ECONOMIC PLANNING AND THEKNOWLEDGE PROBLEM

Israel M. Kirzner

Introduction

Itis nowabout40 years since Professor Hayek decisively Identifiedthe key misconception underlying mainstream welfare economics.This misconception, Hayek argued, was responsible for failure toappreciate the critique ofthe possibility ofrational economic calcu-lation under central planning—a critique stated most forcefully andclearly by Mises, and further developed by Hayek himself. As hasbeen demonstrated by Professor Lavoie (1985), the true import andsignificance ofthe Hayekian lesson was simply not grasped by sub-sequentwelfareeconomists writingon the socialist calculation debate,even though Hayek’s work was widely cited.

In this paper we attempt both to restate and to extend Hayek’sinsight concerning the “knowledge problem” and its implicationsfurcentral economic planning, whether comprehensive in scope orotherwise. In the following paragraphs wecite Hayek’s own formu-lation of his Insight, and make certain observations concernIng It Insubsequent sections of the paper we start from a rather differentpoint of departure, and in this way eventually arrive at our restate-ment and extension of the Hayekian position—spelling out somerather radical Implications ofour restatement,

According to Hayek (1945, pp.71—78):The peculiar characterof the problem of a rational economic orderIs determined precisely by the fret that the knowledge of the cir-cumstances ofwhich wemust make use never exists In concentratedor Integrated form but solely as the dispersed bits of incompleteand frequentlycontradictory knowledge whichall the separate Indi-viduals possess, Theeconomicproblem ofsociety Is thus notmerelya problemofhow to allocate “given” resources—If“given” is taken

CatoJournal,Vol. 4~No~2 (Fall 1984). Copyright0 CabInstitute. All rightsreserved.The author Is Professorof Economics at New York University.

407

CATO JOURNAL

to mean given to a single mind which deliberately solves the prob-lem set by these “data.” ItIs rathera problem ofhow to secure thebest use of resources known to any of the members of society, forends whose relative Importance only these Individuals know. Or,to putItbriefly, It Is aproblem ofthe utilization ofknowledge whichIs notgiven to anyone in its totality.

Let us call this knowledge problem “Hayek’s knowledge prob-lem,” We note, at this initial stage in our discussion, thatthe positionmight be takenthat Hayek’s knowledge problemdoes not, at least atfirst glance, render immediately irrelevant the paramountcy of thesocial efficiency norm. It is true, as Hayek pointed out, that thedispersed character of knowledge means that the economicproblemis not that of allocating “given” resources, where “given resources”means resources given andknown to a singlemind. But surelyknow!-edge itselfis a scarce resource. So that the task ofthecentral plannermay be seen, in the light of Hayek’s knowledge problem, as thatofmaking themost effective use of the available dispersed knowledgeexisting In society at a given moment—with the attainment of suchuse constrained by the communication and search costs made nec-essaryby the dispersed character ofthe available information.

It might seem, therefore, that there is nothing in Hayek’s know!-edge problem which places it outside the scope of economic plan-ning. The knowledge problem, it might be argued, complicates theplanning task: It introduces anewly recognized, subtle, andcomplexresource (knowledge); it compels attention to the special character-istics ofthis resource (its dispersed character); and Itcalls for atten-tion to a specIal class of costs (those required for search and com-munication). But Hayek’s knowledge problem can still, itmight seem,be subsumed under the overall economicproblem, traditionallycon-ceived In resource-allocation terms. The central theme of this paperIs to deny this understanding of the implications of Hayek’s knowl-edge problem.’

‘Hayek (1979, p. 190) has more recentlydeepened ourunderstanding of the problemofdispersed knowledge as going farbeyond thatof “utilizing information aboutpartic-ularconcretefactswhich individuals alreadypossess.” Henowemphasizes theproblemofusingthe abilities thatindividuals possess todlwooerrelevantconcrete Information.This leads Hayek to point out that because aperson“will discover what he knows orcan find out only when faced with aproblem wherethis will helps” he may neverbeable to “pass on all the knowledge he commands. ,..“ In recent unpublished workProfessor Lavole, building on Insights contained in the work of MIchael Polanyi, alsohas emphasized the relevance of”tacltknowledge” for the social problem ofutilizingdIspersed knowledge, The present paper arrives at similar conclusIons but from asomewhat different starting point

408

THE KNOWLEDGE PROBLEM

In this paper I shall argue that this problem of securing the bestuse of dispersed knowledge, In fact cannot be translated into aspecial case of the more genera! problem of securIng an efficientallocation ofsociety’s resources. It will follow that societal planning,by its verycharacter, is Incapable of addressing Hayek’s problem—such planning can only frustrateand hamperthose spontaneous mar-ket forces that are capable of engaging this knowledge problem. Inorder to develop these arguments it will be useful to introducea “knowledge problem” that at first glance seems very differentfrom Hayek’s.

The Individual Plan and the Knowledge Problem

In the course of everydaylife mancontinually plans. Economistshave come to formulate the individual plan as the seekingof amax-Imum: The planner is Intent on arriving at a constrained optimumposition. This is theconcept of theeconomizingdecision articulatedwith special precision and force by Lionel Bobbins in 1932, andwidely adopted since then as the basic building block of mIcro-economic theory. We wish to point out that there is an inescapablepotential “knowledge problem” surrounding this concept of theindividual plan.

The notion of the plan presupposes some deliberately aimed-atentity—say, utility, or profit—that is to be knowingly maximized. Itfurther presupposes known resource constraints. In Bobbins’ termi-nology, both the ends and the means are presumed tobe given. It isthe presumed knowledge of these planning circumstances by theplannerthatpermits theeconomistto perceive theplanas thesolvingofa constrainedmaximization problem. Thevalidityof theplan itself,it should be noted, depends entirely on thevalidity oftheassumptionthattheplanner in fictaccurately knows thecircumstances surround-inghis prospective decisions. If the planner does not know what itIs that he is seeking to achieve, or does not knowwhat resources areathis command, orwhatthe efficacy oftheseresources is with respectto sought-after goals, then his plan—no matter how carefully for-mulated—is unlikelyto result in the best possible outcome.

Wecan now identify the knowledge problem potentially relevantto each individual plan: Because of Inadequacies In the planner’sknowledge of his true circumstances, his plan may fall to yield anattainable optimum. Let us call this knowledge problemthe “basicknowledge problem.” This will distinguish it from what we havecalled “Hayek’s knowledge problem.” Itwill also indicate our inten-tion to demonstrate that Hayek’s knowledge problemcan be consid-ereda specIal case of what we call the “basic knowledge problem.”

409

CATO JOURNAL

To be sure, what we have called the “basic knowledge problem”seems, at first glance, to bear little resemblance to Hayek’s. Hayek’sknowledge problem consists In the dispersed character of availableinformation; ourbasicknowledge problem consists In an Individual’ssimple ignoranceofthe circumstances relevant to his situation. Fur-ther reflection on both Hayek’s knowledge problem and our basicknowledge problem, however, will reveal the important sense inwhich Hayek’s problem is indeed a basic one. Let us first clarify apossible misunderstandingconcerning thebasic knowledge problem.

The Basic Knowledge Problem andthe Economics ofSearch

Itmight be argued that the basic knowledge problem surroundingevery individual plan can be entirely escaped through the additionofnewplanning stages.After all, Ifa plan seems likely to fall becauseof Inadequate availability ofa necessary resource, this threatenedfailure need notbe final; it may possibly be avoided byappropriatepreliminary planning to obtain this resource. Instead ofsimply for-mulating a single plan directed atthe Immediate attainment ofthefinal objective, it is necessary to introduce intermediate objectivesto be pursued in the course ofadditional preliminary plans. Perhaps,then, the basic knowledge problem, too, merely calls for judiciouspreliminary planning.

Fromthisperspectivethebasic knowledge problem would appearmerely to represent an Inadequacy in the available supply of anImportant resource, namely, knowledge. This Inadequacy would thenbe seen to call for a plannedsearch to acquire the necessary infor-mation. In principle, it might then be thought that the basic knowl-edge problem can be escaped, at least to the same extent that anyother problem arising out ofa resource shortagecan be escaped. Tothe extent that It is worthwhile, a preliminary plan ofsearch to over-come theshortage ofnecessary information may totally eliminate thebasic knowledge problem. To the extent that such a costly search isheldnot tobe worthwhile, the basic knowledgeproblem would seemmerely to express theinescapable scarcity constraints inherent in theplanner’s situation. For the economist such an inescapable scarcityproblem means that there is noproblem at all, in therelevant sense,

To theextent that the knowledge problem is escapable, It can (andpresumablywill) be escaped;to the extent that It is notworthwhileescaping, there would appear to be no basic knowledge problem atall—since we defined the basic knowledge problem in terms offall-tire to realize an attainable optimum. If lack ofknowledge renders

410

THE KNOWLEDGE PROBLEM

ahypothetical optimum unattainable, it can generate no basic know!-edge problem. And, Ifthe cost ofacquiring theknowledge is prohib-itive, then the “hypothetical optimum,” while indeed “attainable,”Is in factno optimum at all.

But this line of argument cannot be sustained. The basic knowl-edge problem potentially surrounding each individual plan is by itsnature Inescapable. Certainly a deficiency in knowledge may be ableto be rectified by search, and the individual planner will no doubtconsider, in his preliminary planning, whether such search shouldbe undertaken or not. But the basicknowledge problem—involvingpossible failure to achieve an attainable optimum—remains. In factthe possibility of preliminary planning to acquIre knowledge onlyexpands the scope ofthe basic knowledge problem.

Letus consider an individual engagingin a plan seekingto achievea valued objective. Let us call this “Plan A.” In formulating thespecific steps that should be taken in pursuing Plan A, the individualrealizes that he lacks needed pieces of information. He thereuponengages in a plan to attaIn these preliminary objectives, namely,these missing pieces of information. Let us call thIs search plan,“Plan B.” We may see Plan A as having been expanded (as a resultof realized ignorance) to include the plannedattainment of neededpreliminary objectives, so that Plan B is “nested” within expandedPlanA. And we may identify the later steps to be taken in thecourseofexpanded Plan A—those steps subsequent to theattainment oftheinformation obtained in Plan B—as “Plan A’.” (Plan A’ consists inthe steps that would have made up the originally envisaged Plan Ahad theplanner in fact not lacked the needed information.) We caneasily see that the basic knowledge problem is a potential hazardboth for Plan B and Plan A’. Plan A, which includes Plan B and PlanA’, is ofcourse subject to the vulnerabilities ofbothof them.

Even ifPlan B is completely successful in attaining precisely theoptimal amount of information capable of being searched for (andbelieved to be worth the costs of such search), Plan Ais nonethelessstill subject to the hazards of thebasic knowledge problem. After all,although our decision maker, in originally seeking to formulate PlanA, realized he lacked specific items of information (and thereforeundertook preliminary Plan B), he may have in fact lackedfar moreinfonnatlon than he realized. (Most important, suchunrealized infor-mation may have taken the form ofa firm, buttotallymistaken, beliefin the validity of information that is totally false.) Moreover he maybe mistaken in his belief that the items of information he realizeshe lacks are necessary for the implementation of Plan A. He alsomay be mistaken In his belief that he really lacks these items of

411

CATO JOURNAL

Information (In the relevant sense of lacking). The truth may be thatthese items are already within his grasp.

For example, he may believe that Plan A, in the course of whichhe must communicate with individual Z, requires information con-cerning Z’s telephone number, Information that he believes himselfto lack, so that he undertakes Plan B to search for Z’s telephonenumber. But the truth may be quite different The truth may be thatZ is in fact thewrong person to speak to altogether,or again the truthmay be that Z is now in the very same room wIth our planner, so thatno knowledge ofZ’s telephone number is in fact needed for Plan A.Or, it may be, our planner does not really “lack” Z’s telephonenumber at all; he may In fact have that telephone number clearlywritten and Identified in the listoftelephone numbers that lies nextto the telephone that he uses. Given these possibilities for sheererror that surround Plan A’ and/or Plan B, possibilities In which theplanner is entirely unaware ofthe extent ofhis ignorance, Plan A(because it includes Plan A’ and Plan B) may be far from optimaleven IfPlan B is wholly successful in terms of its own objectives. Inaddition It may be the case that Plan B overlooks more efficientavailable ways of achieving its own objectives (for example, theremaybe less costly methods of search of which the searcher Isunaware.)

To sum up, the possibility ofplanned search for Information per-ceived to be lacking does not eliminate the knowledge problem.First, theplanned search may itselfbe undertakenwithoutawarenessthat more efficient search techniques are easily available. Second,the information sought may in fact not justify the costs of searchbecause the truth (ofwhich the planner Is unaware) is that the infor-mation is not of significance for theattainment of theplanner’s ulti-mate objectives. Third, quite apart from the information that theplanner realizes he lacks and for which he may attempt to search, hemay lack other information that he does not realize he lacks and forwhich he does not think ofundertakingany planned search.

Central Planning and the Knowledge ProblemWe are now in a position to appreciate Hayek’s insight into the

problem of dispersed knowledge as revealing the central planningtask to be one that is deeply and inextdcably bound up in the basicknowledge problem.Letusputourselves in theposition ofthecentralplanners, earnestly and single-mlndedly seeking after the mosteffi-cientpossible pattern ofresource allocation.

Our task as central planners is to formulate a plan for society in amanneranalogous to that in which an individualplans his own course

412

THE KNOWLEDGE PROBLEM

ofactiol. We formulate our socialplan with respect to specific socialobjectives and In the light of specific perceived arrays of availablesocial resources.’ This framework for the central plan is relevant (inprinciple and with the necessary changes having been made) bothfor comprehensive socialplanning and for central planning designedmerely to supplement or modify, rather than totally replace, decen-tralized economic activity. The analogy between the social plan andthe individual plan compels us, ofcourse, to acknowledge therele-vance ofthe basic knowledge problem for the socialplan, In exactlythe same wayas we have seen it to be an inevitable and inescapablehazard for the individual plan. Hayek’s contribution permits us torecognize that the central plan may be subject to hazards (arisingout ofthe basic knowledge problem) that might havebeen escapedby decentralized planning.

That the centralIzed plan is inescapably subject to the hazards ofthe basic knowledge problem follows almost trivially from the verynotion ofthebasic knowledge problem. Because the individualplan-ner may not be aware ofhis true circumstances and may be totallyunaware ofhis ignorance, his best formulated plans mayfall to yieldan attainable optimum. The central planners, too, may be unawareof their own ignorance concerning the true circumstances relevantto thesocialplan. Our understanding oftheimplicationsofdispersedknowledge deepens our appreciation of the seriousness of thebasic knowledge problem, and reveals how the hazards of thisproblem might have been entirely escaped in the absence of thecentralized plan.

Recognition of the fact of dispersed knowledge—especially asregards “theknowledgeofparticularcircumstances oftimeand place”(Hayek, 1945, p.80)—immediately illuminates our understanding ofthe basic knowledge problem that threatens centralplanners, For aplannIng individual, the basic knowledge problem derives from thepossibility that what he thinks he knows about his circumstancesmaydifferfrom whatbe mighthave known (without additional resourceexpenditure) had he beenmorealert or aware ofthe true environment

The same possibility, ofcourse, is fully relevant for a central plan-ner, but it Is heightened by the central planner’s peculiar predica-ment What the central planner thInks he knows about the relevantcircumstances must necessarily take theform ofwhat he thinks heknowsabout the availability ofdispersed bitsofknowledge thatcan

‘we avoid here raising any ofthe well-known dimeultios that surround (a) thenotionof ahierarchy of social— analogous to aranking of individual objectives, and (b)therelated notions of socialefilelency and socialchoice,

413

CAm JOURNAL

somehow, at some cost, be mobilized In formulating and Imple-menting the social plan. There is little chancethatthecentral plannercan everknow where to find, or how to search for, all the items ofdispersed Information known somewhere in the economic system.Moreover there seems little chance that thecentral planner can everbe fully aware ofthe nature ofextent ofthespecific gaps In his ownknowledge in thisregard. lie mayrealize, in a generalway, that thereis Information the locationofwhich he is ignorant, butthis gives hunnoclueon where to look, The endresult isthat theplanner Is unlIkelyto be able to exploit all the information that is within his command.Clearly the dispersal ofinformation Is responsible for a new dimen-sIon ofapplIcation for the basic knowledge program.

EarlIer we raised the possibility that Hayek’s knowledge prob-lem—despite its novelty—might be subsumed under the generaleconomic problem, traditionally conceived in terms ofachieving anefficient allocation of given resources (with available Informationincluded as an important given resource). We can now see howInappropriate it is for us to consider central planners as being ableto grapple with Hayek’s knowledge problemin terms ofconventionalplanning to achieve a constrained optimumpattern ofresource allo-cation. The unknown ignorance that is the heart ofthe knowledgeproblem createdbythe dispersal of information defies its being ableto be squeezed into the Procrustean bed ofthe allocation plan. Justas the individual planner is unable to grapple deliberately with thebasIc knowledge problem surrounding alldecision making, so too isthe centralplanner unable to Invoke planning techniques to grappledeliberately with Hayek’s knowledge problem.

Whatrenders theHayekian knowledge-problem critiqueofcentralplanning so devastating is the circumstance that in a market system,with decentralized decision making, the insoluble knowledge prob-lem confronted by central planners tends to dissolve through theentrepreneurial-competitive discovery procedure.

The Entrepreneurial-CompetitiveDiscovery Procedure

The Hayekian case for decentralized decision making has fre-quently been misunderstood. Alltoo frequently thiscase is presentedas arguingonly that the decentralized market economy escapes theproblem of dispersedknowledge becauseprices accurately andeco-noinically convey necessary Information to relevantdecisionmakers(replacing any need for them to know all the detailed informationthat is dispersedthroughout thesystem). It must be emphasized that

414

THE KNOWLEDGE PROBLEM

while this line ofreasoning is certaInly present In Hayek’s work, ftfitils to doJustice to the hill implications ofthatwork

To argue that market prices convey Information, and thus directlyovercome the problem ofdispersedknowledge, is to make a case formarkets thatdepends upon adubious assumption; namely, thatmar-kets are always ator close to equilibrium. It is only in equilibriumthat ft can be claimed that a market participant guided by marketprices is automatically steered toward those actions that will coor-dInate smoothly with the actions ofall the other (similarly guided)market participants. Moreover, to make the assumption that marketsare close to equilibrium is essentially (quite apart from our otherreasons for feeling uncomfortable concerning the realism of thisassumption) to beg (rather than to overcome) the Hayekian problemof dispersed knowledge. After all, just as the phenomenon of dis-persedknowledge offers a formidable challenge to central planners,thisphenomenon offersmarkets awholly analogous challenge, namely,that ofachieving, in fitct, those arrays ofmarket prices that will clearmarkets.

One does not “solve” the problem of dispersed knowledge bypostulatingprices thatwill smoothly generate dovetallingdecisions.Dispersed knowledge is precisely the reason for the very realisticpossibility that market prices at a given date are unable to clearmarkets and to ensure the absence of wasted resources. The truthIs that the marketdoes possess weapons to combat (If not wholly toconquer) the problem of dispersed knowledge. These weapons areembodied in theworkingsofthepricesystem, but not in theworkingsof a hypothetical system of equilibrium prices. The importance ofprices for coping with the Hayeklan knowledge problem does notlie In the accuracy of the information which equilibrium prices con-vey concerning the actions of others who are sImilarly informed.Rather, its importance lies In the ability of disequilibrium prices toofl~rpure profit opportunIties that can attract the notice of alert,profit-seekIngentrepreneurs. Where market participants have i~iledto coordinate their activities because of dispersed knowledge, thisexpresses itself in an array of prices that suggests to alert entrepre-neurs where they may win pure profits.

We know very little about the precise way in which pure profitopportunities attract entrepreneurial attention. Butthere can be littledoubtabout thepowerful magnetism which such opportunIties exertTo say that pure profit opportunities attract attention is not at all tosay that awareness ofthese opportunities is secured by deliberate,costly search on the part of entrepreneurs. Rather ft is to recognizethat the lure ofpure profit is what permits an individual decision

415

CA1o JOURNAL

maker to transcend the limits of a given,perceived planning frame-work, and to escape, to some extent, the basic knowledge problemthatsurrounds all individualdecision making,Man’s entrepreneurialalertness operates at all times top~ his narrow planning activitieswithin the broader framework of human action.3 At the very sametime as man is routinely calculating the optimal allocation of givenresources with respect to given competing ends, he keeps an entre-preneurial ear cocked for anything that might suggest that the avail-able resources are different from what had been assumed, or thatperhaps a differentarrayofgoals might be worth striving for.

This entrepreneurial elementIn humanaction Is what responds tothe signals for pure profit that are generated bythe errors that ariseout ofthe dispersed knowledgeavailable in society. It Is this yeastthat forments the competitive-entrepreneurial discovery process,tendingtoreveal to marketparticipants more and more ofthe relevantinformation scattered throughout the market It is this entrepreneu-rial-competitive process that thus grapples withthatbasic knowledgeproblem we Ibund Inescapably to confront central planning author-ities. Tothe extentthatcentral planning dispaces theentrepreneurialdiscoveryprocess, whether onthe society-wide scaleofcomprehen-sive planning or on the more modest scale of state piecemeal inter-vention In an otherwise free market, the planners are at the sametime both smothering the market’s abIlity to transcend the basicknowledge problem and subjecting themselves helplessly to thatvery problem. The problem’s source is Hayek’s dispersed knowl-edge: Central planning has no tools withwhich to engage theprob-lem ofdispersed knowledge, and Its very centralization means thatthemarket’s discovery process has been Impeded, Ifnot brought toa hillhalt

Markets, Firms, and Central PlanningAtleastasfarbackasCoase’sl937paperonthetheoryoftheflrm,

it has been recognized that each firm in a market economy is anIsland of local “central planning” In a seaof spontaneously seethIngcompetitive market forces. Withinthe firm, activities are coordinatedby central direction, not by market competition via aprice mecha-nism. Our discussion in thispaper throwslight, perhaps,on the forcesgoverning the location of the boundaries separating the realm offreely adopted “central planning” from that of thecompetitive pricesystem.

‘On this compare Mlses (1949, pp. 253-54).

416

THE KNOWLEDGE PROBLEM

We have seen that the replacement of market discovery (workingthrough entrepreneurial alertness to profitopportunities) bycentralplanning generates a new scope of potency for the basic knowledgeproblem arisingout ofthe dispersal of knowledge. In a free market,therefore, any advantages that may be derived from “central plan-ning” (for example, the avoidance of “wasteful” duplication oftenapparently present In situations of market rivalry) are purchased attheprice ofan enhanced knowledge problem. We may expect firmsspontaneouslyto tendto expandto thepoint whereadditional advan-tages of “central” planning are just about offset by the incrementalknowledge difficulties that stem from dispersed information. On asmall scale these latter difficulties may be Insignificant enough to beworth absorbing in order to take advantage ofexplicitly coordinatedorganization.Knowledge dispersedoverasmall geographicalororga-nizational area may mean a Hayekian knowledge problem that is,unlIke that relevant to large, complex entities, solvable throughdeliberate search. Beyond some point, however, theknowledge dif-ficultieswill tend to reduce the profitabilityoffirms thatare too large.Competition between firms of different sizes and scope will tend,therefore, to reveal the optimal extent of such “central planning.”

On the other hand ifcentral planning is Imposed on an otherwisefree market, whether In comprehensive terms or not, such planningwill almost always involve the knowledge problem, and to an extentnot likely to be justified by anyadvantages that centralization mightotherwIse afford. Governmentally enforcedcentralplanning sweepsaway the market’s delicate and spontaneous weapons for grapplingwith the knowledge problem. Such centralized planning is by Itsvery nature, and the nature of the knowledge problem, unable tooffer any substitute weapons of its own.

ConclusionWe should remember that the nature of theknowledge problem is

such that its extent and seriousness cannot be known in advance.Part of the tragedy of proposals for industrial policy and economicplanning is that their well-meaningadvocates are totallyunaware ofthe knowledge problem—the problem arising out ofunawareness ofone’s ignorance.

ReferencesCoase, Ronald H. “The Natureofthe Firm.” Economica (NS) 4(1931): 386—

405.Hayek, Friedrich A. “The Use of Knowledge In Society.” Atnesican Eco-

nomic RevIew 35 (September 1945): 519-30. Page rekrences are to the

411

CATO JOURNAL

reprinted version In Friedrich A. Hayek, Individualinn and EconomicOrder, London: Routledge and Kegan Paul, 1949.

Hayek, FriedrlchA. Law,LegWationandLiberty,VoI. 3, The PoliticalOrderofa Free People. Chicago: University ofChicago Press, 1979.

Lavole, Donald. Rivalry andCentral Planning. Cambridge: Cambridge Uni-versity Press, 1985 (forthcomIng).

Mises, Ludwig. Human Action.New Haven: Yale University Press, 1949.Robbins, Lionel. An Essay on the Nature and Sign4ficance of Economic

Science. London: Macmillan and Co., 1932. (Second EditIon, 1935).

418

“ECONOMIC PLANNING ANDTHE KNOWLEDGE PROBLEM”:

A COMMENTLeonid Hurwicz

My comments are dIrected to issues raised by Professor lCirzner’sanalysis’ rather than the various current proposals for a nationalIndustrial policy, but I hope that they may be relevant as a back-ground for analyzing such proposals. And, since I shall be indicatingmydisagreementwith some ofthepoints made byProfossor Kirzner,letme stress that I am in complete sympathy withhis poInt of depar-ture, namely, the emphasis on the dispersion of Infonnation amongeconomic decision-making units (called by him, “Hayek’s knowl-edgeproblem”) and theconsequentproblemof transmission ofinfor-mation among those units.

Much of my ownresearch work since the 1950s has been focusedon issues in welfare economics viewed from an Informational per-spective. The ideas of Hayek (whose classes at the London Schoolof Economics I attonded during the academic year 1938—39) haveplayed a major role In Influencing my thinking and have been soacknowledged. But my ideas have also been influenced by OskarLange (University of Chicago, 1940—42), as well as by Ludwig vonMises in whose Geneva seminar I took partdurIng 1938—49.

By now there is a considerable literature in this area.’ A carefulperusal ofthis literature,! believe, would show thatProfessor Kirz-ner’s opening statement (that “the Rayekian lesson was simply notgrasped by subsequent welfareeconomists”r does not apply topres-ent-day mainstream welfare economics, whether or not it applies toearlier work in this area.

CatoJounud,Vol.4, No.2(FaU 1984). Copyright CCato Institute.All rights reserved.The author Is aRegents’ Professor ofEconomics at the Universityof Minnesota.

~ (1984).

‘Anexcellentpresentation ofmany recentIdeas Is foundIn helter (19W).3Kinnor (1984, p~401).

419

CATO JOURNAL

Let me make clear at this point that I do not intend to argue theadvantages or disadvantages ofwhatever may be meant by “centralplanning” or“industrial policy.” Rather, mypurposeis to instill someskepticism with respect to oversImplified arguments sometimes usedin this area. I shall argue that thepaper before us—despite its manyvaluable insights—does not providean adequate basis for formingajudgment concerning the respective merits of “the free market,”“centralplanning,” or other forms of government intervention in theeconomicprocess. (This is the Scottish verdict: neither“guilty,” nor“notguilty,” but“notproven.”)This is so for severalreasons, includ-ing the ambiguity of the terms used, the Implicit assumptions pos-tulating a “classical economic environment” (to be defined below),problems of incentives,andvaluejudgments transcendingefficiency.

Terms such as “central planning” and “free market” have manyinterpretations. In analyzIngthemerits andweaknesses ofthemarketprocess it is important to distinguish perfectly competitive marketsfrom those that are monopolistic, oligopolistic, or otherwise imper-fect. For example, in adecreasing cost industry only a few firms maysurvive even though there is freedom of entry. Such a market maybe called “free,”but it is oligopolistic rather than perfectlycompetitive.

The well-known theorem of welfareeconomics asserts the Pareto-optimality (efficiency) ofperfectlycompetitive equilibrium. Butthereis no theoretical basIs for asserting thatmonopolistic or oligopollstlcmarkets result in efficient resource allocation, Indeed, elementaryanalysis shows that uniform price monopoly or oligopoly is, in gen-eral, Pareto-inefficient.4 Moreover, under conditions of increasingreturns, perfectly competitive equilibrium is, in general, Impossiblebecauseprofitmaximization withparametrically treatedprices wouldcall foreither a zero or infinite level ofoutput.Therefore, it is difficultto see how one could justifr, In the presence of Increasing returns,the claim of efficiency of “free markets,” whether the latter term IsInterpreted as perfect competition or merely free entry.

Difficulties with Increasingreturns constitutebuta special case ofa more general problem. The theorem guaranteeIng the optimalityof perfectly competitive equilibrium assumes the absence of exter-nalities, which also excludes public goods.’

fl’heso”callcd Cease Theorem, as Iunderstand I~merelyexplores theImplications ofpostulating that freelyacting well-Informed economicagents will arriveMa Pareto-opthnal allocation. Butuniformpricemonopolyoroligopoly does notsatlsfrthispostulate.‘According to the usual definition of a public good, the utilization of Its services bypersonA does not detract from the possibility of utilization by person B. Governmentoraprivate party may supplya public good.

420

COMMENT ON KIBZNER

Furthermore, as seen above, there are circumstances (such asIncreasing returns) where no set of prices can balance supply anddemand; hence, perfectly competitive equilibrium Is impossible.Thus, to guarantee both the possibility of existence of balancingprices—technically known as the existence ofperfectly competitiveequilibrium—and of theoptlmailty of perfectly competitive equilib-ria, the relevant theorems make a series of assumptions, excludingsuch phenomena as externalities, public goods, increasing returns,indivisibilities, and so on, When all these assumptions (which ruleout the “troublesome” phenomena) are satisfied, we speak ofa alas-slcal economy or aclassical environment. The theorems guarantee-ing thepossibility and optimality of perfectly competitive equilibriatherefore presuppose a classical environment

In practice, however, one often encounters nonclassical environ-ments. Pollution is an example of an important negative externalitywhile information derivable from new inventions or pleasure deriv-able from musical compositions illustrate positive externalities orpublic goods. Nationaldefense is another example ofavitally impor-tant public good. Bridges and dams exemplifr indivisibilities, andthereare many instances ofeconomies of scale--known as “increas-ing returns (to scale).” I know of no basis for claiming that, in suchsituations, the free-market process (however defined) would yieldoptimal resource allocation.

Ithas beenshown in anumberofcontributions (Mount and Reiter1974; Osana 1978; Hurwicz 1977) that in classical environments theperfectly competitiveprice mechanism uses a minimal size messagespace; that is, ituses the minimum numberofvariables for transmit-ting information between economic units. This confirms Hayek’sview concerning the Informational efficiencyofthepricemechanism.Butithas beenshownbyexamples (Hurwicz 1977; Calsamiglia 1977)that In the absence of convexity,6 it may be Impossible to find anyefficient decentralized mechanism using a finite-dimensional mes-sage space.

In addition to the difficulties In achievIng efficiency innonclassicalenvironments, one must note thatefficiency is only one of the pos-sible criteria on which valueJudgments concerning economic sys-tems are based. Somepeople may be prepared to sacrifice efficiencyfor the sake of egalitarian ideals; far them, the fect that the marketprocess yields efficiencymaybe insufficient—even ifone can assumeaclassical environment Ofcourse, this attitude need not lead to the

°Inoreaslngreturnsare aspecial case of the nonconvexity of a production possibilityset.

421

CATO JOURNAL

discarding of the market process, butperhaps to a supplementationwith such devices as taxes or subsidies. Hence a case for some roleofthe government may be made on value Judgment grounds even ifit were admitted that govermnent intervention results in loweringtheefficiency ofthe system.

I also see another problem in relating Professor ICirzner’s argu-ments to the above-quoted theorems concerning the optimality ofcompetitive equilibria.The paper before us is emphatic in avoidingreliance on markets actually being In equilibrium. But it is only theposition of competitive equilibrium which, under classical assump-tions, is guaranteed to be optimal. So the best one can say for dis-equilibrium situations is that they may tend to an equilibrium. Infect~a studybyArrow,Block, and thepresentwriter (1959) identifiedcertain classes of situations where such a tendency toward equilib-rium (that is, stability) is present But subsequent research (for exam-ple, Scarf 1960) has shown that this tendency is not always presenteven in perfectly competitive markets. In any case, it Is difficult tosee how, in the absence ofstabilizing forces, a theoretical claim canat all be made that markets produce efficiency.7

The paper’s major emphasis Is on what the author calls “the basicknowledge problem.” To the extent that this goes beyond Hayek’sdispersion of knowledge, this “basic knowledge problem” seemssimply tobe thefrctthatmost decisions—whetherbyplanners, firms,or individuals—must be madewithout complete and accurate infor-mation.6 There is no disagreement on this point But some of thediscussion seems to imply that such uncertainty makes any rationalbehavior logically impossible.

With that I cannot agree. There do exist well-developed theoriesofrationalbehaviorunderuncertainty, Including thetheory ofsearch.Statistical decision theory is but one branch of this discipline. Buteven if one accepts thepracticaldifficulties ofoptimal search behav-ior, one is thenled to the framework of so-calledboundedrationality(Simon 1912; Badner 1915).

I would, ofcourse, agree that people often act on beliefs that arefectually Incorrect At best, one can only hope foraction that Is rationalIn the light of foresight—not ofhindsight But this difficulty arisesfor everyone,not Just for planners. True, if’the planner’s information

‘I stress the term “theoretical”because neither Professor iCinnernor Iattempt to dealwith the empirical evidence concerning theactual performance of dlfi~renttypes ofeconomic systcms.‘SeeKlr.cner (1984~p.410): “blurbasic knowledge problemconsists Man Individual’ssimple Ignoranceof thecircumstances relevantto his situation.”

422

COMMENT ON ICIRZNER

or beliefs are based on imperfect transmission, this does constitutean additional source oferror. But that, again, Is the Hayekproblem!

As mentioned above, the market mechanism does minimIze therequired message space, but its claims are based on theassumptionof classical environments.

In nonclassical environmentsor wherevalues otherthan efficiencyare important, acase may be made, at least; for the inadequacy of the“freemarket” process, and possiblyIn favor of a role forgovernmentIntervention. But such a role should not be identified with centralplanning. In fact, this role may be confined to Introducing andenforc-big what may be called “rules of the game.” In particular, this mayInvolvethe creationof property rights through patents or copyrights.Creating such rights doesamount to government Intervention in thefree marketprocess butdoes not constitute what I would callcentralplanning. Similarly, government’s role In enforcing Income transfersthrough taxes and subsidies constitutes intervention but not centralplanning. The latter term should perhaps be reserved for the type ofintervention that might be called micro-targetIng—of which Indus-trial policy or price controls and rationingmay be examples, and inwhich the government makes decisions concerningoutputs, inputs,or prices of specific commodities or groups of commodities. Eventhenit Is important to distinguish betweenall-encompassing centralplanning (attempted In the Soviet Union) on the one hand, and ele-ments of planning grafted onto an otherwise market-type economy(as is typical of Western countries) on the other. Thus one shouldrecognIze that between a iaissez-falre economy and an all-inclusivecentral planning (micro-targeting) system, there Is a spectrum ofIntermedIate possibilities, some involving partial micro-targeting andsome involving rules-of-the-game government intervention (with noelement ofplanning or mIcro-targeting).

Personally, I tend to agree with Professor Itiszner that a largemodern state Is above optImal size as a micro-targeting unit Myreasons—in addition to those In the sphere of knowledge cited byProfessor Iclrzner—have to dowith the discouragementofIndividualincentives’ toward efficiency, due to micro-targeting type planningas in the Soviet Union or China. But it does not follow that laissez-faire constitutes auniversal panacea.

In particular, aproofhas yet to be given that (asclaimed on p.417)“Competition between firms of different sizes and scope will tend

to reveal the optimal extent of such ‘central planning.” In‘ProfessorKlrznerdoes recognize the role of Incentives in thesearchfor knowledge.!

ash referring here toIncentives for elficlent behavior given the avaUable knowledge.

423

CATO JOURNAL

classical environments this might indeed be the case, but one isentitled to question whether, for instance, thepresentmerger trendsin the United States are logically boundto push theeconomy closerto optimality. Ifmonopoly results, inefficiency might follow.

Let me also enter a dissent to the final statement in the paper—that the advocacy of industrial policy or central planning is neces-sarily rooted In the lack of awareness of the knowledge problem(“...their well-meaning advocates are totally unaware oftheknowl-edge problem”).” In my opinion, lack of appreciation of the impor-tance of incentives may be a more serious problem. (China is anexample ofa centrally managed economy which has come torecog-nize the importanceofincentives and the merits ofdecentralization.)Some advocates see Industrial policy as a second-best solution inview of the imperfection of domestic and international markets anddespite the difficulties due to the j,roblem of knowledge. One maydisagree with thejudgment that this is Indeed a second-best solution,but without imputing to its advocates lack of awareness of some ofits disadvantages.

The problemof the approprIate role ofmarkets andof governmentintervention is complex, and, in my view, panaceas are not to befound at either end of the spectrum. Dispassionate analysis—to whichProfessor Klrzner’s paper isa valuable contribution—shows themer-its as well as the deficiencies of polar solutions; it points to thestrengths ofthe market processas well as those imperfectionswhichJusti& the search for supplementary institutional devices involvIngpublic intervention. It Is likely that the answers will not please theideologues of either persuasion.

ReferencesArrow, IC J.; Block, H. D.; and Hurwlcz, L. “On the Stability of Competitive

Equilibrium IL” Econometrica27 (January 1959): 82-109.Calsamiglia, X. “Decentralized Resource Allocation and Increasing Returns.”

Journal ofEconomic Theory 14 (April 1977): 263-83.Hurwlcz, L. “On the Dimensional Requirements ofIn%nnatlonally Decen-

tralized Pareto-Satisfactory Processes.” In Studfrs In Resource AllocationProcesses, pp. 413-24. EdIted by K. J. Arrow and L. Hurwlez. CambridgeUniversity Press, 1977,

Klxzner, Israel M. “Economic Planning and the Knowledge Problem,” CatoJournal 4 (Fall 1984): 407—18.

Mount, K., and Reiter, S. “The InformationalSize ofMessage Spaces.”Jour-isa! qfEconotnlc Theory 8 (June 1974): 161-92.

“Klrznor (LUMP p. 417).

424

COMMENT ON ICIRZNER

Osana, H. “On the InformationalSize of Message Spaces fOrResource Allo-cation Processes,” Journal of EconomIc Theory 17 (February 1978):66—78.

Radner, Roy. “Satisficing.” Journal of Mathematical EconomIcs 2 (1972):z~t

Reiter, S. “Information and Performance Inthe (Newr Welfare Economics.”AmericanEconomicRevIew 67 (February 1977): 226-34~

Scare H. “Some Examples of Global Instability ofthe Competitive Equilib-rium.” International Economic RevIew 1 (September 1960): 151—72.

Simon, Herbert A. “Theories of Bounded Rationality.” In Decision andO,ianlzatlon, chap. 8. Edited by C. B. McGulre and ft Radner. Amster-dam: North-Holland, 1972.

425