You are on page 1of 21

American Economic Association

Appropriation and Efficiency: A Revision of the First Theorem of Welfare Economics Author(s): Louis Makowski and Joseph M. Ostroy Source: The American Economic Review, Vol. 85, No. 4 (Sep., 1995), pp. 808-827 Published by: American Economic Association

Accessed: 09/12/2010 04:37

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aea.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

information about JSTOR, please contact support@jstor.org. American Economic Association is collaborating with JSTOR

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic Review.

http://www.jstor.org

Appropriationand Efficiency:

A Revisionof the FirstTheorem of WelfareEconomics

By LouisMAKOWSKIANDJOSEPHM. OSTROY*

TheFirstTheoremof WelfareEconomicsrestson theassumptionthatindividuals haveneitherprice-makingnor market-makingcapacities.Weoffera revisionin which individualshave such capacities.The revisionemphasizestwo keysfor market efficiency: (i) the need to align private rewards with social contributions calledfull appropriation,and (ii) the needfor an assumptionto counterthepossibilityof coordinationfailuresin the choice of producedcom- modities-called noncomplementarity.Wealsoemphasizethatinformationabout prices of unmarketedcommoditiesinvolvesdecentralizedknowledgeavailable onlytoproductinnovatorsandthatpecuniaryextemalitiesareimportantpotential sourcesof marketfailure.(JEL D51, D60, D62)

The FirstTheoremof WelfareEconomics providesa set of sufficientconditionsfor a price system to efficiently coordinate eco- nomicactivity.It is a beautifulresult,with a strikinglysimple proof. But its reliance on price-taking and complete markets con- tributes to a lack of explicit emphasis on strategic/incentiveissues. This paperoffers an alternative,complementaryset of suffi- cient conditions for efficient coordination, one that emphasizesthe importanceof full appropriationrather than price-takingbe- havior. Once appropriationis given center stage, our understandingof the reasons for eco- nomic efficiencydeepens. For example,the fit between the theory of market failure (whichalreadyemphasizesproblemsof ap- propriability,in the form of externalities) and the theory of marketsuccess becomes tighter.To give a second illustration,in the First Theoremthe set of availablemarkets

*Departmentof Economics,Universityof Califor- nia, Davis,CA 95616,and Departmentof Economics, UCLA, Los Angeles, CA 90024, respectively.Thanks to BryanEllicksonand anonymousreferees for their helpfulcomments.Thisresearchwas supportedin part by a grantfromthe NationalScienceFoundation.This is a revisedversionof ourworkingpaper,"TheMargin of Appropriationand the First Theoremof Welfare

Economics,"August1991.

808

must be complete; hence, product innova- tion cannot occur. By contrast,in a model for achievingeconomic efficiencybased on appropriation,innovativeactivitymaybe re- gardedas endogenous.The extent that the innovatorcan fully appropriatethe conse- quencesof his innovationsbecomesthe cen-

tral question, as far as efficiency is con- cerned. As this suggests, complete marketswill not be a maintained assumptionin what follows. There is, however, an important limitation:the model below does not ad-

dress issues associated with

andadverseselection.In laterworkwe hope to show how these too can be usefullyre- gardedas appropriationproblems.To givea suggestiveillustration,it is well knownthat "residualclaimantcontracts"can efficiently

resolvemoral-hazardproblemswhen agents are risk-neutral.By making the agent the residualclaimant,one forces him to appro- priatefullythe consequencesof his actions.1 The First Theoremis based on the Wal- rasian model of economic coordination,

moral hazard

1J. G. Head (1962) also arguesfor the centralim- portanceof appropriationfor welfare economics.He pursuesthe argumentbroadlyandvigorously.Herewe take a complementaryand moreformalroute,relating appropriationto the FirstTheorem.

VOL. 85 NO. 4

MAKOWSKIAND OSTROY APPROPRL4TIONAND EFFICIENCY

809

whereindividualshaveneitherprice-making nor market-makingcapabilities.By contrast, our revisionis based on an extension that we call a model of occupationalchoice.It is related to mechanism design in that the marketoutcomecan be describedas a Wal- rasianmechanismin whichpricesas well as marketedcommoditiesrespondto individu- als'occupationalchoices.Thus,in the model of occupationalchoice there is:

price-making:individualsmaybe able to in- fluence market-clearingprices by their choice of occupations;and market-making:individualsdeterminethe set of available markets by their choice of

occupations.

To illustrateits workings,productionat dif- ferent scales can be modeled as the choice of different occupations, so the producer maybe able to influencepricesby choosing to operateat a smallerscale. Or to illustrate market-making,different occupations may involve the introductionof different new commodities (in the model, markets are open only for commodities that can be actuallysupplied given individuals'occupa- tional choices; hence the choice of occu- pations has a market-makingrole). An equilibrium in the model is called an occupationalequilibrium. Our main result identifies conditionsun- der which occupational equilibriawill be efficient, in spite of the greater scope for individualchoice and self-interestedbehav- ior. We show that, if the followingtwo con- ditions are met, then the allocationof re- sourceswill be Paretoefficient:

fuillappropriation(FA):each individual'spri- vate benefitfromanyoccupationalchoice coincideswith his/her social contribution in that occupation;and

noncomplementarity(NC): a

conditionis satisfiedamongoccupational

subadditivity

choicesmade by differentindividuals.

The centralcondition,full appropriation, represents an extension of Pigou's "ap- propriation logic" which underlies both marketsuccessand failure.Specifically,FA requires private and social benefits to be aligned. Its role is to give individualsthe right incentives in their occupational choices, and hence in both their price-mak- ing and market-making. Because market-makingis endogenous, even with full appropriation,coordination failures can occur. (Early examples of this phenomenon are underinnovation traps pointed out by Tibor Scitovsky[1954].)The noncomplementaritycondition rules these

cases out. This resultis based on conditions

identified

by

Oliver Hart

(1980)

and

Makowski(1980b)as sufficientfor efficient product innovationunder perfect competi- tion. While both of these studies contain heuristicspointingto the importanceof ap- propriation,theirprimaryfocus is on a par- ticularapplication,ratherthan on incorpo- rating their findings into standardwelfare economics(i.e., the FirstTheorem). Some recent developmentsin macroeco- nomicsand industrialorganizationstudythe implicationsof strategiccomplementaritiesin imperfectlycompetitivemodels(see Russell

Cooper and Andrew John, 1988; Xavier Vives, 1989; Walter P. Heller, 1986; Paul MilgromandJohnRoberts,1990).3 In terms of our revisionof the FirstTheorem,ineffi- ciencies due to strategiccomplementarities arise from an amalgamof failures of FA and failuresof NC. We call our main result a "revision"be- cause its two assumptionsare stated in a languageunlikethat used in the FirstTheo- rem. Instead of emphasizingprice-taking and completemarkets,FA and NC directly describethe structureof individualpayoffs that give good incentives.Such payoffsmay arise in either a thick-marketor a thin- market setting. In either setting, perfect

2 The model of occupationalchoice collapses to a standardWalrasianmodel when each individualhas onlyone occupationalchoice(i.e., no choiceat all).

3The term"strategiccomplementarities"is foundin JeremyI. Bulowet al. (1985).

810

THEAMERICAN ECONOMIC RE1VIEW

SEPTEMBER 1995

competition is a key ingredient for effi- ciency:it leads to FA. By "perfectcompeti- tion" we mean not just price-taking,but somethingstronger:that individualsactually face perfectlyelastic demandsand supplies (PEDS).

Consider a thick-marketssetting first, where all commoditiesare standardizedand there are many buyers and sellers of each commodity.In such a setting, intense com- petition among buyers and among sellers will typicallylead to PEDS. This is probably the most natural setting for applyingthe standardFirst Theorem: both price-taking and completemarketsmakesense here. We show that in a thick-marketssetting, FA and NC are satisfied. Thus, in this setting the revision complementsthe First Theo- rem by makingexplicit the rewardscheme thatinducesefficiency.Of course,one could say that the standardpresentation,where price-takingis a shortcut for PEDS and completemarketsis a shortcutfor standard-

ization, yields a

shorterproof. Our claim is that takingthe

shortcutmeans bypassingthe central issue of appropriabilityunderlyingmarket effi- ciency. (See Remark 3 for an important instancewhere the link between appropri- abilityand efficiencywas bypassed.)

simpler statement and

Consider next a

more dynamic thin-

markets setting, where product innovation is an issue. Supposethat, amongall conceiv- able commodities (a huge set), most are neverinnovated.A difficultywith interpret-

ing the standardFirst Theorem in such a setting is its "uneconomical"use of price information:why should there be market

prices for all conceivablecommodities?Al- ternativelyput, while it mightbe an appeal- ing fiction to have an auctioneerannounce

prices of

auctioneermayhave no idea of what needs to be priced in a world of personalized commodities. The model of occupationalchoice per- mits a more appealing, decentralized de- scriptionof pricing in such a setting. We

assume that each seller only has access to the prices of currentlymarketedcommodi- ties and the prices of commoditiesthat he

can

standardized commodities, the

innovate.For example, a supplierof a

word-processingprogramwho is capableof also producinga (not yet existing)spread- sheet programknowsthe price at whichthe latter could be sold because that is part of his decentralizedknowledge:it comes from participatingin his segmentof the software market.But he mayhave no idea aboutthe potential price of a new item of clothing apparelor even the potential price of new computerhardwarebecausethat is not part of his technologicalexpertise.We call this "local price information." Markets are "complete"in the sense that the potential price of any new commodityis known by someone,butpriceinformationis decentral- ized because-assuming that any one seller can innovateonly a narrowrange of com-

modities-no

Local price informationmay be too de- centralizedto reflect accurate information about complementaritiesamong the com- moditiesnot currentlymarketed.The diffi- culty can be traced to a kind of pecuniary externalityin whichinnovationsby one indi- vidual affect the marketvaluationsof oth- ers' potential innovations.Nevertheless,we showthatFA andNC (hence efficiency)will result under perfect competition,provided that local price informationsatisfies a sistencycondition.As this suggests,in con- trast to a thick-marketssetting, achieving FA and NC in a thin-marketssetting is much more delicate. With complete mar- kets, all pricesare commonknowledge,and hence, price consistency occurs automati- cally. One could argue that the complete- markets assumptionin the standardFirst Theorem is a shortcutwhich yields a sim- pler statementand shorterproof;and again our claim is that takingthe shortcutmeans bypassinganotherissue:the pricedecentral- izationproblemsassociatedwith the alloca- tion of nonstandardizedcommodities(see

Remark4).

In this paper we do not strive for the utmost generality,preferringto emphasize principles.One simplifyingassumptionde- serves special mention. We shall assume that individuals have quasi-linearprefer- ences. This allowsfor cardinalmeasuresof individuals'privaterewardsand their social marginalproducts;hence, it greatly facili-

one knowsmost prices.

con-

VOL. 85 NO. 4

MAKOWSKIAND OSTROY:APPROPRIATIONANDEFFICIENCY

811

tatesemphasizingthe appropriabilitytheme.

From our work on the no-surplus approach

to perfect competition in both its ordinal and cardinal versions (Makowski, 1980a; Ostroy,1980;Makowskiand Ostroy,1987), we stronglysurmisethat there are ordinal analogues of our current results, just as there is both an ordinal and cardinalver- sion of the no-surpluscondition. There is also a simplifiedtreatmentof firms in this paper,relativeto the Arrow-Debreuversion of the Walrasianmodel.Whilethe "individ- uals"in the model of occupationalchoice maypossessproductionpossibilitiesandmay be interpretedas single proprietaryfirms, the model does not includefirmswith mul- tiple shareholders.This is just to avoid the notationalcomplicationsinvolvedin includ- ing shareholdingsand the requiredredistri- butionsof profits,complicationsthat would distractfrom our main goal: an alternative presentationof the FirstTheorem. The contentsof the rest of the paperare as follows.The modelof occupationalchoice and a basic example are describedin Sec- tion I. SectionII givesthe mainresult.First, it is provedthat rewardingindividualswith their social marginalproducts(full appro- priation)is good for incentives:it leads to efficient occupational choices, excepting perhaps for some coordinationproblems. Second,when the changesin the gainsfrom trade are subadditive(the noncomplemen- taritycondition),no coordinationproblems will arise. Section III gives the thick- and thin-marketsapplicationsof our revision. The ultimategoal of anyformalizationof an invisible-handtheorem is to guide our un- derstandingof marketsuccess/failure.With this in mind, Section IV concludeswith a brief discussionof some implicationsof our methodof proof.

I. TheModeland an Example

Althoughthe Walrasianmodel permitsa broadrangeof possibleinterpretations,the Walrasianconceptionof the coordinationof economic activityfosters a certainpoint of view that mightbe termeda "thick-markets mentality."According to this vision, the worldis describedby a fixedset of commod-

ity marketsas the paved highwaysof eco- nomic travel. In contrast to this, we will take a "thin-markets"approach.What we meanby this is thatwe shalltryto avoidthe fixed set of roads upon which individuals travel. The aim is to portray a world in whicheconomicactorsare connectednot by

several main highways,but by a myriadof individualbywaysof theirown construction. It is this alternativevisionthatunderliesthe

following.4

We pose the problemof the coordination of economicactivityby supposingthat each individualcan be one of several different types.Callthese typesthe possible"occupa- tions" for the individual. More formally, there are n individuals,indexed by i. For each individual i there is a given set of possibleoccupations Vi fromwhichhe must choose exactlyone. An assignmentof indi-

viduals to occupations is a v =

v,

set of

In order to include both pure exchange and production-and-exchangeeconomies,it will be simplerto workin tradespace.Thus we leave implicit i's consumptionand pro- ductiondecisions,which are his privatein- formation,to focus on what is essentialfor the model, his trade relationships.A trade for individuali is a point z i E Rewith the sign convention that positive (negative) components of zi represent his purchases (sales). Observe that i's preferences over tradeswill generallychangewhen his occu- pation changes, even if his consumption tastesremainconstant(e.g., if he becomesa baker then he will value the purchase of 1,000 bushels of wheat more than if he becomesa candlestick-maker).Thus,when i chooses an occupation vi E Vi, he chooses both a tradingpossibilityset Z(vi) and pref- erences over the possibletradesin Z(vi). To captureboth aspects of occupational choice,we view an occupationalchoice vi as an extended real-valued function (i.e.,

vi: Re , DRU { - oo)). Our convention is that

(vl,

,

vn) E X Vi.

-- xi Vi represents the

all possible assignments.

4Formally, thick markets will be a special case; see Section III-A.

812

THEAMERICAN ECONOMIC RE

IEW

SEPTEMBER 1995

those trades zi E R' whichare infeasiblefor

i are assigned a utility level of - oo.Thus, i's

trading possibility set in occupation vi is given by the effective domainof vi, that is,

Z(vs) = {zi: vi(zi) >

-oo}-

Thus the function vi does double duty: it identifiesboth i's tradingpossibilitiesin oc- cupation vi and also his preferences over possible trades. Examples illustratingthe flexibilityof the setup will be given. Notice that since we are in trade space, the zero vectorin RDcorrespondsto no trade,which we shall assumeis alwaysan option.' In additionto tradein the ecommodities, there is also a money commoditythat the individualcan use to establishquidpro quo

in exchange.Utility from these e+ 1 com-

moditiesdepends only on

i's characteristics

vi because all individuals have quasi-linear

utility functionswith respect to the money

commodity.That is, i's utility from

(zi, mi)

vi is

elR x DRwhen he is in occupation

givenby

Vi(Zi) + mi.

To preservethe quasi-linearityof the model we put no limitation on the amount of money i can supply (the spirit is that i never hits the boundaryof his money en- dowment).

5To give an illustration,supposethat individuali's consumptionset is Re andthat he has an endowment ei ElRe+ and preferencesover consumption bundles

givenby ui: lR +

a baker,his

productionpossibilitiesset willbe Yi c Re.

11;.Nowsupposethat,if he becomes

Thenforanygiventrade zi E Re, zi is not feasiblefor i

as a bakerif it calls on him to delivermore of some good than he could possiblysupply as a baker-for example,some candlesticks(assuminghe has no en-

dowment of candlesticks). That is vi(zi) = - oo if and

only if wi+ zi + yi

Yi E Yl. On the otherhand,if a trade zi is feasiblefor

i,

maximumutilityin consumptionhe can achievegiven

lRe for any productiondecision

his utilityfrom zi if he is a baker, vi(zi), is simplythe

the trade,that is,

vi(zi) =

max uj(wj + zi + yi)-

yi e

Y

The set of commodities which can be potentiallysupplied in the economy is re- strictedby individuals'occupationalchoices. To express this formally,let h index com-

modities,h =1,.,

. So for anygiventrade

x =

(xl,

,

x,

,xe)GERe

Xh

represents

the amount of commodityh purchased(if Xh> 0) or sold (if Xh< 0). Let

H(v) = {h: Zih< O

for some i and some trade zi E Z( Vi)}

represent the set of commoditiesthat can

be potentiallysupplied in v. We make the harmless assumptionthat all commodities

can be

{1,

,

potentially supplied: U, E e}.Define the subspace

H(v)=

Re(v) ={xE Re:Xh = 0

for all h - H(v)}.

Once the assignmentv to occupations is

made, trading is restricted to DRe(v).

are feasible for v if each

and Eizi = 0. Let Z(v) be

Trades z = (zi)

Re(v)

zieZ(vi)n

the set of all such trades.

Definition:Given an assignmentv, a Wal- rasianequilibriumfor v is a pair (z,p) such

that z

is feasible for v,p E DRe,and for all i,

vi(zi) -pzi > vi(z') -pz'

for all z' E Re(v).

That is, i maximizes vi(z') + m' subject to the (trading)budgetconstraintpz'.+ m' = 0.

Note that z' belongs to 1Re(v);therefore, the

values of Ph for h e H(v) are irrelevant since z' is zero there.

Exploiting the

quasi-linearity of

the

model, define the maximumpotential gains

from trade in v as follows:6

g(v) = max( Evi(zi):

z EZ(v)}.

6We assumethroughoutthatfor all i andall vuE V-, vi is continuouson Z(vi),Z(vi) is closed,and0 E Z(vi). We also assumefor all v E V, Z(v) is compact.Thus the maximum(in the definition)exists since Evi is a continuousfunctionon a compactandnonemptyset.

VOL. 85 NO. 4

MAKOWSKIAND OSTROY APPROPRL4TIONAND EFFICIENCY

813

As is well known,in quasi-lineareconomies maximizingthe gains from trade is both

necessaryand sufficient for

ciency.

achievingeffi-

Definition:The tradez is efficientfor v (syn- onymously,"efficient relative to v") if z is feasible for v and Evi(zi) = g(v). An alloca- tion (v,z) is (globally)Paretoefficientif z is efficientfor v and g(v) 2 g(v') for all v' E V.

As an applicationof the standardFirst Theorem of Welfare Economics, we have the following.

PROPOSITION1: If (z,p) is a Walrasian equilibriumfor v thenz is efficientfor v.

PROOF:

Let z' be anyother feasibleallocationfor

v. Then from the condition for Walrasian equilibrium,summingover the i and recall- ing Ez' = 0 since z' is feasible:

Evi(zJ) > Evi(z')

That is, Evi(zi) = g(v).

for all feasiblez'.

Nevertheless,a Walrasianequilibriumfor v

can evidentlybe very inefficient-not glob-

ally Pareto efficient-since

ble tradesmaybe restrictedto a veryineffi-

cient subsetof commodities:people maybe in the wrongoccupations.We will be inter- ested in how the "invisiblehand"may be able to lead the economy to a Pareto- efficientoutcome.

the Nash

equilibriumoutcome of a game in which people hold rationalconjecturesabout how Walrasian prices will change when they

the set of feasi-

Suppose occupationalchoice is

change occupations.

Let

i&: V--

Re be

a

Walrasianprice selection in the sense that

for each v E

(z, 6(v)) is a Walrasianequilibriumfor v; and let

, there are tradesz such that

Xi(v) = max{vi(zi) -

(v)zi: zi E Re(v))

representi's payoff(synonymously,"profit" or "utility')in the assignmentv underprices

#o(v).

Definition: An

(OE) is a triple(&,v, z) such that (z, #(v)) is a Walrasianequilibriumfor v, and for all i and all vl E Vi,

occupational equilibrium

IrT(V)2 ?i(vz,Vi)

wherevi

, signmentv with individuali omitted; and consequently,(vl,v') representsthe assign- ment v with only i's occupation changed

from vi to v.

(vl,

vi-1+vi1,

v,,)is the as-

The displayed condition expresses the idea that v is a Nash equilibriumin occupa- tional choice. In terms of traditionaleco- nomics, it picks up the idea of resources flowinginto their(privately)most profitable

uses. We will be interested in

conditionsunderwhich OE'sare Pareto ef- ficient. Note that if (,v,vz) is an occupa-

identifying

tional

equilibrium,

then

7ri(v) = vi(zi) -

O(V)zi.

In an occupationalequilibrium,the mar- ket outcomefor v is obtainedfroma prede-

termined selection among the Walrasian, and thereforeprice-taking,equilibriafor v. This should be regarded as a convenient

simplification in

which we

ignore the

monopolyproblemsin a givenv to focus on the monopolyissues across V. Note, how- ever, that the more variationthere is in the

choice of "occupations,"the closer this fic- tion will come to mimickingconventional

monopoly. For

with occupations/activitiesthat distinguish between different quantities of the same good supplied.Then, the seller can observe the Walrasian outcome from selling one unit, from selling two units, and so on (i.e.,

the seller can observe the aggregate de- mand schedulejust as a simple monopolist would).If buyersare permittedto have sim- ilarquantity-varying"occupations,"theywill attemptto exercisetheirmonopsonypower. An illustrationalongthese lines follows.

Example1 (Simplemonopolyas an occupa-

and partition

tional equilibrium):Let

individualsinto one seller, s, and B n -1 buyers indexed by b. The seller only likes

example, consider a seller

e=1

814

THEAMERICAN ECONOMICREVIEW

SEPTEMBER1995

Price supply in k* marginal cost a ,5((k')= a- B k/ seller's B ~~~~~inverse =a
Price
supply in k*
marginal
cost
a
,5((k')= a- B k/
seller's
B
~~~~~inverse
=a
-~q
demand
k*
\q
Quantity (q)
marginalrevenue

FIGURE

1.

THE

OCCUPATIONAL

IN EXAMPLE

1

EQUILIBRIUM

money.His possibleoccupationsare param- eterized by ke [0,K]; when in occupation k, he can supplyup to k units of the com- modity at a cost of 'q2 for any q E [0,k].

Thus,

V9={vS:

vk(z5)

=

Vs(Zs)\-o

(

kE[O,K])

2Zs

where

if

ZE[-k,O]

otherwise.

Buyers are identical. Each buyer has no initial endowment of the commodityand values its consumption according to a quadraticutility function;further,we view buyers as passive here, and so we model them with only one occupation.Thus, for

each buyer b, Vb

Vb(Zb)

(b

a

a

-

{Vb),where

X

1

72

2CZb

if Zb

0

otherwise

and where a and c are positiveconstants. Let q* be the output where the seller's inversedemandcurveintersectshismarginal cost curve (see Fig. 1), and let us assume

K > a. Writing 0(k) for i9( kV,S)

it is easy

to checkthat M(k)is uniqueand givenby

a

--

a -

c

B

k

q*

if

k <q*

otherwise.

For efficiency,we want the seller to pro- duce q* and thus to choose an occupation

k ? q*. But the seller's profit,

mizedin occupation

aB

k* = 2c+B

-(* ), is maxi-

where his marginal revenue equals his marginalcost (againsee Fig. 1). His equilib- riumoccupationalchoice exhibitsthe usual inefficiency associated with simple mono- poly:he caninfluencemarket-clearingprices 1?(k) by his choice of occupation(quantity). Hence, he enters the wrong occupation (undersupplies).

II. TheMainResult

A. A DivergencebetweenPrivateProfit and Social Benefit

The marketfailurethat Example1 illus- tratesmaybe explainedin termsof a failure of appropriationat the individualmargin (i.e., as arisingfrom a divergencebetween the seller's private reward and his social marginal product). To see this, we shall need some new terminology. As a preliminaryobserve that, for any individuali and any assignmentto occupa- tions v, the maximumpotential gainsfrom

tradein v withouti is given by

gi(vi) = max{ E vj(zj):

joi

Ezi

j*i

O}

[Recall that

representsthe occupationsof all individuals except i.] Thus, individuali's contribution to societyis naturallydefined as the differ- ence betweenthe gainsfromtradewithhim andwithouthim.

vi=(vj,

,vj_j,vj+,

,

V0)

Definition: The (social) marginalproduct of

individuali in occupation vi when others

are in occupationsv' is givenby

MPi(v) = g(v) -

gi(vi).

By contrast,

individuali

the private marginalproduct of

in occupation vi when others

VOL. 85 NO. 4

MAKOWSKIAND OSTROY:APPROPRI

TIONAND EFFICIENCY

815

Price marginal cost a average seller's inverse / ' \ / demand q* 0 Quantity
Price
marginal
cost
a
average
seller's
inverse
/
'
\
/
demand
q*
0
Quantity
occupationalequilibrium
when sellerfacesa downward
sloping demand curve
efficientlevel of production
when sellerfacesa downward
sloping demand curve

FIGURE

2.

VARIANT

OF EXAMPLE

1 WITH A FIXED

COST

are in occupationsvi is givenby

PMPi(v)

=

XiT(V)*

In an occupationalequilibrium(0, v,z),

Y(v))is Walrasianfor v, z is effi-

since (z,

cient relative to v; that is, Evi(zi) = g(v).

Thus, EPMPi(v) = g(v). Further, we have the following.

THEOREM1 (InappropriabilityTheorem):

If (z, #a(v)) is a Walrasianequilibriumfor v then,for each individuali,

PMPi(v) < MPi(v)

Thus, EMPi(v) 2 g(v).

PROOF:

Let z' be any set of tradesthat are feasi-

ble without i (i.e., that satisfy E.

1z =0).

Then as in the proof of Proposition1, from the definition of a Walrasianequilibrium for v,

E

ji

vj(zj) -

E

j$i

t(V)zj

E

j$i

vj(zj

Thus, E

v.(z

) - E

,(v)zj

gi(vi). Mul-

tiplyingioth sides of this inequalityby -1

and adding

n

E

j=l

v(zi)

> 1vj(zj) to both sides shows

-

E

j

i

vj(zj)

+

E

j0i

(V)Zj < g(v) - g'(v).

However,recallingthe feasibilityof z, the

left-hand side just

equals vi(zi) - M(v)zi, that

is, PMPi(v); while the right-handside equals

MPi(v).

claimed.The second assertionof the theo-

rem now follows immediatelyfrom the fact

that EPMPi(v) = g(v).

Hence,

PMPi(v) < MPi(v),

as

So "at best"in an OE, everyonewill be rewardedwith his full socialmarginalprod- uct. We call the resultthe "inappropriabil- itytheorem"to emphasizethatusuallysome individualswill be rewardedwith strictly less than their MP's.This was illustratedin Example1. In this example,for any assign-

ment v, the seller'ssocial marginalproduct in v is the wholegain fromtrade in v since no one else has any of the commodityto trade;for example,when k = k*, then the seller's MP is the entire shaded area in

816

THEAMERICANECONOMIC REVIEW

SEPTEMBER 1995

Figure 1. But his PMP, his profit, is just a fractionof g(v) since he faces a downward- sloping demandcurve and so must give up some of g(v) to the buyers as consumer surplus;for example,when k = k*, then he must give the darker-shadedconsumersur- plus trianglein Figure1. Thus,in the exam- ple, the seller appropriatesless than his MP. This explainswhy he undersuppliesin the OE: beyond k*, the changein his PMP is negative,even though the change in his socialmarginalproductis still positive. Whilethe undersupplyequilibriumin Ex- ample 1 is bad, thingscould get worse:the unique seller may not want to produce at all. Specifically,considerthe variantof Ex- ample 1 in which the seller, in additionto

his marginalcost, has a fixed cost C that he mustsufferif he enters anyoccupationk >

0. Supposethis fixed cost exceedshis equi-

librium profit in Example 1; that is, his (now) U-shaped average cost curve lies strictlyabove his downward-slopinginverse demandcurve(see Fig. 2). Thus,while i(k) remains unchanged from Example 1, the unique occupational equilibrium now in- volves autarky:the seller does not produce

any of the commodity.7But also suppose that the sum of producer and consumer surpluswould be strictlypositive for some output levels (i.e., the area of the dark

7The reader may have expected a nonexistence problem.Indeed, such a problemdoes occur in the Walrasianversionof thisvariantbecauseof the discon- tinuityin the firm'ssupplycurvecaused by the fixed cost. In the Walrasianversionthe firm'soccupational

choices are trivial, say Vs = {v }, where

v;(zs)={o

|2z52-C

00 oo-

if z E[-K ,O)

if zs=0

otherwise.

Thereis no nonexistenceproblemin the occupational- choiceversionbecausethe firmtakesinto accountthat the equilibriumprice will change when it changes occupations(quantity);andwhen it entersanyoccupa- tion k > 0, the fixed cost C is a bygonecost for the firm,so its supplycurvein any givenoccupationk is continuous. Nevertheless, the occupational-choice model does not guaranteeexistenceof equilibriaeven if each vi is concave (see Roberts and Hugo

Sonnenschein,1979).

shaded rectangle in Fig. 2 [his losses in occupation q*] is smaller than the shaded consumersurplustrianglein the figure),so the no-productionequilibriumis Pareto in- efficient. In accord with traditionalteach- ing, the sourceof the inefficiencyis that the seller cannot appropriatethe consumersur- plus his commoditywould produce. Or, in our language, he would not get the full social marginalproductof his commodity.

Remark1 (Imperfectcompetitionand appro- priationlogic): As is well known,the market failuresillustratedin Figures1 and 2 would disappearif we allowedthe seller to act as a perfectlydiscriminatingmonopolist,not just as a simple monopolist.But this extension of appropriationlogic to imperfectcompeti- tion is somewhat misleading. Apart from the well-knowninformationaldemandscon- fronting the perfectly discriminating monopolist,Theorem1 can be used to show a fundamentaldifficulty.To illustrate,con- sider the case of bilateralmonopoly.While each of the two parties could appropriate all the surplus from the other, certainly both could not simultaneouslyappropriate. That is, the sum of their MP's is strictly greaterthan the total gains from trade be- tween them-there just is not enough sur- plus to go around.This is alwaysthe case when there is imperfect competition (see Makowskiand Ostroy,1987).

B. GivingIndividualsTheirMarginal ProductsIs Goodfor Incentives

Traditionalappropriationlogic, as amen- ded here to emphasize individualsrather thancommodities,saysthat anydiscrepancy between private and social marginalprod- ucts will typicallybe accompaniedby mar- ket inefficiency,as illustratedby Example1 and its variant.But it also saysthat, if there is no such discrepancy,private initiative leads to sociallyefficient allocations.Let us now formallyexaminethis second assertion, that givingindividualstheir marginalprod- ucts is good for incentives.Accordingly,let us suppose that, in an OE (, v,z), private and social marginalproductscoincide at v in the sense that the followingconditionis

VOL.85 NO. 4

MAKOWSKIAND OSTROYAPPROPRIATIONANDEFFICIENCY

817

met:

Fullappropriation(FA):For everyindividual

i such that Vi * {vi} and every vi E Vi,

PMPi(vj',vi) = MPi(vj',v i).

(Notice that any individual for whom Vi=

{vi}, a singleton,cannot influence prices by his occupationalchoice since his choice set is trivial;hence, we need not worryabout his incentives.) Introducethe following suggestivenota- tion. Denote a change from vi to some other occupation vi' by Avi. Let

APMP. =PMP1(v' v) -PMP1(v) Avj

and let

AMPv

Av

=MPi (vi', v i)-

mpi (V) i

In this notation,FA impliesthat,in anyOE,

APMPi

Avj

AMPS

- for all Av1.

AvI

But notice that

AMP.

A

-MRi(

Vi', v i) )-MPi

( v)

= [g(v'vi)

-

[g(v)

-

9=gv'v)g(v)-

- gi(vi)]

gi(vi)]

=

Ag(V)

;v

Hence, FA impliesthat, in any OE,

APMPi

Avj

Ag(V)

= for all Avi.

Avj

But in any OE, individualschoose their occupationsto maximizetheir privatepay- offs; hence, in any OE satisfyingFA,

APMPi

Ag(v)

Avj

-

A

Avi

<0

forall Avi.

That is, the assignment v is not Pareto- dominatedby any other assignmentv' in- volvingan occupationalswitchby only one individual, Avi. Stated as a theorem, we have provedthe following.

THEOREM 2 (Partial Optimality):If it is the case that0, v,z) is an occupationalequi- libriumsatisfyingFA, thenfor all i and all

v' GE Vi,

g(v)

2 g( vl,vi) (

The word "partial"in the name of the theorem is to suggest two ideas. First and mostobvious,the theoremis onlya "partial" optimalityresult in that it does not claim that v is globallyPareto efficient. Second, the word "partial"suggestsin what sense v is efficient;here the word is intendedto be suggestiveof partialderivatives.The assign- ment v cannotbe Pareto-dominatedby any changesof occupationin the individual"co- ordinatedirections"(i.e., by any Avi); but it may be Pareto-dominatedby coordinated

changesin the "diagonaldirections"(i.e., by

some Av=

eral individualschangingoccupationssimul- taneously).Thus the theoremdoes not pre- clude the possibilityof coordinationfailures in an OE, even when everyoneis rewarded with his or her social marginalproduct.We will examinethis possibilityin Section II-C below.

Remark2 (The mechanism-designconnec- tion): Readers familiar with Vickrey- Clarke-Groves mechanisms (William Vickrey, 1961; Edward H. Clarke, 1971; Theodore Groves,1973)from the theoryof mechanismdesignwill see an intimatecon- nection between the proof of Theorem 2

and the proof that such mechanismseffi- cientlysolve the revelationproblem.This is no accident; see Makowski and Ostroy (1987, 1992) for an interpretationof these mechanismsas mimickingthe logic of the perfectly competitive market. The differ- ence is that, while in the mechanismlitera-

ture there is a central allocator who can

costlesslyfind an efficientallocationonce it

(Av1, , Avn) that involves sev-

818

THEAMERICANECONOMIC REVIEW

SEPTEMBER 1995

knows the true types of individuals,here individualsmust find such an allocationon their own. Thus there is the possibilityof coordinationfailures,to be discussedin the next subsection. CombiningProposition1 and Theorem2, we are immediatelyable to state the follow- ing corollary.

COROLLARY 1 (A Partial Extension of

the First Theorem): Supposethat only one

individual has

a

nontrivial occupational

choice, that is, Vi = {vi}, a singleton, for all

individualsexceptone. Then,anyoccupational equilibriumsatisfyingFA is Paretoefficient.

The corollaryimplies that we can con- structan exampleof a Pareto-efficientOE by modifyingExample 1 so that the seller always earns his social marginalproduct. Sincethe discrepancybetweenhis PMPand his MP resulted from facing a downward- slopingdemandcurve,hence havingto-,give up a partof g(v) to the buyersas consumer surplus,it should suffice if we modify the exampleso that the seller faces a perfectly elastic demandfor his product.

Example2 (An efficientoccupationalequilib- rium):This exampleis the same as Exam- ple 1 except that each buyer'spreferences now exhibita constantmarginalutilityfrom consumingthe commodityequal to a for the first d/B units, where d > K the seller's maximumpotential supply.8 Since d > K, the seller'sinverse demandcurve is now perfectlyelastic in his operatingrange; that is,

i9(k)=a

forall ke[O,K].

8That is, now Vb = {V'b}, where

Vub(Zb)=

aZb

a

Zb

- X0

-

d

J-

-

1

2CZb-

d \2

B J

rdl

J

if ZbE [o

if Zb>

d

B

otherwise.

Price

a

0

/

/

marginal

cost

I

.

q

|

'

d

demand in

Example 2

Quantity

efficient innovation underperfectcompetition

FIGURE3. THE OCCUPATIONALEQUILIBRIUMIN

EXAMPLE2

Hence, the seller'sprofit lTs(14,vs) 1Ts(k) is maximizedby choosingany occupationk E

[q,

cost curve intersects the perfectly elastic portionof his demandcurve(see Fig. 3). So, in accordwith Corollary1, the equilibrium is efficient.Notice that FA is satisfiedsince for any occupationk he maychoose

K] and producingwhere his marginal

iTs(k)

=

g(vsk,vs)

= MPs(vsk,vs).

It is interestingto observethat this effi- cient outcomeis the limitingoutcomeof the occupationalequilibriain Example1 as one replicatesthe numberof buyers.As B in- creases,the inversedemandcurvein Figure 1 rotates around point a on the vertical axis, becoming more and more elastic. Hence asymptoticallythe seller's profits would equal his full socialmarginalproduct (the entireshadedareain Fig. 1). Giventhis context, one can regard Example 2 as a finite "magnification"of the limitingecon-

omy (notice that the length of the flat seg- ment in any buyer'sutility function, d/B, goes to zero as B approachesinfinity;hence,

buyers'preferencesapproachVb,the prefer-

ences of the buyers in Example 1, as B approachesinfinity).A similar,but asymp- totic, exampleappearsin Hart(1979).

VOL.85 NO. 4

MAKOWSKIANDOSTROY:APPROPRIATIONANDEFFICIENCY

819

Either the finite "magnification"or the asymptoticversion of the example tells an interestingmoral:a uniqueseller of a prod- uct maystill be a perfectcompetitor(in the

sense of facing a perfectly elastic demand for his productin the relevantregion),pro- vided his desired supply is less than the

demands of

highest-valuing buyers.

Viewingthe seller as innovatinga new com- modity, the example illustrates that the phrase "a perfectly competitiveinnovator" is not an oxymoron.This is the lesson of the

literatureon productinnovationunderper- fect competition (e.g., Hart, 1979, 1980;

Makowski,1980b,1983).

C. TheCoordinationProblem

the

Continuingwith our suggestivenotation,

let

Ag(v)/Av

= g(v') -

g(v),

where

Av =

(Av1,

some other assignmentv' =(v1,

is an equilibriumassignment,FA ensures

, Avn) denotes a change from v to

,vn).

If v

A(V) < 0

for all Avi

but it does not ensure

Ag(V) < 0

Av

for all Av.

For some Av the changesin the gains from trade may be strictlysuperadditive:

Ag(v)

iAvj

Ag(v)

AV

Example3 (Computerhardwareandsoftware):

Suppose a computer hardware manufac- turercould supplya powerfulmachine,the value of which would be enhanced by a sophisticatedgraphicsprogram,and a soft- ware manufacturercould supplya sophisti- cated graphicsprogramwhose value would increaseif operatedon a powerfulmachine. But either one without the other is suffi- cientlycostlythat it does not coverthe price buyersarewillingto pay.Therefore,neither commodityis produced, although the sum

of their costs is less than the value of joint innovation.(A numericalillustrationwill be given in Section III below.) The upshot is

that if v

is the assignmentin which neither

hardwarenor softwareis innovatedand i is a potentialhardwareinnovator,then

A(V) < 0

Avj

for all Av1

where Avi is any occupationalswitch that involvesi innovatinghardware.Thus, i will stayout of the hardwarebusiness,even if he can fully appropriatehis contribution.A

similar statement holds if

softwareinnovator.Nevertheless

i is a potential

Ag(v)

Av

> 0

for some Avthatinvolvesboth the hardware and software innovatorsproducing;so the OE is inefficient.

D. A Revisionof theFirstTheorem

Ourmainresultsaysthat,providedevery-

one is rewarded with his or her social marginalproduct,suchsuperadditivityis the only possiblesourceof inefficiency.Saythat the changes in the gains from trade are subadditiveat v or, synonymously,satisfythe noncomplementarityconditionif the follow- ing holds:

Noncomplementarity(NC): For any assign-

ment switch Av= (Av1,

, AVn),

Ag(v)

Ag(v)

Avj

AV

THEOREM3 (A Revisionof the FirstThe- orem of Welfare Economics):Any occupa- tional equilibrium(0, v,z) satisfyingFA and NC is globallyParetoefficient.

PROOF:

By definitionof an OE, for all i and all

Avi,

APMP~ I ?0.

Avj

820

THEAMERICANECONOMICREVIEW

SEPTEMBER1995

Thus,

APMPi

Au. <ppi0.

-

But by FA, the

E AMPS /Avi = E Ag(v)/Avi. NC,

left-hand side

equals

Thus, using

Ag(v)

0.

That is, no assignmentv' Pareto-dominates v. Hence, since z is efficient for v, (v,z) is globallyParetoefficient.

III. TwoSettingsfortheRevision

FA and NC characterizea rewardscheme that gives individualsgood incentives.What marketstructuresinduce such payoffs?We highlighttwo: a thick-marketssetting and a thin-marketssetting. The former builds a bridge to the standardFirst Theorem;the lattertakesus into a moredynamicenviron- ment. Commonto both settings,perfectcompe-

tition is

a key ingredientfor efficiency. It

leads to

FA. The followingdefinition cap-

tures the notion that there is perfect com-

petitionin an occupationalequilibrium.

Definition:All individualsface perfectlyelas- tic demandsand supplies(PEDS) in the oc- cupationalequilibrium(0, v,z) if there ex- ists a price vector p such that for all i and all vi' E Vi,

Uh(Vi,V)

=Ph

for all commoditiesh E H(vi',v ).

PEDS saysthat no one individualcan influ- ence market-clearingpricesby switchingoc- cupations.It is strongerthan the hypothesis of price-taking:Since U(vi', v) is a Wal-

rasian selection, PEDS

switched occupations from vi to vi', then prices actually would not change; so his price-takingis rational.

implies that if

i

To establishthe link from PEDS to FA, we will need a technical assumption,that individuals'occupationalchoices Vi are suf-

ficientlyrichin variety.A precisestatement of the assumptionappearsin the Appendix, precedingthe proof of Theorem4. We call an occupationalequilibriumregularif it sat- isfies the richness assumption. (The Ap- pendixcontainsthe proofs of all remaining results.)

THEOREM4: For anyregularoccupational equilibrium,

PEDS=FA.

Thus, under perfect competition, each economic agent is rewardedwith his full social contributionin whateveroccupation he mayenter:

97Ti( *)

MPi(

*)-

The injunctionto "profit-maximize"(i.e., to seek to maximize one's selfish interests) agrees with the injunction to "maximize one's contributionto society."As we have alreadyemphasized,such a rewardscheme gives good incentives, absent coordination problems(failuresof NC). As we are about to show,suchproblemscannotarisein thick markets.

A. ThickMarkets

Supposethereis a fixednumberof homo- geneous commodities traded, with many buyersand sellers of each. In such a thick- marketssetting, competitionamong buyers and sellers implies that no one individual will be able to influence market-clearing prices. Price-takingbehaviorand complete markets-the twin assumptionsthat drive the standardFirst Theorem-make sense. We will showthat

thick markets =: FA and NC =: efficiency.

If we interpretthe standardFirst Theorem as saying that thick markets lead to effi- ciency (the first and last items in the schema), then in this setting the revision

VOL.85 NO. 4

MA4KOWSKIANDOSTROY:APPROPRIATIONANDEFFICIENCY

821

may be viewed as supplementingthe First Theorem by specifyingthe rewardscheme that inducesefficiency.

Definition:Given an occupational equilib- rium(#, v,z), we will say that all commodi- ties are standardizedif

H(v) = H(vl,v')

=

.

.

.

for all i and all vl e Vi.

Marketsare thickif both (i) all commodities are standardizedand (ii) all individualsface PEDS.

The formaldefinitionhighlightstwo fea- tures of thick markets:(i) all commodities can be suppliedirrespectiveof anyone indi- vidual'soccupationalchoice,and(ii) there is perfectcompetition.Condition(i) is weaker than the hypothesis of many buyers and sellers of each commodity.As emphasized above, manybuyersand sellers would lead to PEDS. Thus, the two conditions are at least informallylinked. We have alreadyseen that PEDS implies FA. The next theorem states that when PEDS is combinedwith standardizedcom- moditiesthen NC will be satisfiedautomati- cally. That is, in the absence of product innovation, coordination failures cannot arise.This helps explainthe notableneglect of such failures in discussions of welfare economics starting from a thick-markets perspective.

THEOREM 5: In any regularoccupational equilibrium(0, v,z),

thick markets =* FA and NC.

Remark 3 (Marketsocialism):The thick- marketsinterpretationof the standardFirst Theorem is perhapsthe most naturalone. But it is not the only interpretation.In the market-socialismtradition, price-taking is assumedto be independentof any market structure(e.g., even in one-firm,socialized sectors). The First Theorem is interpreted as showingthat marketsocialismwill lead

to

efficiency.The outcomeresultsfrom the

fact that Walrasianprices measure the so- cial values of all resources at the

margin-even in the

priation. Therefore, the price system will coordinate an efficient outcome, provided agentsact as price-takers.To illustrate,con- siderthe Walrasianversionof Example1 in whichthe firmdoes not considershadingits production;hence its occupationalchoices

absence of full appro-

are

trivial,

Fs = {vK}.

The

Walrasian

equilibriumfor this economy will be effi- cient-the firmwillproduceq* units-since

it is requiredto act as a price-takereven

thoughin realityit facesa downward-sloping demand curve. Thus, the standard First Theorem"goes through"in spite of the fact that the firmonly appropriatesa smallfrac- tion of its social marginalproductwhen it produces q* in the Walrasianequilibrium. Thisillustratesthat"coordinationviaprices"

is possible even in the absence of "full

appropriationvia prices"-provided agents act as price-takerseven if it is not in their self-interest.It is throughthe logic of coor- dination via prices, not appropriationvia prices, that the First Theorem is tradition- ally proved.For a critiqueof marketsocial- ism as relyingtoo heavilyon coordination

rather than appropriation logic, see Makowskiand Ostroy(1993).

B. ThinMarkets

Supposenowthatcommoditiesare not all standardized;rather, commoditiesare het-

erogeneousbecause sellers have the ability

to personalizetheir productlines. Call this

a thin-markets setting. Further, suppose there is only a limitednumberof commodi- ties producedin any assignmentto occupa- tions (e.g., specializingin the productionof software A means not specializingin the productionof softwareB), and as a result a large number of commodities,even most, are not tradedin equilibrium. Becausethe set of all conceivableinnova-

tions is huge, the hypothesis of complete marketsbecomesproblematic.Further,with decentralized knowledge, although each agent knows best what products he could innovate, the set of possible innovationsis not commonknowledge.In such a setting,

822

THEAMERICANECONOMICREVIEW

SEPTEMBER1995

to assumethata Walrasianauctioneercould know of all possible commoditiesand pub-

licly announcea price for each is not only

inconsistent with the hy-

pothesisof decentralizedknowledge. In a thin-marketssetting, it is more ap- propriate to view each individualas only possessing"local price information."Recall that at any assignmentv, individualscan

heroic, but also

only register demands for

H(v);it is onlyas sellersthattheycanchange

the commodity space from WeMv)to Re(u,v).

Therefore,as buyers,the relevantpricesare Vh(O), h E H(v).We shallassumethe follow-

ing:

Local PriceInformation:At v, seller i only

has

commoditiesin

access

to

prices

10h(v,

Vi),

h E

Uv,{H(v,v9}.

An individualthereforeknowsthe pricesof commoditiesin H(v) as well as the pricesof the other commoditieshe can supply.As-

sumingPEDS, the idea of local price infor- mation can also be expressed in terms of the price vector p: any seller i only knows the prices, Ph, of the commoditieshe can innovate.Thisinformationmightcome from test marketingor simplyfrom an accurate estimatepicked up as a result of being "in the business."Note that when U, {H(v,vi)}

is only a small subset of {1, ,

countera problemnot presentin the thick- marketssetting:no individualhas access to most prices. Nevertheless,as we shall see, efficient outcomes may occur, providedin- dividuals'local price informationis consis- tent. The possibilityof inconsistentlocal price informationis intimatelyconnectedwiththe possibilityof coordinationfailures (failures of NC). We can illustrate using the hard- ware-software example introducedin Sec- tion 1I-C. Suppose that individual 1 can produce only hardware,while individual2 can produce only software.Assume that it takes 1.25 units of money to produce each unit of hardwareor software.All potential buyersof hardwareand softwarehave iden-

tical tastes:

e},we en-

vi(r, s) = min(r + 2s, s + 2r).

That is, each buyeris willingto pay $1 per unit of hardware(r) or software(s) if the other commodityis unavailable;but each buyer is willing to pay $3 per hardware- softwarepackage. It is easy to check that no innovationof either hardwareor softwareis an occupa- tional equilibrium.In the absence of soft- ware,individual1 perceivesthat he can only get $1 per unit of hardware,less than his marginalcost. Similarly,in the absence of hardware,individual2 perceivesthat he can only get $1 per unit of software.Both re- main out of business, in spite of the fact thatthe value of a hardware-softwarepack- age exceeds the cost of such a package. Observe that each individual'slocal price informationis correct: $1 is the market- clearingprice per unit of hardware(in the absenceof software),and similarly$1 is the market-clearingprice of software (in the absence of hardware).But the individuals' local priceinformationis inconsistentin the sense that if they pieced their information together, the price vector p = (1,1) would not clear the marketfor bothhardwareand software.Indeed,since buyersare willingto pay $3 for each hardware-softwarepack- age, at p = (1,1) there would be an excess demandfor both commodities. The examplemotivatesthe followingdefi- nition.

Definition:Suppose (0,v,z)

tional equilibriumsatisfyingPEDS. Individ-

uals'local price informationis consistentif

the

fies the followingconditionfor each individ-

ual i and each possibleoccupationvi':

is an occupa-

price vector p (defined in PEDS) satis-

7ri(vi,,v9 2 vi'(z'i)-pzf

for all z'iE-0Re.

Consistencysays that local price informa- tion can be pieced together to form one vector of prices which, if it were knownto all, wouldnot changetradedecisionseven if each individualcould trade any combina- tionof commoditieshe wishes(notjustthose in H(vi', v). In the hardware-softwareex- ample, PEDS is satisfied:individual1 can sell as muchhardwareas he likes at a price

VOL. 85 NO. 4

MAKOWSKIAND OSTROY APPROPRIATIONAND EFFICIENCY

823

of $1 (if softwareis unavailable).9A similar statement holds for individual2. But the consistencyconditionis violated by buyers:

they could increase their payoffs if they could trade anywhere,in D2 at prices p = (1,1).See Hart(1980)andMakowski(1980b) for sufficientconditionsfor consistency;not surprisingly,differentiabilityof preferences (absenceof strictcomplementarities)playsa key role.

THEOREM

equilibrium(,

6: In any regular occupational

v,z),

PEDS and price consistency

FA and NC.

The hypotheses of Theorem 6 are strictly

weaker than those of Theorem 5: in thick- market environments,price consistency is

triviallysatisfied since

Basi-

cally, with thick markets,local and global

H(vi', v) = R.

price information coincide.

The (more general) thin-marketcondi- tions for efficiencymay be fruitfullyinter- pretedusingthe languageof externalities.If an individualimposes externalitieson oth- ers, then he does not fully appropriatethe consequencesof his actions. The model of occupational choice does not include the possibilityof real externalities,but it does allowindividualsto imposepecuniaryexter- nalities:anyindividual,by switchingoccupa- tions, may affect the terms of trade that othersface and hence, indirectly,affect oth- ers'welfare. It is useful to distinguishtwo sorts of pecuniaryexternalities,"market-price"and "reservation-price"externalities.The signif- icanceof PEDS (i.e., perfectcompetition)is that it rules out the possibilityof market- price externalities:under perfect competi-

9PEDSimpliesthat,as in Example2, anyinnovator of any commodityh will alwaysreceive the buyers' reservationprice for his commodity,no matter how manyunitshe sells.The intuitionis thatsincethe price Ph mustcontinueto clearthe marketfor h even if the innovatorswitchedto occupationsthat allow him to

produceless andless of it (hence,occupationsin which

h is getting scarcerand scarcer)Ph must equal the

buyers'reservationprice. The argumentis formalized

in the proofof Theorem4.

tion no one individualcan affect the market prices others face. With thick markets,this is the only type of pecuniary externality possible. But with innovation,anothertype of pecuniaryexternalitymayoccur:an indi- vidual, by innovatingone commodity,may be able to influence the reservationprices other innovatorssee for their potential in- novations.To illustrate,in the case of the hardware-softwareexample,if individual1 marketed his hardwareand sold it at $1

each-even

doing so-then individual2 would see that buyers'willingnessto pay (i.e., reservation price) for his softwarehas increased from $1 to $2, and so he would find it profitable to innovatehis software.The hypothesesof Theorem6, namely,PEDS andpriceconsis- tency, rule out, respectively,market-price and reservation-priceexternalities. Let us say that there are no pecuniary externalitiesin an occupationalequilibrium if it satisfies both PEDS and price consis- tency. Then, our analysismay be summa- rizedby:

though he would lose money

absenceof pecuniaryexternalities

FA and NC

efficiency.

That is, marketsettingsthat precludepecu- niaryexternalities(e.g., all thick-marketset- tings and some thin-marketsettings) will induce payoffsconsistentwith efficient be- havior.

Remark4 (Decentralizedknowledgeofpnces):

Thereis an interestingcontrastbetweenour concept of local price information and FriedrichA. Hayek'sview of the price sys- tem in "The Use of Knowledgein Society" (1945). Hayek, while stressing the local characterof economic knowledge of time and place, views the price system as com- mon knowledge guiding individualsin the

sociallyefficient use of tion (e.g., his famous

different individualswould cope with an economy-widescarcityof tin reflectedin its

higherprice).In a worldof nonstandardized commodities,however,the sharpdistinction

their local informa- illustration of how

824

THEAMERICANECONOMICREVIEW

SEPTEMBER 1995

between local knowledgeof individualcir- cumstancesbut"global"knowledgeof prices

needs to be blurredto recognizelocal price information. Local price information, al- though it economizeson price information,

introduces the possibility of

inconsistent

perceptions of the value of innovations. Hence the possibilityof coordinationfail- ures.

IV. ConcludingRemarks

In what sense is this is a revision? By substitutingFA (or PEDS) for price-taking and by substitutingNC (or consistentlocal priceinformation)for completemarkets,we took a muchlonger route to get to more or

less the same conclusionsas the FirstTheo- rem. This was our goal. In our view, how that theoremis provedis at least as impor-

tant as what is

proved. The currentstate-

ment and proof of the FirstTheoremis too concise;the argumentdoes not exhibitsuffi- cient potential complicationsto allow one to grasp the essentials of why competition leads to efficiency.We call attentionto two featuresof our proof:one is how individual behavioris modeled, and the other is the role of pecuniary externalities. Below we briefly indicate why these features are im- portant.

IndividualBehavior.-In the last decade, a gap has developedbetween general-equi- libriumtheory and other branchesof eco- nomicsdue largelyto differencesin sophis- tication about the meaning of "pursuitof self-interest."With the recent spread of game/informationtheoretictechniques,the price-takingbehaviorof generalequilibrium appearsto be naivelysimplistic(cf. Samuel Bowles and Herbert Gintis, 1993; Joseph Stiglitz,1993).Fromthe more sophisticated perspective,generalequilibriumwouldseem to be fine for the more traditionalissues of determiningthe relativeprices of standard- ized commodities (e.g., in the Hecksher- Ohlin approachto internationaltrade),but when it comes to the manyeconomic phe- nomena based on contracting,asymmetric information, and strategic behavior, one must look elsewhere.In our view, this per- spective is incorrectbecause general equi-

libriumneed not be identified with naive price-takingbehavior.More importantly,it is also ill-advised:the full appropriationun- derpinningof perfect competitionprovides both a canonicalmodelof the kindof incen- tive system that efficientlychannels poten- tially opportunistic behavior and also a

canonical reason why-in

full appropriation-such behavior can be-

come sociallyinefficient.

the absence of

Pecuniaryand Real Externalities.-Ingen- eral equilibrium,inappropriabilityis associ- ated with real externalitiesand is modeled as the incompletenessof marketsassociated with incompletelydefined propertyrights. In the revision,propertyrights to all con- ceivablecommoditiesare well-defined.The only kind of inappropriabilitypermitted is of the pecuniary-externalitieskind, associ- ated either with the absence of PEDS or the absence of consistentlocal price infor- mation. Scitovsky(1954) to the contrarynotwith- standing, pecuniary externalitieshave not been takenvery seriouslysince A. C. Pigou (1912) mistakenlyidentified as efficiency- reducing appropriability problems what

turned out

changes(AllynYoung, 1913;FrankKnight, 1924).The moraldrawnfrom Pigou'serror was that pecuniaryexternalitiesshould be distinguishedfrom welfare-relevantowner- ship externalities