Foundations of International Macroeconomics

1
Workbook
2
Maurice Obstfeld, Kenneth Rogo!, and Gita Gopinath
Chapter 2 Solutions
1. (a) The current account identity can be written as B
s+1
= (1+r)B
s
+TB
s
.
Now just plug in the assumed trade balance rule.
(b) Using the answer to part a, for any ! > 0,
!
!
X
s=t

1
1 +r

s"t
!rB
s
= !
!
X
s=t

1
1 +r

s"t
!r[1 + (1 !!)r]
s"t
B
t
=
!!rB
t
1 !
1 + (1 !!)r
1 +r
= !(1 +r)B
t
.
(c) Under the rule above, debt grows without bound if ! < 1. But once the
debt is as big as Y/r, the country can honor its foreign commitments only
if debt stops growing and consumption is zero forever. Thus, the suggested
rule must entail negative consumption levels at some point, which are inad-
missible. To see directly why, consider the constant-output case, in which
TB
s
= Y ! C
s
= !!rB
s
so that the payback rule implies C
s
= Y + !rB
s
.
Notice that since B
s
" !#, C
s
must at some point become negative. The
rule therefore is consistent with intertemporal solvency only if we counterfac-
tually allow for negative consumption levels: the price of high consumption
today would be infeasibly high trade surpluses later on. In general, suppose
output grows at the gross rate 1 +g, so that Y
s
= (1 +g)
s"t
Y
t
. Unless 1 +g
1
By Maurice Obstfeld (University of California, Berkeley) and Kenneth Rogo! (Prince-
ton University). c MIT Press, 1996.
2
c MIT Press, 1998. Version 1.1, February 27, 1998. For online updates and correc-
tions, see http://www.princeton.edu/ObstfeldRogo!Book.html
14
is at least as great as the gross growth rate of debt, which was shown to be
1 + r(1 ! !) in part a, the external debt-output ratio is unbounded. Thus
the minimal payback fraction ! consistent with intertemporal solvency and
positive consumption is ! = 1 ! (g/r) (which is positive if we assume that
g < r).
2. (a) The expected utility E
t
U
t
is a weighted average over di!erent life
spans, with weights equal to the survival probabilities:
E
t
U
t
= (1 !") [u(C
t
)] +"(1 !") [u(C
t
) +#u(C
t+1
)] +
+"
2
(1 !")
h
u(C
t
) +#u(C
t+1
) +#
2
u(C
t+2
)
i
+....
(b) The result follows simply by expanding the expression in part a and
grouping terms together.
3. Recall that with isoelastic utility,
u(C) =
!
"
"
#
"
"
$
C
1"
1
!
1 !
1
!
, $ > 0,
log C, $ = 1.
Using the intertemporal Euler equation, we thus obtain,
(1 +r)# = 1 =
1
E
t
(

C
t+1
C
t

"1/!
). (1)
Since consumption has a conditional lognormal distribution, the natural log
of the gross consumption growth rate is conditionally normally distributed:
log
C
t+1
C
t
$ N

E
t

log
C
t+1
C
t

, Var
t

log
C
t+1
C
t

.
Thus
E
t
!
#
$

C
t+1
C
t

"
1
!
%
&
'
= E
t

exp

!
1
$
log

C
t+1
C
t

= exp

!
1
$
E
t

log
C
t+1
C
t

+
1
2$
2
Var
t

log
C
t+1
C
t

. (2)
15
[Consult footnote 41 on p. 313 of the book. Equation (2) follows from com-
puting the mean and variance of the random variable (!1/$) log(C
t+1
/C
t
),
which is normally distributed when log(C
t+1
/C
t
) is.] Combining eqs. (1) and
(2) above and taking natural logs of the result, we arrive at
E
t

log
C
t+1
C
t

=
1
2$
Var
t

log
C
t+1
C
t

or
log C
t+1
!log C
t
=
1
2$
Var
t
{%
t+1
} +%
t+1
,
where %
t+1
% log C
t+1
! E
t
{log C
t+1
}. Since %
t+1
is a normal random vari-
able that is uncorrelated with past information (because it is a pure fore-
cast error), it is also statistically independent of that information on the
assumption that the past information itself is generated by a jointly normal
(i.e., Gaussian) stochastic process. In that case the conditional variance in
the preceding equation actually is a time-invariant constant, so the natural
log of consumption follows a random walk with a constant drift equal to
1
2!
Var {%
t+1
} .
4. (a) Using eq. (32) in Chapter 2, we can write
C
t+1
!C
t
= r(B
t+1
!B
t
)
+
r
1 +r
!
#
$
!
X
s=t+1

1
1 +r

s"(t+1)
E
t+1
Y
s
!
!
X
s=t

1
1 +r

s"t
E
t
Y
s
%
&
'
.
The current account identity gives
B
t+1
!B
t
= Y
t
+rB
t
!C
t
= Y
t
!
r
1 +r
!
X
s=t

1
1 +r

s"t
E
t
Y
s
,
which can be substituted into the previous equation for consumption to give
the result that the change in consumption equals the present value of changes
in expected future output levels.
16
(b) If the process for output follows
Y
t+1
!Y
t
= & (Y
t
!Y
t"1
) +
t+1
,
then (E
t+1
!E
t
)Y
t+1
=
t+1
, (E
t+1
!E
t
)Y
t+2
= (1 +&)
t+1
, (E
t+1
!E
t
)Y
t+3
=
(1 +& +&
2
)
t+1
, and so on. Therefore, for s > t,
(E
t+1
!E
t
)Y
s
=
1 !&
s"t
1 !&

t+1
.
(c) Substituting the last expression into the equation for the change in con-
sumption derived in part a, we get the following
C
t+1
!C
t
=
r
1 +r

t+1
"
1
1 !&
!
+

1
1 !&
!
1
1 +r
+...
!
&
1 !&
!
&
2
(1 +r)(1 !&)
!...
#
=
r
1 +r

t+1
"
(1 +r)
(1 !&)r
!
&
1 !&

1 +r
1 +r !&
!#
=
1 +r
1 +r !&

t+1
. (3)
As a result, provided that 0 < & < 1, the desire to smooth consumption
makes consumption innovations more variable than output innovations.
(d) The current account identity for date t + 1 is
CA
t+1
= B
t+2
!B
t+1
= Y
t+1
+rB
t+1
!C
t+1
.
Because Y
t+1
!E
t
Y
t+1
=
t+1
and, by eq. (3) from part c above,
C
t+1
!E
t
C
t+1
= C
t+1
!C
t
=
1 +r
1 +r !&

t+1
,
the preceding current account identity gives a current account innovation of

t+1
!
1 +r
1 +r !&

t+1
=
!&
1 +r !&

t+1
< 0.
17
Thus, a positive output innovation leads to a current account decit, as
claimed at the end of section 2.3.3 in the book.
5. Work backward from the equation
CA
t+1
!"Z
t+1
!(1 +r)CA
t
=
t+1
,
where
t+1
is uncorrelated with date t or earlier information. Taking expec-
tations with respect to date t information yields
E
t
CA
t+1
!E
t
"Z
t+1
!(1 +r)CA
t
= 0.
The previous equation can be rearranged to express CA
t
as
CA
t
=
1
1 +r
E
t
CA
t+1
!
1
1 +r
E
t
"Z
t+1
.
Through forward recursive substitution (and using the law of iterated condi-
tional expectations ) we obtain
CA
t
= !
!
X
s=t+1

1
1 +r

s"t
E
t
"Z
s
[because as j "#,

1
1+r

j
E
t
CA
t+j
"0]. This is Campbells (1987) saving
for a rainy day equation, eq. (43) in Chapter 2. The equation can alterna-
tively be derived using the lag and lead operator methodology described in
supplement C to Chapter 2. Start again with
CA
t+1
!"Z
t+1
!(1 +r)CA
t
=
t+1
and take expectations with respect to date t information to get
E
t
CA
t+1
!E
t
"Z
t+1
!(1 +r)CA
t
= 0.
Using the lead operator we write this as
L
"1
CA
t
!L
"1
"Z
t
!(1 +r)CA
t
= 0,
18
or, dividing by 1 +r and rearranging, as

1 !
1
1 +r
L
"1

CA
t
= !
1
1 +r
L
"1
"Z
t
.
Inversion of the lag polynomial on the left-hand side above gives
CA
t
= !
1
1 +r

1 !
1
1 +r
L
"1

"1
E
t
"Z
t+1
= !
1
1 +r
!
X
s=t

1
1 +r

s"t
E
t
"Z
s+1
= !
!
X
s=t+1

1
1 +r

s"t
E
t
"Z
s
.
To derive the converse, that the last equation implies CA
t+1
!"Z
t+1
!(1 +
r)CA
t
=
t+1
, one can simply reverse the steps above.
6. Write the expression for the current account as follows
CA
t
= Z
t
!E
t
e
Z
t
= Z
t
!
r
1 +r
E
t
!
X
s=t

1
1 +r

s"t
Z
s
= Z
t
!
r
1 +r

1 !
1
1 +r
L
"1

"1
E
t
Z
t
,
where the last equality is suggested in the hint. Then, multiplying both sides
by 1 !
1
1+r
L
"1
, we have

1 !
1
1 +r
L
"1

CA
t
= !
1
1 +r
(E
t
Z
t+1
!Z
t
) = !
1
1 +r
L
"1
E
t
"Z
t
Therefore,
CA
t
= !
1
1 +r
L
"1
E
t
!
X
s=t

1
1 +r

s"t
"Z
s
= !
!
X
s=t+1

1
1 +r

s"t
E
t
"Z
s
.
19
7. Equation (75) in appendix 2A of the book shows that as time T passes
and as B/Y "# along its unstable increasing trajectory,
B
t+T+1
/Y
t+T+1
B
t+T
/Y
t+T
"
(1 +r)
!
#
!
1 +g
.
Because Y
s+1
/Y
s
= 1+g, however, it follows that B
t+T+1
/B
t+T
"(1+r)
!
#
!
as T "#, that is, net foreign assets grow asymptotically at the growth rate
of consumption. But the gross rate of consumption growth, (1+r)
!
#
!
, must
be strictly below 1 + r if, as we have assumed, an individual optimal plan
exists (see the discussion on p. 117 of the book). Thus the asymptotic gross
growth rate of foreign assets is below 1 +r and the transversality condition
lim
T#!
(1 +r)
"T
B
t+T+1
= 0
therefore is satised.
8. If permanent output uctuates more than current output, as is the case
when output is a nonstationary random variable, then, as shown in exercise
4(d), a positive output innovation implies a decline in the current account:
the intertemporal approach can therefore yield a countercyclical current ac-
count. Also, in the presence of investment which enters the current account
with a negative sign, the current account can worsen following a positive
output innovation if & is su#ciently large. Refer to pp. 86-87 in the book for
details.
9. (a) Di!erentiating the rms objective function with respect to I
s
and
K
s
, we get, respectively,
q
s
= 1 +'I
s
and
q
s
=
1
1 +r
[A
s+1
F
0
(K
s+1
) +q
s+1
] .
20
(b) Combining the preceding two rst-order conditions, we get the following
dynamic system for q and K:
K
t+1
!K
t
=
q
t
!1
'
,
q
t+1
!q
t
= rq
t
!A
t+1
F
0

K
t
+
q
t
!1
'
!
.
(c) The phase diagram looks qualitatively the same as gure 2.9 in the book,
which is based on eqs. (66) and (67) in Chapter 2. The dynamics of q and
K and the slopes of the corresponding schedules are, however, quantitatively
di!erent. For example, the slope of the "q = 0 schedule now is
dq
dK

!q=0
=
AF
KK

K +
q !1
'
!
r !

1
"

AF
KK

K +
q !1
'
!
< 0,
not just in a neighborhood of the steady state, but globally. (Compare with
the slope given on p. 109 of the book.) As in the slightly more complex q
model of Chapter 2, the steady state is given by q = 1, AF
K
(

K, L) = r. The
steady state is independent of the adjustment costs, which determine only
the speed of transition to the steady state.
(d) Figure 2.9 can be used for the exercise. An unanticipated permanent rise
in A shifts the "q = 0 schedule immediately and permanently to the right,
raising the steady state capital stock to

K
0
. The unique convergent saddle-
path SS also shifts to the right, becoming S
0
S
0
. Because the initial capital
stock is given as

K <

K
0
, q rises in the short run (to place the economy on
its new saddle-path) and investment surges. Over time, however, q falls back
to 1 and investment decreases as K "

K
0
.
(e) The principle for analyzing anticipated shocks is the same as that ap-
plied in gure 2.11. The rm learns on date t that productivity A will rise
permanently to A
0
at a known future date T. Thus, the shifts described in
21
schedules described in part d occur only on date T. Nonetheless, the rm
will adjust in anticipation so as to smooth its investment costs; and between
dates t and T, q and K therefore will follow the original equations of motion
(those involving A rather than A
0
). Thus, q jumps up initially and, until
date T, continues rising as capital is accumulated. The rm reaches the new
saddle-path S
0
S
0
precisely on date T, and thereafter q falls toward 1 and K
rises to its new steady state. (An anticipated future fall in A would induce
a path qualitatively similar to the path shown in gure 2.11.)
(f) Marginal q does not equal average q because the installation cost function
assumed in this exercise is not linear homogeneous in K and I.
22