00 american economic review - contracts and technology adoption

8/9/2019 00 American Economic Review - Contracts and Technology Adoption

1/28

Contracts and Technology Adoption

By DARON ACEMOGLU, POL ANTRAS, AND ELHANAN HELPMAN*

We develop a tractable framework for the analysis of the relationship betweencontractual incompleteness, technological complementarities, and technology adop-tion. In our model, a firm chooses its technology and investment levels in contract-ible activities by suppliers of intermediate inputs. Suppliers then choose investmentsin noncontractible activities, anticipating payoffs from an ex post bargaining game.We show that greater contractual incompleteness leads to the adoption of lessadvanced technologies, and that the impact of contractual incompleteness is more

pronounced when there is greater complementary among the intermediate inputs.We study a number of applications of the main framework and show that themechanism proposed in the paper can generate sizable productivity differencesacross countries with different contracting institutions, and that differences incontracting institutions lead to endogenous comparative advantage differences.(JEL D86, O33)

There is widespread agreement that differ-ences in technology are a major source of pro-

ductivity differences across firms, industries,and nations.1 Despite this widespread agree-ment, we are far from an established frameworkfor the analysis of technology choices of firms.In this paper, we take a step in this direction anddevelop a simple model to study the impact ofcontracting institutions, which regulate the re-

lationship between the firm and its suppliers, ontechnology choices.

Our model combines two well-establishedapproaches. The first is the representation oftechnology as the range of intermediate in-puts used by firms; a greater range of interme-diate inputs increases productivity by allowinggreater specialization and thus corresponds tomore advanced technology.2 The second isSanford J. Grossman and Oliver D. Harts

* Acemoglu: Department of Economics, Massachusetts In-stitute of Technology, E52-380b, Cambridge, MA 02139, Na-tional Bureau of Economic Research, Centre for EconomicPolicy Research, and Canadian Institute for Advanced Re-

search (e-mail: [email protected]); Antras: Department of Eco-nomics, Harvard University, Littauer 319, Cambridge, MA02138, NBER, and CEPR (e-mail: [email protected]);Helpman: Department of Economics, Harvard University, Lit-tauer 229, Cambridge, MA 02138, NBER, CEPR, and CIAR(e-mail: [email protected]). A previous version of thispaper was circulated under the title Contracts and the Divisionof Labor. We thank Gene Grossman, Oliver Hart, KalinaManova, Damian Migueles, Giacomo Ponzetto, Richard Rog-erson, Rani Spiegler, two anonymous referees, participants inthe Canadian Institute for Advanced Research and MinnesotaWorkshop in Macroeconomic Theory conferences, and semi-nar participants at Harvard University, MIT, Tel Aviv Univer-sity, Universitat Pompeu Fabra, University of Houston,University of California, San Diego, Southern MethodistUniversity, Stockholm School of Economics, StockholmUniversity (IIES), ECARES, University of ColoradoBoul-der, the Kellogg School, New York University, Universityof British Columbia, Georgetown University, the WorldBank, University of Montreal, Brandeis University, Univer-sity of Wisconsin, UCBerkeley, Haifa University, Syra-cuse University, Universitat Autonoma de Barcelona, DukeUniversity, Columbia University, and Cornell Universityfor useful comments. We also thank Davin Chor, AlexandreDebs, and Ran Melamed for excellent research assistance.Acemoglu and Helpman thank the National Science Foun-dation for financial support. Much of Helpmans work forthis paper was done when he was Sackler Visiting Professorat Tel Aviv University.

1 Among others, see Peter J. Klenow and Andres Rodri-guez-Clare (1997), Robert E. Hall and Charles I. Jones(1999), or Francesco Caselli (2005) for countries, and Tor J.Klette (1996), Zvi Griliches (1998), John Sutton (1998), orKlette and Samuel S. Kortum (2004) for firms.

2 See, among others, Wilfred J. Ethier (1982), Paul M.Romer (1990), and Gene M. Grossman and Helpman (1991)for previous uses of this representation. See also the text-book treatment in Robert J. Barro and Xavier Sala-i-Martin(2003). There is a natural relationship between this view oftechnology and the division of labor within a firm. Weinvestigated this link in the earlier version of the currentpaper, Acemoglu, Antras, and Helpman (2005), and do notelaborate on it here.

916

Reprinted from the AMERICAN ECONOMIC REVIEW, copyright 2007 by the American Economic

Association. All rights reserved.


2/28

(1986) and Hart and John H. Moores (1990)approach to incomplete-contracting models ofthe firm. We study technology choice of firmsunder incomplete contracts, and extend Hartand Moores framework by allowing contractsto be partially incomplete. This combinationenables us to investigate how the degree ofcontractual incompleteness and the extent oftechnological complementarities between inter-mediate inputs affect the choice of technology.

In our baseline model, a firm decides ontechnology (on the range of specialized inter-mediate goods), recognizing that a more ad-vanced technology is more productive, but alsoentails a variety of costs. In addition to thedirect pecuniary costs of engaging more suppli-

ers (corresponding to the greater range of inter-mediate inputs), a more advanced technologynecessitates contracting with more suppliers.All of the activities that suppliers undertake arerelationship-specific, and a fraction of those isex ante contractible, while the rest, as in thework by Grossman-Hart-Moore, are nonverifi-able and noncontractible. The fraction of con-tractible activities is our measure of the qualityof contracting institutions.3 Suppliers are con-tractually obliged to perform their duties in thecontractible activities, but they are free to

choose their investments in noncontractible ac-tivities and to withhold their services in theseactivities from the firm. This combination ofnoncontractible investments and relationship-specificity leads to an ex post multilateral bar-gaining problem. As in Hart and Moore (1990),we use the Shapley value to determine the di-vision of ex post surplus between the firmand its suppliers. We derive an explicit solutionfor this division of surplus, which enables usto develop a simple characterization of theequilibrium.

A suppliers expected payoff in the bargain-ing game determines her willingness to invest in

the noncontractible activities. Since she is notthe full residual claimant of the output gainsderived from her investments, she tends to un-derinvest. Greater contractual incompletenessthus reduces supplier investments, making moreadvanced technologies less profitable. Further-more, a greater degree of technological comple-mentarity reduces the incentive to choose moreadvanced technologies. Although greater tech-nological complementarity increases equilib-rium ex post payoffs to every supplier, it alsomakes their payoffs less sensitive to their non-contractible investments, discouraging invest-ments and, via this channel, depressing theprofitability of more advanced technologies.

An advantage of our framework is its relative

tractability. The equilibrium of our model canbe represented by a reduced-form profit func-tion for firms given by

(1) AZ FN CN w0N,

where N represents the technology level, C(N)is the cost of technology N, A is a measure ofaggregate demand or the scale of the market,and w0N corresponds to the value of the Nsuppliers outside options. F(N) is an increasingfunction that captures the positive effect of

choosing more advanced technologies on reve-nue. The effects of contractual incompletenessand technological complementarity are summa-rized by the variable Z. This variable affectsproductivity and is decreasing in the degreeof contract incompleteness and technologicalcomplementarity. Moreover, the elasticity of Zwith respect to the quality of contracting insti-tutions is higher when there is greater comple-mentarity between intermediate inputs. This lastresult has important implications for equilib-rium industry structure and the patterns of com-

parative advantage, because it implies thatsectors (firms) with greater complementaritiesbetween inputs are more contract dependent.

We use this framework to show that thecombination of contractual imperfections andtechnology choice (or adoption) may have im-portant implications for cross-country incomedifferences, equilibrium organizational forms,and patterns of international specialization andtrade.

First, we show that our baseline model cangenerate sizable differences in productivity from

3 Eric S. Maskin and Jean Tirole (1999) questionwhether the presence of nonverifiable actions and unfore-seen contingencies necessarily leads to incomplete con-tracts. Their argument is not central to our analysisbecause it assumes the presence of strong contractinginstitutions (which, for example, allow contracts to spec-ify sophisticated mechanisms), while in our model afraction of activities may be noncontractible, not becauseof technological reasons but because of weak contract-ing institutions.

917VOL. 97 NO. 3 ACEMOGLU ET AL.: CONTRACTS AND TECHNOLOGY ADOPTION


3/28

differences in the quality of contracting institu-tions. In particular, using a range of reasonablevalues for the key parameters of the model, wefind that when the degree of technologicalcomplementarity is sufficiently high, relativelymodest changes in the fraction of activities thatare contractible can lead to large changes inproductivity.

Second, we derive a range of implicationsabout the equilibrium organizational form.Our results here show that the combination ofweak contracting institutions and credit mar-ket imperfections may encourage greater ver-tical integration.

Third, we present a number of general equi-librium applications of this framework. The

general equilibrium interactions result from thefact that an improvement in contracting institu-tions does not increase the choice of technologyin all sectors (firms). Instead, more advancedtechnologies are chosen in the more contract-dependent sectors. We show that in the contextof an open economy, this feature leads to anendogenous structure of comparative advan-tage. In particular, among countries with iden-tical technological opportunities (productionpossibility sets), those with better contractinginstitutions specialize in sectors with greater

complementarities among inputs. These predic-tions are consistent with the recent empiricalresults presented in Nathan Nunn (2006) andAndrei A. Levchenko (2003).

As mentioned above, our work is related totwo strands of the literature. The first investi-gates the determinants of firm-level technology(including the division of labor) and includes,among others, Gary S. Becker and Kevin M.Murphy (1992) and Xiaokai Yang and Jeff Bor-land (1991) on the impact of the extent of themarket on the division of labor, and Romer

(1990) and Grossman and Helpman (1991)on endogenous technological change. Noneof these studies investigates the effects of con-tracting institutions on technology choice. Thesecond literature deals with the internal organi-zation of the firm. It includes the papers byGrossman-Hart-Moore discussed above, as wellas those of Benjamin Klein, Robert G. Craw-ford, and Armen A. Alchian (1978) and OliverE. Williamson (1975, 1985), who emphasizeincomplete contracts and hold-up problems.Here, the papers by Lars A. Stole and Jeffrey

Zwiebel (1996a, b) and Yannis Bakos and ErikBrynjolfsson (1993) are most closely related toours. Stole and Zwiebel consider a relationshipbetween a firm and a number of workers whosewages are determined by ex post bargainingaccording to the Shapley value. They show howthe firm may overemploy in order to reducethe bargaining power of the workers, anddiscuss the implications of this framework fora number of organizational design issues.Stole and Zwiebels framework does not haverelationship-specific investments, however,which is at the core of our approach, and theydo not discuss the effects of the degree ofcontractual incompleteness and the degree ofcomplementarity between inputs on the equi-

librium technology choice.

4

Finally, our paper is also related to the liter-ature on the macroeconomic implications ofcontractual imperfections. A number of papers,most notably Erwan Quintin (2001), Pedro S.Amaral and Quintin (2005), Andres Erosa andAna Hidalgo (2005), and Rui Castro, Gian LucaClementi, and Glenn MacDonald (2004), inves-tigate the quantitative impact of contractual andcapital market imperfections on aggregate pro-ductivity.5 Our approach differs from these pa-pers since we focus on technology choice and

relationship-specific investments, and becausewe develop a tractable framework that can beapplied in a range of problems. Although we donot undertake a detailed calibration exercise asin these papers, the results in Section IVA sug-gest that the economic mechanism developed inthis paper can lead to quantitatively large ef-fects. Finally, Levchenko (2003), Arnaud Costi-not (2004), Nunn (2006), and Antras (2005)also generate endogenous comparative advan-tage across countries from differences in con-tractual environments. Among these papers,

Costinot (2004) is most closely related, since healso develops a model of endogenous compar-

4 Another related paper is Olivier J. Blanchard and Mi-chael Kremer (1997), which studies the impact of inefficientsequential contracting between a firm and its suppliers in amodel where output is a Leontief aggregate of the inputs ofthe suppliers.

5 Acemoglu and Fabrizio Zilibotti (1999), David Marti-mort and Thierry Verdier (2000, 2004), and Patrick Fran-cois and Joanne K. Roberts (2003) study the qualitativeimpact of changes in the internal organization of firms oneconomic growth.

918 THE AMERICAN ECONOMIC REVIEW JUNE 2007


4/28

ative advantage based on specialization, thoughhis approach is more reduced-form than ourmodel.

The rest of the paper is organized as fol-lows. Section I introduces the basic environ-ment. Section II characterizes the equilibriumwith complete contracts. Section III introducesincomplete contracts into the framework of Sec-tion I, characterizes the equilibrium, and derivesthe major comparative static results. Section IVconsiders a number of applications of our basicframework. Section V concludes. Proofs of themain results are provided in the Appendix.

I. Technology and Payoffs

Consider a profit-maximizing firm facing ademand function q Ap1/(1) for its finalproduct, where q denotes quantity and p denotesprice. The parameter (0, 1) determines theelasticity of demand, while A 0 determinesthe level of demand or the market size. Thefirm treats the demand level A as exogenous.This form of demand can be derived from aconstant elasticity of substitution preferencestructure for differentiated products (see Sec-tion IVC) and it generates a revenue function

(2) R A1

q

.

Production depends on the technology choice ofthe firm. More advanced (more productive)technologies involve a greater range of interme-diate goods and thus a higher degree of special-ization. The technology level of the firm isdenoted by N

, and for each j [0, N],

X(j) is the quantity of intermediate input j.Given technology N, the production function ofthe firm is

(3) q N 11/0N

X(j) dj1/

,

0 1, 0.

A number of features of this production func-tion are worth noting. First, determines thedegree of complementarity between inputs;since (0, 1), the elasticity of substitutionbetween them, 1/(1 ), is always greater thanone. Second, we follow Jean-Pascal Benassy

(1998) in introducing the term N11/ infront of the integral, which allows us to sepa-rately control the elasticity of substitution be-tween inputs and the elasticity of output withrespect to the level of the technology. To seethis, consider the case where X(j) X for all j.Then the output of technology Nis q N1X.Consequently, both output and productivity (de-fined as either q/(NX) or q/N) are independent of and depend positively on N, with elasticitydetermined by the parameter .6

There is a large number of profit-maximizingsuppliers that can produce the necessary inter-mediate goods, each with the same outside op-tion w0. For now, w0 is taken as given, but it willbe endogenized in Section IVC. We assume that

each intermediate input needs to be produced bya different supplier with whom the firm needs tocontract.7

A supplier assigned to the production of anintermediate input needs to undertake relation-ship-specific investments in a unit measure of(symmetric) activities. There is a constant mar-ginal cost of investment cx for each activity.

8

The production function of intermediate inputsis Cobb-Douglas and symmetric in the activi-ties:

(4) Xj exp01

ln x(i, j) di,where x(i, j) is the level of investment in activityi performed by the supplier of input j. Thisformulation will allow a tractable parameteriza-tion of contractual incompleteness in SectionIII, whereby a subset of the investments neces-sary for production will be nonverifiable andthus noncontractible.

Finally, we assume that adopting (and using)

a technology N involves costs C(N), and im-pose:

6 In contrast, with the standard specification of the CESproduction function, without the term N11/ in front(i.e., 1/ 1), total output is q N1/X, and the twoelasticities are governed by the same parameter, .

7 A previous version of the paper, Acemoglu, Antras,and Helpman (2005), also endogenized the allocation ofinputs to suppliers using an augmented model with addi-tional diseconomies of scope.

8 One can think of cx as the marginal cost of effort; seethe formulation of this effort in utility terms in Section IVC.



5/28

ASSUMPTION 1:

(a) For all N 0, C(N) is twice continuouslydifferentiable, with C(N) 0 and C(N) 0.

(b) For all N 0, NC(N)/[C(N) w0] [( 1) 1]/(1 ).

The first part of this assumption is standard;costs are increasing and convex. The secondpart will ensure a finite and positive choice ofN.

Let the payment to supplier j consist of twoparts: an ex ante payment (j) before theinvestment levels x(i, j) take place, and a pay-ment s(j) after the investments. Then, the payoff

to supplier j, also taking account of her outsideoption, is

(5)

x j max(j) s(j) 0

1

cxx(i, j) di, w0.Similarly, the payoff to the firm is

(6) R0N

j sj dj CN,

where R is revenue and the other two terms onthe right-hand side represent costs. Substituting(3) and (4) into (2), revenue can be expressed as

(7) RA1 N 11/

0N

exp

01

ln x(i, j) di

dj

/

.

II. Equilibrium under Complete Contracts

As a benchmark, consider the case of completecontracts (the first best from the viewpoint ofthe firm). With complete contracts, the firm hasfull control over all investments and pays eachsupplier her outside option. In analogy to ourtreatment below of technology adoption underincomplete contracts, consider a game form

where the firm chooses a technology level Nandmakes a contract offer [{x(i, j)}i[0,1], {s(j),(j)}] for every input j [0, N]. If a supplieraccepts this contract for input j, she is obliged tosupply {x(i, j)}

i[0,1]as stipulated in the con-

tract in exchange for the payments {s(j), (j)}.A subgame perfect equilibrium of this game is astrategy combination for the firm and the sup-pliers such that suppliers maximize (5) and thefirm maximizes (6). An equilibrium can be al-ternatively represented as a solution to the fol-lowing maximization problem:

(8)

maxN,xi,ji,j ,sj,jj

R

0N

j sj dj CN

subject to (7) and the suppliers participationconstraint,

(9) sj j cx 0

1

xi, j di

w0 for all j 0, N.

Since the firm has no reason to provide rents to thesuppliers, it chooses payments s(j) and (j) thatsatisfy (9) with equality.9 Moreover, since thefirms objective function, (8), is (jointly) concavein the investment levels x(i, j), and these invest-ments are all equally costly, the firm chooses thesame investment level x for all activities in allintermediate inputs. Now, substituting for (9) in(8), we obtain the following simpler unconstrainedmaximization problem for the firm:

(10)

maxN,x

A1 N 1x cxNx CN w0N.

From the first-order conditions of this problem,we obtain

9 With complete contracts, (j) and s(j) are perfect sub-stitutes, so that only the sum s(j) (j) matters. This willnot be the case when contracts are incomplete.



6/28

(11) N* 1 1/1 A 1/1 cx /1

CN* w0 ,

(12) x* CN* w0cx

.

Equations (11) and (12) can be solved recur-sively. Given Assumption 1, equation (11)yields a unique solution for N*, which, togetherwith (12), yields a unique solution for x*.10

When all the investment levels are identicaland equal to x, output equals q N1x.Since NX Nx inputs are used in the pro-duction process, we can define productivity as

output divided by total input use, P N

. Inthe case of complete contracts, this produc-tivity level is

(13) P* N*,

which is increasing in the level of technology.In the next section we compare this to equilib-rium productivity under incomplete contracts.11

The next proposition describes the keyproperties of the equilibrium (proof in the

Appendix).

PROPOSITION 1: Suppose that Assumption 1holds. Then with complete contracts there existsa unique equilibrium with technology and in-vestment levels N* 0 and x* 0 given by(11) and (12). Furthermore, this equilibriumsatisfies

N*

A 0,

x*

A 0,

N*

x*

0.

In the case of complete contracts, the size ofthe market (as parameterized by the demandlevel A) has a positive effect on investments bysuppliers of intermediate inputs and productiv-ity. The other noteworthy implication of thisproposition is that, under complete contracts,the level of technology and thus productivitydoes not depend on the elasticity of substitutionbetween intermediate inputs, 1/(1 ).

III. Equilibrium under Incomplete Contracts

A. Incomplete Contracts

We now consider the same environment un-der incomplete contracts. We model the imper-

fection of the contracting institutions byassuming that there exists a [0, 1] suchthat, for every intermediate input j, investmentsin activities 0 i are observable andverifiable and therefore contractible, while in-vestments in activities i 1 are not con-tractible. Consequently, a contract stipulatesinvestment levels x(i, j) for the contractibleactivities, but does not specify the investmentlevels in the remaining 1 noncontractibleactivities. Instead, suppliers choose their invest-ments in noncontractible activities in anticipa-

tion of the ex post distribution of revenue, andmay decide to withhold their services in theseactivities from the firm. We follow the incom-plete contracts literature and assume that the expost distribution of revenue is governed by mul-tilateral bargaining, and, as in Hart and Moore(1990), we adopt the Shapley value as the so-lution concept for this multilateral bargaininggame (more on this below).12

The timing of events is as follows:

The firm adopts a technology N and offers a

contract [{xc(i, j)}i0 , (j)] for every inter-mediate input j [0, N], where xc(i, j) is aninvestment level in a contractible activity and

10 We show in the Appendix that the second-order con-ditions are satisfied under Assumption 1.

11 This measure of productivity implicitly assumes thatall the investments x(i, j) are measured accurately. Theremay be some tension between this assumption and theassumption that the cost of these investments is not pecu-niary. For this reason, in the previous version we consideredanother definition of productivity: output divided by thenumber of suppliers, P q/N (N)x. The ranking ofproductivity levels between complete and incomplete con-tracts is the same under both definitions, and the quantitativeeffects of changes in contractibility on productivity aresmaller with the definition used here.

12 The incomplete contracts literature also assumes thatrevenues are noncontractible. As is well known, with bilat-eral contracting or with multilateral contracting and a bud-get breaker (e.g., Bengt R. Holmstrom 1982), contracting onrevenues would improve incentives. In our setting, we donot need this assumption, since each firm has a continuumof suppliers, and contracts on total revenues would notprovide additional investment incentives to suppliers.



7/28

(j) is an upfront payment to supplier j. Thepayment (j) can be positive or negative.

Potential suppliers decide whether to applyfor the contracts. Then the firm chooses Nsuppliers, one for each intermediate input j.

All suppliers j [0, N] simultaneouslychoose investment levels x(i, j) for all i [0,1]. In the contractible activities i [0, ]they invest x(i, j) xc(i, j).

The suppliers and the firm bargain over thedivision of revenue, and at this stage, suppli-ers can withhold their services in noncon-tractible activities.

Output is produced and sold, and the revenueR is distributed according to the bargainingagreement.

We will characterize a symmetric subgameperfect equilibrium (SSPE for short) of thisgame, where bargaining outcomes in all sub-games are determined by Shapley values.

B. Definition of Equilibrium andPreliminaries

Behavior along the SSPE can be described bya tuple {N, xc, xn, } in which N represents the

level of technology, xc the investment in con-tractible activities, xn the investment in noncon-tractible activities, and the upfront payment toevery supplier. That is, for every j [0, N], theupfront payment is (j) , and the investmentlevels are x(i, j) xc for i [0, ] and x(i, j)

xn for i (, 1]. With a slight abuse of termi-nology, we will denote the SSPE by {N, xc, xn}.

The SSPE can be characterized by backwardinduction. First, consider the penultimate stageof the game, with N as the level of technologyand xc as the level of investment in contractible

activities. Suppose also that each supplier otherthan j has chosen a level of investment in non-contractible activities equal to xn(j) (these areall the same, because we are constructing asymmetric equilibrium), while the investmentlevel in every noncontractible activity by sup-plier j is xn(j).

13 Given these investments, the

suppliers and the firm will engage in multilat-eral Shapley bargaining. Denote the Shapleyvalue of supplier j under these circumstances bysx[N, xc, xn(j), xn(j)]. We derive an explicitformula for this value in the next subsection.For now, note that optimal investment by sup-plier j implies that xn(j) is chosen to maximizesx[N, xc, xn(j), xn(j)] minus the cost of invest-ment in noncontractible activities, (1 )cxxn(j).In a symmetric equilibrium, we need xn(j)

xn(j) or, in other words, xn needs to be afixed-point given by14

(14) xn arg maxxn j

sx N, xc , xn , xn j

1 cxxn j.

Equation (14) can be thought of as an incentivecompatibility constraint, with the additionalsymmetry requirement.

In a symmetric equilibrium with technologyN, with investment in contractible activitiesgiven by xc and with investment in noncontract-ible activities equal to xn, the revenue of thefirm is given by R A1(N1xc

xn1).

Moreover, let sx(N, xc, xn) sx(N, xc, xn, xn);then the Shapley value of the firm is obtained as

a residual:

sq N, xc , xn A1 N 1xc

xn1

Nsx N, xc , xn .

Now consider the stage in which the firm choosesN suppliers from a pool of applicants. If sup-pliers expect to receive less than their outsideoption, w0, this pool is empty. Therefore, forproduction to take place, the final-good pro-

ducer has to offer a contract that satisfies theparticipation constraint of suppliers under in-complete contracts, i.e.,

13 More generally, we would need to consider a distri-bution of investment levels, {xn(i, j)}i(,1] for supplier j,where some of the activities may receive more investment

than others. It is straightforward to show, however, that thebest deviation for a supplier is to choose the same level ofinvestment in all noncontractible activities. For this reasonwe save on notation and restrict attention to only suchdeviations.

14 This equation should be written with instead of. However, we show below that the fixed point xn in(14) is unique, justifying our use of .



8/28

(15) sx N, xc , xn , xn cxxc

1 cxxn w0

for xn that satisfies 14.

In other words, given N and (xc, ), each sup-plier j [0, N] should expect her Shapley valueplus the upfront payment to cover the cost ofinvestment in contractible and noncontractibleactivities and the value of her outside option.

The maximization problem of the firm canthen be written as

maxN,xc ,xn ,

sq N, xc , xn N CN

subject to 14 and 15.

With no restrictions on , the participationconstraint (15) will be satisfied with equality;otherwise the firm could reduce without vio-lating (15) and increase its profits. We cantherefore solve from this constraint, substitutethe solution into the firms objective function,and obtain the simpler maximization problem:15

(16) maxN,xc ,xn

sq N, xc , xn Nsx N, xc , xn , xn

cxxc 1 cxxn CN w0N

subject to 14.

The SSPE {N, xc, xn} solves this problem, andthe corresponding upfront payment satisfies

(17) cxxc 1 cxxn w0

sx N, xc , xn , xn .

C. Bargaining

We now derive the Shapley values in thisgame (see Lloyd S. Shapley 1953, or Martin J.

Osborne and Ariel Rubinstein 1994). In a bar-gaining game with a finite number of players,each players Shapley value is the average ofher contributions to all coalitions that consist ofplayers ordered below her in all feasible permu-tations. More explicitly, in a game with M 1players, let g { g(0), g(1), ... , g(M)} be apermutation of 0, 1, 2, ... , M, where player 0 isthe firm and players 1, 2, ... , M are the suppli-ers, and let zg

j {jg(j) g(j)} be the set ofplayers ordered below j in the permutation g.We denote by G the set of feasible permutationsand by v:G 3 the value of the coalitionconsisting of any subset of the M 1 players.16

Then the Shapley value of player j is

sj 1M 1! gG

vzgj j vzg

j .

In the Appendix, we derive the asymptoticShapley value of Robert J. Aumann and Shap-ley (1974) by considering the limit of this ex-pression as the number of players goes toinfinity.17 Leaving the formal derivation to theAppendix, here we provide a heuristic deriva-tion of this Shapley value.

Suppose the firm has adopted technology N,

all suppliers provide an amount xc of everycontractible activity, and all suppliers other than

j invest xn(j) in every noncontractible activity,while supplier j invests xn(j). To compute theShapley value for supplier j, first note that thefirm is an essential player in this bargaininggame. (If a coalition does not include the firm,then its output equals zero regardless of itssize.) Consequently, the supplier js marginalcontribution is equal to zero when a coalitiondoes not include the firm. When it does includethe firm and a measure n of suppliers, the mar-

ginal contribution of supplier j is m(j, n) R/n, where

15 Note that, as in the case with complete contracts, the firmchooses its technology and investment levels to maximize salerevenues net of total costs. The key difference is that withincomplete contracts, this maximization problem is con-strained by the incentive compatibility condition (14).

16 In our game, the value of a coalition equals the amountof revenue this coalition can generate.

17 More formally, we divide the interval [0, N] into Mequally spaced subintervals with all the intermediate inputsin each subinterval of length N/M performed by a singlesupplier. We then solve for the Shapley value and take thelimit of this solution as M3 . See Aumann and Shapley(1974) or Stole and Zwiebel (1996b).



9/28

R A1 N 11/

0

n

exp

0

1

ln x(i, k) di

dk

/

is the revenue derived from the employment ofn inputs with technology N, and the last inputk n is provided by supplier j. Since x(i, k)

xc for all 0 i and all 0 k n, x(i, k) xn(j) for all i 1 and all 0 k n, andx(i, k) xn(j) for all i 1 and k n,evaluating the previous expression enables us towrite the marginal contribution of supplier j as

(18) mj, n

A1 N 11/

xn jxn j1

xcxn j

1 n/.

The Shapley value of supplier j is the average ofher marginal contributions to coalitions that con-sist of players ordered below her in all feasibleorderings. A supplier who has a measure n ofplayers ordered below her has a marginal contri-bution of m(j, n) if the firm is ordered below her

(probability n/N), and 0 otherwise (probability1 n/N). Averaging over all possible orderings ofthe players and using (18), we obtain

sx N, xc , xn j, xn j1

N0

NnNmj, n dn

1 A1 xn jxn j1

xc

xn j

1

N

1 1

,

where

(19)

.

We therefore obtain the following lemma (seethe Appendix for the formal proof):

LEMMA 1: Suppose that supplier j investsxn(j) in her noncontractible activities, all the

other suppliers invest xn(j) in their noncon-tractible activities, every supplier invests xc inher contractible activities, and the level of tech-nology is N. Then the Shapley value of supplier

j is

(20) sx N, xc , xn j, xn j

1 A1 xn jxn j1

xcxn j

1 N 1 1,

where is defined in (19).

A number of features of (20) are worth not-ing. First, in equilibrium, all suppliers investequally in all the noncontractible activities, i.e.,

xn(j) xn(j) xn, and so

(21) sx N, xc , xn sx N, xc , xn , xn

1 A1 xcxn

1 N 1 1

1 R

N,

where R A1xcxn

(1)N(1) is the totalrevenue of the firm. Thus, the joint Shapleyvalue of the suppliers, Nsx(N, xc, xn), equals thefraction 1 of the revenue, and the firmreceives the remaining fraction , i.e.,

(22) sq N, xc , xn A1 xc

xn1 N 1

R.

This is a relatively simple rule for the divisionof revenue between the firm and its suppliers.

Second, the derived parameter /( )represents the bargaining power of the firm; it isincreasing in and decreasing in . A higherelasticity of substitution between intermediateinputs, i.e., a higher , raises the firms bargain-ing power, because it makes every supplier lessessential in production and therefore raises theshare of revenue appropriated by the firm. Incontrast, a higher elasticity of demand for thefinal good, i.e., higher , reduces the firms



10/28

bargaining power, because, for any coalition, itreduces the marginal contribution of the firm tothe coalitions payoff as a fraction of revenue.18

Furthermore, when is smaller, sx[N, xc,xn

(j), xn

(j)] is more concave with respect toxn(j), because greater complementary betweenthe intermediate inputs implies that a givenchange in the relative employment of two inputshas a larger impact on their relative marginalproducts. The impact of on the concavity ofsx will play an important role in the followingresults. The parameter , on the other hand,affects the concavity of revenue in output (see(2)) but has no effect on the concavity ofsx withrespect to xn(j), because with a continuum ofsuppliers, a single supplier has an infinitesimal

effect on output.

D. Equilibrium

To characterize an SSPE, we first derive theincentive compatibility constraint using (14)and (20):

xn arg maxxn j

1 A1 xn jxn 1

xcxn

1 N 1 1 cx 1 xn j.

Relative to the producers first-best choice char-acterized above, we see two differences. First,the term (1 ) implies that the supplier is notthe full residual claimant of the return from her

investment in noncontractible activities andthus underinvests in these activities. Second, asdiscussed above, multilateral bargaining distortsthe perceived concavity of the private returnrelative to the social return. Using the first-ordercondition of this problem and solving for thefixed point by substituting xn(j) xn yields aunique xn:

(23) xn xn N, xc 1

cx 1xc

A1 N 1 11/1 1 .

Note that xn(N, xc) is increasing in xc; since themarginal productivity of an activity rises withinvestment in other activities, investments in

contractible and noncontractible activities arecomplements.19 Another implication of (23) isthat investment in noncontractible activities isincreasing in . Mathematically, this followsfrom the fact that (1 ) /( ) isincreasing in . The economics of this relation-ship is the outcome of two opposing forces. Theshare of the suppliers in revenue, (1 ), isdecreasing in , because greater substitutionbetween the intermediate inputs reduces thesuppliers ex post bargaining power. But agreater level of also reduces the concavity of

sx in xn, increasing the marginal reward frominvesting further in noncontractible activities.Because the latter effect dominates, xn is in-creasing in .

Now, using (21), (22), and (23), the firmsoptimization problem (16) can be expressed as

(24) maxN,xc

A1 xcxn N, xc

1 N 1

cx Nxccx N(1)xn (N, xc )

CN

w0N,

18 To clarify the effects of and on , let us return tothe derivation in the text and note that the marginal contri-bution of the firm to a coalition of n suppliers, each oneinvesting xc in contractible activities and xn in noncontract-ible activities, can be expressed as

mq n A1 N 1xc

xn1 n

N

/

nN

/

R,

where R A1N(1)xcxn

(1) is the equilibrium levelof the revenue (when n N). The expression (n/N)/ isdecreasing in and increasing in for all n N. Theparameter is given by the average of the (n/N)/ terms,

1

N0

N

nN

/

dn

,

and is also decreasing in and increasing in .

19 The effect of N on xn is ambiguous, since investmentin noncontractible activities declines with the level of tech-nology when ( 1) 1 and increases with N when( 1) 1. This is because an increase in N has twoopposite effects on a suppliers incentives to invest: agreater number of inputs increases the marginal product ofinvestment due to the love for variety embodied in thetechnology, but at the same time, the bargaining share of asupplier, (1 )/N, declines with N. For large values of,the former effect dominates, while for small values of, thelatter dominates.



11/28

where xn(N, xc) is defined in (23). Substituting(23) into (24) and differentiating with respect to

N and xc results in two first-order conditions,which yield a unique solution (N, xc) to (24):

20

(25) N 1 1/1 A 1/1 cx /1

1 1 1 1 1 1 1 /1

11 1 /1

CN w0 ,

(26) xc CN w0

cx.

As in the complete contracts case, these twoconditions determine the equilibrium recur-sively. First, (25) gives N and then, given N,(26) yields xc. Moreover, using (23), (25), and(26) gives the level of investment in noncon-tractible activities as

(27) xn 1 1 1

1 1 1

CN w0

cx .Comparing (12) to (26), we see that, for a given

N, the implied level of investment in contract-ible activities under incomplete contracts, xc, isidentical to the investment level in contractibleactivities under complete contracts, x*. Thishighlights the fact that differences in the invest-ment in contractible activities between theseeconomic environments result only from differ-

ences in technology adoption. In fact, compar-ing (11) with (25), we see that N and N* differonly because of the two bracketed terms on theleft-hand side of (25). These represent the dis-tortions created by bargaining between the firmand its suppliers. Intuitively, technology adop-tion is distorted because incomplete contractsreduce investment in noncontractible activities

below the level of investment in contractibleactivities, and this underinvestment reducesthe profitability of technologies with high N. As 3 1 (and contractual imperfections disap-pear), both these bracketed terms on the left-handside of (25) go to 1 and (N, xc) 3 (N*, x*).

21

The impact of incomplete contracts on produc-tivity follows directly from their effect on thechoice of technology. In particular, productivityunder incomplete contracts, P N, is alwayslower than P* as given in (13), since N N*.

E. Implications of Incomplete Contracts

We now provide a number of comparative

static results on the SSPE under incompletecontracts, and compare the incomplete-contracts equilibrium technology and invest-ment levels to the equilibrium under completecontracts. The comparative static results are fa-cilitated by the block-recursive structure of theequilibrium; any change in A, , or thatincreases the left-hand side of (25) also in-creases N, and the effect on xc and xn can thenbe obtained from (26) and (27). The main re-sults are provided in the next proposition (proofin the Appendix).

PROPOSITION 2: Suppose that Assumption 1holds. Then there exists a unique SSPE underincomplete contracts, {N, xc, xn}, characterizedby (25), (26), and(27). Furthermore, {N, xc, xn}satisfies N, xc, xn 0,

xn xc ,

N

A 0,

xc

A 0,

xn

A 0,

N

0,

xc

0,

xn/xc

0,

N

0,

xc

0,

xn/xc

0.

20 See the Appendix for a more detailed derivation andfor the second-order conditions.

21 Note that as 3 1 the investment level xn does notconverge to x*, because the effect of distortions on thenoncontractible activities does not go to zero. What goes tozero, however, is the importance of noncontractible activi-ties in the production of final goods.



12/28

The main results in this proposition are intu-itive. Suppliers invest less in noncontractibleactivities than in contractible activities, in par-ticular,

(28)xn

xc

1 1 1

1 1 1 1,

which follows from equations (26) and (27) andfrom the fact that (1 ) /( ) (recall (19)). Intuitively, the firm is the fullresidual claimant of the return to investments incontractible activities, and it dictates these invest-ments in the contract. In contrast, investments innoncontractible activities are decided by the sup-

pliers, who are not the full residual claimants ofthe returns generated by these investments (recall(21)) and thus underinvest in these activities.

In addition, the level of technology and in-vestments in both contractible and noncontract-ible activities is increasing in the size of themarket, in the fraction of contractible activities(quality of contracting institutions), and in theelasticity of substitution between intermediateinputs.22 The impact of the size of the market isintuitive; a greater A makes production moreprofitable and thus increases investments and

equilibrium technology. Better contracting in-stitutions, on the other hand, imply that a greaterfraction of activities receives the higher invest-ment level xc rather than xn xc. This makesthe choice of a more advanced technology moreprofitable. A higher N, in turn, increases theprofitability of further investments in xc and xn.Better contracting institutions also close the(proportional) gap between xc and xn becausewith a higher fraction of contractible activities,the marginal return to investment in noncon-tractible activities is also higher.

A higher , i.e., lower complementarity be-tween intermediate inputs, also increases technol-

ogy choices and investments. The reason is relatedto the discussion in the previous subsection whereit was shown that a higher reduces the share ofeach supplier but also makes sx less concave.Because the latter effect dominates, a lower de-gree of complementarity increases supplier invest-ments and makes the adoption of more advancedtechnologies more profitable.

The reduced-form profit function depictedin equation (1) can also be derived at thispoint. Combining (23), (24), and the condi-tion for the firms optimal choice of xc, thefirms payoff can be expressed as (see theAppendix):

(29) AZ, N1 1 1/1

CN w0N,

where

(30) Z, 1 /1

1 1 /1

1 1 1 1 1 1 1 /1

cx /1

represents a measure of derived efficiencyand captures the distortions arising from incom-plete contracts. The extent of these distortionsdepends on the models parameters as shown inProposition 2. In addition, we have the follow-ing lemma (proof in the Appendix):

LEMMA 2: Suppose that Assumption 1 holds.Let (, ) [ Z(, )/]/Z(, ) bethe elasticity of Z(, ) with respect to and let(, ) [ Z(, )/]/Z(, ) be theelasticity of Z(, ) with respect to . Then, wehave that

(a) (, ) 0 and(, ) 0; and(b) (, )/ 0 and(, )/ 0.

Part (a) of this lemma implies that bettercontracting institutions and greater substitutability

22 Equation (13) above implies that the measure of pro-ductivity, P q/(NX), has the same comparative statics astechnology. The same results also apply if we were to defineproductivity as P q/N, that is, as output divided by thenumber of suppliers. Yet, in this case firms operating underdifferent contracting institutions would have different pro-ductivity levels, not only because they choose differentlevels of technology, but also because they have differentinvestment levels. See Acemoglu, Antras, and Helpman(2005) for more details on this point.



13/28

between intermediate inputs lead to higher levelsofZ(, ). Part (b), on the other hand, implies thatthe proportional increase in Z(, ) in response toan improvement in contracting institutions isgreater when there is more technological comple-mentarity between intermediate inputs. The intu-ition is that contract incompleteness is moredamaging to technologies with greater comple-mentarities, because there are more significantinvestment distortions in this case. This lastresult implies that sectors with greater comple-mentarities are more contract dependent, a fea-ture that will play an important role in thegeneral equilibrium analysis in Section IVC.

Finally, to compare the complete and incom-plete contracts equilibria, recall that they both

lead to the same allocation as 3

1. Togetherwith Proposition 2, this implies (proof in theAppendix):

PROPOSITION 3: Suppose that Assumption 1holds. Let {N, xc, xn} be the unique SSPE withincomplete contracts, and let {N*, x*} be theunique equilibrium with complete contracts.Then

N N* and xnxcx*.

This proposition implies that since incompletecontracts lead to the choice of less advanced(lower N) technologies,23 they also reduce pro-ductivity and investments in contractible andnoncontractible activities.

IV. Applications

We next discuss a number of applications ofthe basic framework developed so far. These

applications emphasize both the potential quan-titative effects of our main mechanism and itsimplications for equilibrium organizationalchoices and endogenous patterns of compara-tive advantage.

A. Quantitative Implications

We first investigate whether our baselinemodel can generate significant productivity dif-ferences from cross-country variation in con-tracting institutions. Our purpose here is not toundertake a full-fledged calibration, but to givea sense of the empirical implications of ourbaseline model for plausible parameter values.To obtain closed-form solutions for the equilib-

rium value of N and for the productivity mea-sure P N, we assume that the function C(N)is linear in N, C(N) N. Equation (25) thenimplies24

P A w0

1 /1 1

/1 1cx/1 1

1 1 1

1 1 1 1 /1 1

1 1 /1 1

.

Our focus is on the impact of the quality ofcontracting institutions, , on firm-level pro-ductivity. For this purpose, consider the ratioof productivity in two economies with thefraction of contractible tasks given by 1 and

0 1:23 It is useful to contrast this result with the overemploy-ment result in Stole and Zwiebel (1996a, b). There are twoimportant differences between our model and theirs. First,our model features investments in noncontractible activities,which are absent in Stole and Zwiebel. Second, Stole andZwiebel assume that if a worker is not in the coalition of thebargaining game, she receives no payment. Since in ourmodel investment in noncontractible activities is not verifi-able, a supplier receives the upfront payment indepen-dently of whether she is in the coalition, and she receivesonly the rest of the payment, i.e., sx, through bargaining.The treatment of the outside option in Stole and Zwiebel isessential for their overemployment result (see Catherine C.de Fontenay and Joshua S. Gans 2003).

24 Recall from the discussion in footnote 11 that thereare alternative ways of measuring productivity in themodel, and which one of these is more appropriate willdepend on how productivity is measured in practice. If,for example, we were to measure productivity as outputdivided by the number of suppliers, the impact of im-provements in contracting institutions on productivitywould be even larger, because such improvements raiseinvestment in contractible and noncontractible activitiesas well.



14/28

(31)P1

P0

1 1 1 1 1 11 1 1 1

11

1 1 1 0

1 10 1 1 0

11

11 01

11 ,

where, recall, /( ). Equation (31)shows that the proportional impact of onaggregate productivity depends only on the val-ues of the parameters , , and .25

The parameter governs the elasticity of

substitution between final-good varieties, andalso determines the markup charged by final-good producers. In our benchmark simulation,we set this parameter equal to 0.75. This impliesan elasticity of substitution between final-goodvarieties equal to four, which corresponds to themean elasticity estimated by Christian Brodaand David E. Weinstein (2006) using US importdata. This value of also implies a markup overmarginal cost equal to 1 1 0.33, whichis comfortably within the range of availableestimates.26

We choose the parameter to match MarkBils and Klenows (2001) estimates of varietygrowth in the US economy from Bureau ofLabor Statistics (BLS) data on expenditureson different types of goods. Their findingsindicate that the growth of the number ofvarieties in the US economy was around 2percent per annum between 1959 and 1999. Ifwe assume that the variety of intermediategoods grew at the same rate over this period,

and also that this variety growth was the mainsource of productivity (TFP) growth in the USeconomy, we can use our model to back out avalue of . We take the rate of TFP growthper annum to be 0.5 percent and use the factthat productivity is given by P N. Thisimplies that (Nt1

Nt)/Nt

0.005 andgenerates a value of ln(1.005)/ln(1.02) 0.25.27

The value of is harder to pin down. Sincethe main theoretical ideas in the paper relate tocomplementarity between different inputs, alow value of may be natural. Nevertheless,since we have no direct way of relating toaggregate data, we show the implications ofchanges in the quality of contracting institutionsfor different values of.

Figure 1 depicts the value of the ratio in (31)for different values of when 0.75 and 0.25. We consider three experiments. Themiddle curve depicts a shift from 0 0.25 to25 In deriving equation (31), we assume that the remain-

ing parameters are held fixed when changes. We shouldthus interpret our results as reflecting the partial-equilibriumresponse of productivity to changes in contractibility.

26 On the lower side, Catherine J. Morrison (1992) findsmarkups ranging from 0.1 in 1961 to 0.29 in 1970, while onthe higher side Julio J. Rotemberg and Michael Woodford(1991) use an estimate of 0.6. See Susanto Basu (1995) fora discussion of the sensitivity of the estimated markups tomodel specification and calibration. It is interesting to notethat Morrison (1995) finds markups between 0.1 and 0.39 inJapan and 0.1 and 0.2 in Canada, Mark J. Roberts (1996)finds average markups between 0.22 and 0.3 in Colombia,and Jean-Marie Grether (1996) finds average markups be-tween 0.34 and 0.37 in Mexico.

27 A somewhat different number is obtained fromBroda and Weinsteins (2006) estimate that the numberof varieties in US imports between 1972 and 2001 grewat an annual rate of 3.7 percent. Using the same formula,this would imply a value of 0.135 for . Note, however,that this is likely to be an overestimate of variety growthand thus an underestimate of because Broda and Wein-stein consider the same good imported from differentcountries as different varieties, and the number of ex-porting countries to the United States increased overtime. For this reason, we choose 0.25 as our bench-mark parameter value.

FIGURE 1. RELATIVE PRODUCTIVITY FOR 0.75 AND 0.25

http://www.atypon-link.com/action/showImage?doi=10.1257/aer.97.3.916&iName=master.img-000.png&w=191&h=143


15/28

1 0.75, the lowest curve depicts a moremodest increase from 0 13 to 1

23 ,while the highest curve illustrates the most ex-treme shift from 0 0 to 1 1.

The patterns in Figure 1 confirm the results inLemma 2 and show that the impact of an in-crease in on productivity is larger for lowervalues of. More importantly, the figure showsthat the quantitative effects can be sizable for alarge range of values of. For example, when 0.25, productivity increases by a factor of2.5, 4.1, and 19.7 in the three experiments. As

increases and intermediate inputs become moresubstitutable, however, the magnitude of thequantitative effects diminishes. For example,when 0.75, productivity increases by afactor of 1.4, 1.7, and 3.2 in the three cases.Though smaller, these are still sizable effects.

We next discuss the sensitivity of these quan-titative results to alternative parameter values.While markup estimates are consistent with avalue of 0.75, other evidence suggestssomewhat higher or lower values of . Figure2 shows the results of an increase in from

0.25 to 0.75 when 0.78 and when 0.7,while leaving at 0.25.28 The figure shows thatthe quantitative effects are considerably largerin the case in which 0.78, even for largevalues of . For example, with 0.75, theimprovement in contracting institutions leads toa proportional increase in productivity of 4.4,which is a very sizable effect. Setting equal to

0.7 reduces the effect of improved contractibil-ity on productivity, but for 0.25 we stillfind that the experiment increases productivityby a factor of 1.8.

Finally, Figure 3 shows the results for alter-native values of ( 0.2 and 0.3) whileholding at 0.75. When 0.2, the estimatedeffects are somewhat smaller than in our bench-mark simulation, but still sizable; with low val-ues of, the impact of an improvement in thequality of contracting institutions on productiv-

ity is still large. For example, with 0.25,productivity increases by a factor of 2.0. On theother hand, greater values of lead to moresignificant responses of productivity to in-creases in . Even for a very high value of such as 0.75, productivity increases by a factorof 4.8 as increases from 0.25 to 0.75.

Overall, this simple quantitative evaluationsuggests that the mechanism highlighted in ourmodel is capable of generating quantitativelysizable effects from relatively modest variationsin contracting institutions.

B. Choice of Organizational Forms

An important insight of the incomplete con-tracts literature, and especially of Grossman andHart (1986) and Hart and Moore (1990), is toconsider organizational forms (and the owner-ship of assets) as a choice variable affecting exante investments. Our framework is tractableenough to allow these considerations and can beused to discuss issues of vertical integrationversus outsourcing, and how the employment

28 Only values of less than 0.8 are consistent withAssumption 1 combined with 0.25.

FIGURE 2. RELATIVE PRODUCTIVITY P(0.75)/ P(0.25) FOR 0.25 AND ALTERNATIVE S

FIGURE 3. RELATIVE PRODUCTIVITY P(0.75)/ P(0.25) FOR 0.75 AND ALTERNATIVE S

http://www.atypon-link.com/action/showImage?doi=10.1257/aer.97.3.916&iName=master.img-002.png&w=191&h=130http://www.atypon-link.com/action/showImage?doi=10.1257/aer.97.3.916&iName=master.img-001.png&w=191&h=128


16/28

relationship between the firm and its suppliersshould be organized. We now briefly discusshow this can be done. A more detailed treatmentis presented in our working paper version, Ace-moglu, Antras, and Helpman (2005).

We have thus far assumed that the threatpoint of a supplier is not to deliver the noncon-tractible activities, whichin view of the Cobb-Douglas structure of the production function ofthe intermediate input in (4)is equivalent toassuming that the supplier does not deliver X(j)at all. Assume, instead, that the threat point of asupplier is not to deliver a fraction 1 of her

X(j), where 0 1. The magnitude of depends, among other things, on whether thesupplier is an employee of the firm or an outside

contractor. Our analysis above corresponds tothe case 0. The following lemma general-izes the Shapley value to the case where 0(proof in the Appendix):

LEMMA 3: Suppose that supplier j investsxn(j) in her noncontractible activities, all theother suppliers invest xn(j) in their noncon-tractible activities, every supplier invests xc inher contractible activities, and the level of tech-nology is N. Then, the Shapley value of supplier

j is given by (20), where

(32) 1

1 .

Lemma 3 implies that the formula for theShapley value is the same as before, except thatnow the firms share in the bargaining game, ,depends on . Clearly, in (32) equals in (19)when 0, but is greater when 0. This isnatural, since, with 0, the bargaining gameis more advantageous to the firm. In the limit, as

goes to 1, the firms share also goes to 1.Using this lemma, the working paper version

demonstrated that all the results in Propositions2 and 3 hold for any [0, 1) and employedthis generalization to analyze the choice be-tween integration and outsourcing. Briefly, sup-pose that for an integrated firm, we have 0,while with outsourcing 0.29 Then, when the

choice of the upfront payment is not restricted,it can be shown that the firm always prefersoutsourcing to vertical integration.30 In contrast,when suppliers face credit constraints (in thesense that the upfront payment is restricted tobe nonnegative), integration may be preferableto outsourcing. In particular, when credit con-straints are present and 1, there exists (0, 1) such that vertical integration is pre-ferred by the firm for and outsourcing ispreferred for . Furthermore, is decreas-ing in , so that integration is more likely whenthere is greater complementarity between inter-mediate inputs. This result implies that verticalintegration is more likely when both contractualfrictions and credit market imperfections are

present.

31

C. General Equilibrium

We now discuss how the technology choicecan be embedded in a general equilibrium ver-sion of our model. Besides verifying that gen-eral equilibrium interactions do not reverse ourpartial equilibrium results, this analysis is usefulas a preparation for the results in the nextsubsection, where we investigate endogenouspatterns of comparative advantage across coun-

tries. We start with an equilibrium with a givennumber of producers (final goods) and thenendogenize this with free entry. In this and thenext subsection, we focus on the case where0 1, and 0, as in the baseline model.

Assume that there exists a continuum of finalgoods q(z), with z [0, Q], where Q representsthe number (measure) of final goods. All con-sumers have identical preferences,

29 An integrated firm has 0 because in this case thefirm owns all the intermediate inputs. In this event, the most

a supplier can do is not to cooperate in the use of herintermediate input, which will reduce the efficiency withwhich the firm can employ this input, but may not reducethis efficiency to zero. See Grossman and Hart (1986).

30 This is because the firm does not undertake any rela-tionship-specific investments. Since suppliers are the onlyagents undertaking noncontractible relationship-specific in-vestments, it is efficient to give them as much bargainingpower as possible. This will not necessarily be the case ifthe firm were to also make relationship-specific invest-ments. See, for example, Hart and Moore (1990), Antras(2003, 2005), and Antras and Helpman (2004).

31 The positive effect of complementarity on the integra-tion decision has been derived before in the property-rightsliterature (see Hart and Moore 1990).



17/28

(33) u 0

Q

q(z) dz1/

cx e, 0 1,

where e is the total effort exerted by this indi-vidual, the elasticity of substitution betweenfinal goods, 1/(1 ), is greater than 1, and cxrepresents the cost of effort in terms of realconsumption. These preferences imply the de-mand function

qz pzpI 1/1 S

pI,

where p(z) is the price of good z, S is the

aggregate spending level, and

pI 0

Q

p(z)/(1 ) dz 1 /

is the ideal price index, which we take to be thenumeraire, i.e., pI 1. The implied demandfunction A[p(z)]1/(1) for each firm is there-fore identical to the demand function used in theprevious sections, with A S.

Recall that each firm in this economy solves the

maximization problem (24) and its reduced-formprofit function is given by (29). We assume thatthe degree of technological complementarity, ,varies across firms (or sectors) with its supportgiven by a subset of (0, 1). We denote its cumu-lative distribution function by H(). This formu-lation implies that if Q products are available forconsumption, a fraction H() of them are pro-duced with elasticities of substitution smaller than1/(1 ).

The key general equilibrium interaction re-sults from competition of producers for a scarce

resource, labor. Assume that labor is in fixedsupply L. A firm that adopts technology N em-ploys N individuals as suppliers and CL(N)workers in the process of technology adoption(implementation, use, or creation of the tech-nology). We assume that these are the only usesof labor (Assumption 1 now applies to CL(N)).Denoting the wage rate in terms of the nu-meraire by w, this implies that the total cost ofadopting technology N is C(N) wCL(N). Thewage rate w is taken as given by each firm, butis endogenously determined in equilibrium.

The first-order condition of the maximizationof (29) yields

(34)

1

AZ, N 1 1/1

wCL N w0 ,

for all 0, 1.

Since each individual can be employed at thewage w in the process of technology adoption,their outside option as a supplier is w0 w.Equation (34) then implies that the technologychoice depends on a A/w, which is the in-verse real cost of technology adoption (or the

equilibrium market size). Let the equilibriumtechnology choice of N as a function of a, ,and implied by (34) be N(a, , ). Definingtotal labor demand by a firm with technology Nas CT(N) CL(N) N, condition (34) impliesthat N(a, , ) is implicitly defined by

(35)

1 aZ, Na, , 1 1/1

CTNa, , ,

for all 0, 1.

Since firms with higher elasticities of substitu-tion choose higher N(Proposition 2), this equa-tion implies that N(a, , ) is increasing in .The demand for labor by a firm choosing tech-nology N(a, , ) is CT[N(a, , )], thus labormarket clearing can be expressed as32

(36) Q 0

1

CTNa, , dH L,

where the left-hand side is total demand for laborand L is labor supply. Since N(a, , ) is increas-ing in a, this condition uniquely determines theequilibrium value ofa, i.e., our measure of the realdemand level. The relationship between a and themodels parameters, as embodied in this equation,

32 Since it is straightforward to verify that the wage isalways strictly positive, (36) is written as an equality ratherthan in complementary slackness form.



18/28

illustrates the key general equilibrium feedback inour model. Proposition 2 implies that N(a, , ) isincreasing in , so we may expect better contract-ing institutions to encourage all firms to adoptmore advanced technologies. The resource con-straint, (36), implies, however, that not all the

N(a, , )s can increase with Q constant. Con-sequently, the equilibrium value of a has to adjustto clear the labor market when rises.

To derive the implications of an increase in on the cross-sectional distribution of technologychoices, recall that the number of products, Q, isgiven, and differentiate the first-order condition(35) to obtain

(37)

a , Na, , Na, , ,

where y, defined as d y/y, represents the propor-tional rate of change of variable y, (, ) isthe elasticity of Z(, ) with respect to , and(N) is the elasticity of the marginal cost curveCT(N) minus [( 1) 1]/(1 ), i.e.,

N CTNN

CTN

1 1

1 .

Assumption 1 implies that (N) 0. Moreover,as proved in Lemma 2, (, ) 0 and (, )is decreasing in . Next, differentiating (36), weobtain a relationship between and a:

(38) 0

1

L Na, , dH 0,

where L() CT[N(a, , )]N(a, , ). Sub-stituting (37) into (38) then yields

a0

1

L, Na, , 1 dH

0

1

LNa, , 1 dH

.

Since the term in front of on the right-handside of this equation is negative, an improve-ment in contracting institutions increases wagesrelative to expenditure and reduces a.

Equation (38) shows that the proportionalchange in Nin response to , N(a, , ), can bepositive only for some s (again because of theresource constraint). Since (, )/ 0(from Lemma 2), the left-hand side of equation(37) is decreasing in . Consequently, thereexists a critical value such that N(a, , ) 0 for all and N(a, , ) 0 for all . This implies that low firms, with greatertechnological complementarity between inputs,are more contract dependent. This establishesthe following proposition (proof in the text):

PROPOSITION 4: Suppose Assumption 1 holds.Then there exists (0, 1) such that in the

general equilibrium economy with Q constant,an increase in raises the level of technologyN(a, , ) in all firms with and reducesit in all firms with .

We next discuss how the number of products,Q, can be endogenized with free entry. To dothis in the simplest possible way, suppose thatan entrant faces a fixed cost of entry wf, where

f is the amount of labor required for entry. Thiscost is borne in addition to the cost of tech-nology adoption. Moreover, in the spirit of

Hugo A. Hopenhayn (1992) and Marc J. Melitz(2003), suppose that an entrant does not know akey parameter of the technology prior to entry.While in their models the entrant does not knowits own productivity, we assume instead that itdoes not know , but knows that is drawnfrom the cumulative distribution function

H(). Since the relationship between andproductivity is determined in general equilib-rium, the distribution of productivity is en-dogenous.

After entry, each firm learns its and maxi-

mizes the profit function (29). This maximiza-tion leads to the choice of technology as afunction of a, , and , i.e., N(a, , ), in (35).Let us define

a, , aZ,

Na, , 1 1 1/1 CTNa, , ,

where Z(, ) is given by (30). This is anindirect profit function, with profits measured in



19/28

units of labor. It is straightforward to verify thatit is increasing in a, , and .

Free entry implies that expected profits mustequal the entry cost wf, or

(39) 0

1

a, , dH f.

This free-entry condition uniquely determinesthe equilibrium value of a, without reference tothe labor market clearing condition (36). Con-sequently, given the equilibrium value of a, thedistribution of induces a distribution of pro-ductivity and firm size in the economy (from thefirst-order condition (35)): firms with larger s

choose more advanced technologies and theyare more productive.33

Finally, the labor market clearing conditiondetermines the number of entrants. Since labordemand now includes individuals working inthe founding of the firms, the market clearingcondition (36) has to be replaced with

Q0

1

CT[N(a, , )] dH() f L.Together with N(a, , ) from (35) and a fromthe free entry condition (39), this modified labormarket clearing condition determines the num-ber of entrants Q.

An interesting implication of the generalequilibrium with free entry is that an increase inthe supply of labor L does not create a scaleeffect and is not a source of comparative advan-tage. If two countries that differ only in L freelytrade with each other, their wages are equalized,their technology level is the same in every in-dustry , and they have the same distribution of

productivity and firm size. The only differenceis that the larger country has proportionatelymore final good producers. This result contrastswith the case with an exogenous number ofproducts, where differences in L would generatecomparative advantage.34

D. Comparative Advantage

Perhaps the most interesting general equilib-rium application of our framework is to inter-national trade. Consider a world consisting oftwo countries, indexed by 1, 2. We nowderive an endogenous pattern of comparativeadvantage between these two countries fromdifferences in contracting institutions.35 Sup-

pose that there is a fixed number of products andthat every product is distinct not only from otherproducts produced in its own country, but alsofrom products produced in the foreign country.All products can be freely traded between thetwo countries.

Suppose also that the two countries are iden-tical, except for their contracting institutions. Inparticular, L1 L2, Q1 Q2, and H1()

H2() for all (0, 1), but the fraction ofactivities that are contractible differs acrosscountries. Without loss of generality, we as-

sume that 1

2

, so that country 1 has bettercontracting institutions.

The equilibrium condition for technologyadoption (34) holds in both countries, with dif-ferent wage rates, w, for the two countries (A isthe same for both countries and equal to worldexpenditure). Defining a A/w and N()

N(a, , ), we have

(40)

1 aZ, N 1 1/1

CTN ,

for all 0, 1, 1, 2.33 Note also that the degree of dispersion of productivity

depends on the degree of contract incompleteness; in coun-tries with better contracting institutions there is less produc-tivity dispersion and less size dispersion. This follows fromthe fact that small firms (i.e., low- firms) are larger incountries with better contracting institutions, while largefirms (high- firms) are smaller in those same countries.This prediction of the model is consistent with the empiricalevidence reported in James R. Tybout (2000), which indi-cates that there are significantly fewer medium-sized enter-prises in less-developed economies, which typically haveworse contracting institutions.

34 We have also worked out an endogenous growth modelwith expanding product variety, where the long-run rate ofgrowth depends on the degree of contract incompleteness. Wedo not discuss this model in order to save space.

35 The link between contracting institutions and endog-enous comparative advantage was previously discussed byLevchenko (2003), Costinot (2004), Nunn (2006), andAntras (2005).



20/28

In addition, the labor market clearing condition(36) holds in both countries. Consequently, thecountry with better contracting institutions, coun-try 1, will have higher wages and a lower a.36

The pattern of trade can now be determinedby comparing the revenues of firms with thesame value of in the two countries. We showin the Appendix (see the proof of Proposition 2)that the revenue of a producer with parameter in country is

(41) R AZ, N/1

1 1

1 1 1 ,

which is increasing in total world expenditure,A, in Z(, ), and in the level of technology. Butit is also directly affected by the parametersthrough the last term on the right-hand side.

If

R1

R2

01 R1 dH

01 R2 dH

,

then country 1 is a net exporter of goods with

substitution parameter . Consequently, we sim-ply need to determine the distribution of R1()/

R2() across different s. From (41) we have

(42)R1

R2

Z, 1

Z, 2 N1

N2/1

1 1 1

1 1 2

1 1 2

1 1 1

.

Both Z(, 1)/Z(, 2) and the last term in(42) are decreasing in .37 Second, Proposition

4 implies that there exists an such thatN1() N2() for all and N

1() N2()for all . As a result, country 1, which hasthe better contracting institutions, tends to exportlow- products and import high- products. If, inaddition, (N) is constant, the proportional changein N(a, , ) in response to an increase in isalways smaller when is greater (because thepartial (, )/ is negative; see Lemma 2). Inthis case, R1()/R2() is everywhere decreasingin and there exists (0, 1) such that

R1()/R2() [01R1() dH()]/[0

1R2() dH()]for all , and the opposite inequality holdsfor all . Consequently, country 1, withthe better contracting institutions, is a net ex-porter in low- sectors and a net importer in

high- sectors. This result is summarized in thefollowing (proof in the text):

PROPOSITION 5: Suppose Assumption 1 holdsand(N) is constant. Then there exists (0, 1)such that in the two-country world equilibrium,country 1 with 1 2 is a net exporter of

products with and a net importer of products with .

The most important implication of this resultis that differences in contracting institutions cre-

ate endogenous comparative advantage. A countrywith better contracting institutions gains a com-parative advantage in sectors that are more con-tract dependent, which, in our model, correspondto the sectors with greater technological comple-mentarities between inputs.

The main implications in Proposition 5 re-ceive support from a number of recent empir-ical papers. Nunn (2006) and Levchenko(2003) use international trade data and theclassification of industries into more and lesscontract-dependent groups. Consistent with

Proposition 5, they find that countries withbetter contracting institutions specialize in theexport of goods that are more contract depen-dent. Moreover, the impact of cross-countrydifferences in the contracting institutions onexports is quantitatively large. Nunn, for ex-ample, finds that the estimated impact of thequality of the legal system on the export ofcontract-dependent sectors is comparable tothe effect of human capital abundance on theexport of human capitalintensive products.In other words, according to these estimates,

36 To obtain a contradiction, suppose that this is not thecase; then (40) implies N1() N2() for all s (sinceN() is increasing in for given a), and so labor marketclearing cannot be satisfied in both countries.

37 Recall from Lemma 2 that Z(, ) is increasing in and , and (, )/ 0. Therefore, Z(,

1)/Z(, 2)is decreasing in whenever 1 2. The fact that the lastterm is decreasing in follows from 1 2.



21/28

differences in legal systems are as importantas differences in human capital abundance forexplaining trade flows.

V. Conclusion

In this paper, we developed a tractable frame-work for the analysis of the impact of contractualincompleteness and technological complemen-tarities on the equilibrium technology choice. Inour model, a firm chooses its technology corre-sponding to the range of intermediate inputsused in production, and offers contracts to sup-pliers, specifying the required investments incontractible activities. Investments in the re-maining, noncontractible activities are then cho-

sen by the suppliers in anticipation of the expost bargaining payoffs.We used the Shapley value to characterize

the division of surplus between the firm andits suppliers, and derived an explicit solutionto these payoffs. Using this setup, we estab-lished that greater contractual incompletenessreduces investments in noncontractible andcontractible activities and depresses technol-ogy choice. The impact of contractual incom-pleteness on technology adoption is greater insectors with more complementary intermedi-

ate inputs.The key mechanism developed in the paper

leads to potentially large differences in produc-tivity in response to differences in contractinginstitutions. We also derived implications ofthis mechanism for differences in the internalorganization of the firm across societies withdifferent contracting institutions and a new

mechanism for endogenous comparative ad-vantage. These implications are consistentwith a range of recent empirical results in theliterature.

It is also useful to note that while we haveused a love-for-variety model of technol-ogy, our results do not depend on this specificmodeling approach. In equation (29), N canbe interpreted as a general technological in-vestment, and the linkages between contractsand technology adoption (or investment)highlighted in our analysis would apply inde-pendent of the specific assumptions regardingthe nature of technology.

A number of areas are left for future re-search. These include, but are not limited to,

the following. The model assumes that allactivities are symmetric; an important exten-sion is to see whether similar results hold witha more general production function, where thefirm may wish to treat some suppliers of in-termediate inputs differently from others, de-pending, for example, on how essential theyare for production. Another area for futurestudy is an investigation of the simultaneousdetermination of the range of intermediateinputs used by the firm and the division oflabor among the suppliers. Finally, it is im-

portant to investigate whether the relationshipbetween contracting institutions, technologi-cal complementarities, and the choice of tech-nology is fundamentally different when weuse alternative approaches to the theory of thefirm, such as the managerial incentives ap-proach of Holmstrom and Paul R. Milgrom(1991).

APPENDIX

Second-Order Conditions in the Complete Contracting Case

Let A1N(1)x cxNx C(N) w0N. Using the first-order conditions (11) and (12), the matrix of the

second-order conditions can be expressed as

2

x2

2

xN

2

Nx

2

N2

1 Ncx2Cw0

1 cx 1 1

cx 1 1 N11 1 1C w0 C

.The second-order conditions are satisfied if this matrix is negative definite, which requires its diagonal elements to be

negative and its determinant to be positive. The first diagonal element is negative; the second diagonal element is

negative if and only if



22/28

CN

C w0

1 1 1

,

and the determinant is positive if and only if

CN

C w01 1

1 .

Note that when (1 ) 1, both of these conditions are satisfied, and when (1 ) 1, the second inequality implies

the first inequality. Therefore Assumption 1 is necessary and sufficient for the second-order conditions to be satisfied.

PROOF OF PROPOSITION 1:

The first part of the proposition is a direct implication of Assumption 1. The comparative statics of N* follow from the

implicit function theorem by noting that, except for , N* increases in response to an increase in a parameter if and only if

this parameter raises the left-hand side of (11). Using the results concerning the response of N* to changes in parameters

together with (12) then implies the responses of x* to parameter changes.

PROOF OF LEMMA 1 AND RELATED RESULTS:

We now provide a formal derivation of the Shapley value introduced in the text, building on the work of Aumann

and Shapley (1974). Let there be M suppliers, each one controlling a range N/M of the continuum of intermediate

inputs. Due to symmetry, all suppliers provide an amount xc of contractible activities. As for the noncontractible

activities, consider a situation in which a supplier j supplies an amount xn(j) per noncontractible activity, while the M

1 remaining suppliers supply the same amount xn(j) (note that we are again appealing to symmetry).

To compute the Shapley value for this particular supplier j, we need to determine the marginal contribution of this supplier

to a given coalition of agents. A coalition of n suppliers and the firm yields a sales revenue of

(A1) FINn, N; A1 N 11/xc

n 1xn j1

xn j1 /,

when the supplier j is in the coalition, and a revenue

(A2) FOUTn, N; A1 N 11/xc

nxn j1 /,

when supplier j is not in the coalition. Notice that even when n N, the term N(11/) remains in front,

because it represents a feature of the technology, though productivity suffers because the term in square brackets is

lower.

Following the notation in the main text, the Shapley value of player j is

(A3) sj 1

M 1! gG

vzgj j vzg

j .

The fraction of permutations in which g(j) i is 1/(M 1) for every i. If g(j) 0 then v(zgj

j) v(zgj

) 0, becausein this event the firm is necessarily ordered after j. Ifg(j) 1 then the firm is ordered before j with probability 1/Mand after

j with probability 1 1/M. In the former case v(zgj j) FIN(1, N; ), while in the latter case v(zg

j j) 0. Therefore the

conditional expected value ofv(zgj j), given g(j) 1, is (1/M) FIN(1, N; ). By similar reasoning, the conditional expected

value ofv(zgj ) is (1/M) FOUT(0, N; ). Repeating the same argument for g(j) i, i 1, the conditional expected value of v(zg

j

j), given g(j) i, is (1/M) FIN(i, N; ), and the conditional expected value of v(zgj ) is (i/M) FOUT(i 1, N; ). It follows

from (A3) that

sj 1

M 1M i 1

M

iFINi, N; FOUTi 1, N; 1

N N i 1

M

iFINi, N; FOUTi 1, N; .

Substituting for (A1) and (A2),



23/28

sj A1 N 11/xc

N N i 1

M

iixn j1

xn j1

xn j1 /

A1

N 11/

xc

N N i 1

M

iixn j1

xn j1 /.

The first-order Taylor expansion of the previous expression gives

sj A1 N 11/xc

/xn j1

N N i 1

M

i ixn j1 / o,

or

sj

A

1

N

11/

/xn j

xn j1

xc

xn j

1

N N i 1

M

i/ o

.

Now, taking the limit as M3 , and therefore N/M3 0, the sum on the right-hand side of this equation becomes

a Riemann integral:

limM3

sj

A1 N 11// xn jxn j

1

xcxn j

1

N2 0

N

z/ dz.

Solving the integral delivers

limM3

sj/ 1 A1 xn j

xn j

1

xcxn j

1 N 1 1,

with /( ). This corresponds to equation (20) in the main text, and completes the proof of the lemma.

In addition, imposing symmetry, i.e., xn(j) xn(j), the firms payoff is

s0 A1 N 1xc

xn1

Nsj A1 xc

xn1 N 1,

as stated in equation (22) in the text.

PROOF OF PROPOSITION 2:

First, we verify that the second-order conditions are again satisfied under Assumption 1. To see this, note that the

problem in (14) is strictly concave and delivers a unique xn xn(N, xc), as given in (23). Plugging this expression into

(24), we obtain

(A4) A1 /1 1 N1 1 1/1 1 xc/1 1

cxxcN CN w0N,

where

cx 11 1 /1 1 1 1 1 .

The second-order conditions can be checked, analogously to the case with complete contracts, by computing the Hessian

and checking that it is negative definite.



24/28

Here, we present an alternative proof, which also serves to illustrate how the reduced-form profit function (29) in the main

text is derived. In particular, notice that for a given level of N, the problem of choosing xc is convex (since 1 )

and delivers a unique solution:

(A5) xc A cx

1

1 1

1 1 /1

N 1 1/1 .

Plugging this solution in (A4) then delivers

AZN1 1 1/1 CN w0N,

where

Z cx 1

1 1 /1

1 1 /1 1 1 1

,

which simplifies to equation (30) in the text. This reduced-form expression of the profit function immediately implies

t

00 american economic review - contracts and technology adoption

Documents