Macroeconomic Analysis

Lecture notes (2) on:

Rational expectations and
the "New Classical Macroeconomics"

(F. Bagliano, 2017)

The rationalization of the different nature of the short- and long-run relation
between inflation and unemployment put forward by Friedman and Phelps, cen-
tered around the concept of the natural rate of unemployment, features a key role
for agents’ expectations. However, the process of expectations formation that is
(implicitly) assumed is essentially "adaptive": people (especially workers) form
expectations about the price level (and therefore the real wage) on the basis of
past experience, gradually "correcting" over time any past forecast error. The
immediate consequence of this behavior is the possibility of long periods in which
systematic mistakes occur, with repeated under- (or over-)estimates of the infla-
tion rate.
Since the 1970s a different hypothesis on expectation formation is introduced
into macroeconomic models. The rational expectations hypothesis seems more sat-
isfactory (because it rules out systematic mistakes) and consistent with the degree
of rationality attributed to economic agents in other decision-making processes,
such as profit and utility maximization. The adoption of the rational expectations
hypothesis in models that show a "natural" rate of output has had important im-
plications on business cycle thoery and macroeconomic policy.
These notes start with a comparison between the rational expectations and
the adaptive expectations hypotheses; then, some formal business cycle models
incorporating the rational expectations assumption are illustrated. Such models
are typical of the so-called new classical macroeconomics school usually associated
to the names of R. Lucas, T. Sargent, R. Barro and N. Wallace.
1. Adaptive expectations vs. rational expectations: a com-
parison
The fact that economic agents (consumers, firms, workers, financial investors) base
their current choices on what they expect on the future course of some economic
variables is a key ingredient of many economic theories (concerning, for example,
consumption, investment, labor supply, portfolio allocation). The mechanism
whereby expectations are formulated has been specified in various ways.
A first, simple, hypothesis is that of adaptive expectations, implicitly adopted
in the Friedman-Phelps account of the absence of a long-run inflation-unemployment
trade-off. This assumption is formalized by the following equation

pet, t+1 = pet−1, t + λ (pt − pet−1, t ) 0<λ<1 (1.1)

where pet, t+1 is the expectation of pt+1 (the log of the price level) formulated at
time t. The value of p expected for the next period is modified between t and t + 1
by a fraction λ of the forecast error occurred at t, pt − pet−1, t . Such mechanism
generates the possibility of "systematic" forecast errors.
The alternative assumption of rational expectations has been adopted to
avoid the possibility that agents make systematic mistakes without altering the
expectation formation mechanism in (1.1). Under rational expectations

pet, t+1 = E(pt+1 | Ωt ) ≡ Et pt+1 (1.2)

where Et denotes the mathematical expectation of pt+1 conditional upon the in-
formation set available to agents at time t, Ωt . The key feature of rational expec-
tations is that the forecast error (or "surprise") at time t + 1 has a zero expected
value conditional on t:
Et (pt+1 − Et pt+1 ) = 0
In the models that adopt the rational expectations hypothesis, the precise defi-
nition of the content of the information set Ωt plays a crucial role. Usually, Ωt
includes past and current (time t) values of all variables and the structure of
the economy, as described by the equations of the model. Rationally formed ex-
pectations are therefore consistent with the structure of the model describing the
economy, "as if" agents made use of the model itself in forming expectations on
future values of the endogenous variables.
To clarify this point, highlighting the consequences on the equilibrium dy-
namics of macroeconomic variables under either mechanism, we use a simple,

2
discrete-time, version of the classic hyperinflation model due to P. Cagan (1956),
assuming adaptive expectations, and used by T. Sargent and N. Wallace (1973)
with rational expectations.

1.1. A simple model

The model considers an economy in which in each period a consumption good can
be exchanged for money. In the presence of only two markets (for the only good
and for money) only one market-clearing condition is necessary to describe the
simultaneous equilibrium between demand and supply on both markets. Focusing
on the money market, the equilibrium condition equates the real money supply
to the real money demand as follows:
m − p = −α (pet, t+1 − pt ) α>0 (1.3)
| t{z }t | {z }
money supply money demand

In (1.3) all variables are in logarithms: m is the quantity of nominal money in

the economy, exogenously set by a monetary authority (central bank), and p is
the price level of the consumption good; pet, t+1 − pt is therefore the inflation rate
expected between t and t + 1. Since the model is focused on the dynamics of
the price level, the amount of the consumption good produced in each period is
assumed to be constant (and not affected by monetary variables), and therefore its
level is omitted from money demand. No stochastic elements (random "shocks")
are added to the money demand and money supply functions.1
1
Equation (1.3) may be interpreted more generally as a traditional LM curve, where an
exogenous money supply is equated in equilibrium to a money demand which depends positively
on output and negatively on the nominal interest rate. Letting upper-case letters denote the
levels of the corresponding variables (whereas lower-case letters denote logarithms, with the only
exception of the interest rate) we can write the equilibrium condition at time t as:
Mt e
= Ȳ e−α(r̄+πt,t+1 ) (*)
Pt
where output is assumed to be constant at level Ȳ and the nominal interest rate i = r̄+
π et,t+1 is the sum of the real interest rate r̄ (assumed constant) and expected inflation. The
specific formulation of the money demand function is chosen for analytical tractability. Taking
logarithms of both sides of (*) yields:
¡ ¢
mt − pt = log Ȳ − α r̄ − α π et,t+1

where the term in brackets is a constant and can be omitted from the equation without altering
the economic interpretation of the results. Finally, approximating πet,t+1 with pet, t+1 − pt we get

3
 Rearranging (1.3), the equilibrium price level at time t may be expressed as a
function of current money supply and the price level expected for the next period:
1 α e
pt = mt + p (1.4)
1+α 1 + α t, t+1
In order to solve the model, an assumption on the mechanism that agents use to
form expectations on the future price level is required.

1.2. Adaptive expectations

The adaptive expectations hypothesis (1.1) can be rewritten as:

pet, t+1 = λ pt + (1 − λ) pet−1, t (1.5)

By repeated substitution of past values of the expected price level pe obtained by

lagging (1.5) we get:

pet, t+1 = λ pt + λ(1 − λ) pt−1 + λ (1 − λ)2 pt−2 + ... + (1.6)

T −1 T
+λ (1 − λ) pt−T +1 + (1 − λ) pet−T, t−T +1

Since, for T → ∞, (1 − λ)T → 0, (assuming also a finite value for pet−T, t−T +1 ) the
last term in (1.6) can be neglected, getting:

X
∞
pet, t+1 = λ (1 − λ)i pt−i (1.7)
i=0

Finally, substituting (1.7) into (1.4) and solving for per pt yields the solution for
the equilibrium price level at time t:

1 αλ X ∞
pt = mt + (1 − λ)i pt−i (1.8)
1 + α(1 − λ) 1 + α(1 − λ) i=1

The current price level depends only on its past history and on the current value
of the money stock mt . As a consequence, a transitory change in mt has the same
influence on pt as a permanent change of the same amount (as will be shown in
detail below).
the equilibrium condition (1.3) in the text.

4
1.3. Rational expectations
If the rational expectations hypothesis is introduced in the model as in (1.2), the
resulting price level is:
1 α
pt = mt + Et pt+1 (1.9)
1+α 1+α
It is now necessary to solve the model to get the (model-consistent) expectation
for pt+1 . Leading (1.9) forward by one period and taking the expected value at
time t we get:
1 α
Et pt+1 = Et mt+1 + Et pt+2 (1.10)
1+α 1+α
Note that : (i) the information set Ωt on which the expected value Et is based
includes the current-period values of m and p and the structure of the economy, as
summarized by (1.3); (ii) the law of iterated expectations (whereby Et (Et+1 pt+2 ) =
Et pt+2 ) has been applied to the last term of Et pt+1 . Substituting (1.10) into (1.9)
yields: µ ¶
1 α 1 α
pt = mt + Et mt+1 + Et pt+2
1+α 1+α 1+α 1+α
and repeating the same steps for Et pt+i (i = 2, 3, ..., T − 1) we get:
T −1 µ ¶i µ ¶T
1 X α α
pt = Et mt+i + Et pt+T (1.11)
1 + α i=0 1 + α 1+α
¡ α ¢T
Since, if T → ∞, 1+α → 0, the last term in (1.11) can be neglected (as-
suming that Et pt+T is finite). Then, the equilibrium price level under rational
expectations is:
∞ µ ¶i
1 X α
pt = Et mt+i (1.12)
1 + α i=0 1 + α
No variable dated t − 1 or earlier affects pt in equilibrium: the current price level
is entirely determined by the current value of money supply that agents observe
(Et mt = mt ) and by future values of mt+i (i = 1, ...) expected as of time t. Any
piece of information available to agents at t and useful to forecast the future path
of mt+i has an immediate effect on the price level. Therefore, the solution of the
model with rational expectations is forward-looking, whereas the solution with
adaptive expectations in (1.8), depending only on past and currently observed
variables, is backward-looking.

5
1.4. Examples
Using alternatively (1.8) or (1.12) we can now study the behavior of the price
level in response to changes in the quantity of money m. In particular, we will
examine the following cases: (i) a permanent increase of m at time t1 from m̄ to
m̄ + k; (ii) a temporary increase of m by the same amount from t1 up to t2 ; (iii)
a temporary increase of m (as in (ii)) but previously announced at time t0 (and
therefore expected by agents).
Permanent increase of m (Figure 1). Sarting from an initial situation in which
the quantity of money has always been equal to m̄, a permanent increase from m̄
to m̄ + k occurs at time t1 . With adaptive expectations, up to period t1 − 1 the
price level is m̄; at t1 , following the current-period increase in m, from (1.8) the
price level increases to:
1
pt1 = k + m̄ (1.13)
1 + α(1 − λ)
After t1 a gradual increase in the price level occurs, according to (1.8) with mt =
m̄ + k for t > t1 , and p asymptotically reaches the new stationary equilibrium
level m̄ + k. On the contrary, with rational expectations, according to (1.12), p
immediately adjusts to its new equilibrium level m̄ + k.

Figure 1: Permanent increase of m

6
Temporary increase of m (Figure 2). In this case money is m̄ up to t1 −1, then
increases to m̄+k only from t1 to t2 −1, to finally go back (forever) to the previous
level m̄ at t2 . The temporary nature of the increase of money becomes known to
agents in t1 . Figure 2(a) describes the reaction of p with adaptive expectations. At
t1 the price increase is again given by (1.13): indeed, since agents do not take into
account the future path of m in forming expectations on p, there is no difference
between the response of the price level to a transitory and a permanent money
increase. At t2 , when money decreases to the initial level m̄, the price level reacts
in the opposite way and decreases according to (1.8); after t2 expectations are
gradually adjusted downwards and the price level gradually decreases towards the
initial equilibrium level m̄.
Instead, with rational expectations, the reaction of p at t1 depicted in Figure
2(b) is different from the case of a permanent increase. The increase of p at t1
can be computed using (1.12) evaluated at t1 :
" µ ¶t2 −t1 #
α
pt1 = 1 − k + m̄ (1.14)
1+α

Figure 2: Temporary increase of m. (a) adaptive expect.; (b) rational expect.

Afterwards, between t1 and t2 , the price level gradually decreases up to the initial

7
equilibrium value m̄, that is reached exactly at time t2 . Note that under ratio-
nal expectations the price level "jumps" (displaying discrete changes) only when
unexpected events occur. This is the case of the increase of m at t1 ; on the con-
trary, the decrease of m at t2 has already been incorporated (at t1 ) in the agents’
expectations and does not cause any "jump" in the price level.
Expected temporary increase of m (Figure 3). The temporary increase in
money supply is the same as in the previous case, but now it is announced to agents
in advance, at time t0 . Therefore, when it occurs, the temporary money expansion
is perfectly expected on the basis of the previous (credible) announcement of the
monetary authorities. With adaptive expectations (Fig. 3(a)), the behavior of the
price level is the same as in the previous case: the announcement at t0 of future
events does not change agents’ expectations and therefore has no effect on the
equilibrium price path.
Instead, with rational expectations (Fig. 3(b)), agents formulate their expec-
tations in t0 taking into account the announcement of a future temporary increase
of m. Using (1.12) with m = m̄ + k between t1 and t2 , we obtain the equilibrium
price level at the announcement date t0 :
µ ¶t1 −t0 " µ ¶t2 −t1 #
α α
pt0 = 1− k + m̄ (1.15)
1+α 1+α

Figure 3: Expected temporary increase of m. (a) adaptive exp.; (b) rational exp.
Note: (i) the only "jump" of p occurs at the announcement date t0 (indeed,

8
only the announcement comes as a "surprise" to agents: everything that happens
after the announcement is perfectly anticipated at t0 ); at t1 , when the increase
of m does take place, p is exactly at the level computed in the previous case
(without announcement) and given by (1.14); (ii) the equilibriun response of the
price level implies a period of inflation between t0 and t1 even in the absence of
a monetary expansion: inflation is entirely attributable to the fact that agents
expect money to increase in the future; then, in the very period in which money
is at the higher level (between t1 and t2 ) “deflation” occurs (with the price level
gradually decreasing).

1.5. Extension to stochastic shocks under rational expectations

So far, no stochastic element has been introduced in the model. In this case, if the
process generating money supply is not subject to random shocks, adopting the
rational expectations hypothesis reduces to the (stronger) assumption of perfect
foresight.
The model can be easily extended to allow for stochastic elements both in the
money supply process and on the demand for money side. In the latter case a
shock ut can be added to expected inflation as a determinant of money demand in
each period t, yielding the following equilibrium condition on the money market:
mt − pt = −α (pet, t+1 − pt ) + ut (1.16)
where ut is a money demand shock with the property Et−1 ut = 0 (i.e. unfore-
castable one period earlier).
The model is then solved following the same steps as in the non-stochastic
case. Given the price level at t + 1
1 α 1
pt+1 = mt+1 + Et+1 pt+2 − ut+1 (1.17)
1+α 1+α 1+α
the expectations of pt+1 formed at time t is
1 α
Et pt+1 = Et mt+1 + Et pt+2 (1.18)
1+α 1+α
which coincides with (1.10) since Et ut+1 = 0. Repeating this step for pt+2 , pt+3 ,
... and letting T go to infinity, the final expression for the price level is:
∞ µ ¶i
1 X α 1
pt = Et mt+i − ut (1.19)
1 + α i=0 1 + α 1+α

9
The price level is now affected not only by the expected values of future money
supply, but also by the realization of the money demand shock ut . In each period
t, the realized price level pt will differ from the expected price level Et−1 pt because
of the presence of the disturbance ut , which is unpredictable at t − 1, and because
the process that generates money supply can have a stochastic (unpredictable)
component. Therefore, the forecast error at time t is:
∞ µ ¶i
1 X α 1
pt − Et−1 pt = (Et mt+i − Et−1 mt+i ) − ut (1.20)
1 + α i=0 1 + α 1+α

The price forecast error at time t depends on the current realization of the shock
ut , and on the revisions in expectations on the future path of money supply
occurred between t − 1 and t: Et mt+i − Et−1 mt+i . Such revisions will occur any
time new information (relevant to predict future money) accrues to agents at t,
enlarging the information set on which their (rational) expectations are based,
and leading agents to change their expected values for future money supplies. An
immediate implication of rational expectations is that such forecast error on pt
is unpredictable on the basis of the information set available to agent at t − 1;
taking expectations of (1.20) at t − 1 we get:

Et−1 [pt − Et−1 pt | Ωt−1 ] = 0 (1.21)

References
[1] Cagan P. (1956) "The monetary dynamics of hyperinflation", in M. Friedman
(ed.) Studies in the quantity theory of money, Chicago, University of Chicago
Press

[2] Lucas R.J. (1973) "Some international evidence on the output-inflation trade-
off", American Economic Review, 63

[3] Sargent T.J. and N. Wallace (1973) "The stability of models with money and
growth with perfect forecast", Econometrica, 41, 6

[4] Sargent T.J. and N. Wallace (1976) "Rational expectations and the role of
monetary policy", Journal of Monetary Economics, 2

10
2. Imperfect information and the business cycle: Lucas’ "is-
lands" model (1973)
The rational expectations hypothesis, ruling out agents’ systematic forecast errors,
has been considered more satisfactory from a theoretical viewpoint and has been
adopted by several macroeconomic models aimed at explaining the basic mech-
anisms originating cyclical fluctuations, and particularly the positive correlation
between nominal variables (such as monetary aggregates) and real quantities (out-
put, employment). One of the first business cycle models making use of rational
expectations and attributing the correlation between nominal and real variables
to informational imperfections (of a different variety from those assumed by Fried-
man) is due to R. Lucas (1972, 1973).
The ideal economy described by Lucas is composed of a large number of ge-
ographically dispersed markets, where producers decide the amount of output in
each period on the basis of the comparison between the price of the good in the
local market in which they operate and the general price level in the whole econ-
omy.2 Only an increase of the local price relative to to the economy-wide price
level induces producers to increase output.
In this economy there are two sources of stochastic disturbances, both from
the demand side: (a) an aggregate demand shock, hitting all markets in the same
way; (b) a local demand shock, hitting any individual market. Only aggregate
disturbances affect the economy-wide price level, whereas local prices are affected
by both aggregate and local demand shocks. If producers in each market could
always observe both their local price and the general price level, they would react
only to differences between the two (a "relative" price), that would be a clear
signal that a purely local demand shock has occurred. Instead, the Lucas model
is built on the crucial hypothesis that producers have imperfect information on
price changes: they observe the price on their local market but do not observe
the general, economy-wide, price level. Therefore, output decisions depend on
the difference between the locally observed price and the general price level that
producers (rationally) expect on the basis of their imperfect information. This
insight can be formalized in a simple model (following Lucas 1973).
2
The view of the economy as made up of many separated markets justifies the denomination
of Lucas "islands" model often used is the literature.

11
2.1. The model
The economy is composed of N separated markets, indexed by z = 1, 2, ...N, for
a homogeneous good, operating under conditions of perfect competition. In each
market, a perfectly flexible local price pt (z) ensures equilibrium between demand
and supply in each period t (i.e. there is continuous market-clearing).3 The
amount of output on each market z, yt (z), is given by the following local supply
function
yt (z) = y ∗ + γ (pt (z) − E(pt | It (z))) γ>0 (2.1)
where y ∗ is the "natural" level of output, determined by long-run real driving
forces such as capital accumulation and population growth (here, for simplicity, y ∗
is assumed constant over time and equal across all markets), and γ is a "structural"
parameter, reflecting the properties of technology and workers’ preferences on the
local market (again, those structural factors are assumed to be equal across all
markets). Producers increase output beyond y ∗ if the price observed locally is
higher than the general price level pt expected by local producers on the basis of
their available information. Such information set, denoted by It (z), includes the
observed level of the local price pt (z) and the characteristics of the probability
distributions of pt e pt (z). More specifically, let us assume that

pt = p̄t + vt with vt ∼ N(0, σ2 ) (2.2)

pt (z) = pt + zt with zt ∼ N(0, τ 2 )
= p̄t + vt + zt (2.3)

The general price level pt is normally distributed around a mean p̄t (perfectly
known by local producers); vt is an aggregate demand shock, with zero mean and
variance σ 2 (again, known by producers). The local price pt (z) is the sum of pt
and the local demand disturbance zt (normally distributed with zero mean and
2
variance
PNτ ), with the property that the sum across markets of all local shocks is
zero, 1 zt = 0, and therefore does not affect the general price level; moreover,
there is no correlation between vt and zt . The distribution of pt (z) is then normal
around p̄t and has a variance of σ 2 + τ 2 .
To choose the optimal output level, local producers make use of all available
information on the distribution of pt in (2.2) and the observation of their local
price pt (z) to form an expectation of the general price level. This expectation is
3
In the model, lower-case letters denote logarithms of the corresponding variable: e.g. pt (z) ≡
log Pt (z).

12
rational since it is obtained as the (mathematical) expected value of pt conditional
on pt (z) and on the properties of the relevant probability distributions: we denote
this expectation as E(pt | pt (z)). Given the properties in (2.2) and (2.3) the two
variables are jointly normally distributed as follows:
µ ¶ ∙µ ¶ µ 2 ¶¸
pt p̄t σ σ2
∼N , (2.4)
pt (z) p̄t σ2 σ2 + τ 2

where the matrix collects the variances and the covariance between pt e pt (z).
Given (2.4) we can derive the conditional expected value as :4

σ2
E(pt | pt (z)) = p̄t + 2 (pt (z) − p̄t )
σ + τ2
τ2 σ2
= p̄t + pt (z)
σ2 + τ 2 σ2 + τ 2

≡ θ p̄t + (1 − θ) pt (z) (2.5)

2
where θ ≡ σ2τ+τ 2 , with 0 < θ < 1. To infer the general price level pt producers
use optimally the information contained in the local price. The expected value
of pt is then obtained as a weighted average of the local price pt (z), and the
unconditional mean of the general price level, p̄t . The weight on local price is
2
given by 1 − θ ≡ σ2σ+τ 2 and depends negatively on the variance of the local shock
τ 2 and positively on the variance of the aggregate disturbance σ 2 . Intuitively, if
the aggregate shock has a high variance relative to the local disturbance, the local
price pt (z) contains information which is very useful in forming expectations on pt
and will carry a relatively large weight in (2.5). The opposite case occurs when τ 2
is relatively high: movements in the local price mainly reflect local disturbances
4
To obtain (2.5) we made use of the following property of the joint normal distributionof two
generic random variables x and y:
µ ¶ ∙µ ¶ µ 2 ¶¸
x x̄ σ x σ xy
∼N ,
y ȳ σ xy σ 2y

The distribution of x conditional on y is normal with (conditional) mean and variance given by:
" #
σ xy 2 (σxy )2
x | y ∼ N x̄ + 2 (y − ȳ) , σ x −
σy σ 2y

13
and pt (z) will be given a lower weight in determining pt . Here producers face a
classic problem of signal extraction: they have to "extract" an estimate of the
"signal" they are interested in (in our case pt ) from the observation of the local
price pt (z) which contains also an element of "noise" (the local disturbance zt ).
Substituting (2.5) into the local supply function (2.1) we get:

yt (z) = y ∗ + γ θ (pt (z) − p̄t ) (2.6)

which relates the fluctuations of local output around the natural value to devi-
ations of the local price from the average general price level p̄t . The "slope" of
this supply curve is given by γ θ and therefore depends not only on the structural
parameter γ but also, through θ, on the relative variances of the local and aggre-
gate demand shocks. The relative magnitude of the variability of v relative to the
variability of z determines the effect on output of movements in local prices. This
feature of the model would not arise in the case of perfect information, in which
local producers can directly observe, beside pt (z), also the general price level pt .
To get a measure of the aggregate supply in the whole economy we can average
over Pthe N markets the local output given by (2.6), getting (using the property
that N 1 zt = 0, and (2.2) to set vt = pt − p̄t ):
PN PN
yt (z) (pt (z) − p̄t )
1
= y∗ + γ θ 1
N N
PN PN
∗ 1 vt + 1 zt
= y +γθ
N
= y ∗ + γ θ vt

⇒ yt = y ∗ + γ θ (pt − p̄t ) (2.7)

Equation (2.7), known as the "Lucas supply curve", shows that fluctuations of
output around the natural level depend positively on unanticipated movements in
the aggregate price level pt . As for the local supply functions, also the "slope" of
the aggregate output supply (2.7) depends not only on the structural parameter γ
but also on θ and therefore on the relative variance of demand shocks. The larger
the variance of the aggregate disturbance (relative to the local shock), the more
"vertical" the supply function in a traditional (p, y) plane, since movements in
local prices are mainly attributed to changes in the general price level pt , having
only a limited effect on output.

14
 In modelling the supply side of the economy we assumed that producers base
their expectations on the true probability distribution of the price level pt . To
check the "rationality" of the expectations formed according to (2.5) we have to
get the general price level in equilibrium and derive its distribution. To this aim,
we add to the model an (extremely simple) aggregate demand function
yt = mt − pt (2.8)
capturing the essential property of aggregate demand, that is a negative relation-
ship between output and the price level. mt can be interpreted as the nominal
quantity of money supplied by monetary authorities or,more generally, as a mea-
sure of nominal aggregate demand.5 Equating aggregate supply and demand,
given by (2.7) and (2.8) we get the equilibrium price level:
1 1 γθ
pt = mt − y∗ + p̄t (2.9)
1+γθ 1+γθ 1+γθ
The expected value of the general price level, p̄t ≡ E(pt ), can be obtained by
taking the expected value E(.) of both sides of (2.9) and solving for p̄t :6
p̄t = E(mt ) − y ∗ (2.10)
Thus, the expected price level depends on the anticipated component of money
(or nominal demand), and by the (known) level of natural output. Substituting
(2.10) into (2.9) and rearranging, we get the final form of the equilibrium price
level:
1
pt = E(mt ) − y ∗ + (mt − E(mt )) (2.11)
| {z } 1+γθ
p̄t | {z }
vt

Equation (2.11) gives an economic content to the deviations of pt from its mean
value p̄t : they are determined by the unexpected component of money (or nominal
demand), mt − E(mt ). Finally, using (2.10) and (2.9) into aggregate supply (2.7)
we obtain the equilibrium output:
γθ
yt = y ∗ + (mt − E(mt )) (2.12)
1+γθ
5
Rewriting (2.8) as yt + pt = mt highlights the interpretation of mt as an indicator of the
nominal level of aggregate demand, since yt + pt (≡ log(Yt Pt )) is output in nominal terms. This
interpretation is adopted by Lucas (1973).
6
Note that p̄t is the unconditional expected price level, obtained on the basis of available
aggregate information only, and not conditioned on the observation of a local price level.

15
Only the unexpected component of aggregate demand affects output, whereas
the perfectly anticipated component E(mt ) affects only the price level pt . The
economic mechanism whereby purely nominal disturbances (the unexpected com-
ponent of mt ) have real effects is based on the producers’ imperfect information
about the nature of the shocks hitting the local price level. With imperfect infor-
mation, an unexpected increase of nominal demand is partly attributed to a local
disturbance, making the local price change relative to the general price level. The
optimal response of all producers is then to increase supply, which generates an
aggregate output increase. This response is stronger the higher is the (relative)
variability of local shocks (given by the ratio τ 2 /σ 2 ) and therefore the larger is θ.

2.2. Implications
Lucas model has relevant implications for policy and econometric practice.
The output-inflation trade-off. The model provides microeconomic founda-
tions to Friedman’s insight that informational imperfections may generate a pos-
itive correlation between movements of nominal (money or, more generally, nom-
inal aggregate demand) and real variables (employment, output).
Differently from Friedman’s model, the Lucas model is not based on an adap-
tive expectations mechanism, but assumes that agents form expectations ratio-
nally using all available information. Again, a Phillips curve arises, linking un-
foreseen changes in nominal demand to deviations of output from its natural level.
The adoption of the rational expectations hypothesis leads to the substitution of
the distinction between a short-run and a long.run Phillips curve (as in Fried-
man’s model) with that between a (positively sloped) Phillips curve which applies
to unexpected changes in demand, and a vertical Phillips curve which applies to
anticipated demand changes: the latter curve does not show any trade-off between
output and inflation.
The “Lucas critique”. The derivation of the supply curve and the equilib-
rium levels of output and prices given above showed that the relationship between
changes in nominal and real variables depends on both "structural" and "non
structural" parameters. The structural parameters capture features of the econ-
omy, such as technology and agents’ preferences, that do not vary when demand-
management policies, affecting the process generating aggregate demand, change;
those parameters are then policy-invariant, as γ in the model. Instead, non struc-
tural parameters change when the properties of aggregate demand and economic
policies vary. This is the case of parameter θ in the model, which (together with

16
γ) determines the effect of unanticipated fluctuations in aggregate demand on
output (i.e. the "slope" of supply function). θ depends on the relative variance
of local and aggregate shocks, on which monetary policy can have a direct influ-
ence. In fact, a policy characterized by large unpredictable fluctuations in money
supply determines a relatively large variance of aggregate demand disturbances
σ 2 , thereby affecting the ratio θ. In this case, the price-output relation is not a
policy-invariant feature of the economy but depends also on (in particular mone-
tary) policymakers’ decisions. This basic insight has been put forward by Lucas
(1976) and has become known as the Lucas critique to traditional economic (and
econometric) models, where behavioral relationships were implicitly considered as
invariant to changes in economic policies.
Lucas (1973) provides a simple empirical application which is inspired by his
own critique. From (2.12), the coefficient linking equilibrium output to unexpected
aggregate demand can be expressed (writing θ in terms of the variances σ 2 and
τ 2 ) as:
γθ γ τ2
= 2 (2.13)
1+γθ σ + (1 + γ) τ 2
For given τ 2 and γ this coefficient decreases in σ 2 . The basic idea of the test
is to compare economies characterized by a different variability of aggregate de-
mand (σ 2 ), and see whether there is a systematic negative correlation between
the estimates of σ 2 and the estimates of the "slope" of the supply function (2.12).
Overall, the empirical evidence presented by Lucas supports this implication of
the model.
The key point of the Lucas critique has a general validity, and, in models with
rational expectations, derives immediately from the fact that agents, in form-
ing expectations on the relevant variables, use also what they know about how
economic policies are conducted. It follows that parameters that describe such
policies play a role in determining agents’ behavior. From the perspective of
econometric practice, policy changes affect the parameters of equations describing
agents’ behavior: therefore, values for those parameters estimated on past data
are inappropriate to simulate the effects of new policy measures.

17
3. New Classical Macroeconomics: the policy ineffective-
ness proposition
The fundamental insights of Lucas (1973) model are at the heart of the so-
called "New Classical Macroeconomics" (NCM). This school of thought (associ-
ated mainly to the names of Lucas, Sargent, Wallace and Barro) made an effort to
offer theoretical underpinnings to macroeconomic fluctuations, particularly to the
comovements between output and inflation, entirely grounded on the assumption
of market-clearing on all markets, in open contrast with traditional macroeco-
nomic models of keynesian inspiration, based on price and wage rigidities that
slow down or completely hinder the equilibrium adjustment of demand and sup-
ply. Combining the market-clearing assumption with the rational expectations
hypothesis the NCM derives sharp implications on the ineffectiveness of economic
policies based on aggregate demand management (known as the policy ineffective-
ness proposition).
In general, the essential elements of NCM models are:

• the existence of a natural rate of output (and therefore unemployment) de-

termined exclusively by real forces;

• continuous and instantaneous market-clearing on all markets due to per-

fectly flexible prices;

• rational expectations.

3.1. A typical NCM model

What follows is a simplified version of the model by Sargent and Wallace (1976),
which formally derives the ineffectiveness proposition. Moreover, it provides a
clear application of a solution technique for models with rational expectations
that will be used again later on in the course.
The demand side of the model is based on the elementary macroeconomic
IS − LM structure and on the three assumptions mentioned above. The IS and
LM curves have a traditional form (in the following equations all variables, with
the exception of the nominal interest rate, are in logarithms):

yt = −b (it − Et−1 (pt+1 − pt )) + v1t (3.1)

mt − pt = yt − d it + v2t (3.2)

18
where y, m and p are aggregate output, the quantity of money, and the price level,
respectively. For simplicity, the natural level of output is assumed constant over
time and normalized to zero (in logs): y can then be interpreted as a deviation
from the natural level. Finally, i is the nominal interest rate and v1 and v2 are
two shocks with the property: Et−1 v1t = Et−1 v2t = 0.
Equation (3.1) describes a simple negative relation between demand for goods
and the real interest rate (expressed as the nominal interest rate at time t less the
inflation rate between t and t + 1 rationally expected on the basis of the available
information at the end of period t−1). Equation (3.2) is the equilibrium condition
on the money market, imposing equality between real money supply and real
money demand (the latter depending positively on output y -with unit elasticity-
and negatively on the nominal interest rate i). v1 and v2 have the interpretation
of shocks to the demand for goods and to the demand for money respectively.
Combining the IS (3.1) and the LM (3.2) functions to eliminate the nominal rate
it , we get thw following aggregate demand function (AD):

yt = α (mt − pt ) + β Et−1 (pt+1 − pt ) + vt (3.3)

b bd
¡ 1 ¢
where α ≡ b+d , β ≡ b+d , and vt ≡ b+d (d v1t − b v2t ) is a composite disturbance
to aggregate demand.
The supply side of the economy is described by a Lucas-type aggregate supply
function, with deviations of output from its natural level due only to the un-
expected component of the price level and to a supply disturbance u (reflecting
shocks to technology and preferences), with the property Et−1 ut = 0:

yt = γ (pt − Et−1 pt ) + ut (3.4)

Monetary policy is conducted by setting the nominal money supply mt according

to a feedback rule of the following form:

mt = mt−1 − δyt−1 + εt (3.5)

Money reacts systematically to past deviations of output from its natural level in
a "countercyclical" fashion (given the negative coefficient −δ on yt−1 ); εt denotes
the non-systematic, unpredictable (Et−1 εt = 0) component of money supply.
To solve the model made up of (3.3), (3.4) and (3.5) we employ one of the
standard solution techniques for models with rational expectations, namely the
method of undetermined coefficients.

19
 By equating aggregate demand and supply and using the monetary rule (3.5),
we obtain a first expression for the equilibrium price level:
µ ¶
1
pt = [(γ − β) Et−1 pt + β Et−1 pt+1 + α mt−1 − δα yt−1 + α εt + vt − ut ]
α+γ
(3.6)
Here the price level is a function of past variables (mt−1 and yt−1 ), current dis-
turbances (εt , vt and ut ) and expectations of the price level itself at time t and
t + 1 (formed on the basis of available information at t − 1). The adopted solution
technique assumes a (linear) solution for the price level pt with coefficients to be
determined in order to satisfy equation (3.6).
Given the set of variables which appear on the right-hand side of (3.6), we can
guess a solution for the price level of the following form:

pt = π 1 mt−1 + π 2 yt−1 + π 3 εt + π 4 vt + π 5 ut (3.7)

where π 1 , ... π5 are coefficients to be determined. From (3.7) we can derive the
expected values Et−1 pt e Et−1 pt+1 (using the fact that εt , vt and ut are shock
unpredictable one period earlier):

Et−1 pt = π1 mt−1 + π 2 yt−1 (3.8)

Et−1 pt+1 = π1 Et−1 mt + π2 Et−1 yt
= π1 mt−1 − π 1 δ yt−1 (3.9)

where we used (3.4) and (3.5) to express the expected values at t − 1 of mt and yt .
It is now possible to equate our guess solution (3.7) to (3.6), making use of (3.8)
e (3.9) to substitute for the terms involving the expected price level. We get:

γ−β
π 1 mt−1 + π2 yt−1 + π 3 εt + π 4 vt + π5 ut = (π 1 mt−1 + π 2 yt−1 )
α+γ
β
+ (π 1 mt−1 − π 1 δyt−1 )
α+γ
1
+ (αmt−1 − αδyt−1 + αεt + vt − ut )
α+γ
(3.10)

In the last step of the solution procedure we have to find the values of the (as
yet undetermined) coefficients π i that satisfy (3.10). To this aim, we solve the

20
system of equations obtained by equating coefficients on the same variables on
the left-hand and right-hand sides of (3.10). We then get:
γ−β β α
mt−1 : π1 = π1 + π1 +
α+γ α+γ α+γ

⇒ π1 = 1

γ−β βδ αδ
yt−1 : π2 = π2 − π1 −
α+γ α+γ α+γ

⇒ π 2 = −δ

α
εt : π3 =
α+γ
1
vt : π4 =
α+γ
1
ut : π5 = −
α+γ
Solving this system we obtain the solution for the price level pt :
1
pt = mt−1 − δyt−1 + (α εt + vt − ut ) (3.11)
α+γ
From (3.11) we derive the price “surprise” which, according to (3.4), causes output
to deviate from its natural level:
1
pt − Et−1 pt = (α εt + vt − ut ) (3.12)
α+γ
Finally, substituting (3.12) into the AS function (3.4) we get the final form for
output at time t:
γ
yt = (α εt + vt − ut ) + ut (3.13)
α+γ
Equilibrium output deviates from the natural rate only because of unanticipated
changes in money suply (εt ) and of shocks to aggregate demand and supply (vt
and ut ). The specific monetary policy rule adopted by the policymaker, captured
by the parameter δ (measuring the degree of policy countercyclicality), has no role

21
in determining output. It is then impossible to use systematic monetary policy
for output stabilization purposes.
Note that, if the assumption that shocks follow white noise (totally unpre-
dictable) stochastic processes is relaxed, then only the unanticipated part of εt
and vt would affect output; moreover, as for the supply shock ut , beside the effect
of the unanticipated component acting through the price surprise, as shown in
(3.12), also the anticipated part of the shockwould affect output, since the whole
of the disturbance ut shifts the aggregate supply curve (3.4).
The most widely known proposition associated to the NCM derives directly
from (3.13): systematic monetary policy cannot be used to stabilize output. Any
monetary feedback rule is predictable by agents who base their decisions about la-
bor and output supply on rational expectations that take into account the features
of the rule, and therefore cannot cause deviations of output from the natural level.
Only unexpected changes in the monetary instruments (here, the supply of money)
affect output, at lest in the short run (here, just in the current period). The kind
of "Phillips curve" that comes out from NCM models is then vertical with respect
to the anticipated part of monetary policy and shows a positive output-inflation
correlation only with respect to the unanticipated policy component.

References
[1] Cagan P. (1956) "The monetary dynamics of hyperinflation", in M. Friedman
(ed.) Studies in the quantity theory of money, Chicago, University of Chicago
Press
[2] Lucas R.J. (1972) "Expectations and the neutrality of money", Journal of
Economic Theory, 4
[3] Lucas R.J. (1973) "Some international evidence on the output-inflation trade-
off", American Economic Review, 63
[4] Lucas R.J. (1976) "Econometric policy evaluation: a critique", Carnegie-
Rochester conference series on public policy, 1
[5] Sargent T.J. and N. Wallace (1973) "The stability of models with money and
growth with perfect forecast", Econometrica, 41, 6
[6] Sargent T.J. and N. Wallace (1976) "Rational expectations and the role of
monetary policy", Journal of Monetary Economics, 2

22
[7] Taylor J. B. (1985) "Rational expectations models in macroeconomics", in K.
Arrow and S. Honkapohja (eds.) Frontiers of Economics, Blackwell, Oxford

23
Problems
1. (Expectations and price dynamics) Suppose that the market for a good is
described by the following demand and supply functions and by the market-
clearing condition:
qtD = a0 − a1 pt
qtS = b0 + b1 pet−1,t + ut
qtD = qtS ≡ qt
where qtD and qtS are, respectively, the quantity demanded and supplied of
the good at time t, and ut is a stochastic supply shock which follows a nor-
mally distributed white noise process, ut ∼ N(0, σ 2u ). All parameters in the
model are positive. The demand for the good at t depends deterministically
on price pt (no demand shock is present), whereas supply at t depends pos-
itively on the price expected one period earlier, pet−1,t , and is affected by the
disturbance ut .

(a) Find the equilibrium price for the good pt and explain its dependence
on expected price pet−1,t .
(b) Assuming rational expectations (such that pet−1,t = Et−1 pt ), find the
equilibrium price level pt . What are the determinants of the forecast
error in each period t ?
(c) Find the equilibrium price level pt under the alternative assumption of
adaptive expectations
pet−1, t = λ pt−1 + (1 − λ) pet−2, t−1 0<λ<1
and compare your result with the answer to question (b).

2. (Cagan’s hyperinflation model ) Using the hyperinflation model presented in

Section 1 with the assumption of rational expectations and with a white
noise stochastic disturbance ut added to money demand:

(a) find the price level pt knowing that money supply follows a first-order
autoregressive stochastic process:
mt = ρ mt−1 + εt
where εt is a white noise shock and 0 < ρ < 1.

24
 (b) How and why does the parameter ρ enter the relationship between
money and prices? (comment upon the cases ρ → 1 and ρ → 0).
(c) Find the price forecast error at time t and check that it is unpredictable
given information available at t − 1.

3. (Lucas critique). Consider a simple macroeconomic model composed by the

following aggregate supply (AS) and aggregate demand (AD) functions:

yt = γ (pt − Et−1 pt ) + ut (AS)

yt = mt − pt , (AD)

where all variables are in logs, y is the output level (with ȳ = 0 for simplic-
ity), p is the price level, m is the nominal quantity of money, and Et−1 pt
is the rational expectation of the price level in t formed by agents in the
private sector on the basis of all information available at time t − 1. u is a
supply shock such that Et−1 ut = 0. Suppose that the monetary authorities
set money supply in each period according to the following rule, perfectly
known to agents:
mt = m̄ + pt−1 + εt , (MR)
where m̄ is a constant and ε is a stochastic, unpredictable component
(Et−1 εt = 0).

(a) Prove that only the stochastic component of the monetary rule does
affect output;
(b) find a relationship between output yt and the actual inflation rate πt ≡
pt − pt−1 . What is the nature of the parameters in this relation?

4. (Intertemporal substitution and the Lucas critique) Let the aggregate supply
function be of the following form:

yt = γ (pt − Et pt+1 )

This function captures an "intertemporal substitution" mechanism, whereby

producers increase (decrease) output when the current price level pt is higher
(lower) than the price level expected for the next period (based on the
information available at t). The aggregate demand function takes the simple
form
yt = mt − pt

25
 and money supply is determined by the following stochastic process
mt = ρ mt−1 + εt
where εt is a white noise shock to the (first-order autoregressive) monetary
rule and 0 ≤ ρ ≤ 1.

(a) Solve the model for the equilibrium price and output levels [Hint: use
the method of undetermined coefficients]. How and why does the re-
sponse of output to a monetary shock εt depend on the parameter ρ?
(b) Find the relationship between yt and pt and show that the "slope" of
this traditional aggregate supply function depends on the parameter of
the policy rule ρ (an example where the Lucas critique applies).

5. (New Classical Macroeconomics) Consider a simplified version of the NCM

model studied in Section 3, with the following aggregate demand and aggre-
gate supply equations:
yt = α (mt − pt ) + vt
yt = γ (pt − Et−1 pt ) + ut
where vt ∼ N(0, σ 2v ) and ut ∼ N(0, σ2u ) are uncorrelated white noise de-
mand and supply shocks respectively. Monetary policy sets money supply
according to the following rule:
mt = mt−1 + δ 1 ut − δ 2 vt
(with δ 1 , δ 2 > 0). In this setting, monetary authorities have more infor-
mation than private producers, and can set mt after observing the time t
realization of the two shocks (whereas producers form expectations on pt
only on the basis of information dated t − 1).

(a) Solve the model for the equilibrium price level and output and discuss
their dependence on the parameters of the monetary rule δ 1 and δ 2 ;
(b) suppose that monetary authorities want to stabilize output around the
perfect-information level (with no price surprises), that is yt = ut . To
this aim, they choose the values of the policy parameters δ 1 and δ 2 to
minimize the variance of output fluctuations yt − ut :
min var (yt − ut ) = var (pt − Et−1 pt )
δ 1 ,δ 2

26
 Find the optimal values for the policy parameters and discuss the ef-
fectiveness of monetary policy in stabilizing output with respect to
demand and supply disturbances.

6. (New Classical Macroeconomics) Consider a modified version of the NCM

model of Section 3, where private agents in financial markets have more in-
formation than producers and monetary authorities. To capture this infor-
mation asymmetry the aggregate demand function (obtained by combining
the IS and LM schedules) is modified as:
yt = α (mt − pt ) + β Et (pt+1 − pt ) + vt
where the information set used by agents to form expectations on the infla-
tion rate is dated t (therefore including the current value of the price level
pt ) and vt ∼ N(0, σ 2v ) is a white noise aggregate demand shock. Aggregate
supply is of the traditional (Lucas) "surprise" variety:
yt = γ (pt − Et−1 pt ) + ut
with ut ∼ N(0, σ 2u ) being an aggregate supply disturbance, uncorrelated with
the demand shock. Finally, monetary policy follows a feedback rule which
makes money supply react to t − 1 realizations of the two disturbances:
mt = δ 1 ut−1 − δ 2 vt−1

(a) Solve the model for the price level and output in equilibrium;
(b) suppose (as in Problem 4) that monetary authorities want to stabilize
output around the perfect-information level (with no price surprises),
that is yt = ut . To this aim, they choose the values of the policy
parameters δ1 and δ 2 to minimize the variance of output fluctuations
yt − ut :
min var (yt − ut ) = var (pt − Et−1 pt )
δ 1 ,δ 2

Find the optimal values for the policy parameters and discuss the ef-
fectiveness of monetary policy in stabilizing output with respect to
demand and supply disturbances.
(c) Compare your results with those of Problem 4. Is it possible that
monetary policy is effective in stabilizing output even though mone-
tary authorities have less information than (some) other agents in the
economy? (provide an economic intuition for this result).

27