Trade Agreements as EndogenouslyIncomplete Contracts

Henrik Horn (Research Institute of Industrial Economics, Stockholm)

Giovanni Maggi (Princeton University)

Robert W. Staiger (Stanford University and Wisconsin)

Introduction

Real-world trade agreements display an interesting combination of rigidity anddiscretion. Consider the GATT/WTO:

1. Trade instruments bound; domestic instruments largely left to discretion,but must satisfy National Treatment, and now (WTO) regulation of subsidies.

2. Bindings rigid, but with escape clauses (e.g. GATT Article XIX).

3. Bindings stipulate ceilings, so governments have downward discretion.

Why? Conjecture: Can be understood from incomplete contracts perspective.

A number of possible sources of incompleteness. We focus on two features offundamental importance to trade negotiators:

(1) Wide array of trade-relevant policies: border measures but also internalmeasures. Controlling opportunism requires comprehensive policy coverage.

(2) Uncertainty about future economic/political conditions. This calls for agree-ments that are highly contingent.

Trade-law literature emphasizes contracting implications of costs associatedwith these features:

...The standard trade policy rules could deal with the common type of tradepolicy measure governments usually employ to control trade. But trade can alsobe a¤ected by other domestic measures, such as product safety standards,having nothing to do with trade policy. It would have been next to impossibleto catalogue all such possibilities in advance. (Hudec, 1990).

...Many contracts are negotiated under conditions of considerable complexityand uncertainty, and it is not economical for the parties to specify in advancehow they ought to behave under every conceivable contingency ... The partiesto trade agreements, like the parties to private contracts, enter the bargainunder conditions of uncertainty. Economic conditions may change, the strengthof interest group organization may change, and so on. (Schwartz and Sykes,2002).

We introduce contracting costs explicitly into economic analysis of trade agree-ments, and study their implications for the structure of the optimal (incomplete)agreement. We argue that contracting costs can help explain some of the corefeatures of the GATT/WTO.

The Model

Two countries, H and F. Two non-numeraire goods, 1 and 2, plus numeraire:partial-equilibrium analysis.

H a natural importer of good 1/exporter of good 2. Sectors 1 and 2 are mirror-image, so focus on sector 1. Demand: D(p) = p; D(p) = p.Supply: X(q) = q; X(q) = q.

H chooses tari¤ , separate consumption taxes on domestic and foreign prod-ucts (th and tf ), production subsidy (s). F does not intervene in this sector.

The following must hold: q = p; p = p tf ; q = p th + s.

More compactly: p = p + T ; q = p + T + S; where T + tf andS s th.

Market clearing/arbitrage: p = p(T; S); q = q(T; S); p = q = p(T; S):

Importing country H experiences a negative consumption externality equal to D with > 0. (Later consider production externality/pol.ec. forces).

Governments maximize welfare, so (still maintaining our focus on sector 1):

W = CS + PS + T M S X D; W = CS + PS:

E¢ cient and Nash equilibrium policies

Globally e¢ cient policies maximize WG W +W , yielding

T eff = ; Seff = .

Nash equilibrium policies:

TNE = +E

+ = +

p

SNE = :

Note: TNE > T eff ; SNE = Seff : Nash trade taxes ine¢ ciently high; Nashdomestic instruments set at e¢ cient levels.

Uncertainty

To simplify, focus for now on one-dimensional uncertainty.

Consider two possible sources of uncertainty: consumption externality ( ) andimport demand level ().

Timing: (1) The agreement is drafted; (2) Uncertainty is resolved; (3) Policiesare chosen subject to the constraints set by the agreement.

Let EWG() denote expected gross-of-contracting-costs global welfare.

The costs of contracting

We focus mostly on instrument-based (not outcome-based) agreements.

Key idea: more detailed agreements are more costly (similar to Battigalli andMaggi, 2002).

cp: cost of including a policy variable (; tf ; s; th). cs: cost of including astate variable ( , ). Cost of writing an agreement: C = cs ns + cp np;with ns (np) the number of state (policy) variables in the agreement.

Note: Basic results go through if C is increasing in ns and np. Modeling of Cis slightly more general than Battigalli and Maggi.

Let cp c; cs k c. Fix k. Then C = c (np + k ns) c m, where mis rough measure of complexity of the agreement.

Optimal agreements

An optimal agreement maximizes expected net global welfare, ! C.

Before imposing more structure on uncertainty, three basic results.

First (Lemma1): nec. and su¤. conditions for an agreement to be optimalfor a range of contracting costs. Second (Lemma 2): complexity of optimalagreement rises as level of contracting cost falls. Figure 1.

m

••

•

•

••

•

••

•

0

Ω

ω ( $ )A

$c$A

~A

•

$m′m ′′m~m

To state third result, recall: T = + tf ; S = s th. Hence T and S therelevant policy variables, with cost 2c for each.

Proposition 1: An agreement that constrains the e¤ective subsidy S (evenin a state-contingent way) while leaving the import tax T to discretion can-not improve over the Nash equilibrium, and therefore cannot be an optimalagreement.

Broad intuition: contracting over S alone is useless because ine¢ ciency in theNE concerns T , not S:

In world of costless contracting, Proposition 1 irrelevant. But as we next show,gains relevance when contracting costly.

Uncertainty about the consumption externality

Assume that can take one of two values: or + .

For now consider only agreements that impose separate equality constraints onT and S (e.g. (T = ) or (S = 10)). Call this set A0.

Note: fFBg agreement is fT = ;S = g, which costs (4 + k) c.

If c is small enough fFBg optimal; for large c empty agreement (yielding NEpayo¤s) is optimal. What happens between these two extremes?

Two ways to save on contracting costs relative to fFBg: agreement can berigid (i.e. non-contingent); and/or it can leave some policies to discretion.

By Proposition 1, can focus on three kinds of agreement (aside from fFBgand f;g): fT; Sg; fT ( )g; and fTg.

Note: fT; Sg rigid; fT ( )g discretion; fTg both rigid and discretionary.

Basic trade-o¤: rigid agreement prevents TOT manipulation, but Pigouvianintervention only on average;discretion allows implicit state-contingency butcreates scope for manipulating TOT.

Optimal sequence of agreements as c increases from zero? By Lemma 2, mustbe a subsequence of (fFBg; fT; Sg; fT ( )g; fTg; f;g).

But fT; Sg and fT ( )g cant be part of the same optimal sequence.

Why? Cost of discretion lower in presence of rigidity, as discretion introducesimplicit state-contingency into a rigid agreement, so rigidity and discretion arecomplementary ; this plus Lemma 1 implies Figure 2a/b.

m

m

T ( )T γ

, T S

FB

T

( )T γ

, T S

FB

•

•

•

•

•

•

•

•

• •

Ω

Ω

0

0

2 2 + k 4 4 + k

2 2 + k 4 4 + k

Proposition 2: Assume that only is uncertain. There exist scalars c1, c2, c3and c4 with 0 < c1 c2 c3 c4 <1 such that the optimal agreement inA0 is:(a) the fFBg agreement for c 2 (0; c1);(b) of the form fT; Sg for c 2 (c1; c2);(c) of the form fT ( )g for c 2 (c2; c3);(d) of the form fTg for c 2 (c3; c4); and(e) the empty agreement for c > c4.Moreover, either c2 = c1 or c3 = c2 (or both).

Implications. Non-empty trade agreements must always include commitmentsover T , but not necessarily S. As c rises, several possible paths: may rstintroduce rigidity, then add discretion; or, may rst introduce discretion, thenadd rigidity; or, could skip rigidity completely; but cant oscillate.

Implications. Very di¤erent from Battigalli-Maggi: reects structure of con-tracting problem and non-separability across policy instruments.

As c rises, which path taken? Depends on the cost of discretion versus the costof rigidity.

Determinants of cost of discretion and rigidity reect simple economic intu-itions.

First, discretion costly to gross global welfare when Hs monopoly power intrade is high, and when S a close substitute for T for TOT manipulation.

Second, rigidity costly to gross global welfare when level of uncertainty is high.

So, for example, if monopoly power high (high ) and/or uncertainty low (low ), then as c rises, rst introduce rigidity, then consider discretion; if S apoor substitute for T (low ), then as c rises, rst introduce discretion, thenconsider rigidity.

Comparative Statics: Focus rst on . Recall that, through the monopolypower e¤ect, increasing increases the cost of discretion.

Proposition 3: As the import demand level increases (holding all other para-meters xed): (i) The optimal degree of discretion decreases, in the sense thatthe number of policy instruments specied in the optimal agreement increases(weakly); (ii) The optimal degree of rigidity decreases (weakly).

At a broad level, Proposition 3(i) suggests: (1) possible explanation for newWTO regulation of domestic subsidies; and (2) possible benet of specialand di¤erential treatment for small/developing countries when it comes tocontracting over domestic subsidies.

Proposition 3(ii) is due to the fact that rigidity and discretion are complemen-tary, hence they tend to move together.

Next focus on the impact of uncertainty. As the degree of uncertainty increases, the optimal agreement tends to become more contingent:

Proposition 4: As the degree of uncertainty increases (holding all otherparameters xed), the optimal agreement may switch from a rigid agreementto a contingent agreement, but not vice-versa.

By itself not surprising. But also a more subtle feature: concerns uncertaintyover a state variable that is directly relevant for the setting of T in the fFBgagreement. As we next demonstrate, e¤ects of increasing uncertainty over statevariables (such as ) that are not directly relevant for the setting of T in thefFBg can be very di¤erent.

Uncertainty about the import demand level

Preview: has no impact on the rst-best levels of T or S. Consequently,implications of uncertainty over very di¤erent from uncertainty over .

Assume can take one of two possible values: + or .

Now the fFBg agreement is fT = ;S = g. Note, this is not a contingentagreement.

We can focus on two types of agreement (aside from fFBg and f;g): fT ()g;and fTg.

New possibility: fT ()g, with T () an increasing function. This has avor ofan escape clause, but rationale is novel: to manage distortions in domesticinstruments introduced by trade agreements.

Intuition: an escape clause is appealing when the agreement leaves discretionover S; allowing for a higher T mitigates the incentive to distort S, and this ismore valuable when (and hence underlying import volume) is higher.

Novel feature: rigidity and discretion now substitutes (cost of discretion higherin presence of rigidity) because rigidity precludes escape clause. Hence, natureof interaction between rigidity and discretion depends on source of uncertainty.

Next, What is the optimal sequence of agreements as c increases?

Proposition 5: Assume that only is uncertain. There exist c1, c2, and c3with 0 < c1 c2 c3 <1 such that the optimal agreement in A0 is:(a) the fFBg agreement fT = ;S = g for c 2 (0; c1);(b) of the form fT ()g for c 2 (c1; c2);(c) of the form fTg for c 2 (c2; c3);(d) the empty agreement for c > c3.

Note: if k su¢ ciently low and su¢ ciently low (but strictly positive), thenescape-clause-type agreement fT ()g is optimal for range of c; optimal con-tract is contingent on the wrong state variable.

Finally, note: cost of discretion convex in , so as rises, optimal agreementmay switch from contingent (fT ()g) to rigid (fT; Sg); could never happenfor , so source of uncertainty matters.

The role of the National Treatment (NT) clause

We now extend the feasible set of agreements by allowing for an NT clause, thatis a constraint th = tf , costing 2c. (For simplicity, return to uncertainty).

Denition: an NT-based agreement is an agreement that includes the NTclause. Let ANT denote the class of NT-based agreements.

Note: for NT-based agreements, the price relationships are: p = p + + t;q = p + + s. Recall for non-NT: p = p + T ; q = p + T + S.

An agreement fNT; ; sg costs less than fFBg and ties down the producerprice wedge qp while leaving the consumer price wedge pp to discretion.This type of discretion is not possible with non-NT agreements.

Under what conditions is an NT-based agreement strictly optimal for a rangeof c?

Proposition 6: Consider the agreement class A0 [ ANT . If > 0 and is su¢ ciently small (so that the consumption tax t is a poor substitute forthe tari¤ ), then there is an intermediate range of c for which the optimalagreement includes the NT clause.

NT-based agreement strictly optimal if low substitutability between t and ,and c lies in intermediate range. Describes world in which NT-based agreementgets close to rst best (fteff = ; eff = 0; seff = 0g) by exploiting implicitstate-contingency associated with discretion over internal taxes.

The role of weak bindings

A weak binding is a constraint of the type T T (), which allows for downwarddiscretion. Can this be an optimal way to save on contracting costs?

Note: weak bindings can potentially improve on strong bindings only if theyare rigid, i.e. of the type T T (e.g. a clause of the type T T ( ) cannotimprove on T = T ( )), because downward discretion allows for some implicitstate-contingency which can then be valuable.

Proposition 7: Suppose the optimal agreement in A0 [ ANT includes somerigid strong binding. Then this agreement can be weakly improved upon by re-placing the rigid strong bindings with rigid weak bindings, and the improvementis strict for some congurations of parameters.

Production externalities and political-economy motives

A production externality in H equal to X with > 0: the fFBg agreementtakes the form fT = 0;S = g, so that now only S is state-contingent.

Implication: appeal of fT ()g similar to fT ()g, not fT ( )g.

A political economy extension with extra weight on producer surplus: thefFBg agreement is fT = 0;S = S(; )g, where S(; ) is increasing inboth arguments.

Implication: fT ()g agreement of particular interest here. A tari¤ response totrade shocks in the presence of distributional concerns, but for the purpose ofmitigating distortions in the discretionary use of S.

Conclusion

Our paper takes a rst step in the analysis of trade agreements as en-dogenously incomplete contracts. We have shown that this perspectiveprovides a novel explanation for:

the emphasis on border measures in real world trade agreements;

the appeal of escape clauses in the presence of surging import de-mand;

the appeal of the National Treatment provision in GATT/WTO;

the emphasis on weak bindings.

Possible directions for future research:

consider a broader class of feasible agreements, and in particular outcome-based agreements (e.g. imposing constraints on trade volumes orprices);

consider export-sector policies;

consider a multi-country setting, to examine the potential appeal ofthe MFN rule and exceptions for FTAs/CUs;

consider a commitment role for trade agreements;

consider the potential appeal of a dispute settlement body, as a mech-anism to complete the incomplete contract;

more explicit modeling of contract costs.