Intertemporal Macroeconomics

Gernot Doppelhofer*

May 2009
Forthcoming in J. McCombie and N. Allington (eds.),
Cambridge Essays in Applied Economics, Cambridge UP

This chapter reviews models of intertemporal choice consumption demand and labour
supply. We discuss optimal decisions by individuals at the microeconomic level and
the implications for the aggregate economy. The chapter describes the equilibrium in
a market-clearing neoclassical model and analyses effects of productivity and
government sector shocks on optimal decisions by consumers and workers.
Predictions from the benchmark neoclassical model are contrasted with alternative
theories and aggregate data for the US and UK.

JEL Code: D91, E12, E13, E21, E24, E62, H30

Department of Economics, Norwegian School of Economics and Business

Administration (NHH), Helleveien 30, N-5045 Bergen, Norway; email:
G.Doppelhofer@nhh.no. I thank Xavier Sala-i-Martin, whose macroeconomics
lectures at Columbia University provided inspiration and guidance to the material in
this chapter, Donald Robertson for helpful discussions and the derivation of the
Keynes-Ramsey rule, and several cohorts of students at Cambridge for helpful
feedback. I thank Bhaavit Agrawal for excellent research assistance, the Economics
Department at UW-Madison for their hospitality while completing this chapter and
Trinity College for financial support. All errors are mine.

1
Table of Contents
1 Introduction ................................................................................................................. 1 
2 Static model of consumption and leisure choice......................................................... 2 
2.1  Substitution and income effect....................................................................... 3 
3 Intertemporal Consumption Choice ............................................................................ 5 
3.1  A two period model ....................................................................................... 5 
3.2  Intertemporal substitution and income effects ............................................... 7 
3.3  Intertemporal optimisation and the Euler equation ........................................ 8 
3.4  Consumption functions in multi-period models .......................................... 10 
3.4.1  Modigliani’s life cycle hypothesis (LCH) ............................................... 10 
3.4.2. Friedman’s permanent income hypothesis (PIH) ...................................... 12 
3.5  Role of uncertainty and Hall’s random walk hypothesis ............................. 14 
3.6  Departures from classical consumers........................................................... 15 
4 Intertemporal Labour Demand and Supply............................................................... 18 
4.1 Labour Demand ................................................................................................. 18 
4.2 Intertemporal labour supply ............................................................................... 18 
4.3 Empirical Evidence of Labour Supply ............................................................... 20 
5 Equilibrium in goods and labour market and productivity shocks ........................... 20 
5.1 Aggregation........................................................................................................ 21 
5.2 Productivity shocks ............................................................................................ 22 
5.2.1 Permanent productivity shock..................................................................... 22 
5.2.2 Temporary productivity shock .................................................................... 22 
5.2.3 Theoretical predictions and stylised facts ................................................... 24 
6 Government............................................................................................................... 24 
6.1 Effect of government spending .......................................................................... 25 
6.1.1 Permanent change in government spending ............................................... 26 
6.1.2 Temporary change in government spending ............................................... 27 
6.1.3 Effects of government spending in Keynesian Model ................................ 28 
6.2 Budget Deficits and Ricardian equivalence ....................................................... 28 
6.3. Taxes ................................................................................................................. 29 
7 Concluding Remarks ................................................................................................. 31 
References .................................................................................................................... 32 

2
1 Introduction
This chapter gives an overview of the literature on intertemporal macroeconomics.
We review models of individual, optimising behaviour by consumers and workers,
discuss implications for aggregate variables and contrast the theoretical predictions
with empirical facts for the US and UK economies.

The advantage of using an optimising model is that we gain a better understanding of

the aggregate economy. Instead of simply postulating ad hoc macroeconomic
relations, we understand the microeconomic foundations of aggregate variables such
as consumption and labour supply. Starting from microeconomic foundations ensures
internal consistency of our macroeconomic models. The drawback of a microfounded
approach is that going from individual to aggregate behaviour requires strong
assumptions.

We start by analysing the neo-classical model as a benchmark. Consumers choose

optimally how much to consume and save over time given the path of income and the
interest rate. Workers supply labour (demand leisure) each period, given the path of
wages and the interest rate. Perfectly competitive firms hire labour at the market-
clearing wage and we can derive the equilibrium in goods, labour and bond markets.
Next we can analyse the effects of productivity shocks and changes in government
spending and tax decisions on the model. We contrast the predictions of the model
with empirical observations and discuss implications of alternatives theories.

For reference, we list the following key aggregate quarterly data for the United States
averaged over the 1959-96 period (from Barro 1997, Table 1.1).

Standard Correlation
Components of GDP Share of GDP
Deviation with GDP
GDP 1.000 0.017 1.00
Consumption 0.589 0.008 0.83
Investment 0.203 0.065 0.93
Government spending 0.226 0.017 0.02

Other variables
Employment - 0.010 0.81
Worker hours - 0.015 0.88
Output per employee - 0.011 0.84
Real wage - 0.011 0.48
Real Interest Rate - - 0.23

This chapter is organised as follows: Section 2 briefly describes a simple static model
of consumption and leisure choice. Section 3 discusses models of intertemporal
consumption choice. Section 4 introduces intertemporal supply and demand for labour
and discusses empirical evidence. Section 5 derives the equilibrium in the neoclassical
model and analyses the effect of productivity shocks on optimal choices. Section 6
introduces the government sector into the model and section 7 concludes.

1
2 Static model of consumption and leisure choice
Consider first the optimal demand for consumption and leisure in a highly simplified
static model of an economy. A representative1 individual has preferences over
consumption and leisure are represented by a utility function

(1) u (c,1 − l )

where c represents individual2 consumption and l the supply of labour. The total time
endowment is normalised to 1 and (1-l) represents leisure. We assume that utility
increases in both consumption and leisure, represented by positive marginal utilities
uc>0 and u1-l>0, and that the ratio of marginal utilities uc/u1-l falls as the
consumption/leisure ratio rises. The preferences can be represented by convex
indifference curves shown in Figure 1. The individual prefers more consumption and
more leisure (less work) and moderate amounts of both compared to extremes of only
one of them.

For each unit of labour l the consumer earns a wage w measured in terms of the
homogenous final good3. The set of feasible choices (budget constraint) is therefore
given by

(2) c = wl + b

where b is the stock of wealth which is independent of labour income. The budget
constraint is shown as upward sloping line in figure 1 with slope w. Increases in initial
wealth b correspond to upward parallel shifts of the constraint.

The problem for the consumer is to maximise the utility function (1) subject to the
budget constraint (2). Formally, we can set up the Lagrangian and find the optimal
choices of consumption and leisure

max u (c,1 − l ) + λ [wl + b − c ]

c ,l

The first order conditions are uc=λ and u1-l=wλ, where λ is the Lagrange multiplier
associated with the budget constraint (2). Combining those two conditions implies
that at the optimum the marginal rate of substitution (MRS) between leisure and
consumption equals the wage rate

u1−l
MRS1−l ,c = =w
uc

1
Aggregation of preferences across individuals requires very restrictive assumptions; see Deaton
(1992) and section 5.1 below for a discussion.
2
Throughout this chapter the notation convention is that lowercase denotes individual and uppercase
aggregate variables. Economy-wide prices such as the interest rate are written as lowercase.
3
This chapter abstracts from money and therefore does not discuss the role of nominal rigidities (sticky
prices or wages). For an excellent discussion of such issues see Blanchard (1990).

2
The optimal choice of consumption c* and labour supply l* is shown in Figure 1.
Graphically the optimum can be found at the point where an indifference curve is
tangent to the budget constraint.

c
wl
better-off region

c*

0 l* 1 l

Figure 1: Optimal choice of consumption and labour supply (leisure)

2.1 Substitution and income effect

Consider next the effect of changes of the wage rate on the optimal choice of
consumption and labour supply. Suppose that the wage rate increases to w’ > w which
corresponds to a steeper slope of the budget constraint in Figure 2. The change in
wages can be decomposed into substitution and income effect.

The substitution effect for consumption and leisure demand is associated with a
change in the relative price of labour, the wage rate. We analyse the effect of an
increase in the wage rate to w’ in the neighbourhood of the original optimal choice
(c*,l*). We assume real income to be unchanged so that the initial choice is still
feasible4. Figure 2 shows the new optimum (c’,l’) where a higher indifference curve is
tangent to the dashed line obtained by rotating the budget constraint counter-

4
Alternatively, we could define the substitution effect as change in demand resulting from a change in
relative price while keeping the consumer equally well-off (on the same indifference curve). For small
changes these two definitions coincide.

3
clockwise around the initial point (c*,l*). Intuitively, the substitution effect induces the
consumer to work more (enjoy less leisure) and increase consumption.

The income effect (sometimes called wealth effect) is due to the higher wage at every
level of labour supplied l. Figure 2 shows the parallel shift from the dashed line to the
new (solid) budget constraint. Note that both lines share the same steeper slope w’.
The income effect implies more demand for both consumption and leisure which are
usually assumed to be normal goods (movement from c’,l’ to c**,l**).

The total effect of an increase of wages is the combination of substitution and income
effects: For consumption, the two effects reinforce each other and imply an rise from
c* to c** in Figure 2. For leisure (and labour supply) substitution and income effect
work in opposite directions and the overall effect is ambiguous depending on the
relative strengths of substitution and income effects.

c
w’l

wl

c**

c’
c*

0 l* l’ 1 l
l**

Figure 2: Substitution and income effect on consumption and labour resulting

from wage increase

4
Empirically we observe increased levels of consumption and falling labour supply
(measured by weekly hours) as economies are developing and wages are rising. For
example, the average number of weekly hours in manufacturing has fallen from
between 55 and 60 in 1890 to 42 in 1996 in the United States and has fallen from 60
in 1850 to 44 in 1994 for the UK (see Barro 1997, pp. 76-77). Similar secular declines
of work hours can be observed in other countries over time, however the relationship
appears to become weaker at high levels of income (weekly hours have not fallen in
the US over the last 50 years).

3 Intertemporal Consumption Choice

3.1 A two period model
Consider next the optimal choice of consumption over time. Suppose that individuals
live for two periods, (denoted by subscripts 1 and 2). Each period, individuals earn
income y and chose how much to consume c or alternatively save by purchasing
bonds b paying a constant5 interest rate r. The budget constraint for the two periods
are

y1 + b0 (1 + r ) = c1 +b1
y2 + b1 (1 + r ) = c2 +b 2

where b0 denotes the level of assets at the beginning of the individuals life and b2 at
the end of its lifetime. The budget constraints for each period can be rewritten
alternatively as intertemporal budget constraint

y2 c b
(4) y1 + + b0 (1 + r ) = c1 + 2 + 2
1+ r 1+ r 1+ r

which states that the present value6 of lifetime income and initial wealth equals the
present value of lifetime spending. Individuals live for 2 periods and are assumed to
care only about their own consumption in periods 1 and 2 (we will relax this
assumption in section 3.5.2). Therefore, optimal assets in the final period of
individuals life will not be positive, that is b2≤0, as long as consumers are not satiated.
On the other hand, individuals are not allowed to leave behind debt, so b2≥0. In
equilibrium each generation is therefore born with zero wealth (b0=0) and the
intertemporal budget constraint therefore simplifies to

y2 c
(5) y1 + = c1 + 2
1+ r 1+ r

The intertemporal budget constraint (5) is shown in Figure7 3 as the black solid line
that passes through the endowment point (y1,y2) and has slope –(1+r). Consumers can

5
Throughout this chapter we assume the interest rate to be fixed and known. In a more realistic model,
the interest rate is determined in capital markets and changes over time. For a discussion of variable
interest rates see for example Obstfeld and Rogoff (1996, pp. 76-78).
6
Given interest rate r the present value at time 0 of an amount Xt t periods in the future is defined as
PV(Xt)≡Xt/(1+r)t. The term 1/(1+r)t can be viewed as the relative price of Xt in terms of current goods.
7
Such figures introduced by Irving Fisher (1907) are know as Fisher diagrams.

5
chose any combination of first and second period consumption (c1,c2) on this line and
the optimal choice depends on preferences over consuming in the first or second
period of their life. We assume diminishing marginal utility of consumption and
individuals prefer to smooth consumption between time periods. Indifference curves
in Figure 3 are therefore negatively sloped and convex to the origin.

Figure 3 shows examples of different types of individuals:

1. Saver: consumers are relatively patient, they save in period 1 by chosing c1*<y1 and
consume their income and savings in period 2 giving c2*>y2.
2. Borrower: consumers are relatively impatient, they chose c1*>y1 and borrow in
period 1 which they repay by saving in period 2 and hence c2*<y2.
3. Autarkist: individuals consume at the endowment point and are neither saving nor
borrowing, c1*=y1 and c2*=y2.

c2
slope −(1+r)

c2*=y2

0 c1*=y1 c1

Figure 3: Optimal consumption choice between periods 1 and 2

How does the optimal consumption choice respond to changes in income? Notice that
individuals want to smooth consumption over time (because of diminishing marginal
utility of consumption) and they can borrow and lend freely at interest rate r.
Consumption each period is therefore a function of the present value of lifetime
income, PV(y) ≡ y1 + y2/(1+r) in equation (5). An increase in income in one period
only, say y1, leads to less than proportional increases of consumption each period.
However, consumption moves proportional to changes in the present value of income.

6
3.2 Intertemporal substitution and income effects
Next consider the effect of a change in the interest rate on optimal consumption
choice. A rise in the interest rate to r’>r implies an increase of the relative price of
current (period 1) relative to future (period 2) goods which induces the consumer to
substitute away from current consumption towards future consumption. The
intertemporal substitution effect leads to a fall in current consumption and increase in
future consumption (increase in savings). Figure 4 shows the intertemporal
substitution effect by moving from the initial point (c1*,c2*) to the new optimum at
(c1’,c2’). Notice that the dashed budget constraint with slope –(1+r’) passes through
the initial point (c1*,c2*) which is still feasible.

The intertemporal income effect associated with a rise of the interest rate depends on
whether the consumer is a saver or borrower in the first period. Figure 4 shows the
case of a saver in period 1. In this case the higher interest rate raises lifetime income
(because of higher return on savings) and the intertemporal income effect is positive
leading to more consumption in both periods. For a borrower the intertemporal
income effect of a rise in interest rates is negative (due to higher debt repayment)
leading to less consumption in both periods.

slope −(1+r’)
c2

c2**
c2’
c2*

slope −(1+r)

y2

0 c1’ c1* y1 c1
c1**

Figure 4: Intertemporal substitution and income effect of interest rate increase

on optimal consumption choice in period 1 and 2

The total effect is a combination of intertemporal substitution and income effects. For
a saver, the two effects reinforce each other and imply higher consumption in period

7
2, but have opposite signs, hence ambiguous total effect on consumption in period 1.
The new optimal point is (c1**,c2**) in Figure 4. A borrower on the other hand reduces
c1 since both effects reinforce each other, but the overall change is ambiguous for c2.

For the aggregate economy only the intertemporal substitution effect is relevant since
there is no net saving or borrowing in equilibrium (see section 5 below). Aggregate
consumption demand (denoted by superscript d) depends therefore negatively on the
interest rate and positively on the present value of aggregate income

C1d ( r , PV (Y ),...)
(−) (+)

3.3 Intertemporal optimisation and the Euler equation

We can analyse the consumer’s problem more formally. Suppose that individuals
have additively separable preferences over first and second period consumption:

(6) max u (c1 ) + βu (c2 )

c1 , c2

where the period utility function u(.) is strictly concave u’>0, u’’<0 and β is the
subjective discount factor measuring the degree of impatience of the individual. We
assume that the individual discounts future utility with 0<β<1.

The problem of the consumer is to maximise lifetime utility (6) subject to the
intertemporal budget constraint (5). For example, we can use the substitution method
and substitute for c2 from (5) into (6):

max u (c1 ) + β u[(1 + r )( y1 − c1 ) + y 2 ]

c1

The necessary first-order condition for this problem is also called an intertemporal
Euler equation8

(7) u ' (c1 ) = (1 + r ) β u ' (c2 )

The Euler equation determines optimal consumption choice over time. The left hand
side of equation (7) is the marginal utility of consumption in period 1 which measures
how much the individual values one unit of consumption. If the individual saves one
unit she would receive (1+r) units of consumption next period which would increase
lifetime utility by the marginal utility of consumption (discounted by β) on the right
hand side of (7). Alternatively, the Euler equation can be written as

β u ' (c 2 ) 1
=
u ' (c1 ) 1+ r

which equates the marginal rate of substitution between c1 and c2 on the left hand side
to the relative price 1/(1+r). This optimality condition can also be seen in Figure 3

8
It is named after the Swiss mathematician Euler who investigated dynamic systems more generally.

8
where the indifference curve representing the trade-off between current and future
consumption is tangent to the intertemporal budget constraint with slope –(1+r).

In order to further understand intertemporal consumption choice consider the

following example of a power utility function

c1−θ − 1
(8) u (c ) =
1−θ

where the parameter θ>0 controls the curvature of the utility function and 1/θ
measures the constant intertemporal elasticity of substitution (IES) of consumption
between time periods. Utility function (8) is also called isoelastic utility function. The
marginal utility of consumption in this case is simply u ' (c) = c −θ and the Euler
equation (7) becomes
(c2 / c1 )θ = β (1 + r )
To interprete this condition, rewrite the discount factor as β =1/(1+ρ), where ρ>0 is
the rate of time preference. For small changes in consumption ∆c≡(c2-c1), the left
hand side of this expression can be approximated by

(c2 / c1 )θ = (Δc / c + 1)θ ≅ 1 + θΔc / c

For small values of r and ρ the right hand side of this expressions is approximately
equal to (1 + r ) /(1 + ρ ) ≅ 1 + r − ρ . Combining the two approximations, the Euler
equation (also known as Keynes-Ramsey rule) becomes

Δc 1
(9) = (r − ρ )
c θ

The growth rate of consumption Δc / c is determined by the difference between the

interest rate and the rate of time preference and the strength of the response depends
on the intertemporal elasticity of substitution 1/θ. Hall (1988) tests the empirical
relationship between aggregate consumption and interest rates and finds only a weak
relationship between these two variables.

Hall’s estimates of the intertemporal elasticity of substitution (IES) are shown in

Table 1. The estimates depend on which asset is used to calculate underlying interest
rates. The first three rows of Table 1 are based on annualised postwar data of expected
returns based on inflation expectations from the Livingston survey9. The point
estimates range from 0.066 for returns on 400 S&P stocks to 0.346 for US Treasury
bills. However, the estimated standard errors are so large that the point estimates are
not statistically different from zero. The estimates based on the sample starting in
1924 and realised returns gives marginally significant, negative estimates of the IES.

9
The Livingston survey is based on expectations of many variables by a panel of economists.

9
Table 1: Hall’s (1988) estimates of intertemporal elasticity of substitution (IES)

Interest rate based on IES estimate Standard error Sample

T Bills (expected) 0.346 (0.337) 1959-83

Savings Accounts 0.271 (.330) 1959-83

Stocks 0.066 (0.050) 1959-83

T Bills (realised) -0.40 (0.20) 1924-40, 1950-83

Beaudry and Van Wincoop (1996) use panel data for US States instead of nationwide
data. For the period 1953-91 they strongly reject the zero estimate of the IES and find
estimates close to one. Several studies have also used microeconomic data to address
problems of aggregation and lack of control for demographic and labour market
effects. Attanasio and Weber (1993) and Blundell, Browning and Meghir (1994) find
for both the US and the UK estimates of the IES just below one. Browning, Hansen
and Heckman (1999) survey general equilibrium models and implied empirical
estimates of the IES using micro data (see for example their Tables 3.1 and 3.2). They
argue that the IES parameter is often poorly determined, and that there is evidence
varies with demographic characteristics and wealth.

3.4 Consumption functions in multi-period models

The analysis for the two period case can be generalised to many periods. The next two
sections discuss the case of finite and infinite life-time.

3.4.1 Modigliani’s life cycle hypothesis (LCH)

In the 1950s Franco Modigliani developed with co-authors the life cycle hypothesis
(LCH) to describe consumption and savings behaviour over individuals’ lifetime10.
Suppose that individuals live for T periods and each period t face a budget constraint

yt + bt −1 (1 + r ) = ct +bt

Similar to the two-period case we can derive the intertemporal budget constraint
saying that the present value of lifetime income and initial wealth equals the present
value of lifetime spending

y2 yT
y1 + + ... + + b0 (1 + r ) =
1+ r (1 + r )T −1
c cT bT
c1 + 2 + ... + T −1
+
1+ r (1 + r ) (1 + r )T −1

10
Here we follow the simple version presented in Modigliani’s (1986) Nobel lecture.

10
Since individuals have finite horizons they leave behind no assets as bequests for
future generations and set bT to zero. For simplicity, assume that individuals earn a
constant labour income y until retiring at R years, but no more labour income until
the expected end of life at time T. We assume that individuals prefer a smooth
consumption profile over the life-time.
Figure 5 shows the allocation of consumption and assets over the lifetime. Individuals
accumulate wealth until retirement and draw down the stock of wealth until life ends
to ensure a smooth path of consumption; more formally, the Euler equation (7) shows
that individuals want to smooth marginal utility across time periods. The life cycle
model predicts the following path of consumption and assets: before entry into the
labour market individuals should borrow, they should accumulate savings while
working and dissave after they retire. Figure 5 shows the resulting hump shaped path
of wealth.

wealth

y
c

R T age

Figure 5: Modigliani’s life cycle hypothesis (LCH)

The life cycle hypothesis has other important implications. Consumption responds
little to temporary changes in income and proportionally to permanent changes; also
the marginal propensity to consume out of current income depends on age.

Some empirical observations seem at odds with the simple LCH model. First, young
individuals consume too little compared to expected life-time income. A high
marginal propensity to consume could point to myopia or liquidity constraints (see
section 3.6 for a discussion). Second, consumption seems to first increase and later
fall in line with labour income which appears at odds with consumption smoothing.
Browning and Crossley (2001) argue that precautionary savings (prudence) and
demographic changes (children in family) could explain these changes over the
working life. Third, the elderly dissave too little after retirement and consumption

11
falls discretely at retirement. These observations might be explained by precautionary
savings motive or the importance of bequests (see section 6.2 below).

3.4.2. Friedman’s permanent income hypothesis (PIH)

Suppose instead that generations are linked to each other. Assume that each
generation lives for one period and cares about its own utility and the utility of the
next generation:

t = 1 : u (c1 ) + βu (c2 )
t = 2 : u (c2 ) + βu (c3 )
...

Lifetime utility of generation born at time 1 is therefore

∞
(10) u (c1 ) + β u (c2 ) + β 2u (c3 ) + ... = ∑ β t −1u (ct )
t =1

Since each generations cares about the next, each generation acts as if it was infinitely
lived and had an infinite horizon. Notice that each generation discounts11 future utility
with discount factor β.

Similar to the life cycle model presented in the previous section, we can combine the
budget constraints faced by each generation t = 1,2,.. into the intertemporal budget
constraint

∞ ∞
yt ct bT
∑
t =1 (1 + r )
t −1
+ b0 (1 + r ) = ∑
t =1 (1 + r )
t −1
+ lim
T →∞ (1 + r )T −1

The left hand side shows the present value of current and future income over the
infinite horizon (discounted at relative price r) plus initial assets. On the right hand
side is the present value of consumption spending plus a term showing the present
value of wealth as the horizon is pushed towards infinity.

Remember that for the two-period and life cycle model consumers were not allowed
to leave behind debts (bT < 0). In the infinite horizon case this condition is replaced by
limT →∞ bT /(1 + r )T −1 ≥ 0 which states the present discounted value of assets must be
non-negative and that debt must not grow faster than the interest rate. Otherwise the
consumer could finance infinite consumption by borrowing ever increasing
amounts12. Furthermore, consumers will not find it optimal to accumulate savings at a
faster rate than the interest rate or else the present value of savings would be
unbounded. The optimal consumption path therefore satisfies the so-called
transversality condition

11
Frank Ramsey (1928) viewed such discounting of future welfare as “ethically indefensible” in
normative models. In a positive (descriptive) sense discounting would be consistent with “selfish”
individuals preferring current to future consumption (see section 3.6 below for discussion).
12
This condition is also known as the No-Ponzi-game condition after the swindler Charles Ponzi who
ran a pyramid scheme in Boston in the 1920s.

12
 bT
(11) lim =0
T →∞ (1 + r )T −1

With condition (11) satisfied the intertemporal budget constraint becomes

∞ ∞
yt ct
(12) ∑ (1 + r )
t =1
t −1
+ b0 (1 + r ) = ∑
t =1 (1 + r ) t −1

The problem for an individual born at time 0 is to maximize utility (10) subject to the
intertemporal budget constraint (12). The condition describing optimal consumption
for the infinite horizon problem is the familiar intertemporal Euler equation13

u ' (ct ) = (1 + r ) βu ' (ct +1 )

The intuition is exactly the same as in the two period case: the marginal utility of
consumption between periods t+1 and t equals the market valuation 1/(1+r). Along
the optimal path the consumer is indifferent to consume or save the marginal unit.

Consider the special case of β=1/(1+r). Then the Euler equation simplifies to

(13) u ' (ct ) = u ' (ct +1 )

which implies that consumption is constant over time ct = ct +1 = c . Because

consumers discount future utility at the same rate as the market interest rate,
consumers have no incentive to “tilt” the consumption path over time and hence
consumption is constant (remember the Keynes-Ramsey rule in equation (9)).

Substitute the constant level of consumption c in the intertemporal budget constraint

(12) and solving gives:

∞ ∞
c yt
∑ (1 + r )
t =1
t −1
=∑
t =1 (1 + r ) t −1
+ b0 (1 + r )

which says that the present value of consumption equals the present value of total
wealth which equals the present value of life-time income plus initial wealth. Each
period t=1,2,… individuals consume the annuity value14 of total wealth which Milton
Friedman (1957) calls permanent income yP

r ⎛ ∞ yt + s ⎞
(14) ct = ytp ≡ ⎜⎜ ∑ + bt −1 (1 + r ) ⎟⎟
1 + r ⎝ s =0 (1 + r ) s
⎠

13
It is straightforward to extend the derivation of the intertemporal Euler from the two period model
(section 3.3) to the life-cycle model with finite T periods. A formal derivation for the infinite horizon
case is beyond the scope of this chapter. For a discussion of this case see for example Dixit (1990).
14
Consuming the annuity value leaves total wealth unchanged over time. Notice that we use
∑
∞
t =1
1 /(1 + r ) t −1 = (1 + r ) / r in deriving (14).

13
Income y can be decomposed into permanent income yP and transitory income yT
defined as (y - yP). According to Friedman’s permanent income hypothesis (PIH)
consumption demand responds proportionally15 to changes in permanent income and
not to at all to changes in transitory income.

Friedman’s theory delivers strong empirical predictions. In a deterministic

environment (with no shocks) consumption and current income would be unrelated
over time. Furthermore, there would be a relationship between consumption and
permanent income when observing individuals in a cross-section. This can explain the
lower marginal propensity to consume and higher propensity to save out of current
income among individuals with below average permanent income (for example, the
difference between White and Black population in the United States).

3.5 Role of uncertainty and Hall’s random walk hypothesis

Next let us analyse the choice of intertemporal consumption when consumers face
uncertainty over future income and consumption due to the presence of random
shocks. Suppose that individuals maximise the expected value of lifetime utility

⎧∞ ⎫
max E1 ⎨∑ β t −1u (ct )⎬
c1 ,c2 ,...
⎩ t =1 ⎭

where E1 denotes mathematical expectations conditional on information available at

time 1. The expected value equals the probability weighted average of possible
outcomes. The intertemporal budget constraint is still given by equation (12) stating
that the realised (ex post) present values of consumption equal total realised wealth.

The stochastic version of the Euler equation governing optimal consumption equates
marginal utility of consumption in period 1 and expected discounted marginal utility
in period 2

u ' (c1 ) = (1 + r ) β E1 [u ' (c2 )]

Hall (1978) tests the stochastic version of life-cycle and permanent income
hypotheses theory by making the following simplifying assumptions:

Assumption 1 Suppose the utility function is quadratic

u (c) = c − ac 2 / 2, with a > 0

This assumption implies that marginal utility u ' (c) = 1 − ac is linear in consumption.
A consequence of this assumption is that the individual consumption decision exhibits
certainty equivalence which implies individuals act as if future consumption was at its
conditional mean value and ignore its variation.

Assumption 2 Assume 1+ r = 1/β.

15
In Friedman’s (1957), consumption is proportional to permanent income ctd = k(r,wealth,…) ytP. In
equation (14) the factor of proportionality k equals 1 since r=β. Note that we abstract from uncertainty
about future income (see section 3.6).

14
Assumption 2 ensures that consumers want to hold marginal utility and hence
consumption constant over time. See discussion after equation (13) above.

Under these two assumption the stochastic Euler equation becomes

(15) c1 = E1 (c2 )

Equation (15) says that consumption follows a random walk and the best estimate of
next period’s consumption is the current level of consumption. Intuitively individuals
plan to smooth consumption over time (assumption 2) and use all information
available at time 1 to make optimal forecasts of permanent income.

An implication of this theory is that changes in consumption should be unpredictable

and represent only surprise information (not known at time 1). Hall (1978) tests the
theory using postwar aggregate US data and finds that past levels of consumption and
income have no predictive power for future consumption. However, the pure LCH-
PIH theory fails because lagged levels of S&P stock market prices help to predict
changes in aggregate consumption.

Flavin (1981) looks at the joint behaviour of aggregate consumption and income and
finds that consumption exhibits excess sensitivity to anticipated changes in current
income. On the other hand Campbell and Deaton (1989) find that consumption
exhibits excess smoothness with regard to contemporaneous innovations to income.
Notice that these two observations are not mutually contradicting each other since
excess sensitivity refers to anticipated changes whereas excess smoothness is with
regards to unexpected innovations (see Abel 1990 for discussion).

Hall’s test of the random walk result relies on the maintained assumption of quadratic
utility (assumption 1) and rejection of a test by the data could imply that the
underlying LCH-PIH theories and/or the assumptions are wrong. Quadratic utility
provides a good approximation to preferences if marginal utility is not responding
strongly to fluctuations in consumption and is close to the linear certainty equivalence
case. If marginal utility is very non-linear on the other hand, individuals strongly
dislike uncertainty and will accumulate precautionary savings. Several studies
investigate the importance of precautionary savings empirically and conclude that the
answer depends crucially on preference parameters, the distribution of earnings and
the individual’s age (see Attanasio, 1999).

3.6 Departures from classical consumers

This section looks at deviations from classical (LCH and PIH) models of consumer
behaviour and discusses empirical evidence.

Keynes (1936) postulates a linear consumption function where aggregate

consumption demand depends linearly on current aggregate income:

(16) C1 = α + λY1

15
where α is a constant and λ is the marginal propensity to consume which lies between
zero and one. Keynes thought of this function as representing a “fundamental
psychological law”. Friedman (1957) strongly criticises Keynes and argues that
consumption demand depends on permanent income instead (see section 3.4.2 above).
However the classical theories (LCH and PIH) rely crucially on individuals having
access to credit markets and being able to borrow and lend freely at interest rate r to
smooth consumption over time. What happens if some consumers are liquidity
constraint?

Suppose a fraction of consumers are liquidity constrained. They would like to borrow
against future income, but cannot use it as collateral. Such liquidity constrained
(henceforth LC) consumers are forced to consume at their current income level

(17) c1LC = y1

Figure 6 shows that LC consumers would achieve higher utility by borrowing against
higher future income and consume at (c1*,c2*), but the binding constraint makes this
optimum unfeasible.

-(1+ r’)
c2

LC
y2 = c2

c2 *

-(1+ r)

LC
0 y 1 = c1 c1 * c1

Figure 6: Intertemporal choice of liquidity constrained (LC) consumer

What happens to the choice of LC consumers if the interest rate increases and the
slope of the budget constraint in Figure 6 rises? The answer is that changes in interest
rate r leave constrained consumption demand (c1LC) unchanged. Consumption demand

16
of LC consumers depends only on current income as in equation (17), and changes in
the interest rate16 or future income are not relevant.

Suppose next that a fraction λ of the population is liquidity constrained and behaves
as in (17) and the remainder (1- λ) behaves like classical consumers from section 3.1
with consumption function c1d(r, PV(y),..). Aggregate consumption is the weighted
sum of the two terms

(18) C1Total = λc1LC + (1 − λ )c1d = λy1 + (1 − λ )c1d (r , PV ( y ))

or written more compactly as Keynesian consumption function (16)

C1K ≅ α + λY1

where α ≡ (1 − λ )c1d (r , PV ( y )) is not constant, but function of the interest rate r and λ
measures the fraction of LC consumers in the economy.

Campbell and Mankiw (1989) test the validity of Euler equation (9). As alterantive
they allow for the presence of “rule-of-thumb” consumers whose consumption growth
is also affected by expected changes in income (compare this to Hall’s (1978) test)

(19) Δ log(C t +1 ) = α + σrt +1 + λΔ log(Yt +1 ) + ε t +1

Using 1953-86 US data Campbell and Mankiw estimate the share of rule-of-thumb
consumers λ to be approximately one half. Does this imply that 50% of the US
population is liquidity constrained? Not necessarily, since there are other reasons that
consumption could depend on current income: these include life-cycle effects and
precautionary motives discussed earlier, deviations from “standard” optimization (see
below) or habit formation17. The estimate of σ is not statistically different from zero
indicating little correlation between expected changes in consumption and ex ante real
interest rates. This finding is consistent with Hall’s (1988) results shown in Table 1.

Flavin’s (1981) findings of excess sensitivity of consumption to anticipated changes in

current income are consistent with the presence of liquidity constraints. Hall and
Mishkin (1982) use panel data on food expenditures and find that approximately 20%
of US households are liquidity constrained. Hayashi (1987) points out that even if
consumers are currently not constrained, the possibility of being constrained in the
future shortens their effective planning horizon.

Loewenstein and Thaler (1989) describe several anomalies in intertemporal choice

including time-inconsistent preferences or nominal anchoring. Angeletos et al (2001)
argue that the introduction of time-inconsistent preferences through hyperbolic
discounting produces can help to explain the relatively low levels of liquid assets (and
large credit card debt) over the life cycle. Hyperbolic consumers are less able to
smooth consumption which is consistent with Flavin’s excess sensitivity result.

16
For a sufficiently high interest rate, liquidity constrained consumers become unconstrained and save
in the first period. We assume the liquidity constraint binds also after changing the interest rate.
17
See Attanasio (1999) and Carroll (2001) for recent reviews of the literature.

17
4 Intertemporal Labour Demand and Supply
4.1 Labour Demand
Consider first labour demand by competitive firms. Suppose that each period firms
produce final output using a production function

(20) y = AF(l)

where F(.) exhibits positive, but diminishing returns to labour, F’(l)>0 and F’’(l)<0.
Productivity is measured by the parameter A>0 which represents factors such as the
weather or available knowledge that increase output for given private inputs. For any
given labour input l an increase in productivity A raises output in equation (20).

Suppose that perfectly competitive firms maximise profits

max
d
π = y s − wl d = AF (l d ) − wl d
l

where ld represents the firm’s labour demand. The necessary first-order condition18
implies that firms hire labour up to the point where the marginal product of labour
equals the real wage rate, w*=AF’(l). At this market clearing wage w* there is full
employment. An increase in productivity A implies a higher marginal product of
labour and hence the wage rate increases (see also section 5.2 below).

4.2 Intertemporal labour supply

Consider next the intertemporal choice of labour supply by individuals in a simple 2
period model (which extends straightforward to several periods). Household earn
wage income and consume and save in periods 1 and 2. Preferences over consumption
and leisure are given by

1
(21) max u (c1 ,1 − l1 ) + u (c2 ,1 − l2 )
c,l 1+ ρ

where the period utility function u(c,l) has the same properties as equation (1). We
assume lifetime utility to be additively separable between time period which implies
marginal utilities are not a function of consumption and leisure choices across time.
The discount factor is given by β = 1/(1+ρ), where ρ is the rate of time preference.

Assuming that individuals earn labour income in both periods, the intertemporal
budget constraint (IBC) equals

1 1
(22) c1 + c2 = w1l1 + w2l2
1+ r 1+ r

which implies that the present discounted value of lifetime consumption equals
lifetime labour income, both discounted at the interest rate r.

18
Optimal input demand here is static and depends only on current inputs and prices. They would
become dynamic if inputs hired in previous periods were used to produce output in equation (20).

18
The necessary first-order conditions19 for maximising lifetime utility (21) subject to
the intertemporal constraint (22) are

ul (1 − l1 )
(23) = w1
uc (c1 )

1+ r
(24) uc (c1 ) = u c (c 2 )
1+ ρ

ul (1 − l1 ) 1 + r w1
(25) =
ul (1 − l2 ) 1 + ρ w2

Equation (23) represents the static optimal trade-off between consumption and leisure
analysed in section 2.1. The consumer equalises the marginal rate of substitution
between consumption and leisure to the equilibrium wage rate. Equation (24) is the
same intertemporal Euler equation as equation (7) in section 3.3. Equation (25) is an
intertemporal Euler equation for leisure: at the optimum the ratio of marginal utilities
of leisure each period is set equal to the ratio of wages and a factor depending on the
interest rate and the rate of time preference.

What are the implications of equation (25)? If the interest rate r increases, individuals
substitute out of work tomorrow into work today. Individuals face a higher return on
saving today (period 1) which increases the incentive to work today and increase
lifetime income. Alternatively, the present discounted value of second period wages is
lower inducing individuals to work more today. Individual labour supply (denoted by
superscript s) is therefore a positive function of relative wages w1/w2 and the interest
rate r
l1s( r , w1 /w2 ,..)
(+) (+)

and the response depends on the intertemporal elasticity of substitution (IES) of the
labour supply (Lucas and Rapping 1969).

Notice that the Euler equation (25) can be obtained by substituting the static
consumption/leisure trade-off (23) into the consumption Euler equation (24), so the
three optimality equations are interdependent. Intuitively an optimal choice of leisure
and consumption in each period and optimal consumption path over time implies an
optimal intertemporal labour/leisure choice.

Mankiw, Rotemberg and Summers (1985) test the validity of equations (22)-(25) for
aggregate US data and find that the restrictions implied by the theory are strongly
rejected by the data. These negative results are in line with rejections of tests of the
intertemporal Euler equation (24) for aggregate consumption data (see Attanasio
1999).

19
Partial derivatives of utility with respect to consumption and leisure are denoted by uc ≡ ∂u / ∂c
and ul ≡ ∂u / ∂l , respectively. A good exercise is to verify the first-order conditions (23)-(25), by
setting up the Lagrangian L = u(c1,1-l1) + u(c2,1-l2)/(1+r) + λ[w1l1 + w2l2/(1+r) - c1 - c2/(1+r)].

19
4.3 Empirical Evidence of Labour Supply
Alogoskoufis (1987a,b) finds that the response of labour supply to growth in real
wages is more important for the number of employees than for hours worked. Table 2
summarises the estimates of intertemporal elasticity of substitution (IES) of labour
supply for the United States and the United Kingdom using annual data and total
employee numbers. Alogoskoufis (1987b) argues that the IES estimates of labour
supply appears to be lower than the static elasticity between consumption and
leisure20.

Table 2: Alogoskoufis’ (1987a,b) estimates of elasticity of labour supply*

Elasticity to relative wages w1/w2 real interest rate r

US 1948-82 1.19 0.74

(0.36) (0.18)

UK 1950-82 0.359 0.192

(0.125) (0.063)

Alogoskoufis (1987a), Table 1, row A2 for UK and Alogoskoufis (1987b), Table 1, row B5 for US. Labour supply
equation estimated by instrumental variables using number of employees and average weekly earnings.

Studies based on US microeconomic panel data typically find much smaller (often
statistically insignificant or negative) estimates of the IES of labour supply (see
surveys by Pencavel 1986, and Browning et al. 1999). Mulligan (1995) criticises
these estimates because they fail to distinguish between temporary and permanent
changes of wages. Note that equation (25) predicts that individual labour supply
should respond only to temporary changes Δ(w1/w2). Mulligan suggests to look
instead at episodes of anticipated, temporary high wages. Examples include labour
supply responses of taxi drivers to changing demand conditions, agricultural workers
due to weather conditions, seasonal variations, US labour supply during world war II
and other temporary episodes such as the construction of a gas pipeline in Alaska
(1974-77) or the Exxon Valdez oil spill in 1989. Mulligan finds estimates of
intertemporal labour supply elasticities near two.

5 Equilibrium in goods and labour market and

productivity shocks
In this section we combine consumption demand and labour supply based on
individual behaviour and derive the equilibrium in the aggregate economy. Aggregate
demand can be derived from individual consumption demand by horizontally
summing up individual consumption demand at different levels of the interest rate21

(26) Y1d = C1d ( r , PV (Y ),...)

(−) (+)

Figure 7 shows a negatively sloped aggregate demand schedule when plotted against
the interest rate. Remember that for the aggregate consumption demand only the

20
Compare equations (23) and (25) above.
21
We use of the fact that all consumers face the same economy-wide interest rate.

20
intertemporal substitution effect of changes in the interest rate is relevant since there
is no net borrowing or saving in the aggregate (closed) economy.

Similarly we can derive the aggregate supply schedule by summing individual labour
supply at different interest rates. The aggregate supply function depends negatively on
the interest rate (shown in Figure 7) and positively on relative wages and productivity:

(27) Y1s = A F [ L1s ( r , w1 / w2 ,...)]

(+) (+) (+)

In the neoclassical model the interest rate (the relative price of goods between time
periods) ensures equilibrium in goods and labour markets. Figure 7 shows the
equilibrium, where aggregate demand for goods equals aggregate supply at interest
rate r*
Y1d (r * , PV (Y ),...) = Y1s (r * , w1 / w2 ,...)

The bonds market also clears at r* ensuring that there is no net borrowing or saving in
the aggregate economy, Bd=0. This is an example of Walras’ law that implies that
when the goods and labour market are in equilibrium through adjustment of relative
prices (interest rate and relative wages), the bonds market has to clear too.

r
Y1s(r, w1/w2,..)

r*

Y1d(r,PV(Y),..)

Y1
Y*

Figure 7: Aggregate demand and supply and goods market equilibrium

5.1 Aggregation
Deaton (1992) warns of the various pitfalls when aggregating individual to aggregate
behaviour. Problems include (1) heterogeneity of preferences, (2) aggregation of
information available to individuals to form conditional expectations of future income
and prices and (3) population heterogeneity because of finite lifetimes. Attanasio

21
(1999) criticizes the practice of using aggregate macroeconomic data to estimate
microeconomic (structural) parameters since these estimates are likely to be
systematically biased. Recent business cycle research therefore often goes the
opposite route of imposing parameters estimated from microeconomic relations when
calibrating macroeconomic models. See the Prescott-Summers (1986) debate for a
discussion of the pros and cons of this calibration approach.

5.2 Productivity shocks

What are the consequences of changes in productivity on aggregate demand and
supply and the interest rate in equilibrium. Let us analyse permanent and temporary
productivity changes in turn.

5.2.1 Permanent productivity shock

Consider first a permanent productivity shock -- productivity increases in all periods
from today onwards, ∆A1=∆A2=…>0. This case is shown in Figure 8.

1. The direct (mechanical) effect is an increase in output for every level of the labour
supply and for a given interest rate which is shown as rightward shift of the
aggregate supply function.
2. The indirect effects (associated with economic decisions) are as follows:
- labour supply remains unchanged since relative wages are the same;
- consumption demand C1d increases proportionally with changes in current
income, since income changes are permanent, ΔC1d = ΔY1

The total effect combines direct and indirect effects: output increases and the interest
rate remains unchanged. This is the case because individuals have no incentive to
change intertemporal decisions leaving the relative intertemporal price r unaffected.
Notice that supply and demand curves shift by a similar amounts.

5.2.2 Temporary productivity shock

Consider next a temporary increase in productivity, only ∆A1>0, future productivity
A2,A3,… remains unchanged. Figure 9 shows the combination of the two effects:

1. The direct effect is the same as for a permanent increase in productivity above
leading to a rightward shift of aggregate supply (shift (a) in Figure 9).
2. The indirect effects are as follows:
- labour supply and hence aggregate supply increase because relative wages go up
∆(w1/w2)>0 (shift (b) in Figure 9);
- current consumption and aggregate demand shifts by a small amount, ΔC1d>0,
due to an increase in transitory income.

The total effect is a rise in output and a fall of the interest rate. This is the case
because individuals attempt to smooth consumption by saving some of the temporary
income increases for the future. Notice that in equilibrium the amount consumed
changes much more than the small shift of the consumption demand function22. This
can be explained by the combined effect of lower interest rates and higher labour
income due to higher labour supply. The strength of the effect depends as usual on the
IES (see Table 1).
22
Distinguish between a shift of and a movement along the demand function.

22
 Y1s(r, w1/w2,..)
r

r*

Y1d(r,PV(Y),..)
Y1
Y Y’

Figure 8: Effect of permanent productivity shock on output and interest rate

r Y1s(r, w1/w2,..)

(a) (b)

r’

Y1d(r,PV(Y),..)
Y1
Y Y’

Figure 9: Effect of temporary productivity shock on output and interest rate

23
5.2.3 Theoretical predictions and stylised facts
The model predicts a countercyclical response of real interest rates. For US data (see
table in section1), interest rates are weakly procyclical which can be reconciled with
the model by combining temporary and permanent productivity changes. Real wages
are procyclical which is in line with the stylised facts.

Empirically, the model predicts a procyclical response of labour supply to changes in

productivity. Hansen and Wright (1992) observe that for the US the number of hours
fluctuates much more than productivity. Together with the small variation in wages
this implies a high IES to reconcile theory and aggregate data. Alternatively, Hansen
and Wright argue that the standard neoclassical business cycle model can be
augmented with the following features to produce a better empirical fit: (i) indivisible
labour, (ii) preferences that are non-separable over time, (iii) home production and
(iv) the importance of other shocks such as government spending for business cycles
which we analyse next.

6 Government
This section introduces the government into the intertemporal model. The government
sector (local, state and federal levels combined) is a large part of our economies.
Barro (1997, p.439) shows a ratio of total government expenditures23 to GDP
averaged over the 1970-85 period of 35% for the US and 44% for the UK. Barro also
argues that the size of the government sector strongly increased in the United States
over the last 80 years which is in line with the long-run expansion of the share of
government to GDP in developed economies.

The government spends and collects revenue facing the following budget constraint
each period:

(28) Gt + Vt + rBtg = Tt + ΔBtg + ΔM s / P

Total spending on the left-hand side of (28) consists of spending on goods by the
government G, transfers payments V, and interest payments on the stock of
government bonds rBg. For simplicity we assume that government bonds pay the
same interest rate r as financial markets. Total revenue on the right-hand side of (28)
is given by revenue from taxes T, issuing government debt ΔB, and changing the
money supply ΔM s / P .

To simplify the discussion, we ignore revenue from printing money24 and set ΔMs = 0
in equation (28). Also we do not discuss transfers and set V=0 in (28), even though
they are very important component of total public spending (exceeding 50% in the
US). Large parts of transfers takes the form of redistributing through the social
security system in the US (see also the chapter on “Pensions” in this volume) or
between income groups in many European countries25.
23
Notice that total expenditures also include spending not counted as components of GDP, for example
transfers payments. This explains the difference to the entry in the table of section 1.
24
This “inflation tax” is a relatively small component of revenue for low inflation countries.
25
Such transfers could distort several intertemporal decisions, for example retirement age, personal
savings or decision to take on a job. These issues are however beyond the scope of this survey.

24
6.1 Effect of government spending
This section discusses the effect of government spending on intertemporal decisions
and the equilibrium in the neoclassical model of section 5. For now, assume that all
taxes are lump-sum , that is they are independent of the individuals decision and do
not distort economic decisions. An example of such a tax is the poll tax introduced by
the Thatcher government in the 1980s. The more realistic case of distortionary taxes
will be discussed in section 6.3.

We also assume that government never issues any debt and Btg = 0 in all periods. This
assumption is relaxed in section 6.2. Under these assumptions the government
balances its budget every period and its budget constraint (28) is simply:

Gt = Tt

Some types of government spending are likely to be productive. This would be

reflected through a positive effect of spending G in aggregate supply:

Y = AF ( L, G )

with positive effect on output FG>0 and possibly positive effects on labour
productivity as well. Examples of such productive spending are public goods such as
police, roads and infrastructure or public education.

Some other types of spending are not necessarily productive, but individuals may like
them and they enter into their utility function u(c,G) with uG>0. Examples could be
parks, public museums, spaceflights to the moon etc.

For simplicity, we will focus on the part of government spending that is neither
productive nor desirable, but is completely useless, such as buying goods and
dumping them into the sea. The advantage of this extreme assumption is that it allows
us to focus on the effect of government spending per se without additional effects on
production or preferences26.

Government spending is a component of the aggregate demand:

Y1d = C1d + G1

The intertemporal budget constraint of individuals reflects the reduction in disposable

income from the taxes. Equation (5) becomes:

y 2 − T2 c
y1 − T1 + = c1 + 2
1+ r 1+ r

Lump sum taxes affect individual demand for consumption and supply of labour in
the same way as a fall in income in sections 2.1. Consumption demand falls and
labour supply increases after an increase in taxes.

26
The effects on production and welfare can be added to the model; see for example Barro (1989).

25
6.1.1 Permanent change in government spending
In the neoclassical model the effect of government spending depends on whether the
change in government spending is permanent or temporary. Consider first a
permanent change in government spending shown in Figure 10 below:

1. The direct effect is an increase aggregate demand for a given interest rate which is
shown as rightward shift (a) of the aggregate demand function.
2. The indirect effects are as follows:
- consumption demand C1d and hence aggregate demand fall by the same amount
as government spending and taxes increase since the implied fall in disposable
income is permanent (shown as shift back of aggregate demand);
- labour supply and aggregate supply shift (b) to the right because of the negative
income effect (notice that relative wages are unchanged);
- the resulting increase in permanent income leads to a shift (c) of consumption
and aggregate demand.

The total effect of the permanent increase in government spending is an increase in

output and an unchanged interest rate. The intuition for this result is that individuals
have no incentive to alter their intertemporal decisions. Notice that supply and
demand curves shift by a similar amounts and that despite the rise in output an
increase in government spending and lump sum taxes means lower disposable income
and (through the reduction in consumption and leisure) lower welfare27.

r Y1s
Y1d
(a)
(b)

r*

(c)

Y1
Y Y’

Figure 10: Permanent change in government spending

27
The negative welfare effects of government spending have to be qualified by potential positive
effects on productivity and individual welfare discussed in section 6.1 above.

26
6.1.2 Temporary change in government spending
Consider next a temporary change in government spending shown in Figure 11.

1. The direct effect leading to the increase in aggregate demand is the same as for
permanent changes in government spending. Aggregate demand shifts by the
same amount as before.
2. The indirect effects are as follows:
- consumption demand C1d and hence aggregate demand fall by a small amount
since the fall in disposable income is only transitory (taxes will fall back to the
initial level next period);
- labour supply and aggregate supply shift also by a small amount since the
present value of life-time income is small. Notice also that relative wages are
unchanged since we consider only lump-sum taxes;

The total effect of a temporary increase in government spending is an increase in

output and a rise of the interest rate. Individuals attempt to borrow to smooth
consumption over their lifetime and this drives up the interest rate. Notice that the rise
of the interest rate leads to a further reduction of the amount consumed and an
increase in the amount of labour supplied. This can be seen in a movement along
demand and supply curve in Figure 11. In fact it is precisely the rise in interest rates
(the relative price of goods between time periods) that brings about the equilibrium.

The temporary increase in government spending leads to some crowding out of

consumption due to higher interest rates. The neoclassical model does not predict that
fiscal policy in the form of changes in government spending is neutral.

r
Y1d Y1s

r’
r*

Y1
Y Y’

Figure 11: Temporary change in government spending

27
As an empirical example, consider a temporary increase in government spending
during wartime. The model predicts that GDP increases, private consumption falls
and labour supply increases during wartimes. These predictions are confirmed by
Barro (1997, Table 2.2) when looking at data for the US in four wartime episodes: the
two world-wars, the Korean and the Vietnam war.

The model also predicts an increase of interest rates during wartime and possibly for
some time after the war ends. Barro (1997, pp. 454-59) does not find empirical
support for this prediction when comparing average interest rates during and
immediately after the four war periods of the US cited above and the US Civil War.
However, Barro finds that long-run nominal interest rates are on average 1.0
percentage points higher (relative to an average level of 3.5%) during wars the United
Kingdom was fighting between 1730-1918.

6.1.3 Effects of government spending in Keynesian Model

Consider now the effect of government spending in the Keynesian model where a
fraction λ of consumers is liquidity constrained. As we discussed in section 3.7 the
parameter λ can also be interpreted as the marginal propensity to consume (MPC) out
of current income in the Keynesian aggregate consumption function (16).

If government spending increases and disposable income falls because of higher

lump-sum taxes, consumption demand falls by the MPC times the change in
disposable income λ × Δ (Y1 − T1 ) in the Keynesian model. Aggregate consumption is
predicted to fall by the same amount irrespective of the duration of government
spending increases (temporary or permanent). This stands in contrast to the
neoclassical model that predicts a small shift in consumption demand resulting from
temporary changes in government spending, but a one-for-one shift responding to
permanent changes. The models have therefore also different implications for the
behaviour of the real interest rate after changes in government spending.

6.2 Budget Deficits and Ricardian equivalence

In this section we allow the government to issue debt Bg to finance its expenditures in
the budget constraint (28). Since the government is a large part of the economy one
might expect increases of the interest rate and resulting crowding out as consequence
of more government borrowing.

To focus solely on the effect of government borrowing, consider the following

“Ricardian experiment”:
1. Suppose the government cuts taxes in period 1 ΔT1<0 and raises taxes the following
period to pay for debt and interest (accumulated at rate r): ΔT2 = −(1 + r )ΔT1 ;
2. the path of government spending G1, G2, G3,… is left unchanged;
3. taxes after period 2, T3, T4,… are left unchanged.

What are the effects of this Ricardian experiment in the different models that we
analysed so far? The answer depends crucially on the time horizon of individuals.

28
First, consider individuals with an infinite horizon28 as in the Permanent Income
Hypothesis (PIH) of section 3.5.2. The relevant part of the individuals budget
constraint affected by the Ricardian experiment becomes:

c2 y − ΔT2
c1 + + ... = y1 − ΔT1 + 2 + ...
1+ r 1+ r

Since the tax cut in period 1 is offset by a tax increase in period 2 to pay for the debt
and interest − ΔT1 = ΔT2 /(1 + r ) , the lifetime budget constraint of the individual is not
changed. Alternatively, the present value of future tax increases exactly offsets
current tax cuts. In the PIH model debt issued by the government in period 1 does not
constitute net wealth and leaves individual decision completely unaffected. This
strong neutrality result is called Ricardian equivalence (Barro 1974).

Second, suppose that individuals behave as in the life-cycle hypothesis (LCH) of

section 3.5.1. The important difference to the PIH case above is that some consumers
will die at the end of period 1 and will not be around to pay off the debt in period 2. If
individuals do not care about future generation and do not leave behind any bequests
then the tax cut increases disposable income of those consumers whose horizon is
shorter than the time of future tax increases.

Third, the Keynesian model predicts that liquidity constrained consumers consume all
of the additional disposable income resulting from the tax cut in period 1.
Consumption would increase by the marginal propensity to consume λ and labour
supply would decrease. The empirical relevance of this resukt depends on the
importance of liquidity constraints (see discussion in section 3.7).

Bernheim (1989) criticises the assumption behind the Ricardian equivalence result
that individuals behave as if living in an infinitely lived dynasty. If families are inter-
related by marriage instead this would ultimately imply that all individuals are
mutually related through some distant family link. The surprising conclusion is too
much or super-neutrality (see Bernheim and Bagwell 1988).

The PIH and life-cycle models differ mainly with respect to the empirical importance
of bequests or intergenerational transfers more broadly, for example through parental
investment in education of their children. Kotlikoff and Summers (1981) estimate the
contribution of wealth accumulation due to intergenerational transfers to be sizeable
compared to life-cycle savings. This leaves open the distribution of bequests across
income groups and whether bequests are intentional.

6.3. Taxes
Governments are taxing a variety of economic activities to collect revenue: income,
consumption (VAT), profits, businesses, property, goods etc. Individual income taxes
alone accounted for 42% of federal revenues in the US in 1991 (Barro 1997). In
practice most taxes are distortionary and not lump-sum as we assumed so far.

28
More precisely the horizon over which individuals are planning equals the government’s horizon.

29
With lump-sum as well as distortionary taxes at rate τ, the individual’s budget
constraint (5) becomes:

y 2 (1 − τ ) − T2 c
y1 (1 − τ ) − T1 + ~ = c1 + 2 ~
1+ r 1+ r

where the after-tax interest rate is at rate ~

r = r (1 − τ ) . To focus on the additional
29
effect of the distortion we replace a lump-sum tax by a distortionary income tax.
This will induce a negative substitution effect out of the taxed activity (work) into the
non-taxed one (leisure). Note that this substitution effect comes in addition to a
negative wealth effect from lump-sum taxation.

Figure 11 shows the effect of replacing lump-sum by distortionary income taxes.

First, there are only indirect effects since production is not taxed. Labour supply and
hence aggregate supply shift left by amount (a) because of the negative substitution
effect. Due to the fall in income consumption and aggregate demand shift left by the
same amount (b). The total effect is a reduction in output, labour supply and
consumption. The interest rate stays constant since intertemporal plans are not
affected.

Y1s
(a)

r*

(b)
Y1d
Y1
Y’ Y

Figure 12: Effect of replacing lump-sum by distortionary income tax

29
Combining the results in sections 6.1 to 6.3 we can analyse the effect of a combination of policies.
For example, a temporary increase in government spending in period 1 financed be distortionary taxes
in period 2 can be viewed as (1) ΔG1 = ΔT1 , (2) − ΔT1 = ΔT2 and (3) − ΔT 2 = τY 2 .

30
7 Concluding Remarks
This chapter discusses models of intertemporal models30 of consumption and labour
markets. Microfounded models provide important insights into aggregate economic
behaviour such as aggregate consumption and savings behaviour or labour market
fluctuations. Aggregation problems from microeconomic behaviour are clearly an
issue and should be taken into considerations in empirical work.

The models discussed in this chapter are necessarily very simple and imply strong
restrictions on aggregate data. Several alternatives and generalisations allowing for
more realistic behaviour have been discussed. As usual in economic modelling there
is a trade-off between theoretical simplicity and empirical fit.

30
Due to space constraints this chapter does not discuss investment and its implications for business
cycle behaviour. See for example the chapters on investment in Barro (1997) or Romer (2001).

31
References
Abel, A.B. (1990). “Consumption and Investment.” In B.M. Friedman and F.H. Hahn
(eds.), Handbook of Monetary Economics. Vol. II. North-Holland: Elsevier.

Alogoskoufis, G.S. (1987a). “Aggregate Employment and Intertemporal Substitution

in the UK.” Economic Journal 97 (June): 403-15.

Alogoskoufis, G.S. (1987b). “On Intertemporal Substitution and Aggregate Labor

Supply.” Journal of Political Economy 95 (5), Oct., 938-60.

Angeletos, G.-M., D. Laibson, A. Repetto, J. Tobacman, and S Weinberg (2001).

“The Hyperbolic Consumption Model: Calibration, Simulation, and Empirical
Evaluation.“ Journal of Economic Perspectives 15(3): 47-68.

Attanasio, O.P. (1999). “Consumption.” In J.B. Taylor and M. Woodford (eds)

Handbook of Macroeconomics, Vol. 1. North-Holland: Elsevier.

Attanasio, O.P. and G. Weber (1993). “Consumption Growth, the Interest Rate and
Aggregation.” Review of Economic`Studies 60: 631-49.

Barro, R.J. (1974). “Are Government Bonds Net Wealth?” Journal of Political
Economy 82: 1095-117.

Barro, R.J. (1989). “The Neoclassical Approach to Fiscal Policy.” In R.J. Barro (ed.)
Modern Business Cycle Theory. Cambridge: Harvard Univ. Press.

Barro, R.J. (1997). Macroeconomics. 4th edition. Cambridge: MIT Press.

Beaudry, P. and E. Van Wincoop (1996). “The intertemporal elasticity of

substitutions: an exploration using a US panel of state data.” Economica 63: 495-512.

Bernheim, B.D. (1989). “A Neoclassical Perspective on Budget Deficits.” Journal of

Economic Perspectives 3(2): 55-72.

Bernheim, B.D. and K. Bagwell (1988). “Is Everything Neutral?” Journal of Political
Economy 96: 308-38.

Blanchard, O.J. (1990). “Why Does Money Affect Output? A Survey.” In B.M.
Friedman and F.H. Hahn (eds.), Handbook of Monetary Economics. Vol. 2. North-
Holland: Elsevier.

Blundell, R., M. Browning and C. Meghir (1994). “Consumer Demand and the
Lifetime Allocation of Consumption.” Review of Economic`Studies 61:57-80.

Browning, M and T.F. Crossley (2001). “The Life-Cycle Model of Consumption and
Saving.” Journal of Economic Perspectives 15(3): 3-22.

32
Browning, M, L.P. Hansen and J.J. Heckman (1999). “Micro Data and general
Equilibrium Models.” In J.B. Taylor and M. Woodford (eds.), Handbook of
Macroeconomics. Vol. 1A. North-Holland: Elsevier.

Campbell, J.Y. and A. Deaton (1989). “Why Consumption Is So Smooth.” Review of

Economic Studies 56: 357-74.

Campbell, J.Y. and N.G. Mankiw (1989). “Consumption, Income and Interest Rates:
Reinterpreting the Time Series Evidence.” In: O.J. Blanchard and S. Fischer (eds.),
NBER Macroeconomics Annual 1989: 185-216. Cambridge: MIT Press

Campbell, J.Y. and A. Deaton (1989). “Why is Consumption so Smooth?” Review of

Economic Studies 56: 357-74.

Carroll, C. D. (2001), A Theory of the Consumption Function, With or Without

Liquidity Constraints. Journal of Economic Perspectives 15 (3): 23-45.

Deaton, A. (1992). Understanding Consumption. Oxford: Clarendon Press.

Dixit, A. (1990). Optimization in Economic Theory. 2nd edition. Oxford: Oxford Univ.
Press.

Fisher, I. (1907). The Rate of Interest, New York: Macmillan.

Flavin, M. (1981). “The Adjustment of Consumption to Changing Expectations About

Future Income.” Journal of Political Economy 89: 974-1009.

Friedman, M. (1957). A Theory of the Consumption Function. Princeton Univ. Press.

Hall, R.E. (1978). Stochastic Implications of the Life Cycle-Permanent Income

Hypothesis: Theory and Evidence. Journal of Political Economy 86(6), December,
971-87.

Hall, R.E. (1988). “Intertemporal Substitution in Consumption.” Journal of Political

Economy 96: 339-57.

Hall, R.E. and F.S. Mishkin (1982). “The Sensitivity of Consumption to Transitory
Income: Estimate From Panel Data on Households.” Econometrica 50: 461-81.

Hansen, G.D. and R. Wright (1992). “The Labor Market in Real Business Cycle
Theory.” Federal Reserve Bank of Minneapolis Quarterly Review, Vol. 16 (2),
Spring. Available at http://minneapolisfed.org/research/qr/

Hayashi, F. (1987). “Tests for Liquidity Constraints: A Critical Survey and Some
New Observations.” In: T.F. Bewley (ed.) Advances in Econometrics: Fifth World
Congress, Vol. II: 91-120.

Keynes, J.M. (1936). The General Theory of Employment, Interest, and Money.
London: Macmillan.

33
Kotlikoff, L.J. and L.H Summers (1981). “The Role of Intergenerational Transfers in
Aggregate Capital Accumulation.” Journal of Political Economy 89(4): 706-32.

Loewenstein, G. and R.H. Thaler (1989). “Anomalies: Intertemporal Choice.” Journal

of Economic Perspectives 3(4): 181-93.

Lucas, R.E. Jr. and L. Rapping (1969). “Real Wages, Employment and Inflation.”
Journal of Political Economy 77, September/October, 721-54.

Mankiw, N.G., J.J. Rotemberg, and L.H. Summers (1985), “Intertemporal

Substitution in Macroeconomics.” Quarterly Journal of Economics 100: 225-51.

Modigliani, F. (1986). “Life Cycle, Individual Thrift, and the Wealth of Nations.”
American Economic Review, 76(3), June, 297-313.

Mulligan, C.B. (1995). "The Intertemporal Substitution of Work—What Does the

Evidence Say?" University of Chicago, Population Research Center. Discussion-
Paper 95-11. June.

Mulligan, C.B. (1998). “Pecuniary Incentives to Work in the United States during
World War II.” Journal of Political Economy 106(5): 1033-77.

Obstfeld, M. and K. Rogoff (1996). Foundations of International Macroeconomics.

MIT Press.

Pencavel, J. (1986). “Labor Supply of Men: A Survey.” In: by: O.C. Ashenfelter and
R. Layard (eds.) Handbook of Labor Economics, Vol 1. North-Holland: Elsevier.

Prescott-Summers Debate (1986). Federal Reserve Bank of Minneapolis Quarterly

Review, Vol. 10(4), Fall. Available at http://minneapolisfed.org/research/qr/

Ramsey, F. (1928). “A Mathematical Theory of Saving.” Economic Journal 38: 543-

59.

Romer, D. (2006). Advanced Macroeconomics. 3rd edition. McGraw-Hill.

34