The Purpose of Trade Agreements
Gene M. Grossman
Princeton University
March 2016
Abstract
This paper reviews the literature on governments’motivations for negotiating and joining
international trade agreements. I discuss both normative explanations for trade agreements and
explanations based on political-economy concerns. Most of the paper focuses on the purpose of
multilateral agreements, bu t I do discus s brie‡y the reasons we might see governments forming
preferential or regional trade agreements that exclude some countries.
Keywords:
Trade agreements, trade pacts, international cooperation, multilateralism, regionalism
JEL Classi…cation:
F13, F53, K33
This paper was prepared for the Handbook of Commercial Policy, edited by Kyle Bagwell and Robert Staiger.
I am grateful to Da niel Goetz for excellent research assistance and to the editors fo r their comments on an earlier
draft. I am a lso gratef ul to Henrik Horn, with whom I have discussed many of these issues before; se e Gros sman and
Horn (2013).
1 Introduction
Britain and France inked the …rst modern trade agreement on January 23, 1860. The so-called
Cobden-Chevalier Treaty promised that France would eliminate all import prohibitions on British
manufactured goods while capping most duties at 30 percent (25 percent after 1865). Britain in
turn agreed to remove import b arriers entirely from all but 48 French commodities while reducing
dramatically its tari¤s on French wine and brandy (Ashley, 1904). Notably, each country p romised
to grant the other most-favored-nation (MFN) consideration with regard to any tari¤ concessions
it might subsequently grant to other trading partners. There followed a veritable explosion of
bilateral trade pacts, with an additional 56 treaties having been signed within …fteen years. By
1875, virtually all of Europ e was party to a low-tari¤ zone by dint of a web of agreements that
included the linchpin MFN clause.
When the General Agreeme nt on Tari¤s and Trade (GATT) went into e¤ect on January 1,
1948, it marked the …rst of a sequence of multilateral trade agreements. The GATT incorporated
more than 45,000 tari¤ c once ssion s by its original 23 signatories, while also providing a b roader
framework for regulating international trade. Seven subsequent “rounds” of negotiations by these
and additional participants led to innumerable further tari¤ cuts and to the introduction of rules
governing various non-tari¤ barriers to trade. The Uruguay Round, which was signed by 123
“contracting parties” and took e¤ect on January 1, 1995, created the World Trade Organization
(WTO), while also extending trade rules to many services, harmonizing treatment of intellectual
property, and establishing procedures for dispute settlement. By November 30, 2015, the WTO
had grown to include 162 members that together conduct more than 96 percent of world trade.
1
Meanwhile, the multilateral agreement lives side-by-side with 267 di¤erent bilateral and regional
trade agreements that the WTO reported to be in e¤ect as of February 1, 2016.
2
This chapter reviews the economics literature that poses the question, Why do countries nego-
tiate and accede to international trade agreements? The chapter focuses mostly on the motivation
for multilateral agreements, in orde r to minimize overlap with Chapter 14 by Limão (2016) on pref-
erential agreements. However, Section 5 does cover some literature that addresses the incentives
countries have to negotiate bilateral or regional agreements alongside (or instead of) multilateral
agreements in a many-country world. Also, this chapter focuses on the broad purpose of trade
agreements, leaving discussion of their design for Chapter 8 by Bagwell and Staiger (2016).
3
I
do consider both the incentives that large countries have to create trade pacts de novo and the
incentives that small countries have to sign existing agreements.
The chap ter is organized as follows. Se ction 2 discusses research that sees trade agreements
as addressing international externalities that arise in competitive economic environments. It in-
cludes the case of both welfare-maximizing governments and politically-motivated governments,
1
Current members are listed at http://www.wto.org/english/thewto_e/whatis_e/tif_e/org6_e.htm. Trade cov-
erage of WTO members is reported at http://www.wto.org/english/thewto_e/acc_e/cbt_course_e/c1s1p1_e.htm.
2
See http://www. wto.org/english/tratop_e/region_e/region_e.htm.
3
This chapter covers some of the same ground as the excellent survey by Maggi (2014) in the Handbook of
International Economics, vol. 4.
1
both in situations with well-functioning markets and with distorted markets. One controversy in
the literature concerns whether the sole motivation for trade agreements in competitive markets is
to eliminate the temptation governments have to manipulate their terms of trade. I discuss this
debate, concluding that it is more a matter of semantics than substance. In Section 3, I review
the literature that identi…es the di¤erent types of international externalities that can arise in im-
perfectly competitive market environments. These externalities re‡ect governments’incentives to
in‡uence …rm location, to shift or e xtract pro…ts away from foreign monopolists or oligopolists, or to
a¤ect imperfect contracting in international outsou rcing relationships. I brie‡y discuss agreements
to protect intellectual property that are motivated by externalities in the international innovation
process.
In Section 4, I turn to an alternative purpose that has been suggested for trade agreements,
namely to aid governments in committing not to intervene in favor of domestic special interests. The
commitment motive arises when optimal policies are not time consistent; that is, when governments
know they would be tempted to adjust p olicies ex post away from the levels that they prefer ex
ante. I argue that commitment is unlikely to be the reason that two governments will sit d own
to negotiate a trade agreement, bu t that it might very well explain why some countries accede to
existing agreements.
Finally, in Section 5, I turn from the purpose of multilateral agreements to that of regional
and bilateral agreements. In this section I discuss only research that bears on the motivation that
governments have to negotiate preferential agreements in addition to— or instead of— multilateral
agreements, in a many-country world. Preferential agreements may serve to promote allocative
e¢ ciency among signatories, to improve members’terms of trade vis-à-vis nonmembers, to transfer
rents to special interests via trade diversion, or to facilitate a dynamic process of multilateral trade
liberalization.
2 International Externalities from Unilateral Trade Policies
Trade treaties are a formal expression of intergovernmental cooperation. Governments relinquish
their sovereign rights to choose their own trade (and other) policies in exchange for similar con-
cessions by others. Why might a government be willing to compromise its sovereignty? In a word,
the answer is interdependence. The policies imposed by any government a¤ect the well-being not
only of its own citizens, but also of those in other countries. No matter what the objectives of
the policy makers— be they benevolent, autocratic, or politically motivated— each has an interest
in the choices mad e by its trading partners. With unilateral policy choices, governm ents may fail
to take into account the impact of their actions on interests abroad. A trade agreement provides
a means to internalize thes e externalities. Of course, to identify the incentives for concluding a
treaty, one must begin by identifying the nature of the potential externalities, that is, by predic ting
the trade policies that would prevail in the absence of cooperation.
2
2.1 Welfare-Maximizing Policy Makers
Harry Johnson (1953) was the …rst to analyze the strategic interdependence between countries’
tari¤-setting decisions. Johnson conceived of tari¤s as being the outcome of a static game played
by a pair of welfare-maximizing governments and he proceeded to provide an early application of
the then-novel concept of a Nash equilibrium.
Suppose initially that there are two countries and two goods. The goods are competitively
produced in each country by …rms that have access to strictly convex technologies. Suppose further
that aggregate welfare in each country can be represented by a strictly quasi-concave function of
the country’s aggregate consumption bundle. Let t be the ad valorem tari¤ rate imposed by the
home country on its import good. Let t
be the ad valorem rate imposed by the foreign country
on its respective import good, which of course is the home country’s export good. We can solve for
the competitive equilibrium as a function of t and t
and then write W (t; t
) and W
(t; t
) as the
resulting aggregate welfare levels in the home and foreign countries, respectively.
Figure 1 depicts the best-response functions of the two welfare-maximizing policy makers, with
t on the horizontal axis and t
on the vertical axis. The inverted u-shaped curves such as the
one labelled W W represent iso-welfare loci for the home country. These curves peak at the tari¤
rates that maximize W (t; t
), given the corresponding values of t
. The peaks generally fall in the
positive quadrant, because the Mill-Bickerdike argument for an optimum tari¤ implies that, for any
given foreign policy and economic conditions, home households can gain from a tari¤ that optimally
exploits the cou ntry’s monopoly power in trade.
4
The curve RR that connects the set of peaks is
the home country’s best-response function; i.e., a function that gives the home country’s optimal
tari¤ in response to an arbitrary level of the foreign tari¤, t
. Among any pair of iso-welfare loci
for the home country, the curve that lies above the other represents a lower level of home welfare
in view of the fact that a departure from the best response (a horizontal move to the right or to
the left of RR) must result in a welfare loss for this country.
Similarly, the right-parentheses-shaped curves such as W
W
represent iso-welfare loci for the
foreign country. These curves “peak” in the horizontal direction at the optimal tari¤s for the
foreign country given the corresponding rates of the home tari¤, t. The curve that conn ects these
peaks, R
R
, represents the foreign country’s best-response function. Among any pair of foreign
iso-welfare curves, that to the right represents the lower level of foreign welfare.
As Johnson …rst noted, a Nash equilibrium occurs at point E, where each government’s tari¤
choice is a best response to that of the other. At this point, neither government can raise aggregate
welfare by unilaterally altering its trade policy. The pair of Nash equilibrium tari¤s in the Johnson
equilibrium are both non-negative, except possibly in a perverse case such as that described in
footnote 4.
5
4
Actually, Kemp (1967) shows th at, for some values of t
, the best response by the home country might be an
import subsidy; i.e ., t < 0. T his can aris e o nly if the foreign o¤er curve is multi-valued , which in turn requires that
the home country’s import good is su¢ ciently inferior in the foreign c ountry’s preferences. For ease of exposition, I
wi ll negle ct this rather obscure possibility.
5
Fo r th e ta ri¤s in the Johnson equi librium to be non-negative, it is su¢ cient that dema nds in each country can
3
t*
t
R
W*
W
R*
R
W
W*
R*
E
Figure 1: Nash Equilibrium of a Tari¤-Setting Game
Figure 2 outlines a “len s”to the southwest of point E. At any point in this lens, the aggregate
welfare levels in both countries are higher than at point E. In other words, points in this set
represent pairs of tari¤ rates that are Pareto-preferred by the two welfare-maximizing governments
to the noncooperative outcome at E. A trade agreement— if one could be negotiated and enforced—
that achieves any pair of tari¤ rates in this lens would be one that both governments prefer to the
outcome that occurs without cooperation.
The basis for a trade agreement in this rather simple setting is the negative externality that
each government imposes on households in the other country when it imposes its optimal tari¤.
At point E, a small reduction in the home tari¤ has virtually no e¤ect on aggregate welfare in the
home cou ntry, because the optimal tari¤ just balances on the margin the positive terms-of-trade
gain with the negative volume-of-trade loss (see, for example, Dixit, 1985). Any redu ction in home
welfare generated by a small departure from the best response is second-order small. Meanwhile, a
reduction in the home tari¤ generates a …rst-order welfare gain for the foreign country inasmuch as it
improves that country’s terms of trade withou t generating any distortion of its resource allocation.
In other words, each country’s optimal tari¤ is a beggar-thy-neighbor policy that achieves gains for
its own citizens (holding constant the other’s policy) at the expense of citizens elsewhere and global
e¢ ciency. When governments behave unilaterally and noncooperatively, they ignore the harm that
their policies impose on citizens outside their borders. This creates the opportunity for mutually
bene…cial exchange of tari¤ cuts; points to the southwest of E have approximately the same terms
be derived from tho se of a representative agent and both goods are normal; see Bo nd (1990). Dixit (1987) points
out that there generally exists another Nash equilibri um, one with prohibitive tari¤s in both countri es. When one
country chooses a prohibitive tari¤, it is always a best response for the other to do likewise.
4
t*
t
E
C
C
Figure 2: E¢ cient Agreements
of trade as at E, but higher volumes of trade and a more e¢ cient allocation of the world’s resources.
The curve labelled CC in Figure 2 connects points of tangency between iso-welfare curves of
the home and foreign countries. At such points, it is not possible to raise aggregate welfare for
one country without reducing it for the other; i.e., CC is a locus of Pareto-e¢ cient tari¤ rates. As
Mayer (1981) pointed out, the tari¤ rates on this curve satisfy
(1 + t) (1 + t
) = 1, (1)
because the relative price of the home import good in the home country is (1 + t) p
w
, where p
w
is the “world” relative price of this good, while th e relative price of the home import good in the
foreign country is p
w
= (1 + t
).
6
A necessary and su¢ cient condition for global e¢ ciency is that
the internal relative prices in the two countries are the same, s ince this ensures equality worldwide
of marginal rates of transformation and marginal rates of substitution.
The …gure also shows a portion of this curve in bold. The northwest endpoint of this bold
segment corresponds to a pair of tari¤ rates that yield the home country the same level of aggregate
welfare as in the Nash equilibrium at E. The southeast endpoint gives the foreign country the same
welfare as at E. Therefore, all points along the bold portion of CC yield a Pareto improvement
for the two ben evolent governments relative to the noncooperative outcome at E. Mayer (1981)
concludes that, with e¢ cient and costless bargaining, the governments should agree to some pair
of tari¤s along CC that leave each at least as well o¤ as at E. The particulars of the agreement
6
The world pr ice , p
w
, is the relative price at a …ctitious o¤shore port where goods are free of all trade taxes. A
tari¤ imposed by the hom e country raises the internal relative price of this good by a factor 1 + t, whereas a tari¤
imposed by the fo reign country (on the “other” good) r educes the inter nal relative price there by a factor 1 + t
.
5
will depend on the bargaining protocol and the countries’ relative negotiating pro…ciency. But,
in all cases, both tari¤ rates are lower under an e¢ cient agreement than at E; in other words,
the cooperative agreement entails trade liberalization (or promotion) by both countries.
7
The
noncooperative tari¤s are both positive in the Nash equilibrium as each country attempts to exploit
its monopoly power in beggar-thy-neighbor fashion. When the countries coop e rate fully, they will
choose either free trade, or else a positive tari¤ in one and a negative tari¤ in the other in order to
achieve the same allocation of resources as with free trade together with a transfer of government
revenue.
Some features of the Johnson (1953) and Mayer (1981) analyses are special to the two-good,
two-country setting that they consider. Graaf (1949) observed that, with many goods, the vector of
a country’s optimal tari¤s and export taxes could include negative eleme nts, i.e., some goods m ay
be subject to import or export subsidies. This conclusion carries over to the Nash equilibrium of a
tari¤-setting game, and so an e¢ cient trade agreement need not entail a reduction in all trade tax
rates. But Bond (1990) shows that if the foreign o¤er curve is monotone— such that an increase in
price reduces foreign imports on average— and if foreign excess demands can be derived from the
preferences of a representative agent for whom all goods are normal, then the optimal trade policy
for any country generates non-negative tari¤ revenue. Under these con ditions, a move to universal
free trade (which is always e¢ cient with perfect competition and no market d istortions) entails
an overall reduction of trade taxes and thus a liberalization of trade. Fukushima and Kim (1989)
provide conditions, later relaxed by Turunen-Red and Woodland (1991), under which an equipro-
portionate reduction in all (spe ci…c) trade taxes and subsidies (i.e., a radial movement toward zero)
must raise global welfare in a world with an arbitrary number of goods and subsidies. Under these
conditions, multilateral trade liberalization of th is sort shifts the world’s utility possibility frontier
outward, but international transfers of purchasing power might still be needed to ensure Pareto
gains for all countries.
2.2 Why a Formal Agreement?
Johnson assumed that, in the absence of any international treaty, policy makers would set their
country’s tari¤s once and for all at the noncooperative levels identi…ed by the static Nash equilib-
rium. In reality, of course, trade policy decisions are made repeatedly over time. This obs ervation
raises the question of whether an actual agreement is needed to achieve the gains from cooperation,
or whether a cooperative outcome might be achieved as an equilibrium (without explicit cooper-
ation) in a repeated tari¤-setting game. Dixit (1987) discusses the in…nitely-repeated tari¤ game
involving a pair of welfare-maximizing governments.
The “folk theorem”of repeated games [see Aumann (1985)] ensures that, if the policy makers’
discount rates are su¢ ciently low, any point that is Pareto preferred to point E can be sustained as
7
In t he …gur e, the free trade point at the origin gene rates a Pareto improvement, and all other points of possible
ag reement involve a positive tari¤ by one country and an import subsidy by the other. It is possible, however, tha t
one country will fare worse at the f ree trade point than at E. In s uch circumstances, a Pareto-improving agreement
requi res an impo rt subsidy in th e other country.
6
a subgame-perfect equilibrium in the in…nitely repeated game. Each government understands that
it is expected to set a given tari¤ rate repeatedly, such that the pair (t; t
) is preferred by each to E.
Each government further understands that any deviation by the other should invoke it to retaliate
by setting its Nash-equilibrium tari¤ forever afterward. The anticipated punishment is credible,
because if on e government is expected to play its Nash equilibrium tari¤ in every period, the best
the other can do is to respond similarly, and an inde…nite repetition of th e static equilibrium is
itself an equilibrium of the in…nitely-repeated game. With this expected punishment, a government
that contemplates a deviation will compare the one-period gain from behaving opportunistically
against the sustained loss that results from foregoing the Pareto gains from cooperation forever
afterward. With a small enough discount rate, the loss must dominate, and so any opportunistic
behavior is deterred.
8
It follows that, if discount rates are low enough, the points along the bold portion of CC in
Figure 2 can be achieved by tacit cooperation in a repeated, noncooperative game, without the need
for a formal treaty. So what role does a treaty play? One might think that a f ormal agreement
is needed to sustain high levels of cooperation (i.e., points on or close to CC) in situations where
the discount rates are not so low, so that governments would otherwise be tempted to behave
opportunistically in order to capture short-run gains. However, this answer is not compelling,
because if a given cooperative outcome cannot be sustained by self interest in a repeated game
with punishments, it is unclear how a trade agreement would solve the problem. Governments have
sovereign rights under international law and there is no higher authority to which the y can appeal
in case their trade partners behave opportunistically. Rather, a formal international agreement or
an international agency can at most threaten poor behavior with (credible) punishments. In this
way of thinking, a f ormal agreement can only achieve those outcomes that are sustainable in an
in…nitely repeated game.
One role that a formal agreement might play is that of coordination. The in…nitely-repeated
game has many equilibria involving di¤erent degrees of coope ration. Any one of them c an be
sustained as long as both policy makers know what is expected of them and stand ready to punish if
the other deviates from its tacitly-agreed behavior. How will the governments know and agree upon
what behavior is expected of them and what actions constitute deviations that warrant retaliation?
After all, there are many tari¤ rates that are consistent with some cooperative behavior along
some equilibrium path of the repeated game. And the governments’interests diverge with respect
to which of the many equilibria they would like to see played. A formal agreement can be used to
achieve a mutual understanding of what is expected, so that misunderstandings do not invoke a
return to noncooperative play.
Maggi (1999) explores another potential rationale for a formal institution such as the World
8
Bond and Park (2002) show that, if coo per ation via repeated play is intended to yield an as ymmetric divi sion o f
surplus relative to the Johnson equilibrium and if the discount rat e is not too high, then achie ving the Pareto frontier
of national welfare le vels require s a gradually declining rate of protection i n the countr y that enjoys the lion’s share of
the gains from cooperation. They o¤er this observation as an explanation for trade agreements t hat involve gradual
liberalization d espite a stationary economic environment.
7
Trade Organization (WTO). His argument explicitly takes into account that multilateral agreements
have more than two participants. Maggi argues that an institution such as th e WTO can play an
informational role by verifying violations of the agreement and by informing third countries when
they have occurred. By sharing information about deviations that take place in particular bilateral
relationships, the organization can facilitate multilateral punishments for opportunistic behavior
that collectively o¤er greater deterrence than bilateral punishments. Maggi has in mind policy
deviations that can be observed directly only by the injured parties (e.g., an exporter knows when
its market access has been restricted) but not by uninvolved, third parties. As he notes, import
tari¤s are su¢ ciently transparent that presumably rate hikes can be observed outside the particular
relationship. But other forms of non-tari¤ barriers to trade are more sub tle and more di¢ cult for
third countries to perceive.
Let us compare the co operation that can be achieved when a country can observe only the
barriers imposed on its own exports with the cooperation that can be achieved when an institution
such as the WTO publicizes to all members any violation of the agreement. Conside r a stylized
trading environment with three countries in which each pair of countries exchanges two good s that
are neither produced nor consumed in the third country. I n this setting, no country is directly
harmed by opportunistic behavior on the part of its trade partners in its bilateral relations with
the third country. Nevertheless, in some circumstances, every country can bene…t from an agree-
ment that calls for multilateral punishment for any deviation. Such punishments would prescribe
reversion to Nash equilibrium tari¤s in bilateral relations with the violating country not only by
the country that is directly harmed by its actions, but also by the third c ountry.
The key consideration for whether multilateral punishments facilitate greater cooperation than
do bilateral punishments alone is the existence of imbalances of power in the bilateral trading
relationships. Maggi s hows this point by …rst examining a case in wh ich all countries are symmetric
and each has balanced trade with both of its trading partners. In this case, multilateral punishments
are more severe than bilateral punishments, but the temptation to behave opportunistically also
is greater. The set of outcomes that can be su stained by multilateral punishments is the same as
that with b ilateral punishments. On the other hand, if each of the symmetric countries is a net
exporter in one of its bilateral relationships and a net importer in the other and if each can apply
import tari¤s but not export taxes, then each country will be more powerful than its partner in
the relationship in which it imports. The greater power re‡ects the fact that a net importer in
a given bilateral relationship stands to lose less from a reversion to Nash equilibrium tari¤s than
does the net exporter. Then, a regime that pub licizes violations and calls for punishments by third
parties is more e¤ective in deterring opportunistic behavior than one that entails punishments by
only directly injured parties. As Maggi notes, the WTO rules as currently construed do not allow
for explicit third-party punishments, but it is conceivable that such punishments occur implicitly
by a loss of goodwill. In any case, the paper shows that a multilateral institution that monitors
trade behaviors and publicizes violations could play a role in sustaining greater cooperation and
8
freer trade.
9
2.3 Trade Agreements among Politically-Minded Governments
When welfare-maximizing governme nts can act opportunistically in an e¤ort to improve their coun-
try’s terms of trade, the resulting international externalities open the potential for mutually advan-
tageous trade agreements. But governments’trade policies seem to be motivated as much or more
by distributional concerns as by c once rns about aggregate welfare. The assumption that govern-
ments maximize aggregate welfare is convenient but rather unrealistic in a world with asymmet-
rically informed voters and with campaign spending funded by contributions from special interest
groups. Policy makers who are interested in their own political success might be tempted to choose
protectionist policies not (or not only) to manipulate the terms of trade, but rather (or also) to re-
distribute income to swing voters in the electorate or to groups that o¤er campaign support. With
perhaps limited interest in aggregate welfare, would these politicians be willing to negotiate with
foreign counterparts? By doing so, wouldn’t they curtail their own use of a valuable instrument for
currying political favor?
These questions cry out for a political-economic theory of trade agreements. Hillman and Moser
(1995) provided a …rst pass at such a theory [see also Hillman, et al. (1995)]. They considered a
model with two countries, two goods, and two sector-speci…c factors of production, one that derives
income only from a country’s import-competing industry and the other that derives income only
from the country’s export industry. The governments have objective functions that are expressed
in reduced form; each government derives “political support” from the speci…c factors in its own
polity, with overall support an increasing but concave function of support from each group. The
support, in turn, re‡ects the real income of a group of speci…c factor owners. In short, they posit
government objective functions G (I
x
; I
y
) and G
I
x
; I
y
for the home and foreign governments,
where I
j
and I
j
are the real incomes of the speci…c factors used to produce good j in the home
and foreign countries, for j = x; y. The …rst partials of G and G
are assumed positive while
the second partials are assumed negative. Each government can set the level of a single policy
variable, t or t
respectively, which is the tari¤ levied on its own imports. Hillman and Moser rule
out exports subsidies with casual reference to countervailing duty laws that would neutralize their
e¤ects; however, they o¤er no explanation for why export subsidies are subject to countervail but
policy makers are free to choose import tari¤s without institutional constraints.
From here, the story proceeds much as in Johnson (1953) and Mayer (1981). In the absence
of an agreement, each government sets its tari¤ to maximize its political support taking the other
government’s policy choice as given. The noncooperative outcome is the Nash equilibrium pair of
tari¤s, much like in Figure 1. The home tari¤ raises the real income of the speci…c factor in the
home import-competing industry at the expense of the real income of the speci…c factor in the
9
Park (2011) illus tra tes a nother potential role of a formal organization such as the Worl d Trade Organization in a
world with imperfectly observed trade barrie rs. In his model, a third party can publi cize private signals a nd thereby
initiate a punishment phase after a violation of the agreement. This enhances the cooperation that can be achi eved
in repeated pl ay.
9
home export industry. Each government’s choice balances the marginal e¤ects on its own political
support, but neglects entirely the adverse e¤ect that the tari¤ has on the other country’s export
interests (and the positive e¤ect that it has on the other country’s import-competing interests).
The Nash equilibrium occurs at a peak of an inverted-u shaped indi¤erence curve for the home
government drawn in the space of the two tari¤s. It also occurs at the peak of a right-parenthesis
shaped indi¤erence curve for the foreign government. And, as in Figure 2, there is a lens to the
southwest of the Nash equilibrium such that, for all pairs of tari¤ rates in the lens, each government
could achieve greater political support from its own speci…c factors than at the noncooperative
equilibrium.
Several observations are in order. First, the motivation for a trade agreement again re‡ects
international externalities and interdependence. Without the agreement, each government neglects
the impact that its policy choices have on the political welfare of the opposite government. But
the motivation for the agreement is not necessarily based on an improvement in social welfare, nor
are the agreed-upon tari¤s likely to be economically e¢ cient, as they are in Mayer (1981). Rather,
an agreement allows the governments to promote “political e¢ ciency”by an exchange of “market
access.”By reducing its own tari¤ slightly, the home government has a negligible e¤ect on its own
political support, since the Nash tari¤ was chosen to balance the o¤setting e¤ects on domestic
interests (including the bene…ciaries of any tari¤ revenue). But the foreign export interests bene…t
from the improved market access and, although the foreign import-competing interests are harmed,
their income drop is bu¤eted by an improvement in the foreign terms of trade. On the margin, a
reduction in the home tari¤ improves political support for the foreign government at only negligible
cost to the home government. The foreign government can reciprocate by cutting its own tari¤,
thereby improving market access for home exporters and raising the home government’s support
from its export industry.
Grossman and Helpman (1995b) attempt to provide better microfoundations for the govern-
ments’choices in a model with campaign giving by special interest groups. There are n+1 industries,
one of which produces the numeraire good from labor alone while the remaining n industries each
produce with labor and a factor of production that is speci…c to the industry. Household welfare
is a quasi-linear function of consumption of the numeraire good and utility from consuming the
various non-numeraire goods. The home government maximizes a political objective function of
the form G (W; C) = aW + C, where W is aggregate welfare in the home country and C is aggre-
gate contributions from all special interest groups.
10
Meanwhile, the foreign government maximizes
G
(W
; C
), with variables de…ned analogously. The politics are treated as a two-stage game in
which the interest groups o¤er contribution schedules to their own government in the …rst stage
and the governments choose their trade-policy vectors (import and export taxes and subsidies) in
the second stage. Interest groups are assumed to represent the sector-speci…c f actors of production
in a subs et of industries, namely those industries that manage to overcome free-rider problems and
10
This objective function can be derived from a two-party model o f elections with impressionable and non-
impressionable voters, where the former can be in‡uenced by camp aign spending and the latter vote pure ly base d on
self interest; see Gro ssman and He lpman (1996).
10
become politically organized. Each contribution schedule links a group’s campaign contribution to
the government’s multidimensional policy choice. In the competition for in‡uence that arises when
the governments behave noncooperatively, the interest-group contributions are assumed a function
of the own government’s policy vector. If a cooperative trade agreement is anticipated, then an
interest group can also tie its contribution to the other government’s choices in the hope of in‡u-
encing its own government’s posture in the trade negotiations. Finally, the “trade war”outcome is
taken to be the Nash equilibrium of a second-stage noncooperative game between the governments,
whereas the “trade talks”outcome is the Nash bargaining solution to the second stage.
The equilibrium policies in the trade war satisfy
t
i
1 =
I
i
L
a +
L
x
i
m
i
1
e
i
+
1
e
i
(2)
t
i
1 =
I
i
L
a
+
L
x
i
m
i
1
e
i
+
1
e
i
(3)
where t
i
is the import tax at home if good i is imported or the export subsidy if it is exported, I
j
is an
indicator variable that takes on the value one if industry j is politically organized or zero if it is not,
L
is the fraction of the total population that belongs to some organized interest group or another,
x
i
is outpu t in home industry i, m
i
is home imports (possibly negative), e
i
is the elasticity of import
demand (negative) or export supply (positive), as the case may be, and symbols with asterisks are
analogous variables that apply to the foreign country. Grossman and Helpman interpret the …rst-
expression on the right-hand side of each equation as the political motive for noncooperative trade
policy; the competition for in‡uence via contributions induces each government to support those
local industries that are politically organized, that have a high ratio of output to trade and that
are relatively immune to deadweight loss, as re‡ected in a small local trade elasticity. The second
term in each e quation is the standard terms-of-trade motive for p os itive tari¤s and export taxes.
The inverse trade elasticity in the partner country re‡ects a country’s monopoly power in trade.
Notice that, if no industry is politically organized (I
i
= 0 and I
i
= 0 for all i) or if the governments
give negligible weight to campaign contributions relative to aggregate welfare (a = a
= 1) then
the model predicts the vector of Nash-equilibrium optimal tari¤s, as in Johnson (1953).
Starting from the trade policies indicated in (2) and (3), the two governments have an incentive
to negotiate a trade agreement for much the same reason as in Hillman and Moser (1995) or, for that
matter, Mayer (1981). In the Nash equilibrium, the unilateral policies chosen by each government
confer externalities on the other. The pair of non coop erative tari¤ vectors is economically ine¢ cient,
but that is hardly surprising, because the governments are only partly concerned with aggregate
welfare. What motivates their cooperation is rather the lack of political e¢ ciency. By trading policy
concessions, the governments can …nd an agreement such that each achieve s a higher weighted sum
of welfare and campaign contributions as compared to the outcome at the Nash equilibrium.
When the organized interest groups design their contribution schedules in anticipation of an
e¢ cient trade negotiation (one that will yield a Pareto-e¢ cient outcome for the two governments
11
with respect to their objectives, G and G
), the negotiated trade policies satisfy
t
i
t
i
=
I
i
L
a +
L
x
i
m
i
1
e
i
I
i
L
a
+
L
x
i
m
i
1
e
i
: (4)
The trade talks do not determine the levels of the imp ort and export taxes or subsidies in each
country in a given industry, but only the di¤erence between the importing country’s tari¤ and the
exporting countries export subs idy, as re‡ected in (4). This is so, because an equal change in the
import tari¤ rate and the export subsidy rate would not a¤ect internal prices in either country,
and so it would not a¤ect consumer surplus, producer surplus, or deadweight loss. An equal rate
change only a¤ects the distribution of tax revenues (positive or negative) captured by the two
governments, and such a change in one industry can always be compensated by an o¤setting rate
change in another industry. The relative bargaining power of the two governments determines the
aggregate trade tax revenue that each collects, but not the composition of that revenue by source
industry.
11
The net e¤e ct of the negotiated trade policies applied on any good re‡ects the di¤e rence in the
political strength of the industry’s special interests in the two countries. This di¤erence re‡ects
whether or not the speci…c factors in the industry are politically organized, whether price distortions
in the indu stry cause a great or small amount of deadweight loss, and how willing the industry’s own
government is to sacri…ce aggregate welfare in exchange for campaign contributions. The negotiated
policies preserve the political motive that each government has to cater to its local interests, while
accounting for the externality that its policies impose on the other. The components of (2) and (3)
that re‡ect the terms-of-trade motive for unilateral policy intervention do not appear in (4), because
the exercise of monopoly power via an optimal tari¤ or export tax enhances aggregate welfare in
one country at the expense of that in the other, and such beggar-thy neighbor policies have no
place in an e¢ cient bargain between governments, be they economically motivated or otherwise.
2.4 Is it All About the Terms of Trade?
To recap, Mayer (an d subsequent authors) identi…ed a terms-of-trade externality that arises when
governments unilaterally pursue the maximization of constituents’aggregate welfare. This exter-
nality creates an opportunity for mutual gain from cooperation. Hillman and Moser (an d others)
argued that governments are motivated not only by economic concerns but also— or especially— by
political concerns that re‡ect the distribution of income. In their view, a terms-of-trade e xternality
is not the primary motivation for trade agreements. Rather, governments seek agreements in order
to internalize a market-access externality. They argued that when institutional constraints preclude
direct government support for export industries (for some unspeci…ed reason), the governments can
reap mutual political gains by jointly opening their markets, thereby capturing the political rewards
that their respective export interests have to o¤er.
11
The indeterminacy arises for much the same reason that all home and foreign tari¤ rates that satisfy (1) are
e¢ cient when tari ¤s are set by two welfare-m aximizing governments.
12
So, is it a terms-of-trade externality that motivates trade agreements, or something else? Bag-
well and Staiger (1999, 2002) make a strong claim that, even in a world with politically-motivated
governments, trade agreements are all about governments overcoming their temptation to manip-
ulate the terms of trade. But they also claim that, when addressing the terms-of-trade externality,
there is a sense in which governments are inexorably led to grant better market access. In short,
they claim that there is no meaningful di¤erence between a terms-of-trade externality and a market-
access externality.
Bagwell and Staiger be gin by specifying two governments’objective functions in reduced form.
They write the governments’political objectives as G (p; p
w
) and G
(p
; p
w
), where p and p
again
are the internal (or “domestic”) relative prices that producers and consumers face in the home
and foreign countries, respectively, and p
w
is the world relative price of the good imported by the
home country, i.e., the international terms of trade. The wedge between p and p
w
re‡ects the
home country’s trade policies, while that between p
and p
w
re‡ects the foreign country’s trade
policies. A special case of these preferences arises when the governments maximize aggregate
welfare, since all factor prices (and thus market incomes) and all consumption decisions depend
on domestic prices, while a government’s revenue depends on the di¤erence between domestic and
world prices. The Grossman and Helpman (1995b) speci…cation yields a similar reduced form, once
the contributions C and C
that enter directly into the governments’objectives are replaced by the
equilibrium …rst-stage contribution o¤ers of the special interest groups. Indeed, the Bagwell-Staiger
reduced form represents gove rnment preferences in a wide range of political-economy models with
perfect competition, because in all such models the domestic prices determine th e distribution of
factor income and con sum er surplus, while the wedges between the world price and the internal
relative p rices determine the inter-governmental distribution of tax revenues.
Consider the e¤ects of a rise in the world relative price of the home country’s import good, p
w
,
holding local prices in both countries constant. This will not alter factor incomes in either country,
nor will it change the prices faced by any consumers. The only e¤ect is a reduction in the home
government’s revenues from its trade policies and an equal increase in government revenues in the
foreign country. Plausibly, this event sh ould be viewed favorably by the foreign government and
unfavorably by the home government. Accordingly, Bagwell and Staiger posit that G
p
w
> 0 and
G
p
w
< 0. Concerning the e conomic environment, they assume only that the Lerner paradox and
the Metzler paradox do not arise.
12
The market-clearing world price depe nds on the trade policies chosen by the two governments.
Given the trade policies and the market-clearing world price, the pair of domestic prices are deter-
mined mechanically. In such circumstances, one might as well imagine that the governments choose
their domestic prices directly and the respective trade policies are de termined residually, once p
w
12
The Ler ner paradox applies when an increase i n a countries tar i¤ leads to an increase in the world price o f its
imports, hence a deterioration in the terms of trade. This can happen only if the government spends a disproportionate
share of its tar i¤ revenues on the import good. The Metlzer paradox applies when a tari¤ generates such a large
improvement in an importing country’s terms of trade that the domestic relative pric e of the import good actually
falls there. For this to happen in the home country, the foreign excess demand for its import good must be su¢ c iently
inelastic.
13
is determined. Writing p
w
as a function of p and p
to re‡ect the dependence of equilibrium prices
on trade policies, the …rst-order conditions for the governments’optimal choices of domestic prices
at the Nash equilibrium imply
G
p
(p; p
w
) + G
p
w
(p; p
w
)
@p
w
(p; p
)
@p
= 0 (5)
G
p
(p
; p
w
) + G
p
w
(p
; p
w
)
@p
w
(p; p
)
@p
= 0: (6)
As before, this outcome is politically ine¢ cient, inasmuch as both G and G
could be increased by
appropriate adjustments in the domestic prices away from those that prevail in the noncooperative
equilibrium. This can be seen from the fact that the iso-utility curve for the home policy maker
is perpendicular to that for the foreign policy maker at the Nash equilibrium, whereas political
e¢ ciency calls for tangency between these two curves.
Next, Bagwell and Staiger designate a benchmark that they term the “politically optimal tar-
i¤s.”They de…ne these tari¤s implicitly as the wedges that arise between domestic and world prices
when the internal prices satisfy
G
p
(p; p
w
) = 0 (7)
G
p
(p
; p
w
) = 0. (8)
How should we understand these politically optimal tari¤s? They are the tari¤s the governments
would choose if, for some unspeci…ed reason, each were to ignore the e¤ect on its political objective
that results from changes in the terms of trade, and if each expected the other to do similarly; that
is, if both acted as if G
p
w
G
p
w
0. Were the governments hypothetically to behave in such
a manner, there would be no scope for them to negotiate a trade agreement in pursuit of mutual
political gains. To see why, consider the e¤ect of a change in the foreign tari¤, t
, on the political
welfare of the home government, starting from a pair of local prices that satisfy (7) and (8). A
change in t
would induce an equilibrium adjustment of the home price, p, but to …rst order this
alone would neither bene…t nor harm the home government, by (7). Th e change in t
would also
alter the terms of trade. But the indu ced change in p
w
would generate a pure transfer of welfare
between the two governments and so it could not be a basis for mutual gain. In short, if each
government were to set its tari¤ equal to the political optimal, there would be nothing further on
which they could agree.
The politically optimal tari¤s refer to a thought experiment; Bagwell and Staiger view them as
purely hypothetical and do not suggest that governments would have reason to behave this way in
any set of circumstances. Since there is no behavioral justi…cation for these tari¤s, it is di¢ cult to
evaluate whether they provide a sensible benchmark to use in ascribing motives to the governments.
Are terms-of-trade externalities an apt description of the motivation for trade negotiations, because
no negotiations would take place if governments simply ignored their welfare e¤ects? Perhaps, but
without an explanation about why and in what circumstances the governments might do so, this is
largely a matter of semantics. The semantic nature of the debate becomes clear when Bagwell and
14
Staiger are asked, Is there no market-access externality such as the one identi…ed by Hillman and
Moser? Bagwell and Staiger (2002, pp. 28-29) in fact regard the terms-of-trade externality and the
market-access externality as being two sides of the same coin. They associate the market access
that a country a¤ords to its trade partner at world prices p
w
with the import demand function,
for example M (p; p
w
) for the home country. Now, starting from a noncooperative equilibrium,
suppose the home government were to contract its market access marginally by increasing the
domestic relative price of the import good. In th e absence of any Lerne r paradox, this would have
to decrease the world price, p
w
. At the Nash equ ilibrium value of p
, the e¤ect on the political
objective of the foreign government is unambiguously negative. So, a contraction of market access
imposes a negative externality abroad, and a mutually bene…cial trade agreement requires each
government to sh ift out its import demand function, i.e., to grant expanded market access to its
trade partner.
13
The important conclusion is that, with perfect competition in world markets, for a wide range
of political objective functions governments have an incentive to negotiate a trade agreement that
expands world trade. This is the essence of the Bagwell-Staiger insight, more so than whether we
describe the agreement as being one in which each country o¤ers a better world p rice for the initial
volume of trade (improve d terms of trade) or as one in which each o¤ers greater import demand at
the initial world prices (improved market access).
2.5 Market Distortions and Corrective Policies
We have seen that governments have an incentive to negotiate a trade agreement even when their
objectives include the support of local special interests. The politically e¢ cient agreement induces
governments to take into account the international externalities from their policy decisions and
generates improved market access for exporters in both countries. What if local markets do not
function perfectly and governments need to use the policy instruments at their disposal to overcome
market ine¢ ciencies? Or what if governments have idiosyncratic preferences for certain types of
market outcomes, be they economic or non-economic objectives? Trade policies are part of the
arsenal that policy makers have to correct market failures and to generate preferred outcomes. And
other policies besides trade policies— even if “legitimately” motivated — can generate externalities
for other governments. How should we think about the pu rpose of trade agreements in a setting with
market distortions, non-economic objectives, and the possibility that governments might invoke all
manner of domestic economic policies?
Bagwell and Staiger (2001; 2002, ch. 8) o¤er an elegant answer to this question. Let us
now write the objectives of the home and foreign governments as G (p; p
w
; s) and G
(p
; p
w
; s
),
respectively, where p and p
are the relative prices faced by home and foreign consume rs, and s and
s
represent th e levels of two arbitrary domestic policies set by the home and foreign governments.
14
13
Bagwell and Staiger (2 002) make a global ar gume nt. They say that an agreement provide s additional market
access for a country if its trade partner ’s import demand function shifts out for some world prices . They prove that
every mutually bene…cial t rade agreement must secuire additional market access in this sense for both countries.
14
This representation of government object ives is slightly di¤erent from t he one used by Bagwell an d Staiger (2001,
15
The domestic policies might be subsidies to local production of one good or the other, in which case
the prices received by producers would di¤er from those paid by consumers. Or the policies might
represent interventions in local factor markets, either in the form of taxes or subsidies for factor
employment or factor supply, or else measures of direct regulation of factor usage. The policies
could also represent limits on pollution, standards for product quality or safety, restrictions on the
use of underage labor, minimum wage rates, etc. Meanwhile G () and G
() could incorporate
political and distributional objectives of the government, as before, as well as their idiosyncratic
preferences over market outcomes, and the government objectives could re‡ect the extent to which
market failings that can be traced to factor-market rigidities or product-market externalities. We
rule out only the exercise of monopoly power by …rms or households (which we will discuss later)
and international non-pecuniary externalities that arise, for example, from cross-border pollution,
induced climate change, or local concerns about market outcomes in the other country.
15
As before,
we assume away conditions that give rise to the Lerner and Metzler paradoxes, and we assume that
the Marshall-Lerner conditions for market stability (i.e., that a rise in the relative price of a good
causes world excess for the good to fall) are satis…ed.
Now, governments set their trade policies and their “other”policies. Equivalently, they choose
their domestic prices p and p
along with s and s
; in this conceptualization, the trade policies
are determined residually. The choices of (p; s) and (p
; s
) determine home and foreign supplies
and demands, given p
w
, and so the world relative price must settle at the level that clears the
world market. In other words, the equilibrium world price can be written as p
w
(p; s; p
; s
). Each
government has two …rst-order conditions that guide its u nilateral choice of trade and domestic
policies; for example, for the home government, these are
G
p
(p; p
w
; s) + G
p
w
(p; p
w
; s)
@p
w
(p; s; p
; s
)
@p
= 0 (9)
G
s
(p; p
w
; s) + G
p
w
(p; p
w
; s)
@p
w
(p; s; p
; s
)
@s
= 0: (10)
On the other hand, global political e¢ ciency requires that the home government set p and s
to maximize G (p; p
w
; s) + G
(p
; p
w
; s
), where is the relative weight attached to the foreign
government’s objective in the e¢ cient agreement, an indicator of its relative negotiating ability.
Therefore, the globally e¢ cient home policies satisfy
G
p
(p; p
w
; s) +
G
p
w
(p; p
w
; s) + G
p
w
(p
; p
w
; s
)
@p
w
(p; s; p
; s
)
@p
= 0 (11)
G
s
(p; p
w
; s) +
G
p
w
(p; p
w
; s) + G
p
w
(p
; p
w
; s
)
@p
w
(p; s; p
; s
)
@s
= 0: (12)
20 02), but it captures t he esse nce of their arguments about both s tandards and production subsidies provided that
the local prices p and p
are interpreted to be consumer pric es and the governments do not have use of consumption
taxes and subsidies. Bagwell and Staiger (2006) analyze WTO treatment of production subsidies in a setting that
allows for domestic consumption policies.
15
An example of a local concern for market out comes in other countries a rises when citizens disdain the u se of child
labor, even if it occurs outs ide the borders of their c ountry.
16
Clearly, the unilateral policies that the home government would choose in a noncooperative s etting,
as given by (9) and (10), do not satisfy the requirements for political e¢ ciency in (11) and (12). The
ine¢ ciency of the Nash equilibrium arises again from an international e xternality, as re‡ected in the
neglected terms G
p
w
(p
; p
w
; s
) [@p
w
(p; s; p
; s
) =@p] and G
p
w
(p
; p
w
; s
) [@p
w
(p; s; p
; s
) =@s].
And c learly the externalities travel to the foreign country via world prices; when the home govern-
ment acts unilaterally, it ignores the e¤ect that its choices have on world market prices and thereby
on the objectives of the foreign government. The foreign government can react by setting its trade
and domestic policies to further its own political objectives, to pursue its idiosyncratic preferences
over market outcomes, and to o¤set its local market distortions, but ultimately the joint choices in
a Nash equilibrium leave scope for mutual gain.
Bagwell and Staiger make a further observation. Supp os e the governments negotiate over market
access, as re‡ected in the location of each country’s import d emand schedule or, equivalently, as
ultimately re‡ected in the world market price. Once they agree on a value of p
w
, the governments
can (and should) be lef t to determine their own mix of trade and domestic policies. That is,
suppose the governments conclu de that p
w
(p; s; p
; s
) should be equal to p
w
under an international
agreement. Then unilateral policy choices by the home government subject to p
w
(p; s; p
; s
) = p
w
will satisfy
G
p
(p; p
w
; s) +
@p
w
(p; s; p
; s
)
@p
= 0 (13)
G
s
(p; p
w
; s) +
@p
w
(p; s; p
; s
)
@s
= 0; (14)
where is the Lagrange multiplier on the terms-of trade (or market-access) constraint. By judicious
choice of p
w
, can be made to equal G
p
w
(p; p
w
; s) + G
p
w
(p
; p
w
; s
), and then the conditions for
global political e¢ ciency in (11) and (12) will be satis…ed. In short, the international agreement
can leave the details about policy mix to each country, as long as each provides the appropriate
degree of market access.
Lest this tolerance f or national sovereignty be misinterpreted, Bagwell and Staiger emphasize
its limitation. Given an agreement about market access, their analysis s hows that each government
can be allowed to choose its own mix of trade and domestic policies. But this is not the same thing
as saying that the governments can come to an agreement about trade policies while leaving each
one free to choose whatever d omestic policies it prefers. If the governments agree only about trade
policies while making no commitments about their domestic policies, the home government, for
example, will have an incentive to distort its choice of s to satisfy a condition like (10). That is, it
will choose its policy not only to correct the domestic market failure or to achieve its non-economic
objective, but also with an eyes towards the e¤ect on the terms of trade. In s o doing, it will partly
undo any concession about providing market access that is implied by its restricted choice of trade
policy. An agreement that restricts choices of trade policies may be better than no agreement at
all, but without a comm itment to prescribed levels of market access, the resulting outcomes will
not be politically e¢ cient.
17
2.6 Critiques of the Theory
I have presented a theory of trade agreements founded on the potential gains from cooperation
among policy makers that have di¤erent constituencies and serve di¤erent interests. On the eco-
nomic side, the theory assumes convex technologies and perfect competition, but it allows for
various types of domestic market failures. On the political side, it assumes self-interested policy
makers who may have purely benevolent motivations or political motivations and may pursue non-
economic objectives. The theory rests on the sole premise that a n oncoop erative Nash equilibrium is
ine¢ cient for the two sets of policy makers, because unilateral policy choices generate international
externalities. The theory presumes governments negotiate trade agreements in order to internalize
these externalities.
The theory has been criticized on various grounds, most notably by Ethier (2007, 2013) and
Regan (2006, 2015). The criticisms take three, interrelated forms. First, whereas the theory
highlights the role of terms-of-trade manipulation in motivating trade negotiations, the rhetoric
surrounding actual negotiations makes virtually no reference to the terms of trade and focuses
instead on governments’aims to reduce “protectionism.”Second, whereas the theory assumes that
governments value tari¤ revenue, the discussions surrounding trade agreements do not highlight
such a concern. Finally, the theory takes the set of political actors that lobby for trade policy as
given, whereas it seems that the prospect of a trade negotiation mobilizes export interests that
might otherwise remain on the political sideline. In this section, I will describe these criticisms in
more detail and outline potential responses.
Recall that Bagwell and Staiger (1999, 2002) identify terms-of-trade manipu lation as the problem
that trade agreements are meant to solve. They argue that, in a hypothetical situation in which
governments act unilaterally but b e have as if they see no bene…t (political or otherwise) f rom
any terms-of-trade improvement, the outcome would be politically e¢ cient and ad mit no potential
gains from cooperation. For this reason, they point to governmental concerns about terms-of-trade
manipulation as the motivating force be hind trade negotiations. Grossman and Helpman (1995b)
make no such claim, but they do observe that the politically-e¢ cient trade policies strip away the
optimum-tari¤ components of the noncooperative policies while preserving the components that
re‡ect the di¤erence in the political strengths of the special interests in the import-competing
and exporting indus tries. One might say that the trade talks in Grossman and Helpman also are
necessary only to rid th e world economy of terms-of-trade manipulation.
Regan (2015) describes a “practitioner’s understanding” of trade agreements, based lo osely
on his reading of historical accounts by participants in trade negotiations. Trade practitioners,
he asserts, do not mention any concern about the terms of trade. The words never appear in
their accounts of their bargaining experiences. Rather, the practitioners tell a story of reducing
“protectionism,” which he de…nes as “any unilateral trade policy that restricts imports to get
political support for the government from import-competing producers.”In his view, policy makers
are willing to forego protectionism in exchange for similar concessions by their partners, because
the political support each can attract from export interests exceeds what each stands to lose from
18
its import-competing interests.
16
Regan further notes that a reduction of protectionism in this
sense can play no role in motivating trade negotiations in Bagwell or Staiger (1999, 2002) or
Grossman and Helpman (1995b), because protectionism remains unabated under the politically-
e¢ cient agreement.
Next, Regan and Ethier (2013) point to tari¤ revenues. These are the other side of the coin
from terms-of-trade improvements. When a trade tax causes the world price of a country’s imports
to fall or the world price of its exports to rise, and domestic prices move in the opposite direction
in the absence of any Lerner paradox, the terms-of-trade gain is realized in the revenues collected
by the government. Were the government not to care about these revenues, it also could not value
the change in world p rices. Put di¤erently, a policy that raises domestic prices while collecting no
revenue (such as a “voluntary export restraint”) can only worsen th e terms of trade, not improve
them. But Regan and Ethier see no evidence in their reading of practitioners’ accounts of any
negotiator’s interest in the revenues generated by import and export taxes.
Finally, Regan and Ethier argue that the key to understanding trade agreements is the boon
they provide to export interests. As they recognize themselves, the ir preferred account relies on an
assumption that export interests have no way to capture policy rents in a world of unilateral policy
setting in a manner akin to what import-competing interests manage to do. Import-competing
interests exchange political support in the form of campaign contributions or otherwise for pri-
vate bene…ts from protection. Export interests, they claim, cannot play this game. According to
these authors, a trade negotiation is fundamentally about e mpowering export interests to counter
protectionist forces.
17
Let us evaluate these criticisms in terms of what they imply about the modeling of trade
agreements. The fact that practitioners do not mention a concern about the terms of trade per se
is not determinative about their role in the workings of formal models of international externalities.
The terms of trade are instrumental in these models, not a direct policy goal. Th e Grossman-
Helpman model assumes that policy makers are concerned with the welfare of their constituents
and with campaign contributions, and that th e latter are o¤ered by interest groups to further their
economic interests. The politicians need only recognize that a lower price of imports contributes to
the welfare of consumers and that a higher price of exports adds to the income of (some) producers
in order to behave “as if”they are concerned about the terms of trade. What policy maker would
not p refer that imports be cheaper and that exports command higher prices, all else the same?
This logic underlies the reduced-form speci…cation of pref eren ces in Bagwell and Staiger. The
legitimacy of their rather general formulation of government preferences does not rest on whether
practitioners use the words “terms of trade”to de scribe their bargaining goals.
The modeling of trade agreements by Grossman and Helpman and by Bagwell and Staiger does
16
Rega n (2015) cites Hillman et al. (1995) as an e xample of what he has in mind, although, as noted above, the
welfare e¤ects of term-of-trade changes are essential to the Hillman et al. demonstration th at the political support
ga ins from export interests exceeds the loss from import-competing intere sts .
17
Neither Ethier nor Regan provides a logically-consistent formal model of this process. Similar informal statements
about this role of trade negotiatio ns can be found in the writ ings of some trade-focused politic al scientists and l egal
scholars, such as for e xample, Hudec (1993) , Destler (2005) and Pauwelyn (2008).
19
rely on the assumption that governments care about the …scal consequences of their policy choices.
To see why, consider a Grossman-Helpman world in which a negligible fraction of the population
is represented by an interest group (
L
= 0) and in which the government for some reason places
no value of tax revenue, positive or negative. In such a setting, the equilibrium contribution
schedules would induce the government to act as if it were maximizing G =
P
(1 + aI
i
)
i
(p
i
) +
S
i
(p
i
), where a is the weight that the policy maker attaches to campaign contributions relative to
producer-plus-consumer surplus (but with no revenue term),
i
(p
i
) and S
i
(p
i
) are producer and
consumer surplus from the consumption of good i when the domestic price is p
i
, and I
i
again is a
dichotomous variable that indicates whether industry i is politically organized or not. Note that
@G=@p
i
= (1 + aI
i
) x
i
(p
i
) c
i
(p
i
) ; where x
i
and c
i
are output and consumption, respectively, and
that @
2
G=@p
2
i
= (1 + aI
i
) x
0
i
(p) c
0
i
(p) > 0. Evidently, there can be no inte rior solution in this
setting; each government tries to set as high a domestic price as possible, which it can do by raising
toward in…nity its export subsidy. Needless to say, this is not a sensible prediction of the model and
does not provide a benchmark against which to consider governments’incentives for negotiating a
trade agreement. In a political-economy setting in which governments literally do not care about
the …scal implications of their policies, their search for contributions or other forms of political
support leads them to subsidize lavishly whichever group has more at stake in the policy decision
and a better means of playing the political game.
Regan and Ethier recognize that an as sum ption of no governmental concern for …scal de…cits
generates nonsensical predictions, so they argue instead for a formulation in which the government
places no value on any positive revenues generated by trade taxes, but b ears a prohibitive cost
of making public outlays for import or export subsidies. Regan in particular argues that such an
assumption— admittedly inconsistent with the belief that policy makers are “rational actors”— is
descriptively realistic. Bu t, even if we accept such irrationality, it is not clear how it would rescue
the situation. Suppose a government acts unilaterally to maximize G =
P
(1 + aI
i
)
i
(p
i
)+S
i
(p
i
),
but does so under the constraint that R
i
(p
i
p
wi
) [c
i
(p
i
) x
i
(p
i
)] 0 for all i, where p
wi
is the
world price of good i and therefore R
i
is the revenues collected from trade taxes on good i. The
constraint says that the revenues from any trade tax cannot be negative, i.e., that subsidies to trade
are impossible. There are two cases to consider. Suppose …rst that x
i
(p
wi
) > c
i
(p
wi
), i.e., that the
country would export good i under free trade. In this case, @G=@p
i
= (1 + aI
i
) x
i
(p
i
) c
i
(p
i
) > 0,
so the government would like to raise the domestic price to bene…t the exporters, but it cannot
do so due to the fact that subsidies are forbidden. Alternatively, suppose that c
i
(p
wi
) > x
i
(p
wi
),
i.e., that the country would import good i under free trade. If c
i
(p
wi
) > (1 + aI
i
) x
i
(p
wi
), it would
wish to reduce the domestic price, but cannot do so without subsidizing imports. If c
i
(p
wi
) <
(1 + aI
i
) x
i
(p
wi
), it will instead wis h to raise the domestic price, and will do so until imports are
eliminated. So a government that acts unilaterally and that does not care about positive revenues
from trade taxes but stays clear of trade subsidies has either zero or prohibitive trade taxes in every
sector. This hardly seems a reasonable depiction of the counterfactual outcome in th e absence of a
trade agreement.
20
More promising, perhaps, is the Ethier-Regan suggestion that trade negotiations empower ex-
port interests that otherwise would be shut out of the political process. They o¤er no convincing
reason why import-competing interests are able to organize themselves and lobby for protection in a
noncooperative setting whereas export interests cannot do so. Nor do they o¤er any explanation as
to why the situation suddenly changes when trade negotiations are contemplated.
18
But Grossman
and Helpman (1995b) are silent about which industries are politically organized and which are not,
and Bagwell and Staiger (2002) are silent about the stability of their reduced-form government
preferences. It is certainly possible that new interest groups become organized when trade nego-
tiations are active, so that the induced government preferences change.
19
Future research might
uncover reasons why the onset of trade negotiations encourages the participation of export interests
in lobbying that are otherwise unable to in‡uence trade policy outcomes. If so, then policy makers
might well wish to enter into trade agreements as a way of to mobilize these interests and thereby
counter the forces of protectionism.
20
Regan (2015) and others have o¤ered another critique of the competitive theory of trade agree-
ments, one that focuses on its predictions rather than its assumptions. The critique concerns the
model’s implications for how export subsidies ought to be treated in trade agreements and the
reality of how modern agreements actually treat such subsidies. Indeed, Bagwell and Staiger (2002,
ch.10) have been puzzled by this same issue .
Consider again the trade policies that result when the home and foreign governments choose
t and t
to maximize G(W; C) = aW + C and G
(W
; C
) = a
W
+ C
, respectively, in a
noncooperative equilibrium of the Grossman-Helpman “trade wars” model.
21
Suppose that good
i is an export good for the home country and that, contrary to the discussion in the previous
paragraph, its export industry is politically active even before any trade talks take place, i.e.,
I
i
= 1. The home -cou ntry’s trade policy in the Nash equilibrium might involve an export subsidy
(t
i
> 1) or an export tax (t
i
< 0). The …rst term on the right hand side of (2) captu res the force
for a subsidy, re‡ecting the in‡uence bought by the domestic industry. The second term captures
the force for a tax, inasmuch as an export tax improves the country’s terms of trade and thus
contributes to the country’s aggregate welfare (note that e
i
< 0 when the foreign country imports
good i). The net e¤ect re‡ects, inter alia, the size of a, the government’s weight on contributions
18
Rega n (201 5) does o¤er the argument that government s uppor t for ex porters in the noncooperative setting woul d
requi re export subsidies and that governments …nd such subsidies politically intolerable, for some unexplained reason.
But as we have just s een, the combina tion of indi ¤erence to positive tari¤ revenues and i ntolerance for any trade
subsidies leads inexorably to the conclusion that either trade is free or none takes p lac e.
19
Mitra (1999) develops a mode l in the spirit of Grossman-Helpman (1994) in which interests group endogenously
form prior to the lobbying game by bearing a cost of organizing. However, the equilibrium in his model does have
endogenous org anization by some export industries, just as i t has endogenous organization by some import-competing
industries.
20
This potential argument for a trade agree ment has the same ‡avor as the commitme nt theories that are d iscussed
in Section 4 below. That is, a government that ha s benevolent intentions ex ante may anti cipate that it will be
tempted to cater to organized interes ts ex post. By entering into a trade negoti ation, it encourages the par ticipation
of the exporters in t he political process and th ereby induces an outcome that is closer to its e x ante preferences.
21
A similar argument can be made using instead the Bagwell-Staiger reduced-form speci…cat ion of the governments’
objectives, G (p; p
w
) and G (p
; p
w
), for which the Nash equilibri um is descr ibed by (7) and (8).
21
relative to welfare. When a is large, t
i
< 1, as concerns about welfare carry the day. When a is
small, t
i
> 1, as th e lobbyists prefer.
Suppose the Nas h equilibrium policy involves an export subsidy and that the countries come
together to discuss a cooperative trade agreement. A small change (up or down) in the home
country’s subsidy has no e¤ect on the government’s objective, G, because t
i
was chosen by the home
government to balance the marginal e¤ects of a policy change on contributions and welfare. But a
small increase in the home export subsidy (starting from the Nash equilibrium) raises the objective
G
of the foreign government. An increase in the s ub sidy rate conveys a positive externality to the
foreign government inasmuch as it reduces the world price of one of that country’s import goods,
i.e., it improves the foreign terms of trade. Since the foreign government has chosen t
i
to balance
the marginal in‡uence of its import-competing ind us try against the marginal a¤ect on aggregate
welfare, dG
=dt
i
must be positive at the Nash equilibrium. In short, the model predicts that trade
talks ought to encourage larger export subsidies than in the noncooperative equilibrium. But the
actual rules of the GATT and WTO do much the opposite, in fact they forbid export subsidies
entirely.
22
Unfortunately, the literature o¤ers no compelling reason why trade agreements should outlaw
export subsidies in a trading environment characterized by perfectly competitive markets. Pe rhaps
the world economy is better described by pervasive imperfect competition, in which case the analysis
of Section 3 comes into play. There, as we shall see, limitations on export subsidies can more easily
be explained. Or perhaps further research will uncover other ways to resolve what Maggi (2014)
has termed the “export subsidy puzzle.”
3 International Externalities with Imperfect Comp etition
Until now, I have focused on international externalities that arise in a perfectly-competitive trading
environment. These externalities give rise to gains from cooperation and so provide incentives for
governments to negotiate a trade agreement. I now turn to environments with imperf ect competi-
tion, where a wider s et of externalities prevail.
3.1 Firm-Delocation Externalities
Delocation refers to the exit by producers from some locations coupled with entry by new …rms
in other locations. The mix of producers’locations in an industry can matter for national welfare
whenever …rms set prices above marginal cost and transport costs create price di¤erences across
markets. In such a setting, a government may have an incentive to pursue trade policies that
encourage entry at home and exit abroad, thereby changing the mix of prices face d by local con-
22
As note d previously, the national policies for industry i in the cooperative equilib rium are not well determined;
only the di¤erence in policies is p inned down by the requirements for poli tical e¢ ciency, as expressed in (4). Ac-
cordingly, t
i
might fall as the result of tr ade talks, so long as t
i
does s o as well. Still, it is troubling that the model
suggests that marginal increases in export subsidies starting fr om the Nash equilibrium would enhance the combined
political objectives of the two governm ents and yet the trade agreement calls for zero subsidies.
22
sumers. If the governments act unilaterally, they will neglect the harm that delocation causes to
consumers elsewhere.
Venables (1985, 1987) studied the welfare e¤ects of tari¤s in models with imperfect competition
and free entry. In Venables (1987), for example, two countries produce varieties of a di¤erentiated
product under increasing returns to scale. They also produce a h omogene ous good under constant
returns to scale. The varieties are CES substitutes, as in Krugman (1980), and trade in the
di¤erentiated varieties entails shipping costs. Labor is the only factor of production. A zero-pro…t
condition determines the number of varieties produced in each country. Venables shows that when
the home country levies a small import tari¤, domestic welfare rises even if the tari¤ revenues
confer no social value. The gains come at the expen se of foreign consumers. The mechanism for
these welfare transfers is that of deloc ation: The tari¤ raises the pro…tability of producing in the
home market, which generates additional entry there. Foreign …rms lose directly from the incidence
of the tari¤ and indirectly from competition with a greater number of home …rms. So, some …rms
exit abroad. Since home products do not bear tari¤s or shipping costs in the home market, they
are cheaper than the imports. Therefore, an increase in the relative number of domestic p rodu cers
reduces the home price index, wh ich raises real incomes at home. Just the opposite is true in the
foreign country.
In the model developed by Venables (1987), an export subsidy also can be used to raise home
welfare. A subsidy, like a tari¤, en hanc es pro…tability for home …rms while reducing that of foreign
…rms, so it induces entry at home and exit abroad. Again, the change in the composition of …rms
bene…ts consumers at home while harming those abroad. Although Venables does not discuss the
net e¤ect on foreign welfare, it is straightforward to show that welfare there might fall; the direct
bene…t that foreign consumers receive from the subsidized prices can be more than o¤set by the
harm from delocation.
What are the incentives for forming a trade agreement in the presence of delocation externalities?
Ossa (2011) addresses this question in a model based on Krugman (1980). He considers trade
between two countries that have symmetric preferences and the same production technologies. The
countries may di¤er in their labor endowments and in their trade policies. Consumers have Cobb-
Douglas preferences over a homogeneous good and a CES composite of di¤erentiated varieties. The
homogeneous good is produced with c onstant returns to scale and is freely traded. Varieties of the
di¤erentiated product require a …xed input of labor as well as a constant per-unit variable input.
The di¤erentiated varieties are traded subject to an iceberg shipping cost and an ad valorem import
tari¤. Firms earn zero pro…ts in an equilibrium with free entry.
Ossa assumes that the iceberg trade costs are su¢ ciently high that both countries produce
di¤erentiated products. Also, the labor endowments are su¢ ciently large that both countries
produce positive amounts of the homogeneous good. As in Venables (1987), he …nds that the real
price index in each country is decreasing in the country’s own tari¤ and increasing in the tari¤
of its trading partner, for small enough tari¤s. The explanation again is delocation: A tari¤ at
home generates entry by home …rms and exit by foreign …rms and the direct e¤ect of the higher
23
import prices due to the tari¤ is more than o¤set by the indirect e¤ect of the change in the mix of
producers.
If tari¤ revenues do not enter the government’s welfare calculus, then the pair of welfare-
maximizing governments set prohibitive tari¤s in a Nash equilibrium of the policy game. If the
revenues do …gure in welfare, then the Nash tari¤s are positive but …nite. The interior solution
results from the fact that by further raising an already high tari¤, the government sacri…ce s tax
revenue. This cost must eventually outweigh the bene…t from further delocation. As usual, the
noncooperative tari¤s are detrimental to the joint welfare of the two countries. The bene…t that each
country achieves by delocation comes at the expense of its trading partner and generates further
deadweight loss. E¢ ciency requires tari¤ reductions; e.g., in the case in which tari¤ revenue does
not enter welfare, at least one country’s tari¤ rate must be set to zero under an e¢ cient trade
agreement.
Ossa emphasizes that the externality that arises in the noncooperative Nash equilibrium of his
model should not be called a terms-of-trade externality. Since the CES demand structure implies
a cons tant markup over marginal cost, the ex-factory prices of export goods are independent of
the tari¤ rate. If we de…ne the terms of trade to be the ex-factory price of an imported variety
divided by that of a domestic variety, then the terms of trade are, in fact, independe nt of trade
policy. If, instead, we de…ne the terms of trade as the price inde x for exported varieties divided
by the price index for imported varieties, then— as Ossa shows— a tari¤ that improves a country’s
welfare via d elocation actually generates a deterioration in its terms of trade. The purpose of a
trade agreement, he argues, is to internalize the externality that results when countries use their
import policies to alter the composition of domestic and foreign …rms in the market.
Bagwell and Staiger (2012b) consider a di¤erent but related model with delocation possibilities,
namely one based on Venables (1985). Their model has linear demands for a homogeneous product
and no income e¤ects on demand. Home and foreign …rms engage in Cournot competition in the
two markets, which are assumed to be segmented. Trade costs are positive. As in Venables (1985),
the equilibrium features two-way trade in identical products. A small tari¤ in the home country
increases the number of …rms located there, decreases the number of …rms in the foreign country,
and by altering the intensity of Cournot competition in the two markets, it raises home welfare at
the expense of foreign welfare. In this setting, if governments are limited to using import tari¤s as
instruments for delocation, then— just as in Ossa (2011)— the ine¢ cient Nash equilibrium involves
positive tari¤s in both countries. Moreover, the countries have reason to negotiate a reduction in
these tari¤s, just as Ossa describes. In a symmetric setting, the e¢ cient tari¤s are zero.
In Bagwell and Staiger’s model, just as in Ossa’s, the countries also have incentive to introduce
export policies. A small export subsidy s tarting f rom f ree trade raises a cou ntry’s welfare just as
does a tari¤, and for much the same reason (see Venables, 1985). The entry of …rms in the country
with the subsidy and the exit of …rms in the other country results in a change in the intensity of
competition in the two markets that favors consumers in the country with the active policy.
Bagwell and Staiger proceed to consider the Nash equilibrium of a policy game in which the two
24
governments can implement both import and e xport policies. They …nd, perhaps surprisingly, that
in the Nash equilibrium, each country combines a tari¤ on imports with a tax on exports. The key
to this …nding is their observation that, when deviating from free trade, each government can always
…nd a small tari¤ and a small export tax that together generate the same internal price and the same
consumer surplus as are achieved under free trade but that yield positive tax revenue. In e¤ect,
import tari¤s and export taxes are complements in their model, because a country that imposes a
high tari¤ and thereby induces entry by a large numbe r of local …rms will want to avoid an export
subsidy that would transfer a great deal of revenue to foreign consumers. Put di¤erently, with an
import tari¤ in place, an export tax by the home country that causes entry abroad will increase the
volume of foreign exports on which the tari¤ is levied, thereby increasing the home governme nt’s
revenues. Bagwell and Staiger conclud e that, starting from a noncooperative equilibrium that has
positive export taxes and high import tari¤s, the countries might appear to have no reason to
conclude an agreement that limits the use of export subsidies inasmuch as subsidies are absent
from this equilibrium. Moreover, an e¢ cient agreement is one in which any positive import tari¤
maintained by one country is exactly o¤set by an export subsidy in the other, so that the net
e¤ect of the o¤setting trade policies on world prices is nil. In this sense, export subsidies must
be tolerated (indeed encouraged) to achieve e¢ ciency, unless import tari¤s are f ully eliminated.
However, as the authors emphasize, an e¢ cient agreement can not leave governments free to choose
whatever e xport subsidies they pref er; once tari¤s have been reduced to low levels, the countries
will have incentives to overuse these instruments to encourage delocation. While the mo de l does
not provide an explanation for bans on export subsidies, it does provide a reason why the use of
subsidies should be regulated.
Finally, in Bagwell and Staiger (2015), the authors consider the nature of the externality that
motivates a trade agreeme nt in models of …rm delocation. They show that the globally e¢ cient
policies coincide with the “political optimum,”where the latter is de…ned as the vector of policies
that would maximize G (p; p
; p
w
) and G
(p; p
; p
w
) were the two countries for some reason to ignore
the e¤ect of its trade policy choices on the world price. In this s ens e, they argue, and contrary to
the claim in Ossa (2011), the fundamental purpose of a trade agreement in the delocation model
is really to eliminate manipulation of the terms of trade. Maggi (2014) points out, however, that
their conclusion very much relies on their assumption of no income e¤ects in the demands for the
imperfectly-competitive good. Moreover, the meaning of “terms-of-trade manipulation”in a model
of Cournot competition with free entry is not entirely clear, at least not to me.
3.2 Pro…t-Extracting and Pro…t-Shifting Externalities
International externalities also arise in imperfectly-competitive environments when governments
can use trade policies to extract monopoly rents from foreign producers or to shift pro…ts from
such produce rs to their domestic rivals. Katrak (1977) and Svedberg (1979) were the …rst to
demonstrate that a government might be able to raise national welfare by imposing a tari¤ on a
good imported from a foreign monopolist. They assumed that the home country confronts the
25
foreign monopoly with a linear demand. In such a setting, a small speci…c tari¤ reduces the ex-
factory price charged by the monopoly. Although consumers pay more for the good with the tari¤
in place, part of their payment goes to the home government in the form of tari¤ revenue. The
monopolist’s reduction in the ex factory price corresponds to a terms-of-trade improvement, for
the home country, which boosts its welfare. Brander and Spencer (1984a) extended the Katrak
and Svedberg analyses to include more general demands. When the foreign monopolist operates
subject to a constant marginal cost, a speci…c tari¤ induces a reduction in its ex-factory price if
and only if the demand curve is not too convex. The condition for a terms-of-trade gain in the
policy active country is R mp
00
(m) =p
0
(m) > 1, where p (m) is the home inverse demand for
imports m. If this condition is satis…ed— or equivalently, if the inverse demand curve is ‡atter than
the inverse marginal revenue curve— then a small tari¤ improves the home country’s terms of trade
and the country’s optimal tari¤ is positive. Much like the tari¤ that arises from unilateral welfare
maximization in a competitive setting, the optimal rent-extracting policy is a beggar-thy-neighbor
policy. The revenues captured by the home government come entirely at the expense of the foreign
producer’s pro…ts. Since global e ¢ ciency requires an output greater than what the monopolist sells
under free trade and since the home country’s tari¤ reduces the monopolist’s output, the optimal
tari¤ in fact reduces global welfare.
Brander and Spence r (1984b) extend the analysis by considering a market in which a pair of
…rms with di¤erent national origins compete as duopolists. The …rms compete in segmented markets
by simultaneously choosing their deliveries to the two destinations. In this setting with Cournot
competition, not only does a unilateral tari¤ by some country have the potential to extract rents
from the foreign …rm, to the bene…t of the domestic treasury, but it also changes the outcome of the
strategic competition between th e two …rms, to the bene…t of the domestic producer. Accordingly,
if a tari¤ induces the foreign …rm to reduce the ex-factory price of output destined for the home
market, this is su¢ cient to ensure a welfare gain for the home country. But even if the ex-factory
price rises, home welfare might rise due to the pro…t shifting that results from the changes in market
shares.
Brander and Spencer go on to consider the Nash equilibrium of a noncooperative game between
two welfare-maximizing governments, In the Nash equilibrium, each government sets a positive
tari¤ on imports from the other’s c ountry’s monopolist. In so doing, it captures revenue and shifts
pro…ts towards its domestic producer. But the pro…t-shifting e¤ects in the two markets o¤set
one another and, taken together, the pair of Nash tari¤s reduce world output. The noncoop erative
tari¤s exceed those that maximize world welfare. In this setting, the governments have an incentive
to negotiate mutual trade liberalization and improved market access for their own national …rms
in the other country’s market.
Brander and Spencer (1985) provide the cleanest analysis of the international externalities that
arise from pro…t-shifting trade policies. They assume that there are sin gle suppliers in each of
two countries that produce a common good only for export to a third market. The governments
can subsidize their local …rm’s exports in anticipation of the Cournot competition for third-country
26
sales. Consider the unilateral incentives for the use of trade policy in one of the exporting countries.
There is no consumption there, so no concern about consumer surplus. The subsidy payment
represents a dollar-for-dollar transfer from the government to the domestic …rm, which is neutral
from the point of view of aggregate welfare, if tax revenues and …rm pro…ts are weighted equally.
What remains, then, is only the strategic e¤ect of the e xport subsidy on the outcome of the
oligopolistic competition. When a government o¤ers a subsidy, the local producer h as a greater
incentive to export than otherwise. For any given quantity of its rival’s sales the local prod uce r
sells more than it would without the subsidy. In other words, the …rm’s best response function
shifts outward in the space of the two output levels. If the foreign best-response function slopes
downward— as is commonly assumed for Cournot competition— then the subsidy induces the rival
producer to reduce its exports to the third market. This strategic response increases the market
share for the subsidized …rm and increases its export price relative to what it would be without the
rival’s retreat from the market. The price in the third-country market typically will be lower than
what it would be absent the export subsidy— wh ich implies a terms-of-trade loss for the subsidizing
country. But the extra pro…ts captured by the domestic …rm at the expense of its foreign rival more
than compensate for this. When both best-response fu nc tions slope downward, each government
has a unilateral incentive to subsidize exports in a grab for oligopoly pro…ts.
Again, we recognize an international externality. The pro…t gains for one producer come at the
expense of the other. And the subs idy causes the price in the third country to fall, which means
that joint pro…ts for the two exporters also fall. If both governments were to pursue their unilateral
incentives, the resulting pair of subsidies would roughly neutralize one another, leaving market
shares about whe re they would have been without the interventions. The …rms may bene…t f rom
their governments’largesse, but welfare inclusive of the subsidy costs must fall in at least one of
the exporting countries. Therefore, the two governments of the exporting countries have a shared
incentive to negotiate a trade agreement that limits the use of such strategic subsidies. Notice that
the purpose of such an agreement would not be to limit manipulation of the terms of trade (since
the noncooperative policies are subsidies that, in fact, worsen the exporters’terms of trade), but
rather to prevent the unilateral pursuit of pro…t shifting.
23
As Brander and Spencer (1985) point out, and Bagwell and Staiger (2012a, Sec. 5) further
emphasize, the externality associated with pro…t shifting can explain a trade agreement that limits
or prohibits export subsidies among a pair (or group) of exporting countries, such as the Boeing-
Airbus pact between the United States and Europe that limited the countries’use of credit subsidies
on foreign sales of large passenger aircraft. But the pro…t-shifting externality cannot explain a lim-
23
Eaton and Grossman (1986) have pointed out that the optimality of subsi dies in an export duopoly rests heavil y
on the assumption of Courn ot competition. If, instead, the rival exporters engage in Bertrand (price) compet ition,
the Nash equ ilibrium policies typical involve a pa ir of export t axes. A unila terally-imposed tax serves to temper
competition in the thir d market and to generate a more collusive outcome. In a Bert rand competition— where a tax
on one …rm’s exports typically induc es its rival to raise its price— th e strategic response can generate revenues for
the taxing government that exceeds the loss in its …rm’s pro…ts. Nonetheless, an international externality arises from
unilateral policies, because the governments do not consider the pro…t gain tha t the forei gn …rm enjoys as a result of
their export taxes. In fact, the joint welf are of the two exporti ng countries would be grea ter if the expo rt tax es were
raised above the levels in the Nash equilibrium.
27
itation on export subsidies in the context of a multilateral trade negotiation in which all countries
participate. The bene…t that exporting countries would capture from an agreement that restricts
export subsidies comes at the expense of h igher prices and reduced consume r surplus in the im-
porting country or countries. Bagwell and Staiger examine a Nash equilibrium in a three-cou ntry
model with symmetric exporting …rms in two of the countries and all consumption con…ned to the
third country, much as in B rander and Spencer (1985). The Nash equilibrium in the three-country
policy game involves pro…t-shifting export subsidies in each of the two exporting countries and a
rent-extracting import tari¤ in the importing country. An agreement that achieves e¢ ciency for
the three countries together does not pin down the tari¤ in the importing country or the subsidies
in the exporting c ountries; e¢ ciency only determines the magnitude of the combined policy wedge.
However, it can be shown that the volume of trade is ine¢ ciently low in the Nash equilibrium as
compared to what is required for e¢ ciency. Therefore, a negotiated agreement should either lower
the tari¤ in the importing country or raise the subsidies in the exporting countries. Under an
e¢ cient agreement, the trade policies should be s et so that the …rms in the two exporting countries
receive the same cum-subsidy price as one another and the consumers in the importing country
pay the same price as they would under perfect c ompetition with free trade.
Bagwell and Staiger (2012a) revisit the question of what is the “fundamental purpose” of a
trade agreement in settings with monopoly or oligopoly pro…ts. In other words, the y ask, What
is the nature of the international externality that a trade agreement seeks to correct? Brander
and Spencer might concede that rent extraction is a form of terms-of-trade manipulation inasmuch
as the government captu res revenue from its import tari¤ by inducing a foreign monopolist or
oligopolist to lower its ex-factory price. But what about strategic subsidies that steal pro…ts from
other exporters and generate welfare gains despite causing a deterioration of the exporters terms-
of-trade? Bagwell and Staiger conclude that the identi…cation of a pro…t-shifting motive for a trade
agreement arises only when the governments lack full sets of trade policy instruments to tax or
subsidize imports and exports and when the importing countries are, for some reason, left out of the
calculus. They show that, in a variety of settings with excess pro…ts— albeit all with quas i-linear
utility— a country’s welfare can be written in reduced form as a function G (p; p
; p
w
) of local prices
p at home and p
abroad, and a world price, p
w
.
24
They de…ne again a “p olitical optimum” as
the p olicies that the governments would choose in a Nash equilibrium if, for some reason, they
were to ignore the welfare en han cing e¤ects of a terms-of-trade improvement, given local prices at
home and abroad. That is, their benchmark arises when all governments choose policies to satisfy
their …rst-order cond itions, but in the process act as if G
p
w
(p; p
; p
w
) 0. They show that the
e¢ cient agreement coincides with this benchmark, which leads them to conclude once again that
the fundamental externality operates through the terms of trade.
It is not clear to me why their benchmark is appropriate, especially in a setting such as this
one where the prices in each market are chosen by active playe rs rather than resulting from market
24
The com petitive setting considered in Bagwell and S taiger (1999, 2002) is a spe cial case in which the for eign price
p
does not enter the reduced- for m government objective function.
28
clearing. What does it mean— aside from the formal de…nition in terms of the …rst-order condition—
that the government acts “as if the terms of trade do not a¤ect aggregate welfare given the domestic
prices?” And why is it so important that we pin a name to the externality, be it “pro…t shifting”
or “terms of trade”? Be that as it may, an interesting insight emerges from Bagwell and Staiger’s
analysis. Namely, in a setting in which each traded good is subject both to an export tax or subsidy
imposed by the exporting country and an import subsidy or tax imposed by the importing country,
and when utility is quasi-linear so that the traded good in question is not subject to any income
e¤ects, the policies in the importing and exporting countries become perfect substitutes in terms
of their e¤ects on quantities, consumer surplus, and pro…ts. When one government sets its policy
at a level that the other takes as given, the second government can always “undo” the e¤e cts of
this policy on local prices in each market by an appropriate choice of its own ins trument. In so
doing, it will impact its trade tax revenues or subsidy outlays, but it can achieve its chosen targets
for consumer surplus and local producer pro…ts. This m eans that the governments can each “get
what they want”in the Nash equilibrium, except for the implied revenues and …scal c osts.
25
3.3 Pro…t-Extracting Externalities in International Outsourcing Relationships
A recent paper by Antràs and Staiger (2012) addresses the ine¢ ciencies that result from nonco-
operative policy setting in an environment with international outsourcing. In such circumstances,
governments may use their trade policies both to correct allocative ine¢ ciencies that result from
incomplete contracting and to extract rents from foreign producers. The two governments share
the former objective but not the latter, and a trade agreement may be needed to help them to
achieve their common goals while avoiding the ine¢ ciencies that result from con‡ict.
The economic environment has customized intermediate inputs, two-s ided buyer-supplier rela-
tionships, and incomplete contracts that give rise to hold-up problems. Two countries, Foreign and
Home, are small in relation to the world market for some …nal good. They take the world price
of the …nal good as given. Foreign alone among the two can produce intermediate inputs, with
a unit continuum of potential suppliers. Home alone among the pair can produce the …nal good,
with a unit mass of potential buyers. Each potential supplier in Foreign is matched randomly with
a potential buyer in Home and the two engage in bilateral bargaining in order to work out an
outsourcing arrangement. If the bargaining breaks down, no …nal good can be produced and any
customized inputs produced for the relationship become worthless. Either country can import the
…nal good f rom the rest of the world at the …xed world price and Home can export the …nal good
to the world market at this same price.
Imperfect contracting manifests in that the speci…cations of the input cannot be stipulated
before production. Instead, the supplier must produce some quantity of the intermediate good and
then n egotiate to sell it to its downstream partner. Given that the outside options are zero at the
25
As Maggi (2014) notes, this would not be true if the allocation of revenues had an e¤ect on m arket demands,
as they woul d in a demand system with income e¤ects. It would also not be tr ue if the internat ional distribution
of revenues and outlays had implicat ions for e¢ ciency, for ex ample if raising revenue entails de adweight l osses from
distorting taxes.
29
bargaining stage, the negotiation always results in a transaction, with a division of ex post surplus
dictated by exogenous parameters. But the supplier in any relationship anticipates that it will bear
the full cost of production while capturing only a fraction of the surplus; this hold-up problem
generates underinvestment in the intermediate input and an ine¢ ciently low level of o¤shoring in
a free-trade equilibrium.
Enter trade policies. The government of Foreign can encourage greater investment by subsidizing
exports of the intermediate good. The government of Home can do likewise by subsidizing imports
of the input or by subsidizing exports or taxing imports of the …nal product. But the combination
of these three policies has implications for the distribution of rents among the two …rms and the
two governments.
26
And the governm ents care about national welfare, not global welfare; if they
act unilaterally in a noncooperative equilibrium, they will take actions on the margin th at bene…t
their local producers and their treasury at the expense of pro…ts and revenues abroad.
Consider …rst the incentives facing the government of Home. If its goal were to maximize global
welfare given the foreign export subsidy (or tax), it would allow free trade in the …nal good and
subsidize imports of the intermediate good to induc e the e¢ cient level of input production. Such a
policy would solve the hold-up problem without generating any by-product distortion of consump-
tion decisions. But, by taxing exports or subsidizing imports of the …nal good, the government of
Home can engineer a bargaining outcome that is more favorable to its local …rms at the expense of
their foreign suppliers. The best response of the government of Home to any policy in Foreign is to
combine such a tax on exports or subsidy to imports with an import subsidy for intermediates that
pushes the volume of input trade closer to the e¢ cient level, but not fully there. The government
of Foreign, in tu rn, …nds it optimal to respond to this pair of policies with a tax on exports of
the intermediate input. Two considerations explain why a tax is optimal. First, the foreign …rms
choose their output levels to maximize their own pro…ts, which coincides with the goal that the
foreign government has for its loc al industry. Second, the tax generates revenue for the foreign
government and part of the tax burden is passed on to home …rms in the bargaining process. As
Antràs and Staiger (2012) show, the Nash equilibrium involves an ine¢ ciently low volume of input
trade and an ine¢ ciently low price of the …nal good in Home, which creates the motivation for a
trade agreement.
27
The requisite trade agreement in these circumstances is, however, rather subtle. First, e¢ ciency
requires intervention in the input market, not free trade. The governments must jointly subsidize
the trade in intermediates to overcome the underinvestment associated with the hold-up problem.
But, second, it is not enough that they agree to an appropriate level of joint subsidy to intermediate
26
The government of Fore ign has no reason to place a tari¤ on imports of the …nal good, i nasmuch as this has no
e¤ect on the e¢ ciency of the outsourcing relationships and it cannot alter its terms of tr ade vis -à-vis the rest of the
world.
27
Antr às and Staiger proceed to introduce political-economic concerns in the form of a potential extra weight that
the government attache s to pro…ts relative to consumer su rpl us or tax revenues. They then ask whether the so-called
“political opti mimum ” in this setting, de…ned as elsewhere, achieves (political) e¢ ciency f or the two governments.
They conclude in the a¢ rmative if the governments place no extra weight on pro…ts, but not otherwise. Accordin gly,
they describe the trade agreement that arises with political-e conomic concerns as ad dressing more than just a terms-
of-trade externali ty.
30
trade. If such a joint subsidy were agreed and the government of Home were left free to set its own
trade policy for …nal goods— perhaps because Foreign does not sell this good to Home and it can
always import this good at a …xed terms of trade from th e rest of the world— then the government
of Home would subsidize exports of the …nal good in order to tilt the bargaining between supplier
and buyer in favor of the latter. In fact, an e¢ cient agreement must constraint not only the policies
that directly a¤ect input trade, but also those that a¤ect the outsourcing relationship in other ways.
3.4 International Agreements to Protect Intellectual Property
Most of the literature on the purpose of international agreements focuse s on contracts aimed at
limiting the opportunistic use of trade policies. But the externalities approach that I have outlined
here can also be applied to international agreements that may arise in other policy areas. To illus-
trate, I will brie‡y discuss the purpose of international agreements to protect intellectual property,
such as the TRIPS Agreement in the WTO. A similar approach has also been applied, for example,
to externalities that arise from governments’choices of environmental policies; see the survey by
Barrett (2005) for a review of this research.
Grossman and Lai (2004) consider two countries, North and South, that di¤er in the sizes of
their populations and in their endowments of human capital. The countries use labor alone to
produce varieties of a horizontally di¤erentiated product and a homogeneous good. They use labor
together with human capital to develop new varieties of the di¤erentiated product. North has a
larger endowment of human capital. Preferences are quasi-linear, with each di¤erentiated product
generating some consumer surplus for households. The di¤erentiated products are sold at monopoly
prices by their inventors until imitation takes place due to imperfect patent protection or until the
patent runs out. Every product becomes obsolete after a …xed period of time. There is free entry
into product development, so the dynamic equilibrium is characterized by equality between the
expected pro…ts from a new product over the course of its economic life and the cost of developing
such a product. In the steady state, the rate of invention of new products exactly matches the rate
of obsolescence of old products.
The welfare maximizing governments in North and South choose policies that determine the
degree of their protection of intellectual property. Grossman and Lai de…ne a pair of policy variables
as
N
!
N
(1e
N
)= and and
S
!
S
(1e
S
)=, where !
J
is the instantaneous probability
that a patent used in country J will be violated due to lack of su¢ cient enforcement,
J
is the
duration of patent protection in country J, and is the discount rate. Thus,
J
is a combination
of the length and strength of patent protection in country J. By assumption, these policies are
applied by each government with national treatment ; i.e., local and foreign producers are treated
similarly by the patent enforcement authorities of each country.
There are two international externalities that arise in this setting. First, governments that
maximize national welfare ignore the surplus that consumers in the other country derive from a
new invention over the course of that product’s subsequent economic life. Second, governments that
are concerned only with national welfare neglect the loss of producer surplus (pro…ts) that foreign
31
monopolies su¤er when imitation occurs or patents expire. Both of these externalities point in the
same direction: patents are too short and too weakly enforced in a noncooperative regime of patent
policies compared to the e¢ cient level of protection of intellectual property. The externalities create
the opportunity for a Pareto-improving patent agreement.
If the governments behave noncooperatively, their best-response functions are downward sloping
curves in (
N
;
S
)-space, because the policies implemented by the two governments are strategic
substitutes; when one country a¤ords greater protection of intellectual property, this induces greater
innovation by …rms in both countries and reduces the incentive that the other country has to o¤er
its own inducement for R&D. The Nash equilibrium occurs at the un ique intersection of these
downward sloping curves. Grossman and Lai show that, in a Nash equilibrium, if the size of the
consumer population in North is at least as large as that in S outh and the endowment of human
capital in North is larger that in the South, then
N
>
S
; i.e., North provides greater protection
for intellectual property than South. In such circumstances, North has greater incentive to protect
intellectual property, because it has more consumers who c an enjoy the surplus from new products
and it has more monopolies that stand to lose pro…ts by patent infringement or patent termination.
Notice that these results do not rely on discrimination, because national treatment precludes any
discrimination in the application of patent rules.
Grossman and Lai compare the patent policies that emerge in a Nash equilibrium with those
that would be stipulated by an e¢ cient agreement. The e¢ ciency frontier lies uniformly outside the
two best-response functions, because the two externalities imply that, given any policy of the other
government, each government provides less protection of intellectual property than would maximize
global welfare. Therefore, an international agreeme nt must strengthen patent protection in at
least one country and provide greater incentives for innovation worldwide. An e¢ cient agreement
need not strengthen p atent protection in both countries in order to generate Pareto welfare gains
and “harmonization” of national patent policies is not necessary for global e¢ ciency. In fact, a
continuum of combinations of
N
and
S
can be used to achieve e¢ ciency; all that matters for
e¢ ciency is the aggregate protection of intellectual property in the world economy, whereas the
policies required of each country under an e¢ cient agreement govern the division of welfare between
the two.
4 Trade Agreements as Commitment Devices
In Sections 2 and 3, I reviewed a literature that treats trade agreements as a means to overcome
international externalities. The externalities may arise from the incentives governments have to
manipulate their terms of trade or from other externalities, such as the incentive to relocate …rms
to the local market or to shift pro…ts to local oligopolists. Another strand of literature o¤ers an
alternative explanation for trade agreements, namely that they provide a means for governments
to tie their own hands and resist the temptation to give in to local special interests that advocate
polices inimical to the general good. I prefer to think of the commitment motive as explaining why
32
a country might sign an existing trade agreement, rather than a reason for two countries to get
together to negotiate an agreement de novo. It is not clear to me why governments would prefer
to design a trade agreement to achieve commitment vis-à-vis their own spec ial interests rather
than to self commit by some other, simpler means. Negotiating a new agreement is a complicated
process that involves many compromises; wouldn’t it be easier to do so unilaterally by, for example,
passing a constitutional amendment that restricts the use of trade policy instruments? However, if
an agreement already exists in some form, a country may choose to take advantage of its existence
by acceding to its terms. Be that as it may, the literature o¤ers an interesting answer to the
question, Why might a government be willing to sign a trade agreement?
Pre-commitment is desirable when a policy that the government regards as bene…cial before the
private sector takes some irreversible (or costly to reverse) action no longer is so afterwards. This
situation creates a “time-inconsistency” problem, as famously described by Kydland and Prescott
(1977). The literature o¤ers several examples of economic environments where the trade policy
that is optimal ex ante no longer is so ex post. I will d esc ribe informally a setting akin to that in
Staiger and Tabellini (1987) as an example.
Suppose the world price of some import good falls, which depresses incomes for those who
work in the import-competing sector. This creates an incentive for workers to move from the
adversely-impacted sector to others. Imagine that workers must make a decision at some point in
time whether to move to a new job or not. Moving requires them to incur a sunk cost that cannot
be recouped subsequently. Let there be a distribution of such costs among workers in the import-
competing sector, so that all those with a personal cost below some critical level move, while the
remainder stay in their original j obs, albeit at lower pay. Finally, suppose the government values
high national income, but also has a preference for an equal distribution of that income.
Once the trade shock has occurred and workers have made their irreversible choices to stay or
move, the government will see higher incomes in the export sector than in the import-competing
sector. If resource allocation has been fully determined by this point, the policy makers may …nd
it attractive to use trade policy as a means to redistribute income. The government can impose an
import tari¤ that somewhat restores wages in the import-competing sec tor, thereby narrowing the
income gap that results from the terms-of-trade shock.
However, if the government is free to use trade policy as a redistributive tool, the workers
may anticipate such interventions when they make their decisions whether to move jobs or not. If
the workers understand that the government’s optimal policy will involve intervention that partly
restores wages in the import-competing sector, fewer of them will move to the export sector than
would be the case without such expectations. The government’s ex post response to potential
inequality will come at a high cost in terms of national income, because there will be less movement
of workers than what e¢ ciency requires. Herein lies the potential bene…t from precommitment:
a trade agreement that precludes tari¤ hikes in response to terms-of-trade shocks can increase
allocative e¢ ciency to such an extent as to more than o¤set the perceived social welfare cost from
33
accepting greater inequality.
28
Maggi and Rodriguez-Clare (1998) extend a similar logic to a setting with political-economic
forces at work. Consider a small country that faces given world prices and that has a …xed stock of
capital. Initially, the c apital is malleable and can be allocated to either of the economy’s two, non-
numeraire sectors. Once allocated, however, the capital is specialized such that it no longer can be
used to produce the other good. The timing is as follows. First, capital owners allocate their capital
to one industry or the other. Then interest groups form (exogenously) and the organized groups
lobby for trade policy.
29
In a small departure from Grossman and Helpman (1994), suppose that
the lobbying proceeds by Nash bargaining between the policy maker and the organized groups,
with an exogenous fraction of the surplus accruing to the former. Finally, the government
implements trade policies, competitive …rms hire capital and mobile labor to produce output, and
households devote their after-tax incomes (re‡ecting payments made or rebates received to balance
the government’s budget) to consumption of the three goods. The policy maker has an objective
function of the form G(W; C) = aW + C, while the lobbies, which represent a small fraction of the
total population, seek to maximize their industry’s capital income in view of the owners’negligible
stake in aggregate consumer surplus and in the government budget.
Suppose there is a multilateral trade agreement in e¤ect, but that the government of the small
country has not acceded to it, thereby retaining its sovereign right to set whatever trade policies
it likes. Once the c apital has been allocated and the lobbies have been formed, the organized
groups bargain with the government over trade policies. At this stage, the fallback positions are
zero contributions from any organized group and free trade; the latter policy maximizes aggregate
welfare and therefore the government’s objective G in the absence of any meeting of the minds
about contributions and policies. Let
i
[(1 + t
i
) p
i
; K
i
] be the payments to capital in industry i
when the world price is p
i
, the ad valorem tari¤ or export subsidy applied to this good is t
i
, and the
capital that has previously been allocated to the sector is K
i
. The joint surplus of the organized
lobbies and the policy maker is given by
J (K) = max
[t]
aW (t; K) +
X
i
I
i
i
[(1 + t
i
) p
i
; K
i
] aW (0; K)
X
i
I
i
i
[p
i
; K
i
]
where t is the vector of trade policies applied to the two non-numeraire industries, K is the vector
of capital allocations to the two non-numeraire industries, W (t; K) is aggregate welfare when the
28
One might reasonab ly ask, In w hat sense does a trade agreement prec lude a tari¤ hike? Mightn’t a governme nt
that has signed a trade agreement deci de anyway to rais e its tari¤s? Presumably, a government that is tempted to
use trade policy to redis tribute income might be dissuaded from doing so by its participation in a trad e agreement
only if it fears some sort of retaliation from its agreement partners. Moreover, th e country’s trading partners will not
have a ny incentive to retaliate a gainst an unauthorized tari¤ hike unless there are external e¤ects of trade policies.
In this general sense, the commitment power of a trade agr eement also relies on the presence of intern ational tra de
policy externalities.
29
Mitra (2002) disallows intersec toral capital mo bility both ex ante and ex post but introd uces a …xed cost of
political organization such that lobbies form endogenously. In his setting, as well, a government may wish to sign a
trade agreement in order to prec ommit to free trade in anticipation of how the political ec onomy will play out in the
absence of commitme nt.
34
trade policies are t and the capital allocations are K, W (0; K) is aggregate welfare under free trade
when the capital allocations are K,and I
i
is an indicator variable that takes on a value of one
if industry i is politically organized and zero otherwise. In these circumstances, the government
achieves the political welfare G = aW (0; K)+J (K), in view of the fact that it receives its fallback
level of welfare plus the fraction of the surplus in the lobbying relationships. The net payo¤s for
the organized groups can be calculated similarly.
Still assuming that the government has not acceded to the multilateral trade agreement, we can
solve for the capital allocations in a rational-expectations equilibrium. The equilibrium allocations,
~
K, are those th at equalize the expected n et incomes for capital allocated to the alternative uses,
considering the trade policies that are anticipated as well as the contributions (if any) that are
expected to be made.
Now suppose that the government has the opportunity at the outset to sign a trade agreement
that commits the country to free trade. By doing so, it foregoes the surplus from its political
relationships with the lobbies. Although the government sees free trade as a desirable outcome
for aggregate welfare, given any K it prefers to enact protectionist or export-promoting policies
in exchange for the valuable contributions it can extract from the lobbies. So why would the
government potentially wish to precommit to trade freely? The answer, as before, has to do with
the ex ante allocation of capital. With a trade agreement in place, the capital owners recognize
that lobbying for protection or export subsidies will be futile. Accordingly, they anticipate earnings
of
i
[p
i
; K
i
] in industry i: The allocation that equates the returns in the two industries, K
F T
,
is of course the one that is ex ante most e¢ cient. Accordingly, W (0; K
F T
) W (0;
~
K), with
strict inequality whenever
~
K 6= K
F T
. In deciding whether to accede to the trade agreement, the
government compares aW (0; K
F T
), its expected political welfare when its joins the agreement, to
aW (0;
~
K) + J
~
K
, the net payo¤ including contributions that it achieves when it opts not to
join.
Maggi and Rodriguez-Clare (1998) observe that joining the trade agreement will be attractive
to the government if = 0, but not so if = 1. If the policy maker captures none of the surplus in
its relationship with the lobbies, then the comparison hinges on the aggregate welfare that results
from the initial allocation of capital, and aW (0; K
F T
) > aW (0;
~
K). On the other hand, if the
policy maker captures most of this surplus, the n the capital owners will anticipate a net return
after lobbying that di¤ers little from what they would earn under government’s fallback position of
free trade; accordingly, the allocation
~
K will be very close to K
F T
. Then aW (0;
~
K) + J
~
K
aW (0; K
F T
)+ J
~
K
> aW (0; K
F T
). In short, the government prefers to tie its own hands when
its bargaining position vis-à-vis domestic interest groups is weak, but not when it is strong; in the
latter case, it can use the ‡exibility to implement trade policies to attract contributions from the
lobbies that exceed (in its political assessment) the losses that it policies imp os e on the general
public.
Brou and Ruta (2013) extend the model of Maggi and Rodriguez-Clare (1998) to allow for do-
mestic subsidies. The government can transfer income to special interests in the import-competing
35
sector either by a¤ording protection or by providing production subsidies that are …nanced by
distortionary taxation. An agreement that limits only the use of trade policies will not be very
attractive to the government, because the value of precommitting the use of one instrument is lim-
ited when the interest group knows that the government can readily substitute another. Brou and
Ruta use the model to analyze the Subsidies and Countervailing Measures Agreement in the WTO
system; a government that wishes to tie its hands vis-à-vis domestic lobbies will be more inclined
to accede to an international agreement that limits the use of production subsidies alongside the
use of tari¤s (or export subsidies) than one that restricts only trade policies.
Whereas an agreement that eliminates tari¤s and export subsidies may not be attractive for
commitment purposes if it leaves the gove rnment free to use good substitutes for trade policies like
production subsidies as alternative means to redistribute income, a government may be willing to
sign an agreement that does not constrain the use of more ine¢ cient means of income transfer.
Limão and Tovar (2011) study a government’s willingness to constrain the use of tari¤ policies
when non-tari¤ barriers (NTBs) are available as substitutes. They assume that NTBs, like tari¤s,
transfer rents to domestic special interests in the import-competing sectors. But these policies
dissipate some of these rents, so they are strictly less e¢ cient as tools of redistribution. The
government may be willing to sign a trade agreement that constraints the use of tari¤s for reasons
akin to those o¤ered by Maggi and Rodriguez-Clare while recognizing that the costliness of NTBs
provides some assurance that it will not succumb to that temptation, or at least not do so to any
great extent. Limão and Tovar (2011) o¤er their analysis as an explanation for why the WTO
system binds tari¤s and prohibits production subsidies, but does not constraint the use of a variety
of less e¢ cient policies that can serve as (imperfect) substitutes for the tari¤s and subsidies.
If governments have incentives to negotiate a trade agreement in order to mitigate international
externalities, then the bene…ts from p recommitment vis-à-vis domestic special interests can p rovide
an added inducement for doing so. Maggi and Rodriguez-Clare (2007) incorporate the two distinct
bene…ts of a trade agreement in a single model, which generates some interesting further insights.
There are two symmetrically di¤erent countries and three goods. In the home country, capital
endowment K
1
can be used to produce either the numeraire good or good 1. Capital endowment
K
2
can only be used to produce good 2. In the foreign country, capital endowment K
1
can be
used only to produce good 1. Capital endowment K
2
can be used either to produce good 2 or the
numeraire good. Endowments and deman ds are such that the home country imports good 1 and
exports good 2. Production technologies are linear in capital and only the import-competing sector
in each country is politically organized. The governments have a political motive for providing
protection and a terms-of-trade motive, as in Grossman and Helpman (1995b). The international
externality lends value to a trade agreement, as does the capital misallocation that results from
anticipated protection.
Let the initial allocations of capital and th e initial tari¤ rates emerge from the Nash equilibrium
of a non cooperative game, as in Grossman and Helpman’s trade war. The resulting tari¤s are larger
than the welfare-maximizing rates, and so there is overinvestment in the import-competing industry.
36
Initially there is no thought of a trade agreement and the allocations do not anticipate one being
negotiated. But suddenly, that possibility arises.
30
The lobbies and the governments negotiate over
a campaign contribution and the terms of an agreement, which takes the form of an (endogenous)
cap on tari¤ rates. The agreement maximizes the joint surplus of the two governments and the
two lobbies. After an agreement is signed— if that happens— each owner of a unit of K
1
in the
home country or of type K
2
in the foreign country has the opportunity to move that capital to
the numeraire sector with some exogenous probability z. The parameter z is meant to capture the
degree of capital mobility in the import-competing industry, from complete speci…city (z = 0) to
perfect mobility (z = 1). After any capital reallocation takes place, the lobby in each country and
its government negotiate again about the actual level of the tari¤, but this time subject to the
constraints impose d by the international agreement. Finally, political contributions are paid and
production, trade and consumption take place.
Maggi and Rodriguez-Clare show …rst that the governments and the lobbies (weakly) prefer an
agreement that impos es caps on tari¤s to one that explicitly determines their levels. If an agreement
must stipulate exact levels of the policy instruments, a jointly e¢ cient agreement will reduce tari¤s
from their noncooperative levels, but not to zero, and there will remain distortions in the allocation
of capital and in consumption. Now suppose that there is an option instead to set that same tari¤
as a maximum, rather than as a requirement. Such an agreement would leave discretion for each
government to set the actual tari¤ below the agreed ceiling, and so the lobbies would have to o¤er
contributions to avoid such an outcome . Inasmuch as a tari¤ ceiling imposes an additional burden
of contributions on the capital owners, it reduces the overinvestment in politically organized sectors.
Joint surplus is raised by an agreement that reduces such distortions, and all parties (governments
and lobbies) can share in the gains by appropriate adjustment of contributions in the initial round
of lobbying. Accordingly, the model predicts that a trade agreement will designate tari¤ ceilings
(“bindings”) rather than tari¤ levels, if such a contract is possible.
The model links the size of tari¤s cuts (from the initial Nash levels to ultimate policies that
are set subject to the constraints of the agreement) to the degree of capital mobility, as captured
by the parameter z. In the extreme, if z = 1, the owners of capital in the import-competing
sector can always earn the return promised in the numeraire sector. There are no rents to be
captured in the ex post stage of lobbying, hence the lobbies are not willing to pay anything to their
governments to compensate for long-run distortions associated with protection. Accordingly, the
trade negotiation cuts tari¤s to zero. At the opposite extreme, if z = 0, there is no possibility
to undo the misallocation of capital. The equilibrium agreement eliminates the terms-of-trade
component of the Nash tari¤— as in Grossman and Helpman’s trade talks— but it cannot reduce
the domestic-commitment problem. Maggi and Rodriguez-Clare show that the tari¤ cut relative to
the initial, Nash equilibrium level is monotonically increasing in capital mobility, z. If we interpret
this result as a cross-sectional prediction, it says that tari¤ cuts should be deepe r in those industries
30
Maggi and Rodriguez (2007) show that their insights carry over to a version of the model in which the opportunity
to negotiate a trad e agreeme nt is not a surp rise, but rather is perfectly anticipated when the initial capital allocation
takes place.
37
where capital speci…city is less and outward reallocation is easier.
Finally, th e authors consider how the extent of trade liberalization varies with the extent of the
governments’concern for aggregate welfare relative to campaign contributions. In a setting without
precommitment considerations, such as Grossman and Helpman (1995b), the Nash equilibrium
tari¤s tend to be higher when the governments’concern for aggregate welfare is small. High initial
tari¤s limit the volume of trade and thus weaken the terms-of-trade externalities. We might expect,
therefore, that tari¤ cuts will be smaller the greater is the governments’taste for contributions and
the smaller is its weight on aggregate welfare. Bu t in the model developed by Maggi and Rodriguez
(2007), the opposite can be true when capital is su¢ ciently mobile. In this setting, high initial
tari¤s in the Nash equilibrium (that emerge wh en the government has less concern for aggregate
welfare) imply a large departure from allocative e¢ ciency. If z is large, there is much to be gained
by committing to low tari¤s and thereby inducing a substantial reallocation of capital. Accordingly,
the agreement should call for deep tari¤ cuts when a small government weight on welfare induces a
large initial distortion and a high degree of capital mobility implies that the costs can be reduced
greatly by a commitment to freer trade.
5 Incentives to Form Regional or Preferential Trade Agreements
Until now, I have mostly discussed the purpose of trade agreements in the context of a two-
country world economy. One exception concerned the incentives that exporting countries might
have to limit their use of strategic export subsidies to third-country markets. But bilateral and
plurilateral agreements take place in many other contexts. Indeed, the number of bilateral, regional,
or other preferential trade pacts has been growing in leaps and bounds, giving rise to what Bhagwati
(1995) has called the “spaghetti bowl”of international agreements. Baldwin and Venables (1995),
Panagariya (2000), and Krishna (2005) have written excellent surveys of the theoretical literature
on the economic e¤ects of preferential trade agreements (PTAs), while Limão (2016) reviews the
empirical research in Chapter 14 of this volume. In this section, I will limit myself to those few
articles that address the reasons that governments might choose to negotiate exclusive agreements—
rather than, or in addition to, multilateral agreem ents— in a world econ omy with more than two
countries.
5.1 The Ohyama-Kemp-Wan Theorem
Any discussion of the incentives for trade agreements among a limited set of cou ntries should begin
with the renowned Ohyama-Kemp-Wan theorem (see Ohyama, 1972, and Kemp and Wan, 1976).
These authors proved a striking result: If lump-sum transfers are feasible within a union, any
group of countries can form a customs union and set a common external tari¤ in such a way that
all memb e r countries bene…t and no excluded country is harmed. The logic of the argument is
simple. Let the union choose an external tari¤ that leaves its members’aggregate vector of trades
with the rest of the world unchanged. (We know this always is possible based on results about the
38
existence of market-clearing prices in a competitive equilibrium.) Then markets will clear in the
rest of the world at the prices that prevailed before the union. With the same prices and the same
trades, these countries are exactly as well-o¤ as before. As for the union members, we can treat their
vector of trades as if it were an endowment vector. E¢ ciency within the union requires equalization
of marginal rates of transformation and of marginal rates of substitution across memb er countries.
This is achieved by a common vector of prices, which in turn is guaranteed by internal free trade.
All that remains is to share the e¢ ciency gains, which can be accomplished costlessly when the
countries have access to lump-sum transfers be tween members.
31
Countries have an incentive to form cu stoms u nion in order to achieve allocative e¢ ciency. The
same is true if the governments have non-economic production targets (see Krishna and Bhagwati,
1997). The result also extends to environments in which the policy makers have political objectives
besides aggregate welfare, provided they have access to e¢ cient instruments to redistribute income
to favorite interest groups. However, as Richardson (1995) cautions, the Ohyama-Kemp-Wan the-
orem should be interpreted with care. It cannot be taken to imply that a group of countries can
form a mutually-bene…cial customs union no matter what is the response in the rest of the world.
The proof assumes that the rest of the world responds to the customs union by making the same
vector of trades at the same prices; i.e., the aggregate o¤er curve of the rest of the world is not
a¤ected by the formation of the union. If nonmembers can respond by, for example, setting a new
vector of “optimal”or “politically-guided” tari¤s, then gains for union members are not assured.
5.2 Terms-of-Trade Gains
Just as a single country can gain at the expense of its trading partners by exploiting its monopoly
power in world markets, so too can a group of large countries bene…t by c ooperating to exploit
their joint market power in trade. In fact, two large countries stand to gain by forming a free-trade
area (FTA) even if they d o not alter their external tari¤s vis-à-vis nonmember countries.
Consider a three-country world in which countries A and B form an FTA with distinct external
tari¤s and country C represents the excluded rest of the world. Suppose …rst that A imports some
good from B and C, both of which have upward-sloping, competitive supply curves. Country A
has an external tari¤ of t
A
that applies initially to imports from all sources. Once it forms an FTA
with B, the tari¤ applies only to imports from C. Let p
C
represent the initial price received by
exporters in both B and C for sales in country A, so that p
C
1 + t
A
is the pre-FTA domestic
price in A. When country A eliminates its tari¤ on imports from B, suppliers there can sell in A’s
market at the prevailing domestic price. There is excess supply in the world market at the original
prices, as …rms in B produce more at the higher delivered price. The price of imports from the rest
31
Di xit and Norman (1986) show that lum p-sum transfers are not necessarily to share the gains from tra de, if
countries have access to a full set of consumption and factor taxes and subsidies. Panagariya and Krishna (2002)
extend the Ohyama-Kemp-Wan result to include free trad e areas in which member countries maintain separa te
exter nal t ari¤s vis-à-vis imports from the rest of the world but trade freely within the area. In this cas e, member
tari¤s are chosen to preserve the initial vector of trades by each area country and resulting internal prices are no t
the same in these countries .
39
of the world must fall to clear the world market. The fall in p
C
bene…ts A while harming B, but
since A’s imports are larger than B’s exports, the net e¤ect must be positive. The members of the
FTA capture a terms-of-trade gain as a result of trade diversion.
Now suppose instead that A and C both import from B. The initial price received by exporters
in B is p
B
and the domestic price in A is p
B
1 + t
A
. With the elimination of the barrier to
internal trade, the price in country A falls. This creates excess demand. The supply price from
B must rise to clear the world market. The increase in p
B
bene…ts B while harming A, but since
B’s exports are larger than A’s imports, again the gains outweigh the losses. In this case, trade
creation generates a positive terms-of-trade e¤ect for the me mbe rs of the FTA.
Countries that forge a regional or preferential trading arrangement c an gain even more by
adjusting their external tari¤ or tari¤s. The incentive for doing so is analogous to that for merger
among competing oligopolists; whereas each country can exploit market power on its own, the joint
in‡uence over world prices is greater than for any one alone. Th e potential gains are evident as
a corollary to the Ohyama-Kemp-Wan theorem (or the Panagariya-Krishna extension to FTAs):
If a group of countries can bene…t by forming a customs union with an external tari¤ that leaves
the terms of trade the same as before, then they can ben e…t even more by adjusting their external
tari¤ optimally.
Kennan and Riezman (1990) investigated whether countries can gain by forming a customs
union, once the tari¤ response by nonmember countries is taken into account. They examined an
endowment e conomy with three goods and three countries that are symmetric up to a relabelling
of the endowment good s, with a linear expenditure system in all countries. In this setting, they
compared the outcome in a Nash equilibrium in which two of the countries allow internal free trade
while jointly choosing an optimal external tari¤ to the outcome in a Nash equilibrium without any
cooperation, and th e outcome in an equilibrium with global free trade. Whenever each country’s
endowments of its export good is n ot too large relative to the total world endowment, any pair of
countries fares better in a customs union equilibrium than in one with global free trade. There-
fore, the possibility of forming customs unions undermines the prospects for a multilateral trade
agreement
Kennan and Riez man also examined how size a¤ects the incentives that countries have to form
a customs union and, in particular, whether a pair of smaller countries can gain by joining forces
to enhance their collective market power once retaliation is taken into account. In their examples,
welfare rises in each of a pair of smaller countries when they form a customs union compared to
the outcome with no cooperation, but each fares less well than it would in an equilibrium with
universal free trade. In contrast, larger countries fair better in a Nash equilibrium in which they
are partners in a customs union compared to both the Nash equilibrium without any cooperation
and the equilibrium with global free trade.
Of course, the motivation to form a customs union or FTA in order to exert collective market
power relies on the same beggar-thy-neighbor calculus as does the unilateral imposition of optimal
tari¤s in a setting without trade agreements. The gains to the member countries come entirely at
40
the expense of countries on the outside. Krugman (1991a) began a small literature that addresses
the welfare implications of having a trading system with multiple, non-overlapping blocs in which
each bloc allows internal free trade but behaves non-cooperatively vis-à-vis the others. He considers
a world with a large number of symmetric countries divided into a smaller number of symmetric
blocs. Each country produces a unique good and all such goods are CES substitutes for one another.
He takes the number of such blocs as exogenous. Each bloc sets an external tari¤ that maximizes
the joint welfare of its members, given the tari¤s set by other blocs. In these circumstances, the
height of each cou ntry’s tari¤ grows with the size of the typical bloc. Consolidation of the world into
larger blocs has o¤setting e¤ects on welfare in the typical country; between-bloc trade distortions
grow m onotonically larger with bloc size, but so does the fraction of world trade that takes place
within blocs. Welfare is highest when the entire world comprises one bloc, but also is high when
the world has small bloc s that have little monopoly power and therefore impose low tari¤s. In
between, welfare is a non-monotonic function of bloc size. Krugman notoriously found that, for
many values of the elasticity of substitution, welfare is minimal when the trading systems comprises
three symmetric blocs.
The …ndings in Krugman (1991a) rely heavily on the assumption that countries are symmetric
and form their blocs arbitrarily. In a follow-up paper, Krugman (1991b) discusses informally a case
with “natural trading blocs.”In this setting, geography or other conside rations give certain groups
of countries a greater predilection to trade with one another th an with those outside the group.
If blocs form “naturally” among groups of cou ntries that trade intensively, the free movement of
goods within blocs will cover a majority of world trade and the external trade barriers will apply
to a small volume of trade. In the limit, with very high costs of trade outside a natural grouping,
the formation of trading b locs must be raise welfare for all involved.
32
5.3 Political Incentives for Regional or Preferential Agreements
In addition to the potential improvements to allocative e¢ ciency and to external terms of trade that
can motivate countries to form preferential trade agreements such as FTAs and customs unions,
there are political forces that can explain this outcome.
Grossman and Helpman (1995a) examine the political con‡ict between supporters and op po-
nents of an FTA in a model with industry campaign contributions. There are two small c ountries
that initially have the tari¤s predicted by the lobbying model of Grossman and Helpman (1994).
Each can trade with the rest of the world at …xed terms of trade. The countries have an have
an opportunity to eliminate tari¤s on their internal trade while retaining their initial external
tari¤s for imports from the rest of the world. Export industries o¤er contributions to encourage
the government conclude an agreement in order to e xpand market access in the partner country.
Import-competing interests o¤er contributions to discourage the agreement, in order to protect
32
In another e xtens ion of Krugma n (1991a), Bo nd and Syropo ulos (199 6) e xamine more fully the relationship
between bloc size and the market power exerted by trading blocs, by allowing for blocs of di¤erent sizes and by
allowing for alte rnative endowment structures.
41
their local markets. Each government chooses a stance that maximizes a weighted sum of aggre-
gate welfare and contributions. A FTA is politically viable if and only if it is favored by both
governments. The question is, Under what conditions is an FTA most likely to form?
An FTA has no e¤ect on industries in which both countries export to the rest of the world.
So, Grossman and Helpman focus on industries in which at least one of the member countries has
positive imports in the initial equilibrium. They distinguish two possible outcomes for such an
industry onc e an FTA comes into being. First, the importer might expand its imports from its
partner while continuing to import from the rest of th e world. In this case, the internal price in
the importing country remains equal to the world price augmented by the MFN tari¤. There is no
expansion of its total imports, only diversion of trade from the rest of world to the FTA partner.
Second, the importing country might cease its imports from the rest of the world and import only
from its FTA partner. In this case, the price falls in the importing country and total imports
expand. The internal price in the exporting member of the FTA need not rise very much and
might not rise at all. Grossman and Helpman refer to the former outcome as one with enhanced
protection, which comes hand in hand with trade diversion. The latter outcome is one with reduced
protection, together with trade creation.
Now consider the politics. In the case of enhanced protection, the industry in the exporting
country gains producer surplus from preferential access to the higher internal prices in the partner
country. Meanwhile, the import-competing industry su¤ers no losses, as the internal price there
remains as before. On net, s uch industries contribute to political viability; the export interests lobby
in favor of the agreement and the import-competing interests have no reason to oppose. In contrast,
with reduced protection, the exporting interests stand to gain little or nothing, while the import-
competing interests in the potential partner fac e the prospect of falling local prices. The existence
of such industries bolster opposition to a potential agreement in one country with only modest (or
no) support generated in the other. Overall, an FTA will b e viable if most industries have the
potential for enhanced protection and not so if most face the prospect of reduced protection. But
trade creation enhances economic e¢ c iency in the small countries, whereas trade diversion detracts
from an e¢ cient allocation of resources. This leads Grossman and Helpman to conclude that an
FTA is most likely to be politically viable when it is also economically harmful. Krishna (1998)
comes to much the same conclusion in a model with oligopolistic trade in which trade politics are
driven entirely by e¤ects on …rms’pro…ts.
However, Ornelas (2005a) provides a counterargument. He considers an economy much like
the one in Krishna (1998), with quasi-linear utility, constant returns production of a numeraire
good, and oligopolistic competition in a s econ d sector, with …xed numbers of …rms in each country
and segmented markets. But whereas Krishna assumes that MFN tari¤s remain the same after
any agreement is signed, Ornelas allows the FTA members to adjust their external tari¤s. He
shows that, in this setting, they have incentive to reduce their tari¤s vis-à-vis nonmembers, for two
reasons.
33
First, a government’s incentive to set high tari¤s in order to extract pro…ts from …rms in
33
Ri chardson (1993) shows that countries also have an incentive to reduce their external tari¤s after forming an
42
nonmember countries is mitigated by the FTA, because …rms in partner countries with improved
access to the local market capture more of the pro…ts that are extracted from the outsiders. Second,
the political support for high tari¤s vis-à-vis nonmember countries is reduced by the formation of an
FTA, because politically-active …rms lose sales in their home market to rivals in partner countries,
and so have less incentive to lobby for protection from outsiders. In the economic environment
considered by Ornelas, the fall in external tari¤s that results from an FTA always is deep enough
to generate net trade creation with nonmember countries. Moreover, in his model, an FTA can be
politically viable only if it is e¢ ciency enhancing. Since the FTA reduces political contributions by
local …rms, it can be attractive to the policy makers only if aggregate welfare rises to more than
compensate.
When FTAs cause trade diversion from nonmembers, the politics of preferential trade can be
self-reinforcing, a phenomenon that Baldwin (1993) termed “domino regionalism.” He considers a
world with many potential members of a customs union. E ach government’s stance toward joining
the union re‡ects an internal trade-o¤ between consumer welfare and industry interests, but with
an additional term that re‡ects the country’s exogenous “resistance”(positive or negative) to being
a member of the club. Countries di¤er in this regard, and those that are least res istant are the
…rst to join. The economies produce a homogeneous good and a …xed set of di¤erentiated varieties.
The former good is traded freely, whereas the latter goods are traded subject to iceberg trading
costs. These costs are higher for …rms outside the region than for those that are potential members
of th e customs union. The political contributors in this setting are the owners of the di¤erentiated
varieties.
Baldwin imagines an initial equilibrium in which all countries in the region with resistance below
some critical level are members of the union and those with greater resistance are not. Then, there
is an exogenous shock that reduces trade costs within the region. The decline in within-group trade
costs causes additional governments to apply for membership. But, as membership expands, …rms
in nonmember countries intensify their lobbying, as the potential pro…ts for insiders expand at the
expense of pro…ts for outsiders. An initial round of union expansion alters the political balance in
nonmember countries, such that further expansion follows. The ultimate growth in regionalism,
then, is a multiple of what one might expect from the initial shock. In short, Baldwin argues that
the trade diversion associated with preferential trade agreements can be politically contagious.
34
But once again, the conclusion can be di¤erent if an FTA results in a reduction of external
tari¤s that promotes trade with nonmember countries. Ornelas (2005b) considers one such setting,
FTA when markets are perfectly competitive.
34
Chen and Joshi (2010) examine how the existence of FTAs a¤ects the incentives that countries have to form
new agreements. They consider a model with three countries and two good s. The numeraire good is com petitively
produced while the other g ood is produced by one …rm in each country. Preferences are quasi-l inear, demands
for the oligoply good are linear, and the competiti on features Cournot behavior with seg mented markets. In this
setting, if one country in a pair has an existing FTA with the third c ountry but the other does not , then the existing
FTA strengthens the incentive for th e member to form a nother FTA b ut weakens the incentive for the non-member
country to do so, compare d to a benchmark with no pre-existing F TA. However, if both potential members of a new
FTA part icipate in pr e-existing agreements with the third country, the incentives that both have to conclude a new
ag reement are inevitably stengthene d.
43
using the same economic model as in Krishna (1998) and Ornelas (2005a). The f act that …rms in
nonmember countries bene…t from the reduction in external tari¤s of the FTA partners without
having to reduce their own tari¤s means that they may have less incentive than before to enter
into agreements with the member countries, either as partners to a PTA or as signatories to a
multilateral agreeme nt.
5.4 PTAs as Stepping Stones to Multilateral Free Trade
Another potential purpose of preferential trade agreements is to facilitate a process of multilateral
trade liberalization. The literature identi…es circumstances under which PTAs serve as “building
blocks”for global free trade; that is, a multilateral agreement that implements free trade becomes
achievable as an equilibrium outcome in a dynamic game when PTAs are negotiated along the way,
when such an outcome would not be possible in a negotiation game that precludes preferential
agreements. Of course, PTAs can also represent “stumbling blocks”in some circumstances; that is,
they can impede or prevent the achievement of global free trade in situations wh ere a multilateral
agreement would emerge as an equilibrium if discriminatory trade were prohibited.
35
Indeed, the
Ornelas (2005b) paper mentioned above provides one such example. But, in such cases, one would
not typically regard the “purpose”of the PTA as being to interfere with multilateral negotiations;
rather, the imped iment to free trade would be seen as an unintended consequence. Since this
chapter deals with the purpos e of trade agreements and not their un intended consequences, I will
not review further the research that describes PTAs as stumbling blocks.
One way in which a PTA may facilitate the achievement of global free trade is by raising the
cost of being left out. Saggi and Yildiz (2010) illustrate this idea in a three-country, three-good,
endowment model with linear demands. Each c ountry is endowed with two goods that it exports
to the other two. It has no endowment of the third good, which it imports from its two trade
partners. In a dynamic game that allows for “bilateralism,”each country announces in a …rst stage
the names of any trade partners with whom it is willing to engage in mutual free trade. In the
second stage, the countries set external tari¤s for any an d all countries with whom they have not
entered into an agreement in the …rst stage. In a game that precludes bilateralism, the countries
can only agree to liberalize trade on a multilateral basis. If any country declines to do so, then the
noncooperative tari¤s ensue.
36
Saggi and Yildiz consider …rst a situation in which the countries are symmetric with respect to
their endowment levels. In such circumstances, global free trade is the unique stable equilibrium
under both bilateralism and multilateralism. Accordingly, with symmetry, PTAs have no role to
play in the achievement of global free trade in their mode l; a multilateral agreement is reached
35
The te rminology of building blocks and stumbling blocks was …rst i ntrod uced by Bhagwati (1991).
36
Saggi et. al (2013) compare bilater alism and multilateralism in a model in which countries are free to form
customs unions but n ot FTAs. They show that customs unions, unlike FTAs, can prevent th e achievement of global
free trade, because two countr ies may prefer t o exc lude the third country from mutual free trade in order to expoit
their joint market power. The incentive for exclusion is stronger under a customs union than under a FTA, so the
former can be a stumbling block but not a building block for multilateral trade liberalization in their se tting.
44
even if bilateral agreements are not allowed. But the conclusion can be di¤erent with asymmetric
countries. Then, global free trade emerges as a stable equilibrium for a wider set of p arameter
values with bilateralism than with multilateralism. In other words, for some endowment combina-
tions, a multilateral agreement can be reached only if bilateral agreements represent a permissible
alternative.
The …nding is readily und erstood . Consider a country that fares better in a Johnson-like
Nash equilibrium than under global free trade. In a multilateral process, such a country would
decline to name any of its partners in the negotiation s tage. By submitting a blank sheet, it
could e¤ectively block an agreement and achieve its …rst best. Suppose instead that bilateral
agreements are possible. The country that fares best in the noncooperative equilibrium might not
have this outcome as a viable option. If the other two countries can gain by forming a bilateral
block compared to universal noncooperation, then the relevant comparison for the third country is
between a multilateral agreement and a world with an PTA f rom which it is excluded. Since trade
diversion within the member countries would hurt the excluded country, it might well prefer the
multilateral agreeme nt to being the one country on the outside.
37
In Saggi and Yildiz (2010), the possibility of a P TA facilitates the con clus ion of a multilateral
agreement, but no PTAs need to form for global free trade to be realized. This is an inevitable
consequence of the game structure, wherein either a bilateral agreement or a multilateral agreement
is signed, but not both. Aghion et al. (2007) introduce a sequential structure in which PTAs may
actually form along the equilibrium path. They consider a trade-negotiation process in reduced
form, with the details of the ec onomic environment suppressed. In p articular, they specify payo¤s
for each of three countries under all possible coalition structures: with each country alone; with
all combinations of bilateral agreements; and with a multilateral agreement. They allow for utility
transfers between coalition members, so the relevant payo¤s are those that accrue jointly to the
parties to any trade agreement. The bargaining protocol features an agenda setter that makes take-
it-or-leave o¤ers to the other two countries. In a multilateral negotiation, the o¤ers to engage in
free trade are made simultaneously to the others and they can accept or reject. With bilateralism,
the agenda setter makes the o¤ers sequentially, in whatever order it prefers. If the …rst to receive
an o¤er accepts, a bilateral agreement is formed. If the second also accepts, the agreement becomes
multilateral. Aghion et al. (2007) ask, When does the agenda setter choose the sequential process?
And, When is a sequential process necessary to achieve global free trade?
The authors distinguish several cases. If the payo¤s exhibit “grand coalition superadditivity,”
then the su m of the payo¤s under global free trade exceeds the sum of the payo¤s under any other
coalition structure. Grand coalition superadditivity necessarily applies in a neoclassical economy
with welfare-maximizing governments. But it need not apply if there is imperfect competition in
37
Saggi and Yildiz (20 11) reach much the same conclusion in a model with a di¤erent economic setting. There they
consi der a world economy with Co urnot competition in segmented markets and one …rm per country. In a symmetric
setting in which al l three …rms have the same costs, global free trade emerges as the unique stable equilibrium
wi th b ilateral agreements are possible or not. However, the possib ility of bilateral agreements is neces sary for the
achievement of global free trade in a setting in which one …rm has much hi gher production costs than the others.
45
world markets, or if the governments have political motivations. With or without grand coalition
superadditivity, there can be positive or negative coalition externalities. Payo¤s are characterized
by positive coalition externalities when the welfare of a country that is excluded from a PTA is
higher than it would be without the bilateral agreement. Negative coalition externalities imply just
the opposite.
With grand coalition superadditivity and transferable utility, the agenda setter always prefers
a multilateral agreement to any alternative. Moreover, it can achieve such an outcome by o¤ering
payo¤s to the others that make them indi¤erent between joining and not. The question remains,
How do the structure of coalition externalities a¤ect the agenda sender’s choice between simulta-
neous and sequential negotiation? Aghion et al. show that, if at least one coalition externality is
negative, the agenda setter will opt to negotiate sequentially. First it will form an agreement with
the country (if any) that faces a positive externality. Then it will invite the participation of the
remaining country. By the time the second country receives its o¤er, its fallback position is the
lower level of welfare that it would su¤er if e xcluded from a bilateral pact. So, the agenda setter
can extract surplus from this country by con fronting it with an inferior default option.
Even if the payo¤s do not satisfy grand coalition superadditivity, a sequential negotiation process
might lead to global free trade. A multilateral agreement can be reached even at a cost in terms of
the collective (economic or political) welfare of the three countries unde r conditions that the authors
describe. The outcome becomes possible, because the agenda setter’s option to form a bilateral
agreement …rst with one country allows it to extract enough surplus from the other that it is willing
to begin these negotiations in preference to the status quo. More s peci…cally, let C represent the
agenda setter and A and B represent the other two countries. Suppose C prefers the status quo
with no trade agreements to global free trade. In a simultaneous, multilateral negotiation, C must
o¤er A and B their status quo payo¤s. If the sum of the three status quo payo¤s exceeds the joint
welfare under a multilateral trade agreement, then C will not make any o¤ers under multilateral
bargaining, and the status quo will prevail. But suppose C has already negotiated a bilateral
agreement with A and that the coalition externalities for B are negative. Then C will not need to
o¤er B as much as its status quo level of welfare in order to induce it to join the existing agreement.
With the lesser payment that is needed, C might prefer to have B join once an agreement with
A has been established . Moreover, C might prefer this outcome (with the smaller payment to B)
than what it can attain in the status quo. In these circumstances, C will approach A …rst and then
B, and the bilateral agreement with A serves as a building block for a multilateral agreement with
both A and B.
38
The authors develop in their appendix an example of an e conomic setting with
the requisite payo¤ structure.
38
An additional condition is needed to e nsure t hat c prefers to approach a for a bilateral agreement before nego-
tiating with b. Of course, there are analogous conditions on the payo¤s for which multilateral free trade is achieved
after an initial bilateral agreement between c an d b.
46
6 Conclusion
This chapter has reviewed research that asks about the purpose of trade agreements. Why do
governments willingly give up their sovereign rights to set trade policies and enter into agreements
that restrict their choices? A unifying theme in much of the literature is that they do so in order
to help internalize international externalities. No matter what the governments’ objectives, be
they the welfare of the aggregate polities or of particular cons tituent groups, or even the interests
and well-being of the politicians themselves, interdependence in the trading system implies that
any government’s actions a¤ect outcomes abroad. Each government would like others to take its
concerns into account when setting policy. The only way to secure s uch an outcome is to display a
willingness to take account of other governments’concerns when setting ones own policies.
The literature has u sef ully identi…ed a number of international externalities that can arise in
di¤erent market settings and with di¤erent political institutions. Less fruitful, in my opinion, have
been the e¤orts to pin labels on these externalities. Is the fundamental purpose of the agreement
to eliminate manipulation of the terms of trade or to ensure that domes tic exporters are granted
satisfactory access to foreign markets? It is not clear to me why this distinction is important, so long
as we understand that noncooperative policy setting gives rise to inferior outcomes. We do want
to understand why trade agreements have the features they do and what provisions must be added
or mod i…ed to generate e¢ cient outcomes and mutual gain. In my opinion, future research e¤ort
would be more productively spent understanding and improving the design of trade agreements
(i.e., extending the literature reviewed in Chapter 8) than in worrying about the words that best
describe the purpose of trade agreements.
47
References
[1] Aghion, P., Antràs, P., Helpman, E., 2007. Negotiating free trade. Journal of International
Economics 73 (1), 1-30.
[2] Antràs, P., Staiger, R.W.,2012. O¤shoring and the role of trade agreements. American Eco-
nomic Review 102 (7), 3140-3183.
[3] Ashley, P., 1904. Modern tari¤ history: Germany, United States, France. John Murrary, Lon-
don.
[4] Aumann, R.J., 1985. Repeated games. In: Feiwel, G.R. (Ed.), Issues in Contemporary Micro-
economics and Welfare. State University of New York Press, Albany..
[5] Bagwell, K., Staiger, R.W., 1999. An economic the ory of GATT. American Economic Review
89 (1), 215-248.
[6] Bagwell, K., Staiger, R.W., 2001. Domestic policies, national sovereignty and international
economic institutions. Quarterly Journal of Economics 116 (2), 519-562.
[7] Bagwell, K., Staiger, R.W., 2002. The Economics of the World Trading System. MIT Press,
Cambridge MA.
[8] Bagwell, K., Staiger, R.W., 2006. Will international rules on subsidies disrupt the world trading
system? American Economic Review 96 (3), 877-895.
[9] Bagwell, K., Staiger, R.W., 2012a. Pro…t shifting and trade agreements in imperfectly com-
petitive markets. International Economic Review 53 (4), 1067-1104.
[10] Bagwell, K., Staiger, R.W., 2012b. The economics of trade agreements in the linear Cournot
delocation model. Journal of International Economics 88 (1), 32-46.
[11] Bagwell, K., Staiger, R.W., 2015. Delocation and trade agreements in imperfectly competitive
markets. Research in Economics 69 (2), 132-156.
[12] Bagwell, K., Staiger, R.W., 2016. The design of trade agreements. In: Bagwell, K., Staiger
R.W., Handbook of Commercial Policy. North Holland, Amsterdam.
[13] Baldwin, R.E., 1993. A Domino Theory of Regionalism. NBER Working Paper 4465, National
Bureau of Economic Research.
[14] Baldwin, R.E., Venables, A.J., 1995. Regional economic integration. In: Grossman, G.M.,
Rogo¤, K. (Eds.), Handbook of International Economics, vol. 3. North-Holland, Amsterdam.
[15] Barrett, S., 2005. The theory of international environmental agreements. In: Mäler, K.-G.,
Vincent, J.R., Handbook of Environmental Economics, vol. 3. North Holland, Ams terdam..
48
[16] Bhagwati, J., 1991. The World Trading System at Risk. Princeton University Press, Princeton
NJ.
[17] Bhagwati, J., 1995. U.S. trade policy: The infatuation with free trade areas. In: Bhagwati, J.,
Krueger, A.O. (Eds.), The Dangerous Drift to Preferential Trade Agreements. The AEI Press,
Washington DC..
[18] Bond, E.W., 1990. The optimal tari¤ structure in higher dimensions. International Economic
Review 31 (1), 103-116.
[19] Bond, E.W., Park, J.H., 2002. Gradualism in trade agreements with asymmetric countries.
Review of Economic Studies 69 (2), 379-406.
[20] Bond, E.W., Syropoulos, C., 1996. The size of trading blocks: market power and world welfare
e¤ects. Journal of International Economics 40, 412-437.
[21] Brander, J., Spencer, B., 1984a. Trade warfare: Tari¤s and cartels. Journal of International
Economics 16, 227-242.
[22] Brander, J., Spen cer, B., 1984b. Tari¤ protection and imperfect competition. In: Kierzkowski,
H. (Ed.), Monopolistic Competition and International Trade. Clarendon Press, Oxford.
[23] Brander, J., Spencer, B., 1985. Export subsidies and market share rivalry. Journal of Interna-
tional Economics 18, 83-100.
[24] Brou, D., Ruta, M., 2013. A commitment theory of subsidies. The B.E. Journal of Economic
Analysis and Policy.
[25] Chen, M.X., Joshi, S., 2010. Third-country e¤ects on the formation of free trade agreements.
Journal of International Economics 82 (2), 238-248.
[26] Destler, I.M., 2005. American Trade Politics, 4th Edition. Institute for International Eco-
nomics, Washington DC.
[27] Dixit, A.K., 1985. Tax policy in open economies. In: Auerbach, A., Feldstein, M.J. Handbook
of Public Economics, vol. 1. North Holland, Amsterdam, 313-374.
[28] Dixit, A.K., 1987. Strategic aspects of trade policy. In: T.F. Bewley (Ed.), Advances in Eco-
nomic Theory: Fifth World Congress. Cambridge University Press, Cambridge..
[29] Dixit, A.K., Norman, V.D., 1986. Gains from trade without lump-sum compens ation. Journal
of International Economics 21, 111-122.
[30] Eaton, J., Grossman, G.M., 1986. Optimal trade and industrial policy under oligopoly. Quar-
terly Journal of Economics 101 (2), 383-406.
49
[31] Ethier, W.J., 2007. The theory of trade policy and trade agreements: A critique. European
Journal of Political Economy 23, 605–623.
[32] Ethier, W.J., 2013. The trade-agreement embarrassment. Journal of East Asian Economic
Integration 17, 243-260.
[33] Fukushima, Kim, 1989. Welf are improving tari¤ changes: A case of many goods and many
countries. Journal of International Economics 26, 383-388..
[34] Graaf, J., 1949. On optimum tari¤ structures. Review of Economic Studies 17, 47-59.
[35] Grossman, G.M., Helpman, E., 1994. Protection for sale. American Economic Review 84 (4),
833-850.
[36] Grossman, G.M., Helpman, E., 1995a. The politics of free-trade agreements. American Eco-
nomic Review 85 (4), 667-690.
[37] Grossman, G.M., Helpman, E., 1995b. Trade wars and trade talks. Journal of Political Econ-
omy 103 (4), 675-708.
[38] Grossman, G.M., Helpman, E., 1996. Electoral competition and special interest politics. Re-
view of Economic Studies 63 (2), 265-286.
[39] Grossman, G.M., Horn, H., 2013. Why the WTO? An introduction to the economics of trade
agreements. In: Horn, H.,P.C. Mavroidis (Eds.), Legal and Economic Principles of World Trade
Law. Cambridge University Press, Cambridge.
[40] Grossman, G.M., Lai, E.L.-C., 2004. International protection of intellectual property. American
Economic Review 94 (5), 1635-1653.
[41] Hillman, A.L., Moser, P., 1995. Trade liberalization as politically optimal exchange of market
access. In: Canzoneri, M., Ethier, W.J., Grilli, V. (Eds.), The new transatlantic economy.
Cambridge University Press, Cambridge.
[42] Hillman, A.L., Long, N.V., Moser, P., 1995. Modeling reciprocal trade liberalization: The
political-economy and national welfare perspectives. Sw iss Journal of Economics and Statistics
131, 503-515.
[43] Hudec, R.E. 1993. ‘Circumventing’democracy: The political morality of trade negotiations.
International Law and Politics 25, 311-322.
[44] Johnson, H.G., 1953. Optimum tari¤s and retaliation. Review of Economic Studies 21 (2),
142-153.
[45] Katrik, H., 1977. Multi-national monopolies and commercial policy. Oxford Economic Papers
29 (2), 283-291.
50
[46] Kemp, M., 1967. Notes on the theory of optimal tari¤s. Economic Record 43, 395-403.
[47] Kemp, M., Wan, Jr., H., 1976. An elementary proposition concerning the formation of customs
unions. In: Kemp, M. (Ed.), Three Topics in the Theory of International Trade: Distribution,
Welfare and Unce rtainty. North Holland, Amsterdam.
[48] Kennan, J., Riezman, R. Optimum tari¤ equilibria with customs unions. The Canadian Journal
of Economics 23 (1), 70-93.
[49] Krishna, P., 1998. Regionalism versus multilateralism.: A political economy approach. Quar-
terly Journal of Economics 113 (1), 227-251.
[50] Krishna, P., 2005. The economics of preferential trade agreements. In: Choi, E.K., Harrigan,
J. (Eds.), Handbook of International Trade. Blackwell, Oxford.
[51] Krishna, P., Bhagwati, J., 1997. Necessarily welfare improving customs unions with industrial-
ization constraints: The Cooper-Massell-Johnson-Bhagwati proposition. Japan and the World
Economy 9, 441-446.
[52] Krugman, P.R., 1980. S cale economies, product di¤erentiation, and the pattern of trade. Amer-
ican Economic Review 70 (5), 950-959.
[53] Krugman, P.R., 1991a. Is b ilateralism bad? In: Helpman, E., Razin, A. (Eds.), International
Trade and Trade Policy. MIT Press, Cambridge MA, 9-23.
[54] Krugman, P.R., 1991b. The move toward free trade zone s. In: Policy Implications of Trade
and Currency Zones: A Symposium Sponsored by the Federal Reserve Bank of Kansas City.
Federal Reserve Bank of Kansas City, Kansas City.
[55] Kydland, F.E., Prescott, E.C., 1977. Rules rather than discretion: The inconsistency of optimal
plans. Journal of Political Economy 85 (3), 473-492.
[56] Limão, N., 2016. Preferential trade agreeme nts. In: Bagwell, K., Staiger R.W. Handbook of
Commercial Policy. North Holland, Amsterdam.
[57] Limão, N., Tovar, P., 2011. Policy choice: Theory and evidence from commitment via interna-
tional trade agreements. Journal of International Economics 85 (2), 186-205.
[58] Maggi, G., 1999. The role of multilateral institutions in international trade cooperation.
American Economic Review 89 (1) 190-2014.
[59] Maggi, G., 2014. International trade agreements. In: Gopinath, G., Helpman, E., Rogo¤, K.
(Eds.), Handbook of International Econ omics, vol. 4. North Holland, Amsterdam..
[60] Maggi, G., Rodriguez-Clare, A., 1998. The value of trade agreements in the presence of political
pressures. Journal of Political Economy 106 (3), 574-601.
51
[61] Maggi, G., Rodriguez-Clare, A., 2007. A political-economy theory of trade agreements. Amer-
ican Economic Review 97 (4), 1374-1406.
[62] Mayer, W., 1981. Theoretical considerations on negotiated tari¤ adjustments. Oxford Eco-
nomic Papers 33 (1), 135-153.
[63] Mitra, D., 1999. Endogenous lobby formation and endogen ous protection: A long-run model
of trade policy determination. American Economic Review 89 (5), 1116-1134.
[64] Mitra, D., 2002. Endogenous political organization and the value of trade agreements. Journal
of International Economics 57 (2), 473-485.
[65] Ohyama, M., 1972. Trade and welfare in general equilibrium. Keio Economic Studies 9, 37-73.
[66] Ornelas, E., 2005a. Endogenous free trade agreements and the multilateral system. Journal of
International Economics 67 (2), 471-497.
[67] Ornelas, E., 2005b. Trade creating free trade areas and the undermining of multilateralism.
European Economic Review 49 (7), 1717-1735.
[68] Ossa, R., 2011. A new trade theory of GATT/WTO negotiations. Journal of Political Economy
119 (1), 122-152.
[69] Panagariya, A., 2000. Preferential trade liberalization: The traditional theory and new devel-
opments. Journal of Economic Literature 38 (2), 287-331.
[70] Panagariya, A., Krishna, P., 2002. On n ec ess arily welfare-enhancing free trade areas. Journal
of International Economics 57, 353-367.
[71] Park, J.-H., 2011. Enforcing international trade agreements with imperfect private monitoring.
Review of Economic Studies 78 (3), 1102-1134.
[72] Pauwelyn, J., 2008. New trade politics for the 21st Century. Journal of International Economic
Law 11 (3), 559-573.
[73] Regan, D.H., 2006. What are trade agreements for? Two con‡icting stories told by economists,
with a lesson for lawyers. Journal of International Economic Law 9, 951–988.
[74] Regan, D.H., 2015. Explaining trade agreements: The practitioners’ story and the standard
model. World Trade Review 14 (3), 391-417.
[75] Richardson, M., 1993. Endogenous protection and trade diversion. Journal of International
Economics 34 (3-4), 309-324.
[76] Richardson, M., 1995. On the interpretation of the Kemp/Wan theorem. Oxford Economic
Papers 47 (4), 696-703.
52
[77] Saggi, K., Yildiz, H.M., 2010. Bilateralism, multilateralism, and the quest for global free trade.
Journal of International Economics 81, 26-37.
[78] Saggi, K., Yildiz, H.M., 2011. Bilateral trade agreements and the feasibility of multilateral free
trade. Review of International Economics 19 (2), 356-373.
[79] Saggi, K., Woodland, A., Yildiz, H.M., 2013. On the relationship between preferential and
multilateral trade liberalization: Th e case of customs unions. American Economic Journal:
Microeconomics 5 (1), 63-99.
[80] Staiger, R.W., Tabellini, 1987. Discretionary trade policy and excessive protection. American
Economic Review 77 (5), 823-837.
[81] Svedberg, P., 1979. Optimal tari¤ policy on imports from multinationals. Economic Record 55
(1), 64-67.
[82] Turunen-Red, A.H., and Woodland, A.D., 1991. Strict Pareto-improving reform of tari¤s.
Econometrica 59, 1127-1152.
[83] Venables, A.J., 1985. Trade and trade policy with imperfect competition: The case of identical
products and free entry. Journal of International Economics 19, 1-20.
[84] Venables, A.J., 1987. Trade and trade policy with di¤erentiated products: A Chamberlinian-
Ricardian model. Economic Journal 97, 700-717.
53
