li.'
'
'(
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/tradeintegrationOOkraa
B31
.DEWEY
^415
i
OB
Technology
Department of Economics
Working Paper Series
Massachusetts
Institute of
TRADE INTEGRATION
AND
RISK
SHARING
Aart Rraay
Jaume Ventura
Working Paper 02-08
February 2002
Room
E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http://papers.ssrn. com/paper.taf?abstract_id=xxxxx
[^^y
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
TRADE INTEGRATION
AND
RISK
SHARING
Aart Kraay
Jaume Ventura
Working Paper 02-08
February 2002
Room
E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science
Research Network Paper Collection
at
http://papers.ssrn. com/paper.taf?abstractJd=xxxxx
MASSACHUSEHS
INSTITUTE
OF TECHNOLOGY
MAR
1 5 2002
LIBRARIES
Trade Integration and Risk Sharing
Aart Kraay,
Jaume
The World Bank
Ventura,
MIT
February 2002
What are the effects of increased trade in goods and services on the trade
balance? We study the effects of reducing transport costs in a Ricardian model with
complete asset markets and find that this increases the volatility of the trade balance.
This result applies regardless of whether supply or demand shocks are the main
source of economic fluctuations. Both type of shocks generate fluctuations in the
trade balance that are in part moderated by stabilizing movements in the terms of
trade. Trade integration dampens these terms of trade movements and, for a given
distribution of shocks, amplifies fluctuations in the trade balance. To overturn this
result, one must assume that either trade integration is sufficiently biased towards
goods with strong comparative advantage and/or risk aversion is sufficiently extreme.
We calibrate the model to U.S. data and find that, for reasonable parameter values,
increased trade in services could double the volatility of the trade balance.
Abstract
:
We thank Pol Antras, Antonio
Fatas, Philip Lane and participants
in
the International
Seminar on
College Dublin, June 2001 for their useful comments. We are very
grateful to Bjoern Bruegemann for pointing out a mistake in a previous version of the paper. The views
expressed here are the authors' and do not necessarily reflect those of The World Bank.
Macroeconomics
at University
,
In
the last
more than
thirty
years, the
During this
tripled.
have grown larger and more
volume
same
same
trade
set of
due
is
goods and
economic
volatile.
It
is
in
it
industrial countries
among these
natural to ask whether there
in particular,
is
has
countries
a connection
whether they are both driven by
Many have argued
to a reduction in the technological
services.^ Is
the increase
forces.
among
period, trade imbalances
between these two developments and,
the
of trade
that the increased
and policy-induced costs
volume
of
of trading
possible that these cost reductions are also responsible for
the size and
volatility of
theoretical channels through
which
this
trade imbalances?
If
so,
what are the main
happens? Are these channels
quantitatively
important?
Unfortunately, there
is little in
answer these questions.^ This state
the form of received "wisdom" that can help us
of affairs reflects in part the
absence
of a simple
workhorse model incorporating the main insights of the theories of goods and asset
trade and the key interactions between them. This paper attempts to
strategy
is
to build
in
we assume
assets,
we assume
to
it.
To
motivate trade
in
some goods can be
countries have different industry technologies.
To motivate trade
countries experience imperfectly correlated shocks to
become
that
at prohibitively high transport costs (the
process of trade integration as one
in
which
"nontraded" sector).
some nontraded goods
traded.
The main
volatility
we assume
traded at negligible transport costs (the "traded" sector), while the
can only be traded
We interpret the
gap. Our
goods and
technology (or "supply") and to preferences (or "demand"). As usual,
rest
this
from the classic continuum model of Dornbusch, Fischer and
Samuelson [1977] and add asset markets
services,
fill
theoretical result of this
paper
is
that trade integration increases the
of the trade balance. This result applies regardless of
whether supply or
Baier and Bergstrand (2001) find that about 31 to 45 percent of the increase in the volume of trade can
be explained by reductions in costs of trading goods and services. Of this total, tariff rate reductions and
preferential agreements account for 23 to 26 percent, and transport cost declines for 8 to 9 percent. We
think that these are very conservative estimates and the real numbers might be even higher.
^ A remarkable exception is Cole and Obstfeld
[1991], who provide an example in which a drastic
reduction (from prohibitive to negligible) in the costs of trading all goods and services has no effect on
the trade balance. As we shall see later, their model obtains as a special case of ours.
^
demand shocks
are the main source of economic fluctuations. Both type of shocks
generate fluctuations
in
the trade balance that are
in part
moderated by
stabilizing
dampens these terms
movements
in
movements
and, for a given distribution of shocks, amplifies fluctuations
balance.
To
the terms of trade. Trade integration
overturn this result
is
not easy
cases.
The
goods
with strong comparative advantage.
first
goods must
one requires
in
our framework, but
that trade integration
By
this
be
we mean
extreme. That
is
is,
in
two
that the newly-traded
aversion be
likely to
risk
benchmark.
be empirically relevant.
trade integration increase the effects of supply shocks on the trade
balance? Economy-wide fluctuations
production of each good and also
When
risk
either 'large' or 'small' relative to the logarithmic
Why does
the traded sector
primarily raise the production of
in
is
in
labour productivity lead to fluctuations
in
the
the range of traded goods produced by the
small,
shocks that raise labour productivity
goods already produced
lowers their prices and worsens the terms of trade. This
in
can be done
preferences must exhibit a coefficient of relative
However, these two exceptions are not
increase
the trade
exhibit cross-country differences in productivity that are 'large' relative to
aversion that
country.
in
biased towards
sufficiently
those of existing traded goods. The second case requires that
sufficiently
it
of trade
in
in
the country, and this
turn
moderates the
income and the trade surplus created by the shock.
Similarly,
shocks that
lower labour productivity improve the terms of trade, moderating the resulting trade
deficit.
When
reflected
in
the traded sector
fluctuations
effects on their prices
in
Why
In
shocks
to labour productivity are mostly
the range of goods produced at home, with only small
trade.
By increasing the
size of the traded
decreases the effects of supply shocks on the terms of trade
volatility of
the trade balance.
does trade integration reduce the effects of demand shocks on the trade
balance? The
problem.
large,
and the terms of
sector, trade integration
and so raises the
is
intuition is
simple and
is
based on the classic analysis of the transfer
the presence of transport costs, there
since domestic goods are cheaper at
home
is
a
home
bias
in
consumption
than abroad. Under these conditions,
shocks that increase spending
re-direct
demand towards domestic goods and
improve the terms of trade. This raises income and finances
spending, moderating the trade
deficit.
in
part the increase
Through the same mechanism, shocl<s
in
that
reduce spending worsen the terms of trade, moderating the trade surplus. By
home
weal<ening the
demand shocks on
Armed
bias
in
consumption, trade integration lessens the effects of
the terms of trade and raises the
with these theoretical findings,
we
volatility of
calibrate the
the trade balance.
model
to U.S. data to
provide a quantitative assessment of the effects of further trade integration.
that future trade integration
is likely
to
countries, services account for almost
percent of exports and imports."^
of technological
premise
is
that
70 percent
To some
and policy-induced
some
be concentrated
in
services goes half the
of the
of value
way and
all
in
fall
volatility of
The paper
is
for
20
significantly
Our
over the next
communications technology as well as
We therefore study two scenarios in which trade in
the
way towards
"catching-up" with trade
economy. For plausible parameter values, we
doubles the
added but only
extent this mismatch reflects a wide variety
of these barriers are likely to
regulatory barriers.
services. In industrial
barriers to trade that are specific to services.
decade or two, spurred by improvements
reductions
in
We argue
in
the rest
find that trade integration
almost
the U.S. trade balance.
organized as follows: section one presents a version of the
Dornbusch-Fischer-Samuelson model of Ricardian trade with asset markets. Section
two uses the model to determine the main theoretical effects of trade integration on
the trade balance. Section three calibrates the model to actual data and provides a
quantitative
assessment
of the effect of increased trade in services
on the U.S. trade
balance. Section four concludes.
Moreover, much of existing trade in services is concentrated in transportation and travel. For instance,
the U.S. these two items constitute roughly half of service trade but only five percent of service
^
in
production.
The Dornbusch-Fischer-Samuelson Model
1.
with
Asset Markets
This section presents a simple model designed to study
how the
We build
costs of commodity trade affect the behavior of the trade balance.
classic Ricardian trade
model with a continuum
of
goods due
and Samuelson [1977], and then add asset markets
lasts
one period and consists
endowed
with labour, L
and
of
L*.
two countries:
As usual an
to
it.
to
nature and
on the
Dornbusch, Fischer
We consider a world that
Home and
Foreign.''
Each country
is
asterisk refers to Foreign variables.
Countries use labour to produce a continuum of intermediate goods, indexed by
ze[0,1].
that
is
These intermediates are then combined
used
for
to
produce a nontraded
final
good
consumption.
Countries trade
in
goods
to exploit differences in the
technology used to
produce intermediates. The extent to which they are able to do so depends on the
costs of trading these intermediates.
In particular,
we assume
that a fraction x of the
intermediates can be transported across countries without cost.
We refer to these
intermediates as the traded sector and assign them a low index, Z6[0,t].
the intermediates cannot be transported across countries and
nontraded sector, zg(t,1].
We shall
interpret
changes
in
we
refer to
The
rest of
them as the
the equilibrium as x
increases as the effects of trade integration.
Countries trade
in
assets to insure against
risks.
At the beginning of the
period, countries are uncertain about their labour productivity
consumption. As usual,
we assume
that they
these variables, but not their realizations.
*
It
is
not
difficult to write
assume throughout
know
and
their taste for
the true probability distribution of
In particular,
we assume
a multi-period version of this model. But there
is little
that there are
point
in
S
doing so, since
we
markets are complete and factors of production are nonreproducible. Under the standard assumptions that shocks are independent and identically distributed,
and both counthes have 'ex-ante' identical time-separable and homothetic preferences, there is no
incentive to engage in intertemporal trade and the multi-period model is equivalent to a sequence of oneperiod models.
that international financial
states of nature, indexed by s=1,...,S; and assign state s a probability
tis.
Since
countries are risl<-averse, they have an incentive to share risks before the state of
nature
We
revealed.
is
rule out frictions
in
financial
markets and allow countries to
set of Arrow-Debreu securities.
freely trade a
full
1.1 Firms,
Technology and The Labour Market
Each country contains many competitive firms
intermediates. In the final
good
sector, firms
technology that requires the use of
consumption
good
is
positive
is strictly
positive
means
nontraded, this
in all
Home and
all
that
produce the
final
good and
use a symmetric Cobb-Douglas
intermediates.
We shall see
later that
countries and states of nature. Since the final
in all
that the production of the final
good
is
also
countries and states of nature. Therefore, the prices of the
strictly
final
good
in
Foreign, Ps and Ps*, are given by:
1
P3=exp
(1)
where
ps(z)
and
ps (z) are the prices of variety
Foreign. Since the final
applies
will
if
good
is
z of intermediates
not traded, purchasing
power
Home and
in
parity
(i.e.
the prices of intermediates are equalized across countries.
In
Ps=Ps*)
general, this
not be the case.
In
Foreign
is
The
W^a(z)
—
fg*.
We shall
Foreign firms produce intermediates
cost of producing one unit of intermediate z
and
varies across countries
and
Home and
the intermediates sector.
using only labour.
fs
and
jlnp3{z)-dz
—W*^a
.
(z)
and states
,
respectively.
of nature in a
refer to variation
in fs
and
fs*
in
Home and
We assume labour productivity
way
that
is
captured by the indexes
as technology or supply shocks. This
is
the
source of uncertainty
first
in this
We assume that unit labour
model.
requirements vary across intermediates and across countries
captured by the technology schedules a(z) and
intermediates are produced
intermediates are produced,
only
if
—— > —
a*(z)
a"(z')
a(z)
a(z')
^
-
,
for
all
it
equilibrium.
in
is
Home
Home
produces
the traded intermediates,
To
good
is
final
good,
the traded
rule that z<z'
and
if
exports while traded intermediates with high
assume
in
which one country
—
a*(z)
^-^
that lim
z^o a(z)
that separates
the good for which costs of production
W
that
produce the
them using the
rule out states of nature in
Let Zs be the index of the cutoff
(2)
way
z,z'g [0,t]. This ordering rule implies that traded
indexes are
all
to
To determine where
useful to order
intermediates with low indexes are
imports.
a
a*(z).
Given our assumptions about the technology used
all
in
Home
Home
=
oo
and
—^—
a*(z)
lim
^
=
exports and imports,
and Foreign are the same:
W
prices of traded intermediates are the
same
in
both countries and can be
written as follows:
W
-^a(z)
P3(z)
= p;(z) =
W'
ze[0,Z3]
.
-^a(z)
ze(z3,T]
Nontraded intermediates are produced
might
i.e.
^.a(z3) = ^.a(z3)
The
(3)
.
z^T a(z)
differ:
in
both countries, and their prices
=
p^(z)
(4)
^.a(z)
Equations
(1)-(4)
we need
summarize
W
good. Then,
final
^a*(z)
firm maximization
prices.
(5) states that
we have
ifz£(T,1]
and provide
To complete
clear.
all
the relevant
the production side of
Define Cs as Home's
that:
PC^^
L
,. =Z3+(1-T)W^L,'%,,
+ W, L
.„
Equation
=
ensure that labour markets
to
consumption of the
(5)
p;(z)
between goods and factor
relationships
model,
and
^^
W3L + W3L
,
Home's share
of world labour
income or production
equals the share of world spending on goods produced by Home. The
of the shares of world
latter consists
spending on Home's traded sector and nontraded sector,
respectively.
Equations
(1)-(5)
are almost identical to those of the original Dornbusch-
Fischer-Samuelson model. These equations determine the pattern of labour income
and goods trade as a function
model,
we
that
Home's shares
of world labour
income and spending
most by an exogenously given transfer or trade balance. This modeling
strategy permitted an illuminating analysis of the
After
To complete the
therefore need a consumption side for the model. Dornbusch, Fischer and
Samuelson [1977] assumed
differ at
of the world distribution of spending.
World Wars
I
and
II,
economic
effects of
war
reparations.
cross-border financial transactions were severely restricted
and the transfers imposed on the defeated countries were determined mainly by
political factors.
today,
when
But
this
does not seem
to
be the appropriate modeling strategy
a sophisticated international financial market exists
can buy and
environment,
sell
we
in
which countries
a large array of securities. Recognizing this change
in
the economic
next provide a market-based theory of the determinants of the
transfers or trade balances.
1
.2
Preferences and Asset Markets
Each country contains a representative consumer that maximizes the
following
utility
U=
(6)
where
nature
is s.
variation
yn3- ^--^-
and
ds
function:
-^
and
ds* are variables that
We assume that ds and
U'
=
Vn^
measure the value
^-
of
'^^
"^
consuming when the
ds* vary across states of nature,
and
refer to this
as preference or demand shocks. These shocks are the second and
source of uncertainty
in
state of
final
the model.
Representative consumers obtain income by inelasticaliy supplying a labour
endowment equal
to L
and
L*.
Therefore, their income
is
equal to the
wage times
the
labour endowment, and their budget constraints can be written as follows:
s
s
XQs(PsC3-W3L)<0
(7)
and
s=1
where Qs
state
s.
is
JQ3
s=1
(p;
c; -
Wj
L )<
the price of the Arrow-Debreu security that delivers one unit of income
These
in
constraints simply state that the total (across states of nature) value of
consumption cannot exceed the
total
value of income. Naturally,
in
each state the
values of income and consumption need not be equal. Maximizing (6) subject to (7)
we
^3^
obtain:
QsPsC3
_
v.d3.(Q3.P3)V
^^^
Qs-pjc;
vd;(QsP:)T
^Q3,.W3,.L
|]v.d,-(Q3.p,)'T
XQs-Ws'.l
|^v.d;..(Q3..p;)T
s'=1
s'=1
s'=1
s'=1
Equation
describes
The share
of nature.
(higher
(8)
of
spending
and consumption
tTs)
how consumers
is
higher
yields higher
in
states of nature that are
consumption good
is
ds).
The
ambiguous.
amount
On
usual, the
On
consideration dominates
while the second consideration dominates
the one hand,
in
if
risk
if
risk
aversion
is
in
of nature.
To
which consumption
aversion
is
is
low, y<1,
high, y>^. In the magical
distribution of
not affected by the price of consumption.
Equilibrium
in
international financial markets requires that the world value of
consumption equals the world value of income
(9)
To
which the
case of logarithmic preferences, y-^, the two effects cancel and the
is
between
the other hand, consumers are risk averse and want
should spend more on those states
first
lil<ely
relationship
consumption as evenly as possible across states
attain this objective they
spending
more
of total consumption.
should spend more on those states
cheap.
to distribute their total
As
is
to achieve the highest possible
attain this objective they
expensive.
(higher
utility
spending and the price of consumption (Qs-Ps)
consumers want
spending across states
distribute their
in
each state of nature:
p3C3 + p3*c;=w3l + w;l
Equations
prices of
all
income and
(8)
and
(9) implicitly define the distribution of
Arrow-Debreu securities as a function
consumption and the
of the world distribution of labour
price levels.
This completes the presentation of the model. Equations (1)-(5) and (8)-(9)
describe the solution of the model up to a choice of numeraire. For a given
distribution of spending, the production side of the
(1)-(5))
model (as described by Equations
determines the world distribution of labour income, price levels and the
pattern of
levels, the
goods
trade. For a given world distribution of labour
income and
consumption side of the model (as described by Equations
price
(8)-(9))
determines the world distribution of spending and the pattern of transfers or trade
balances.^
2.
Determinants of the Trade Balance
Our goal
integration
in this
section
is
to
develop results about the effects of trade
on the trade balance. To do
we make
this,
three strategic assumptions:®
A1. (Identical size) L=L*=1;
A2. (Symmetric technologies) Let the relative technology schedule for the traded
sector be A(z;t) =
a*(z)
—
—^
for ze[0,T].
Then, A(z;t)- A(t-z;t) =
1
for
all x;
a(z)
A3. (Symmetric shocks) Let 5^ =
such that
((t)s,5s)=(<j),5),
^
and
=
(t)s
-!-
.
If
there exists a state s with ns-n
then there exists another state
s'
with tis—n such that
((t)s',5s)=(<t)-\5-^).
Assumption A1 simply states that countries have the same size and
normalizes
and forces
it
to one.
Assumption A2 centers the
(log) productivity differences to
relative
technology schedule
at
0.5t
be symmetric. This can be understood as
saying that no country has a superior technology on average. By writing the
'
The Arrow-Debreu model
in different
states of nature.
of asset trade
These
generates predictions about the transfers that countries
predictions are robust
in
the sense that enlarging the
menu
make
of assets
consumers can choose from would not affect these transfers or trade balances. But it is difficult to find
assets that resemble the set of Arrow-Debreu securities that the theory postulates in existing financial
markets. Nevertheless, we do not believe that this observation invalidates the theory. The key
assumption underlying the model's predictions for the trade balance is not that a full set of Arrow-Debreu
securities is actually traded, but instead that
is possible to manufacture them with the available menu
of assets. Whether this assumption provides a reasonably good description of actual financial markets is
a hotly debated question. Nevertheless, we shall adopt in what follows. Since we do not specify the
available menu of assets, we interpret the theory as being silent on what specific assets are used to
implement the equilibrium transfers or trade balances. Recognizing this, we focus only on its predictions
it
it
for the trade balance.
^
We will
relax these
assumptions
in
the next section
when we
10
calibrate the
model
to the U.S. data.
technology schedule as an
we
recognize that
a(z)
and
sectors.
its
explicit function of
the size of the traded sector,
shape depends not only on our technology assumptions,
a*(z); but also
on how we
distribute
defined
this distribution is
absolute shocks. This
generate fluctuations
is in
in
A(.;t),
i.e.
goods between traded and nontraded
Assumption A3 says that both countries face the same
Note that
i.e.
distribution of shocks.
terms of relative shocks as opposed to
in
shocks can
anticipation of the finding that only relative
the trade balance and the terms of trade. Together, these
three assumptions ensure that the two countries are symmetric and this simplifies the
algebra of the problem substantially.
An
implication of
assumptions A1-A3
is
^Q3W3.L
wealth 'ex-ante',
^^|
.
X^s^-ds-(Qs-Ps)'r
^7r3^-d;-(Q3-p:)T
seS
sgS
In this particular
case, Equations (8)-(9) admit a simple and intuitive closed-form solution for
P^ C
share of world spending, 63 =
^
e3=e(co3,S3;x,Y)^
1
where
co^
=
W—
—
W3/f3
This variable
exchange
/f
!
will
!-
is
Home's
^
Ps-c^+pjc;
'
(10)
real
^Q^WJ-L*
=
^^^
i.e.
have the same
that both countries
'
^
+ 63-(03<'-^>T
Home's
relative labour costs or double-factoral
play a crucial role
in
what follows and
rate or ratio of price levels since —I'
lead to an appreciation of the real
exchange
=
cOg^'^
.
is
terms of trade.
closely related to the real
Since higher terms of trade
s
rate, this
means
that
9e
—
^>
3co,
11
if
y>^
,
and
de
—
^<
if
Y<1. Note also that trade integration reduces the effects of
changes
in
the
terms of trade on spending.
Let Xs be the share of traded
defined as A(tXs;x) = c%
function. Then,
.
Assume
X3=x(co3;x)s
goods produced
A(.;t) is invertible
^'
^
in
,
Home,
and
let
i.e.
Xs is implicitly
A"\.;t)
with x(t03;x) = 1-x(a)3V)
be
and
its
—^<O.The
T
effects of
changes
in i
on
distribution of unit labour
X3 are
aco3
ambiguous, since they depend on how the
requirements of the marginal traded goods compares to that
of existing traded goods. In the top (bottom) panel of Figure
which the
inverse
distribution of unit labour
requirements
is
1,
we
depict the case
in
uniformly less (more) dispersed
across marginal goods than across existing traded goods. This leads to a counterclockwise (clockwise) rotation
requirements
is
is
we say that
in
unit labour
in
the top panel
biased towards goods where comparative advantage
the bottom panel
in
the Xs schedule. Since dispersion
the source of comparative advantage,
trade integration
while
in
it
is
is
weak,
biased towards goods where comparative advantage
is
strong.^
Using
this notation,
/l^s-cos
(11)
1
we can now
rewrite Equation (5) to obtain:
=^.x(co3;T) + (1-T).e((03,S3;x,Y)
+ (t)s-«s
This definition is not devoid of some ambiguity, since tliere is no reason for the distribution of unit
labour requirements of the marginal traded goods to always be uniformly more or uniformly less
dispersed than that of existing traded goods. For instance, if trade integration is biased towards goods
weak comparative advantage, the distribution of unit labour
requirements of the marginal traded goods has more mass both in the tails and around the mean than
that of existing traded goods. In this case, the rotation of the Xs schedule is counter-clockwise near the
middle, but clockwise in the extremes. Obviously, more complicated shifts are also possible. To eliminate
any ambiguity, we assume from now on that the distribution of unit labour requirements of the marginal
traded goods is either uniformly more or uniformly less dispersed than that of existing traded goods. It is
straightforward to extend the analysis to the general case.
that exhibit either very strong or very
12
The
left-hand side of Equation (11)
production, while the right-hand side
Equation (11)
implicitly
is
is
Home's share
world spending on
defines the equilibrium value of
balance as a share of world income,
and imports, T(1-Xs)es, we can
ts, is
cos.
income or
of world labour
Home produced
goods.
Since Home's trade
the difference between exports, TXs(l-es),
write:
t3=T-[x(co3;T)-e(co3,53;t,Y)]
(12)
We shall
next use Equations
(1
1
)-(1 2)
to study the cyclical behavior of the
trade balance and the effects of trade integration.
sections 2.1 and 2.2
on spending. That
is,
we assume
we
restrict
To
streamline the discussion,
that fluctuations in the
in
terms of trade have no effect
the analysis to the logarithmic case
in
which y=1-
This restriction leads to a clean description of the effects of trade integration on the
trade balance.
In
section 2.3,
we examine
the consequences of removing this
restriction.
The
2.1
Cyclical Behavior of the
In this
to
world economy, countries use asset markets ex-ante to transfer income
those states of nature
consumption
are
made
Trade Balance
is
high.
in
The
which
their labour productivity is
trade balance records the transfers between countries that
ex-post. In general, the nature of these transfers
distribution of
shocks and
low and the value of
their effects.
next two polar examples where
all
To
isolate the
main forces
the shocks are of the
Consider first an economy where fluctuations
depends on the
in
same
at work,
(j)>1.
We
two states of nature s=H,L; with
refer to
the trade balance are driven
((})H,5H)=((t),1)
and
we assume
((t)L,5L)=((l)"\l)
these as the high and low states, respectively. The
13
study
type.
exclusively by technology or supply shocks. In the top panel of Figure 2,
that there are
we
AS
with
schedule
plots
i.e.
Home's share
of world labour
income
the left-hand side of Equation (11).
The slope
higher terms of trade raise Home's relative
AS
is
schedule
Home's
for different
relative productivity the larger is
Home traded
relative labour
its
i.e.
plotting the trade
Equation
(12).
The
equilibrium value for
to the
cog
TB
Home
in
ts is
in
this
depends on the
AD
The
is
in
low.
If
balances are small
same spending
When Home's
in
these movements
in
which
(i)
large
in
productivity
Home's share
in
if
enough
,
the
both states despite the
on the terms of trade,
is
high,
its
i.e.
on
terms of trade
of world labour
first to
provide an influential example of
to eliminate trade balances. Their
there are only supply shocks,
t<1
in
income and
when Home's
the terms of trade are strong, trade
(ii)
—^
3ws
result also holds
uses
absolute value.
Cole and Obstfeld [1991] were the
movements could be
Home
the high state and a trade deficit
effects of productivity gains
schedule.
quite standard.
lowering the required trade surplus. Naturally, the opposite applies
is
AS and AD
size of the trade balances that are required to
deteriorate moderating the increase
productivity
i.e.
the low state and finances this by selling income
able to achieve the
is
TB
obtained by projecting the equilibrium
the top panel of Figure 2
labour income.
in
the slope of the
in
of this
schedule.
story depicted
fluctuations
model
for different
The slope
obtained by crossing the
cOs is
the high state. By running a trade surplus
achieve
AD
balance for different values of the terms of trade,
asset markets to purchase income
low state.
produced goods
the right-hand side of Equation (11).
schedules, while the equilibrium value for
in
Home
income. The
goods. The top panel of Figure 2 also shows the
schedule
The
share of labour income. The
negative because ceteris paribus higher terms of trade reduce the
is
for
value for
its
across states since, holding constant the terms of trade, the larger
shifts
values of the terms of trade,
demand
positive because, ceteris paribus,
wages and
schedule plots the share of world spending on
schedule
is
values of the terms of trade,
provided that y=^
14
=
and
(ill)
how these terms
example
t=0.
is
of trade
the special case of our
Our model shows
that their
Consider next an economy where fluctuations
exclusively by preference or
assume
shocks.
refer to these
((t)H,5H)=(1 ,5)
is
the
demand
for
Home
shifts
because, ceteris paribus, the larger
lower
is
demand
the
for
Home
and
is
AD
Home's share
nontraded goods. The
Home's share
is
exports and the higher
is
we
((t)L,5L)=(1 ,5'^);
as the high and low states. The
across states because, ceteris paribus, the larger
spending the higher
the trade balance are driven
the bottom panel of Figure 2,
In
there are two states of nature s=H,L; with
and 5>1. Once again, we
shifts
demand
in
TB
schedule
of world
schedule also
of world spending the
demand
the
for
Home
imports.
The
story depicted
Home
uses asset markets
selling
income
surplus
is
high.
in
in
to
purchase income
the low state.
the low state,
The
the bottom panel of Figure 2
in
Home
in
the high state and finances this by
By running a trade
is
also quite standard.
is
deficit in
able to spend more
when
the high state and a trade
the value of consumption
size of the trade balances that are required to achieve this
the effects of increases
the slope of the
AD
in
spending on the terms of trade,
schedule.
When Home spending
is
i.e.
high,
depends on
on both the
its
shift
and
terms of trade
improve, raising Home's share of world labour income and lowering the required trade
deficit.
Of course, the opposite applies when
find that
movements
if
in
Home
spending
is
low.
Once
again,
the terms of trade are strong, trade balances are small
we
in
absolute value.
The
intuitions
developed
in
these two polar examples carry almost directly to
the general case with
many
mixture of supply and
demand shocks.
Instead,
we
does trade
build
on these
integration
states of nature and with
intuitions to
change the
We shall
some
of these states involving a
not pursue this point further though.
address the main question of
cyclical
this
behavior of the trade balance?
15
paper;
how
2.2
The
Effects of
We
Trade Integration
have seen that supply and demand shocks generate fluctuations
trade balance that are moderated
trade. Next,
we show how trade
part
in
by
integration
stabilizing
movements
dampens these terms
and, for a given distribution of shocks, this amplifies fluctuations
To see
this,
in
in
the
the terms of
of trade
movements
the trade balance.
in
apply the implicit function theorem to Equations (11)-(12) with y=1 to
obtain:
dts.
(1
+
^t.
<t>s-»3)'
(13)
dx
^
*s
(1
9Xs
+ <|)s-«s)^
+T
ax.^
"""
dz
J
T
9«s
Our symmetry assumptions A1-A3 imply
that the
average trade balance
is
zero. Therefore, a necessary condition for the volatility of the trade balance to
increase with t
is
that sign(ts)
= sign
^dt ^
—
3x
for
all s.
Since
dx
5.
< o we have
,
3co,
established the following result;
Result #1:
and
Assume
that
trade integration
volatility
is
changes
unbiased,
is
i.e.
—
^ =
dx
.
do not
affect spending,
i.e.
y^1;
Then, trade integration increases the
the main result of the paper and provides theoretical support for the
view that a reduction
in
the costs of trading goods and services
for larger trade imbalances.
distribution of
is
the terms of trade
of the trade balance.
This
number
in
Note that Result #1 does not place any
shocks beyond assumption A3.
of states of nature with
In particular,
part responsible
is in
restriction
there could be any
any mix of demand and supply shocks. The
simple: trade integration both flattens the
AD
16
on the
schedule and makes
it
intuition
less sensitive
to
changes
in
the world distribution of spending.
The
first
of these effects
reduces the
impact of supply shocks on the terms of trade and increases their impact on the trade
balance.
The two
effects
combine also
terms of trade while increasing
is
It
Result #2:
If
trade integration
it
trade integration
reduce the impact of demand shocks on the
impact on the trade balance.
their
also immediate to provide a
comparative advantage,
If
to
first
is sufficiently
qualification to our
result:
biased towards goods with strong
could lead to reduction
is
main
in
the
volatility
of the trade balance.
biased towards goods with strong comparative
advantage, the Xs schedule rotates clockwise.
schedule might become steeper
large and the equilibrium lies
in
If
this effect is strong
the middle range.
in this
Assume
that
AD
enough, the
shocks are not too
range. Then, trade integration increases the
impact of supply shocks on the terms of trade and reduces their impact on the trade
balance. Therefore, the
example
economy
that proves Result #2.
effects of
demand shocks on
It
with supply shocks of section 2.1 provides
is
also possible that trade integration increases the
the terms of trade and reduces their impact on the trade
balance. But since trade integration
makes
the
we need
AD
the world distribution of spending,
in
slope more than "compensates" for the smaller
with
demand shocks
schedule less sensitive to changes
the extra requirement that the increase
in
economy
an
shift.
When
this
happens, the
of section 2.1 provides an additional
example
that also
proves Result #2.^
Is
there any reason to expect trade integration to be biased towards goods
with strong comparative
advantage?
without actual data. But
some
trade integration with
most popular
as a situation
in
its
Naturally, this question
intuition
cannot be answered
can be obtained by comparing our model of
alternative.
We
have modeled trade integration
which the transport costs of a small set of goods
falls
dramatically
in
these two examples, x
—
tg
9X
Note that
^ has different sign and
dx
17
is
larger
in
absolute value than
T
from prohibitive to negligible, without placing any
restrictions
these goods. An alternative way to model trade integration
transport cost for
all
goods and consider a small reduction
goods
high transport costs only
transport costs
fall
goods
well. In this alternative
now
apparent shortly,
integration
cases
it
aversion
AD
cost. In this case, at
in this
start to
be traded as
in
Trade
we remove
in
which changes
the restriction that y=^
.
in
the terms of
As
will
become
not possible to derive general results on the effects of trade
is
first
economy where
the
high, y>1
,
If
the terms of trade and the
respect to the benchmark.
AD
If
aversion
the effects of supply shocks
this
If
i.e.
in
is
low, y<1
,
—^ < —^ <
.
If
of
spending depends negatively on
If
risk
still
is
_
is
Home's share
The
high,
only difference with the
spending becomes counter-
the trade balance with respect to the
low, the opposite applies.
18
of
the qualitative description of
dcOg
applies.
aversion
volatility of
aversion
5.
1-T
9cOs
Figure 2
quantitative.
risk
^ =
dx
benchmark case
the effects of terms of trade changes on
increases the
benchmark case.
dx
—
and the TB schedules rotate counter-clockwise with
9a)g
and
i.e.
we
fluctuations are driven by supply shocks.
rotate clockwise with respect to the
risk
spending are not too strong,
is
simplify matters,
spending depends positively on the terms of trade and both
and the TB schedules
benchmark case
To
this section that trade integration is unbiased,
logarithmic preferences.
cyclical
same
always biased towards goods
of section 2.1 allows us to derive useful intuitions.
Consider
the
the
on the trade balance. However, a detailed examination of the two polar
assume throughout
risk
is
of
more general case
is,
assume
weak.
Terms
return to the
trade affect spending. That
is
is
to
advantage are traded. As
with strong comparative
model, trade integration
which comparative advantage
We
is
weaker comparative advantage
with
2.3 Spending and the
on the characteristics of
shows what can happen
Figure 3
if
the effects of changes
on spending are strong. The top panel shows the case
AD
high that the
3es ^
3cOs
-X
1-T
of trade
is
spending
schedule becomes upward sloping
8X3
10
When Home's
productivity
even larger and, as a
shows the case
in
which
upward sloping
in
some
an improvement
in
aversion
risk
is
so
the region of the equilibrium,
in
high, the deterioration in the
so large than Home's share of labour income
is
which
terms
3(03
result,
in
risk
aversion
range,
is
falls.
However, the decline
in
the qualitative behavior of the terms of trade
and the trade balance remain the same as
case
is
in
the terms of trade
in
in
the
benchmark case. The bottom panel
so low that the
—^ < —^
.
That
is,
TB schedule becomes
the Marshall-Lerner condition that
the terms of trade worsens the trade balance
fails.
This
is
the only
which the qualitative behavior of the terms of trade and the trade balance are
different than in the
benchmark case. When Home's
deterioration in the terms of trade leads to
increase
in
Figure
the
2,
labour income.
the low state.
an increase
The opposite occurs when
economy now runs
An
trade deficits
in
implication of this discussion
Marshall-Lerner condition
productivity
fails,
the
volatility of
in
is
high, the
spending that exceeds the
productivity
is
low. Unlike
the high state and trade surpluses
is that,
unless y
is
the trade balance
In
so small that the
is
increasing
in risk
aversion.
In this
relationship
weakens
more general model, the
size of the traded sector affects the
between spending and changes
this relationship
terms of trade.
In
the
limit
spending regardless of
in
the terms of trade. Trade integration
as the real exchange rate becomes less sensitive to the
as
x^1 changes
,
risk aversion.
This
is
in
a
the terms of trade have no effects on
new channel through which
integration affects the volatility of the trade balance.
If
trade
the Marshall-Lerner condition
holds, through this channel trade integration increases the volatility of the trade
^°
The
AD
schedule might be increasing
analysis of Equation (11) shows that the
as a result, the equilibrium is unique.
in
some ranges and decreasing
AD
in
schedule always crosses the
19
an
schedule from below and,
others. Despite this,
AS
balance
if
aversion
risk
Lerner condition
fails,
balance through
this
If
condition
fails, this
low, but lowers
it
if
by
illustrates this
is
high.
volatility of
is
new channel might be
strong
enough
to
generate a negative
of the trade balance. Figure 4
volatility
plotting the volatility of the trade
values of y that do not depart
lowers the
some ranges
volatility of
The
is
much
much from
balance against the size of the
in
if
which an increase
aversion
risk
in
in
simpler to analyze. Naturally,
risk
if
low that the
Figure 2
is
so high that the
applies. This
benchmark case
is
is
shown
quantitative.
reinforced by improvements
i.e.
—^< —^<
shocks
in
If
risk
aversion
is
trade balance
increasing
relationship holds at
all
in
^, the
1-x
Figure 2
3cOs
still
applies. But
sloping, the qualitative description
5.
aversion
The only
is
high, increases in
benchmark case
in risk
An
in
difference with the
spending are
the terms of trade and this increases the
low, the opposite applies.
is
3(JL)s
schedule becomes upward sloping or so
Figure
in
the trade balance with respect to the
risk
AD
TB schedule becomes upward
still
extreme,
the effects of terms of trade changes on
of spending are not too strong,
aversion
sufficiently
which fluctuations are driven only by demand shocks
qualitative description of the effects of supply
if
is
the size of the traded sector
3cOs
even
that at
the trade balance.
other polar case
Home's share
we find
one, trade integration monotonically
of the trade balance. But
volatility
there might be
the trade
high enough, or else sufficiently low that the Marshall-Lerner
traded sector for different values of risk aversion. Not surprisingly,
increases the
the Marshall-
If
channel.
between trade integration and the
relationship
aversion
risk
trade integration always lowers the
aversion
risk
is
volatility of
of logarithmic preferences.
implication
is
If
that the volatility of the
aversion. Unlike the case of supply shocks, this
levels of risk aversion.
As we have already discussed,
integration has the additional effect of
in
the
more general case
weakening the
20
of this section trade
effects of the
terms of trade on
spending.
In
the
increases the
aversion
volatility of
high.
is
becomes
economy
strong
It
is
with
conceivable that
enough
to create a
the trade balance.
example
which
the
volatility of
this
Result #3:
We
in
in
is
is
high
low, but lowers
enough
6,
risk
it if
this effect
which trade integration reduces the
have however been unable
which
y- In all
is
to find a numerical
analogous to Figure
cases, the relationship
4,
shows how
risk
aversion
is
monotonically increasing.
can now be summarized as follows:
is sufficiently
higher or sufficiently lower than the
benchmark case of logarithmic preferences,
reduction
aversion
aversion
risk
range
happens. Figure
results of this section
If
if
risk
if
the trade balance changes as the size of the traded sector increases
for different values of
The
shocks, through this channel trade integration
the trade balance
volatility of
in
demand
trade integration might lead to a
the volatility of the trade balance.
Figure 4 provides examples that prove this claim.
To sum
up. Result #1 provides theoretical support for the view that trade
integration raises the volatility of the trade balance. Results
view by showing what can go wrong
violated.
3.
An
if
the two conditions stated
These cases however do not seem
Application to Trade
Our
objective
in this
section
is
#2 and #3
likely to
in
in
qualify this
Result #1 are
be empirically relevant.
Services
to develop a
sense
of the quantitative
importance of the effects of trade integration on the trade balance. To achieve
goal,
we
calibrate the
model and use
it
to study the effects of
21
an increase
in
the
this
While further trade integration
tradeability of services."
different industries,
we
think that the
example
is lil<ely
of services
given the glaring mismatch between the share of services
share
in
is
to occur in
many
particularly interesting
in
production and their
international trade, in industrial countries, the service sector accounts for
almost 70 percent of value added but only 20 percent of exports and imports. To a
large extent, this bias
in
trade flows
induced barriers to trade
change
in
is
in ser\/ices.
the near future.
the result of both technological and policy-
There are signs however that
We interpret the two-country model
in
section
1
as describing the United
States (Home) and the rest of the O.E.C.D. countries (Foreign).
model using available data and then consider two scenarios.
increase the share of traded goods
in
we
In
We calibrate the
the
first
services to half of that observed
In
the second one,
observed
in
the rest of the economy.
one,
in
increase the share of traded goods
the economy.
that
this is likely to
^^
we
the rest of
in
services to
3.1 Calibration
To
calibrate the model,
we need
three pieces of information: (1) the size of the
traded sector; (2) the relative technology schedule; and (3) the distribution of shocks.
^^
Throughout
utilities,
this section,
wholesale and
health, education,
we
retail
will
trade,
use the term services
FIRE
to refer to transportation,
(finance, insurance
and
real estate),
communication,
and other services (notably
and other professional services).
As the textbook example of haircuts suggests, many services are inherently more difficult to transport
than manufactures and commodities. Services also tend to be more vulnerable to a wide variety of nonbarriers to trade, such as professional licensing requirements that discriminate against foreigners,
domestic content requirements in public procurement, or poor protection of intellectual property rights.
But this seems to be changing rapidly. The last decade has brought a series of technological
improvements that are making many services increasingly tradeable. As a result of advances in
telecommunications technology, outsourcing abroad of computer programming, data entry, and call
center services is becoming common practice. With the appearance of e-commerce, wholesale/retail
sales and brokerage services can now be offered worldwide online. And the development of new
software has raised the ability of architectural, engineering and other types of consulting firms to better
interact around the globe. But this is not all. Recent multilateral negotiations under the Wodd Trade
Organization's General Agreement on Trade in Services have made substantial progress towards
dismantling a wide array of policy-induced barriers to trade in services. The harmonization of rules and
regulations within the European Union has also contributed to this process.
tariff
22
For reasons of data
how
discuss
Item
availability,
to obtain
The
1
economy as
Our
as the reference year, and
objective here
is
to
the U.S.
This
is
in
the
O.E.C.D.
sum
in
produced either
is
let
at
home
is
is
X/(1-v).
on U.S. exportables plus U.S. spending on these
simply X, and since
equally across goods, the latter
or
v be the share of
1997. O.E.C.D. spending on U.S. exportables
of foreign spending
The former
exportables.
is
We do so by taking seriously
as overall U.S. exports and imports, and
GDP
are traded
determine the size of the traded sector
a whole, and at the industry level.
X and M
some goods
In all industries,
:
the theoretical implication that every traded good
abroad. Define
we
of these three items in turn.
size of the traded sector
while others are not.
for the
each
we choose 1997
all
countries distribute their spending
simply Xv/(1-v). Following the
same argument, we
can calibrate O.E.C.D. spending on U.S. importables as M/v. This means that the
traded sector of the O.E.C.D.
U.S.,
we
is
X/(1-v)+M/v.
in
O.E.C.D. production
the non-traded sector of the O.E.C.D.
X, and
obtain the nontraded sector
in
the
simply take gross output, G, and subtract the traded sector, X/(1-v). Under
the assumption that the U.S. share
GDP,
To
M from the
is
the
(G-X/(1-v))(1-v)/v.
1997 U.S. input-output table
sectors and find that the share of traded goods,
To
is
to
same as
We
economy. The
first
X and M
column of Table
1
for
33
x, is
industries. In services,
3-digit industries
we
^^
repeat the
spanning the entire
reports the share of tradeables
in
in
gross output
the share of traded
which account for 62 percent of production,
tradeables represent only 3 percent of production.
(primarily manufacturing
in
use data on G,
roughly 10 percent.
or production by industry. There are substantial differences
goods across
share
compute the traded and nontraded
obtain the size of the traded sector industry-by-industry,
procedure using data on G,
its
In
the rest of the
economy
which accounts for 28 percent of production), tradeables
represent 22 percent of production.
"
Note that
we
are expressing tradeables as a fraction of production, rather than value added. For the
in value added in 1997 is 0.23.
U.S., the share of exports plus imports
23
The
Item 2.
technology schedule
relative
technologies to match data on trade
in
goods and
impose some additional structure on the data.
relative productivities within
that
it
well
is
come up
variance
(a,)
data to determine
ii,
for
5'^
we
industry
means
that our calibration
each
goods
industry. This
in
each
we do
industries
not
mean
(^ij)
and use the trade
assumption means that we
restrict
the
the 95"^ percentile to be roughly 5 times that of goods
does not seem unreasonable.
industry. This
find that the results are robust to sensible
a*(z)
'^
;
and
have enough
changes
in
the value of o.
Define the relative technology schedule for the traded sector of industry
as A|(z) =
to
treat the distribution of
with two parameters per industry, namely, the
percentile within
Moreover,
we
distribution. This
We therefore set 0,-0-0.5 for all
relative productivity of
the
In particular,
in
To do so we need
services.^"
of log productivity differences. Unfortunately,
information to do so.
in
We calibrate differences
each of the 33 industries as unobservable, but assume
approximated by a lognormal
procedure must
:
and order goods by using the
rule that z<z'
if
and only
i
if
ai(z)
a*(z)
a*(z')
—- Assume that,
>-^
.
ai(z)
within
an industry, the
distributions of log productivities
ai(z')
in
the traded and nontraded sectors are identical. Then, our assumptions imply that
In
Ai(z)
~
N(|j,|,a)
for industry
average the share of exports
i.
in
To
obtain the value of
X|
=
for
each
the traded sector of industry
share of traded goods for which A|(z) >
P[A|(z)>co]=X| where
ix,
co
.
i
industry, note that
on
corresponds to the
We therefore choose \m to ensure that
—
.
Inverting this probability gives:
X|/(1-v) + Mi/v
Could we have estimated productivity differences directly? A vast number of papers report crosscountry estimates of productivity levels. However, only very few provide disaggregated productivity
comparisons based on disaggregated purchasing power parity adjustments for inputs and outputs, which
are essential for meaningful productivity level comparisons (see Harrigan (1999) for a review). Because
of the difficulties in collecting disaggregated price data, these papers focus on only a subset of industries
(for example, Harrigan (1999) reports estimates for machinery and equipment manufactuhng productivity
levels across several O.E.C.D. countries). But without information on productivity levels comparisons for
all sectors and for all potential trading partners, it is impossible to construct a relative technology
schedule.
24
Hi=ln(co)-a*-\l-Xi)
(14)
where
denotes the cumulative normal
0(.)
we need
procedure,
relative
wages,
average
relative
wages by equating
O.E.C.D. income with the observed 40 percent U.S. share
We interpret differences in average
capital
between the U.S. and the
on years of
wage
of
co
total
this
1
GDP
in
1997.
human
in
.
number
productivities, ranging
average
O.E.C.D.
and measure them usingdata
rest of the O.E.C.D.,
at
hand, and the assumption that a=0.5,
|j.iS.
The
results are
There are large differences across sectors
.
in
the U.S. share of
productivity as reflecting differences
(14) to obtain a set of estimates for the
of Table
this
education. ^^ This leads us to an estimate of the productivity-adjusted
= 1 .33
With
To implement
information on the average values of productivity-adjusted
We obtain
co.
distribution function.
in
we use
shown
calibrated
in
Equation
the third column
mean
relative
from 0.47 to 2.68. By construction, these differences
relative productivities reflect differences
share of tradeable production (reported
in
exports as a
the second column of Table
in
the most noticeable feature of column 3
across industries
is
in
Perhaps
1).
again the difference between services and
the rest of the economy. Although tradeables as a fraction of services production
quite small, exports as a share of tradeables
rest of the
economy
productivity
is
much
larger
in
services than
in
is
the
(79 percent versus 30 percent). This implies that average relative
must be substantially higher
in
services than
in
the rest of the
economy
(2.54 versus 1.31).
Given values of
i^i
obtained
in this
way,
we can
construct the empirical analog
of the inverse of the relative technology schedule as a weighted
average of the
industry distributions of relative productivities:
^^
Specifically
we use
the Barro-Lee (2000) data on human capital stocks to find that average total years
then
and in the rest of the O.E.C.D. in 1995 are 12.2 and 8.3, respectively.
We
of education in the U.S.
assume
a Mincer coefficient of 0.1
,
and adjust
relative
25
wages by
a factor of e
"
=
1
.07
.
x(co;t)
(15)
=
^
^~^ ^-^
= Y.h-
'-
i=i
where
denotes the share of industry
t;
procedure are depicted
in
the 95*^ and
in
The schedules
the overall traded sector.
in
5'^
in
productivities.
percentiles
22 percent,
assume
respectively.
way
to the level of the trade
scenario). This
for
the rest of the
in
in
is
The
around
goods
in
7.5.
share
turn increases
tj
in
each industry within services, we
i
that industry increases to half the level of
economy
(in
the
first
scenario),
economy
the rest of the
(the share of industry
each of the industries within services,
in
calibrations indicate that
in
in
technology schedule
schedule was not possible
case
of
Item
to
match the
means
that
remember that our
in
is
substantially higher
our example, trade integration
in
such a rightward
the examples
in
shift in
the
result,
in
is
the
the relative technology
Section 2 where
we
restricted attention
symmetric technology differences across countries.
3.
The
distribution of
shocks
:
We calibrate shocks to demand
cyclical properties of the U.S. trade balance.
know the value
of the risk aversion coefficient,
y.
It
is
section 2 that this coefficient plays an important role
trade balance.
the overall traded sector)
and becomes steeper than
shifts to the right
that
the
all
the second
which the U.S. has comparative advantage. As a
benchmark case. Note also
to the
in
relative productivity in services
the rest of the economy. This
biased towards goods
relative
US
i
(in
and increases
Equation (15). To understand the effects
of trade integration on the relative technology schedule,
than
ratio of productivities
services increases from 2.5 to 10, and
in
In particular, for
that the share of traded
the trade share
results of this
labeled "Scenario 1" and "Scenario 2" correspond to our
assumptions that the share of traded goods
to
The
Figure 7 under the label "baseline". Note that there are
substantial cross-country differences
between goods
i
In
To do
this,
we
and supply
first
need
clear from the discussion
determining the
to
in
volatility of
the
the absence of strong priors on the magnitude of this parameter,
we
26
in
consider scenarios
in
there are two equally
(|)"\5"Y
As
shocks
to supply
in
Section
which y tal<es the values
likely
states of nature s=H,L;
the parameters
2,
cyclical properties of the U.S. trade
and
(2) its
(\)
with
in
and
We then
5.
which
assume
((l)H,5H)=((t),S)
and
that
((t)L,5L)-(
We then choose
(\>
of
and 6 to match the
balance as a share of O.E.C.D. income over the
summarized by
comovement
1
and 5 regulate the standard deviation
and demand, respectively.
period 1970-1999, as
0.5,
(1) its
standard deviation, which
income measured as the slope
0.5 percent;
is
of a regression of the
U.S. trade balance on U.S. income (both as a share of O.E.C.D. income), which
0.3.
We do this for each
that
we
the
-
of the three values of the coefficient of relative risk aversion
consider.
The
In
is
results of this
benchmark case
procedure are presented
in
the
of y=1, a combination of supply
deviation of 2 percent and
demand shocks
first
two columns of Table
2.
shocks with a standard
with a standard deviation of 10 percent
are required to match the cyclical properties of the trade balance. Since the U.S.
trade balance tends to decline
when incomes
shocks and small supply shocks
difference between
and 5 percent,
in
are high,
If
require large
order to match the data,
demand and supply shocks
respectively).
we
y=0.5,
we need
(with
if
y=5,
demand
we need
a smaller
standard deviations of 8 percent
a larger differences
in
the
volatility of
demand and
supply shocks (with standard deviations of 17 percent and 3 percent
respectively)
in
order to match the observed cyclical properties of the trade balance.
To understand these
differences,
remember
that fluctuations in the
magnify (dampen) the effects of demand shocks
smaller
demand shocks
the larger
if
y>1 (y<1)- This
terms of trade
is
why we need
is y-
3.2 Results
The remaining columns
of
Table 2 summarize the result of our trade
integration exercise, reporting the predicted standard deviation of the U.S. trade
27
balance as a fraction of O.E.C.D. income
integration scenarios discussed above.
trade balance
is
equal to
three rows of the
first
in tine
By
baseline 1997 scenario, and the two
construction, the standard deviation of the
observed value of 0.5 percent of O.E.C.D. income
its
column. Moving to the
right illustrates
in all
the effects of goods and
services trade liberalization on asset trade.
In
benchmark case
the
on the
integration
volatility of
of y=1
,
we
find quite substantial effects of further trade
the trade balance, which rises from 0.5 percent to 0.7
percent
when
services are half as tradeable as the rest of the economy, and to 0.9
percent
when
services are equally tradeable than the rest of the economy. This
suggests that the main effect of trade integration on the
summarized
Result #1
in
is
On
quantitatively important.
volatility of
the trade balance
the other hand. Result #2
which qualifies our main result does not appear to be empirically very important.
Although
in this
example trade
has comparative advantage,
To
we find
isolate the effects of this bias,
assumption that the
baseline scenario.
balance
is
relative
In this
biased towards goods
Y=5
we
which the U.S.
technology schedule does not change relative to the
case,
we
find that the
in
standard deviation of the trade
we
allow for a bias
show the
in
trade integration.
effects of trade integration
that the effects of trade integration
integration scenario.
To understand
on the standard deviation of the
this difference,
fluctuations in the terms of trade induce fluctuations
As
fluctuations
in
the trade balance.
fluctuations
in
the terms of trade on spending
^^
Remember
we have
when
the terms of trade also affect spending.^®
trade balance are substantially smaller, with the latter rising to only 0.6 percent
second
trivial.
re-estimate Scenario 2, but do so under the
the rows with y=5 and y=0.5
we find
in
that the quantitative effects of this bias are
allow for the possibility that changes
When
is
0.95%, as opposed to 0.94% when
Finally,
we
integration
in
in
the
remember that when y>1
spending which amplify
trade integration proceeds, the effects of
fall,
and so the
volatility of
the trade
shocks in each row of Table 2 to replicate
changes in the volatility of the trade balance
aross different values of y for a given level of trade integration reflect changes in our assumptions about
both (i) the shocks which drive fluctuations, and (ii) the degree of risk aversion.
that
observed fluctuations
in
re-calibrated the underlying
the trade balance.
As
a result,
28
,
balance increases by less than
much
the benchmark case of y=1
in
larger effect of trade integration
latter rising to
0.9 percent
intuition for this is just
in
the
first
on the
volatility of
.
When y==0.5, we find
a
the trade balance, with the
scenario, and 1.2 percent
in
the second.
The
the converse of the case where y=5: through their effects on
spending, fluctuations
in
the terms of trade attenuate fluctuations
in
the trade
balance, and the importance of this attenuation declines with trade integration.
To sum
integration
example suggests
up, our empirical
on the
trade balance can be substantial, with the
volatility of
the trade balance almost doubling
Although
in
that the effects of trade
in
volatility
of
the benchmark case of log preferences.
our example trade integration
biased towards goods
is
has comparative advantage, the quantitative effects of
this bias
in
which the U.S.
are negligible.
Departures from the assumption of log preferences do affect the quantitative effects
of trade integration, with larger (smaller) increases in the volatility of the trade balance
when
risk
risk
aversion
low (high). However,
in
our calibrations the additional effects of
aversion are never sufficiently strong to reverse our qualitative conclusion that
trade integration
4.
is
will
increase the
volatility of
the trade balance.
Concluding Rennarks
In this
paper,
we have used
analyze the question of
contribution
will
how
a simple model of trade
in
goods and assets
trade integration affects the trade balance.
induce others to explore
this issue
We hope our
from alternative theoretical
perspectives.
We
differences
technology and imperfectly correlated shocks. But one could for
in
have motivated trade
instance allow for differences
in
as an additional motive for trade
in
factor
in
in
to
goods and assets by postulating
endowments and/or
increasing returns to scale
goods. Similarly, one might introduce differences
time preference and/or rates of return to capital as an additional motive to trade
assets.
We conjecture that the
main
results of this
29
paper would survive
in
in
these more
general frameworks, except for extreme cases. The interesting question
new and
would arise that cannot be studied
realistic effects
Finally,
we hope
whether
our simple framework.
in
good
that our contribution also provides a
is
theoretical
grounding to empirical studies of the effects of trade integration. By isolating key
theoretical channels, the
studies on this subject.
model here can sharpen the
By
interpretation of
providing a fully specified model,
we
also
econometric
hope
to aid in the
task of developing quantitative assessments of the effects of trade integration on the
trade balance.
simple
In this
illustration
integration
respect, our calibration exercise should
or example.
A serious
be interpreted only as a
quantitative analysis of the effects of trade
on the trade balance would be a major project on
its
own.
References
Baler, S.L. and J.H. Bergstrand [2001], "The Growth of World Trade: Tariffs,
Transport Costs and Income Similarity," Journal of International Economics 53,
.
1, 1-
27.
and J. Lee [2000], "International Data on Educational Attainment:
Updates and Implications," Harvard University Center for International Development
Working Paper No. 42.
Barro, R.J.
Cole, H. and M. Obstfeld [1991], "Commodity Trade and International Risk Sharing:
How Much Financial Markets Matter?" Journal of Monetary Economics 28, 1, 3-24.
.
J. [1999], "Estimation of Cross-Country Differences
Functions," Journal of International Economics 47, 267-293.
Harrigan,
in
Industry Production
.
Dornbusch, R., S. Fischer and P. Samuelson [1977], "Comparative Advantage, Trade
and Payments in a Ricardian Model with a Continuum of Goods," American Economic
Review 67, 823-839.
,
30
Table
1
:
Trade and Relative Productivity by Industry
Averaae
Tradeables/
Exports/
Relative
Production
Tradeables
Productivitvd)
Agriculture
0.13
0.41
1.34
Metallic ores mining
0.21
0.31
1.18
Coal mining
0.09
0.84
2.46
Crude petroleum and natural gas
0.38
0.03
0.60
Nonmetallic minerals mining
0.11
0.29
1.15
Construction
0.00
0.00
0.47
Manufacturing
Food
Tobacco
0.10
0.39
1.32
0.15
0.74
2.08
Textiles
0.17
0.32
1.19
Apparel
0.44
0.09
0.78
Lumber and wood products
Furniture and fixtures
0.15
0.22
1.03
0.21
0.18
0.96
Paper
0.16
0.36
1.25
Printing
0.05
0.49
1.49
Chemicals
Petroleum Refining
Rubber and Plastics
0.26
0.39
1.30
0.11
0.33
1.22
0.17
0.28
1.13
Leather
0.75
0.06
0.71
Nonmetal Production
0.16
0.23
1.04
Primary Metals
0.22
0.22
1.02
Fabricated metals
0.13
0.31
1.18
0.39
0.43
1.37
0.50
0.31
1.18
Motor Vehicles
0.37
0.16
0.92
Other Transportation Equipment
0.44
0.51
1.53
Other Manufacturing
0.40
0.30
1.16
Transportation Services
0.12
0.77
2.17
Communications
0.00
0.00
0.89
Electric Utilities
0.01
0.15
0.89
Gas
0.00
0.00
0.89
Trade
0.05
0.71
2.00
FIRE
0.02
0.88
2.68
Other Services
0.01
0.77
2.19
0.10
0.37
1.49
0.22
0.30
1.31
0.03
0.79
2.54
Agriculture and Mining
Construction
Industrial
Electrical
Machinery
machinery
Services
Utilities
Weighted Averages
(2)
Overall
Rest of
Economy
Services
Notes:
(1)
This column reports E[Aj] = e"'*°
(2)
The
first
column uses shares
in total
production as weights; the remaining columns uses shares
tradeables production as weights.
31
in
Table
2:
standard Deviation of
Calibration
Res ults
US Trade Balance/OECD GDP
After
1
^
5
0.5
1.03
1.17
1
1.02
1.10
5
1.05
1.08
Before
Scenario
0.50%
0.50%
0.50%
0.91%
0.74%
0.58%
1
Scenario 2
1.24%
0.94%
0.60%
Notes:
(1)
To obtain the standard
deviation of the trade balance as a fraction of U.S.
32
GDP simply
multiply by 2.5.
Figure
1
:
Effects of Trade Integration
Case
2
on Relative Technology Schedule
Trade Integration Weakens Comparative Advantage
1:
n
High
\
.t--"^
>v
\low t
x'^
b\^^ \
^^^"-s,^\^
3
1?
1
-
<
^^"^"^^^r-^B.
N
0.5
1
X
Case
2:
Trade Integration Strengthens Comparative Advantage
2Low t\.
^^-'^^
\ High
T
XA\_^^\bNw
^"^-^^S.
3
1?
X
<
1
-
^^X;:;;^.^^^
^Vv^^^~~--~-,.^A'
^"\>\
>
0.5
X
33
1
Figure
2:
Effects of Supply and
Demand Shocks,
y=^
Supply Shocks
TB
CO
fASH
El
Eh
Trade Balance
AS,
AD
AS,
AD
Demand Shocks
ADh
TBh
/
Trade Balance
Note:
These
x=0.5; 4)H=1 .5
figures are
in
drawn under the assumption
the top panel and 5h=1
.5 in
that x(co)=co"7(1+co") with a=0.2;
the bottom panel;
34
AS
and y=1
.
Figure
3:
Effects of Supply
Shocks
Y»1
ASH
AS,
Trade Balance
AD
Y«1
\
(1)
^^
Ypr
El\
/
Eh\
/y
Note:
These
(t)H=1-5;
figures are
and y=5
/ASH
/
X^"
\ AD
AS,
Trade Balance
T=0.5;
Z^^'"
\
in
AD
drawn under the assumption that x(co)=co7(1+co"°) with a=0.2,
the top panel and y=0.2 in the bottom panel.
35
Figure 4: Effects of Trade Integration with Supply Shocks
0.03
Note: This figure
(t)H=1 .5;
and
is
drawn under the assumption that
for the indicated values of
y.
36
x(co)=co 7(1 +co") with
a=1; t=0.5;
Figure
5:
Effects of
Demand Shocks
Y»1
TBh
CO
AS,
Trade Balance
TBii
\tBh
AS,
Trade Balance
Note:
These
T=0.5; 5h=1
AD
drawn under the assumption that x(co)=co""/(1+(jl)'") with a=0.2;
and y=5 in the top panel and y=0.2 in the bottom panel.
figures are
.5;
AD
37
Figure
0.05
*•*
0.04
6:
Effects of Trade Integration with
Demand Shocks
^
n
-
4-
o
o d CO
Deviation
ard
o d K3
c
0.01
-
n
'
0.4
0.2
0.6
0.8
1
T
Note: This figure
6h=1
.5;
and
is
drawn under
tiie
assumption that
for the indicated values of
y-
38
x(co)=co""/(1+co"") with
a=1 T=0.5;
Figure
5
7:
Estimated Relative Technology Schedule
n
yi
4.5
-
4
-
3.5
-
1
\
—
"^
Baseline
—
Scenario
~-»
Scenario 2
1
3
.
'P
<
2
-
1.5
-
1
-
^--C>-:^^^.
'
0.5
i~^iii""
-
N
U
3
(
0.1
0.2
0.3
0.4
0.5
X
39
0.6
0.7
0.8
0.9
1
Date Due
^1
P^lo^
MIT LIBRARIES
3 9080 02246 0981