You are on page 1of 4

# Submit Assignment For Help

14.462

Problem set 4

## Consider the following model

pt = αpet+1 + εt
e
Where pt is the endogenous variable (say an asset price), pt+1 its expected value,
2
and εt an iid shock with zero mean and variance σ , which is not known a period
1. What is the rational expectations equilibrium (REE)?
Assume that instead of rational expectations, we have a learning process
deﬁned by
1
pet+1 = pte + (pt−1 − pte ),
gt
where gt is a deterministic sequence of strictly positive numbers/
We will say that the learning process converges to the REE if

## lim E(pet+1 − Et,REE pt+1 )2 = 0,

t→+∞

where E is the unconditional expectations operator, and Et,REE pt+1 is the value
of pet+1 in the rational expectations equilibrium, given the history of shocks.
2. Show that a necessary and suﬃcient condition for convergence to the
REE is
lim mt = 0, (1)
t−→+∞

where
t
� h2 ...h2
1 s t
mt = 2 + 2 ,
gt s=1
g s−1

and
ht = 1 + (α − 1)/gt .
3. Show that mt satisﬁes

2
(2)

## 4.Show that for (1) to hold it must be that

α < 1

lim gt = +∞
and
t

lim ht = 0
s=0

1
https://www.economicshomeworkhelper.com
5. Show that the latter condition is asymptotically equivalent to
+∞

1/gt = +∞.
0

## 6. Does the learning process converge to the REE is gt is a constant? If

gt = 1/(t + 1)2 ?
7. Assume gt = 1/(t + 1). Show that (1) is satisﬁed (hint: use (2) and look
for β > 0 such that one can show by induction that mt < At−β for t large
enough).
8. Intuitively, what are the problems if gt converges ”too fast” or ”too
slowly” to inﬁnity?

2
Spring 2004

## Problem Set 4 Solution

1. It is easy to check that pt = εt is the (bubbleless) REE equilibrium. Next consider the
learning process. Substituting the model equation into the learning equation yields
1 εt−1
pet+1 = pte + (αpte + εt−1 − pet ) = ht pet + .
gt gt
By induction one obtain
t
� �t+1 t
εs−2 �
pet+1 = p0e hs +

s=0 s=1
g s−1 τ =s

Using this formula and the fact that the εt are iid, we can compute

�t t+1
� t t

2 1
� 2 �
E
pet+1 − Et,REE pt+1 = (pe0 )2 h2s + σ 2 2
h τ = (p e 2
0 ) h2s + σ 2 m
t
s=0
g
s=1 s−1 τ =s s=0

�t
2. Necessity is obvious. To show suﬃciency notice that mt ≥ τ =1 h2τ , so limt→∞ mt = 0
insures that the term in front of pe0 vanishes as t → ∞.

3. We have
1
h2t+1 mt + 2
gt+1
t+1
� t t+1 t+1 t+2 t+1
1 � 1 � 1
� 2 1 � 1
� 2
= h2t+1 2
h2τ + 2
=
hτ + 2 =
hτ = mt+1
s=1
gs−1 τ =s
gt+1 g2
s=1 s−1 τ =s
gt+1 g2
s=1 s−1 τ =s

4. First notice that mt ≥ g12 so limt→∞ mt = 0 clearly requires that limt→∞ gt = +∞.
�t t
Also mt ≥ g21 2
τ =s τ for all s ∈ {1, . . . , t + 1}. It follows that limt→∞ mt = 0
h
s−1
requires
�t
lim hτ = 0
t→∞
τ =s

for all s ≥ 1, which is stronger than the requirement stated in the problem set since
it must hold for all s ≥ 1. Finally if α ≥ 1, then ht ≥ 1 for all t which is inconsistent

with limt→∞ tτ =s hτ = 0. Thus it must be the case that α < 1.

1
5. Suppose
t

lim hτ = 0
t→∞
τ =s
�+∞ 1
for all s ≥ 1. The goal is to show that 0 gt
= +∞.

6. Pick T such that t ≥ T implies 1−α < 12 and deﬁne an = g1−α for n ≥ 0. We want to
�∞ qt �∞ T +n
show that n=0 (1 − an ) = 0 if and only if n=0 an = +∞. This follows from some
results on convergence of inﬁnite products,. see the book Mathematical Analysis, 2ed
by Apostol, pp. 206–209. One direction is not so diﬃcult: as 1 − an ≤ e−an , we have
�∞ �∞
− n=0 an
n=0 (1 − an ) ≤ e . At least to me it seems that the other direction is a bit
more involved, but perhaps you found an easy proof. Otherwise take a look at the
book by Apostol.
1
�+∞ 1
7. If gt is a constant, then limt→∞ gt = +∞ fails. If gt = (t+1) 2 , then 0 gt
= +∞
fails.

8.