)
=
0
(:
|
)
_
c
I
|
(:
|
) j
I
|
(:
|
)

s
t+1
]s
t
0
(:
|l
)
_
c
I
|l
(:
|l
) j
I
|l
(:
|l
)
0
(:
|l
)
|
(:
|
) a
(:
|
)
(3.47)
With a
(:
) = 0 for all :
we have, via continuing substitution
0
(:
T
)a
(:
T
) =

|=T
s
t
]s
J
0
(:
|
)
_
c
I
|
(:
|
) j
I
|
(:
|
)
(3.48)
5
The last equality is valid whenever the series converges. One can circumvent the problem
of possible nonconvergence by adjusting the denition of Arrow Debreu equilibrium: one
denes the Arrow debreu budget constraint as
j
0
(c
i
j
i
) 0
where
j
0
c = lim
T!1
inf
T
t=0
s
t
jt(c
t
)ct(c
t
)
Then all arguments go through even if, for a given price process j, the innite sum does not
converge for some process c. For details see Wright (1987), in particular footnote 6.
But for each :
|
, a
(:
|
) is a sequence in . Now dene
a
|
(:
|
) = lim
o
inf a
(:
|
) (3.49)
Taking liminf
t
s on both sides of (8.46) yields
a
|
(:
|
) = c
I
|
(:
|
) j
I
|
(:
|
)
st+1
|
(:
|
, :
|l
)a
|l
(:
|
, :
|l
) (3.50)
and hence the sequential budget constraint. Taking liminf
t
s on both sides of
(8.48) yields
0
(:
T
)a
T
(:
Tl
, :
T
) = lim
o
inf

|=T
s
t
]s
J
0
(:
|
)
_
j
I
|
(:
|
) c
I
|
(:
|
)
(3.51)
and hence (c
I
, a) as constructed above satises the no Ponzi scheme condition.
This proposition shows that each Arrow-Debreu equilibrium can be imple-
mented as a sequential markets equilibrium with a full set of one-period ahead
Arrow securities only (and vice versa, each sequential markets equilibrium is
implementable as an Arrow Debreu equilibrium).
The no Ponzi scheme condition (8.20) may be hard to check as it requires
knowledge of either asset holdings in the time limit or knowledge of the en-
tire future consumption plan (or both), in addition to prices. An alternative,
equivalent no Ponzi condition (equivalent in the sense that exactly the same
proposition as before can be proved) is
0
(:
T
)a
T
(:
Tl
, :
T
) _ \(:
T
) (3.52)
where
\(:
T
) =
o
|=T
s
t
]s
J
0
(:
|
)j
I
|
(:
|
) (3.53)
is the Arrow-Debreu future wealth of the agent.
6
The no Ponzi condition, stated
this way, is sometimes called the natural debt limit. It is easier to check as it
only involves the endowment process and prices (either Arrow securities prices
or Arrow Debreu prices).
Suppose the stochastic process :
|
is Markov, and let (:
t
[:) denote the
transition probabilities of the Markov chain. Then the consumer problem, un-
der the assumption that Arrow security prices take the form
7
|
(:
|
, :
|l
) =
(:
|l
[:
|
), can be written recursively as
(a, :) = max
]o
0
(s
0
)]
s
0
2:
_
l
_
j(:) a
s
0
a
t
(:
t
)(:
t
[:)
_
,
s
0
(:
t
[:)(a
t
(:
t
), :
t
)
_
(3.54)
6
We again assume that this sum converges.
7
Under assumption 2 and the assumption that ct is Markov, this assumption is valid.
The rst order and envelope condition are
(:
t
[:)l
t
_
j a
s
0
a
t
(:
t
)(:
t
[:)
_
= ,(:
t
[:)
o
(a
t
(:
t
), :
t
)
o
(a, :) = l
t
_
j a
s
0
a
t
(:
t
)(:
t
[:)
_
(3.55)
Combining yields a recursive version of the Euler equation, to be solved for the
policy function a
t
(a, :; :
t
)
(:
t
[:)l
t
_
j(:) a
s
0
a
t
(a, :; :
t
)(:
t
[:)
_
= ,(:
t
[:)l
t
_
j(:
t
) a
t
(a, :; :
t
)
s
00
a
t
(a
t
(a, :; :
t
), :
t
; :
tt
)(:
tt
[:
t
)
_
We will use these equations in our discussions of complete markets models with
informational and/or enforcement frictions in later chapters of these notes.
3.1.2 Empirical Tests of the Complete Consumption In-
surance Hypothesis
Some of the most important empirical papers that try to assess whether the
complete consumption insurance hypothesis is a good description of the data are
Mace (1991), Cochrane (1991), Townsend (1994), Attanasio and Davis (1996)
and Hayashi, Altonji and Kotliko (1996).
Derivation of Maces Empirical Specications
The paper by Mace presents the most straightforward test of the complete con-
sumption insurance hypothesis. The main implication of perfect risk sharing is
that individual consumption should respond to aggregate income (or consump-
tion) shocks, but not to idiosyncratic income shocks, no matter whether these
idiosyncratic shocks are unanticipated or not anticipated (this is important in
Using (3.41) and (3.21) we nd
q(c
t
, c
t+1
) =
j
t+1
(c
t+1
)
jt(c
t
)
= o
t+1
(c
t+1
)
t(c
t
)
_
c
t+1
(c
t+1
)
ct(c
t
)
_
= o(c
t+1
[ct)
_
j
t+1
(c
t+1
)
jt(ct)
_
= q(c
t+1
[ct)
discriminating the complete markets model from the PILCH type models to be
discussed next).
Under assumptions 1 and 2, equation (8.18) implies that
^lnc
I
|
= ^lnc
|
(3.56)
that is, the growth rate of individual consumption ^lnc
I
|
= lnc
I
|
(:
|
)lnc
I
|l
(:
|l
) -
c
1
t
c
1
t1
c
1
t1
should equal the growth rate of aggregate consumption ^lnc
|
, but be
independent of individual income growth or other individual-specic shocks.
That is, in the regression
^lnc
I
|
= c
l
^lnc
|
c
2
^lnj
I
|
c
I
|
(3.57)
under the null hypothesis of complete consumption insurance c
l
= 1 and c
2
= 0.
In this regression the error term c
I
|
captures (thus far unmodeled) individual
preference shocks as well as measurement error in income and or consumption
growth.
We can generalize assumption 2 and still arrive at an empirical specica-
tion similar to equation (8.7) that can be used to test for complete insurance.
Suppose we assume that preferences are of the form
l
I
(c
I
|
(:
|
), :
|
) = c
b
1
(s
t
)
c
I
|
(:
|
)
lc
1
1 o
(3.58)
where /
I
(:
|
) capture preference shocks such as changes in family size (which are
taken as exogenous at this point). Taking rst order conditions in the planners
problem and then taking logs of these conditions yields
ln(c
I
|
(:
|
)) =
1
o
/
I
(:
|
)
1
o
ln(c
I
)
1
o
ln
_
`(:
|
)
,
|
|
(:
|
)
_
(3.59)
Taking averages of (8.0) over all agents yields
1
ln(c
|
(:
|
)) =
1
o
(:
|
)
1
o
ln(c
)
1
o
ln
_
`(:
|
)
,
|
|
(:
|
)
_
(3.60)
and substituting back yields
ln(c
I
|
(:
|
)) =
1
o
_
_
/
I
(:
|
)
1
(:
|
)
_
_
1
o
_
_
ln(c
I
)
1
ln(c
)
_
_
ln(c
|
(:
|
))
=
1
o
_
/
I
(:
|
) /
o
(:
|
)
_
1
o
_
ln(c
I
) ln(c
o
)
_
ln(c
o
|
(:
|
)) (3.61)
where
/
o
(:
|
) =
1
(:
|
)
ln(c
o
) =
1
ln(c
)
ln(c
o
|
(:
|
)) =
1
ln(c
|
(:
|
)) (3.62)
are population averages (note that the last two expressions are sums of logs
rather than logs of sums). Finally, taking rst dierences of (8.61) yields (again
suppressing dependence on :
|
)
^ln(c
I
|
) = ^ln(c
o
|
)
1
o
_
^/
I
|
^/
o
|
_
(3.63)
Thus, with this specication of preferences we can test the complete insurance
hypothesis with the empirical specication
^ln(c
I
|
) = c
l
^ln(c
o
|
) c
2
^ln(j
I
|
) c
I
|
(3.64)
where the error c
I
|
now captures individual and aggregate changes in preference
shocks as well as potentially measurement error.
8
Under the null hypothesis of
complete insurance we have c
l
= 1 and c
2
= 0. Thus we can carry out the same
test of complete consumption insurance as before, keeping in mind that the log
of average consumption c
o
|
is dened slightly dierent form ln(c
|
).
Finally, we can derive a similar specication under the assumption that
preferences take Constant Absolute Risk Aversion (CARA) form. If
l
I
(c
I
|
(:
|
), :
|
) =
1
c
~(c
1
t
(s
t
)b
1
(s
t
))
(3.65)
where is the coecient of absolute risk aversion, then, using the same manip-
ulations as before one obtains
^c
I
|
= ^c
n
|

_
^/
I
|
^/
n
|
_
(3.66)
where
c
n
|
=
1
c
I
|
/
n
|
=
1
/
I
|
(3.67)
8
Note that, if we dene
c
i
t
= b
i
t
b
a
t
then the expectation of c
i
t
is zero by denition of b
a
t
.
Thus, under exponential utility the empirical specication used to test complete
insurance is
^c
I
|
= c
l
^c
n
|
c
2
^j
I
|
c
I
|
(3.68)
with the null hypothesis c
l
= 1 and c
2
= 0.
Note that, instead of individual income growth other variables capturing
idiosyncratic risk, such as changes in employment status could be used (and
were used by Mace (1991) as well as Cochrane (1991) in their tests for complete
insurance.
Results of the Tests
Mace uses CEX data from 1980 to 1983 to estimate equations (8.64) and (8.68).
We focus our discussion on the specications in which individual income growth
is used as measure of idiosyncratic risk, rather than changes in unemployment
status.
In order to implement the regressions household level data on consumption
and income are needed. For each household, since a specication in rst dier-
ences is employed (in order to remove the individual xed eects), we require
at least two observations on both income and consumption. Thus, although
the CEX has no deep panel dimension, the fact that it contains income and
consumption data for four quarters for each household makes this data set an
appropriate (if not ideal) source for Maces application.
The total number of households in the CEX sample that Mace employs is
10, 60. Each household contributes one dierenced observation; Mace chooses
to use the information from the rst and the forth interview of each house-
hold (remember that each household stays in the sample for only 4 quarters)
to obtain dierenced consumption and income observations. For the growth
rate specication households that report nonpositive income or consumption in
any of the two quarters have to be excluded, so that the sample size for this
specication is smaller.
Mace employs several measures of consumption in her analysis; we will report
results for total consumption expenditures, for nondurables, for services and for
food expenditures. The income variable is disposable income, dened as before-
tax income minus income taxes, deductions for social security and pension plans
as well as occupational expenses (e.g. union dues).
In Table 1 we present the results from the estimation of equation (8.64),
where Mace included a constant to the regression. The rst three columns
present the OLS estimates (with standard errors in parentheses) for the con-
stant, the coecient on aggregate consumption growth and on individual income
growth. The forth column gives the test statistic of an F-test
9
of the joint hy-
9
The test statistic for the F-test is distributed as an 1(2, . 3), where . is the number
of usable household observations in the sample. For the dierence specication . = 10, 695
whereas for the growth rate specication it varies by consumption category because households
with nonpositive consumption in either interview have to be excluded.
A + after the test statistic indicates that the null hypothesis can be rejected at the 95%
condence level.
pothesis that c
l
= 1 and c
2
= 0 and the last column shows the 1
2
of the
regression.
Table 1: Growth Rate Specication
Cons. Measure c
0
c
l
c
2
F-test 1
2
Total Consumption
0.04
(0.01)
1.06
(0.08)
0.04
(0.007)
14.12
+
0.021
Nondurables
0.02
(0.01)
0.07
(0.07)
0.04
(0.006)
22.60
+
0.027
Services
0.02
(0.01)
0.08
(0.10)
0.04
(0.01)
12.44
+
0.011
Food
0.02
(0.01)
0.01
(0.07)
0.04
(0.006)
18.67
+
0.020
The results in Table 1 are, for the most part, inconsistent with the com-
plete insurance hypothesis. For all consumption groups the F-test is decisively
rejected at the / condence level (in fact it would be rejected at the 1/ con-
dence level). This is mostly due to the fact that individual income growth
does help to explain individual consumption growth, even when aggregate con-
sumption growth is accounted for. Note that for all consumption measures,
based on a simple t-test the null hypothesis that c
l
= 1 cannot be rejected at
conventional condence levels, but the hypothesis that c
2
= 0 can be soundly
rejected for all consumption measures. Also note that the constant is on the
edge of being signicant for most specications and that the 1
2
is quite low for
all specications.
10
In Table 2 we report the results for the specication relying on CARA utility,
i.e. the estimates of equation (8.68), again including an intercept.
Table 2: First Dierence Specication
Cons. Measure c
0
c
l
c
2
F-test 1
2
Total Consumption
77.87
(10.82)
1.06
(0.11)
0.08
(0.02)
1.27 0.008
Nondurables
18.07
(8.88)
0.00
(0.06)
0.01
(0.008)
7.71
+
0.028
Services
80.47
(16.47)
1.01
(0.10)
0.01
(0.007)
1.14 0.000
Food
7.46
(2.12)
1.01
(0.08)
0.00
(0.002)
2.2 0.020
For this specication the results are more supportive of perfect consumption
smoothing. In particular, the F-test is rejected only for nondurable consump-
tion expenditures, again primarily due to the fact that individual consumption
growth is sensitive to individual income growth. Note that here the intercept is
signicant for almost all measures of consumption (in theory it should be zero).
10
Remember, though, that we estimate an equation in rst dierences (of logs).
It is these results that lead Mace to the conclusion that perfect risk sharing is
a good rst description of the data, at least for exponential utility.
However, as Nelson (1994) argues, if one excludes households classied as
incomplete income reporters by the CEX (households that either did not
respond to certain income questions or gave inconsistent answers) and uses
as consumption measures consumption expenditures for the entire quarter (as
opposed to Mace who used only the expenditure in the month preceding the
interview) one nds a strong rejection of complete consumption insurance even
for the CARA utility specication. Thus, it is somewhat surprising that the
literature tends to cite Maces paper as supporting the complete risk sharing
hypothesis.
The Problem of Measurement Error
Mace (1991) discusses several econometric issues arising in her regression analy-
sis. Somewhat surprisingly the paper does not include an extended discussion
of measurement error, which is a severe issue in micro data sets in general and
for the CEX in particular. Given the survey design of the CEX measurement
error in the income variable is most likely the biggest concern; as we will show
it may bias the test of perfect risk sharing in favor of the null hypothesis of
perfect consumption insurance.
To make things simple let us suppose we want to estimate the equation
^c
I
|
= c
2
^j
I
|
c
I
|
(3.69)
i.e. for the moment we ignore the presence of aggregate consumption growth.
Note that if we have a single cross-section of households (e.g. a single quarter
of data), this would be the regression we would run (see Cochrane (1991) for a
discussion why the pooling of cross-sectional and time series data is not free of
problems and a test of complete risk sharing using purely cross-sectional data).
The null hypothesis of perfect risk sharing implies c
2
= 0. For ease of exposition
we assume c
0
= 0.
Suppose we have household observations for one period t, and suppose
that the income variable is measured with error
^.
I
|
= ^j
I
|
^
I
|
(3.70)
where ^.
I
|
is the measured income change, ^j
I
|
is the true income change and
^
I
|
is an additive measurement error satisfying 1
_
^
I
|
_
= 0, where the expec-
tation is the cross-sectional (across households) expectation; remember that we
keep t xed. Furthermore assume that \ ar
_
^
I
|
_
= o
2
u
, that \ ar
_
^j
I
|
_
= o
2
and that ^
I
|
, c
I
|
and ^j
I
|
are all mutually independent (the assumption that c
I
|
and ^j
I
|
are independent is needed already to make the OLS estimate for c
2
consistent even in the absence of measurement error).
The equation we estimate is then
^c
I
|
= c
2
^.
I
|
c
I
|
(3.71)
and the OLS estimate for c
2
is
c
2
=

l
=l
^c
|
^.
=l
_
^.
|
_
2
(3.72)
We now show that c
2
is an inconsistent estimator of c
2
, with persistent bias
towards zero. We have that, multiplying out and using Slutskys theorem
plim
o
c
2
= plim
o
=l
^c
|
^.
=l
_
^.
|
_
2
= plim
o
=l
_
c
2
^j
I
|
c
I
|
_
_
^j
|
^
|
_
=l
_
^j
|
^
|
_
2
= c
2
plim
o
=l
_
^j
I
|
_
2
plim
o
=l
_
^j
|
_
2
plim
o
=l
_
^
|
_
2
= c
2
o
2
o
2
o
2
u
= c
2
1
1
c
2
r
c
2
(3.73)
Thus, without measurement error (o
2
u
= 0) c
2
estimates c
2
consistently; the
higher the ratio between measurement noise and signal,
c
2
r
c
2
the more is the

estimator c
2
asymptotically biased towards zero; in the presence of substantial
measurement error of income we may not reject the null of perfect risk sharing
even when it should be rejected had we observed the data without measurement
error.
This discussion is meant to give an indication for why even the results in
Mace supporting perfect risk sharing should be taken with caution; they are also
meant to motivate similar analyses like the one in Cochrane (1991) that focus on
measures of idiosyncratic income risk which may be less prone to measurement
error and correlation with the error term than changes in or growth rates of
individual income.
3.2 Permanent Income-Life Cycle Hypothesis (PILCH)
Models
Permanent income type models assume that agents do not have access to a
complete set of contingent consumption claims. The main dierence between
this type of models and the complete markets model thus manifests itself in
the budget constraints. Whereas the complete markets model in sequential
formulation has budget constraints of the form (8.24), for PILCH models we
have
c
|
(:
|
)
|
(:
|
)a
|l
(:
|
) = j
|
(:
|
) a
|
(:
|l
) (3.74)
Here
|
(:
|
) is the price at date t, event history :
|
, of one unit of consumption
delivered in period t 1 regardless of what event :
|l
is realized. Notation
is crucial here: a
|l
is a risk-free IOU and is a function of :
|
and not of :
|l
whereas in the complete markets model with a full set of Arrow securities these
assets are indexed by (:
|
, :
|l
) = :
|l
and each asset pays o only at a particular
event history :
|l
. Obviously each of these assets has a potentially dierent price
|
(:
|
, :
|l
)
The basic income uctuation problem is to maximize (8.2) subject to (8.74)
and a short-sale constraint on assets which rules out Ponzi schemes, but is
assumed to be suciently wide so as to never bind at the optimal allocation.
Let the initial wealth of an agent be denoted by a
0
(:
l
) = a
l
. At this point
we make no assumptions about a
l
.
3.2.1 A Simple 2 Period Toy Model
To gain some intuition and dene some concepts, let us look at the simplest
example of a life cycle model, in which T = 2 and there is no uncertainty. The
agents problem becomes
max
c1,c2,o1
l(c
l
) ,l(c
2
) (3.75)
s.t.
c
l
a
l
= j
l
(3.76)
c
2
= j
2
a
l
(3.77)
c
l
, c
2
_ 0 (3.78)
One can consolidate the budget constraints to
c
l

c
2
1 r
= j
l

j
2
1 r
(3.79)
where 1 r =
l
j
is the gross real interest rate. Equation (8.70) together with
the Euler equation
l
t
(c
l
) = ,(1 r)l
t
(c
2
) (3.80)
uniquely determine the optimal consumption allocation over the life cycle. An
increase in the interest rate r has three eects on consumption allocations
3.2. PERMANENT INCOME-LIFE CYCLE HYPOTHESIS (PILCH) MODELS35
1. An increase in r makes period 2 consumption relatively less expensive
compared to period 1 consumption and hence reduces current consumption
and increases saving and future consumption. This is the substitution
eect.
2. An increase in r reduces the price of second period consumption in absolute
terms, acting like an increase in income. This is the income eect.
3. Provided that the household has income in the second period, an increase
in r reduces the present value of lifetime income and hence reduces current
consumption. This is the human capital eect.
Which eect dominates depends on the particular form of the utility function
and the income prole. In macroeconomics we often use C11 period utility.
At this point it is useful to discuss its basic properties.
Remark 8 Constant relative risk aversion utility functions l(c) =
c
1o
l
lc
have the following important properties.
11
First, dene as o(c) =
I
00
(c)c
I
0
(c)
the
(Arrow-Pratt) coecient of relative risk aversion. Hence o(c) indicates a house-
holds attitude towards risk, with higher o(c) representing higher risk aversion.
For CRRA utility functions o(c) is constant for all levels of consumption, and
for l(c) = ln(c) it is not only constant, but equal to o(c) = o = 1. Second, the
intertemporal elasticity of substitution i:
|
(c
2
, c
l
) measures by how many percent
the relative demand for consumption in period 2, relative to demand for con-
sumption in period 1,
c2
c1
declines as the relative price of consumption in 2 to
consumption in 1, =
l
l:
changes by one percent. Formally
i:(c
2
, c
l
) =
_
J(
c
2
c1
)
c
2
c
1
_
_
J
1
1+r
1
1+r
_ =
_
J(
c
2
c1
)
J
1
1+r
_
_
c
2
c1
1
1+r
_
But from (8.80) we have that
,l
t
(c
2
)
l
t
(c
l
)
=
1
1 r
which, for l(c) =
c
1o
l
lc
becomes
c
2
c
l
=
_
1
,(1 r)
_
1
o
11
CRRA utility function belong to the general class of hyperbolic absolute risk aversion
utility functions, given by the general form
l(c) =
1 j
j
_
cc
1 j
+.
_
where j, c, . are parameters. It is easy to show that, up to irrelavant constants, CRRA is a

special case (as are CARA and quadratic utility functions).
and thus
d
_
c2
c1
_
d
l
l:
=
1
,o
_
1
,(1 r)
_
1
o
l
and therefore
i:(c
2
, c
l
) =
_
J(
c
2
c1
)
J
1
1+r
_
_
c
2
c
1
1
1+r
_ =

l
oc
_
l
o(l:)
_
1
o
l
(1 r) +
_
l
o(l:)
_
1
o
=
1
o
Therefore for C11 utility both risk aversion (that is, attitudes towards risk)
as well as intertemporal elasticity of substitution (that is, attitudes towards con-
sumption over time) are governed by exactly the same parameter. This turns
out to be very problematic for asset prices (see Mehra and Prescotts (1985) eq-
uity premium puzzle). To explain the large excess returns on stocks over bonds
one requires a large risk aversion o. But with C11 preferences this necessar-
ily implies low intertemporal elasticity of substitution; that is, households really
like at consumption proles. In a growing representative agent economy, in or-
der for households to tolerate a growing consumption prole then requires very
large risk-free interest rates, much larger than the ones actually observed (this
is the so-called risk-free rate puzzle). There are thousands of papers trying to
jointly explain the equity premium and the risk-free rate puzzle. One attempt
is to disconnect the link between risk aversion and intertemporal elasticity of
substitution in preferences by letting them be independently controlled by two
parameters. Epstein and Zin (1989) postulated the following recursive prefer-
ences. These preferences are, of course, only useful when there is uncertainty,
so let us assume that second period consumption is stochastic (say, because send
period income is).
12
Then recursive preferences are given by
l(c
l
, c
2
(:)) =
_
_
[c
l
[
l
1
,
_
s
(:) [c
2
(:)[
lc
_
1
1
1o
_
_
1
1
1
It is straightforward to see that for o = this formulation coincides with stan-

dard C11 preferences. Also, measures the intertemporal elasticity of substi-
12
For an innite horizon economy, recursive preferences over a consumption allocation
c = ct(c
t
) are dened as follows: let \ (c, c
t
) denote the expected lifetime utility from
consumption allocation c, from node c
t
onwards. Then recursive preferences are given by
\ (c, c
t
) =
_
_
_
_
_
_
_
_
_
_
ct(c
t
)
_
1
1
+o
_
_
s
t+1
(c
t+1
[c
t
)
_
\ (c, c
t+1
)
_
1
_
_
1
1
1o
_
_
_
_
_
_
_
_
_
1
1
1
The recursive nature of the denition of lifetime utility gives these preferences its name. It
should be clear (and we will discuss this later) that recursive preferences lend itself very well
to recursive numerical methods.
tution (if there were no uncertainty in the second period, equals the IES) and
o measures risk aversion (if there were no rst period, then o is the coecient
of relative risk aversion on second period consumption gambles).
After this digression we now discuss the optimal solution to the two-period
problem for the case of C11. It is given by
c
l
=
_
1 ,
1
o
(1 r)
1
o
l
_
l
_
j
l

j
2
1 r
_
(3.81)
Hence
0c
l
0(1 r)
=
j
2
(1 r)
2
_
1 ,
1
o
(1 r)
1
o
l
_
l
_
j
l

j
2
1 r
_
_
1 ,
1
o
(1 r)
1
o
l
_
l
_
,
1
o
_
1
o
1
_
(1 r)
1
o
2
_
(3.82)
=
_
1 ,
1
o
(1 r)
1
o
l
_
l
_
j
2
(1 r)
2

_
1
o
1
__
j
l

j
2
1 r
_
_
,
1
o
(1 r)
1
o
2
_
_
The rst term in the rst expression the human capital eect, the second term is
the combination between income and substitution eect. We immediately have
the following results
1. If o = 1 (log-case, iso-elastic utility), income and substitution eect can-
cel out and current consumption declines because of the human capital
eect. If j
2
= 0, then an increase in the interest rate has no eect on the
intertemporal consumption allocation
2. If 0 < o < 1 (high intertemporal elasticity of substitution), then the sub-
stitution eect dominates the income eect and hence current consumption
declines as reaction to an increase in the interest rate
3. If o 1 (low intertemporal elasticity of substitution), then the income
eect dominates the substitution eect and the total eect of an increase
in the interest rate on current consumption is ambiguous.
It is also obvious that in this simple model an increase in current income (as
well as in future income) will increase both current and future consumption; the
magnitude of this increase, in this simple model, depends only on the interest
rate and the time discount factor. With stochastic income, the magnitude of
the increase in consumption due to an increase in current income also crucially
depends on how current income aects expectations about future income.
3.2.2 Many Periods and Certainty Equivalence
Now consider the problem of maximizing (8.2) subject to (8.74). For this sub-
section we impose
Assumption 3: The price of a one period bond =
l
l:
is nonstochastic
and constant over time.
Nonstochastic Income
First assume that the income process of the individual is nonstochastic and given
by j
|
T
|=0
. Also assume that :
|
potentially aecting individuals preferences
is a deterministic sequence. When T = we impose a short sale constraint
on IOUs that prevents Ponzi schemes, but is loose enough to allow optimal
consumption smoothing.
13
Dene as
\
0
= a
l

o
|=0
j
|
(1 r)
|
(3.84)
the present value of all future labor income, including initial wealth. We assume
that \
0
is nite (otherwise the households problem does not have a solution).
Forming the Lagrangian, taking rst order conditions and combining yields the
standard Euler equation
l
c
(c
|
, :
|
) = ,(1 r)l
c
(c
|l
, :
|l
) (3.85)
and hence
l
c
(c
|
, :
|
) =
_
1 j
1 r
_
|
l
c
(c
0
, :
0
) (3.86)
Several Implications of the rst order condition immediately arise
1. Suppose that j = r, so that the right hand side of (8.86) is constant over
time. In periods in which the preference shock :
|
makes marginal utility
high, consumption also has to be high, since marginal utility is decreasing
in consumption. This may provide us with a theory of some life-cycle
phenomena of consumption
(a) If l is separable between consumption and :
|
then we immediately
obtain that household consumption is constant over the life cycle,
c
|
= c
|l
(3.87)
something that seems counterfactual in light of the discussion in the
last chapter. Under the separability assumption, even if j ,= r,
consumption does not display a life-cycle hump shape, but rather
monotonically trends upwards (if r j, i.e. if incentives to postpone
consumption dominate impatience) or monotonically downwards (if
j r, i.e. if impatience dominates incentives to postpone consump-
tion). For more realistic assumptions on the structure of preference
shocks this need not be the case, however.
13
A shortsale constraint of the form
o
t+1
sup
t
1
=t+1
j
(1 +v)
t
(3.83)
would accomplish this, provided that the right hand side is nite, which we shall assume. A
sucient condition for this is that v 0 and the sequence jt to be bounded above.
(b) With more people in the household, an increase in household con-
sumption expenditures yields higher marginal utility. Consider a
simple example: suppose that :
|
represents the number of people in
the household, and suppose that the period utility function takes the
form
l(c
|
, :
|
) =
_
ct
st
_
lc
1 o
with o 1 (3.88)
i.e. period utility depends on per-capita consumption expenditure
(remember that c
|
is household consumption expenditure) and fea-
tures low intertemporal elasticity of substitution
l
c
< 1. The Euler
equation reads as
c
|
c
0
=
_
:
|
:
0
_
l
1
o
(3.89)
As family size increases (i.e. :
|
:
0
), so will optimal household con-
sumption. We saw in Section 2 that consumption is hump-shaped
over the life cycle. With hump-shaped family size, this example pro-
vides the starting point for a simple explanation of this observa-
tion, investigated further in Attanasio et. al. (1995) and Fernandez-
Villaverde and Krueger (2001).
(c) Suppose :
|
reects the amount worked in period t, taken to be ex-
ogenous for now. If the marginal utility from consumption is high
when agents work a lot (that is, consumption and leisure are substi-
tutes), then consumption tends to be high when labor supply is high
(this point was rst highlighted in Heckman (1974)). Since hours
worked tend to be characterized by a hump-shape over the life cy-
cle, so should consumption, if more hours worked indeed increase the
marginal utility from consumption. Again a simple example is given
by
l(c
|
, :
|
) =
_
c
~
|
(1 :
|
)
l~
_
lc
1
1 o
with o 1 and (0, 1)
(3.90)
where is a share parameter measuring the importance of consump-
tion relative to leisure 1 :
|
in the utility function. The Euler equa-
tion becomes
c
|
c
0
=
_
1 :
|
1 :
0
_
(1)(1o)
1+o
(3.91)
Thus, in periods where labor supply :
|
is high, so should consumption
be. Also note that since, at retirement, the number of hours worked
decreases sharply, consumption should fall at retirement as it does
in the data. Again, this argument is the starting point for potential
explanations for why consumption drops at retirement; see Heathcote
(2001) for further details.
2. A decline in the interest rate makes the life cycle consumption prole less
steep, as current consumption increases relative to future consumption,
ceteris paribus (i.e. for unchanged structure of preference shocks :
|
). The
reverse is true for an increase in the interest rate. Note that this is not
a statement about the level of the consumption prole, because this level
is inuenced by income and human capital eects as in the simple two
period model. It is simply a statement about the life cycle consumption
prole.
Before making the income process stochastic we note that under assumption
2 (CRRA separable utility) consumption is given by
c
0
=
1
1
Tl
\
0
(3.92)
c
|
=
_
1 r
1 j
_t
o
_
1
1
Tl
_
\
0
= '1C(t, T) + \
0
(3.93)
where =
l
l:
_
l:
l
_1
o
is assumed to be less than 1. We nd that for given
lifetime income \
0
1. an increase in the lifetime horizon T reduces the marginal propensities to
consume out of lifetime wealth '1C(t, T) at all ages t
2. if j = r, then '1C(t, T) is constant across age, as is consumption. Fur-
thermore =
l
l:
and hence '1C(t, T) =
:
l:(
1
1+r
)
J
. If T = ,
'1C(t, ) =
:
l:
, and consumption in each date equals permanent in-
come
:V0
l:
. The last statement is the purest version of the permanent
income hypothesis: consumption at each date should equal permanent
income.
Stochastic Income and Quadratic Preferences
Now let us consider maximizing (8.2) subject to (8.74), with income fully sto-
chastic. Again, to make the problem well-dened for T = we need to impose
some no Ponzi condition, which, for now, we take as
a
|l
(:
|
) _
o
r=|l
s
:+1
s
t
rl
(:
rl
)
(1 r)
r(|l)
j
rl
(:
rl
) (3.94)
Attaching Lagrange multiplier `(:
|
) to the event history :
|
budget constraint
and taking rst order conditions with respect to c
|
(:
|
) and c
|l
(:
|l
) yields
,
|
|
(:
|
)l
c
(c
|
(:
|
), :
|
) = `(:
|
) (3.95)
,
|l
|l
(:
|l
)l
c
(c
|l
(:
|l
), :
|l
) = `(:
|l
) (3.96)
Taking rst order conditions with respect to a
|l
(:
|
) yields
`(:
|
) =

s
t+1
]s
t
`(:
|l
) (3.97)
Combining yields
,
|
|
(:
|
)l
c
(c
|
(:
|
), :
|
) =
1
s
t+1
]s
t
,
|l
|l
(:
|l
)l
c
(c
|l
(:
|l
), :
|l
) (3.98)
or
l
c
(c
|
(:
|
), :
|
) =
_
1 r
1 j
_

s
t+1
]s
t
|l
(:
|l
[:
|
)l
c
(c
|l
(:
|l
), :
|l
)
=
_
1 r
1 j
_
1
_
l
c
(c
|l
(:
|l
), :
|l
)[:
|
_
(3.99)
or, more compactly,
l
c
(c
|
, :
|
) =
_
1 r
1 j
_
1
|
_
l
c
(c
|l
, :
|l
)
_
(3.100)
where 1
|
is the expectation of :
|l
conditional on :
|
. This is the standard
stochastic Euler equation for PILCH type models. We immediately see that if
j = r, then the marginal utility of consumption follows a martingale, i.e. its
period t expectation of the t 1 variable equals the t variable.
Note that one would, under the assumption
14
that
|
(:
|l
) =
tt+1(s
t+1
]s
t
)
l:
obtain Euler equations of the form
l
c
(c
|
(:
|
), :
|
) =
_
1 r
1 j
_
l
c
(c
|l
(:
|l
), :
|l
) for all :
|l
(3.101)
for the complete markets model, i.e. in that model the Euler equation holds state
by state :
|l
, whereas in the PILCH type model it only holds in (conditional)
expectation. This is again due to the fact that in the complete markets model
there is a full set of contingent claims available, whereas in the PILCH type
models assets are usually restricted to a single uncontingent bond.
Now we make further assumptions to derive Halls Martingale Hypothesis.
Assumption 4: The period utility function is quadratic and separable in
consumption
l(c
|
(:
|
), :
|
) =
1
2
_
c
|
(:
|
) c
_
2
(:
|
) (3.102)
14
We will soon construct general equilibrium versions of both models in which the assump-
tions about the Arrow securities prices (or the price of the uncontingent consumption claim)
are in fact warranted.
where c is the bliss level of consumption, assumed to be large relative to an
agents stochastic income process.
15
Under assumptions 3 and 4 the Euler equation (8.100) then becomes
1
|
c
|l
= c
l
c
2
c
|
(3.103)
with c
l
= c
_
1
l
l:
_
and c
2
=
l
l:
. With j = r, the previous equation
becomes
1
|
c
|l
= c
|
(3.104)
i.e. consumption itself follows a martingale. We summarize the most important
implications of (8.108) or (8.104) as:
1. The agents optimal consumption decisions obey certainty equivalence;
comparing the rules for optimally allocating consumption over time in the
certainty and the uncertainty case (8.87) and (8.104), we see that they are
identical (of course the realized consumption path in the uncertainty case
will ex post dier from the consumption path without uncertainty).
2. Consumption should obey the regression
c
|l
= c
l
c
2
c
|
n
|l
(3.105)
where n
|l
is a random variable satisfying 1
|
n
|l
= 0. The main impli-
cation of the PILCH model in its standard, certainty equivalence form is
that period t 1 consumption c
|l
is perfectly predicted by period t con-
sumption c
|
and no other variables that are in the households information
set at period t should help predict it. In particular, once c
|
is included
in the regression, current income j
|
, current assets a
|
, past consumption
c
|l
or other variables should not enter with a signicant coecient.
3. The income realization in period j
|l
does aect c
|l
, but only that part
that is an unexpected change from income in period t. The component of
income j
|l
that is already predicable by j
|
or any other variable that is
in the period t information set should not aect consumption in period t.
Note that with complete markets not even unexpected changes in income
should aect consumption, since agents can perfectly insure against these
unexpected changes.
15
We assume that c is so large that the agent cannot aord, given his income process,
to obtaim ct(c
t
) = c for all c
t
. Note that it is easy to pick c large enough for T nite,
since o
T+1
(c
T
) = 0 is required. For T = o it must be chosen large enough so that the
consumption allocation ct(c
t
) = c leads to asset holdings that eventually violate the no Ponzi
scheme condition. But note that we explicitly wanted to pick the no-Ponzi scheme condition
wide enough to never be binding. One solution to this problem is to allow the consumer to
contemplate only consumption-asset allocations that lie in a particular commodity space, say
|1. Then with an appropriately high choice (which may depend on the income process) of c
agents will optimally choose consumption-asset sequences that dont have ct(c
t
) = c for all t,
all c
t
.
3.2.3 Empirical Tests of the Martingale Hypothesis Using
Macroeconomic Data
The last two implications were empirically tested rst in a seminal paper by
Robert Hall (1978). He used aggregate consumption and income data (the
corresponding empirical evidence from micro data will be discussed extensively
later) to run the two regressions, motivated by equation (8.10)
c
|l
= c
l
c
2
c
|
c
3
c
|l
c
d
c
|2
c
5
c
|3
n
|l
(3.106)
c
|l
= ,
l
,
2
c
|
,
3
j
|

|l
(3.107)
If the martingale hypothesis is true, then we expect c
3
= c
d
= c
5
= 0 and ,
3
=
0. As data Hall used for c
|
per capita nondurable consumption expenditures
(including services) in constant 1972 dollars from NIPA. His income data is
total nominal disposable income per capita from NIPA, divided by the implicit
deator for nondurable consumption and services. Again, in order to stress
this, these are aggregate data, although the theory has a single household as its
unit of analysis (we havent talked about aggregation yet for this model, but
exact aggregation as in the complete markets model does not obtain in PILCH
models). Hall basic results from regression for quarterly data from 1948-1977
are summarized in Table 3
Table 3: Test of Martingale Hypothesis
Regression Parameter Estimates and Standard Errors (in Parenthesis) 1
2
(8.106)
c
l
= 8.2
(8.8)
c
2
= 1.18
(.002)
c
3
= .04
(.142)
c
d
= .08
(.142)
c
5
= .118
(.008)
.0088
(8.107)
,
l
= 16
(11)
,
2
= 1.024
(.044)
,
3
= .01
(.082)
.0088
The basic message of Table 3 is that the martingale hypothesis cannot be
rejected. The coecients on lagged consumption and current income are not
signicantly dierent from 0, and 1-tests indicate that the hypothesis that they
are jointly equal to zero cannot be rejected at standard condence levels. The
inclusion of further lags in consumption or income does not change the result.
This is strong evidence in favor of the martingale hypothesis. When Hall in-
cludes current stock prices (the real value of the S&P 500) as a proxy for current
wealth into the regression he nds that they are statistically signicant, formally
rejecting the martingale hypothesis. He therefore concludes that this hypoth-
esis is rejected by the data. It is my reading of the subsequent perception of
his paper, however, that it was interpreted as strong evidence in favor of the
martingale hypothesis, with future rejections mostly deriving from studies that
used micro data rather than aggregate data, as economic theory suggests.
3.2.4 Consumption Responses to Income Shocks under
Certainty Equivalence
Let us now use (8.104) to explicitly solve for the consumption prole with un-
certainty (remember that we are now assuming j = r).
16
Consolidating the
budget constraints from period t onwards yields
1
|
T|
s=0
c
|s
(1 r)
s
= 1
|
T|
s=0
j
|s
(1 r)
s
a
|
= \
|
(3.108)
where the dependence of all variables on :
|
is understood (in fact, a
|
= a
|
(:
|l
))
and the expectation is conditional on :
|
. This can be seen as follows. Take the
budget constraint at node :
|
c
|
(:
|
)
a
|l
(:
|
)
1 r
= j
|
(:
|
) a
|
(:
|l
).
In the next period it reads as
c
|l
(:
|l
)
a
|2
(:
|l
)
1 r
= j
|l
(:
|l
) a
|l
(:
|
)
Solve for a
|l
(:
|
) and multiply by
tt+1(s
t+1
]s
t
)
l:
to obtain
|l
(:
|l
[:
|
)a
|l
(:
|
)
1 r
=
|l
(:
|l
[:
|
)
c
|l
(:
|l
)
ot+2(s
t+1
)
l:
j
|l
(:
|l
)
1 r
Note that this equation is valid for every node :
|l
following :
|
and note that
s
t+1
]s
t
|l
(:
|l
[:
|
)a
|l
(:
|
)
1 r
=
a
|l
(:
|
)
1 r
Substituting this back into the rst budget constraint yields
c
|
(:
|
)
a
|l
(:
|
)
1 r
= j
|
(:
|
) a
|
(:
|l
)
c
|
(:
|
)

s
t+1
]s
t
|l
(:
|l
[:
|
)a
|l
(:
|
)
1 r
= j
|
(:
|
) a
|
(:
|l
)
c
|
(:
|
)

s
t+1
]s
t
|l
(:
|l
[:
|
)
_
c
|l
(:
|l
)
ot+2(s
t+1
)
l:
_
1 r
= j
|
(:
|
)
s
t+1
]s
t
|l
(:
|l
[:
|
)j
|l
(:
|l
)
1 r
a
|
(:
|l
)
16
It is easy to check that there is another solution to the Euler equation (3.103), namely
ct(c
t
) = c for all t and c
t
. In fact, if this allocation is not ruled out by an appropriate borrowing
constraint or other tricks, it is the optimal allocation for the household. Halls (1978) paper
implicitly assumes that this allocation is not attainable.
Continuing this procedure and interpreting the conditional expectation appro-
priately yields equation (8.108).
Now since 1
|
c
|l
= c
|
and, by the law of iterated expectations 1
|
c
|2
=
1
|
1
|l
c
|2
= 1
|
c
|l
= c
|
we have, using (8.108), that
c
|
=
_
0
l
|
:Vt
l:
if T <
:
l:
\
|
if T =
(3.109)
where 0
|
=
_
1
l
(l:)
Jt+1
_
. Compare this to the consumption function under
certain income. Of course, the exact form that \
|
takes depends on the under-
lying stochastic labor income process. Given the consumption function we can
also determine how realized consumption changes with uctuations in income
(we already know that ex ante consumption should be constant in an expected
sense).
Rewrite (8.100) as
0
|
c
|
=
r\
|
1 r
(3.110)
0
|l
c
|l
=
r\
|l
1 r
(3.111)
and use the period t 1 budget constraint
a
|
= (1 r) (a
|l
j
|l
c
|l
) (3.112)
to obtain
0
|
c
|
= r (a
|l
j
|l
c
|l
)
r
1 r
1
|
T|
s=0
j
|s
(1 r)
s
(3.113)
0
|
c
|
rc
|l
1 r
=
r
1 r
(a
|l
j
|l
)
r
(1 r)
2
1
|
T|
s=0
j
|s
(1 r)
s
(3.114)
Subtracting 0
|l
c
|l
from both sides and using (8.111) one gets
0
|
c
|
rc
|l
1 r
0
|l
c
|l
=
r
(1 r)
2
1
|
T|
s=0
j
|s
(1 r)
s
r
1 r
j
|l
r
1 r
1
|l
T|l
s=0
j
|ls
(1 r)
s
(3.115)
Working on the left hand side we arrive at
0
|
c
|
rc
|l
1 r
0
|l
c
|l
=
0
|
c
|
1 r

_
1
1
(1 r)
T|2

r
1 r
_
c
|l
=
0
|
c
|
1 r

1
1 r
_
1 r
1
(1 r)
T|l
r
_
c
|l
=
0
|
^c
|
1 r
(3.116)
where we dene ^c
|
= c
|
c
|l
to be the realized change in consumption. Thus
0
|
^c
|
=
r
1 r
1
|
T|
s=0
j
|s
(1 r)
s
rj
|l
r1
|l
T|l
s=0
j
|ls
(1 r)
s
=
r
1 r
1
|
T|
s=0
j
|s
(1 r)
s
r1
|l
T|l
s=l
j
|ls
(1 r)
s
=
r
1 r
1
|
T|
s=0
j
|s
(1 r)
s

r
1 r
1
|l
T|l
s=l
j
|ls
(1 r)
sl
=
r
1 r
1
|
T|
s=0
j
|s
(1 r)
s

r
1 r
1
|l
T|
s=0
j
|s
(1 r)
s
=
r
1 r
T|
s=0
(1
|
1
|l
)j
|s
(1 r)
s
= j
|
(3.117)
where j
|
is the annuity value of innovations to future income between period
t1 and t. Again the exact value of j
|
depends on the particular income process.
Examples of Particular Income Processes
Let us look at some examples. Suppose the income process of an individual is
specied as
j
|
= j
|
n
|
(3.118)
j
|
= j
|l

|
(3.119)
where j
|
is the permanent part of current income, n
|
is the transitory part
and
|
is the innovation to the permanent part of income. We assume that n
|
and
|
are uncorrelated iid random variables with 1
|
n
|s
= 1
|
|s
= 0 for
: 0. This process can be rewritten as
j
|
= j
|l
n
|
n
|l

|
(3.120)
and thus
j
|s
= j
|l
n
|s
n
|l

|s
r=|
r
(3.121)
We can now demonstrate that agents obeying the PILCH model react quite
dierently to permanent income shocks
|
(getting the rst job as an assistant
professor) and transitory income shocks n
|
(win in the lottery). We have that
1
|
j
|s
= j
|l
n
|l

|

_
n
|
if : = 0
0 if : 0
(3.122)
1
|l
j
|s
= j
|l
n
|l
0 0 (3.123)
and hence
(1
|
1
|l
) j
|s
=
_
n
|

|
if : = 0
|
if : 0
(3.124)
Therefore we nd that the realized change in consumption is given by
0
|
^c
|
= j
|
=
r
1 r
n
|
0
|
|
(3.125)
^c
|
=
r0
l
|
1 r
n
|

|
(3.126)
and hence households adjust their consumption one for one with the permanent
shock
|
, but only change their consumption mildly, by
:0
1
t
l:
as response to a
temporary shock n
|
.
This discussion of course assumes that households can distinguish permanent
from transitory income shocks. Now suppose that households are unable to make
this distinction, and let income take the form
j
|
= j
|l
-
|
-
|l
(3.127)
with [0, 1[. Comparing (8.127) to (8.120) we see that the case of = 0
corresponds to only permanent shocks, whereas = 1 corresponds to only
transitory shocks. With (0, 1) shocks may be transitory or permanent, but
the agent cant tell. With this income process we nd that
(1
|
1
|l
) j
|s
=
_
-
|
if : = 0
(1 )-
|
if : 0
(3.128)
and thus
0
|
^c
|
=
r
1 r
-
|
(1 )0
|
-
|
(3.129)
Hence the agent tries to decompose the new information about income con-
tained in -
|
into a component reecting new information about the permanent
component of income and into a component reecting the transitory innovation
to income. His consumption response is accordingly. Again, for = 1 the agent
responds as to transitory shocks and with = 0 she responds as to permanent
shocks, if the shock may reect both shocks (from the perspective of the agent),
her response is a convex combination.
Finally, for the simple AR(1) process
j
|
= cj
|l
-
|
(3.130)
with 0 < c < 1 we have that
0
|
^c
|
=
r-
|
1 r
T|
s=0
_
c
1 r
_
s
(3.131)
and we see that the change in consumption in reaction to a persistent (but not
permanent) shock -
|
lies quantitatively in between that induced by a purely
transitory shock (which equals
:0
1
t
l:
), since

T|
s=0
_
o
l:
_
s
1 and that of a
permanent shock (which equals 1), since c < 1.
Chapter 4
Partial Equilibrium
Extensions of PILCH
Models
In this chapter we look at partial equilibrium extensions of the basic PILCH
model. That is, we will continue to assume that the interest rate process that
the agent faces is exogenously given (partial equilibrium) and that the agent
can only self-insure against income uctuations by trading one-period risk free
bonds (i.e. by borrowing and lending at a risk free rate).
We will relax two crucial assumptions underlying the martingale hypothesis
from last chapter, however. First, we will discuss how the relaxation of the
assumption of linear marginal utility aects consumption choices under uncer-
tainty. In particular we want to incorporate a precautionary savings motive
into the model, so that agents reduce current consumption (and hence increase
saving) as reaction to an increase of uncertainty with respect to future labor in-
come. For this part the key references include Kimball (1990), Barsky, Mankiw
and Zeldes (1986), Deaton (1991) and Carroll (1997).
Second, so far we assumed that agents can always borrow as much as de-
sired to smooth consumption over time (of course, subject to a non-binding no
Ponzi condition). This may be an assumption with little empirical appeal and
its relaxation (i.e. the imposition of some form of potentially binding borrowing
constraint, a so-called liquidity constraint) will also invalidate the martingale
hypothesis. An empirical test for the presence of liquidity constraints is pre-
sented by Zeldes (1989) and will be discussed in class.
Finally we will discuss optimal consumption choices with both a precaution-
ary saving motive and binding borrowing constraints. As we will see, most of
the analysis relies on numerical approximations of the consumption function,
as analytical solutions for this problem, in contrast to the certainty equivalence
case, are usually not available. The few theoretical results can be found, un-
der varying assumptions about the stochastic income process and the relative
49
50CHAPTER 4. PARTIAL EQUILIBRIUMEXTENSIONS OF PILCHMODELS
size of interest rate and time discount rate, in Schechtman (1976), Schechtman
and Escudero (1977), Yaari (1977), Sotomayor (1984), Deaton (1991), Huggett
(1993) and Chamberlain and Wilson (2000).
[Also discuss Carroll and Kimball (1996, 2001)]
4.1 Precautionary Savings
Certainty equivalence explicitly rules out precautionary saving. If we look at
the consumption function under certainty equivalence (8.100), and remember
the denition of \
|
as the discounted sum of expected future labor income,
we see that only the conditional (on period t information) rst moment of the
j
|s
s matters for the consumption choice, but not the conditional variance of
future labor income. Hence a change in the uncertainty of future labor income
(with unchanged mean) leaves consumption choices unchanged under certainty
equivalence (almost by denition).
In this chapter, unless otherwise noted, we continue to assume that the in-
terest rate r is constant and thus nonstochastic. For a general utility function
the Euler equation for PILCH models with uncertainty (and no liquidity con-
straints) is, repeating equation (8.100)
l
c
(c
|
, :
|
) =
_
1 r
1 j
_
1
|
_
l
c
(c
|l
, :
|l
_
) (4.1)
4.1.1 A Simple Model and a General Result
Let us consider a simple two-period model where in period 0 income is known
and equal to j
0
and period 1 income is stochastic.
1
First assume j = r = 0
(the results go through unchanged with j ,= 0 and/or r ,= 0, but the algebra
becomes more messy). Let denote by 1
l
the random income in period 1. It will
be convenient to write
1
l
= j
l

1
l
(4.2)
where j
l
= 1
0
1
l
is the (conditional) expectation of 1
l
at period 0 and

1
l
is a
random variable with 1
0
(
1
l
) = 0. Let us also dene
n = j
0
j
l
(4.3)
as expected present discounted value of lifetime labor income.
Consumption at date 1 is then given by (remember r = 0 is assumed)
c
l
= n c
0

1
l
(4.4)
i.e. is stochastic and varies with the realization of 1
l
. Let by : = nc
0
denote
the saving of the consumer. The Euler equation, ignoring preference shocks,
becomes
l
c
(c
0
) = 1
0
l
c
(n c
0

1
l
) (4.5)
1
The discussion of the simple model is motivated by Barsky, Mankiw and Zeldes (1986).
4.1. PRECAUTIONARY SAVINGS 51
We want to determine how consumption in period 0 (and hence saving in period
0) varies with the degree of uncertainty about labor income in period 1, i.e. with
o
2
= \ ar(
1
l
). For concreteness assume that
1
l
=
_
-
-
with prob.
l
2
with prob.
l
2
(4.6)
with 0 < - < j
l
, so that o
2
= -
2
. Totally dierentiating equation (4.) with
respect to - yields
l
cc
(c
0
)
dc
0
d-
=
1
2
_
l
cc
(n c
0
-))
_
dc
0
d-
1
_
l
cc
(n c
0
-))
_
dc
0
d-
1
__
(4.7a)
and thus
dc
0
d-
=
l
2
(l
cc
(n c
0
-) l
cc
(n c
0
-))
l
cc
(c
0
)
l
2
[l
cc
(n c
0
-)) l
cc
(n c
0
-))[
(4.8)
The denominator of this expression is unambiguously negative (as we assume
that l is strictly concave). The nominator is positive if and only if
l
cc
(n c
0
-)) l
cc
(n c
0
-)) (4.9)
is positive.
But this is true for arbitrary - 0 if and only if l
ccc
(c) 0. Hence con-
sumption in the period 0 declines in reaction to an increase in uncertainty about
period 1 income if and only if the third derivative of the utility function is pos-
itive. Thus a sucient (and necessary) condition for precautionary saving (i.e.
increased saving as reaction to increase uncertainty of future labor income) is
strictly convex marginal utility. Kimball (1990) has dened the term prudence
to mean the propensity to prepare and forearm oneself in the face of uncer-
tainty and hence the intensity of the precautionary saving motive (although
the concept of prudence applies to other decisions in uncertain environments).
From this denition it is clear that the term prudence characterizes preferences,
whereas precautionary saving characterizes behavior: prudence leads to precau-
tionary saving, but both concepts should be kept separately.
Note that prudence and risk aversion are not the same thing; risk aversion
is controlled by the concavity of the utility function whereas prudence (the
precautionary savings motive) is controlled by the convexity of the marginal
utility function.
[Here insert a brief discussion of Barsky et al.s Ricardian tax experiment
in this simple model; you would think there is Ricardian equivalence, but the
precautionary savings motive kills this, since future tax increases may make
future income less risky]
In correspondence to the Arrow-Pratt measures of absolute and relative risk
aversion
r(c) =
l
cc
(c)
l
c
(c)
(4.10)
o(c) =
cl
cc
(c)
l
c
(c)
(4.11)
Kimball (1990) denes the index of absolute prudence
j(c) =
l
ccc
(c)
l
cc
(c)
(4.12)
and one could also dene the index of relative prudence
jr(c) =
cl
ccc
(c)
l
cc
(c)
(4.13)
Both indices measure the intensity of prudence, i.e. the intensity of the precau-
tionary savings motive. Kimball shows that, drawing on results by Pratt (1964)
the following. Suppose there is no uncertainty, i.e. o
2
= 0 and let the consump-

tion function be denoted by c
0
(n). Now for small (marginal) uncertainty o
2
the
resulting consumption function c
0
(n) satises
c
0
(n (o
2
, :)) = c
0
(n) (4.14)
where
(o
2
, :) =
1
2
o
2
l
ccc
(:)
l
cc
o(o
2
) (4.15)
and : = :(n) = n c(n) is the level of saving in the absence of uncertainty.
The residual term o(o
2
) satises
lim
c
2
0
o(o
2
)
o
2
= 0. (4.16)
Note that (o
2
, :) 0 if and only if l
ccc
0. We can interpret (o
2
, :) as the
rightward shift in the consumption function (as a function of human wealth n)
induced by small uncertainty: if without uncertainty a wealth level n leads to
consumption c = 10 in the rst period, now with uncertainty a higher wealth
level n(o
2
, :) is needed to induce that same consumption c = 10. Since saving

is simply wealth n minus rst period consumption c
0
, the reverse relations apply
to saving, i.e. the saving function shifts to the left if and only if l
ccc
0.
This result indicates that the magnitude of the shift in the saving function
is proportional to the index of absolute prudence, which justies this index as
a quantitative measure of prudence (the intensity of the precautionary savings
motive). It is reminiscent of the result proved by Pratt (1964) that the risk
premium 0(o
2
, c), for small, zero mean uncertainty, dened implicitly as

1
0
l(c

1
l
) = l(c 0(o
2
, c)) (4.17)
4.1. PRECAUTIONARY SAVINGS 53
is explicitly given by
0(o
2
, c) =
1
2
o
2
l
cc
(c)
l
c
o(o
2
) (4.18)
and hence justies the coecient of absolute risk aversion as measure to quantify
willingness to pay to avoid (small) risk.
For the special case of constant relative risk aversion utility l(c) =
c
1o
lc
we
nd that
j(c) =
o 1
c
and jr(c) = o 1 (4.19)
whereas for constant absolute risk aversion l(c) =
l
~
c
~c
we have
j(c) = and jr(c) = c (4.20)
and in both cases risk aversion and prudence is controlled by a single parameter
(o and , respectively).
4.1.2 A Parametric Example for the General Model
Now let us consider the general model with many periods. In general nothing
analytical can be said about the time (age) prole of consumption. Under
particular assumptions, however, we can derive qualitative results. Let us again
start with the Euler equation
l
c
(c
|
, :
|
) =
_
1 r
1 j
_
1
|
_
l
c
(c
|l
, :
|l
_
) (4.21)
Assuming separability and CRRA utility this becomes
_
1 r
1 j
_
1
|
_
_
c
|l
c
|
_
c
_
= 1
1
|
_
c
In
(
1+r
1+,
)(
c
t+1
c
t
)
o
_
= 1
1
|
_
c
c In(ct+1)
c
c In(ct)In(l:)In(l)
_
= 1
c
c In(ct)In(l:)In(l)
1
|
_
c
c In(ct+1)
_
= 1 (4.22)
Now assume that ln(c
|l
) is normally distributed with mean j = 1
|
ln(c
|l
)
and variance o
2
c
, which requires appropriate assumptions on the underlying sto-
chastic income process.
2
Then we note that since
1
|
_
c
c In(ct+1)
_
=
_
o
o
c
cu
c
(r,)
2
2o
2
c
_
2o
c
dn (4.23)
2
For the two period model below, this requires that the random variable
Z = ln(i +
~
Y
1
)
is normally distributed, where i = & c
0
is a constant.
where n is the dummy argument for ln(c
|l
) in the integration. Now
1
|
_
c
c In(ct+1)
_
=
_
o
o
c
cu
c
(r,)
2
2o
2
c
_
2o
c
dn
= c
(,oo
2
c
)
2
,
2
2o
2
c
_
o
o
c
[
r(,oo
2
c
)
[
2
2o
2
c
_
2o
c
dn
= c
(,oo
2
c
)
2
,
2
2o
2
c
= c
1
2
c
2
c
2
c
c
= c
1
2
c
2
c
2
c
cJt In(ct+1)
(4.24)
Thus equation (4.22) becomes
c
cJt.In(ct+1)In(l:)In(l)
1
2
c
2
c
2
c
= 1 (4.25)
and upon taking logs
1
|
^ln(c
|l
) =
1
o
[ln(1 r) ln(1 j[
1
2
oo
2
c
(4.26)
Remember that without uncertainty and CRRA utility the Euler equation be-
comes
_
1 r
1 j
__
c
|l
c
|
_
c
= 1 (4.27)
or
^ln(c
|l
) =
1
o
[ln(1 r) ln(1 j[ (4.28)
We make three important observations
1. The consumption decision do not obey certainty equivalence. As CRRA
utility induces a precautionary savings motive to households, from (4.26)
we see that uncertainty about future consumption, captured in the last
term of (4.26) tilts the consumption prole upward in expectation: con-
sumption growth is higher in the uncertainty case than in the certainty
case, meaning that households tend to postpone consumption for precau-
tionary motives.
2. The degree to which consumption is postponed (i.e. the degree to which
there is precautionary saving) as reaction to future uncertainty is deter-
mined by the parameter o controlling prudence for the CRRA utility func-
tion, as we did show earlier was the case for the two-period model.
3. All variables
|
that, at period t, help to predict the variability of future
consumption o
2
c
, will help to predict expected consumption growth. For
example, agents with a higher level of current assets or income, since
they may be better able to smooth future uncertainty, should have lower
4.2. LIQUIDITY CONSTRAINTS 55
future consumption variability and thus, according to equation (4.26),
lower consumption growth. This point was made, among others, by Carroll
(1992). Thus it may be awed to run the regression
lnc
|l
= c
l
c
2
lnc
|
c
3
|
-
|
(4.29)
where
|
may be current wealth, and interpret a statistically signicant
estimate of c
3
as evidence against the PILCH model with CRRA utility.
4.2 Liquidity Constraints
So far we have assumed that the household can borrow up to some arbitrarily
large amount, up to the no-Ponzi scheme condition which was assumed to be
generous enough never to be binding. Apart from the fact that borrowing
constraints seem empirically plausible and formal econometric tests seem to
indicate not only their presence, but also their eect on consumption allocations
(see Zeldes (1989)), let us briey see what can happen in the absence of binding
borrowing constraints. Assume that households have quadratic utility so that
certainty equivalence holds. Also suppose that the household is innitely lived.
The budget constraint gives
a
|l
(1 r)a
|
= (1 r)(j
|
c
|
) (4.30)
Using the optimal consumption rule under certainty equivalence
c
|
=
r
1 r
_
1
|
o
s=0
j
|s
(1 r)
s
a
|
_
(4.31)
we nd that
a
|l
a
|
= ^a
|l
= (1 r)j
|
(1 r)c
|
ra
|
= (1 r)j
|
r1
|
o
s=0
j
|s
(1 r)
s
= j
|
r1
|
o
s=l
j
|s
(1 r)
s
(4.32)
= j
|
1
|
o
s=l
_
j
|s
(1 r)
sl

j
|s
(1 r)
s
_
(4.33)
= 1
|
o
s=l
^j
|s
(1 r)
sl
(4.34)
Hence, if income shocks are iid we have
j
|
= -
|
. (4.35)
Then, since 1
|
^j
|s
= -
|
for : = 1 and 1
|
^j
|s
= 0 for : 1 we have that
^a
|l
= -
|
(4.36)
i.e. assets follow a random walk, with the property that with probability 1
asset holdings will exceed any given borrowing limit eventually. This may not
violate the no Ponzi scheme condition if we specify it very carefully, but calls
into question the assumption that liquidity constraints are never binding, a
maintained assumption so far.
4.2.1 The Euler Equation with Liquidity Constraints
Now let us assume that there exist borrowing constraints, and for simplicity let
us follow Deaton (1991), Schechtman (1976), Aiyagari (1994) and many others
and assume that agents cannot borrow at all, i.e. face the constraint a
|l
(:
|
) _ 0
for all :
|
. These constraints may or may not be binding, depending on the
realizations of the labor income shock, but we have to take these constraints
into account explicitly when deriving the stochastic Euler equation. Let us
attach Lagrange multiplier j(:
|
) to the borrowing constraint a
|l
(:
|
) _ 0 at
event history :
|
. The rst order conditions with respect to consumption, (8.0)
and (8.06) remain unchanged. The rst order condition with respect to a
|l
(:
|
)
now becomes
`(:
|
)
1 r
j(:
|
) =

s
t+1
]s
t
`(:
|l
) (4.37)
with complementary slackness conditions
a
|l
(:
|
), j(:
|
) _ 0
a
|l
(:
|
)j(:
|
) = 0 (4.38)
Combining the rst order condition yields
l
c
(c
|
(:
|
)) j(:
|
)(1 r) = (1 r),

s
t+1
]s
t
,
|l
(:
|l
[:
|
)l
c
(c
|l
(:
|l
), :
|l
)
(4.39)
or in short
l
c
(c
|
) _
1 r
1 j
1
|
l
c
(c
|l
) (4.40)
=
1 r
1 j
1
|
l
c
(c
|l
) if a
|l
0
This Euler equation, which sometimes holds with inequality, depending on
whether the liquidity constraint is binding, is the basis for empirical tests for liq-
uidity constraints. In particular, Zeldes (1989) subdivides the sample of house-
holds into two groups, depending on their current wealth positions, with one
group composed of households whose liquidity constraint is most likely not bind-
ing (high wealth households) and the other group composed of households whose
4.2. LIQUIDITY CONSTRAINTS 57
constraint is most likely binding (low wealth households). The empirical test
then consists of running a regression like Hall (1978), but for both groups sep-
arately (and with micro data rather than macro data) and ask whether current
income helps forecast future consumption growth. for both groups As evidence
for the presence of liquidity constraints would be interpreted a nding which
shows that current income helps predict consumption growth for the low-wealth
group, but not for the high-wealth group. Zeldes (broadly) nds such evidence
and concludes that borrowing constraints are important in shaping consumption
choices, at least for the currently wealth-poor. In evaluating his result, how-
ever, you should keep in mind that we argued above that with preferences that
exhibit prudence, current income may help predict consumption growth, if it
contains information about the extent of future income uncertainty. While it is
not entirely obvious why this should be the case for low-wealth households and
not for high-wealth households, note that wealth accumulation is an endogenous
choice, so the low wealth-group sample is not simply a random sample. There-
fore it is not inconceivable that low-wealth households have income processes
with very dierent stochastic properties that high-wealth households, including
predictability of future income risk by current income.
We now want to analyze (4.40) a bit further. Note from the budget constraint
that
c
|
= j
|
a
|
a
|l
1 r
_ j
|
a
|
(4.41)
because of the constraint a
|l
_ 0. Thus either a
|l
= 0, therefore c
|
= j
|
a
|
and
l
c
(c
|
) = l
c
(j
|
a
|
)
or a
|l
0, therefore c
|
< a
|
j
|
and thus (using strict concavity of the utility
function)
l
c
(j
|
a
|
) < l
c
(c
|
) =
1 r
1 j
1
|
l
c
(c
|l
)
Hence the Euler equation can be compactly written as
l
c
(c
|
) = max
_
l
c
(j
|
a
|
),
1 r
1 j
1
|
l
c
(c
|l
)
_
(4.42)
4.2.2 Liquidity Constraints with Quadratic Preferences:
Precautionary Saving without Prudence
Before summarizing the theoretical and quantitative implications of liquidity
constraints in the presence of prudence, let us briey investigate how liquidity
constraints aect consumption allocations if all other assumptions required for
certainty equivalence are in place. With quadratic preferences (4.42) becomes
(c
|
c) = max
_
(j
|
a
|
c),
1 r
1 j
(1
|
c
|l
c)
_
(4.43)
For simplicity assume j = r, and remember that max )(r) = min)(r). Then
(4.48) can be rewritten as
c
|
= minj
|
a
|
, 1
|
c
|l
(4.44)
= minj
|
a
|
, 1
|
minj
|l
a
|l
, 1
|l
c
|2
and so forth. We obtain several implications

1. If the certainty equivalence solution (4.81) (or its nite horizon counter-
part) has associated asset holdings that satisfy a
|
_ 0 with probability
1, then it is the optimal consumption allocation even in the presence of
borrowing constraints. Whether or not this is the case obviously depends
crucially on the income process. If asset holdings follow a random walk,
as above with iid income shocks, this assumption is obviously violated.
2. In the absence of borrowing constraints we know from our discussion of
certainty equivalence that
c
|
= 1
|
(c
|s
)
for all : 0. Suppose now there exists an : 0 (e.g. : = 1) such that for
some realization of the income shock j
|s
(with positive probability) the
household is borrowing constrained, and thus j
|s
a
|s
< 1
|s
c
|s
. Then
equation (4.44) implies that (using the law of iterated expectations) c
|
<
1
|
(c
|s
). Thus, even if the liquidity constraint is not binding in period t,
future potential borrowing constraints aect current consumption choices.
This argument also demonstrates that even if econometric studies such
as Zeldes (1989) do not nd evidence for currently binding borrowing
constraints (for particular groups), one cannot conclude that borrowing
constraints are absent in general or do not aect consumption decisions.
3. Now suppose the variance of future income increases (say for t1), making
more low realizations of j
|l
possible or likely (and more high realizations
also). If the set of j
|l
values for which the borrowing constraint binds be-
comes bigger, 1
|
minj
|l
a
|l
, 1
|
c
|2
declines and so does c
|
, since in
more instances the min is the rst of the two objects. Saving increases in
reaction to increases in uncertainty about future income, because agents,
afraid of future contingencies of low consumption (agents with quadratic
utility are risk-averse) and aware of their inability to smooth low income
shocks via borrowing, increase their precautionary savings. Note that we
obtain precautionary savings behavior without prudence (i.e. without a
precautionary saving motive) purely from the existence of liquidity con-
straints (and risk aversion). Hence when observing increases in saving
as reaction to increased income uncertainty, this may have a preference-
based interpretation (agents are prudent: l
ccc
0) or an institution-based
interpretation (credit markets prevent or limit uncontingent borrowing).
4.3. NUMERICAL SOLUTIONS OF MODELS WITHPRECAUTIONARYSAVINGANDLIQUIDITYCONSTRAINTS59
4.3 Numerical Solutions of Models with Precau-
tionary Saving and Liquidity Constraints
The analysis in the previous section is as far as one can push the model an-
alytically. Apart from very special cases (see Caballero (1990) for an explicit
solution when utility is CARA, agents life for nite time, the interest rate is zero
and income follows a random walk with normally distributed iid disturbances)
the model cannot be solved analytically.
For computational purposes we want to make the problem a consumer faces
recursive. From now on, for the rest of this chapter, unless otherwise noted,
we assume that the stochastic process governing labor income is described by a
nite state, stationary Markov process with domain j 1 = j
l
, . . . , j
and
transition probabilities (j
t
[j), where we assume that j
l
_ 0 and j
Il
j
I
. As
state variables we choose current asset holdings and the current labor income
shock (a, j). In order to make the problem well-behaved we have to make sure
that agents dont go into debt so much that they cant pay at least the interest
on that debt and still have non-negative consumption. Let

be the maximum
amount an agent is allowed to borrow. Since consumption equals c = ja
o
0
l:
,
we have, for an agent that borrowed to the maximum amount a =
, received
the worst income shock j
l
and just repays interest (i.e. a
t
=
):
c = j
l

1 r
_ 0 (4.45)
Non-negativity of consumption implies the borrowing limit
=
1 r
r
j
l
(4.46)
which we impose on the consumer. Since

also equals the present discounted
value of future labor income in the worst possible scenario of always obtaining
the lowest income realization j
l
, we may call this borrowing limit the natural
debt limit (see Aiyagari (1994)). Note that, since borrowing up to the bor-
rowing limit implies a positive probability of zero consumption next period (if
(j
l
[j) 0 for all j 1 ), this borrowing constraint is not going to be binding,
as long as the utility function satises the Inada conditions.
For any a [
, ) and any j 1 we then can write Bellmans equation

as
|
(a, j) = max
.o
0
(l:)(o)
_
_
_
l
_
j a
a
t
1 r
_
,
0
(j
t
[j)
|l
(a
t
, j
t
)
_
_
_
(4.47)
The rst order conditions to this problem is
1
1 r
l
t
_
j a
a
t
1 r
_
= ,
0
(j
t
[j)
t
|l
(a
t
, j
t
) (4.48)
where
t
|l
is the derivative of
|l
with respect to its rst argument. The
envelope condition is
t
|
(a, j) = l
t
_
j a
a
t
1 r
_
(4.49)
Combining both conditions yields
l
t
_
j a
a
t
1 r
_
= ,(1 r)
0
(j
t
[j)l
t
_
j
t
a
t
a
tt
1 r
_
(4.50)
Dening
c
|
= j a
a
t
1 r
(4.51)
c
|l
= j
t
a
t
a
tt
1 r
(4.52)
we obtain back our stochastic Euler equation (8.100). There are several remarks
to be made for the computation of these models
1. If labor income shocks are iid over time with probability density , then
one can reduce the state space to a single variable, so called cash at hand
r = a j. The budget constraint then becomes
c
a
t
1 r
= r (4.53)
and since
r
t
= a
t
j
t
(4.54)
we have the associated Bellman equation
|
(r) = max
.o
0
(l:)r
_
_
_
l
_
r
a
t
1 r
_
,
0
(j
t
)
|l
(a
t
j
t
)
_
_
_
2. Independent of whether the state space consists of cash at hand r or (a, j),
asset holdings and the current income shock we can either work on the
value function directly or use the Euler equation to solve for the optimal
policies. Depending on whether the time horizon of the household is nite
or innite, dierent iterative procedures can be applied
(a) Finite time horizon T, value function iteration. We want to nd
sequences of value functions
|
(a, j)
T
|=0
and associated policy func-
tions c
|
(a, j), a
t
|
(a, j)
o
|=0
. Given that the agent dies at period T we
can normalize
Tl
(a, j) = 0. Then we can iterate backwards on the
equation (4.47): at period t we know the function
|l
(., .), hence can
solve the maximization problem to nd functions a
t
|
(., .),
|
(., .) and
c
|
(., .). Note that for each t there are as many maximization problems
to solve as there are admissible (a, j)-pairs.
(b) Finite time horizon, policy function iteration on the Euler equation.
We are looking for sequences of policy functions c
|
(a, j), a
t
|
(a, j)
o
|=0
.
Again, given that the agent dies at period T 1 we know that at
period T all income will be consumed and nothing we be saved, thus
c
T
(a, j) = a j (4.55)
a
t
T
(a, j) = 0 (4.56)
The Euler equation between period t and t 1 reads as
l
t
_
j a
a
t
|
(a, j)
1 r
_
= ,(1r)
0
(j
t
[j)l
t
_
j
t
a
t
|
(a, j)
a
t
|l
(a
|
(a, j), j
t
)
1 r
_
(4.57)
where a
t
|l
(., .) is a known function from the previous step and we
want to solve for the function a
t
|
(., .). Note that for a given (a, j) no
maximization is needed to nd a
t
|
(a, j) as in the previous procedure,
one just has to nd a solution to a (potentially highly nonlinear)
single equation.
(c) Innite Horizon T = , value function iteration. Now we look for a
time invariant value function (a, j) and associated policy functions
a
t
(a, j) and c(a, j). We need to nd a xed point to Bellmans equa-
tion, since there is no nal period to start from. Thus we make an
initial guess for the value function,
0
(a, j) and then iterate on the
functional equation
n
(a, j) = max
.o
0
(l:)(o)
_
_
_
l
_
j a
a
t
1 r
_
,
0
(j
t
[j)
nl
(a
t
, j
t
)
_
_
_
(4.58)
until convergence, i.e. until
_
_
nl
_
_
_ - (4.59)
where - is the desired precision of the approximation. Note that at
each iteration we generate policy functions a
tn
(a, j) and c
n
(a, j). We
know that under appropriate assumptions (, < 1, l bounded) the
operator dened by Bellmans equation is a contraction mapping and
hence convergence of the iterative procedure to a unique xed point
is guaranteed.
(d) Innite horizon T = , policy function iteration on the Euler equa-
tion. Again we have no nal time period, so we guess an initial policy
a
t0
(a, j) or c
0
(a, j) and then iterate on
l
t
_
j a
a
tn
(a, j)
1 r
_
= ,(1r)
0
(j
t
[j)l
t
_
j
t
a
tn
(a, j)
a
tnl
(a
tn
(a, j), j
t
)
1 r
_
(4.60)
until
_
_
a
tn
a
tnl
_
_
_ -
o
(4.61)
Deaton (1992) argues that under the assumption ,(1 r) < 1 the
operator dened by the Euler equation is a contraction mapping, so
that convergence to a unique xed point is guaranteed.
3. Deaton (1991) demonstrates how to tackle problems in which income of
households follow a random walk. For agents living forever, there is no
hope to obtain nite bounds on the amount of assets that are potentially
accumulated, hence the state space is unbounded, which poses problems
for the computation of optimal policies (and the value function). It should
also be noted that in such an economy there is in general no hope for a
stationary general equilibrium, since the income distribution is nonsta-
tionary and thus in general the wealth distribution will be for any xed
interest rate r. Hence when dealing with general equilibrium models with
innitely lived agents, one of the maintained assumptions is stationarity
of the individual income processes. Note that the same comment does not
apply for economies in which agents die with probability 1 in nite time
(see, e.g., Gourinchas (2000)). The partial equilibrium problem with non-
stationary income process and innite horizon can be tackled, however,
using the following trick by Deaton (1991). Assume that period utility is
of C11-form and suppose that the income process of the individual is
such that the growth rate of income
.
|l
=
j
|l
j
|
(4.62)
is a sequence of iid random variables. This implies that
log(j
|l
) log(j
|
) (4.63)
is equal to an iid random variable, and thus implies that the natural
logarithm of income follows a random walk, potentially with drift
log(j
|l
) = j log(j
|
) -
|
(4.64)
where -
|
is a sequence of iid random variables with 1(-
|
) = 0. Note
that this implies strictly positive income in all periods, with probability
one. In order to overcome the problem that the state space for assets is
unbounded, one may try to normalize all variables by the current level of
income
0
|
=
c
|
j
|
(4.65)
n
|
=
r
|
j
|
=
a
|
j
|
j
|
(4.66)
The budget constraint then becomes
c
|

a
|l
1 r
= j
|
a
|
c
|

r
|l
j
|l
1 r
= r
|
0
|

n
|l
1
1 r
.
|l
= n
|
n
|l
=
(1 r)(n
|
0
|
)
.
|l
1 (4.67)
and, by dividing both sides of the Euler equation by j
c
|
we obtain
0
c
|
= max
_
n
c
|
,
_
1 r
1 j
_
1
|
_
_
c
|l
j
|
_
c
__
= max
_
n
c
|
,
_
1 r
1 j
_
1
|
_
(0
|l
.
|l
)
c
_
_
(4.68)
We are searching for a time-invariant policy function 0(n) solving this
Euler equation. Note that, written in recursive formulation 0
|
corresponds
to 0(n) and 0
|l
to 0(n
t
) where n
t
=
(l:)(u0(u))
:
0
. Assuming that .
t
=
.
|l
can take on only a nite number of values we obtain
0
c
|
=
_
max n
c
,
_
1 r
1 j
_
:
0
(.
t
)
_
0
_
(1 r)(n 0(n))
.
t
1
_
.
t
_
c
_
(4.69)
This Euler equation can be solved numerically for the optimal policy func-
tion 0(n) of consumption per current income (the average propensity to
consume) by doing policy iteration on
3
0(n)
c
= max
_
n
c
,
_
1 r
1 j
_
:
0
_
0
_
(1 r)(n 0(n))
.
t
1
_
.
t
_
c
_
(4.71)
Note that Deaton (1991) is not able to develop a condition under which
the state space for n is bounded above. It is obviously bounded below
by 1 since assets a
|
are restricted to be nonnegative. In accordance to
the model with stationary income he establishes a threshold n such that
0(n) = n and n
t
(n) = 1 for all n _ n; i.e. for low levels of normalized
cash at hand it is optimal for the consumer to eat all cash at hand today
and save nothing for tomorrow.
3
Deaton and Laroque (1992) show that with nonstationary income the operator associated
with the Euler equation is a contraction if
1 +v
1 +j
1(:
) < 1 (4.70)
4. Exact methods of how to implement these algorithms vary a great deal.
They range from discretization of the state space for assets and income
(a, j) and approximation of the functions only on the grid, to approxima-
tion of the functions over the entire real line by piecewise linear function,
to more sophisticated approximations with Chebychev polynomials and
other methods. You will adopt some of these methods in your research
project, but Ill leave it to you which one to use. The best source of in-
formation about the available procedures as well as their advantages and
disadvantages can be found in Ken Judds Numerical Methods in Eco-
nomics which should be required reading for every student working on
computational projects. Refer to his Chapters 10-12 for all the details.
5. Once one has solved for the policy functions one can simulate time paths
for consumption and asset holdings for a particular agent. Start at some
asset level a
0
, possibly equal to zero, then draw a sequence of random in-
come numbers j
I
1
|=0
according to the specied income process. This will
usually require rst drawing random numbers from a uniform distribution
and then mapping these numbers into the appropriate labor income real-
izations. Now one can generate a time series for consumption and asset
holdings (for the innite horizon case) by
c
0
= c(a
0
, j
0
) (4.72)
a
l
= a
t
(a
0
, j
0
) (4.73)
and recursively
c
I
= c(a
I
, j
I
) (4.74)
a
Il
= a
t
(a
I
, j
I
) (4.75)
The same applies to the nite time horizon case, but here the policy
functions being applied also vary with time.
4.3.1 Implementation of Particular Algorithms
[Piecewise linear approximation, Finite Element Methods in General, Weighted
Residual Methods, especially Chebychev collocation]
4.4 Combining Prudence and Liquidity Constraints
In this section we summarize the main theoretical results that are known about
the income uctuation problem with liquidity constraints and potentially non-
quadratic preferences. Important references include Schechtman (1976), Yaari
(1976), Schechtman and Escudero (1977), Sotomayor (1984) and Chamberlain
and Wilson (2000). As it will turn out, the results depend crucially on assump-
tions about the relative magnitudes of the interest rate and the time discount
factor. This will have important implications for the relationship between r and
j in the general equilibrium models in the next chapter, in which the interest
rate is determined endogenously.
4.4. COMBINING PRUDENCE AND LIQUIDITY CONSTRAINTS 65
4.4.1 Case T nite
If T is nite, obviously c
|
and a
|
remain bounded. In general equilibrium, in-
terest rates may exceed the subjective time discount factor, depending on the
age prole of labor income. General equilibrium models with many overlapping
generations, each of which faces a income uctuation problem with nite hori-
zon, have become popular tools to analyze policy reforms, from social security
reform to fundamental tax reform. We will come back to some of these studies,
which are purely numerical in nature.
4.4.2 Case T = and < r
In this case it turns out to be optimal for households to accumulate assets
indenitely and consume c
|
= as t . The basic intuition comes from
equation (4.40). Under the assumption that j < r we have
l
c
(c
|
) _
1 r
1 j
1
|
l
c
(c
|l
) (4.76)
1
|
l
c
(c
|l
)
Hence marginal utility, which is strictly positive, follows a supermartingale.
4
By the martingale convergence theorem
5
it follows that the sequence of random
variables l
c
(c
|
) converges almost surely to some limit random variable l
c
(c).
Since for all t
1
|
l
c
(c
|l
) _
_
1 j
1 r
_
|l
l
c
(c
0
) (4.77)
and l
c
is strictly positive, it must be the case that not only does l
c
(c
|
)
converges almost surely to some limit random variable l
c
(c), but it converges
almost surely to l
c
(c) = 0. From strict concavity and the Inada conditions of
the utility function it follows that
lim
|o
c
|
= almost surely (4.78)
Obviously, to nance innite consumption in the time limit assets a
|
have to
satisfy
lim
|o
a
|
= almost surely (4.79)
i.e. for j < r agents run up their assets indenitely to nance higher and
higher consumption. This result also shows that with innitely lived agents
there will not be a stationary asset distribution if j < r and thus no stationary
general equilibrium with j < r will exist. This result was proved, under various
assumptions, by Sotomayor (1984) and Chamberlain and Wilson (2000). Note
4
In fact, for the seqence of random variables lc(ct) to be a supermartingale we need
that lc(ct) is nite, which is assured if the endowment process jt is such that positive
consumption is is possible with probability 1 (or alternatively, if lc(0) < o, violating the
Inada condition).
5
The theorem was rst proved by Doob (1953) in Stochastic Processes.
that Sotomayor does not assume l to be bounded, but makes the assumption
that the income process is iid with support j
|
[a, [ where a _ 0 and < .
On the other hand, Chamberlain and Wilson assume l to be bounded, but
consider arbitrary income processes (and can allow the interest rates r
|
to
follow a stochastic process, too).
4.4.3 Case T = and = r
Sotomayor (1984) and Chamberlain and Wilson (2000) show that the previ-
ous result goes through even for j = r, provided that the income process is
suciently stochastic. In particular, suppose that the endowment process is a
deterministic sequence j
|
o
|=0
. Then we have the following proposition, due to
Chamberlain and Wilson
Proposition 9 Dene
r
|
=
r
1 r
o
r=|
j
r
(1 r)
r|
(4.80)
Then
c = lim
|o
c
|
= sup
|
r
|
= r (4.81)
The proposition says that consumption converges to some nite limit sup
|
r
|
.
The interpretation of this limit goes as follows. Suppose an agent arrives at date
t with a
|
= 0, i.e. with zero assets. If the borrowing constraint is never binding
again in the future he should consume
c
|
= r
|
(4.82)
as a standard permanent income hypothesis consumer would do. The proposi-
tion basically says that the eects of potentially binding borrowing constraints
never vanish until the period with the highest annuity value of future labor
income, r
|
, is reached.
Note that for the case j = r and no uncertainty the Euler equation implies
c
|
= c
|l
or (4.83)
c
|
_ c
|l
and a
|l
= 0
Proof. See Sargent and Ljungquist, Chapter 13, p. 356. Note that for this
result l need not be bounded, a maintained assumption of Chamberlain and
Wilson (2000).
Now lets turn to the more interesting case in which the endowment process
is stochastic. Here the result that consumption and asset holdings diverge as
time extends to innity is restored, provided that, for any event history :
|
the
present discounted value of future income is suciently stochastic.
Proposition 10 Suppose that either (a) l is bounded and j
|
is a stochastic
process that satises the following condition: there exists - 0 such that for all
c R
jro/
_
c _
o
r=|
j(:
r
)
(1 r)
r|
_ c -
:
|
_
< 1 - (4.84)
or (b) j
|
is a sequence of nondegenerate iid random variables with support
j
|
[a, [ with a _ 0 and < . Then, almost surely,
lim
|o
c
|
= (4.85)
lim
|o
a
|
= (4.86)
Note from (b) that Sotomayor (1984) needs no assumptions on the bounded-
ness of the utility and the level of variability of the income process (other that
it has to be nondegenerate) to prove the result, but needs the iid assumption.
Dispensing with the iid comes at the cost of having to make l bounded and
the income process suciently stochastic so that present discounted value of
future income leaves set [c, c-[ with probability of at least -. So unfortunately
the theorem does not apply to an economy with both serially correlated shocks
and standard utility functions (C11 or C1).
Note that one can show that condition (4.84) holds for income following a
nite-state Markov chain with H(j) 0 for all j and (j
t
[j) 0 for all j, j
t
. The
previous theorem implies that, under the appropriate conditions, the state space
for assets a
|
is unbounded. In particular, since with probability 1 asset holdings
explode, the demand for assets in a steady state of an economy composed of
agents with j = r (and facing the income uctuation problem discussed here)
is innite and thus no steady state equilibrium can exist in which the interest
rate, endogenously determined in general equilibrium, will satisfy j = r.
4.4.4 Case T = and > r
At this stage it will prove helpful to formulate the income uctuation problem
with borrowing constraints recursively. For the general problem in which income
follows a stationary, nite state Markov chain
6
the Bellman equation is
(a, j) = max
o
0
,c0
_
_
_
l(c)
1
1 j
0
(j
t
[j)(a
t
, j
t
)
_
_
_
(4.87)
s.t. c
a
t
1 r
= j a (4.88)
As rst order condition we obtain
l
c
(c) _
1 r
1 j
0
(j
t
[j)(a
t
, j
t
) (4.89)
= if a
t
0
6
We will deal with the case of nonstationary income below.
The envelope condition reads as
t
(a, j) = l
c
(c) (4.90)
so that the Euler equation once again becomes
l
c
(c) _
1 r
1 j
0
(j
t
[j)l
c
(c
t
) (4.91)
= if a
t
0 (4.92)
or
l
c
(c) = max
_
_
_
l
c
(j a),
1 r
1 j
0
(j
t
[j)l
c
(c
t
)
_
_
_
(4.93)
Both equations (4.87) and (4.08) can be used to compute optimal policy func-
tions a
t
(a, j) and c(a, j) as well as the value function (a, j) using exactly the
same iterative procedures as described in the last section, just with the modi-
cation that the Bellman equation has the additional constraint a
t
_ 0 and the
Euler equation has two parts now.
Income Process IID
If the income process is iid we can reduce the state space from two to one
dimension by introducing the variable cash at hand r = a j. The Bellman
equation becomes
(r) = max
0o
0
(l:)r
_
_
_
n
_
r
a
t
1 r
_
1
1 j
0
(j
t
)(a
t
j
t
)
_
_
_
(4.94)
with Euler equation
l
c
_
r
a
t
(r)
1 r
_
= max
_
_
_
l
c
(r),
1 r
1 j
0
(j
t
)l
c
_
a
t
(r) j
t
a
t
(r
t
)
1 r
_
_
_
_
(4.95)
= max
_
_
_
l
c
(r),
1 r
1 j
0
(j
t
)l
c
_
a
t
(r) j
t
a
t
[a
t
(r) j
t
[
1 r
_
_
_
_
Schechtman and Escudero (1977) provide the following characterizations of the
optimal policy function a
t
(r) for next periods asset holdings and for consump-
tion c(r). We rst consider the deterministic case.
Proposition 11 Let T = and H(j) = 1 (no uncertainty). Then there exists
an r such that if r _ r, then
r
t
= a
t
(r) j < r (4.96)
A liquidity constrained agent remains so forever, i.e. a
t
(j) = 0 and c(j) = j.
Consumption is strictly increasing in cash at hand, or
dc(r)
dr
0 (4.97)
There exists an r _ j such that a
t
(r) = 0 for all r _ r and a
t
(r) 0 for all
r r. Finally
Jc(r)
Jr
_ 1 and
Jo
0
(r)
Jr
< 1.
Proof. If a
t
(r) 0 then from envelope and FOC
t
(r) =
1 r
1 j
t
(r
t
)
<
t
(r
t
) (4.98)
since
l:
l
< 1 by our maintained assumption. Since is strictly concave (
inherits all the properties of l) we have r r
t
. Obviously, if a
t
(r) = 0 and
r j, then r
t
= j < r. Thus pick any r j.
For second part, suppose that a
t
(j) 0. Then from the rst order condition
and strict concavity of the value function
t
(j) =
1 r
1 j
t
(a
t
(j) j)
<
t
(a
t
(j) j)
<
t
(j) (4.99)
a contradiction. Hence a
t
(j) = 0 and c(j) = j.
The last part we also prove by contradiction. Suppose a
t
(r) 0 for all
r j. Pick arbitrary such r and dene the sequence r
|
o
|=0
recursively by
r
0
= r (4.100)
r
|
= a
t
(r
|l
) j _ j (4.101)
If there exists a smallest T such that r
T
= j then we found a contradiction,
since then a
t
(r
Tl
) = 0 and r
Tl
0. So suppose that r
|
j for all t. But
then a
t
(r
|
) 0 by assumption. Hence
t
(r
0
) =
1 r
1 j
t
(r
l
)
=
_
1 r
1 j
_
|
t
(r
|
)
<
_
1 r
1 j
_
|
t
(j)
=
_
1 r
1 j
_
|
n
t
(j) (4.102)
where the inequality follows from the fact that r
|
j and the strict concavity of
. the last equality follows from the envelope theorem and the fact that a
t
(j) = 0
so that c(j) = j.
But since
t
(r
0
) 0 and n
t
(j) 0 and
l:
l
< 1, we have that there exists
nite t such that
t
(r
0
)
_
l:
l
_
|
n
t
(j), a contradiction.
The proof that the derivatives of the policy functions have the asserted
properties follow from the strict concavity of the value function and the rst
order condition.
This last result bounds the optimal asset holdings (and hence cash at hand)
from above for T = . Since computational techniques usually rely on the
niteness of the state space we want to make sure that for our theory the
state space can be bounded from above. For the nite lifetime case there is no
problem. The most an agent can save is by consuming 0 in each period and
hence
a
|l
(r
|
) _ r
|
_ (1 r)
|l
a
0

|
=0
(1 r)
j (4.103)
which is bounded for any nite lifetime horizon T < .
The last theorem says that cash at hand declines over time or is constant at
j, in the case the borrowing constraint binds. The theorem also shows that the
agent eventually becomes credit-constrained: there exists a nite t such that
the agent consumes his endowment in all periods following t. This follows from
the fact that marginal utility of consumption has to increase at geometric rate
l:
l
if the agent is unconstrained (and thus consumption to decline) and from
the fact that once he is credit-constrained, he remains credit constrained forever.
This can be seen as follows. First r _ j by the credit constraint. Suppose that
a
t
(r) = 0 but a
t
(r
t
) 0. Since r
t
= a
t
(r) j = j we have that r
t
_ r. Thus
from the previous proposition a
t
(r
t
) _ a
t
(r) = 0 and hence the agent remains
credit-constrained forever.
For the innite lifetime horizon, under deterministic and constant income
we have a full qualitative characterization of the allocation: If a
0
= 0 then the
consumer consumes his income forever from time 0. If a
0
0, then cash at hand
and hence consumption is declining over time, and there exists a time t(a
0
) such
that for all t t(a
0
) the consumer consumes his income forever from thereon,
and consequently does not save anything.
We now consider the stochastic case with income being iid over time. The
proof of the following proposition is identical to the deterministic case. Remem-
ber the assumption that the minimum income level j
l
_ 0.
Proposition 12 Consumption is strictly increasing in cash at hand, i.e. c
t
(r)
(0, 1[. Optimal asset holdings are either constant at the borrowing limit or strictly
increasing in cash at hand, i.e. a
t
(r) = 0 or
Jo
0
(r)
Jr
(0, 1)
It is obvious that a
t
(r) _ 0 and hence r
t
(r, j
t
) = a
t
(r) j
t
_ j
l
so we have
j
l
0 as a lower bound on the state space for r. We now show that there is a
level r j
l
for cash at hand such that for all r _ r we have that c(r) = r and
a
t
(r) = 0
Proposition 13 There exists r _ j
l
such that for all r _ r we have c(r) = r
and a
t
(r) = 0
Proof. Suppose, to the contrary, that a
t
(r) 0 for all r _ j
l
. Then, using
the rst order condition and the envelope condition we have for all r _ j
l
(r) =
1 r
1 j
1
t
(r
t
) _
1 r
1 j
t
(j
l
) <
t
(j
l
) (4.104)
Picking r = j
l
yields a contradiction.
Hence there is a cuto level for cash at hand below which the consumer
consumes all cash at hand and above which he consumes less than cash at hand
and saves a
t
(r) 0. So far the results are strikingly similar to the deterministic
case. Unfortunately here it basically ends, and therefore our analytical ability to
characterize the optimal policies. In particular, the very important proposition
showing that there exists r such that if r _ r then r
t
< r does not go through
anymore, which is obviously quite problematic for computational considerations.
In fact we state, without a proof, a result due to Schechtman and Escudero
(1977).
Proposition 14 Suppose the period utility function is of constant absolute risk
aversion form n(c) = c
c
, then for the innite life income uctuation problem,
if H(j = 0) 0 we have r
|
almost surely.
Proof. See Schechtman and Escudero (1977), Lemma 3.6 and Theorem 3.7
Fortunately there are fairly general conditions under which one can, in fact,
prove the existence of an upper bound for the state space. Again we will refer to
Schechtman and Escudero for the proof of the following results. Intuitively why
would cash at hand go o to innity even if the agents are impatient relative to
the market interest rate, i.e. even if ,(1r) < 1? If agents are very risk averse,
face borrowing constraints and a positive probability of having very low income
for a long time, they may nd it optimal to accumulated unbounded funds over
time to self-insure against the eventuality of this unlikely, but very bad event
to happen. It turns out that if one assumes that the risk aversion of the agent
is suciently bounded, then one can rule this out.
Proposition 15 Suppose that the marginal utility function has the property
that there exist nite c
u
0 such that
lim
co
(log
c
n
t
(c)) = c
u
0 (4.105)
Then there exists a r such that r
t
= a
t
(r) j
_ r for all r _ r.
Proof. See Schechtman and Escudero (1977), Theorems 3.8 and 3.9
The number c
u
0 is called the asymptotic exponent of n
t
. Note that if the
utility function is of CRRA form with risk aversion parameter o, then since
log
c
c
c
= o log
c
c = o (4.106)
we have c
u
0 = o and hence for these utility function the previous proposition
applies. Also note that for CARA utility function
log
c
c
c
= c log
c
c =
c
ln(c)
(4.107)
lim
co
c
ln(c)
= (4.108)
and hence the proposition does not apply.
So under the proposition of the previous theorem we have the result that
cash at hand stays in the bounded set A = [j
l
, r[.
7
Consumption equals cash
at hand for r _ r and is lower than r for r r, with the rest being spent on
capital accumulation a
t
(r) 0. Figure 4.4.4 below shows the situation for the
case in which income can take only two possible realizations 1 = j
l
, j
.
45 degree line
45 degree line
_ ~
y x x x
1
y
1
y
N
a(x)
c(x)
x=a(x)+y
N
x=a(x)+y
1
7
If a
0
= o
0
+j
0
happens to be bigger than ~ a, then pick ~ a
0
= a
0
.
Income Serially Correlated, but Stationary
Now consider the case where income is correlated over time and follows a Markov
chain with transition . Now the trick of reducing the state to the single variable
cash at hand does not work anymore. This was only possible since current in-
come j and past saving a entered additively in the constraint set of the Bellman
equation, but neither variable appeared separately. With serially correlated
income, however, current income inuences the probability distribution of fu-
ture income. There are several possibilities of choosing the state space for the
Bellman equation. One can use cash at hand and current income, (r, j), or
asset holdings and current income (a, j). Obviously both ways are equivalent
and I opted for the later variant, which leads to the functional equation (4.04).
What can we say in general about the properties of the optimal policy functions
a
t
(a, j) and c(a, j)? Huggett (1993) proves a proposition similar to the ones
above showing that c(a, j) is strictly increasing in a and that a
t
(a, j) is con-
stant at the borrowing limit or strictly increasing (which implies a cuto a(j)
as before, which now will depend on current income j). What turns out to be
very dicult to prove is the existence of an upper bound of the state space,
a such that a
t
(a, j) _ a if a _ a. Huggett proves this result for the special
case that income can only take two states j j
l
, j
|
with 0 < j
l
< j
|
and
(j
|
[j
|
) _ (j
|
[j
l
), and CRRA utility. See his Lemmata 1-3 in the appendix.
I am not aware of any more general result for the non-iid case. With respect to
computation in more general cases, we have to cross our ngers and hope that
a
t
(a, j) eventually (i.e. for nite a) crosses the 4
0
-line for all j.
Chapter 5
PILCH Models in General
Equilibrium
In this section we will look at general equilibrium versions of the PILCH model
discussed in the last chapter. In general equilibrium the interest rate(s) and
real wage(s) are determined endogenously within the model. Three reasons for
considering general equilibrium come to my mind
1. It imposes theoretical discipline. As we saw above, the behavior of con-
sumption and saving over time depends crucially on the relative size of
the interest rate r and the time discount factor j. In the partial equilib-
rium analysis so far both r and j were exogenous parameters, to be chosen
by the model builder. In general equilibrium, once j is chosen, r will be
determined endogenously. The relationship between j and r shifts from
being arbitrarily chosen to an equilibrium relationship; as we will see, only
a stationary equilibrium with r < j can exist in these models.
2. It gives rise to an endogenously determined consumption and wealth distri-
bution and hence provides a theory of wealth inequality. To be successful,
this model should be able to reproduce stylized facts of empirical wealth
distributions. Note that it is not a theory of the income distribution, as
the stochastic income process is an input to the model (and needs to be
chosen by the model builder), rather than a result of the model.
3. It enables meaningful policy experiments. Partial equilibrium models,
by keeping interest rates and wages xed when analyzing the change in
household behavior due to changes in particular policies (tax reform, social
security reform, welfare reform), may over- or understate the full eects
of such a reform. A claim, e.g., that a reform of the social security system
towards a system that invests more funds in the stock market is welfare
improving may ignore the fact that, once large additional funds are in-
vested in the stock market, the excess return of stocks over bonds (or the
75
76 CHAPTER 5. PILCH MODELS IN GENERAL EQUILIBRIUM
population growth rate plus the growth rate of productivity), may dimin-
ish or vanish altogether. Only in cases in which one can reasonably expect
relative prices to remain uninuenced by policy reforms (for a small open
economy, say) or one can convincingly argue that price eects are quanti-
tatively unimportant
1
would a partial equilibrium analysis yield unbiased
results.
5.1 A Model Without Aggregate Uncertainty
5.1.1 The Environment
But let us leave the realm of ideology and describe the model. The economy is
populated by a continuum of measure 1 of individuals that all face an income
uctuation problem of the sort described above. Each individual has the same
stochastic labor endowment process j
|
o
|=0
where j
|
1 = j
l
, j
2
, . . . j
. A
households labor income in a particular period is given by n
|
j
|
where n
|
is
the real wage that one unit of labor commands in the economy, which will be
constant in a stationary equilibrium. Hence j
|
can be interpreted as the number
of eciency units of labor a household can supply in a given time period. The
labor endowment process is assumed to follow a stationary Markov process. Let
(j
t
[j) denote the probability that tomorrows endowment takes the value j
t
if todays endowment takes the value j. We assume a law of large numbers to
hold: not only is (j
t
[j) the probability of a particular agent of a transition
from j to j
t
but also the deterministic fraction of the population that has this
particular transition.
2
Let H denote the stationary distribution associated with
, assumed to be unique. We assume that at period 0 the income of all agents,
j
0
, is given, and that the distribution of incomes across the population is given
by H. Given our assumptions, then, the distribution of income in all future
periods is also given by H. In particular, the total labor endowment in the
economy (in eciency units) is given by
1 =
jH(j) (5.1)
As before, let denote by
|
(j
|
[j
0
) the probability of event history j
|
, given initial
event j
0
. We have
|
(j
|
[j
0
) = (j
|
[j
|l
) + . . . + (j
l
[j
0
) (5.2)
Hence, although there is substantial idiosyncratic uncertainty about a particu-
lar individuals labor endowment and hence labor income, the aggregate labor
endowment and hence labor income in the economy is constant over time, i.e.
1
It is not clear to me, though, how one would arrive at such a conclusion without actually
rst doing the general equilibrium analysis.
2
Whether and under what conditions we can assume such a law of large numbers created
a heated discussion among theorists. See Judd (1985), Feldman and Gilles (1985) and Uhlig
(1996) for further references.
5.1. A MODEL WITHOUT AGGREGATE UNCERTAINTY 77
there is no aggregate uncertainty. Without this assumption there would be no
hope for the existence of a stationary equilibrium in which wages n and interest
rates r are constant over time. In the next section we will consider a model that
is similar to this one, but will include one source of aggregate uncertainty, in
addition to the idiosyncratic labor income uncertainty present in this model.
Each agents preferences over stochastic consumption processes are given by
n(c) = 1
0
o
|=0
,
|
l(c
|
) (5.3)
with , =
l
l
and j 0. As before the agent can self-insure against idiosyn-
cratic labor endowment shocks by purchasing at period t uncontingent claims
to consumption at period t 1 at a price
|
=
l
l:t+1
. Again, in a stationary
equilibrium r
|l
will be constant across time. The agents budget constraint is
given by
c
|
a
|l
= n
|
j
|
(1 r
|
)a
|
(5.4)
Note that we slightly change the way assets are traded: instead of zero
coupon bonds being traded at a discount =
l
l:
we now consider a bond that
trades at price 1 today and earns gross interest rate (1 r
|l
) tomorrow. We
do this for the following reason: the asset being traded will be physical capital,
with the interest rate being determined by the marginal product of capital. As
long as the interest rate is constant as in Aiyagari (1994) both formulations
are equivalent (if you derive the corresponding Euler equations for both for-
mulations, youll nd that theyre identical). With aggregate uncertainty as
in Krusell and Smith (1998), however, it would make a substantial dierence
whether agents can trade a risk free bond or, as Krusell and Smith assume,
risky capital. Thus, in order to be consistent with their formulation, I opted for
changing the budget constraint. Evidently all theoretical results from the last
section about the income uctuation problem still apply (as long as the interest
rate is nonstochastic), because the optimality conditions for the households are
identical across both formulations.
We impose an exogenous borrowing constraint on asset holdings: a
|l
_ 0.
Aiyagari (1994) considers several alternative borrowing constraints, but uses the
no-borrowing constraint in his applications. The agent starts out with initial
conditions (a
0
, j
0
). His consumption at period t after endowment shock history
j
|
has been realized is denoted by c
|
(a
0
, j
|
) and his asset holdings by a
|l
(a
0
, j
|
).
Let 1
0
(a
0
, j
0
) denote the initial measure over (a
0
, j
0
) across households. In ac-
cordance with our previous assumption the marginal distribution of 1
0
with
respect to j
0
is assumed to be H. At each point of time an agent is character-
ized by her current asset position a
|
and her current income j
|
. These are her
individual state variables. What describes the aggregate state of the economy
is the cross-sectional distribution over individual characteristics 1
|
(a
|
, j
|
). This
concludes the description of the household side of the economy.
On the production side we assume that competitive rms, taking as given
wages n
|
and interest rates r
|
, have access to a standard neoclassical production
technology
3
1
|
= 1(1
|
, 1
|
) (5.5)
where 1 C
2
features constant returns to scale and positive but diminishing
marginal products with respect to both production factors. We also assume the
Inada conditions to hold. Firms choose labor inputs and capital inputs to max-
imize the present discounted value of prots. As usual with constant returns to
scale and perfect competition the number of rms is indeterminate and without
loss of generality we can assert the existence of a single, representative rm. In
this economy the only asset is the physical capital stock. Hence in equilibrium
the aggregate capital stock 1
|
has to equal the sum of asset holdings of all
individuals, i.e. the integral over the a
|
s of all agents. We assume that capital
depreciates at rate 0 < c < 1. The aggregate resource constraint is then given
as
C
|
1
|l
(1 c)1
|
= 1(1
|
, 1
|
) (5.6)
where C
|
is aggregate consumption at period t.
We are now ready to dene an equilibrium. We rst dene a sequential mar-
kets equilibrium, then a recursive equilibrium and nally a stationary recursive
equilibrium. The rst denition is done in order to stress similarities and dier-
ences with the complete markets model, the third because this is what we will
compute numerically, and the second is presented because the third is a special
case, and with aggregate uncertainty we will need the more general denition
anyway.
Denition 16 Given 1
0
, a sequential markets competitive equilibrium is allo-
cations for households c
|
(a
0
, j
|
), a
|l
(a
0
, j
|
)
o
|=0,
t
Y
t
, allocations for the rep-
resentative rm 1
|
, 1
|
o
|=0
, and prices n
|
, r
|
o
|=0
such that
1. Given prices, allocations maximize (.8) subject to (.4) and subject to
the nonnegativity constraints on assets and consumption and the non-
negativity condition on assets
2.
r
|
= 1
|
(1
|
, 1
|
) c (5.7)
n
|
= 1
J
(1
|
, 1
|
) (5.8)
3
In the next section aggregate uncertainty is introduced by incorportating stochastic pro-
ductivity into the aggregate production function, in the tradition of real business cycles.
3. For all t
1
|l
=
_

t
Y
t
a
|l
(a
0
, j
|
)(j
|
[j
0
)d1
0
(a
0
, j
0
) (5.9)
1
|
=

1 =
_

t
Y
t
j
|
(j
|
[j
0
)d1
0
(a
0
, j
0
) (5.10)
_

t
Y
t
c
|
(a
0
, j
|
)(j
|
[j
0
)d1
0
(a
0
, j
0
) 1
|l
= 1(1
|
, 1
|
) (1 c)1
|
(5.11)
The last three conditions are the asset market clearing, the labor market
clearing and the goods market clearing condition, respectively
Now let us dene a recursive competitive equilibrium. We have already
conjectured what the correct state space is for our economy, with (a, j) being
the individual state variables and 1(a, j) being the aggregate state variable.
First we need to dene an appropriate measurable space on which the measures
1 are dened. Dene the set = [0, ) of possible asset holdings and the
set 1 of possible labor endowment realizations. Dene by T(1 ) the power set
of 1 (i.e. the set of all subsets of 1 ) and by E() the Borel o-algebra of .
Let 7 = 1 and E(7) = T(1 ) E(). Finally dene by / the set of all
probability measures on the measurable space ' = (7, E(7)). Why all this?
Because our measures 1 will be required to be elements of /. Now we are
ready to dene a recursive competitive equilibrium. At the heart of any RCE
is the recursive formulation of the household problem. Note that we have to
include all state variables in the household problem, in particular the aggregate
state variable, since the interest rate r will depend on 1. Hence the household
problem in recursive formulation is
(a, j; 1) = max
c0,o
0
0
n(c) ,

0
Y
(j
t
[j)(a
t
, j
t
; 1
t
) (5.12)
s.t. c a
t
= n(1)j (1 r(1))a (5.13)
1
t
= H(1) (5.14)
The function H : // is called the aggregate law of motion. Now let us
proceed to the equilibrium denition.
Denition 17 A recursive competitive equilibrium is a value function : 7
/1, policy functions for the household a
t
: 7/1 and c : 7/1,
policy functions for the rm 1 : / 1 and 1 : / 1, pricing functions
r : /1 and n : /1 and an aggregate law of motion H : // such
that
1. , a
t
, c are measurable with respect to E(7), satises the households
Bellman equation and a
t
, c are the associated policy functions, given r()
and n()
2. 1, 1 satisfy, given r() and n()
r(1) = 1
|
(1(1), 1(1)) c (5.15)
n(1) = 1
J
(1(1), 1(1)) (5.16)
3. For all 1 /
1(H(1)) =
_
a
t
(a, j; 1)d1 (5.17)
1(1) =
_
jd1 (5.18)
_
c(a, j; 1)d1
_
a
t
(a, j; 1)d1 (5.19)
= 1(1(1), 1(1)) (1 c)1(1)
4. The aggregate law of motion H is generated by the exogenous Markov
process and the policy function a
t
(as described below)
Now let us specify what it means that H is generated by and a
t
. H basically
tells us how a current measure over (a, j) translates into a measure 1
t
tomorrow.
So H has to summarize how individuals move within the distribution over assets
and income from one period to the next. But this is exactly what a transition
function tells us. So dene the transition function Q
+
: 7 E(7) [0, 1[ by
4
Q
+
((a, j), (/, )) =

_
(j
t
[j) if a
t
(a, j; 1) /
0 else
(5.20)
for all (a, j) 7 and all (/, ) E(7). Q
+
((a, j), (/, )) is the probability
that an agent with current assets a and current income j ends up with assets
a
t
in / tomorrow and income j
t
in tomorrow. Suppose that is a singleton,
say = j
l
. The probability that tomorrows income is j
t
= j
l
, given todays
income is (j
t
[j). The transition of assets is non-stochastic as tomorrows assets
are chosen today according to the function a
t
(a, j). So either a
t
(a, j) falls into
/ or it does not. Hence the probability of transition from (a, j) to j
l
/ is
(j
t
[j) if a
t
(a, j) falls into / and zero if it does not fall into /. If contains
more than one element, then one has to sum over the appropriate (j
t
[j).
How does the function Q
+
help us to determine tomorrows measure over
(a, j) from todays measure? Suppose Q
+
were a Markov transition matrix for
a nite state Markov chain and 1
|
would be the distribution today. Then to
gure out the distribution 1
|
tomorrow we would just multiply Q by 1
|
, or
1
|l
= Q
T
+t
1
|
(5.21)
where here T stands for the transpose of a matrix. But a transition function is
just a generalization of a Markov transition matrix to uncountable state spaces.
4
Note that, since o
0
is also a function of , Q is implicitly a function of , too.
Hence we need integrals:
1
t
(/, ) = (H(1)) (/, ) =
_
Q
+
((a, j), (/, ))1(da dj) (5.22)
The fraction of people with income in and assets in / is that fraction of
people today, as measured by 1, that transit to (/, ), as measured by Q
+
.
In general there no presumption that tomorrows measure 1
t
equals todays
measure, since we posed an arbitrary initial distribution over types, 1
0
. If the
sequence of measures 1
|
generated by 1
0
and H is not constant, then obvi-
ously interest rates r
|
= r(1
|
) are not constant, decision rules vary with 1
|
over
time, and the computation of equilibria is dicult in general. We would have to
compute the function H explicitly, mapping measures (i.e. innite-dimensional
objects) into measures. This is exactly what we are going to discuss in the next
section. In this section we are interested in stationary Recursive Competitive
Equilibria.
Denition 18 A stationary recursive competitive equilibrium is a value func-
tion : 7 1, policy functions for the household a
t
: 7 1 and c : 7 1,
policies for the rm 1, 1, prices r, n and a measure 1 / such that
1. , a
t
, c are measurable with respect to E(7), satises the households
Bellman equation and a
t
, c are the associated policy functions, given r and
n
2. 1, 1 satisfy, given r and n
r = 1
1
(1, 1) c (5.23)
n = 1
J
(1, 1) (5.24)
3.
1 =
_
a
t
(a, j)d1 (5.25)
1(1) =
_
jd1 (5.26)
_
c(a, j)d1
_
a
t
(a, j)d1 = 1(1, 1) (1 c)1 (5.27)
4. For all (/, ) E(7)
1(/, ) =
_
Q((a, j), (/, ))d1 (5.28)
where Q is the transition function induced by and a
t
as described above
(not indexed by 1 anymore)
Note the big simplication: value functions, policy functions and prices are
not any longer indexed by measures 1, all conditions have to be satised only
for the equilibrium measure 1. The last requirement states that the measure
1 reproduces itself: starting with measure over incomes and assets 1 today
generates the same measure tomorrow. In this sense a stationary RCE is the
equivalent of a steady state, only that the entity characterizing the steady state
is not longer a number (the aggregate capital stock, say) but a rather compli-
cated innite-dimensional object, namely a measure.
5.1.2 Theoretical Results: Existence and Uniqueness
In this section we want to summarize what we know about the existence and
uniqueness of a stationary RCE. We rst argue that the problem of existence
boils down to the question of whether one equation in one unknown, namely
the interest rate, has a solution. To see this, rst note that by Walras law
one of the market clearing conditions is redundant; so lets ignore the goods
market equilibrium condition. The labor market equilibrium is easy and gives
us 1 =

1 and

1 is exogenously given, specied by the labor endowment process.
It remains the asset market clearing condition
1 = 1(r) =
_
a
t
(a, j)d1 = 1a(r) (5.29)
where 1a(r) are the average asset holdings in the economy. This condition
requires equality between the demand for capital by rms and the supply of
capital by households (last periods demand for assets, with physical capital
being the only asset in the economy).
From (.28) it is clear that the capital demand of the rm 1(r) is a function
of r alone, dened implicitly as
r = 1
|
(1(r),

1) c (5.30)
since labor supply 1 =

1 0 is exogenous. Furthermore we know from the
assumptions on the production function that 1(r) is a continuous, strictly de-
creasing function on r (c, ) with
lim
:o
1(r) = (5.31)
lim
:o
1(r) = 0 (5.32)
It is obvious that the average capital supply (asset demand) by households 1a(r)
satises 1a(r) [0, [ for all r in (c, ). It is our goal to characterize 1a(r).
In particular, if 1a(r) is continuous and satises
lim
:o
1a(r) < (5.33)
lim
:o
1a(r) 0 (5.34)
then a stationary recursive competitive equilibrium exists. Furthermore, if
1a(r) is strictly increasing in r (the substitution eects outweighs the income
eect), then the stationary RCE is unique.
The rst thing we have to argue is that average capital supply (asset demand)
is in fact a function of r alone and that it is well dened on the appropriate
range for r. First we note that the wage rate n is a function solely of r via
n(r) = 1
J
(1(r),

1) (5.35)
As r increases, 1(r) decreases, the capital-labor ratio declines and with it the
wage rate. Hence, once r is known, the recursive problem of the household can
be solved, because all prices are known.
The rst step is to ask under what conditions the recursive problem of the
household has a solution and under which conditions the support of asset hold-
ings is bounded from above (it is bounded from below by 0). Various assumptions
give us these results, but here are the most important ones (note that Aiyagari
(1994) does not prove anything for the case of serially correlated income and
assumes bounded utility, Huggett (1993) does, but with additional assumptions
on the structure of the Markov process).
Proposition 19 (Huggett 1993) For j 0, r 1 and j
l
0 and C11
utility with o 1 the functional equation has a unique solution which is strictly
increasing, strictly concave and continuously dierentiable in its rst argument.
The optimal policies are continuous functions that are strictly increasing (for
c(a, j)) or increasing or constant at zero (for a
t
(a, j))
Similar results can be proved for the iid case and arbitrary bounded l with
j r and j 0, see Aiyagari (1994).
With respect to the boundedness of the state space, as seen in the previous
section we require j r and additional assumptions. Under these assumptions
there exists an a s.t. a
t
( a, j
) = a and a
t
(a, j) _ a for all j 1 and all
a [0, a[ = . From now on we will restrict the state space to 1 and it will
be understood that 7 = 1. Thus, under the maintained assumptions by
Huggett or Aiyagari there exists an optimal policy a
t
:
(a, j) indexed by r. The
next step is to ask what more is needed to make aggregate asset demand
1a(r) =
_
a
t
:
(a, j)d1
:
(5.36)
well-dened, where 1
:
has to satisfy
1
:
(/, ) =
_
Q
:
((a, j), (/, ))d1
:
(5.37)
and Q
:
is the Markov transition function dened by a
:
as
Q
:
((a, j), (/, )) =

_
(j
t
[j) if a
t
:
(a, j) /
0 else
(5.38)
Hence all is needed for 1a(r) to be well-dened is to establish that the operator
T
+
:
: // dened by
(T
+
:
(1)) (/, ) =
_
Q
:
((a, j), (/, ))d1 (5.39)
has a unique xed point (that T
+
:
maps /into itself follows from SLP, Theorem
8.2). To show this Aiyagari (in the working paper version, and quite loosely
described) draws on a theorem in SLP and Huggett on a similar theorem due
to Hopenhayn and Prescott (1992). In both theorems the key condition is a
monotone mixing condition that requires a positive probability to go from the
highest asset level a to a intermediate asset level in periods and an evenly
high probability to go from 0 assets to an intermediate asset level also in
periods. More precisely stated, the theorem by Hopenhayn and Prescott states
the following. Dene the order _ " on 7 as
. _ .
t
i [(.
l
_ .
t
l
and .
2
= .
t
2
) or (.
t
= c = (0, j
l
)) or (. = d = ( a, j
))
(5.40)
Under this order it is easy to show that (7, _) is an ordered space, 7 together
with the Euclidean metric is a compact metric space, _ is a closed order, c 7
and d 7 are the smallest and the largest elements in 7 (under order _) and
(7, E(7)) is a measurable space. Then we have (see Hopenhayn and Prescott
(1992), Theorem 2)
Proposition 20 If
1. Q
:
is a transition function
2. Q
:
is increasing
3. There exists .
+
7, - 0 and such that
1
(d, . : . _ .
+
) - and 1
(c, . : . _ .
+
) - (5.41)
Then the operator T
+
:
has a unique xed point 1
:
and for all 1
0
/ the
sequence of measures dened by
1
n
= (T
+
)
n
1
0
(5.42)
converges weakly to 1
:
Here 1
., ?) is the probability of going from state . to set ? in steps.

Instead of proving this result (which turns out to be quite tough) we will explain
the assumptions, heuristically verify them and discuss what the theorem delivers
for us. Assumption 1 requires that Q
:
is in fact a transition function, i.e.
Q
:
(., .) is a probability measure on (7, E(7)) for all . 7 and Q
:
(., ?) is a
E(7)-measurable function for all ? E(7). Given that a
t
(a, j) is a continuous
function, the proof of this is not too hard. The assumption that Q
:
is increasing
means that for any nondecreasing function ) : 7 R we have that
(T)) (.) =
_
)(.
t
)Q
:
(., d.
t
) (5.43)
is also nondecreasing. The proof that Q
:
satises monotonicity is straight-
forward, given that a
t
(a, j) is increasing in both its arguments
5
, so that big-
ger .s make Q
:
(., .) put more probability mass on bigger .
t
. Together with
) being nondecreasing the result follows. Finally, why is the monotone mix-
ing condition 8. satised? Pick (this is going to be very heuristic here) .
+
=
(
l
2
(a
t
(0, j
) a) , j
l
). Starting at d, with a sequence of bad shocks j
l
one con-
verges to assets 0 monotonically and from c one converges to a with a sequence
of good shocks j
. Hence with probability bigger than zero one goes with nite
steps
l
or less from d to something below .
+
and with nite steps
2
or less
from c to something above .
+
. Take = max
l
,
2
.
The conclusion of the theorem then assures the existence of a unique invari-
ant measure 1
:
which can be found by iterating on the operator T
+
:
. Convergence
is in the weak sense, that is, a sequence of measures 1
n
converges weakly to
1
:
if for every continuous and bounded real-valued function ) on 7 we have
lim
no
_
)(.)d1
n
=
_
)(.)d1
:
(5.44)
The argument in the preceding section demonstrated that the function 1a(r)
is well-dened on r [c, j). Since a
t
:
(a, j) is a continuous function jointly in
(r, a), see SLP, Theorem 3.8 and 1
:
is continuous in r (in the sense of weak con-
vergence), see SLP, Theorem 12.13, the function 1a(r) is a continuous function
of r on [c, j). Note that the real bottleneck in the argument is in establishing
an upper bound of the state space for assets, as 7 needs to be compact for the
theorem by Hopenhayn and Prescott to work. To bound the state space an
interest rate r < j is needed, as the previous sections showed.
The remaining things to be argued are that lim
:o
1a(r) < and lim
:
1a(r)
1(j). The rst condition is obviously satised whenever a bound on the state
space for assets, a, can be found. So this part of the problem is easily resolved.
We also know from the previous section that for j = r all consumers accumulate
innite assets eventually, so that, loosely speaking 1a(j) = . What is asserted
(but not proved) by Aiyagari is that as r approaches j from below, 1a(r) goes
to by some form of continuity argument (but note that, strictly speaking,
1a(j) is not well-dened).
This is a so far missing step in the existence proof. Otherwise we have
established that both 1(r) and 1a(r) are continuous functions on r (c, j),
that for low r we have that 1a(r) < 1(r) and for high r < j we have 1a(r)
1(r). These arguments together then guarantee the existence of r
+
such that
1(r
+
) = 1a(r
+
) (5.45)
5
For o
0
(o, j) to be increasing in j we need to assume that the exogenous Markov chain
does not feature negatively correlated income.
and hence the existence of a stationary recursive competitive equilibrium.
With respect to uniqueness the verdict is negative, as we cant prove the
monotonicity of the function 1a(r) which is due to osetting income and sub-
stitution eects of saving with respect to the interest rate directly, as well as
the indirect eect that the interest rate has on the wage rate. Even harder is
the question about the stability of the stationary equilibrium. We know that,
provided the economy starts with initial distribution 1
0
= 1
:
, then the econ-
omy remains there forever. The question arises whether, for an arbitrary initial
distribution 1
0
,= 1
:
it is the case that 1
|
1
:
and r
|
r
+
(either locally
or globally). Note that to assess this question one has to examine either the
sequential equilibrium or the law of motion H in the full-blown RCE. To the
best of my knowledge no stability result has been established for these types of
economies as of today.
5.1.3 Computation of the General Equilibrium
Since nding a stationary RCE really amounts to nding an interest rate r
+
that clears the asset market, consider the following algorithm
1. Fix an r (c, j). For a xed r we can solve the households recursive
problem (e.g. by value function iteration or policy function iteration).
This yields a value function
:
and decision rules a
t
:
, c
:
, which obviously
depend on the r we picked.
2. The policy function a
t
:
and induce a Markov transition function Q
:
.
Compute the unique stationary measure 1
:
associated with this transition
function from (.72)
3. Compute excess demand for capital
d(r) = 1(r) 1a(r) (5.46)
Note that d(r) is just a number. If this number happens to equal zero, we
are done and have found a stationary RCE. If not we update our guess for
r and start from 1. anew. If 1a(r) is increasing in r (which means that
the substitution eect outweighs the income eect and potential indirect
income eects from changes in wages) we know that d(r) is a strictly
decreasing function. The updating should then be such that if d(r) 0,
increase the guess for r, if d(r) < 0 decrease the guess for r.
5.1.4 Qualitative Results
One general qualitative result of these models is excess saving and the overaccu-
mulation of capital, compared to a complete markets model in which a full array
of insurance contracts against idiosyncratic income uncertainty is at the disposal
of households. Remember that equilibria in this model are Pareto ecient, from
which it readily follows that the stationary equilibrium in the present model (or
any equilibrium, for that matter) is suboptimal (in an ex-ante welfare sense, of
course). As seen in Chapter 2, without aggregate uncertainty consumption of
all agents would be constant in any steady state of a complete markets model.
From the complete markets Euler equation it follows right away that the station-
ary equilibrium interest rate with complete markets satises r
c1
= j. But this
means, since r
+
< j that the steady state capital stock under complete markets
satises 1
c1
< 1
+
, i.e. there is overaccumulation of capital in this economy,
compared to the rst-best. Put another way, since total gross investment (and
total saving) in the steady state of both economies equal
o
c1
= c1
c1
(5.47)
o
+
= c1
+
c1
c1
= o
c1
(5.48)
i.e. agents in this model oversave because of precautionary reasons. Only po-
tentially binding liquidity constraints are needed for this result (in non of the
discussion above did we ever assume l
ccc
0), with prudence strengthening the
quantitative importance of the result. One of the main objectives of Aiyagaris
quantitative analysis is to assess whether this excess precautionary saving is
quantitatively important, i.e. whether PILCH people, in general equilibrium,
accumulate signicantly too many assets compared to the benchmark complete
markets model. As metric for this the aggregate savings rate and real inter-
est rate is used. Note for future reference that the aggregate saving rate in a
stationary equilibrium is given by
: =
o
1
= c
1
1
(5.49)
5.1.5 Numerical Results
Calibration
Before solving the model Aiyagari has to specify functional forms of preferences
and technology. Period utility is assumed to be C11 and Aiyagari experi-
ments with values o = 1, 8, . The model period length is taken to be one
year, and he picks j = 0.0416 (, = 0.06). For the complete markets model
this yields an (empirically reasonable) annual real interest rate of 4.16/, and
we know that in the present model the interest rate will be lower.
The aggregate production function is assumed to take Cobb-Douglas form
with share parameter c = 0.86. Since c equals the capital share of income
and capital income as share of GDP is roughly 86/ on average for US postwar
data (the number is somewhat sensitive to how the public capital stock and
consumer durables are treated). The annual depreciation rate is set at c = 8/,
values usually used in these types of calibration exercises (Cooley and Prescott
(1995) provide a careful description of the calibration procedure for a standard
RBC model).
Compared to standard RBC theory the only nonstandard input the model
requires is the stochastic labor income process, where we note that labor income
is just labor endowment scaled by a constant. Aiyagari poses the following
1(1) process for the natural logarithm of labor earnings
log(j
|l
) = 0 log(j
|
) o
:
_
1 0
2
_
1
2
-
|l
(5.50)
Note that for this process we nd
corr(log(j
|l
), log(j
|
)) =
Co(log(j
|l
), log(j
|
))
_
\ ar(log(j
|
)) + \ ar(log(j
|l
))
= 0 (5.51)
\ ar(log(j
|l
)) = o
2
:
(5.52)
For this income process log(j
|
) can take any value in (, ). Our theory
developed so far has assumed a nite state space for j
|
1 or equivalently
1
Iog
for log(j
|
). The transformation of processes with continuous state space
into nite state Markov chains was pioneered in economics by George Tauchen
(1986) and roughly goes like this. First we pick the nite set 1
Iog
. Aiyagari
picks the number of states to be 7. Since log(j
|
) can take any value between
(, ), rst subdivide the real line into 7 intervals
1
l
= (,
2
o
:
)
1
2
= [
2
o
:
,
8
2
o
:
)
1
3
= [
8
2
o
:
,
1
2
o
:
)
1
d
= [
1
2
o
:
,
1
2
o
:
)
1
5
= [
1
2
o
:
,
8
2
o
:
)
1
6
= [
8
2
o
:
,

2
o
:
)
1
7
= [
2
o
:
, ) (5.53)
and take as state space for log-income the set of midpoints of the intervals
1
Iog
= 8o
:
, 2o
:
, o
:
, 0, o
:
, 2o
:
, 8o
:
(5.54)
The matrix of transition probabilities is then determined as follows. Fix a state
:
I
= log(j) 1
Iog
today. The conditional probability of a particular state
:
= log(j
t
) 1
Iog
tomorrow is then computed by
(log(j
t
) = :
[ log(j) = :
I
) =
_
1
c
(olog())
2
2o
2
_
2o
dr (5.55)
where o
= o
:
_
1 0
2
_
1
2
. That is, we integrate the Normal distribution with
mean log(j) and variance o
2
:
_
1 0
2
_
over the interval 1
. Doing this for all

states today and tomorrow (this integration has to be done either numerically
or one has to use tables for the pdf of a Normal distribution) yields the transition
matrix . Now we can nd the stationary distribution of , hopefully unique,
by solving the matrix equation
H =
T
H (5.56)
for H. Given that the states for log(j) are given by 1
Iog
, we nd the state space
for levels of income as
1 = c
3cz
, c
2cz
, c
cz
, 1, c
cz
, c
2cz
, c
3cz
(5.57)
Now we compute the average labor endowment as
j =

Y
jH(j) (5.58)
and normalize all states by j to arrive at
1 = j
l
, . . . , j
7
(5.59)
c
3cz
j
,
c
2cz
j
,
c
cz
j
,
1
j
,
c
cz
j
,
c
2cz
j
,
c
3cz
j

and now the average (and aggregate) labor endowment equals
Y
jH(j) = 1. (5.60)
Of course, this normalization of aggregate labor supply to 1 only xes the units
of account for this economy and thus is completely innocuous. How good the
approximation of the continuous process with its nite discretization is depends
on the parameter values that one chooses for the persistence 0 and dispersion o
:
(and of course on the number of states allowed in the discretization). Although
Aiyagaris Table 1 seems to indicate that the approximation works ne for the
parameter values he considers, it is my experience that for high values of 0 (close
to 1) the discretization has a harder and harder time getting the persistence
correct, in addition to the dispersion being somewhat o already for 0 = 0.0.
Aiyagari computes stationary equilibria for parameter values in the range
0 0, 0.8, 0.6, 0.0 (5.61)
o
:
0.2, 0.4 (5.62)
Quite a few labor economists these days believe, however, that the persistence
parameter should be higher than 0 = 0.0, closer to 1 (what happens if 0 = 1?).
But then there is substantial debate whether the simple 1(1) process for the
log of labor earnings is the appropriate stochastic process to model this variable.
Instead let us turn to the results.
Results
First note that for the Cobb-Douglas production function and

1 = 1 we have
1 = 1
o
and
r c = c1
ol
(5.63)
Thus
r c =
c1
1
=
cc
:
(5.64)
where : is the aggregate saving rate. Thus
: =
cc
r c
(5.65)
and there is a one-to-one mapping between equilibrium interest rates r and
equilibrium aggregate savings rates :. For the benchmark case of complete mar-
kets we have r
c1
= j = 4.16/ and thus, given the other parameters chosen,
: = 28.7/. Aiyagaris Table II contains aggregate savings rates and interest
rates for the incomplete markets general equilibrium model under various as-
sumptions about risk aversion (also prudence) o, persistence 0 and dispersion
o
:
. The most important ndings are:
Keeping prudence and dispersion xed, an increase in the persistence of
the income shock leads to increased precautionary saving and bigger over-
accumulation of capital, compared to the complete markets benchmark.
Keeping xed persistence and dispersion in income, an increase in pru-
dence o leads to more precautionary saving and more severe overaccumu-
lation of capital
Keeping prudence and income persistence constant, an increase in the
dispersion of the income process leads to more precautionary saving and
more severe overaccumulation of capital.
Quantitatively, the biggest eect on precautionary saving seem to stem
from the persistence parameter. In particular, for high income persistence
(and high dispersion) the aggregate saving rate is 14 percentage points
higher than in the complete markets benchmark.
The other implications of the model that Aiyagari discusses, the welfare cost
of idiosyncratic income uctuations (as well as the welfare benet from access to
self-insurance), as well as the predictions of the model with respect to the wealth
and consumption distribution will be addressed once the model is enriched by
aggregate uctuations.
5.2. UNEXPECTEDAGGREGATE SHOCKS ANDTRANSITIONDYNAMICS91
5.2 Unexpected Aggregate Shocks and Transi-
tion Dynamics
In this section we consider hypothetical thought experiments of the following
form. Suppose the economy is in a stationary equilibrium, with a given govern-
ment policy and all other exogenous elements that dene preferences, endow-
ments and technology. Now, unexpectedly, either government policy or some
exogenous elements of the economy (such as the labor productivity process)
change. This change was completely unexpected by all agents of the economy
(a zero probability event), so that no anticipation actions were taken by any
agent. The exogenous change may be either transitory or permanent; for the
general discussion to follow this does not make a dierence. We want to study
the transition path induced by the exogenous change, from the old stationary
equilibrium to a new stationary equilibrium (which may coincide with the old
stationary equilibrium in case the exogenous change is of transitory nature, or
may dier from it in case the exogenous change is permanent.
For concreteness, suppose that the economy to start with is the standard
Aiyagari economy we studied in the previous section. As an example, we con-
sider as exogenous unexpected change the sudden permanent introduction of a
capital income tax at rate t. The receipts are rebated lump-sum to households
as government transfers T. The initial policy is characterized by t = T = 0. Ob-
viously, since a tax on capital income changes households savings decisions, we
expect that, due to the policy change, individual behavior and thus aggregate
variables such as the interest rate, wage rate and the capital stock change. We
would also hope that, over time, the economy settles down to its new station-
ary equilibrium associated with the capital income tax. But since the economy
starts, pre-reform, with an aggregate state (aggregate capital, wealth distrib-
ution) not equal to the nal stationary equilibrium, one would expect that it
requires time for the economy to settle down to its new stationary equilibrium.
In other words, there will be a nontrivial transition path induced by the reform.
5.2.1 Denition of Equilibrium
We now want to dene an equilibrium that allows for such transition path and
then outline how one would possibly compute such a transition path. As should
be clear from the previous discussion that no analytical characterization of the
transition path is available in general, so that we have to rely on computational
analysis.
Since, on the aggregate level, the transition path is characterized by a de-
terministic sequence of prices, quantities and distributions we will cast the de-
nition and solution of the model in sequential notation, with the household
decision problem still being formulated recursively. Let 7 = 1 R
be the set
of all possible (j
|
, a
|
). Let E(R
) be the Borel o-algebra of R
and T(1 ) be
the power set of 1 . Finally let E(7) = T(1 ) E(R
) and M be the set of all

nite measures on the measurable space (7, E(7)).
First lets write down the household problem
|
(a, j) = max
c0,o
0
0
n(c) ,

0
Y
(j
t
[j)
|l
(a
t
, j
t
)
s.t. c a
t
= n
|
j (1 (1 t
|
)r
|
)a T
|
Note that value functions are now functions of time also, since aggregate prices
and policies may change over time.
Denition 21 Given the initial distribution 1
0
, and scal legislation t
|
o
|=0
a competitive equilibrium is a sequence of individual functions for the house-
hold
|
, c
|
, a
|l
: 7 MR
o
|=0
, sequences of production plans for the rm
|
, 1
|
o
|=0
, factor prices n
|
, r
|
o
|=0
, government transfers T
|
o
|=0
, and a se-
quence of measures 1
o
|=l
such that, for all t,
1. (Maximization of Households): Given n
|
, r
|
and T
|
, t
|
the functions
|
solve Bellmans equation for period t and c
|
, a
|l
are the associated
policy functions
2. (Marginal Product Pricing): The prices n
|
and r
|
satisfy
n
|
= 1
J
(1
|
, 1
|
) (5.66)
r
|
= 1
1
(1
|
, 1
|
) c. (5.67)
3. (Government Budget Constraint):
T
|
= t
|
r
|
1
|
(5.68)
for all t _ 0.
4. (Market Clearing):
_
c
|
(j
|
, a
|
)d1
|
1
|l
= 1(1
|
, 1
|
) (1 c)1
|
(5.69)
1
|
=
_
j
|
d1
|
(5.70)
1
|l
=
_
a
|l
(j
|
, a
|
)d1
|
(5.71)
5. (Aggregate Law of Motion):
6
1
|l
= I
|
(1
|
) (5.74)
6
The functions t can be written explicitly as follows. Dene Markov transition functions
Qt : Z B(Z) [0, 1] induced by the transition probabilities and the optimal policy
o
t+1
(j, o) as
Qt((j, o), (], ,)) =

y
0
2Y
_
(j
0
[j) if o
t+1
(j, o) t
0 else
(5.72)
for all (j, o) Z and all (], ,) B(Z). Then
t+1
(], ,) = [t(t)] (], ,) =
_
Qt((j, o), (], ,))ot (5.73)
for all (], ,) B(Z).
Denition 22 A stationary equilibrium is an equilibrium such that all elements
of the equilibrium that are indexed by t are constant over time.
5.2.2 Computation of the Transition Path
We are interested in computing the following equilibrium. At time 0 the econ-
omy is in the stationary equilibrium associated with t
0
= 0 and associated
distribution 1
0
(from this distribution all other aggregate variables, such as
1
|
, r
|
, n
|
can easily be derived) and associated value function
0
and policy
functions c
0
, a
l
. Now at time t = 1 the policy changes permanently, t
|
= t 0
for all t _ 1. Let the new stationary equilibrium associated with t be denoted
by 1
o
,with associated value function
o
and policy functions c
o
, a
o
In the
previous section we discussed how to compute such stationary equilibria. Now
we want to compute the entire transition path and trace out the welfare conse-
quences of such a policy innovation.
The key idea is to assume that after T periods the transition from the old
to the new stationary equilibrium is completed. We will discuss below how to
choose T, but heuristically it should be large enough so that the economy had
enough time to settle down to the new stationary equilibrium. The key insight
then is to realize that if
T
=
o
, then for a given sequence of prices r
|
, n
|
T
|=l
the household problem can be solved backwards (note that this is true indepen-
dent of wether people live forever or not). This suggests the following algorithm.
Algorithm 23 1. Fix T
2. Compute the stationary equilibrium 1
0
,
0
, c
0
, a
0
, r
0
, n
0
, 1
0
associated with
t = t
0
= 0
3. Compute the stationary equilibrium 1
o
,
o
, c
o
, a
o
, r
o
, n
o
, 1
o
associ-
ated with t
o
= t. Assume that
1
T
,
T
, c
T
, a
T
, r
T
, n
T
, 1
T
= 1
o
,
o
, c
o
, a
o
, r
o
, n
o
, 1
o
4. Guess a sequence of capital stocks
1
|
Tl
|=l
Note that since the capital
stock at time t = 1 is determined by decisions at time 0,

1
l
= 1
0
. Also
note that 1
|
= 1
0
=

1 is xed. Thus with the guesses on the capital stock
we also obtain r
|
, n
|
Tl
|=l
determined by
n
|
= 1
J
(

1
|
,

1)
r
|
= 1
1
(

1
|
,

1) c
T
|
= t
|
r
|

1
|
.
5. Since we know
T
(a, j) and r
|
, n
|
,

T
|
Tl
|=l
we can solve for
|
, c
|
, a
|l
Tl
|=l
backwards.
6. With the policy functions a
|l
we can dene the transition laws
I
|
Tl
|=l
.
But since we know 1
0
= 1
l
from the initial stationary equilibrium, we can
iterate the distributions forward
1
|l
=

I
|
(
1
|
)
for t = 1, . . . , T 1.
7. With
1
|
T
|=l
we can compute
|
=
_
ad
1
|
for t = 1, . . . , T.
8. Check whether
max
l|<T

1
|
< -
If yes, go to 9. If not, adjust your guesses for
1
|
Tl
|=l
in 4.
9. Check whether

T
1
T
< -. If yes, we are done and should save

|
, a
|l
, c
|
,

1
|
, r
|
, n
|
,

1
|
. If not, go to 1. and increase T.
This procedure determines all the variables we are interested in along the
transition path: aggregate variables such as r
|
, n
|
, 1
|
, 1
|
and individual deci-
sion rules c
:
, a
|l
. It turns out that the value functions
|
enable us to make
statements about the welfare consequences of a tax reform for which we just
have computed the transition path.
5.2.3 Welfare Consequences of the Policy Reform
From our previous algorithm we obtain the sequence of value functions
|
T
|=0
.
Remember the interpretation of the value functions:
0
(a, j) is the expected
lifetime utility of an agent with assets a and productivity shock j at time 0 in
the initial stationary equilibrium, that is, for a person that thinks he will live in
the stationary equilibrium with t = 0 forever. Similarly
l
(a, j) is the expected
lifetime utility of an agent with assets a and productivity j that has just been
informed that there is a permanent tax change. This lifetime utility takes into
account all the transition dynamics through which the agent is going to live.
Finally
T
(a, j) =
o
(a, j) is the lifetime utility of an agent with characteristics
(a, j) born in the new stationary equilibrium (i.e. of an agent that does not live
through the transition).
So in principle we could use
0
,
l
and
T
to determine the welfare conse-
quences from the reform. The problem, of course, is that utility is an ordinal
concept, so that comparing
0
(a, j) with
l
(a, j) we can only determine whether
agent (a, j) gains or loses from the reform, but we cannot meaningfully discuss
how big these gains or losses are.
Now suppose that the period utility function is of C11 form
l(c) =
c
lc
1 o
Consider the optimal consumption allocation in the initial stationary equilib-
rium, in sequential formulation, c
s
o
s=0
, where for simplicity we have suppressed
the explicit dependence of the history of productivity shocks. Then
0
(a, j) = 1
0
o
s=0
c
lc
|
1 o
Now suppose we increase consumption in each date, in each state, in the old
stationary equilibrium, by a fraction q, so that the new allocation is (1
q)c
s
o
s=0
. The lifetime utility from that consumption allocation is
0
(a, j; q) = 1
0
o
s=0
[(1 q)c
|
[
lc
1 o
= (1q)
lc
1
0
o
s=0
c
lc
|
1 o
= (1q)
lc
0
(a, j)
Obviously
0
(a, j; q = 0) =
0
(a, j). If we want to quantify the welfare conse-
quences of the policy reform for an agent of type (a, j) we can ask the following
question: by what percent q do we have to increase consumption in the old
stationary equilibrium, in each date and state, for the agent to be indierent
between living in the old stationary equilibrium and living through the transi-
tion induced by the policy reform.
7
This percent q solves
0
(a, j; q) =
l
(a, j)
or
(1 q)
lc
0
(a, j) =
l
(a, j)
q(a, j) =
_
l
(a, j)
0
(a, j)
_ 1
1o
1
Evidently, this number is bigger than zero if and only if
l
(a, j)
0
(a, j), in
which case the agent benets from the reform, and q(a, j) measures by how
much, in consumption terms. Note that the number q(a, j) depends on an
agents characteristics (one would expect households with a lot of assets to
lose badly, households with little assets to lose not much or to even gain -
remember that taxes are lump-sum redistributed). Also note that we only need
to know
0
(a, j) and
l
(a, j) to compute this number, but our computation of
the transition path gives us the value functions
0
,
l
anyhow.
The computation of consumption equivalent variation, explicitly taking into
account the transition path, is the theoretically correct way to assess the welfare
7
Lucas (1978) proposed this consumption equivalent variation measure in order to assess
the welfare costs of business cycles.
consequences of a reform. Often studies try to assess the steady state welfare
consequences of a policy reform, in particular if these studies do not explicitly
compute transition paths. Even though I think that these welfare numbers are
often not very meaningful, let us quickly describe the procedure. Obviously one
can compute
q
ss
(a, j) =
_
T
(a, j)
0
(a, j)
_ 1
1o
1
with the interpretation that this number is the welfare gain of an agent being
born with characteristics.(a, j) into the new as opposed to the old stationary
equilibrium. If we want to study welfare consequences of a policy reform for
households before their identity is revealed, we may dene as
q
ss
=
__

T
(a, j)d1
T
_

0
(a, j)d1
0
_
1
1o
1
the expected steady state welfare gain. Here
_

T
(a, j)d1
T
is the expected
lifetime utility of an agent in the new stationary equilibrium, before the agent
knows with what characteristics he or she will be born (behind the veil of ig-
norance). The quantity
_

0
(a, j)d1
0
is dened correspondingly. Even though
these measures are not hard to compute (in fact, one does not need to compute
the transition path to compute these numbers), they do ignore the fact that
it takes time to go from the old stationary equilibrium to the new one. For
the policy example at hand, the increase of the capital tax is likely to induce a
decline in the capital stock, thus lower aggregate consumption and thus steady
state welfare losses. What these losses ignore is that along the transition path
part of the capital stock is being eaten, with associated consumption and welfare
derived from it. Therefore welfare measures based on steady state comparisons
may give only fairly limited information about the true welfare consequences of
policy reforms.
5.3 Aggregate Uncertainty and Distributions as
State Variables
In section 5.1 we discussed a model where agents faced signicant amounts
of idiosyncratic uncertainty, which they could, by assumption, only self-insure
against by precautionary saving via risk-less bonds. In the aggregate, the econ-
omy was stationary in that output, investment and aggregate consumption were
constant over time. Thus by construction the model could not speak to how ag-
gregate consumption, saving and investment uctuates over the business cycle.
In this section we introduce aggregate uncertainty, in addition to idiosyn-
cratic uncertainty, into the model. In contrast to idiosyncratic uncertainty,
which is in principle insurable (but insurance against which we ruled out),
there is no mutual insurance against aggregate shock, unless one analyzes a
two-country world where the two countries have aggregate shocks which are
imperfectly correlated.
5.3. AGGREGATE UNCERTAINTYANDDISTRIBUTIONS AS STATE VARIABLES97
5.3.1 The Model
In the spirit of real business cycle theory aggregate shocks take the form of
productivity shocks to the aggregate production function
1
|
= :
|
1(1
|
, 1
|
) (5.75)
where :
|
is a sequence of random variables that follows a nite state Markov
chain with transition matrix . We will follow Krusell and Smith (1998) and
assume that the aggregate productivity shock can take only two values
:
|
:
b
, :
= o (5.76)
with :
b
< :
and denote by (:
t
[:) the conditional probability of the aggregate
state transiting from : today to :
t
tomorrow. Krusell and Smith (1998) interpret
:
b
as an economic recession and :
as an expansion.
The idiosyncratic labor productivity j
|
is assumed to take only two values
j
|
j
u
, j
t
= 1 (5.77)
where j
u
< j
t
stands for the agent being unemployed (having low labor pro-
ductivity) and j
t
stands for the agent being employed. Krusell and Smith
attach the employed-unemployed interpretation to the idiosyncratic labor pro-
ductivity variable, an interpretation which will be important in the calibration
section. Obviously the probability of being unemployed is higher during re-
cessions than during expansions, and the Markov chain governing idiosyncratic
labor productivity should reect this dependence of idiosyncratic uncertainty
on the aggregate state of the economy.
In particular, let denote by
(j
t
, :
t
[j, :) _ 0 (5.78)
the conditional probability of an agent having individual productivity tomorrow
of j
t
and the aggregate state being :
t
tomorrow, conditional on the individual
and aggregate state j and : today. For example (j
u
, :
b
[j
t
, :
) is the probability
of getting laid o tomorrow in a recession if the economy is booming today and
the individual is employed. Consistency with aggregate transition probabilities
requires that
0
Y
(j
t
, :
t
[j, :) = (:
t
[:) for all j 1 and all :, :
t
o (5.79)
i.e. the probability

0
Y
(j
t
, :
t
[j, :) of going from a particular individual
state j and aggregate state : to some individual state j
t
and a particular ag-
gregate state :
t
equals the aggregate transition probability (:
t
[:).
We again assume a law of large numbers, so that idiosyncratic uncertainty
averages out, and only aggregate uncertainty determines the number of agents
in states j 1. Obviously this number will vary with the aggregate state :.
Let by H
s
(j) denote the fraction of the population in idiosyncratic state j if
aggregate state is :; e.g. H
s
|
(j
u
) is the deterministic number of unemployed
people if the economy is in a recession. For consistency, the probabilities have
to satisfy
H
s
0 (j
t
) =

Y
(j
t
, :
t
[j, :)
(:
t
[:)
H
s
(j) for all :, :
t
o (5.80)
Suppose the economy is in a boom today, : = :
. The fraction of people unem-

ployed in a recession tomorrow, H
s
0
=s
|
(j
t
= j
u
) equals the fraction of employed
people today H
s=s
(j = j
t
) who get laid o in the recession, (j
u
[:
b
, j
t
, :
) =
t(r,s
|
]c,s)
t(s
|
]s)
, plus the fraction of unemployed people today H
s=s
(j = j
u
) who
remain unemployed, (j
u
[:
b
, j
u
, :
) =
t(r,s
|
]r,s)
t(s
|
]s)
. The same restriction ap-
plies for all other states.
The exogenous Markov chain driving the economy is therefore described by
the two sets of cardinality 2, o and 1 and the joint 4 4 transition matrix for
idiosyncratic and aggregate productivity
=
_
_
_
_
(j
u
, :
b
[j
u
, :
b
) (j
u
, :
b
[j
t
, :
b
) (j
u
, :
b
[j
u
, :
) (j
u
, :
b
[j
t
, :
)
(j
t
, :
b
[j
u
, :
b
) (j
t
, :
b
[j
t
, :
b
) (j
t
, :
b
[j
u
, :
) (j
t
, :
b
[j
t
, :
)
(j
u
, :
[j
u
, :
b
) (j
u
, :
[j
t
, :
b
) (j
u
, :
[j
u
, :
) (j
u
, :
[j
t
, :
)
(j
t
, :
[j
u
, :
b
) (j
t
, :
[j
t
, :
b
) (j
t
, :
[j
u
, :
) (j
t
, :
[j
t
, :
)
_
_
_
_
(5.81)
In the previous section we dened, proved existence of and computed a station-
ary equilibrium, in which prices and the cross-sectional distribution of assets
(i.e. all macroeconomic variables of interest) were constant over time. Obvi-
ously, with aggregate shocks there is no hope of such an equilibrium to exist.
This makes our extension interesting in the rst place, because it allows business
cycle uctuations, but also will impose crucial complications for the computa-
tion of the economy.
Since the cross-sectional distribution 1 of assets will vary with the aggre-
gate shock, we have to include it as a state variable in our formulation of a
recursive equilibrium. The aggregate state variables also include the aggregate
productivity shock :, since this shock determines aggregate productivity and
hence aggregate wages and interest rates.
For the recursive formulation of the household problem we note that the
individual state variable is composed of individual asset holdings and income
shocks (a, j), whereas the aggregate state variables include the aggregate shock
and the distribution of assets (:, 1). Hence the households problem in recursive
formulation is
(a, j, :, 1) = max
c,o
0
0
_
_
_
l(c) ,

0
Y
s
0
S
(j
t
, :
t
[j, :)(a
t
, j
t
, :
t
, 1
t
)
_
_
_
:.t. (5.82)
c a
t
= n(:, 1)j (1 r(:, 1))a (5.83)
1
t
= H(:, 1, :
t
) (5.84)
Note that :
t
is a determinant of 1
t
since it species how many agents have
idiosyncratic shock j
t
= j
u
and how many agents have j
t
= j
t
. We have the
following denition of a recursive competitive equilibrium
Denition 24 A recursive competitive equilibrium is a value function : 7
o / 1, policy functions for the household a
t
: 7 o / 1 and
c : 7 o / 1, policy functions for the rm 1 : o / 1 and
1 : o /1, pricing functions r : o /1 and n : o /1 and an
aggregate law of motion H : o /o/ such that
1. , a
t
, c are measurable with respect to E(o), satises the households Bell-
man equation and a
t
, c are the associated policy functions, given r() and
n()
2. 1, 1 satisfy, given r() and n()
r(:, 1) = 1
1
(1(:, 1), 1(:, 1)) c (5.85)
n(:, 1) = 1
J
(1(:, 1), 1(:, 1)) (5.86)
3. For all 1 / and all : o
1(H(:, 1)) =
_
a
t
(a, j, :, 1)d1 (5.87)
1(:, 1) =
_
jd1 (5.88)
_
c(a, j, :, 1)d1
_
a
t
(a, j, :, 1)d1 (5.89)
= 1(1(:, 1), 1(:, 1)) (1 c)1(:, 1) (5.90)
4. The aggregate law of motion H is generated by the exogenous Markov
process and the policy function a
t
(as described below)
Again dene the transition function Q
+,s,s
0 : 7 E(7) [0, 1[ by
Q
+,s,s
0 ((a, j), (/, )) =

_
(j
t
, :
t
[j, :) if a
t
(a, j, :, 1) /
0 else
(5.91)
The aggregate law of motion is then given by
1
t
(/, ) = (H(:, 1, :
t
)) (/, ) =
_
Q
+,s,s
0 ((a, j), (/, ))1(da dj) (5.92)
Before describing the computational strategy for the recursive equilibrium some
comments are in order:
1. One should rst dene a sequential markets equilibrium. Such an equilib-
rium will in general exist, but it is impossible to compute such an equilib-
rium directly. No claim of uniqueness of a sequential markets equilibrium
can be made.
2. For the recursive equilibrium dened above Krusell and Smith assert that
the current cross-sectional asset distribution and the current shock are
sucient aggregate state variables. This is not formally proved, and such
a proof would be very hard. What one has to prove is that a recur-
sive equilibrium generates a sequential equilibrium in the following sense:
Cross-sectional distributions are generated, starting from the initial con-
dition (:
0
, 1
0
) as follows
1
l
(:
l
) = H(:
0
, 1
0
, :
l
) (5.93)
1
|l
(:
|l
) = H(:
|
, 1
|
(:
|
), :
|l
) (5.94)
prices are generated by
r
|
(:
|
) = r(:
|
, 1
|
(:
|
)) (5.95)
n
|
(:
|
) = n(:
|
, 1
|
(:
|
)) (5.96)
and, starting from initial conditions (a
0
, j
0
), an individual households
allocation is generated by
c
0
(a
0
, j
0
, :
0
) = c(a
0
, j
0
, :
0
, 1
0
) (5.97)
a
l
(a
0
, j
0
, :
0
) = a
t
(a
0
, j
0
, :
0
, 1
0
) (5.98)
and in general recursively
c
|
(a
0
, j
|
, :
|
) = c(a
|
(a
0
, j
|l
, :
|l
), j
|
, :
|
, 1
|
) (5.99)
a
|l
(a
0
, j
|
, :
|
) = a
t
(a
|
(a
0
, j
|l
, :
|l
), j
|
, :
|
, 1
|
) (5.100)
Similarly optimal choices of the rm can be generated. Thus, a given re-
cursive equilibrium generates sequential allocations and prices; it remains
to be veried that these prices are in fact a sequential equilibrium.
3. Finally, the issue of existence of a recursive equilibrium arises. We know
that, since a sequential equilibrium in general exists, there is a state space
large enough such that a recursive equilibrium (recursive in that state
space) exists. So the issue is whether a recursive equilibrium in which the
aggregate state only contains the current shock and the current wealth dis-
tribution does exist. Although this state space seems natural in some
sense, there is no guarantee of existence of such a recursive equilibrium.
The analysis of this economy is purely computational in spirit as neither
the existence, uniqueness, stability or qualitative features of the equilib-
rium can be theoretically established.
5.3.2 Computation of the Recursive Equilibrium
The key computational problem is the size of the state space. The aggregate
wealth distribution 1 is an innite-dimensional object, the aggregate law of
motion therefore maps an innite-dimensional space into itself. What we look
for is a low-dimensional approximation of the wealth distribution. Why do
agents need to keep track of the aggregate wealth distribution? In order to
gure out todays interest and wage rate the aggregate (average) capital stock
(i.e. the rst moment of the wealth distribution) is sucient. Thus, to forecast
tomorrows factor prices, all the agent needs to forecast is tomorrows aggregate
capital stock. The need to keep track of the current wealth distribution stems
from the fact that the entire wealth distribution 1 today, not only its rst
moment 1, determines tomorrows aggregate capital stock via
1
t
=
_
a
t
(a, j, :, 1)d1 (5.101)
Another way of putting it, the average capital stock tomorrow is not equal to the
saving function of some average, representative consumer, evaluated at todays
average capital stock. If the optimal policy function for tomorrows assets, would
feature a constant propensity to save out of current assets and income (a
t
being
linear in a, with same slope for all j 1 ), then exact aggregation would occur
and in fact the average capital stock today would be a sucient statistic for the
average capital stock tomorrow. This insight is important for the quantitative
results to come.
The computational strategy that Krusell and Smith (and many others since)
follow is to approximate the distribution 1 with a nite set of moments. Re-
member that 1 is the distribution over (a, j). Obviously, since j can only take
two values, the second dimension is not the problem, so in what follows we
focus on the discussion of the distribution over assets a. Let the :-dimensional
vector : denote the rst : moments of the asset distribution (i.e. the marginal
distribution of 1 with respect to its rst argument).
We now posit that the agents use an approximate law of motion
:
t
= H
n
(:, :) (5.102)
mapping the rst : moments of the asset distribution today, :, into the rst :
moments of the asset distribution tomorrow, :
t
. Note that by doing so agents
are boundedly rational in the sense that moments of higher order than : of
the current wealth distribution may help to more accurately forecast the rst :
moments tomorrow. It is the hope that, according to some metric to be discussed
later, agents dont make severe forecasting errors of tomorrows average capital
stock (the only variable they care about for forecasting future prices), and that
the resulting approximate equilibrium is in some sense close to the rational
expectations equilibrium in which agents use the entire wealth distribution to
forecast tomorrows prices.
The next step in the computation of an approximate equilibrium is to choose
the number of moments and the functional form of the function H
n
. Since the
agents only need to know next periods average capital stock, Krusell and Smith
rst pick : = 1 and pose the following log-linear law of motion (remember the
law of motion for the stochastic neoclassical growth model with log-utility and
Cobb-Douglas production)
log(1
t
) = a
s
/
s
log(1) (5.103)
for : :
b
, :
. Here (a
s
, /
s
) are parameters that need to be determined. The
recursive problem of the household then becomes
(a, j, :, 1) = max
c,o
0
0
_
_
_
l(c) ,

0
Y
s
0
S
(j
t
, :
t
[j, :)(a
t
, j
t
, :
t
, 1
t
)
_
_
_
:.t. (5.104)
c a
t
= n(:, 1)j (1 r(:, 1))a (5.105)
log(1
t
) = a
s
/
s
log(1) (5.106)
Note that we have reduced the state space from something that includes the
innite-dimensional space of measures to a four dimensional space (a, j, :, 1)
R 1 o R. The algorithm for computing an approximate recursive equi-
librium is then as follows:
1. Guess (a
s
, /
s
)
2. Solve the households problem to obtain decision rules a
t
(a, j, :, 1)
3. Simulate the economy for a large number of T periods for a large number
of households (in their exercises Krusell and Smith pick = 000 and
T = 11000: start with initial conditions for the economy (:
0
, 1
0
) and for
each household (a
I
0
, j
I
0
). Draw random sequences :
|
T
|=l
and j
I
|
T,
|=l,I=l
and use the decision rule a
t
(a, j, :, 1) and the perceived law of motion
for 1 to generate sequences of a
I
|
T,
|=l,I=l
. Aggregate to nd the implied
sequence of aggregate capital stocks
1
|
=
1
I=l
a
I
|
(5.107)
Discard the rst t periods, because of dependence on initial conditions.
4. With the remaining observations run the regressions
log(1
t
) = c
s
,
s
log(1) (5.108)
to estimate (c
s
, ,
s
) for : o. If the 1
2
for this regression is high and
(c
s
, ,
s
) - (a
s
, /
s
) stop. An approximate equilibrium is found. Otherwise
update guess for (a
s
, /
s
). If guesses for (a
s
, /
s
) converge, but 1
2
remains
low, try to add higher moments to the aggregate law of motion and/or ex-
periment with a dierent functional form for the aggregate law of motion.
5.3.3 Calibration
In this economy no analytical results can be proved and one has to resort to
numerical analysis. Krusell and Smith take the model period to be one quarter
(note that this is a business cycle model). As preferences they assume C11
utility with rather low risk aversion of o = 1 (i.e. log-utility). The time dis-
count factor is chosen to be , = 0.00
d
= 0.06 on an annual basis, which implies
a yearly subjective time discount rate of j = 4.1/ as in Aiyagari. Also similar
to Aiyagari they take the capital share to be c = 0.86. As annual depreciation
rate they choose c = (1 0.02)
d
1 = 0.6/, slightly higher than Aiyagari,
but within the range of values commonly used in real business cycle studies.
The remaining parameters describe the joint aggregate-idiosyncratic labor pro-
ductivity process. Unfortunately the paper itself does not contain a precise
discussion of the parameterization, so that the discussion here relies partly on
Krusell and Smiths information from the paper, partly on Imrohoroglus (1989)
paper to which they refer to and partly on the FORTRAN code posted on Tony
Smiths web site.
The calibration strategy is to rst calibrate the aggregate component of the
productivity process, i.e. the set o and the 2 2 matrix (:
t
[:). Remember
that the aggregate state represents recessions and expansions. With respect to
o they choose
o = 0.00, 1.01 (5.109)
I would think that the standard deviation of the technology shock is a bit on
the small side with 0.01. In fact, using Cooley and Prescotts (1995) continuous
state process and discretizing into a two state chain yields a standard deviation
of the shock of about 0.02. Krusell and Smith claim that with their aggregate
process they are able to generate aggregate uctuations of output similar to
US data, which, given the information in the paper I wasnt able to verify.
Given that Cooley and Prescott need suciently more variance in the technol-
ogy shock to generate business cycles of reasonable size, this must mean that
the economy with heterogenous agents and uninsurable idiosyncratic risk, for
a given aggregate shock variance, is more volatile than its representative agent
counterpart.
As for the transition matrix for the aggregate shock the rst assumption
made is symmetry, so that (:
[:
) = (:
b
[:
b
). Krusell and Smith choose
(:
[:
) such that, conditional on being in the good state today, the expected
time in the good state are 8 quarters, or
8 = [1 (:
[:
)[
_
1 2(:
[:
) 8(:
[:
)
2
. . .
(5.110)
or
8 =
1
1 (:
[:
)
(:
[:
) =
7
8
(5.111)
so that
(:
t
[:) =
_
7
S
l
S
l
S
7
S
_
(5.112)
It follows that, conditional on being in the bad state today, the expected time
of staying there is also 8 quarters.
With respect to idiosyncratic labor productivity, the state space is chosen
as
1 = 0.2, 1 (5.113)
Remember that the idiosyncratic states are meant to represent employment
and unemployment, so that it is assumed that an unemployed person makes
2/ of the labor income of an employed person. Imrohoroglu (1989) argues
that, although the average level of unemployment compensation is higher, a
large fraction of the unemployed do not receive benets (Imrohoroglu quotes a
number of 61/), so that the 2/ gure is argued as reasonable. Krusell and
Smith adopt this value. The level of productivity for an employed person is free
for normalization, and for the sake of the discussion I set it to 1 (doubling both
states in 1 obviously only changes units, but does not aect the results).
With respect to the transition probabilities for the idiosyncratic productivity
we note that
(j
t
[:
t
, j, :) =
(j
t
, :
t
[j, :)
(:
t
[:)
(5.114)
or
(j
t
, :
t
[j, :) = (j
t
[:
t
, j, :) + (:
t
[:) (5.115)
So in order to specify (j
t
, :
t
[j, :), given our previous work we have to specify
the four (for each pair (:
t
, :)) 2 2 matrices (j
t
[:
t
, j, :) indicating, conditional
on an aggregate transition from : to :
t
, what the individuals probabilities of
transition from employment to unemployment are. These should vary with the
aggregate transition (:
t
, :).
The calibration of these matrices are governed by the following observa-
tions. In an expansion the average time of unemployment, conditional on being
unemployed today, is equal to 1. quarters. This implies that
1. = [1 (j
t
= j
u
[:
t
= :
, j = j
u
, : = :
)[
o
I=l
i + (j
t
= j
u
[:
t
= :
, j = j
u
, : = :
)
Il
=
1
1 (j
t
= j
u
[:
t
= :
, j = j
u
, : = :
)
(j
t
= j
u
[:
t
= :
, j = j
u
, : = :
) = 1
1
1.
=
1
8
(5.116)
and hence (j
t
= j
t
[:
t
= :
, j = j
u
, : = :
) =
2
3
. Similarly, to match the
fact that in bad times the average time of unemployment, conditional on un-
employed, equals 2. quarters we nd (j
t
= j
u
[:
t
= :
b
, j = j
u
, : = :
b
) = 0.6
and hence (j
t
= j
t
[:
t
= :
b
, j = j
u
, : = :
b
) = 0.4. The authors then assume
that the probability of remaining unemployed when times switch from expan-
sion to recession is 1.2 times the same probability when the economy was
already in a recession, implying (j
t
= j
u
[:
t
= :
b
, j = j
u
, : = :
) = 0.7
and thus (j
t
= j
t
[:
t
= :
b
, j = j
u
, : = :
) = 0.2. Finally, the proba-

bility of remaining unemployed when times switch from recession to expan-
sion is assumed to be 0.7 times the same probability when times were al-
ready good. This gives (j
t
= j
u
[:
t
= :
, j = j
u
, : = :
b
) = 0.2 and
(j
t
= j
t
[:
t
= :
, j = j
u
, : = :
b
) = 0.7.
The nal transition probabilities are derived from the facts that in recessions
the unemployment rate is H
s
|
(j
u
) = 10/ and in expansions it is H
s
(j
u
) = 4/.
With the probabilities already derived and equation (.80) this uniquely pins
down the remaining probabilities
(j
t
= j
u
[:
t
= :
, j = j
t
, : = :
) = 0.028
(j
t
= j
t
[:
t
= :
, j = j
t
, : = :
) = 0.072
(j
t
= j
u
[:
t
= :
b
, j = j
t
, : = :
b
) = 0.04
(j
t
= j
t
[:
t
= :
b
, j = j
t
, : = :
b
) = 0.06
(j
t
= j
u
[:
t
= :
b
, j = j
t
, : = :
) = 0.070
(j
t
= j
t
[:
t
= :
b
, j = j
t
, : = :
) = 0.021
(j
t
= j
u
[:
t
= :
, j = j
t
, : = :
b
) = 0.02
(j
t
= j
t
[:
t
= :
, j = j
t
, : = :
b
) = 0.08 (5.117)
In short, the best times for nding a job are when the economy moves from a
recession to an expansion, the worst chances are when the economy moves from
a boom into a recession. Combining the aggregate transition probabilities with
the idiosyncratic probabilities, conditional on the aggregate transitions, nally
yields as transition matrix (.81)
=
_
_
_
_
0.2 0.08 0.0087 0.0000
0.8 0.84 0.0812 0.111
0.0812 0.002 0.202 0.024
0.0087 0.122 0.88 0.80
_
_
_
_
(5.118)
With this parameterization we are ready to report results on computed approx-
imate equilibria.
5.3.4 Numerical Results
The model generates three basic entities of interest: an aggregate law of motion
:
t
= H
n
(:, :), (5.119)
individual decision rules a
t
(a, j, :, :) and time-varying cross-sectional wealth
distributions 1(a, j). First lets focus on the aggregate law of motion. Remember
that we allowed agents only to use the rst : moments of the current wealth
distribution to forecast the rst : moments of tomorrows wealth distribution,
i.e. agents are boundedly rational. In particular, the aggregate law of motion
perceived by agents may not coincide with the actual law of motion, which is
just another way of saying that we are NOT computing a rational expectations
equilibrium.
Since agents only have to forecast tomorrows aggregate (average) capital
stock (i.e. the rst moment of 1
t
), Krusell and Smith start out with : = 1.
After repeated updating of the coecients in (.108) the coecients converge
and one obtains as perceived law of motion of agents
log(1
t
) = 0.00 0.062 log(1) for : = :
(5.120)
log(1
t
) = 0.08 0.06 log(1) for : = :
b
(5.121)
How irrational are agents? We would have computed a perfectly legitimate
rational expectations equilibrium if the actual law of motion for the capital
stock follows (.120) to (.121) exactly. Remember that at each step, to update
the coecients in (.120) to (.121), a simulated time series for the aggregate
capital stock in conjunction with a sequence of aggregate shocks (:
|
, 1
|
T
|=0
is
generated. By dividing the sample into periods with :
|
= :
b
and :
|
= :
one
then can run the regressions
log(1
|l
) = c
log(1
|
) -
|l
for periods with :
|
= :
(5.122)
log(1
|l
) = c
b
,
b
log(1
|
) -
b
|l
for periods with :
|
= :
b
(5.123)
In fact, given the computational algorithm, for the last step of the iteration
on the perceived law of motion they did obtain c
= 0.00, c
b
= 0.08 and
= 0.062,

,
b
= 0.06. From econometrics we remember that with regression
errors
-
|l
= log(1
|l
) c
log(1
|
) for , = q, / (5.124)
we dene the standard deviation of the regression error as
o
=
_
1
T
|r
_
-
|
_
2
(5.125)
where t
is the set of time indices for which :

|
= :
and T
is the cardinality of
that set. Note that, since the regression is on log-income, o
can be interpreted
as the average percentage error for the prediction of the capital stock made by
the regression. The 1
2
of the regression, as alternative measure of t, is dened
as that fraction of the variation in tomorrows log-capital stock that is explained
by the variation of todays log-capital stock, or
1
2
= 1
|r
_
-
|
_
2
|r
_
log 1
|l
log 1
_
2
(5.126)
If o
= 0 for , = q, / or equivalently, if 1
2
= 1 for , = q, / then households

perceived aggregate law of motion is exactly correct for all periods, i.e. agents
do not make forecasting errors and are perfectly rational. Low o
and high 1
2
are taken as evidence that the actual law of motion generated by individual
behaviors and aggregation does not depart much from the perceived law of
motion, i.e. that agents make only small forecasting errors.
Krusell and Smith obtain 1
2
= 0.000008 for , = /, q and o
= 0.0028,
o
b
= 0.0086, i.e. extremely low average forecasting errors made by agents.
They calculate maximal forecasting errors for interest rates 2 years into the
future of 0.1/ for their simulations. Unfortunately they do not report magni-
tudes of corresponding utility losses from forecasting errors (say, in consumption
equivalent variation), but assert that given the small magnitude of forecasting
errors even for the far future these utility losses from bounded rationality are
negligible.
Before giving intuition for the results a word of caution is in order. In all
of computational economics, since computer precision is limited, the best one
can achieve is to compute an approximate equilibrium, i.e. allocations and
prices in which markets almost clear (excess demand is - away from zero). This
approximate equilibrium may be arbitrarily far away from a true equilibrium
(i.e. prices and allocations for which markets clear exactly). The study by
Krusell and Smith is not unusual in this respect. But even if this general problem
were absent, Krusell and Smiths equilibrium is at best an approximation to a
rational expectations equilibrium and the true rational expectations equilibrium
may be arbitrarily far away from the computed equilibrium. Krusell and Smiths
agents make small forecasting errors (according to their o
, 1
2
metric), but the

aggregate law of motion at which agents make no forecasting errors (i.e. a
rational expectations equilibrium) will in general involve higher moments and
may look very dierent from the one they found.
5.3.5 Why Quasi-Aggregation?
Suppose all agents, for all interest rates r(1) and wages n(1) have linear
savings functions with the same marginal propensity to save, so that
a
t
(a, j, :, 1) = a
s
/
s
a c
s
j (5.127)
Then
1
t
=
_
a
t
(a, j, :, 1)d1 = a
s
/
s
_
ad1 c
s
1
= a
s
/
s
1 (5.128)
since 1 =
_
ad1. Therefore exact aggregation obtains and the rst moment of
the current wealth distribution (which equals the current capital stock) is in fact
a sucient statistic for 1 for forecasting the aggregate capital stock tomorrow.
In their Figure 2, Krusell and Smith plot savings functions (note: for a par-
ticular current aggregate stock and a particular shock :). We see that they are
almost linear with same slope for j = j
u
and j = j
t
. The only exceptions are
unlucky agents (j = j
u
) with little assets which are liquidity constrained and
hence have a low (zero) marginal propensity to save. However, since these (few)
agents hold a negligible fraction of aggregate wealth, they dont matter for the
aggregate capital dynamics. All other agents have almost identical propensi-
ties to save, thus individual saving decisions almost exactly aggregate, and the
current aggregate capital stock is almost a sucient statistic when forecasting
tomorrows capital stock: quasi-aggregation obtains.
The key question is why individual savings functions a
t
are almost linear in
current assets a at just about all current asset levels. From Figure 2 of Krusell
and Smith we see that the slope of a
t
when plotted against a is roughly equal
to 1 for all but very low asset levels. Remember from the PILCH model with
certainty equivalence that optimal consumption was given by
c
|
=
r
1 r
_
1
|
T|
s=0
j
|s
(1 r)
s
a
|
_
(5.129)
and hence agents consume a fraction
:
l:
of current assets. That is, they save
out of current assets for tomorrow
a
|l
1 r
=
_
1
r
1 r
_
a
|
G(j) (5.130)
where G(j) is a function of the stochastic income process. Thus under certainty
equivalence
a
|l
= a
|
H(j) (5.131)
and thus the saving function a
t
has slope 1 under certainty equivalence (and
j = r). In Krusell and Smiths economy agents are prudent and face liquidity
constraints, but almost act as if they are certainty equivalence consumers. Why?
A few reasons come to my mind:
1. Agents are prudent, but not all that much. A o = 1 is at the lower end
of the empirical estimates for risk aversion. As we saw from Aiyagaris
(1994) paper, the amount of precautionary saving increases signicantly
with increases in o, and so should the deviation of agents decision rules
from certainty equivalence.
2. The unconditional standard deviation of individual income is roughly 0.2,
at the lower end of the estimates used by Aiyagari. As we saw, higher
variability, induces more quantitatively important precautionary saving.
3. Probably most important, negative income shocks (unemployment) are in-
frequent and not very persistent, so that they dont force a large departure
in behavior from certainty equivalence. Krusell and Smiths agents never
become permanently disabled and dont face permanent wage declines af-
ter being laid o (once they nd a new job, their relative wage is as high
as before they were laid o)
Krusell and Smith consider various sensitivity tests with respect to these and
other dimensions (change in time discount rates, endogenous labor-leisure choice
etc.), and claim that the result of quasi-aggregation (only the mean capital stock
matters for forecasting tomorrows mean capital stock) is robust to changes in
the model parameterization.
5.3.6 Rich People are Not Rich Enough
The model generates an endogenous consumption and wealth distribution, some-
thing that all of representative agents macroeconomics is silent about. Note
that the income distribution is, by specifying the income process that house-
holds face, an input into the model. To the extent that the income process
and hence the cross-sectional income distribution is realistic, one would hope
that the resulting wealth distributions (remember that this distribution changes
over time) is on average consistent with the cross-sectional wealth distribution
in the data. Unfortunately, the model does a fairly bad job reproducing the
US wealth distribution, in particular it fails to generate the high concentration
of wealth at the upper end of the distribution. In the data, the richest 1/ of
the US population holds 80/ of all household wealth, the top / hold 1/ of
all wealth. For the model described above the corresponding numbers are 8/
and 11/, correspondingly. In the model people save to buer their consump-
tion against unemployment shocks, but since these shocks are infrequent and of
short duration, they dont save all that much.
There are several ways of making a small fraction of the population save
a lot, and hence making them accumulating a large fraction of overall wealth.
The repair job that Krusell and Smith propose is to make some agents (sto-
chastically) more patient than others. In particular, they assume that the time
discount factor of agents , is stochastic and follows a three state Markov chain
with , 1 and transition probabilities (,
t
[,). The dynamic programming
problem of agents then becomes
(a, j, ,, :, 1) (5.132)
= max
c,o
0
0
_
_
_
l(c) ,

0
Y
s
0
S
o
0
1
(j
t
, :
t
[j, :)(,
t
[,)(a
t
, j
t
, ,
t
, :
t
, 1
t
)
_
_
_
:.t.
c a
t
= n(:, 1)j (1 r(:, 1))a (5.133)
log(1
t
) = a
s
/
s
log(1) (5.134)
They pick 1 = 0.088, 0.0804, 0.008. Hence annual discount rates dier
between 2.8/ for the most patient agents and .6/ for the most impatient
agents. As transition matrix Krusell and Smith propose (remember that the
model period is one quarter)
=
_
_
0.00 0.00 0
0.00062 0.0087 0.00062
0 0.00 0.00
_
_
(5.135)
which implies that 80/ of the population has discount factor in the middle
and 10/ are on either of the two extremes (in the stationary distribution). It
also implies that patient and impatient agents remain so for an expected period
of 50 years. De facto this parameterization creates three deterministic types
of agents. The patient agents are the ones that will accumulate most of the
wealth in this economy. With this trick of stochastic discount factors Krusell
and Smith are able to approximate the US wealth distribution to a reasonably
accurate degree (see their Figure 3): in the modied economy the richest 1/
of the population holds 24/ of all wealth, the richest / hold / percent of
all wealth. Also note that the quasi-aggregation result extends to the economy
with stochastic discount factors.
Chapter 6
Complete Market Models
with Frictions
In this chapter we will consider models that, in spirit, build on the complete
markets model considered in Chapter 3, in the same fashion the models in
Chapter 4 and 5 built on the simple Pilch model in Chapter 3. Most empirical
studies reject the complete insurance hypothesis and thus cast doubt upon the
complete markets model as a reasonable description of reality.
Therefore models are desired that predict some, but not perfect risk shar-
ing. The Pilch model (in its general equilibrium form) generates some risk
sharing via self-insurance: agents smooth part of their income uctuation by
asset accumulation and decumulation, with the part being determined by pref-
erences and the nature of the income shocks. But remember that there are no
formal risk-sharing arrangements in the Pilch model, as explicit contingent in-
surance contracts, which agents in the model would have an incentive to trade
and nancial intermediaries would have an incentive to oer, are ruled out by
assumption, without any good reason from within the model.
1
The models we study in this chapter will allow a full set of contingent con-
sumption claims being traded (in the decentralized version) or being allocated
by the social planner (in the centralized version of the model). That full insur-
ance does not arise as optimal and equilibrium allocation is due to informational
and/or enforcement frictions, which are explicitly modeled. These models ad-
here to the principle, most forcefully articulated by Townsend that the world
is constrained ecient, and that is up to the researcher/modeler to nd the
right set of constraints that give rise to model allocations which are in line with
empirical consumption allocations.
We will study two such sets of constraints: the rst stemming from the fact
1
Townsend (1985) and Cole and Kocherlakota (1997) study environments with private
information and show that the optimal contract between agents and nancial intermediaries
is a simple uncontingent debt contract, closely resembling the one-period uncontingent bonds
that we let agents trade in the Pilch model.
111
112 CHAPTER 6. COMPLETE MARKET MODELS WITH FRICTIONS
that in actual economies nancial contracts are only imperfectly enforceable
(because there exist explicit bankruptcy provisions in the legal code or it is too
costly to always enforce repayment), the second deriving from the fact that in-
dividual incomes and/or actions are imperfectly observable, so that contingent
claims payments cannot be directly conditioned on these. Both types of models
will deliver allocations characterized by some, but (depending on parameteriza-
tions) imperfect insurance, which is due explicitly to these frictions.
6.1 Limited Enforceability of Contracts
We start with models in which perfect insurance is prevented because at each
point of time every agent can default on her nancial obligations, with the en-
suing punishment being modeled as exclusion from future credit market partic-
ipation. The seminal contributions to this literature include Kehoe and Levine
(1993), Kocherlakota (1996) and Alvarez and Jermann (2000). Kehoe and
Levine (1993), building on earlier work of Eaton and Gersowitz (1981), develop
a competitive equilibrium model with a small number of agents incorporating
imperfect enforceability of contracts (in the standard Arrow Debreu model one
of the crucial implicit assumptions is perfect enforcement of contracts); their
2001 Econometrica paper compares this model to a standard Pilch model in
a fairly accessible way. Kocherlakota studies the same environment, but lets
agents interact strategically; the no-default constraints become restrictions de-
rived from the requirement of subgame perfection and constrained eciency.
Finally, Alvarez and Jermann (2000) show how to, under weak assumptions,
decentralize constrained-ecient allocations arising in Kehoe and Levine and
Kocherlakota as sequential markets equilibrium with a full set of Arrow securi-
ties and state-contingent borrowing constraints that are not too tight, in the
sense that agents can borrow up to the amount in which they are indierent
between repaying and defaulting (and being excluded from future intertemporal
trade).
Lately some authors have extended these models to settings with a contin-
uum of agents (the counterpart of Aiyagaris model), see Krueger (1999) and
used them to address empirical questions (Krueger and Perri, 2005, 2006a,b).
Lustig (2001) develops a method to handle aggregate uncertainty in these mod-
els (as in Krusell and Smith), and uses it to study asset prices.
2
In this chapter we will rst study how to characterize (and compute) con-
strained ecient allocations for a model with 2 agents, how to decentralize it
as a competitive Arrow-Debreu, a sequential markets and a subgame perfect
equilibrium. We then use it to study the relationship between the variability of
income and the resulting consumption distribution. We then will discuss how
to deal with a continuum of agents, both in terms of theory and in terms of
2
An OLG-version of the Kehoe and Levine model is studied by Azariadis and Lambertini
(2001) and applied to the study of social security by Andolfatto and Gervais (2000). Devel-
opment economists applied these class of models to study formal and informal risk-sharing
arrangements in rural Africa (see Udry 1995) and India (see Ligon, Thomas and Worrall 2000).
6.1. LIMITED ENFORCEABILITY OF CONTRACTS 113
computation, and discuss other applications.
6.1.1 The Model
The model is populated by 2 (types of) innitely lived households, i = 1, 2,
that, in each period, consume a single perishable consumption good. Let by
c
I
|
= c
I
|
(:
|
) 0 denote the stochastic symmetric endowments governed by the
random variable :
|
o, where the set o is nite. As before let denote by
:
|
= (:
|
, . . . :
l
) o
|
denote a history of shocks. The stochastic process is
assumed to be Markov with transition probabilities (:
|l
[:
|
) and invariant
distribution H. For given initial shock :
0
, whose distribution is given by H, the
probabilities of endowment shock histories are given by
(:
|
) = (:
|
[:
|l
) + . . . (:
l
[:
0
) (6.1)
Assume that the endowment processes are symmetric in that if c
l
|
(:
|
) = c
l
|
(:
|
) =
c, then there exists a :
|
with (:
|
) = ( :
|
) and c
2
|
(:
|
) = c
2
|
(:
|
) = c. That is,
both agents face identical stochastic endowment processes.
A consumption allocation is denoted by (c
l
, c
2
) = c
l
|
(:
|
), c
2
|
(:
|
)
o
|=l,s
t
S
t
and agents are assumed to have preferences over consumption streams given by
l(c) = (1 ,)
o
|=0
s
t
S
t
,
|
(:
|
)n(c
I
|
(:
|
)) (6.2)
where , (0, 1) and n is strictly increasing, strictly concave and C
2
and satises
INADA conditions.
The denition of a Pareto ecient allocation is standard, but we repeat it
here for the sake of refreshing memory.
Denition 25 An allocation (c
l
, c
2
) is Pareto ecient if it is resource feasible
c
l
c
2
= c
l
c
2
(6.3)
for all t, all :
|
, and there is no other feasible allocation ( c
l
, c
2
) such that l( c
I
) _
l(c
I
), with at least one inequality strict.
Note from Chapter 2 that an allocation is Pareto ecient if and only if it is
resource feasible and satises
n
t
(c
l
|
(:
|
))
n
t
(c
2
|
(:
|
))
= for all t, all :
|
, (6.4)
for some 0.
In this economy there are strong incentives to share the endowment risk both
agents face. The rst question we want to pose and answer is what is the extent
of risk sharing possible if agents cannot commit to long term contracts that are
not individually rational? In order to do so we will compute constrained ecient
allocations using recursive contract techniques, in which promised utility acts
as a state variable. Then we decentralize these allocation within a
1. Dynamic transfer game: individual rationality translates into subgame
perfection (Kocherlakota 1996)
2. Arrow Debreu competitive market. Individual rationality constraints enter
individual consumption sets (Kehoe and Levine 1993, 2001)
3. Sequential market where agents enter (long-term) relationship with nan-
cial intermediaries that set appropriate borrowing constraints for Arrow
securities (Alvarez and Jermann, 2000).
Now let us formalize the individual rationality constraints that agents face
in the light of their inability to commit to repaying their debt. First dene the
continuation lifetime expected utility from allocation (c
l
, c
2
) for agent i in node
:
|
of the event tree as
l(c
I
, :
|
) = (1 ,)n(c
I
|
(:
|
))
(1 ,)
o
r=|l
s
:
]s
t
,
r|
(:
r
[:
|
)n(c
I
r
(:
r
)) (6.5)
We now impose the following individual rationality constraints on allocations
l(c
I
, :
|
) _ l(c
I
, :
|
) = l
I,.u|
(:
|
) (6.6)
These constraints on allocations say that, at no point of time, no contingency
any agent would prefer to walk away from the allocation (c
l
, c
2
), with the con-
sequence of living in nancial autarky and consuming her own endowment from
that node onward.
6.1.2 Constrained Ecient Allocations
Let I denote the set of all allocations that satisfy the resource constraints (6.8)
and the individual rationality constraints (6.6). Also dene

l(:
0
) = max
(c
1
,c
2
)I
l(c
l
, :
0
)
and l(:
0
) = l
l,.u|
(:
0
) as the upper and lower bounds of lifetime utility agent
1 can obtain from any allocation that is feasible and satises the individual ra-
tionality constraints (remember that endowments are symmetric). We have the
following denition.
Denition 26 An allocation (c
l
, c
2
) I is constrained ecient if there is no
other feasible allocation ( c
l
, c
2
) I such that l( c
I
) _ l(c
I
), with at least one
inequality strict.
For given initial :
0
, the constrained Pareto Frontier \
s0
: [l(:
0
),

l(:
0
)[
[l(:
0
),

l(:
0
)[ is dened as
\
s0
(l) = max
(c
1
,c
2
)I
l(c
2
, :
0
)
:.t. l(c
l
, :
0
) _ l (6.7)
The number \
s0
(l) is interpreted as the maximal lifetime utility (conditional
on :
0
having been realized) agent 2 can obtain from any constrained-feasible
allocation if agent 1 is guaranteed at least lifetime utility l [l(:
0
),

l(:
0
)[.
We immediately have the following
Proposition 27 An allocation (c
l
, c
2
) is constrained ecient if and only if it
solves the above maximization problem, for some l [l(:
0
),

l(:
0
)[
6.1.3 Recursive Formulation of the Problem
We now want to construct the constrained ecient consumption allocation
(sometimes called a recursive contract, because it will, as we will see, have
the nature of a long-term risk sharing contract. The basic idea for doing so
goes back to Spear and Srivastava (1987) and Abreu (1988): since the individ-
ual rationality constraints constrain continuation utilities of allocations, make
continuation utility a state variable in the recursive problem.
The constrained Pareto frontier was dened as maximizing the lifetime utility
of the second agent, subject to guaranteeing the rst agent at least a certain
lifetime utility l. For the recursive problem this lifetime utility becomes a state
variable (together with the current exogenous state of the world :). In order to
notationally distinguish between the recursive and the sequential problem let
the lifetime utility of agent 1 in the recursive formulation be denoted by n
(obviously we can switch the roles of agent 1 and 2).
Thus state variables for the recursive problem are (n, :) and the Bellman
equation reads as
3
\ (n, :) = max
(c1,c2,]u
0
(s
0
)]
s
0
2:
)
_
(1 ,)n(c
2
) ,

s
0
S
(:
t
[:)\ (n
t
(:
t
), :
t
)
_
s.t. (6.8)
c
l
c
2
= c
l
(:
|
) c
2
(:
|
) (6.9)
\ (n
t
(:
t
), :
t
) _ l
2,.u|
(:
t
) (6.10)
n
t
(:
t
) _ l
l,.u|
(:
t
) (6.11)
n
t
(:
t
) _

l(:
t
) (6.12)
n = (1 ,)n(c
l
) ,

s
0
S
(:
t
[:)n
t
(:
t
) (6.13)
This is a standard dynamic programming problem that can be solved with
standard techniques. Note that rst the values of autarky have to be determined,
which is done by solving the equations
l
I,.u|
(:) = (1 ,)n(c
I
(:)) ,
s
0
(:
t
[:)l
I,.u|
(:
t
) (6.14)
3
I deviate from the timing convention in Kocherlakota, who lets the economy start in
period 1, after an initial c
0
has been drawn, and analyzes allocations from period 1 onward.
I formulate the problem so that consumption also takes place in period 0, after the period 0
shock has been realized. This is more consistent with my discussion in earlier chapters and
will help to simplify the discussion of the continuum economy.
These are =card(o) equations in unknowns l
I,.u|
(:), : o. Once we
have the values of autarky, we can solve the dynamic programming problem
4
for policy functions c
I
(n, :) and n
t
(n, :; :
t
).
Sequential consumption allocations are derived from the recursive policy
functions c
I
(n, :) and n
t
(n, :; :
t
) as usual. The initial conditions for the se-
quential problem are (l, :
0
). Then we derive the consumption allocation as
follows:
c
I
0
(:
0
) = c
I
(l, :
0
)
n
l
l
(:
l
) = n
t
(l, :
0
; :
l
)
n
2
l
(:
l
) = \ (n
l
l
(:
l
), :
l
) (6.15)
and recursively
n
l
|
(:
|
) = n
t
(n
l
|l
(:
|l
), :
|l
; :
|
)
n
2
|
(:
|
) = \ (n
l
|
(:
|
), :
|
)
c
I
|
(:
|
) = c
I
(n
I
|
(:
|
), :
|
) (6.16)
The remaining question is whether these allocations in fact solve the sequen-
tial planning problem and whether the function \
s0
satises \
s0
(l) = \ (l, :
0
)
for all :
0
o. But this is nothing else but proving Bellmans principle of optimal-
ity, which in fact has been done by Thomas and Worrall (1988). Kocherlakota
provides a partial characterization of constrained ecient allocations; I will de-
fer the discussion of this to the paper presentation and will discuss a simple
example below.
6.1.4 Decentralization
Since both the papers by Kocherlakota and by Kehoe and Levine will be pre-
sented in class, I will skip the discussion of how constrained ecient allocations
can be decentralized within a subgame perfect equilibrium of a transfer game or
a competitive equilibrium with enforcement constraints. I will instead present a
decentralization of constrained-ecient allocations as a sequential markets equi-
librium with state-contingent borrowing constraints that is due to Alvarez and
Jermann (2000).
This sequential markets equilibrium is almost identical to the one discussed
in Chapter 3; there, however, we picked borrowing constraints that were very
loose in that they just prevented Ponzi schemes, but did allow rst-best con-
sumption smoothing. Now we are looking for borrowing constraints that have
more bite, in that they resemble exactly the incentive constraints from the social
planners problem.
4
Strictly speaking one also needs the upper bound on utilities

l(c). One guesses these
bounds and then iterates between guesses and solutions of the Bellman equation, until one
attains \ (
l(c), c) = l
2;Aut
(c) for all c.
Let by
|
(:
|
, :
|l
) denote the price of the Arrow security at node :
|
that
pays out one unit of the consumption good if tomorrows shock is :
|l
. Also
denote by a
I
0
the initial conditions of asset holdings for agent i, where of course
I
a
I
0
= 0. Note that there exists a one-to-one mapping between the (a
l
0
, a
2
0
) and
the position on the Pareto frontier l. By

\ (a
I
|
(:
|
), :
|
) let denote the continuation
utility of agent i when entering node :
|
with assets a
I
|
(:
|
). It satises
\ (a
I
|
(:
|
), :
|
) = max
c
1
,o
1
l(c
I
, :
|
) (6.17)
s.t.
c
I
|
(:
|
)
st+1
|
(:
|
, :
|l
)a
I
|l
(:
|
, :
|l
) _ c
I
|
(:
|
) a
I
|
(:
|
) (6.18)
a
I
|l
(:
|l
) _

I
(:
|l
) (6.19)
Evidently

\ (a
I
0
, :
0
) is the lifetime utility of agent i with initial conditions (a
I
0
, :
0
)
which faces the borrowing constraints

I
(:
|l
). Alvarez and Jermann dene
borrowing constraints that are not too tight as satisfying
\ (

I
|
(:
|
), :
|
) = l
I,.u|
(:
|
) (6.20)
That is, the borrowing limits are such that agents that have borrowed up
the maximum are exactly indierent between repaying their debt and default-
ing, with the punishment of default being specied as nancial autarky. The
denition of a sequential market equilibrium with borrowing constraints that
are not too tight is standard and hence omitted.
Dene Arrow securities prices as
|
(:
|
, :
|l
) = max
I=l,2
_
,(.
t
[.)
n
t
(c
I
|l
(:
|
, :
|l
))
n
t
(c
I
|
(:
|
))
_
(6.21)
and implied Arrow-Debreu prices as
Q(:
|
[:
0
) =
0
(:
0
, :
l
) +
l
(:
l
, :
2
) + . . . +
|l
(:
|l
, :
|
) (6.22)
For any allocation (c
l
, c
2
), the implied interest rate are said to be high if
|0
s
t
Q(:
|
[:
0
) +
_
c
l
|
(:
|
) c
2
|
(:
|
)
_
< (6.23)
Alvarez and Jermann then prove the following results:
Proposition 28 Suppose a constrained ecient consumption allocation (c
l
, c
2
)
has high implied interest rates. Then it can be decentralized as sequential markets
equilibrium with some initial conditions and borrowing constraints that are not
too tight.
Proposition 29 Suppose that a constrained ecient allocation (c
l
, c
2
) has some
risk sharing, that is, for all t and all :
|
there exists an i such that
l(c
I
, :
|
) l(c
I
, :
|
) (6.24)
Then the implied interest rates are high.
Combining these two results we have the following
Corollary 30 Any nonautarkic constrained ecient consumption allocation can
be decentralized as a sequential markets equilibrium with borrowing constraints
that are not too tight.
The main implication of this result is that one can solve for the entire set
of potential equilibrium allocations by solving the recursive planning problem
and then nd Arrow securities prices and borrowing constraints from (6.20) and
(6.21) that, together with the consumption allocation, form a sequential markets
equilibrium.
6.1.5 A Simple Application: Income and Consumption In-
equality
In this section we will consider a simple example of the model below and use
it to study how, in the model, an increase in income inequality aects the
consumption distribution. The aggregate state of the world can take two values
:
|
o = 1, 2. Let individual endowments be given by the functions
c
l
(:
|
) =
_
1 -
1 -
if :
|
= 1
if :
|
= 2
(6.25)
c
2
(:
|
) =
_
1 -
1 -
if :
|
= 1
if :
|
= 2
(6.26)
That is, if :
|
= 1 then agent 1 has currently high endowment 1 -, if :
|
= 2
then agent 2 has currently high endowment. Here - measures the variability
of individual endowment and also turns out be the (unconditional) standard
deviation of the cross-sectional income distribution.
Which agent is rich follows a stochastic process that is assumed to be Markov
with transition probabilities
=
_
c 1 c
1 c c
_
(6.27)
where c (0, 1) denotes the conditional probability of agent 1 being rich tomor-
row, conditional on being rich today (and symmetrically for agent 2). Thus c is
a measure of persistence of the income process, with c =
l
2
denoting the iid case
and c = 1 reecting a deterministic income process (that is, permanent shocks).
The stationary distribution associated with , for any given c (0, 1) is given
by H(:) =
l
2
for all : o, and we assume that H(:
0
) =
l
2
for all :
0
o. This
assumption makes agents ex ante identical and allows us to restrict attention to
symmetric equilibrium allocations.
The good thing about this model is that we can solve for the continuation
expected discounted utility from autarky analytically. Since there are only two
aggregate states at each point of time, the continuation value from the autarkic
allocation can take two values, one for the currently rich agent, denoted by
l(1 -), and one for the currently poor agent, denoted by l(1 -). These two
values solve the following recursion
l(1 -) = (1 ,)n(1 -) cl(1 -) (1 c)l(1 -)
l(1 -) = (1 ,)n(1 -) cl(1 -) (1 c)l(1 -) (6.28)
Solving these two equations in two unknowns yields
l(1 -) =
1
1
(1 ,) n(1 -) ,(1 c) [n(1 -) n(1 -)[
l(1 -) =
1
1
(1 ,) n(1 -) ,(1 c) [n(1 -) n(1 -)[(6.29)
where
1 =
(1 ,c)
2
(, ,c)
2
1 ,
0 (6.30)
The utility from autarky is the weighted sum of the utility from consuming the
endowment today, n(1 -) (or n(1 -) for the currently poor agent) and the
expected future utility, which is proportional to n(1 -) n(1 -). A change
in - thus changes current consumption and the risk of future consumption.
We are interested in how the variability of the income process - aects the
consumption distribution. Since the only reason perfect insurance is not pos-
sible is the presence of the individual rationality constraints, we rst want to
characterize the right hand side of these constraints l(1 -) and l(1 -), as
functions of -.
Let by l
J1
= n(1) denote the lifetime utility from the rst best, perfect
risk sharing allocation. This lifetime utility would be obtained by both agents
if the individual rationality constraints were absent, i.e. with complete markets
and no commitment problem.
By simple inspection of (6.20) we can prove the following
Lemma 31 The continuation utilities from the autarkic allocation satisfy the
following properties
1. l(1 -)[
:=0
= l(1 -)[
:=0
= n(1) = l
J1
2. l(1-) and l(1-) are strictly concave and dierentiable for all - [0, 1)
3.
JI(l:)
J:
< 0 for all - [0, 1)
4.
JI(l:)
J:
:=0
0 and lim
:l
JI(l:)
J:
< 0
5. There exists a unique -
l
= aig max
:|0,l)
l(1 -). For all - [0, -
l
) we
have
JI(l:)
J:
0 and for all - (-
l
, 1) we have
JI(l:)
J:
< 0
6. There exists at most one -
2
(0, 1) such that l(1 -
2
) = n(1). If -
2
exists, it satises -
2
-
l
.
The proof of this lemma is straightforward and hence omitted. The interpre-
tation is also fairly simple: without income uctuations the autarkic allocation
is rst-best; since an increase in income variability reduces current consumption
and increases future consumption risk, it reduces the continuation utility for the
currently poor agent; for the rich agent initially the direct eect of currently
higher consumption dominates the risk eect, but as - becomes large the risk
eect becomes dominant; the last two properties follow from strict concavity of
l(1 -) and the signs of the derivatives at - = 0 and - 1. The rst gure at
the end of this chapter graphically summarizes the lemma.
We now want to characterize constrained-ecient consumption allocations in
this model and analyze how they change with changes in -. For this we note that
in any insurance mechanism as the one described in this model it is ecient to
transfer resources from the currently rich to the currently poor agent. Therefore
the constrained-ecient consumption distribution features maximal insurance,
subject to delivering at least the continuation utility of autarky to the currently
rich agent. This argument motivates the following proposition, which is due to
Kehoe and Levine (2001):
5
Proposition 32 The constrained-ecient consumption distribution is completely
characterized by a number -
c
(-) _ 0. The agent with income 1 - consumes
1-
c
(-) and the agent with income 1- consumes 1-
c
(-) regardless of her past
history. The number -
c
(-) is the smallest non-negative solution of the following
equation
max(l
J1
, l(1 -)) = l(1 -
c
(-)) (6.31)
The interpretation of this result goes as follows: if max(l
J1
, l(1 -)) =
l
J1
, then -
c
(-) = 0 and the resulting allocation is the rst-best, complete risk
sharing allocation. From the previous lemma we know that this case applies to
all - _ -
2
, if such -
2
exists. Now suppose that max(l
J1
, l(1 -)) = l(1 -).
Obviously one solution for -
c
(-) = -, i.e. the autarkic consumption allocation.
We note from the previous lemma that if - _ -
l
, then this is in fact the resulting
consumption allocation. In this case, for small -, there is no risk sharing possible
whatsoever, since at any level of risk sharing the rich agent would have an
incentive to default. However, if - (-
l
, -
2
) then there exists an -
c
(-) < -
l
such
that l(1 -) = l(1 -
c
(-)). In this case the resulting consumption allocation
features some risk sharing, but not complete risk sharing. The second gure
at the end of this chapter shows a sample path for the income process and the
resulting constrained-ecient sample path for the consumption allocation.
How does the dispersion of the consumption distribution -
c
(-) vary with an
increase in income dispersion. From the previous discussion we immediately
obtain the following proposition (see Krueger and Perri, 2006a):
5
This proposition is less general than Kehoe and Levines original proposition, because their
environment also contains a long-lived asset. It is more general since, making agents ex ante
identical enables us to establish the characterization of the constrained-ecient consumption
distribution from the initial model period onward, and not just after a transition period after
which initial conditions cease to matter, as in Kehoe and Levine.
Proposition 33 For given c, starting from a given income dispersion - a mar-
ginal increase in - leads to a decrease in consumption inequality if and only if
-
c
(-) < - (in the initial equilibrium there is positive risk sharing). The decrease
is strict if and only if 0 < -
c
(-) < -
0
(in the initial equilibrium there is positive,
but not complete risk sharing).
This discussion is summarized in the third gure at the end of this chap-
ter. Similarly one can establish how the consumption distribution varies with
changes in the persistence of the income process. See Kehoe and Levine (2001)
or Krueger and Perri (2006a) for a proof.
Proposition 34 For a given income dispersion - a marginal increase in per-
sistence c leads to an increase in consumption inequality. The increase is strict
if and only if 0 < -
c
(-, c) < - (in the initial equilibrium there is positive, but
not complete risk sharing).
So for this simple example we have a complete characterization of the cross-
sectional consumption distribution and how it varies with the parameters of the
model. However, the resulting consumption distribution is somewhat trivial.
since the model consists of only two agents. For serious quantitative work we
therefore would like a similar model, but with a large number of agents, in
some sense a counterpart of Aiyagaris (1994) or Huggetts (1993) model for
a world with limited commitment. The formulation and numerical solution of
such model is discussed next.
6.1.6 A Continuum Economy
In this section we formulate a version of the above model with a continuum of
innitely lived agents. Let j
|
denote an agents idiosyncratic income process,
which is assumed to be a nite state Markov chain with transition probabilities
and associated invariant distribution H. As before an endowment shock history
is denoted by j
|
, with probability of that history occurring, conditional on
the initial income shock j
0
, equal to (j
|
[j
0
). We again assume a law of large
numbers so that (j
|
[j
0
) is also the deterministic fraction of the population that
started with j
0
and has experienced history j
|
. It is also assumed that the initial
income distribution (and hence the income distribution at any future date)
is given by H. Agents in this continuum economy face the same commitment
problem as the agents in the simple economy before.
Let 1(a
0
, j
0
) denote the initial distribution over assets and income. We
rst want to compute and characterize constrained-ecient allocations and then
decentralize them as equilibria with borrowing constraints that are not too tight,
in the spirit of Alvarez and Jermann (2000). Note that an agents initial wealth
level (and initial income realization) determines how much expected discounted
lifetime utility this agent can obtain. Let this utility level be given by n
0
; we
will formally establish the mapping between (a
0
, j
0
) and (n
0
, j
0
) below. Let the
initial distribution of utility promises and income be denoted by w(n
0
, j
0
).
Previously we solved for the constrained-ecient consumption allocation by
maximizing one agents expected utility, subject to the promise-keeping con-
straint of delivering at least a certain minimum utility level to the other agent
(and, of course, subject to the individual rationality constraints). With a con-
tinuum of agents this is evidently impossible, since there would be a continuum
of promise-keeping constraints. Atkeson and Lucas (1992, 1995) propose a dual
approach to overcome this problem: instead of maximizing someones utility,
subject to enforcement constraints, the resource constraint and individual ratio-
nality constraints one minimizes the cost of delivering a given level of promised
utility to a particular agent; the distribution of utility promises is then adjusted
so as to preserve resource feasibility.
Atkeson and Lucas rst formally dene constrained eciency (in their dual
sense), then formulate a sequential social planners problem that solves for
constrained-ecient allocations and then make this problem recursive and nally
prove that the policy functions from the recursive problem induce constrained-
ecient sequential consumption allocations c
|
(n
0
, j
|
). We will directly go to
the recursive formulation; readers interested in the details may consult Krueger
(1999).
In order to make the cost minimization operative we rst have to endow the
social planner with a time discount factor that measures the relative price of
resources today versus tomorrow. Let this shadow interest rate be denoted by
1 (1,
l
o
). We will see later that this 1 will correspond to the risk-free interest
rate in the decentralization of the allocations. Also dene the function C = n
l
denote the inverse of the period utility function n. The entity C(n) is inter-
preted as the amount of current consumption needed to generate period utility
n. The key idea of Atkeson and Lucas is to realize that, since the individual
rationality constraints involve continuation utilities, one may let the planner al-
locate utilities instead of consumption and use promises of expected discounted
utility as state variables.
6
The function C may then be used to translate utility
allocations back into consumption allocations.
So let the individual state variables in the dynamic programming problem
be (n, j), that is, the current promise of expected discounted future utility an
agent enters the period with and the current income realization j. In order to
satisfy this utility promise the planner can either give the agent current utility
/ or expected utility from tomorrow onward, conditional on tomorrows income
shock j
t
, q(j
t
). The dynamic programming problem the planner solves is then
6
Of course this is not their idea; it goes back at least to Abreu (1986) and Spear and
Srivastava (1987).
given by
\ (n, j) = min
|,(
0
)
_
1
1
1
_
C(/)
1
1
0
Y
(j
t
[j)\ (q(j
t
), j
t
) (6.32)
s.t.
n = (1 ,)/ ,

0
Y
(j
t
[j)q(j
t
) (6.33)
q(j
t
) _ l
.u|
(j
t
) (6.34)
Here \ (n, j) is the total (normalized) resource cost the planner has to min-
imally spend in order to fulll his utility promises n, without violating the
individual rationality constraint of the agent: he cant promise less from tomor-
row onwards, in any state of the world, than the agent would obtain from the
autarkic allocation. As before, multiplying the current cost by the factor 1
l
1
expresses total cost in the same units as current cost; it is an innocuous nor-
malization that just changes the units in which \ (n, j) is measured. Note that
if the income process is iid, the variable j does not appear in the minimization
problem and hence the value function \ does not depend on j.
It is a standard exercise to show that the operator induced by this functional
equation is a contraction, that the resulting unique xed point \ is dierentiable
and strictly convex (strict convexity is not that straightforward, but note that
since n is strictly concave, C is strictly convex) and that the resulting policies
/(n, j) and q(n, j; j
t
) are single-valued continuous functions. Furthermore / is
strictly increasing in n and q is either constant at l
.u|
(j
t
) or strictly increasing
in n as well.
Attach Lagrange multiplier ` to the promise-keeping constraint () and mul-
tipliers (j
t
[j)j(j
t
) to the individual rationality constraints (6.84). The rst
order conditions and envelope conditions read as
_
1
1
1
_
C
t
(/) = `(1 ,) (6.35)
1
1
\
t
(q(j
t
), j
t
) = ,` j(j
t
) (6.36)
\
t
(n, j) = ` (6.37)
Combining equations (6.86) and (6.87) and the complementary slackness con-
ditions yields
\
t
(q(j
t
), j
t
) = 1,\
t
(n, j) j(j
t
) (6.38)
or
\
t
(q(j
t
), j
t
) _ 1,\
t
(n, j)
= if q(j
t
) l
.u|
(j
t
) (6.39)
Now suppose that income is iid and that 1, < 1, something that was true
in general equilibrium in the standard incomplete markets model in the last
chapter and is true (under weak conditions) in this model as well (see again
Krueger (1999) or Krueger and Perri (2006b)). Then the Euler equation (6.80)
becomes
\
t
(q(j
t
)) < 1,\
t
(n)
= if q(j
t
) l
.u|
(j
t
) (6.40)
Since \ is strictly convex, this implies that either q(n; j
t
) < n or q(n; j
t
) =
l
.u|
(j
t
). That is, for states in which the individual rationality constraints are
not binding utility promises for tomorrow are lower than for today; for states
in which the constraint is binding utility promises are raised suciently in
order to prevent the agent from reneging. Dene n = min
Y
l
.u|
(j) and
n = max
Y
l
.u|
(j). It is clear that for all n we must have q(n; j
t
) _ n.
On the other hand we have that for j
nax
= aig max
Y
l
.u|
(j) we have
q(l
.u|
(j
nax
), j
nax
) = l
.u|
(j
nax
) and for all n l
.u|
(j
nax
) we have q(n, j
t
) _
q(n, j
nax
) _ l
.u|
(j
t
). Thus for iid income shocks the state space for promised
utility n, denoted by \, is bounded: \ = [n, n[. For 2 income shocks the
fourth gure at the end of this chapter demonstrates the situation; for a formal
proof of the assertions above see Krueger and Perri (2006b). You may want it
instructive to follow an agent with given initial utility promise n
0
through his
life with a sequence of income shocks: one positive income shock brings utility
to n = n, with a sequence of bad income shock making the agent move down
in promised utility, until he hits n = n, with a single good shock putting him
back at n = n.
So far resource feasibility and the determination of the distribution of promised
utilities has not been discussed; in fact, the beauty of Atkeson and Lucas (1992,
1995) approach is that it separates the dynamic programming problem the plan-
ner has to solve for a given individual (n, j) from the problems solved for other
individuals in society. We want to nd a gross real shadow interest rate and as-
sociated invariant measure over utility promises such that the resources available
to the planner equals the resources given out by the planner.
The remaining discussion of this model has strong similarities with the dis-
cussion in Aiyagari (1994) and Huggett (1993) for standard incomplete markets
models. First, for a given interest rate 1, the dynamic programming problem
(??) delivers value function \ (n, j) and policy functions /(n, j) and q(n, j; j
t
).
In order to nd the associated stationary distribution over utility promises and
income shocks w
1
we rst determine the Markov transition function Q
1
in-
duced by and q(n, j; j
t
). First one has to prove that there is an upper bound
for utility promises n, denoted by n (this is straightforward for being iid, but
not for the general case). Then denote the state space for utility promises by
\ = [n, n[, dene 7 = \ 1 and E(7) = E(\) T(1 ). Then the transition
function Q
1
: 7 E(7) [0, 1[ is given by
Q
1
((n, j), (, )) =

_
(j
t
[j)
0
if q(n, j; j
t
)
else
(6.41)
The invariant measure w
1
over (n, j) then satises
w
1
(, ) =
_
Q
1
((n, j), (, ))dw
1
(6.42)
Of course one has to prove that such invariant measure exists and is unique,
the proof of which for the iid case is contained in Krueger (1999). For the
general, non-iid case Im not aware of any such result, which is similar to the
status of theoretical results available for the standard incomplete markets model
discussed in the previous chapter.
The last gure at the end of this chapter shows the stationary consump-
tion distribution associated with w
1
(n, j), again for the iid case with only two
possible income shocks. Since consumption C(/(n, j)) is strictly increasing in
utility promises n, the consumption distribution mimics the utility distribution
w
1
. In particular, all agents with currently high shock consume the same, and
the maximum in the consumption distribution, with agents with a sequence
of bad shocks working themselves down through the consumption distribution
until they have hit rock bottom.
So for a given intertemporal shadow interest rate 1 we now know how de-
termine the associated stationary utility promise distribution. So far nothing
assures that the social planner can satisfy this distribution of utility promises
with the aggregate resources available to society. Total resources available are
_
jdw
1
=
jH(j) (6.43)
How about total resources needed by the planner? An agent that enters this
period with utility promises n and current income j is awarded current utility
/(n, j). Thus requires, in terms of resources, C(/(n, j)). Thus total resources
required to satisfy utility distribution w
1
in the current period (and by station-
arity in each period) are
_
C(/(n, j))dw
1
(6.44)
and therefore the excess resource requirements, as a function of the gross shadow
interest rate 1, are given by
d(1) =
_
(C(/(n, j)) j) dw
1
(6.45)
To nish the determination of a stationary constrained-ecient allocation re-
quires to nd an 1
+
such that d(1
+
) = 0. Computationally this can be done
by searching over 1 (1,
l
o
) for such 1
+
. Theoretically, for the iid case one
can show that d(1) is a continuous and increasing function in 1 on (1,
l
o
).
With additional assumptions on n and the income process one can show that
lim
1
1
{
d(1) 0 and that lim
1l
d(1) _ 0, proving existence of a stationary
constrained-ecient consumption allocation. Since it is hard to prove that d(1)
is strictly increasing, uniqueness is hard to establish (obviously the set of e-
cient 1
t
: is convex). Again see Krueger (1999) or Krueger and Perri (2006b)
for the details.
Finally we can decentralize a constrained ecient stationary consumption
allocation as a sequential markets equilibrium with borrowing constraints that
are not too tight, in the spirit of Alvarez and Jermann. Agents trade a full set of
Arrow securities; in this model these are securities that pay out conditional on
individual income realizations, rather than aggregate realizations. The determi-
nation of the borrowing constraints that are not too tight is as before and hence
omitted. Finally, the price of an Arrow security bought/sold today by an agent
with current individual income shock j that pays out one unit of consumption
if tomorrows income shock for that agent is j
t
turns out to be
(j
t
[j) = (j
t
[j) =
(j
t
[j)
1
(6.46)
which justies that 1 is in fact called a shadow interest rate: it turns out to
be the equilibrium risk free interest rate in the corresponding equilibrium with
borrowing constraints that are not too tight.
Finally we want to discuss the relationship between initial promised utilities
n
0
and initial assets a
0
. So suppose we have found a stationary constrained-
ecient utility distribution w(n, j), a corresponding 1
+
and associated value
and policy functions \ (n, j), /(n, j) and q(n, j; j
t
). First, analogously to (6.1)
and (6.16) from the recursive policy functions we can construct sequential constrained-
ecient consumption allocations c
|
(n
0
, j
0
) for an agent with initial condi-
tions (n
0
, j
0
). What is missing is the connection between initial conditions
(n
0
, j
0
) and (a
0
, j
0
) (and the corresponding relationship between w(n
0
, j
0
) and
1(a
0
, j
0
)). Dene Arrow-Debreu prices associated with (6.46) as
Q(j
0
) = H(j
0
)
Q(j
|
) = (j
|
[j
|l
) + . . . (j
l
[j
0
)H(j
0
) (6.47)
Then initial assets associated with (n
0
, j
0
), denoted by a
0
(n
0
, j
0
), are given
as
a
0
(n
0
, j
0
) =
o
|=0
t
]0
Q(j
|
)
_
c
|
(n
0
, j
|
) j
|
_
(6.48)
and the associated equilibrium consumption allocations are given by
c
|
(a
0
, j
|
) = c
|
(a
l
0
(a
0
, j
0
), j
|
) (6.49)
where n
0
= a
l
0
(a
0
, j
0
) is the inverse function of (6.48) with respect to the rst
argument (note that this function is well-dened because a
0
(n
0
, j
0
) is strictly
increasing in n
0
). The distribution 1(a
0
, j
0
) is then determined as
1(a
0
, j
0
) = w(a
l
0
(a
0
, j
0
), j
0
) (6.50)
What one then can prove (see Krueger (1999) is that for an initial distrib-
ution 1(a
0
, j
0
) given by (6.0), the allocation determined by (6.40) and prices
6.2. PRIVATE INFORMATION 127
(6.46) are a competitive equilibrium with borrowing constraints that are not
too tight (or alternatively, form a competitive equilibrium in the spirit of Kehoe
and Levine (1993), where equilibrium prices are given by (6.47)).
A few nal comments: the method of using utility promises as state variables
to make dynamic contracting problems recursive has been used in a number
of applications, for example in the study of optimal unemployment insurance
(Shavell and Weiss 1979, Hopenhayn and Nicolini 1997, Zhao 2000), asset pric-
ing and the equity premium (Alvarez and Jermann 2000, Lustig 2001), sovereign
debt (Atkeson 1991, Kletzer and Wright 2000), redistributive taxation (Krueger
and Perri 2006b, Attanasio and Rios-Rull 2001), time-consistent monetary pol-
icy (Chang 1998), time-consistent scal policy (Phelan and Stacchetti 1999,
Sleet 1998) and risk sharing in village economies (Udry 1994, Ligon, Thomas
and Worrall 2001).
Also, instead of using utility promises Marcet and Marimon (1999) have
developed a recursive method that use (cumulative) Lagrange multipliers as
state variables. As before, the same problems with the curse of dimensionality
when the number of heterogeneous agents becomes large arises.
So what next: on the methodological side one would like to gure out how
to handle models with a large number of agents and aggregate uncertainty, both
theoretically and numerically. Lustig (2001) takes a big step in that direction.
On the substantial side a careful study of redistribution and insurance over the
business cycle, of optimal insurance of large income risks in the future, both
by government policies and private arrangements, and of the dynamics of the
income, consumption and wealth distribution seems to be fruitful avenues for
future research.
6.2 Private Information
In this section we consider another friction that may prevent perfect risk sharing
from occurring as a result of ecient or equilibrium consumption allocations. As
in standard Arrow Debreu theory now households and nancial intermediaries
can write legally binding and enforceable contracts. However, we now assume
that individual income realizations (and individual consumption) is private in-
formation of the agent. Financial intermediaries or the social planner have to
rely on reports of income by agents. Consumption allocations therefore have to
structured in such a way that households nd it optimal to tell the truth about
their income realization, rather than to lie about it. We will rst consider the
problem of a nancial intermediary dealing with a single agent in isolation, be-
fore discussing a model with many agents, an aggregate resource constraint and
an endogenous interest rate
6.2.1 Partial Equilibrium
Our treatment of the partial equilibrium case is motivated by the seminal pa-
pers by Green (1987) and Thomas and Worrall (1990), which in turn is nicely
discussed in Ljungqvist and Sargents book. Consider a risk-neutral nancial
intermediary that lives forever and discounts the future at time discount factor
, (0, 1). This nancial intermediary (sometimes called principal) engages in
a long-term relationship with a risk averse household (sometimes called agent)
that also discounts the future at factor ,. The agent faces a stochastic income
process j
|
, assumed to be iid with nite support 1 = j
l
, j
2
, . . . j
and
probabilities (
l
,
2
, . . . ,
); and seeks insurance from the risk-neutral princi-

pal. Both parties can commit to long-term contracts, so that in the absence of
private information the optimal consumption allocation for the agent, subject
to the nancial intermediary breaking even, is
c
|
(j
|
) = 1(j
|
) = 1(j) = 1
where the last equality is by assumption (that is, we normalize mean income to
1). Thus the agent hands over his realized income in every period and receives
constant consumption equal to mean income back from the nancial intermedi-
ary. His lifetime utility equals
l(c) = (1 ,)
o
|=0
|
(j
|
),
|
n(c
|
(j
|
)) = n(1)
and expected utility of the nancial intermediary (or prots) equal
\(c) = (1 ,)
o
|=0
|
(j
|
),
|
_
j
|
c
|
(j
|
)
_
= (1 ,)
o
|=0
,
|
|
(j
|
) (j
|
1(j
|
))
= 0
The big problem with this consumption allocation is that the agents in-
come realizations are private information. If the agent is promised constant
consumption independent of his income report, then he would always report
j
|
= j
l
, keep the dierence between his true income and the report, receive
1 from the nancial intermediary and do strictly better. Then, however, the
principal would lose money, since his prots would be
\ = (1 ,)
o
|=0
|
(j
|
),
|
(j
l
1)
= j
l
1 < 0.
Thus the tension in the current environment is between the provision of insur-
ance (without the informational frictions it is ecient for the nancial inter-
mediary to insure the agent, since he is better taking risk because of his risk-
neutrality) and the provision of incentives to tell the truth (and we saw above
that without providing these incentives the principal is going to lose money).
What we want to construct in the following is the ecient long-term in-
surance contract between the two parties, explicitly taking into account the
informational frictions in this environment. We will again immediately proceed
to the recursive formulation of the problem, understanding that in principle
one should rst write down the sequential problem, then the recursive problem
and then prove the principle of optimality.
7
The recursive formulation of the
contracting problem again makes use of promised lifetime utility n as a state
variable. Let us rst pose the dynamic programming problem and then explain
it:
\ (n) = min
]|s,us]
1
s=1
s
[(1 ,)t
s
,\ (n
s
)[ (6.51)
s.t.
n =

s
[(1 ,)n(t
s
j
s
) ,n
s
[ (6.52)
(1 ,)n(t
s
j
s
) ,n
s
_ (1 ,)n(t
|
j
s
) ,n
|
\:, / (6.53)
t
s
[j
s
, ) (6.54)
n
s
[n, n[ (6.55)
We cast the optimal contracting problem as the problem of minimizing cost of
the contract for the principal, subject to delivering lifetime utility n to the agent
(as assured by the promise-keeping constraint (6.2)), subject to the agent hav-
ing the correct incentives to report income truthfully (see constraints (6.8)),
and subject to the domain restrictions (6.4) and (6.) which we will discuss be-
low.
8
The choice variables are the state-dependent transfers the principal makes
to the agent today, t
s
, and the state-dependent continuation utility promises n
s
from tomorrow onwards.
The constraints (6.8) say the following: suppose state : with associated
income realization j
s
is realized. If the agent reports the truth, he receives
transfers t
s
and continuation utility n
s
, for total lifetime utility
n(t
s
j
s
) ,n
s
.
If, on the other hand, he falsely reports j
|
, he receives transfers t
|
and contin-
uation utility n
|
, for total lifetime utility
n(t
|
j
s
) ,n
|
.
Note that even when mis-reporting income, his true lifetime utility depends on
his true, rather than his reported income j
s
. The constraints (6.8) simply state
7
In fact, the incentive constraints in the sequential formulation are much more complex.
That they can be simplyed to period-by-period incentive constraints (so-called temporary
incentive compatibility constraints, in Greens (1987) language) is one of the major results
proved in Greens paper. Only because of this result do we obtain a simple recursive formu-
lation of the problem.
8
We can restrict ourselves to contracts that induce truthtelling behavior because the reve-
lation principle assures that whatever insurance can be provided with an arbitrary mechanism
can also be provided with a mechanism (contract) in which agents report their income directly
(a direct mechanism) and are provided with the appropriate incentives to report income truth-
fully.
that it has to be in the agents own interest to truthfully reveal his income.
The domain restriction on transfers is self-explanatory: the principal can make
arbitrarily large transfers, but cannot force the agent to make payments bigger
than their current (truthfully reported) income j
s
. With respect to the bounds
on promised utility we note the following. Let n = sup
c
n(c) and n = inf
c
n(c);
evidently it is not possible to deliver more lifetime utility that n even with
innite resources and it is not possible to deliver less lifetime utility than n even
without giving the agent any consumption even. We explicitly allow n =
and n = , but will make more restrictive assumptions later. For now we
assume that
Assumption: The utility function n : [0, ) R is strictly increasing,
strictly concave, at least twice dierentiable and satises the Inada conditions.
So far is has been left unclear what determines the n the agent will start
the contract with. A higher initial n means more lifetime utility for the agent,
but less lifetime utility for the principal. The expected lifetime utility for the
principal, conditional on delivering n to the agent, is given by
\(n) = 1 \ (n)
since \ (n) measures the lifetime expected costs from the contract and 1 mea-
sures the expected revenues from the agent, equal to his expected per period and
(by normalization of 1 ,) lifetime revenue. Thus varying the n traces out the
constrained utility possibility frontier between principal and agent. One partic-
ularly important n is that n that solves \ (n) = 1, that is, that lifetime utility
of the agent that yields 0 prots for the principal. If \ is strictly increasing in
n, such n is necessarily unique.
Properties of the Recursive Problem
By the standard contraction mapping arguments a unique solution \ to the
functional equation exists Let us rst bound \ (n). The cheapest way to pro-
vide lifetime utility n is by delivering constant consumption. This may not be
incentive compatible, in fact it will not be, but surely provides a lower bound
on \ (n). So dene c
}b
(n) as
n(c) = n
or c
}b
(n) = n
l
(n). The value of the transfers needed to deliver the constant
consumption stream c
}b
(n) are given by
\ (n) = c
}b
(n) 1.
Obviously
\ (n) _ \ (n)
with strict inequality if any of the incentive constraints is binding. On the other
hand, the principal can decide to give constant transfers t
s
= t (and constant
continuation utility n
s
= n) Surely, since transfers are independent of reported
income, this induces truth-telling. In order to obtain the utility promise n, the
transfer has to satisfy

s
s
n(j
s
t) = n
Denote this transfer by

t(n). The cost of this policy is given by
\ (n) =

t(n)
and evidently
\ (n) _

\ (n)
with strict inequality whenever the distribution of income shocks is not degen-
erate.
Observe the following facts. Since c
}b
(n) is strictly increasing and strictly
convex, so is \ (n). Furthermore, as long as n satises the Inada conditions we
have \
t
(n) = 0 and \
t
( n) = , since n(.) and c
}b
(.) are inverse functions
of each other. Finally we have \ (n) = 1 and \ ( n) = . Similarly, the
function

t(n) and thus

\ (n) are strictly increasing and strictly convex, with
\ ( n) = . While this does not necessarily mean that \ (n) is strictly increasing
(it is straightforward to show that it is) and strictly convex (it is harder to show
that it is), this discussion gives us a fairly tight bound on \ (n).
The fact that \ (n) is strictly increasing in n is intuitive: delivering higher
lifetime utility to the agent requires higher transfers on average by the principal.
Since marginal utility is declining, a given additional unit of transfers increases
utility by less and less, thus one would expect the cost function to be strictly
convex. We will go ahead and assert this without proof.
Now we discuss further properties of the optimal contract. Consider the
constraints
(1 ,)n(t
s
j
s
) ,n
s
_ (1 ,)n(t
sl
j
s
) ,n
sl
(6.56)
(1 ,)n(t
sl
j
sl
) ,n
sl
_ (1 ,)n(t
s
j
sl
) ,n
s
(6.57)
Adding them results in
n(t
s
j
s
) n(t
sl
j
sl
) _ n(t
sl
j
s
) n(t
s
j
sl
)
or
n(t
s
j
s
) n(t
sl
j
s
) _ n(t
s
j
sl
) n(t
sl
j
sl
)
Since n is strictly concave and j
s
j
sl
we have t
sl
_ t
s
(it helps to draw a
picture to convince you of this). From (6.6) it then easily follows that n
s
_
n
sl
. That is, it is ecient to give households with lower income realizations
higher transfers. But in order to provide the incentives of not always claiming to
have had low income realizations, these households are punished with lower
continuation utilities. We summarize this in the following
Proposition 35 In an ecient contract, lower income reports are rewarded
with higher current transfers, but lower continuation utilities: t
sl
_ t
s
and
n
s
_ n
sl
for all : o
Second we claim that the principal only has to worry about the agent lying
locally, that is, if income is j
s
, the principal has to only worry about the agent
reporting j
sl
or j
sl
. As long as the agent does not have an incentive to lie
locally, he does not want to lie and report j
s2
or j
s2
or even more extreme
values. Formally
Proposition 36 Suppose that the true income state is : and that all local con-
straints of the form
(1 ,)n(t
s
j
s
) ,n
s
_ (1 ,)n(t
sl
j
s
) ,n
sl
(6.58)
(1 ,)n(t
s
j
s
) ,n
s
_ (1 ,)n(t
sl
j
s
) ,n
sl
(6.59)
hold. Then all other incentive constraints are satised.
Proof. We will show that, if these constraints are satised, then a household
would not want to report :1 when the true state is :1. A simple repetition of
the argument will then show that the household would not want to report : 1
for any state / _ :. The case of mis-reporting upwards is, of course, symmetric
and hence omitted.
Since : : 1, we have, from the previous result, t
s
_ t
sl
and thus (again
by concavity of n)
n(t
s
j
sl
) n(t
s
j
s
) _ n(t
sl
j
sl
) n(t
sl
j
s
) (6.60)
Multiplying (6.60) by (1 ,) and adding it to (6.0) yields
(1 ,)n(t
s
j
sl
) ,n
s
_ (1 ,)n(t
sl
j
sl
) ,n
sl
But
(1 ,)n(t
sl
j
sl
) ,n
sl
_ (1 ,)n(t
s
j
sl
) ,n
s
(use (6.0), but for : 1) and thus
(1 ,)n(t
sl
j
sl
) ,n
sl
_ (1 ,)n(t
s
j
sl
) ,n
s
_ (1 ,)n(t
sl
j
sl
) ,n
sl
That is, if the household does not want to lie one state down from : to : 1,
he does not want to lie two states down, from : 1 to : 1 either.
This result reduces the set of incentive constraints substantially. It turns
out that we can reduce it even further. Without proof we state the following
proposition (if you are interested in the proof, consult Thomas and Worrall
(1990) or the discussion of the same paper in Sargent and Ljungqvists book;
note that the proof requires strict convexity of the cost function \ ).
Proposition 37 The local downward constraints (6.0) are always binding, the
local upward constraints (6.8) are never binding.
Proof. Omitted
Intuitively, this result makes a lot of sense. Remember, the purpose of the
contract is for the principal to insure the agent against bad income shocks, with
high transfers in states with bad income realizations. This, of course, triggers
the incentive to report lower income than actually realized, rather than higher
income. Thus the lying-down constraints are the crucial constraints. With this
result the dynamic programming problem becomes more manageable. In par-
ticular, if = 2, there is only one incentive constraint and the promise keeping
constraint. It is then quite feasible to characterize the qualitative properties of
the optimal contract in more detail, which we leave to the discussion of Atkeson
and Lucas (1992) in class.
Let us consider instead a simple example with = 2 income states. From
our characterization of the binding pattern of the constraints the dynamic pro-
gramming problem becomes
\ (n) = min
]|1,|2,u1,u2]
[(1 ,)t
l
,\ (n
l
)[ (1 ) [(1 ,)t
2
,\ (n
2
)[
s.t.
n = [(1 ,)n(t
l
j
l
) ,n
l
[ (1 ) [(1 ,)n(t
2
j
2
) ,n
2
[
(1 ,)n(t
2
j
2
) ,n
2
_ (1 ,)n(t
l
j
2
) ,n
l
where is the probability of low income j
l
. Let us proceed under the assumption
that \ is dierentiable (which is not straightforward to prove). The the rst
order and envelope conditions read as
(1 ,) `(1 ,)n
t
(t
l
j
l
) j(1 ,)n
t
(t
l
j
2
) = 0
(6.61)
(1 )(1 ,) `(1 )(1 ,)n
t
(t
2
j
2
) j(1 ,)n
t
(t
2
j
2
) = 0
(6.62)
,\
t
(n
l
) `, j, = 0
(6.63)
(1 ),\
t
(n
2
) `(1 ), j, = 0
(6.64)
\
t
(n) = `
(6.65)
Rewriting equations (6.68) and (6.64) and using (6.6) yields
` = \
t
(n) = \
t
(n
l
)
j
= \
t
(n
2
)
j
1
Thus, since j 0, we have (by strict convexity of the cost function \ ) that
n
l
< n < n
2
, that is, utility promises spread out over time. Rewriting (6.61)
and (6.62) yields
1 = `n
t
(t
l
j
l
)
j
n
t
(t
l
j
2
)
=
_
`
j
_
n
t
(t
l
j
l
)
j
[n
t
(t
l
j
l
) n
t
(t
l
j
2
)[
_
`
j
_
n
t
(t
l
j
l
)
1 =
_
`
j
1
_
n
t
(t
2
j
2
)
and thus
1 = \
t
(n
2
)n
t
(c
2
) = \
t
(n
l
)n
t
(c
l
)
j
[n
t
(c
l
) n
t
(t
l
j
2
)[
\
t
(n
l
)n
t
(c
l
)
The rst equality (and again using strict convexity of the cost function) show
that if future promises n
2
are increased (in response to, say, an increase in n),
then it is necessarily optimal to also increase c
2
. That is, the principal should
spread costs over time, increasing both current costs and well as future costs.
[To be completed]
6.2.2 Endogenous Interest Rates in General Equilibrium
See the papers by Phelan and Townsend (1991) and Atkeson and Lucas 1992,
1995 should be discussed
Bibliography
[1] Aiyagari, R. (1994): Uninsured Risk and Aggregate Saving, Quarterly
Journal of Economics, 109, 659-684.
[2] Aiyagari, R. (1995): Optimal Capital Income Taxation with Incomplete
Markets, Borrowing Constraints, and Constant Discounting, Journal of
Political Economy, 103, 1158-1175.
[3] Aiyagari, R. and McGrattan, E. (1998): The Optimum Quantity of
Debt, Journal of Monetary Economics, 42, 447-469.
[4] Altug, S. and Miller, R. (1990): Household Choices in Equilibrium,
Econometrica, 58, 543-70.
[5] Atkeson, A. and Lucas, R. (1992): On Ecient Distribution with Private
Information, Review of Economic Studies, 59, 427-453.
[6] Atkeson, A. and Lucas, R. (1995): Eciency and Equality in a Sim-
ple Model of Ecient Unemployment Insurance, Journal of Economic
Theory, 66, 64-88.
[7] Attanasio, O. (1992): Intertemporal Substitution in Consumption: The-
ory and Evidence, Carnegie Rochester Series on Public Policy, supple-
ment to Journal of Monetary Economics.
[8] Attanasio, O. (1999): Consumption, in Handbook of Macroeconomics,
J. Taylor and M. Woodford (eds.), Elsevier Science.
[9] Attanasio, O. and De Leire, T. (1994): IRAs and Household Saving
Revisited: Some New Evidence, NBER Working Paper 4900.
[10] Attanasio, O. and Weber, G. (1993): Consumption Growth, the Interest
Rate and Aggregation, Review of Economic Studies, 60, 631-649.
[11] Attanasio, O. and Weber, G. (1995): Is Consumption Growth Consistent
with Intertemporal Optimization? Evidence from the Consumer Expen-
diture Survey, Journal of Political Economy, 103, 1121-1157.
135
136 BIBLIOGRAPHY
[12] Attanasio, O., J. Banks, C. Meghir and G. Weber, (1999): Humps and
Bumps in Lifetime Consumption, Journal of Business and Economic
Statistics.
[13] Barro, R. (1974): Are Government Bonds Net Wealth?, Journal of Po-
litical Economy, 82, 1095-1117.
[14] Ben-Porath, Y. (1967), The Production of Human Capital and the Life
Cycle of Earnings, Journal of Political Economy, 75, 352-365.
[15] Bernheim, D. (1991): How Strong are Bequest Motives? Evidence Based
on Estimates on the Demand for Annuities, Journal of Political Economy,
99, 899-927.
[16] Blanchard, O. and Mankiw, G. (1988): Consumption: Beyond Certainty
Equivalence, American Economic Review Papers and Proceedings, 78,
173-77.
[17] Browning, M.. and Lusardi, L. (1996): Household Saving: Micro Theories
and Micro Facts, Journal of Economic Literature, 34, 1797-855.
[18] Browning, M. and Meghir, C. (1991): The Eects of Male and Female
Labour Supply on Commodity Demands, Econometrica 59, 925-51.
[19] Browning, M., Deaton, A. and Irish, M. (1985): A Protable Approach
to Labor Supply and Commodity Demands over the Life-Cycle, Econo-
metrica, 53, 503-43.
[20] Caballero, R. (1990): Consumption Puzzles and Precautionary Savings,
Journal of Monetary Economics, 25, 113-36.
[21] Campbell, J. and Deaton, A. (1989): Why is Consumption so Smooth?,
Review of Economic Studies, 56, 357-74
[22] Campbell, J. and Mankiw, G. (1989): Consumption, Income, and Inter-
est Rates: Reinterpreting the Time Series Evidence, in NBER Macroeco-
nomics Annual, The MIT Press.
[23] Carroll, C. (1992), The Buer-Stock Theory of Saving: Some Macroeco-
nomic Evidence, Brookings Papers on Economic Activity, 61-135.
[24] Carroll, C. (1997): Buer-Stock Saving and the Life Cycle/Permanent
Income Hypothesis, Quarterly Journal of Economics, 112, 1-55.
[25] Carroll, C. and Samwick, A. (1997): The Nature of Precautionary Sav-
ing, Journal of Monetary Economics, 40, 41-71.
[26] Caselli, F. and Ventura, J. (2000), A Representative Consumer Theory
of Distribution, American Economic Review, 90, 909-926.
BIBLIOGRAPHY 137
[27] Chamberlain, G. (1984): Panel Data, in Handbook of Econometrics, vol.
2, Zvi Griliches and Michael D. Intriligator (eds.) Elsevier Science.
[28] Chamberlain, G. and C. Wilson (2000): Optimal Intertemporal Con-
sumption under Uncertainty, Review of Economic Dynamics, 365-395.
[29] Chatterjee, S. (1994), Transitional Dynamics and the Distribution of
Wealth in a Neoclassical Growth Model, Journal of Public Economics,
54, 97-119.
[30] Cochrane, J. (1991), A Simple Test of Consumption Insurance, Journal
of Political Economy, 99, 957-976.
[31] Conesa, J. and Krueger, D. (1999): Social Security Reform with Hetero-
geneous Agents, Review of Economic Dynamics, 2, 757-795.
[32] Conesa, J. and Krueger, D. (2006): On the Optimal Progressivity of the
Income Tax Code, Journal of Monetary Economics, 53, 1425-1450.
[33] Cooley, T. and Soares, J. (1999): A Positive Theory of Social Security
Based on Reputation, Journal of Political Economy, 107, 135-160.
[34] Cox, D (1987): Motives for Private Income Transfers, Journal of Polit-
ical Economy, 95, 508-46.
[35] Deaton, A. (1985): Panel Data from Time Series of Cross-Sections,
Journal of Econometrics, 30, 109-26.
[36] Deaton, A. (1991): Saving and Liquidity Constraints, Econometrica,
59, 1221-1248.
[37] Deaton, A. (1993) Understanding Consumption, Oxford University Press.
[38] Deaton, A. (1997): The Analysis of Household Surveys, The Johns Hop-
kins University Press.
[39] Deaton, A. and Paxson, C (1994): Intertemporal Choice and Inequality,
Journal of Political Economy, 102, 384-94.
[40] Deaton, A. and Paxson, C. (1997): The Eects of Economic and Popula-
tion Growth on National Saving and Inequality, Demography, 34, 97-114.
[41] Diamond, P. and Hausman, J. (1984): Individual Retirement and Saving
Behaviour, Journal of Public Economics, 23, 81-114.
[42] Dynan, K. (1993): How Prudent are Consumers?, Journal of Political
Economy, 101, 1104-13.
[43] Engen, E., Gale, W. and Scholz, J. (1996): The Illusory Eect of Saving
Incentives on Saving, Journal of Economic Perspectives, 10, 113-38.
138 BIBLIOGRAPHY
[44] Farber, H. and Gibbons, R (1996), Learning and Wage Dynamics, Quar-
terly Journal of Economics, 111, 1007-1047.
[45] Fernandes, A. and Phelan, C. (2000): A Recursive Formulation for Re-
peated Agency with History Dependence, Journal of Economic Theory,
91, 223-247.
[46] Fernandez-Villaverde, J. and D. Krueger (2002), Consumption and Sav-
ing over the Life Cycle: How Important are Consumer Durables? Pro-
ceedings of the 2002 North American Summer Meetings of the Economet-
ric Society: Macroeconomic Theory.
[47] Fernandez-Villaverde, J. and D. Krueger (2006), Consumption over the
Life Cycle: Facts from the Consumer Expenditure Survey forthcoming,
Review of Economics and Statistics.
[48] Flavin, M. (1981): The Adjustment of Consumption to Changing Ex-
pectations About Future Income, Journal of Political Economy, 89, 974-
1009.
[49] Gal, J. (1990): Finite Horizon, Life-Cycle Savings and Time Series Evi-
dence on Consumption, Journal of Monetary Economics, 26, 433-52.
[50] Gollier, C. (2002): What Does the Classical Theory Have to Say About
Household Portfolios?, in Household Portfolios, L. Guiso, M. Haliassos,
T. Jappelli (eds.), MIT Press, Cabridge, MA.
[51] Gourinchas, P. and Parker, J. (1999): Consumption over the Life Cycle,
Econometrica 70, 47-89.
[52] Green, E. (1987): Lending and the Smoothing of Uninsurable Income,
in: Prescott, E. and Wallace, N. (eds) Contractual Arrangements for In-
tertemporal Trade, University of Minnesota Press, 3-25.
[53] Guiso, L., Jappelli, T. and Terlizzese, D. (1992): Earnings Uncertainty
and Precautionary Saving, Journal of Monetary Economics 30, 307-37.
[54] Guvenen, F. (2005): An Empirical Investigation of Labor Income
Processes, mimeo, University of Texas
[55] Guvenen, F. (2006): Learning your Earning: Are Labor Income Shocks
Really Very Persistent? forthcoming, American Economic Review.
[56] Haliassos, M. and Bertaut, C. (1995): Why Do so Few Hold Stocks?,
Economic Journal, 105, 1110-29.
[57] Hall, R. (1978): Stochastic Implications of the Life Cycle-Permanent
Income Hypothesis: Theory and Evidence, Journal of Political Economy,
86, 971-987.
BIBLIOGRAPHY 139
[58] Hall, R. (1988): Intertemporal Substitution in Consumption, Journal
of Political Economy, 86, 971-87.
[59] Hansen, L. and Singleton, K. (1983): Consumption, Risk Aversion and
the Temporal Behavior of Stock Market Returns, Journal of Political
Economy 91, 249-265.
[60] Hayashi, F. (1982): The Permanent Income Hypothesis: Estimation and
Testing by Instrumental Variables, Journal of Political Economy, 90,
895-916.
[61] Hayashi, F. (1985): The Permanent Income Hypothesis and Consump-
tion Durability: Analysis Based on Japanese Panel Data, Quarterly Jour-
nal of Economics, 100, 1083-113.
[62] Hayashi, F. (1987): Tests for Liquidity Constraints: A Critical Survey
and Some New Observations, in: Bewley, T. (ed.) Advances in Economet-
rics, Vol. 2, Fifth World Congress, Cambridge University Press, 91-120.
[63] Hayashi, F. (1997) Understanding Saving, MIT Press.
[64] Hayashi, F., Altonji, J. and Kotliko, L. (1996): Risk Sharing Between
and Within Families, Econometrica, 64, 261-94.
[65] Heathcote, J. (2003): Fiscal Policy with Heterogeneous Agents and In-
complete Markets, forthcoming in Review of Economic Studies.
[66] Heathcote, J., K. Storesletten and G. Violante (2004), The Welfare Im-
plications of Rising Wage Inequality in the US, Mimeo, New York Uni-
versity.
[67] Heaton, J. (1995): An Empirical Investigation of Asset Pricing with
Temporally Dependent Preference Specications, Econometrica, 63, 681-
717.
[68] Heckman, J. (1974), Life Cycle Consumption and Labor Supply: An
Explanation of the Relationship between Income and Consumption Over
the Life Cycle, American Economic Review, 64, 188-194.
[69] Hubbard, G., Skinner, J. and Zeldes, S. (1994): The Importance of
Precautionary Motives in Explaining Individual and Aggregate Saving,
Carnegie Rochester Series on Public Policy, 40, 59-125.
[70] Hubbard, G., Skinner, J. and Zeldes, S. (1995): Precautionary Saving
and Social Insurance, Journal of Political Economy, 103, 360-99.
[71] Huggett, M. (1993): The Risk-Free Rate in Heterogeneous-Agent
Incomplete-Insurance Economies, Journal of Economic Dynamics and
Control, 17, 953-969.
140 BIBLIOGRAPHY
[72] Huggett, M. and Ventura, (1999): On the Distributional Eects of Social
Security Reform, Review of Economic Dynamics, 2, 498-531.
[73] Hurd, M. (1987): Savings of the Elderly and Desired Bequests, Ameri-
can Economic Review, 77, 298-312.
[74] Hurd, M. (1989): Mortality Risk and Bequests, Econometrica, 57, 779-
813.
[75] Kantor, S. and Fishback, P. (1996): Precautionary Saving, Insurance,
and the Origin of Workers Compensation, Journal of Political Economy,
104, 419-42.
[76] Kehoe, T. and D. Levine (2001): Liquidity Constrained Markets versus
Debt Constrained Markets, Econometrica, 69, 575-598.
[77] Kimball, M. (1990): Precautionary Saving in the Small and in the Large,
Econometrica, 58, 53-73.
[78] King, M. and Dycks Mireaux, L. (1982): Asset Holdings and the Life-
Cycle, Economic Journal 92, 247-67.
[79] Klein, P. and Rios-Rull, V. (1999): Time-Consistent Optimal Fiscal Pol-
icy, International Economic Review, 44, 1217-1246.
[80] Kocherlakota, N. (1996a): Implications of Ecient Risk Sharing without
Commitment, Review of Economic Studies, 63, 595-609.
[81] Kocherlakota, N. (1996b): The Equity Premium: Its Still a Puzzle,
Journal of Economic Literature, 34, 42-71.
[82] Kotliko, L. (1988): Intergenerational Transfers and Savings, Journal
of Economic Perspectives, 2, 41-58.
[83] Kotliko, L. and Summers, L. (1981): The Role of Intergenerational
Transfers in Aggregate Capital Accumulation, Journal of Political Econ-
omy, 89, 706-32
[84] Krueger, D. and H. Lustig (2006): The Irrelevance of Market Incom-
pleteness for the Price of Aggregate Risk, NBER Working Paper 12634.
[85] Krueger D. and F. Perri (2004), On the Welfare Consequences of the
Increase in Inequality in the United States, 2003 NBER Macroeconomics
Annual.
[86] Krueger D. and F. Perri (2005), Understanding Consumption Smoothing:
Evidence from the US Consumer Expenditure Survey, Journal of the
European Economic Association, 3, 340-250.
[87] Krueger D. and F. Perri (2006a), Does Income Inequality Lead to Con-
sumption Inequality: Evidence and Theory, Review of Economic Studies,
73, 163-193.
BIBLIOGRAPHY 141
[88] Krueger, D. and F. Perri (2006b): Public versus Private risk Sharing,
mimeo, University of Pennsylvania.
[89] Krusell, P., Quadrini, V. and Rios-Rull, V. (1997): Politico-Economic
Equilibrium and Economic Growth, Journal of Economic Dynamics and
Control, 21, 243-72.
[90] Krusell, P. and Smith, A. (1998): Income and Wealth Heterogeneity in
the Macroeconomy, Journal of Political Economy, 106, 867-896.
[91] Krusell, P. and Smith, A. (2003): Consumption and Savings Decisions
with Quasi-Geometric Discounting, Econometrica, 71, 365-375.
[92] Laibson, D. (1997): Golden Eggs and Hyperbolic Discounting, Quar-
terly Journal of Economics, 112, 443-77.
[93] Lambertini, L. (2003): Endogenous Debt Constraints in Life Cycle
Economies, Review of Economic Studies, 70, 1-27.
[94] Lettau, M. and Uhlig, H. (1999): Rules of Thumb versus Dynamic Pro-
gramming, American Economic Review, 89, 148-174.
[95] Mace, B. (1991), Full Insurance in the Presence of Aggregate Uncer-
tainty, Journal of Political Economy, 99, 928-956.
[96] MaCurdy, T. (1982): The Use of Time Series Processes to Model the
Error Structure of Earnings in a Longitudinal Data Analysis, Journal of
Econometrics, 18, 82-114.
[97] Mankiw, G. and Zeldes, S. (1991): The Consumption of Stockholders
and Non-Stock-Holders, Journal of Financial Economics, 29, 97-112.
[98] Mariger, R. (1987): A Life-Cycle Consumption Model with Liquidity
Constraints: Theory and Empirical Results, Econometrica, 55, 533-57.
[99] Meghir, C. and Weber, G. (1996): Intertemporal Non-Separability or
Borrowing Restrictions? A Disaggregate Analysis Using a US Consump-
tion Panel, Econometrica, 64, 1151-82.
[100] Mehra, R. and Prescott, E. (1985): The Equity Premium: A Puzzle,
Journal of Monetary Economics, 15, 145-161.
[101] Modigliani, F. (1986): Life Cycle, Individual Thrift, and the Wealth of
Nations, American Economic Review, 76, 297-312.
[102] Modigliani, F. (1988): The Role of Intergenerational Transfers and Life
Cycle Saving in the Accumulation of Wealth, Journal of Economic Per-
spectives, 2, 15-40.
[103] Pischke, J. (1995): Individual Income, Incomplete Information, and Ag-
gregate Consumption, Econometrica, 63, 805-39.
142 BIBLIOGRAPHY
[104] Pistaferri, L. (2001): Superior Information, Income Shocks and the Per-
manent Income Hypothesis, Review of Economics and Statistics, 83, 465-
476.
[105] Quadrini, V. (1997): Entrepreneurship, Saving and Social Mobility, In-
stitute for Empirical Macroeconomics Discussion Paper 116.
[106] Runkle, D. (1988): Liquidity Constraints and the Permanent Income
Hypothesis, Journal of Monetary Economics, 27, 73-98.
[107] Sargent, T. (1993) Bounded Rationality in Macroeconomics, Clarendon
Press.
[108] Schechtman, J. (1976): An Income Fluctuation Problem, Journal of
Economic Theory, 12, 218-241.
[109] Schechtman, J. and Escudero, V. (1977): Some Results on An Income
Fluctuation Problem, Journal of Economic Theory, 16, 151-166.
[110] Shea, J. (1995): Union Contracts and the Life-Cycle Permanent Income
Hypothesis, American Economic Review, 85, 186-200.
[111] Shorrocks, A. (1975): The Age Wealth Relationship: A Cross-Sectional
and Cohort Analysis, Review of Economics and Statistics, 57, 155-63.
[112] Stoker, T. (1993), Empirical Approaches to the Problem of Aggregation
Over Individuals, Journal of Economic Literature, 31, 1827-1874
[113] Townsend, R. (1994): Risk and Insurance in Village India, Economet-
rica, 62, 539-91.
[114] Udry, C. (1994), Risk and Insurance in a Rural Credit Market: An Em-
pirical Investigation in Northern Nigeria, Review of Economic Studies,
61, 495-526.
[115] Wilson, R. (1968): The Theory of Syndicates, Econometrica, 36, 119-
132.
[116] Zeldes, S. (1989): Consumption and Liquidity Constraints: an Empirical
Investigation, Journal of Political Economy, 97, 305-46.

Cons Book

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Cons Book

Transféré par

Droits d'auteur :

Formats disponibles

Consumption and Saving: Theory and Evidence

is total private consumption expenditures and 1

) denote the price at period t of a contract that pays out

. These contracts are often called Arrow securities, contingent claims

= sup&[an < & for at most nite number of as

we have, via continuing substitution

the more is the

where j, c, . are parameters. It is easy to show that, up to irrelavant constants, CRRA is a

It is straightforward to see that for o = this formulation coincides with stan-

= 0 and let the consump-

, :) is needed to induce that same consumption c = 10. Since saving

, c), for small, zero mean uncertainty, dened implicitly as

and so forth. We obtain several implications

, ) and any j 1 we then can write Bellmans equation

., ?) is the probability of going from state . to set ? in steps.

. Doing this for all

) be the Borel o-algebra of R

) and M be the set of all

< -. If yes, we are done and should save

. The fraction of people unem-

) = 0.2. Finally, the proba-

is the set of time indices for which :

= 1 for , = q, / then households

= 0.000008 for , = /, q and o

metric), but the

); and seeks insurance from the risk-neutral princi-

Vous aimerez peut-être aussi