Language Planning and Economics

François Grin 1 University of Geneva and Education Research Unit, Geneva, Switzerland

Thispaper proposes a comprehensiveoverview of the‘ economicsof language’. This fieldof research,which isgrounded in thediscipline of economics,displays a strong interdisciplinaryorientation, which placesit on thefringes of mainstreameconomics. Itstudies the ways in which linguisticand economicprocesses influence one another.It isalso well placed to contributeto theevaluation of public policiesregarding language, becauseit offers analytical tools for the systematic identification and measurementof theadvantages and drawbacksof policy alternatives.I begin by discussingthe reasons why aneconomicperspective on languageis scientifically and politicallyrelevant; I thenreview some of theattendant epistemological and methodologicalissues, before moving on to anoverviewof themain lines of researchin languageeconomics. A full sectionof thispaper is devoted to theeconomic approach to languagepolicy, and anotherfocuses on anapplication to educationpolicy, detailing the economics of secondor foreignlanguage education. The paperaims at making thekey economic toolsaccessible to readersfrom varied academic backgrounds; italso emphasises the needto combine disciplinesto develop aninclusive methodology forthe selection, design, and implementation of language policies.

Keywords: Language economics,language policy,policy analysis,diversity, rates of return on language skills, costs of language policy

Introduction Becauselanguage issuesare present in every aspectof human experience, they areof interest not only tolanguage specialists,but topractitioners of all social sciences.This wide relevance oflanguage issuesas objects of study is evidenced by the attentionthey havereceived, firstand foremost, on the partof sociologists. This hascontributed to the emergence of ‘sociolinguistics’as a full-fledged field of specialisationwhich researchers may access not only throughlinguistics, but also throughsociology. To a lesserdegree, the disciplines ofanthropologyand political science havealso been takingan increasing interest in language matters. Overthe past35 years,however, economics has also been used toinvestigatea varietyof language problems.The ‘economicsof language’, or‘ language economics’, asa field ofresearch plays a marginalrole in academia,but an increasingly importantone in practice.The goalof thisessay is to takestock of the contributionsof language economicsand to assesstheir relevance, withparticu- lar emphasis on applications to language planning. The paper isorganised as follows: The secondsection, ‘ The Need foran EconomicPerspective onLanguage Issues’, discussesthe reasonsfor developing aneconomicperspective onlanguage issues.The thirdsection, ‘ Language:Too Much Alive forthe DismalScience?’ , addressesepistemological and methodologicalproblems raised by the applicationof economics to language. The fourth section,‘ The Economicsof Language:History and Main Strands of Research’ ,

Language Planning and Economics 2 Current Issues in Language Planning containsa brief historicalaccount of the development oflanguage economics, followedby areview ofthe literaturearranged in five thematicareas. The fifth section,‘ The Economicsof Language Policy’, focuseson the economicsof language policy,examining in particularthe reasonsfor state intervention, the ‘value’of language, optimaldiversity, and principles ofpolicy evaluation.The sixthsection, ‘ AnApplication:Language EducationPolicy’ , examinesat closer range the applicationof economic analysis to language educationpolicies. The concluding section is devoted to a brief critical assessment.

The Need for an Economic Perspective on Language Issues

The rise in subjective diversity While the field oflanguage economicsis epistemologically anchored in the discipline ofeconomics,it remains relegated tothe outerfringes ofthe latter.If a field ofspecialisationmay be saidto have a centre ofgravity,it couldin thiscase be locatedsomewhere in avirtualterritory bordered notonly byeconomics,but alsoby policy analysis,the sociologyof language, sociolinguistics,and the educationsciences. This relative marginalisation, however, is mitigated by the factthat the need foran economicperspective onlanguage mattersappears to be onthe rise.Various global trends in the areasof language, ethnicity andculture aregiving increasingrelevance tothe examinationof specific problemswhere the economics of language can help. Four such trends should be mentioned: · First,the geopoliticalrecomposition of Eastern Europe andthe Common- wealthof Independent States(CIS) 2 followingthe fall ofthe Berlin Wallon 9 November 1989has made room for the reassertionof awide range oflocal andregional identities which had been suppressed fordecades. To the extent thatthese identitiesare associated with certain languages, the latter aremoregenerally visible nowthan they were in the 1960s,when the first papers in language economics started to appear. · Second, the end ofthe 20thcentury andthe beginning ofthe 21stcentury arewitnessing migration flows of unprecedented magnitudeand diversity;the pairsof countries defined by amigrant’s countryof origin and countryof destinationare more varied than ever before. Populationmove- ment,far from being strictlyassociated with economic hardship or political oppression,increasingly concernshighly trainedprofessionals as well. Taken together,these trendsresult in linguistically,ethnically andcultur- ally more diverse environments, particularly in urban settings. · Third,the development ofsupra-nationalinstitutions such as the European Union (EU) give riseto new language contexts.These maybe described as additional layers in our‘ linguistic environment’(a term discussed in more detailbelow), where these layersare characterised by the diversityof the languagesat hand and the potentialrivalry or even conflictbetween the different languages spoken by different groups of users. · Fourth,the complexweb ofprocessesoften subsumed under the general term‘ globalisation’manifests itself, among others, by anincrease in the shareof internationaltrade in worldproduction, as well asby adecline in the relativecost of traveland telecommunications. Both evolutions increase Language Planning and Economics 3

the likelihood ordecreasethe costof regular contactwith a wide range of people using a wide range of languages. Itis therefore useful tomakea distinctionbetween ‘objective’and ‘ subjective’ (or‘ perceived’) diversity,and to notethat this rise in subjective diversityis fully compatiblewith a decline in objective linguistic diversityaround the world,to whichmany commentators have been alertingus (e.g. Crystal,2000; Grimes, 1998;Skutnabb-Kangas, 2000).Yet what matters here –particularlyfrom the standpointof economics – isthat the increasein subjective diversitywe are confronted with raises policy questions that need to be solved.

Conflict and threat Againstthis backdrop, language economicsis gaining recognition,because linguistic diversityis both conflictual and threatened.Since bothdimensions are often addressedseparately, it is useful topause fora momentand consider why both need to be taken into account. The conflictualnature of linguistic diversitystems from the factthat groups using different languagesoften competefor access to material and symbolic resources,and someauthors (e.g. Nelde, 1997)point out that there isno language contactwithout language conflict– itbeing understoodthat conflict is not neces- sarilyarmed or even violent,and may be confined topolitical rivalry in the frameworkof democraticinstitutions. Language need noteven be the majorfault line separatingparties in aconflict– othermarkers of identity,such as religion, often prove much moredivisive. Rather, one mightsay that allother things being equal,language contacttends towards conflictuality. Rivalry or conflictdoes not, of course,occur between languagesbut ratherbetween users of different languages, and it has more than one economic dimension. Three maintypes ofconnectionbetween the spheres oflanguage andeconomics may be identified, and each provides a context for rivalry or conflict: · First,economic factors (which maybe treatedas ‘ variables’or more complex ‘processes’– adistinctionI shallreturn toshortly) affect the fortunesof different languages,both at the microand at the macrolevel. Forexample, atthe microlevel, the relativeprices ofdifferent language-specific goods andservices (such asbooks in different languages) arelikely toaffect patternsof language use, whether directlythrough consumer decisions to purchasethem or not,or indirectly throughthe greateror lesserprovision ofpublic servicesin certainlanguages, to the extent thatsuch provision is sensitiveto costs. At the macrolevel, patternsof international trade are likely toinfluence language dynamics,both informally, in interaction between economicactors and formally, through the choicesmade by large organisations(such asmultinational corporations) to favour certain languagesfor internal communication). · Second andreciprocally, language variables(and/ orprocesses) have an impacton economicvariables (and/ orprocesses)– again,at the microas well asmacrolevel. Forexample, atthe microlevel, aperson’s language skillsmay give riseto wage rate differentials thatreward those skills. Skills arebetter rewardedin somelanguages than others, illustrating a formof competitionbetween languages.At the macrolevel, the demographicsize of 4 Current Issues in Language Planning

alanguage communityis likely tostimulate aggregate demand for language-specific goodsand services consumed by thatcommunity. The microand macro levels arealso linked toone another:some form of ascen- dancyover business competitors(a micro-level link between firms,possibly achieved throughthe dominantposition of one’ s language) willgenerally facilitateaccess to economic resources or provide better conditionsfor secur- ing economicgain, which will ultimatelybe observedat the macrolevel. Third,economic arguments may be invoked tosupportclaims made by different partiesin language conflict,as well asin the morespecific contextof debate overcompeting language policy options.Consider, for example, theissue ofsupportto minoritylanguages. Some claimthat the necessarymeasures are too expensive (thereby bringing economic considerationsinto play) while otherswill argue thatthey arewell worth it(thereby resorting,deliberately ornot,to aquintessentially economic reasoningthat hinges onthe weighing of advantagesand drawbacks, even if thoseare neither materialnor financial). Practically, authorities developing language policy plansare also confronted with the need to assesstheir costsand bring the latterin line withbudget constraints.This requires someevaluation work, in whicheconomists have to enter the fray. Letus now review these links between language andthe economywith respectto the ‘threatened’character of diversity.We haverecalled above that this stateof affairs is now amply documented. It also raises economic questions, many ofwhichare much the sameas thosethat arise from conflict, and reflect the same three types oflinkages. For example, asregardsthe ‘economy ®language’cate- goryof issues, one mightask how economic processes contribute to the demise of smalllanguages, and whether economicforces are intrinsically responsible forthis stateof affairs, or whether these forcesare merely aconduitthrough which independent language dynamicsoperate. As to the reciprocaltype ofrelationships (‘language®economy’), somecontend, for example, thatminority language pres- ervationmakes a net contributionto economicwelfare –isthis indeed the case? Finally, asregards policy matters,should scarce resources be devotedto minority language protectionand promotion, and if so,how much? The aboveis but asampleof the manyways in which language andeconomics areconnected, and which the economicsof language canhelp toinvestigate. Such investigationis taking on increasing strategic importance, not primarily because economicanalysis might help tounderstand reciprocal causal links between economicand language-related processes,but because owingto its conflictual and threatenedcharacter, linguistic diversity calls for some kind of intervention– whichtakes the formof languagepolicies .Wehavealso observed thatlanguage policies,which mobilise material and symbolicresources and redi- rectthem in certainways, have direct economic implications. Consequently, a language economicsviewpoint is more frequently required toevaluate possible courses of action in language policy.

A supporting rather than central role In ageneral introductionto the economicsof language published somesix yearsago, I suggested thatthe economicsof language shouldbe seen asafield of Language Planning and Economics 5 specialisationin the making(Grin, 1996a). I believe thisobservation to remain,by andlarge, true today.What has happened in recent yearsis a moregeneral recognitionof the usefulness ofan economicperspective onabroadrange oflanguage problems,with a generally morefavourable climate (albeit moderatelyso) towards interdisciplinary work. Since thispoint is not always apparent in the language economicsliterature, let itbe madehere withoutfurther ado:in myview,the chief usefulness ofan economicperspective onlanguage probably isnotso much thatit helps tounder- standlanguage-related processes as such.Economicsas a discipline doesnot havemuch ofaclaimto being particularlywell-suited tothistask, although some interesting insightsinto patterns of individual language behaviour ormeso-level patternsof language dynamicscan be acquiredthrough economic modelling restingon the conceptof ‘networkexternalities’ (see the sectionheaded ‘The Economicsof Language:History and Main Strands of Research’below). The chief usefulness oflanguage economicsis derived fromits capacity to formulate, evaluateand compare policy optionsregarding language, andhence toassistin decision-making. In relationto this, it is important to stressthat economic tools do not replace politicaldebate orcontributionsfrom other disciplines. The economicanalysis of language-related processes,as these areaffected, amongother factors, by language policies,does require reference tosome concepts developed in other fields, suchas the sociologyof language. Thisalone indicates that the endeavour mustbe interdisciplinary.A noless important point is that language policies ulti- matelyrequire political decisions.Even if the treatmentof language economics proposedhere emphasisesits application to policy aspects(which presupposes that political choiceshave been madeupstream), the policy discussioncannot takethe place ofthe politicalone. Hence, propositionsderived froma language economicsperspective onlanguage policies areonly intended asaninput ina wider political debate on ‘language-in-society’. Thispaper attemptsto provide abalancedoverview of thisfield ofinvestigation.Given thatother contributions, some published in recent years,contain fairly extensive reviewsof language economics,I shallavoid repeating whatthe readercan easily find elsewhere (Grin,1994, 1996b; Grin & Vaillancourt,1997; Vaillancourt,1985a). Moreover, striking developments in language economics overthe pastfive yearshave been relativelyfew, withnew resultsmainly in two areas:the theoreticaland empirical assessment of language policies,and the estimationof rates of return onforeign language skillswith novel, sometimes tailor-madedata sets. For these variousreasons, the restof thispaper, ratherthan focusing on a literature review, will emphasise analysis and commentary.

Language: Too Much Alive for the Dismal Science?

The meaning of scarcity Economicsis sometimes called the ‘dismalscience’ . Asoften isthe casewith wittyyet unflattering labels,it does hold some truth, but somecommentary is needed inorderto assessits actual import. Such commentarycan alsoprovide a suitableentry pointinto a discussionon the natureof economicsas adiscipline and on its applicability to language issues. 6 Current Issues in Language Planning

Whatmakes the exercise particularlyapposite is the factthat one sometimes encounters,among scholars from the language disciplines orother social sciences,what amounts to a formof a prioriaversion towards economics. This hostilityis perhaps notentirely undeserved. However,it is not always grounded in arobustgrasp of economicanalysis, and I submitthat, if appropriatelyunder- stoodand put in itsproper place,economics can offer analyticaland method- ologicaltools relevant to the studyof language, in particularlanguage policy, and offer insights that other approaches do not provide. The paternityof the label ‘dismalscience’ is generally attributedto the British historianThomas Carlyle (1795– 1881), who witnessed the wretchedliving condi- tionsof much ofthe Britishworking class and observed that this state of affairs wassadly compatible with economists’ analyses of economicprocesses. Some, however,trace the expressionback to the predictionby Malthus,at the end ofthe 18thcentury, thatthe productionof goods(in particularfood) would always be outpacedby populationgrowth, dooming large tracts of mankindto perpetually hoveron the edge ofstarvation.Malthus’ s predictionshave been disproved,as shownby the growthof per capitaproduct in the long run,which has allowed an average level ofprosperityfar beyond the expectationsof most‘ classical’economists.3 Nonetheless,the derogatorylabel hasremained, unchallenged even by mosteconomists, many of whomapparently assumethat being dismalis some- how a proof of seriousness and scientific accomplishment. 4 Whatcertainly is true isthat economics constantly reminds us of limitations – moreprecisely, ofthe factthat resources are limited, and that this imposes constraintson humanaction. This is reflected in the moderndefinition ofthe discipline. Definitions havechanged overtime. In the 4thcentury BC, Aristotle haddefined economicsas the science ofthe acquisitionof wealth.In the introductionto his weighty tomeon the historyof economicthought, Blaug (1985) recordsthree types of early definitions,namely: the studyof the natureand causesof the wealthof nations;the studyof the lawsgoverning the distribution ofthe productsof agricultureand mining; andthe studyof the lawsof capitalism. More recent (thatis, post-1870s) definitions, listed by Becker (1976:3) typically include the studyof the allocationof material goods to satisfymaterial wants; the marketsector; and the allocationof scarce means to satisfy competing ends.This latterdefinition iscertainly the mostprofound –andprobably the mostwidely accepted.One ofits chief virtuesis that it avoids an obvious problem withmore substantivedefinitions ofeconomics,which stemsfrom the factthat demarcating anyparticular phenomenon as‘economic’is purely arbitrary.For example, as manyeconomics textbooks point out, inflation, unemployment and consumption,which are regarded as quintessentially economic questions, also are political,sociological or psychological ones. Conversely, typically ‘ political’ processes,such as votingbehaviour or the adoptionof apublic policy,present indisputably economicdimensions. Characterising this discipline withrespect tootherstherefore callsfor a shiftof emphasisfrom subject matterto approach, andthis is precisely whatthe lastof the abovedefinitions achieves:economics is fundamentally a way oflookingat humanexperience asatensionbetween ends and means. Thisdefinition wasfirst formulated expressly by Robbins,who described economicsas ‘ ascience whichstudies human behaviour as a relationship Language Planning and Economics 7 between ends andscarce means that have alternative uses’ (Robbins, 1935: 16–17). In otherwords, economics as a discipline existsonly because ofscarcity:in aworldof boundless plenty, resourceswould not be scarce,and the problem of their allocationwould simply notarise. If, therefore, the discipline ofeconomics is,in essence, nothingbut aconstantreminder ofthe prevalence ofsome‘ principle ofreality’ (perhaps oversome less stern ‘ principle ofpleasure’ ), itis not surprising thatmany should see itas dismal.This, however, may not be the chief reasonfor its unpopularity withsome practitioners of otherdisciplines in the socialsciences, whose determinism (as with much of structuralisttheory in mainstreamsociology) can be seen asno less pessimistic, and just as reductionist. Aswe shallsee shortly,the causesof someof the hostilitymay have their roots elsewhere. Fornow, we cannonetheless take‘ scarcity’as the startingpoint and ultimate justification for all economic theorising.

Scarcity and choice One questionto address is whether scarcityis a relevantissue tobegin with. There canbe littledoubt that it is, and this point should not detain us too long: for mostpeople, needs andaspirations exceed resources,which logically implies thatsome form of scarcityexists. 5 Recallfirst that nothing, in economictheory, mandatesa restrictionto material or financial resources,and that intangibles like socialnetworks and interpersonal affection are, from the perspective of economic analysis, perfectly relevant resources. Thatresources are at least perceived aslimitedcan be illustratedby the following experiment: mosthuman beings, if askedto choose between having‘ more’or ‘less’resources than they currently enjoy, will express apreference for‘ more’ over‘ less’, whether the resourceat stake is money, love, health or friendship. Furthermore,even apersonblessed onall these countswill be confrontedwith one absolutelimitation, namely, time. Wealth may give accessto better health andhence increaselife expectancy,but thisis true only withinlimits, and (beyond acertainage) witha tendency todecreasing physical and mental abili- ties.Moreover, as Linder onceobserved, even immortalitywould not solve the problem of scarcity: in aneconomicheaven, the problem oftime will be particularlypressing. We will find there aninfinite volumeof consumptiongoods, which plea- sure-hungry angelswill feverishly trytoexploit during the limited timeat their disposal per day.Thatone mayin thisheaven enjoy eternallife asa consumerfails to alterthe situation.This can increase the totalsatisfaction derived overthe courseof centuries.What we areinterestedin, however,is the yield pertime unit .Tomaximizethis, time must be carefully stewarded by the students of Epicureanism. (Linder, 1970: 13; emphasis mine) Inthemoremundane contextin whichmosthumanslive,scarcity,whether seriousand life-threat ening ormerely annoying,is an everyday reality.If scarcityis sucha pervasiveaspect of humanexperience, itslogical, unavoidable consequenceisthe need tomakechoices in orderto use scarceresources wisely.Choices may be influenced, orin large partconstrai ned by social norms(possibly mediatedthrough an individual ly developed habitusin Bourdieu’s sense), butsomespaceforchoice willnonethelessremain,and itis 8 Current Issues in Language Planning thisspace that economists are interested in.Hence, economicsdoes not merely (andnegatively) confine usto a depressing worldof scarcityand frustration – it alsostresses (positively) the roleof choice,in whichactors have a notinsignifi- cantdegree ofpersonal freedom. Assuch, it contains a full-blown theoryof agency,6 andthe emphasison a certain‘ approach’(as distinct from a certain subject matter)as the truly distinguishing feature ofeconomics as afield ofscien- tificinquiry doesnot strip it of allsubstantive content. It is worth, at thispoint, repeating Becker’s famousobservation that when there areno alternatives, there isno choice, and where there isno choice, there isno economic problem; put differently, thissimply meansthat more than most social scientists, economists arewedded tothe notionof the autonomyof actors.The strengthsand weak- nessesof economictheory as atheoryof humanbehaviour (as compared to other theoriesof agency developed in otherdisciplines, in particularsociology) raise questionswe shallreturn tolater; the importantpoint here isthat the subject matterof economics is how humans make choices,7 andthat the need tomake choices is a result of scarcity.

The relevance of rationality The economicapproach to human behaviour is based on one very simple, possiblysimplistic hypothesis, namely, that actors use their limitedresources rationallyby makingchoices regarding the allocationof these resourcesin sucha wayas tomaximisetheir satisfaction(or, in economicparlance, their ‘utility’). Thisis assumed to apply to any actor,and given economists’concern with generality,they areless interested in particularsocial actors operating in agiven contextthan in the generic actoror agent,often calledthe ‘homoœ conomicus’. Thisapproach has strengths as well aslimitations.Its strongest suit is that owing toits very generality, itcan make a defensible claimat grounding ageneral, deductive theory ofhumanbehaviour, rather than case-specific, ad hoc interpreta- tionsof the idiosyncraticaction of actorsin aparticularsituation. Another strong pointis that the twinassumptions of ‘rationality’and ‘ utility maximisation’are sogeneral thatthey aredifficult tofaultsubstantively. This point has often been madeby scholarsin othersocial sciences, such as the anthropologistCancian, who observed almost forty years ago that: [utility] maximizationis one ofthe standardrestatements of the a priori truththat all human behavior is patterned; that all human behavior has a reason.The use ofmaximization as a scientificstrategy involves seeking outthe motives(or whatever the investigatorsees as the impetus ofbehav- ior)and attempting to rank-order them so as to see the behavioras the (consciousor unconscious) maximization of these things.They becomethe ends being maximized. ..Itis in thissense thatall people alwaysmaximize or economize. There can be no argument about it. (Cancian, 1966: 47) Itis easyto criticisethis approach as ‘reductionist’, but itisfar from clear that itis–andsuch criticism is sometimes uttered by commentatorswho use theoriesof actors’behaviour thatare no less reductionist, and possibly even more so,without their being awareof it. 8 Aproperly applied economicanalysis, precisely because itstartsout from anassumptionabout human behaviour that issogeneral astobe almosttautological, forces the analystto be precise about Language Planning and Economics 9 the logicallinks thatenable her tolink the general theoreticalplane ofthe ratio- nalityassumption with observed patterns of behaviour.In thisexercise, successive reductions froma complex,multi-faceted realityto asimpler (but, one hopes, heuristicallyvaluable) abstractionneed tobe identified andexpressly formulated. The rationalityhypothesis, in thatit relates ends andscarce resources, is actu- allyextremely malleable andopen toimportsfrom other disciplines. For example, the precise natureof the ‘ends being maximised’(to use Cancian’s phrase) will varyaccording to timeand place, history, custom and culture. Economists canuse informationprovided by anthropologists,sociologists, political scien- tistsand psychologists regarding notonly howthese ends arechosen,but what they areand, consequently, what isbeing maximised.In the sameway, findings fromother social sciences provide informationabout the constraints under which ‘utility’is maximised – these constraints,after all, are merely amanifestationof the scarcityof resources, or ofrestrictionson the waysin which resourcesmay be used.Hence, ashasalready been noted,both the ends andthe meansconnected by autility-maximisingbehaviour can be ofany nature – tangible orintangible, financialor symbolic. Rationality is procedural rather than substantive. This makesit possible foreconomists to study production, consumption, markets, prices,wages or interest rates, but alsoto apply their analyticaltools to health, education, the environment – or language. The rationalcalculus of utilitymaximisation is, of course,not synonymous withegoism, and Schenk notesthat ‘ [i]t iswell-known in economicsthat the assumptionof self-interest canlead to cooperativeand altruistic behavior [The lesswell-known conversecan be shown]that altruism can lead to behavior whichappears based on self-interest’(Schenk, 1987:192). Economics is mainly basedon philosophical utilitarianism, but authorslike Nobel Prize laureate AmartyaSen usefully remind us(1985) that this intellectual tradition(whose classicalformulation is attributed to Jeremy Bentham)comes in different and subtle variants,and that it incorporates a distinctiveethical perspective. Notonly isit, therefore, afarcry from what some deride ascrassmaterialism, but there are noconvincing grounds to construe ‘ utilitarianism’as aninsult,although some do– sometimes,one suspects,with a limitedgrasp of whatthe discipline of economics actually is about. Nonetheless,the economicapproach to human behaviour does have limitations.These limitations,however, are notconsubstantial with the discipline, and shouldrather be associatedwith two featuresof a very different logicalstanding. The firstlimitation is associated with the methodologythat economists generally favour;the secondhas to dowiththe praxisof someeconomistsas socialactors in the academic context. Let us examine these limitations in turn.

Formal modelling and its limitations Formalmodelling isused in alldisciplines in the socialsciences; however, it hasa distinguished traditionin economics,which prides itself onitsextensive use ofit,particularly in theoreticalas distinctfrom empirical research. Models areusually defined asabstract,simplified representationsof the problem tobe studied.What economists regard as amodel,however, usually requires amore elaborateand formal construction than in mostother disciplines in the social 10 Current Issues in Language Planning sciences.For example, atypologyof language functions,because itrests on the analyticaldevelopment ofcategories of language behaviour,may be described by sociolinguistsas amodel(e.g. Appel &Muysken, 1987),but itwould not, in economics,be calledone. An economicmodel is always made up oftwocatego- riesof ingredients, namely,variables and relationships between them;and at leastsome of these relationshipsshould be causalones. 9 These relationshipsneed notbe formulatedin greatdetail; however, what is usually expected ofthemis someinformationas towhether a(positiveor negative) change in the value ofan independent variablewill induce a(positiveor negative) change in the value ofa dependent variable.Suppose forexample thatY standsfor the labourincome of a generic actor(the ‘homoœ conomicus’) while Qstandsfor his demand of a partic- ulargood or classof goods (say, clothing). One wouldnormally expect aperson’s demandfor the goodto be anincreasingfunction ofhisincome, allother things being equal.Hence, one ofthe mostelementary modelsof individual demandas a function of labour income would be of the form:

Q = f(Y) where the gradientof the relationshipbetween Q and Y isassumed to be positive; put differently, the firstderivative of Q withrespect to Y islarger than zero. This general analytical form can be given a specification such as:

Q = a + bY where a ³ 0 and 0 < b < 1.Thisspecification would imply thateven when incomeis negligible orzero, the actor’s demandfor the goodwill be positive,and that increasesin hisincome will give riseto less-than-proportionalincreases in his demandfor the good(for example, if b = 0.8,a10%increase in incomewill result inan8% increasein demand).Hence, itis notsufficient, in economicmodelling, tostatethat a certainvariable ‘ will depend’on another– forexample, tostatethat aperson’s consumptionchoices ‘ will depend’on socialclass, or that a bilingual’s choiceto use one language oranother ‘ will depend’on respective positionsof powerbetween himself andan interlocutor. Quite obviously itwill, and the real questionsare ‘ how?’and ‘ howmuch?’ . Barringduly argued exceptions,the direction(if notthe magnitude) of the implied cause-and-effect relationships must be specified. Nomatterthe discipline in whichit is applied, thisinsistence on the specifica- tionof relationshipsoffers the notableadvantage of forcing the analystto spell outher assumptions,which gives riseto a degree oftransparencyand logical rigourthat other analytical devices donotalways guarantee (on thisquestion andits application to the studyof language problems,see Pool,1991a). Modelling ‘makesexplicit assumptionsthat would otherwise remain hidden andhence unexamined’(Mayer, 1993:123). Yet this type oftransparencyand rigourdoes not come without a price, fortwo reasons. First, the approachis epistemologicallyperilous. Owing totheir formalelegance, algebraicmodels can be seductive,and it is all too easy to assume that beauty ofform will necessarily be matchedby relevance ofcontent.This is obviously a logicalerror of major proportions,but one towhich manyusers of formal modelling fall prey. If only because economistsmore systematically use thistype oftool,this error occurs Language Planning and Economics 11 morefrequently amongeconomists than among practitioners of other social sciences.Therefore, itbears repeating thata modelis nothing but aformof discourseabout reality. This restriction should be takenone step further by stressingthat the rationaleof thistype ofdiscourse is not to resemble,let aloneto replicate reality,but toconstitute a useful instrumenthelping usto reflect upon andunderstand reality. When seen in thislight, analytical, possibly formalised (algebraic) modelscan be uniquely powerful andtransparent tools in the study ofselected relationshipsbetween selected setsof variables.Models serve to clar- ify the logicalimplications of arelationshipbetween pairsof variables,and force analyststo apply demanding standardsof logicalconsistency. However, these benefits will accrueonly if the limitationsof formalmodelling areclearly acknowledged. Second, because formalmodelling, particularlyin the formfavoured by economists,insists on ‘ signing the effects’(that is, giving the sign,positiveor negative, ofthe derivativeof a dependent withrespect to an independent variable– or combinationof primary variables in amodel),both categories of variables must lend themselvesto a quantitativeinterpretation. In otherwords, the method demandsthat it always be possible toexpresseach variable in termsof ‘more’or ‘less’, thatis, to project iton a quantitativescale. This is not a problem when studying relationshipsbetween variablesthat are primarily defined by their quantitativeaspect, such as the unit price ofagoodor the quantityof thatgood exchanged ona marketduring acertainperiod. Hence, economicanalysis is well-suited tothe studyof markets whose equilibrium isdefined in termsof quantityand price. Itis,however, much lesssatisfactory in the caseof variables thatmay have a quantitativedimension (ultimately, all variablesdo), but whose quantitativedimension is neither particularlyinteresting in itself,nor particu- larlypertinent in agiven analyticalendeavour. For example, when studying the link between aperson’s socialclass and their education,representing the latteras acertainnumber ofyearsof schooling may be useful, but will only allowinvesti- gationof asmallpart (not necessarily the mostrelevant one) ofthe links between socio-economicbackground and education, because itwill fail toaccount for many of the qualitative dimensionsof that person’s education– ascomparedto the qualitativedimensions of the educationreceived by another person from anothersocio-economic background. The applicationof multivariateanalysis to quality data sets may mitigate this problem – but then only to a certain extent. Still, aneconomist’s pathof leastresistance is towards variables that have a moreor lessobvious quantitative interpretation and can be connectedwith one anotherusing mathematicallanguage. Thisunavoidably induces abiasin favour ofthe studyof relationshipsbetween variablesthat can more easily be formulatedin thisway. It follows that variables that do notlend themselvesto quantita- tiveexpression will often be simply ignored,or flattened outto the extent that mostof the phenomenologicaldepth issqueezed outof them. Formal modelling hasa tendency toproduce interesting analysesof the relationships between variables,but often terribly dry (andcertainly reductionist) characterisations of the variables themselves.That these problems arenot always solved to satisfactionis reflected in the ambiguity ofeconomists’relationship to methodology.On the one hand,there isa constantinsistence on the analyticalrigour with which the algebraicformulation of relationships between variablesis dissected. On the 12 Current Issues in Language Planning otherhand, one often encountersa certainreluctance to seriously address more fundamentalquestions pertaining tothe soundnessof the assumptionsmade aboutthese relationships,and even moreso about the variableswhich these rela- tionshipsconnect. Taken toextremes,this can result in disappointinganalyses, whichare far removed from perceived humanexperience. The economistDaly oncederisively observedthat ‘ Layerupon layerof abstruse mathematical modelswere erected higher andhigher abovethe shallowconcrete foundation of fact.The behaviorof a peasantselling acowwas analyzedin termsof the calculus of variations and Lagrangian multipliers . . . ’ (Daly, 1977: 3).

Is economics exclusionary and imperialistic? The cruxof the matter,therefore, isto use economicswhere itis useful, recog- nising itscontribution to the analysisof someproblems, but notto stretch it beyond whatit can actually say. Unfortunately, economists do notalways exer- cisesimilar caution, and before closingthis section, it is useful todiscuss the ‘dismal’epithet in relationwith the practicesof economists as members ofan academic profession. Economists’work often lends itself tocriticismfor the reasonsjust discussed; yet whatis at stakemay be notso much the contentsof economicresearch as the philosophy withwhich it issometimes carried out. As we haveseen, modern mainstreameconomics makes abundant use of formalmodelling. However, whereassimple modelscan be illuminating andreveal aspectsthat other forms ofanalysiswould have left in the dark, 10 manymathematically refined models (quite apartfrom the questionof their topicalrelevance –see the quoteby HermanDaly justabove) aresimply inaccessibleto the layman– and,increasingly, toother economists who are not specialists of the specific classof models concerned.The inner dynamicsof the profession,which primarily rewards dazzlingmathematical prowess (usually aconditionfor publishing in the most prestigiousacademic journals) encourage economiststo prioritise the displayof mathematicalapparatus, to the pointthat some economic research is closer to applied mathematicsthan to an investigationof the actualprocesses of allocation of scarceresources. In sodoing, economists themselves disregard the lawof diminishing returns,which, of course,also applies tomathematicalsophistica- tionas atoolof scientific inquiry. The actualpractice of many economists some- timesends up sterilisingnot just the answers,but the questionsthemselves. Furthermore,it is difficult toshake off the accusationof exclusiveness, or even arrogancethat can be levied againstsome members ofaprofessionwhose choice oflanguage (mathematics)shuts other scholars out: although the reluctancethat economicssometimes inspires insomequarters may, in part,be due tosimple frustrationat notbeing able tofollowits methodology, it probably alsoreflects an understandableimpatience in the face ofa sometimesdeliberately esoteric discourse. The issue of‘economicimperialism’ is quite different, andcan arise in two distinctguises. Onthe one hand,one mightbrand as ‘imperialistic’the behaviourof researchers(say, economists) who work on issuesthat were tradition- allytreated by scholarsfrom other disciplines (sociologists,anthropologists, historians,etc.). Such criticism,however, is hard to countenance,unless one is toadopta deeply corporatistview ofresearch,in which certaindisciplines hold Language Planning and Economics 13 amonopolyover certain types ofissues. A sounderperspective mustbe torecog- nise the essentialcomplementarity of different disciplines.As aconsequence, no discipline isintrinsically more or less legitimate than any other to produce scien- tificdiscourse about any particular set of issues;simply, some disciplines will havemore to saythanothers about them. Gary Becker, whosemany publications haveoften been citedby criticsas instances of crass imperialism, because, in particular,of his work on the family,accepts the label of‘economicimperialist’ if itrefers tohis application of aneconomic approach to domains traditionally studied in other disciplines (see Swedberg, 1990, Chap. 1). Bycontrast,the stanceadopted by someeconomists with respect to thisissue isquestionable. One suchexample isa recent essayby Lazear(2000) in one ofthe mostprestigious journals of the profession,which starts out by proclaimingthat ‘economicsis not only asocialscience, it isa genuine science’(read: ‘ othersocial sciencesare not’ ), andgoes on to assert that ‘ economicsis the premier social science’. In supportof these claims,Lazear describes findings fromeconomic modelling ondemography,discrimination, social interaction, religion, law,etc., suggesting thatthese findings aremore important and relevant than those which otherdisciplines haveyielded. The argument,however, remains unconvincing. Hisbrief account,among the variousareas addressed, of economiccontributions tolanguage (which, surprisingly, ignoresmost of the workproduced in language economicsover the past35 years)highlights only acouple ofresults fromthis literature, some truly important(e.g. Church &King, 1993),some not, 11 andit should be acausefor concern that a paper devotedto establishing the supe- riorityof economicsover other social sciences purports to makethis case almost withouta single reference (outof 110)to publicationsfrom disciplines otherthan economicsor business.Readers may then legitimately wonderon whatbasis the critic judges the vast tracts of scholarly work he presumes to criticise. Amorereasonable path certainly lies faraway from such extremist views. It requires neither awholesalecondemnation of all the workof economists as empty formalismand reductionist nonsense, or the dismissalof allsociological workas hopelessly illogicalramblings and flights of metaphor.It stands to reasonthat any social process or any aspect of human behaviour can be approachedfrom the complementaryperspective ofthe varioussocial sciences. Noprocess can be fully understoodby using one perspective alone,and, depending onthe caseat hand,some disciplines will haverelatively more to saythan others.Most commentators will probably agree thatthe relativecontribution of economics(by comparisonwith that of otherdisciplines) ishigh in the studyof marketequilibrium, andlow in the studyof intra-family exchange. Yetthis does notmean that sociologists have nothing to sayabout market equilibrium, orthat economistscannot offer insightsinto intra-family exchange. Quite simply,the relativecontribution ofeachapproach will varyfrom case to case. This general conclusioncarries over to the studyof language issues:economists will have manyuseful thingsto say regarding somelanguage issues,and little regarding others. Fortunately,recent evolution,particularly in the secondhalf ofthe 1990s, suggeststhat entrenched battlesbetween academicdisciplines areprogressively makingway for a morebalanced recognition of complementarities.In hisinter- view withSwedberg (1990:33ff.), Becker notesa greaterreadiness to exchange 14 Current Issues in Language Planning between economistsand sociologists in the United States.Mayer (1993)stresses, withineconomics itself, the legitimacyof two different strandsof research, namely formalisticand empirical research. The (stillmodest) influence of socio-economicswhich, less than a field ofinvestigation,describes a philosophy ofresearchin the socialsciences, may also be seen asapromisingsign ofmore integrative work in years to come. Forthe mostpart, economic research on language andlanguage planning isa deeply interdisciplinaryendeavour. Having described the epistemologicalback- ground inwhichit originates, it is now time to moveon toanoverviewof the literature in the economics of language.

The Economics of Language: History and Main Strands of Research

Historical overview 12 The historyof the economicsof language asa field ofresearchon the fringes of economicsas adiscipline datesback to the mid1960s. For a long timeafter, economists’contributions on language issuesremained unrelated toeach other, and it isonly in recent yearsthat a greaterdegree ofinterconnection has appeared, reflected in morefrequent cross-referencing. Furthermore,early studies in the field directlyresponded tothe socialand political issues their authorswere confrontedwith – namely,the relativesocio-economic position of Latinosin the United States,or English–French earningsdifferentials in Canada,particularly Québec. The beginnings ofthe economicsof language canbe represented in termsof three generationsof studies. Since these arereflected in currentstrands of researchto be reviewed inthe followingsub-section, references tothe relevant literature are, for the most part, deferred until then. The firstgeneration of studiesis mostly made up ofempirical(that is, statistical)work and largely coincideswith the 1960s,although its roots can be traced backto Becker’ s (1957)economic analysis of (racial) discrimination, and its formalanalytical application to language proposedby Raynauldand Marion (1972).It views language primarilyas an ethnic attribute :havinga particular language asone’s mothertongue ascribes apersonto a particulargroup, andthis language-based ascriptionmay have an effect onthatperson’ s socio-economic status,particularly his or her earnings.This approach, which clearlyharks back tothe ‘identity’function oflanguage regularly mentioned in the sociolinguistic literature,has been used toanalyse earnings differences between blackand white United Statesresidents, or between anglophonesand francophones in Canada.In termsof economicanalysis, it is closely related to the modernstudy of discrimination,which stresses empirical and methodological aspects and emphasisesissues of gender andrace(see e.g. Oaxaca,1973), more than analytical aspects in relation to language. Asecondgeneration of studies, which largely coincideswith the 1970sand early1980s, emphasises the humancapital natureof language, aboutwhich more will be saidin the followingsection. This opens the wayto adifferent perspective onlanguage, favouring,in particular,linkages with educationeconomics , which hadalso emerged in the 1960s.The analysis,which is examined atcloserrange Language Planning and Economics 15 below under the sectionheading ‘An Application:Language EducationPolicy’ , goesas follows:particular language skillscan be interpreted, in the sameway as othertypes ofskills,as aformof capitalbecause itis anarea in whichindividuals andsocieties can profitably invest. Deliberately acquiredlanguage skillscan therefore be seen asa source ofeconomicadvantage. This approach harks back to the generic communicationfunction oflanguage encountered in sociolinguistic work,and has often been applied in empiricalwork on the socio-economicstatus ofimmigrants in the United States(particularly native speakers of Spanish) to assess the value to them of acquiring English. Athirdgeneration of studies, opened up by Vaillancourt(1980), considers bothlanguage functionsjointly asdeterminants of labourincome. Languages are seen notonly aselements ofidentity oraspotentiallyvaluable communication skills,but asasetof linguistic attributes (embodied in individuals)which together influence actors’ socio-economic status. Although the historicalsources in language economicscan, for the most part,be arrangedin the categoriesjust described, someisolated contributions since the early 1960sdo notfit intothis pattern. This is, for example, the caseof anoften-mentioned paper byMarschak(1965), simply entitled ‘Economicsof language’, which discussesthe internalevolution of linguistic codesas a permanent drive towardsmore effective communication(where effectiveness isunderstood as requiring the lowestpossible timeinput tocommunicate a certainmessage). Therefore, apartfrom the chronologicaldistinctionjust presented, apossiblymore fundamental categorisat ioncan be proposed,not only inorderto accommodatethe progressiveexpansion of language econom- icsto abroaderrange ofquestionsgoing beyond the studyof earnings,but also toproposea general definition oflanguage economicsas afield ofresearchin itsown right.

A definition Letus firstobserve thatfor the mostpart, the studiesmentioned sofarare duetoNorthAmerican scholars. They stressthe roleoflanguage asan explanatory factorofeconomicvariables –usually asadeterminantoflabourincome. Inthelate1980s,economistsfromother partsoftheworldbecame moreactive intheinvestigationof the language–economy relations hip, often focusing on the reverse causation,namely,the roleof economicvariables as explanatory factorsof linguistic variables.Examplesinclude the effect ofprices orearn- ings onindividualpatternsof language use oronlanguage dynamics– with anoccasionalemphasison thecaseofregionalor minoritylanguages inWest- ern Europe. Thiscontrast suggests that a general definition oflanguage economicsmay be articulatedin termsof the causaldirections investigated. Furthermore, particu- larlysince the 1980sand more so since the 1990s,a risingnumber ofcontribu- tions,both in Europe andNorth America, started looking at the roleof economics asatoolfor evaluating the effects oflanguage policies,particularly in termsof the costsand benefits ofdifferent policy options. 13 In thiscase, standard economic variablessuch as prices,earnings, transaction costs, etc. do not necessarily intervene, andthe ‘economic’nature of the approachis evidenced by the factthat the advantagesand drawbacks of language scenariosare weighed againsteach 16 Current Issues in Language Planning other.This line ofresearchis a directillustration of the economicapproach as defined by Robbins.Using thisbreakdown, a definition firstproposed a few yearsago(Grin, 1994a) and used againin laterwork seems to remainappropriate today: The economicsof language refers tothe paradigmof mainstreamtheoreti- caleconomicsand uses the conceptsand tools of economics in the studyof relationshipsfeaturing linguisticvariables. It focuses principally, but not exclusively, onthoserelationships in whicheconomic variables also play a part. The ‘paradigmof mainstream theoretical economics’ to which this definition refers isencapsulated in theRobbinsdefinition presented abovein the sectionheaded ‘The Need foran Economic Perspective onLanguage Issues’, andstresses the utility-maximisingchoices made by actorswhose resources are scarce. The secondhalf ofthe definition, whichrefers torelationships that do not necessarily feature economicvariables, may deserve additionalcommentary. For mostpeople, whether economistsor not,economic variables typically include, in micro-economics,variables like prices,quantities supplied ordemanded,wages, earnings,the return oncapital,etc., and the wayin which these variablestend towardsan equilibrium value onmarketsfor given goodsand services. In the realmof macro-economics, typical economic variables include aggregateslike consumption,investment, exports and imports, their combinationinto an aggre- gateproduct (for example GDP), the rateat which the latterchanges (growth), an aggregateprice level andits rate of change (inflation), unemployment, etc. However,such variables need notbe present in aneconomic analysis of language policy.Consider, for example, the questionof whether apolicy tointro- duce agiven minoritylanguage asa schoolsubject shouldprioritise the formal trainingof language teachersor the development ofteachingmaterials – bothof whichcan be consideredexplanatory factors in the studyof the resulting language skillsof learners.None ofthe variableslinked by thiscausal relationship arestandard economic variables, yet the treatmentof the questionmay be essentiallyeconomic, in thatit puts forward a certainnotion of effectiveness, bothas anoperationalconcept and as a relevant evaluationcriterion. Incidentally, evaluationsof thiskind makeup one ofthe corework areas of educationecono- mists– anareato whichwe shallreturn in ‘AnApplication:Language Education Policy’. Finally, readerswill observethat the definition makesuse ofthe conceptof ‘variable’. There isno question that variables are analytical constructs with a certaindegree ofarbitrariness,and that they shouldbe understoodas such,along withtheir limitations,as pointed outat the end ofthe sectionabove on ‘The Need foran Economic Perspective onLanguage Issues’. Nonetheless,they remain heuristicallyrelevant concepts, which forthe sakeof definitional crispnessmay be preferable tothe notionof ‘processes’. The reasonis that the characterisation ofaprocessnecessarily requires anoless arbitrary definition anduse (whether explicitly ornot)of variables. Hence, astrictreference to‘ processes’could turn outto be merely euphemistic; itseems therefore preferable torefer to‘variables’, itbeing clearthat they haveno existence of their ownand are, quite simply, useful. Language Planning and Economics 17

Main research directions Letus now turn to a review ofthe maindirections of researchin language economics.For reasons explained in the introduction,this overview is kept delib- eratelyshort and the listof authors mentioned, though fairly extensive, isnot intended asexhaustive.The interested readeris directed to otherwork (in partic- ularBreton, 1999a, 1999b; Grin, 1996b; Grin & Vaillancourt,1997; Vaillancourt, 1985a)for additional references. Twomajor topics, namely, the economicsof language policy andthe economicsof language skills,which areintroducedonly briefly in thissection, are examined atcloserrange below under ‘The Economics ofLanguage Policy’, and‘ An Application:Language EducationPolicy’ . In this literaturereview, allformal notation is avoided. While thiswill prevent usfrom presenting the detailof the analyticalmodels, it will suffice tooffer anaccountof the main ideas on which they rest. Language and labour income The studyof the effect oflinguisticattributes (that is, a person’s L1 and/or their skillsin otherlanguages, which will be referred toasaperson’s L2or setof L2s) onearnings (or, equivalently, ‘labourincome’ ) remains,throughout the history oflanguage economics,the single mostimportant area of research.It is also one which,as we shallsee in ‘AnApplication:Language EducationPolicy’ , hasdirect relevance forlanguage educationchoices. Its basic idea is that linguistic attributes can influence earnings in two different ways. First, membership ofa language group maygive somean edge in the labour marketand put othersat adisadvantage.This is the situationwhen aperson’s L1 canbe seen asa determinant oflabourincome, and where the disadvantage(or privilege) ofbirth isreflected, in particular,in lower(or higher) wagerates. 14 Of course,if educationallevels orothersocially legitimate determinants of income arecorrelated with linguistic attributes, members ofaparticularlanguage group will logicallytend toearn less thanmembers ofanotherlanguage group, independently of their language attributes.The questiontherefore iswhether language itself, other things being equal , does result in earnings differentials. In orderto treatthis question theoretically, economists have tried out a wide range ofmodels.Let us first consider the caseof L1as a determinantof labour income.Such language-based differentials mayresult from a deliberate intent, by members of another(presumably dominant)language group, toexert discrimination,possibly by manipulatingthe rateat which the goodsprimarily produced by one ofthe twogroups are bought andsold (Raynauld & Marion, 1972).If the relativeprice oflabour-intensive goodsgoes down, and if the domi- natedgroup hasrelatively more labour than capital to contributeto the economy, the shareof aggregateincome accruing to this group will godown. This will also tend toreduce the incomeaccruing to the ownersof capital in the dominant group (because the combinationof capitaland labour in productionwill diverge fromwhat it would have been in anon-discriminatoryoptimum), but thisis a price thatthey maybe willing topay,just as Becker (1957)assumed that some people havea ‘tastefor discrimination’ which will be satisfiedat the costof some forgone income. Avariantof this explanation, relying onanotherand presumably lessdeliber- atediscrimination strategy, has todowiththe existence ofdifferent networksof 18 Current Issues in Language Planning accessto employment (Migué, 1970).Suppose thatthe populationis made up of twolanguage groups, A and B.The employer belonging tothe A group may prefer tohire workforcefrom the samegroup, because culturalproximity will makeit easier to assess ex ante the employee’s productivity(this strategy is some- timescalled ‘ statisticaldiscrimination’ ). In arelatedmodel, Lang (1986) explains earningsdifferentials notas the resultof anyprejudice ordiscriminatoryintent, but simply asthe consequence of communicationcosts between A-speaking employers and B-speaking employees; thiscommunication cost will largely be borne by the latter,whereas A-speaking employees, withwhom employers are able to interact at no extra cost, are exempted from it. Second (orforeign) language skillscan be aprofitable investmentfor a number ofreasons,in particularif employers rewardthem. They will dosoif a bilingual employee, allother things being equal, ismore productive than a unilingual one –typicallybecause he orshe cancarry out duties that a unilingual couldnot. Beyond thisgeneral interpretationderived frombasic labour economics,the investigationof the micro-level processesthat explain the value ofsecond language skillsremains limited to exploratory models (e.g. Sabourin, 1985) studying,for example, the roleof matchingbetween the linguistic characteristics ofa job andthe language attributesof aworker.The importanceof language skillsin amatchingprocedure thatstarts at hiring isconfirmed by (non- representative) survey resultsin Canada(Chorney, 1998),but questioned by surveysof Australian business firms(ALLC, 1994).For the mostpart, however, the study of the role of L2s as determinants of income remains empirical. Independently ofthe reasons behind language-based earningsdifferentials, their existence hasbeen shownempirically in avarietyof contexts. However, it is difficult in practiceto separate empirical work focusing onthe effect ofL1(in the sense ofethnic membership), oronthe effect ofL2 skills (as a formof humancapital)or on the effect ofbothlanguage attributes.As noted at the beginning ofthis section,almost all of the empiricalwork published after1980 takes account both ofL1andL2 asdeterminantsof labourincome, and most of the workmentioned inthissection does. Itis therefore convenient toarrangeempirical research on the relationshipbetween language andearnings in four groupsaccording to the main (but not necessarily exclusive) focus of the various contributions. 15 · A.The measurementof language-baseddiscrimination accordingto aperson’s firstlanguage hasbeen studiedin the United States(Fogel, 1966)and Canada(Raynauld et al.,1969),and more recently in Switzerland(Grin, 1997b;Grin & Sfreddo, 1998).This research confirms the presence ofearnings differentials between personsbelonging todifferent language communities, even aftercontrolling for their second language skills,including in the dominantlanguage ofthe countryor region considered. · B.The estimationof the valueof second language skills ,when the second language in questionis demolinguisticallydominant in the region orcountry considered,makes up the bulk ofthe economicsliterature (particularly the workpublished in mainstreameconomic journals). This situation, of course, mostdirectly reflects the experience ofimmigrants. It has been studiedwith sampledata on immigrants on the United States(see pioneering workby Grenier, 1984;for an overview,see Bloom& Grenier, 1996;for a setof very Language Planning and Economics 19

recent results,see Chiswick& Miller, 2002;for a moredetailed examination ofthe link withskills levels,see Dávila & Mora,2000), Canada (Chiswick & Miller, 2000),Israel (Chiswick & Repetto,2001), Australia (Chiswick & Miller, 1985);and Germany (Dustmann, 1994; Dustmann & vanSoest, 2001); forsome international comparisons, see Chiswick& Miller (1995).Unsur- prisingly, these publicationsindicate that immigrants benefit significantly fromknowing the dominantlanguage oftheir countryof residence. · C.The estimationof the valueof second or foreignlanguage skills , when the language in questionis not demolinguisticallydominant in the region orcoun- tryconsidered. This case can be further subdivided in twotypical situa- tions:one concernsthe residentsof aparticularcountry (say, Austria) who havelearned a foreign language (say,English orSpanish); the other concernsthe members ofone establishedlanguage communityin amulti- lingual country(say, French-speaking Canadiansor Flemish-speaking Belgians) whohave learned anotherofficial or nationallanguage (in these examples,English orFrench respectively). Researchinto this question is very much aCanadiantradition reflected in alargenumber ofpapers, offi- cialreports and books (for anextensive review, see Vaillancourt,1996; see alsoChristofides & Swidinsky, 1998)mainly using censusresults. The focusof the Canadianresearch is on earnings differentials between anglophonesand francophones, with particular attention to Québec. Cana- dianresults indicate significant variation between genders andacross provinces.The ratesof return toFrench–English bilingualism arehigh in Québec, but they tend tobe modestelsewhere; they arealso much more likely to be significant for men than for women. Swissresults using survey dataon earnings and skills in the country’s three mainofficial languages (German, French andItalian),plus aforeign, non-official language (English) similarlyreveal considerable variation depending onarespondent’s gender, the language region where heorshe resides,and the targetlanguage considered.For example, althoughthe ratesof return onEnglish-language skillsare high throughoutthe country, they comefirst (before French) in German-speaking Switzerland,but second(after German)in French-speaking Switzerland(Grin, 1999c, 2001). Thiscase will be examined in moredetail in ‘AnApplication:Language Education Policy’. One rareexample ofresearchon thisset of issuesin adeveloping country isa paper by Chiswick et al. (2000)using survey dataon Bolivia containing informationabout respondents’ linguistic profile as‘ monolingualSpan- ish’, ‘monolingualindigenous andbilingual (which mayin factmean competence in morethan one indigenous language plus Spanish); results indicatethat monolingual speakers of Spanish fare better thanbilinguals – and,needless tosay,monolingual speakers of indigenous languages.The moststraightforward interpretation of this finding shouldbe thatthis stronglysuggests discrimination against speakers of languages like Aymara, Quechua or Guaraní. However,much hinges onthe definition andreliability oflanguage skills categoriesin the dataset. Most of the studiesquoted here rely onfairly 20 Current Issues in Language Planning

vague categories;these categoriessometimes refer tolanguage use rather thanlanguage skills.Tomy knowledge, the mostdetailed data base remains the Swissone, where adistinctionis made between four different types of skill (listening comprehension,speaking, reading andwriting) andfour different skillslevels (fluent, good,basic, none), withreference toa self-evaluationgrid inspired by the levels used in the EuropeanLanguage Portfolio.16 · D.Finally, there are(still exceedingly rare)investigations of the rates of return on immigrantlanguages in their new countryof residence. They raise the questionof whether –contraryto what seems assumed a priori by much ofthe researchin group Babove– immigrantlanguages might be anasset ratherthan a hindrance.Both a Canadianstudy using censusdata covering 13non-official languages(Pendakur &Pendakur, 1998)and a Swissstudy using survey datacomparing Italian- and Turkish-speaking immigrants (Grin et al.,2002)conclude thatthis value is,at best,very low;most of the time,coefficients forimmigrant language skillsin earningsregressions are either notstatistically significant, or negative; some niche effects canbe found where immigrantlanguages translate into economic advantage, but it would be hasty to generalise from mainly anecdotal evidence. In the contributionsfalling in the BorC groups(and, to someextent, the D group aswell, since competence inanimmigrantlanguage mayrequire adelib- eratechoice to maintain this competence), languagesare primarily seen as elements ofhuman capital in whichindividuals invest. This interpretation, however,requires qualification,because the investmentmay not be entirely voluntary.Learners acquire other languages more or lesssuccessfully; language acquisitionmay also occur through informal learning, throughresidence in a foreign country,marriage with an L1 speaker ofanotherlanguage, orsimply asa resultof the factthat they areled tomake this investment because the authorities require particularlanguages to be taughtas schoolsubjects aspart of their education policy. The strategicimportance of empirical work on language-based earnings differentials, however,lies notonly in itscapacity to provide estimates of the actualeffect oflanguage attributeson earnings. It also enables usto reconsider criticallysome commonly held views,such as the oft-encountered metaphorof ‘language asvalue’, which,tempting asitis, usually fallsshort of areliable guide forpolicy action(see ‘The ‘value’of language’below). Arelatedline ofinquiry focusing onthe statisticallink between linguistic identity (L1) andbusiness ownershipmust be mentioned here (see e.g. Vaillancourt& Leblanc,1993). It generates politicallyimportant information about the relativesocio-economic positionof different language communities,showing, in the casequoted, that francophonecontrol over the economyof the Provinceof Québec hasmarkedly increasedsince 1961.However, the scarcityof dataseverely restrictspossibilities to replicate this research elsewhere. Language dynamics Sociolinguiststhemselves readily concede thattheir discipline hasnot produced ageneral theoryof language dynamics 17 (e.g. Appel &Muysken, 1987). Understandably,this theme keeps attractingconsiderable attention among Language Planning and Economics 21 language specialists,particularly in the caseof language decline andlanguage death(Crystal, 2000; Robins & Uhlenbeck, 1991;Skutnabb-Kangas, 2000),or in the caseof the spreadof English (Ammon,1994; Crystal, 1997; Graddol, 1997; Pennycock, 1994;Phillipson, 1992;Watts & Murray,2001). A general theorystill needs tobe developed, althoughan essential milestone was reached with Fishman’s analysisof language revitalisation(under the expressionof reverse language shift ; see Fishman, 1991). Economistshave also tried their handat this problem, developing various modelsof language behaviour yielding implicationsfor language dynamics, startingwith Hocevar (1983). Some modelsare static, in thatthey assumea one-period calculusin which(often bilingual) individualswill decide which language touse in different activities(e.g. Grin,1990); other models are dynamic, in thatthe economicmodel of language use attime t influences language use at time t+1, then t+2,etc.(Grin, 1992; Pool,1991b); these modelslend supportto the tentativeconclusion proposed by Nettle (2000),who suggests that economic incentives determine language choices rather than the other way around. Aparticularlyinteresting range ofmodelsexamines the network effects associatedwith languages (apart from Pool, 1991b, see Church &King, 1993;Selten & Pool,1991): one intriguing dimensionof languages,which setsthem apart from mostother ‘ commodities’in aneconomicsense, is that when morepeople use a language, the more useful itbecomes, all other things being equal, to other people. Thisis markedly different from,say, a public transportsystem (which becomes lessand less usable the morepeople travelwith it) andof coursefrom a standard privategood (say, an apple), which cannotbe eaten (used) by different people simultaneously. 18 Anotherkey implicationof the networkdimension of language spread throughdeliberate language learning isthat because someof the benefits that resultfrom language acquisitiondo notaccrue to the individual(whereas learning costdoes), people maybe led to under-invest insecondlanguage acquisition. Thisprovides justification for state support to language learning notfor political orcultural reasons, but onthe basisof hard-nosedeconomic considerations. The readeris cautioned, however, that in the mostdirect applications of the approach,this argument usually yields the recommendationthat non- anglophonesshould be encouraged tolearn more English, rather than the other wayaround. A relatedresult obtains in Choi’s (2002)very abstractmodel of the influence ofinternationaltrade on language spread,which amounts to naturalis- ing the dominanceof English. Therefore, further modeldevelopment, witha morecareful treatmentof benefits andcosts, is necessaryto understandnetwork effects andto derive their policy implications.This more sophisticated treatment,which shouldtake non-market componentsof value intoaccount (see ‘The Economicsof Language Policy’below), mayjustify statesupport for the learning of languages other than the most dominant one. There islittledoubt that the networkeffects oflanguage play amajorrole in language dynamics,and hence in the attractivenessoflearning particular languages.Hence, they alsoinfluence the contextof language educationpolicies.Network effects raisehighly complextechnical problems, which atthis timeare notsolved in theliterature(or canbe solvedonly inpart,at thecostof severely restrictiveassumptions which detractfrom the practicalusefulness of 22 Current Issues in Language Planning theanalysis).Further researchinto the ‘networkexternality’ effects oflanguages andtheir implicationsfor spontaneous or policy-induced language dynamics shouldbe considereda priority,particularly in orderto explore long-term dynamics,whether in general termsor in relationto the macro-levelprocess of globalisation (Grin & Rossiaud, 1999; see also Bruthiaux, this issue).

Language and economic activity Thiscategory contains extremely diverse,even heterogeneous, lines ofwork, which focuson the processesof production,consumption and exchange and examine the roleof language in them.Production does, of course, address the behaviourand inner workingsof the firm,whether asaseller offinal goodsand services,as aseller andbuyer ofsemi-finished goods,as abuyer ofproduction factors,notably labour, and as aborrowerof capital. 19 Consumptionharks back tothe behaviourof the householdor individualperson, both as abuyer ofgoods andservices and as asupplier ofproductionfactors, first and foremost labour but alsocapital. 20 Lastly,exchange takesplace onmarkets for final goodsand services,for intermediate goods and services, and for production factors (labour andcapital). These marketsare characterised not only by the interactionof supply anddemand and by the actualexchange, overa certainperiod oftime,of acertainquantity of agoodat acertainprice, but alsoby their structure.Each marketmay be moreor lesscompetitive (which implies alarge number ofbuyers andsellers) or,on the contrary,monopolistic (single seller) ormonopsonistic (single buyer), with many gradations in between. Despite the factthat research in thisfield couldbank onawhole arrayof core conceptsin economictheory, it remains a relativelyunder-developed areaof language economics.This may be tracedback to apointmade above under ‘The Need foran Economic Perspective onLanguage Issues’: economicanalysis is well-suited tothe studyof relationships between variablesthat have a clear quantitativeinterpretation, which typically is not the casefor language. Hence, it shouldnot come as asurprisethat most of the workin thisarea proposes little in the wayof general theory,and not much morein the wayofformal applications ofthe standardeconomic theory of production, consumption and exchange. Rather,most publications are organised around more inductive approaches (offering atheoreticallyplausible explanationfor an observed consumption, productionor exchange situationin whichlanguage appearsto make a difference), orsimply describe ordocument patterns of language use in those economicactivities. However, two important theoretical contributions focusing onthe productionside arethose by Hocevar(1975), who analyses changes in productioncost functions depending onthe language characteristicsof the outputs,and Sabourin (1985)mentioned earlier,who examines the processof ‘matching’between employees withgiven linguisticattributes and specific posi- tionsin afirm,which require ahigher orlesser degree oflanguage skills(practically,more or lessbilingualism). The lines ofinquiry opened up by these two contributionsremain in need offurther workand empirical testing. It would also be useful togather and confront descriptive evidence onlanguage use in the workplacein different countries,since barringa few notableexceptions such as Lambert and Moore (1990), it remains rather scattered and difficult to access. Issuesof consumptionare difficult toseparatefrom issues of exchange in the Language Planning and Economics 23 language economicsliterature. Using dataon the 22Organisationfor Economic Cooperationand Development (OECD) member countries,Helliwell (1999) finds thatthe existence ofacommonlanguage between twotrading partners mayhave a positiveeffect onthe volumeof trade. 21 Cremer andWilles (1991)use small-scalesurvey datato analyse language use in tradingactivities in the Far East,showing that such trade can take place witha remarkably low degree of secondlanguage competence amongthe varioustrading partners. Some work attemptsto model language use in advertisingand consumer relations (Grin, 1994b),but mostof the few contributionson thistopic 22 areempiricalor descriptive,documenting ageneral preference by bilingual customersin Cataloniaor Québec (Vaillancourt,1985b) for being offered goodsand services in their own language (even if they canperfectly well understandanother – usually dominant – language). Separate mentionshould be madeof someresearch grounded in ‘intercultural management’, which isinspired lessby the paradigmof economics than by conceptsfrom business administrationand management, sometimes also by economicgeography. Connectionswith language economicsarise, however, fromthe factthat these contributionssometimes discuss the roleof minority language maintenanceas afactorof regional economic vitality (e.g. Price, 1994; Taddei& Antomarchi,1997). Generally, the conceptualbasis of thisline ofwork replicatesthat of researchon immigrationand ‘ ethnic business’(see e.g. Berset et al.,2000)and does not single outanalytically the roleof language asanexplana- toryvariable in acausalprocess. Instead, it points to the importanceof culture (which, ultimately,turns out to be in partdefined by language) in given aspects of production and exchange. Progressivelymoving from micro-economic issues (to which most of language economicsbelongs) tomore macro-economic issues, some of the contributionsin thisgroup tryto quantify the impact,on the regionaleconomy, of the presence of language-specific activities(e.g. ÓCinnéide &Keane, 1988;Sproull, 1996). Othershave attempted to relatethe linguistic profile ofworkersin aneconomyto the economy’s externaltrade patterns (Stanton & Lee, 1995);the implied theoreti- callinks are,at this time, not elucidated, and no empirical results stand out. Arcand’s work(1996), though primarilymicro-economic in conceptsand methodology,also addresses macro-economic implications, revisiting economic modelsof development andgrowth in ThirdWorld countries to discuss the possibilitiesof integrating language asan explanatory factor in development processes.23 Interestingly, the production,consumption and exchange of‘language goods’ and‘ language services’(such ascourses,books or other cultural products in a particularlanguage, etc.)do not constitutea significantpart of language economics.One mayeven argue thatthere wouldbe noreason at allfor such topics to be included in language economics.This is because the production,consumption andexchange ofsuchcommodities is not markedly different fromthe production,consumption and exchange of other (i.e. non-linguistic )commodities,and therefore doesnot justify specific analysis– otherthan, possibly, in the formof casestudies on aspecific market.Generally, the analyticalconcepts of supply, demandand market for any given goodor servicealso apply tolanguage goods. The situationis different, ofcourse, when talkingabout supply anddemand for a 24 Current Issues in Language Planning

[particular]language (thatis, when the commodityis a language takenas a whole,not specific goodsand services in thislanguage) whichare, of course, perfectly relevantareas of researchfor language economics,but where itis importantto reconstruct the conceptsof supply anddemandcarefully (we shall return tothis question later on in thissection). However, one range ofservices thatdo not merely havelinguistic dimensions (like books inlanguageX ), but are intrinsically linguistic havebeen inadequately studiedin economicperspective anddeserve seriousstudy – namely,the production,consumption and exchange oftranslationand interpretation. 24 In the sameway, there is,to my knowledge, hardlyany theoretical or empirical work on the economicsof terminological development or, more generally, on the economics of language corpus.

The economics of language policy Becausea sectionbelow isdevoted to the economicperspective onlanguage policy,I shallconfine myself here toa brief account.Let us first recall the point madeearlier thatif economicscan make useful contributionsto the analysisof language policy,it is not so much because itbrings linguisticand economic vari- ablesin relationwith each other (with causallinks flowing in either direction), but ratherbecause ithelps tolookat different choicesabout language in termsof advantagesand drawbacks. This does not imply lookingat languages in a narrowlymaterialistic perspective, orgauging themstrictly in termsof the monetaryadvantages or money costsassociated with them. Nonetheless, economicsis particularly useful because ithelps, if sodesired, to re-express advantagesand drawbacks in monetaryterms, thus facilitating comparison. Thateconomists should enter the frayof language policy isa logicalconsequence ofthe factthat society is confronted with choices regarding language (or,more to the point, languages)andhas to make decisions in thisarea – justas itdoes regarding transport,health, the environment,etc. Economics is relatively better equipped thanother social sciences, conceptually and perhaps methodologi- cally, to process decision-making problems. The whole endeavour,however, also raises the questionof the nature of language policy andwhat language policy isexpected tomodify. Most of the economicwork on language policy addressesthe positionof one language vis-à-vis otherlanguages, or the broaderquestion of linguisticdiversity . Status issues(in abroadsense) aretherefore centralin the economicapproach to language policy,whereas corpus questions have practically never been studied; the approachis mainly about how human action can affect our linguisticenvironment.The conceptof linguistic environmentused here aswell asin someearlier work(Grin, 1997a, 1999b) may differ fromthat encountered elsewhere, forexample in the literatureon linguistic ecology(Mü hlhä usler, 1996, 2000). I donot use it toimply anybiological parallel, but because ithelps toformulate the links between languagepolicy onthe one hand,and the field of policyanalysis on the otherhand. The basicidea is that language policy aimsto change the linguistic dimensions of the surroundings (or of the environment) in which we live. Clearly,much ofthe theoreticalliterature on the mutualinfluences between linguistic andeconomic variables (in particularwork quoted elsewhere in this paper, e.g. Carr,1985; Church &King, 1993;Hocevar, 1983; Pool, 1991b, 1991c) hasdirect implications for language policy,and a concernfor language policy Language Planning and Economics 25 implicationshas been present in the literaturefor a long time.At the sametime, the connectionto language policy,in mostof the earlier language economics literature,was somewhat derivative, because itresponded tospecific develop- mentsof whichauthors had personal experience. 25 Thisfocus on policy issues hasbecome much moreexplicit in recent years.Possibly echoing callsto this effect morethan 30 yearsago (Jernudd, 1971;Thorburn, 1971), much ofthe ongoing economicwork on language policies addressesthe benefits andcosts which characterise policy options(Grin &Vaillancourt,1999; Pool, 1996) or attemptsto transposebasic decision-making criteria in the contextof language policy (Grin, 2003).The thrustof mostrecent researchin thisarea is on the identificationof the mainsources of benefits andcosts, from the perspective ofindividuals and of society,attaching to variouspolicy alternatives,and to propose estimates of the order of magnitude of such benefits and costs. Mostof the literaturefocuses on ‘allocative’rather than ‘ distributive’issues, and this important distinction requires a few words of commentary. The allocative level focuseson efficient resourceallocation, but noton who gainsand who loses as aresultof apolicy choice.It is merely concerned with ensuring thata policy generates againin aggregatewelfare ,opening the possibility forwinners to compensatelosers, so that no one isworseoff atthe end ofthe process.By contrast,the distributive level ofanalysisdoes raise the issueof fair- nessand focuses on the identificationof winnersand losers, as well asonthe estimationof respective gainsand losses. This examination is a preconditionfor proposingsystems of compensationas well asfor investigating ways of ensuring compliance in the case of incentive-based policies. Followingpioneering workby Pool(1991c), 26 there hasbeen renewed interest in distributiveissues in recent years(Grin &Vaillancourt,2000; Van Parijs,2001), anda closeranalysis of distributivedimensions is undoubtedly amongthe most challenging anduseful lines ofinquiry topursue. In particular,it has direct rele- vanceto pressing problems suchas the choiceof official languages in multilingual structures like the European Union. The fundamentalconcepts of language policy evaluationin economic perspective arepresented in the sub-sectionon ‘ Language policy:Diversity, and the “fundamentalrule” ’below. However, it is useful topointout at thisstage that educationeconomics alsoprovides relevant inputs, principally because education tendsto be the single mostimportant channel ofgovernment intervention in the sphere oflanguage. Otherchannels of interventiondo of courseexist, such as the regulationof language use inthe administrationor in the judicial system,state financing orat least state support to audio-visualmedia, cultural policy in literature andthe fine arts,etc. Still, educationin general remainsa majorarea of competence ofstates, typically representing between 10%and 20% ofaggregate 27 governmentexpenditure indeveloped countries.As such, it usually represents the single largestbudget itemand is often the mostimportant vehicle oflanguage policy.Accordingly, the benefits andcosts associated with education-based language policies alsotend tobe larger,in monetaryterms, than those that proceed fromintervention in otherdomains. Given the importanceof education, the whole ofthe sectionbelow ‘AnApplication:Language EducationPolicy’ is devoted to the principles guiding the evaluation of language education policy. 26 Current Issues in Language Planning Other research directions Thisbrief overviewdoes not exhaust the field oflanguage economics.In addi- tionto further detailand references which couldbe provided foreach of the abovecategories, other, most of the timeanalytically isolated, contributions on otherdimensions of the language experience alsoexist. Marschak’ s (1965) attemptto explain patternsof internal language change asadrive towardseffi- ciency in communicationhas been mentioned earlier. 28 Colomer’s (1991)models ofthe choiceof language in conversationsbetween people havingdifferent linguistic attributesexplains howthe questfor effectiveness will,depending on the numbers ofthosepresent, favourthe use ofone oranotherlanguage. Using a similartype ofmodel, Colomer (1996a) argues that under very general assumptions,generalised foreign (orsecond) language learning, which generates wide- spreadindividual multilingualism, is likely tobe amorecost-effective solutionto the problem ofcommunicationin plural societiesthan systematic resort to trans- lationand interpretation facilities; these approachesare combined in Colomer (1996b). Awordalso needs tobe saidof the language-as-currency analogy,because the metaphoris proving tobe avery tempting (usually fornon-economists more thanfor economists), but also,potentially, a very misleadingone. From the non-economicside, this analogy was put forwardby Rossi-Landiwell over30 yearsago, when he compareda linguisticcommunity to ‘asortof huge marketin whichwords, expressions and messages circulate as commodities’(Rossi-Landi, 1968:49).Seductive asit may appear,the metaphoris not sound. A marketin the economicsense emerges fromthe existence ofsupply anddemand functions. Supply isdefined asthe willingness by producersto offer acertainquantity of a certaingood or service at acertainunit price overa certainperiod. Demand is defined asthe willingness by consumersto buy acertainquantity of thatgood or serviceat a certainunit price overa certainperiod. Normally, supply isan increasingfunction of price, while demandis a decreasingfunction of price. Hence, the supply curve andthe demandcurve will intersectin a two-dimensional{price-quantity} space,determining an equilibrium level both for quantity (q*) and price (p*), as shown in Figure 1.

Figure 1 The market diagram Language Planning and Economics 27

Eachof these elements ismissing from Rossi-Landi’ s account,making the reference toa ‘market’a purely metaphoricalone withlittle heuristic pertinence. Tosome extent, the samecan be saidof Bourdieu’s (1982)often-quoted essayon ‘linguisticexchange’ . Bourdieu’s textmakes constant reference to‘ markets’, ‘profit’and ‘ capital’, thereby creatingwith some readers the impressionthat his is an economic theoryof language use. Itwould be moreappropriate to describe hiscontribution as a sociologicalone in whichstandard economic terms are given another,somewhat idiosyncratic interpretation. Bourdieu’ s analysis certainlyamounts to excellent sociologyoflanguage –but froman economic standpoint,his use of economicterminology is no less metaphorical than Rossi Landi’s,29 and does not amount to ‘economics of language’. The true meaning ofapossible parallelbetween language andcurrency has been developed by Carr(1985). Carr shows that the parallelto be drawnis not between language onthe one handand currency onthe otherhand, but between the factthat there existdifferent languagesand the factthat there existdifferent currencies.In the sameway asthe use ofa commoncurrency facilitatesexchange by reducing tradingcosts, so does the use ofacommonlanguage. The actual extentof costreduction, incidentally, is an empirical question that crucially depends onthe tradingpartners’ reliance onlinguisticcommunication. As noted before, the evidence ispatchy, and more detailed examination remains necessary.30 Anotheroft-repeated (andsometimes hasty) analogy is that of language asa formof ‘ wealth’. Since thisquestion is crucial to the whole language policy endeavour, its treatment is deferred until the next section. Finally, the readershould be awarethat there existsa strandof literatureon the languageof economics .However,it largely consistsin ananalysisof economics discourse,from an economic (McCloskey, 1990)or linguistic perspective (Henderson et al.,1993).As such,it clearlyfalls outside the definition oflanguage economics proposed earlier and will not be discussed in this paper.

The Economics of Language Policy

A working definition of language policy Language planning andlanguage policy arefast developing asmajorareas ofanalytical and practical work in the language disciplines,particularly sociolinguisticsand applied linguistics.The sameistrue in language economics, andthe morevaluable contributions that an economic analysis can make to language issuesare those that directly or indirectly help tomake informed choicesabout language issues– andultimately, making informed choicesis precisely whatlanguage planning orpolicy (asdistinct from language politics; see below) are about. Thissection does not attempt to discuss language policy andplanning in sociolinguisticperspective, insteadreferring the readerto the growingliterature in thisfield (e.g. Calvet,1996; Cooper, 1989; Fishman, 1991; Kaplan & Baldauf, 1997;Labrie, 1993; Maurais, 1987; Schiffman, 1996;various contributions published by the Generalitat deCatalunya ,1997,1999; or in EuropeanCultural Foun- dation, 1998; etc.). Nodistinction is made here (noris it made in the economicsliterature) 28 Current Issues in Language Planning between ‘language policy’and ‘ language planning’. In thispaper, bothexpres- sions refer to: asystematic,rational, theory-based effort atthe societallevel tomodify the linguistic environmentwith a view toincreasing aggregate welfare. Itis typicallyconducted by officialbodies or their surrogatesand aimed at part orall of the populationliving under their jurisdiction.(Grin, 1999a: 18, adapted from Cooper, 1989, Chap. 2) A few comments about this definition are in order. ‘Systematic’ The use ofthisadjective means that language policy isassumed to be (even if informally) organisedaccording to acertainmethod and aims at certaingoals, whichmay be overtor covert.The assumptionof rationalityfollows logically: to the extent thata certainmethod is applied toreach certain goals with certain means,rationality implies thatthe latterare judiciously applied toachieve the former.Of course,this does in no way imply thatno improvements could be made. Rationalityis a process,not anachievedstate ofaffairs .Itonly implies that actorsinvolved in the selection,design andimplementation of language policy areusing certainmeans towards certain ends in awaythat is, to the bestof their (necessarily incomplete) knowledge, rational. ‘Linguistic environment’ The conceptof linguistic environmenthas been introducedin the preceding section.‘ Linguisticenvironment’ must be understoodas the object of language policy.Whether itaims to increase the visibility ofan autochthonousminority language in culturallife, todevelop foreign language skillsamong the school-age population,or even to suppress the use ofcertainlanguages, 31 language policy is intended to modify the linguistic features of our environment. 32 ‘Welfare’ Whether thisis explicitly statedor not, the goalof all language policy isto increase‘ welfare’– andthis, of course,opens the Pandora’s boxof the questionof what welfare isor, more precisely, whatcontributes to make welfare higher. Obviously,there need notbe agreement onthis matter, but fundamental economicanalysis normally eschews this issue and adopts a deeply liberal stance:welfare isseen assomethingthat only individualscan define forthem- selves,and it is not up tothe economist(or, for that matter, the sociologist,the politician,the priestor the guru) todecide whatwill makesomeone else, orsoci- ety asawhole,happier. One canonly assume,in ananalyticallydistinct step, that mostindividuals will generally feel happier (thatis, their level ofwelfare will be higher) if they arehealthier, if they aremore socially integrated, if they enjoy higher levels ofconsumption of goodsand services, if the airthey breathe is cleaner,their roadssafer, and soon.Dismissing this analytical approach as ‘utilitarianism’, asnoted in ‘The Need foran Economic Perspective onLanguage Issues’above, would be either absurdor perfectly trite.One mayof course reject the notionthat individuals enjoying allthe abovewill begenuinely happier, but then one will be hardput tocomeup witha better,more credible assumption about human preferences. Language Planning and Economics 29

Atthis point, it is important to note once more that no restrictions apply regarding the sourcesof the ‘welfare’referred toin economicsin general orin the definition oflanguage economicsjust proposed. In line withfundamental economictheory, welfare isnot just a matterof material consumption or well-being. Itcan also include anynon-material elements, including afeeling of safetyin one’s identity orthe preservationof markersof culturalheritage. This, ofcourse, establishes links withother branches of economics in which non-materialor symbolic values are taken into account, such as ‘ cultural economics’. However,contrary to a frequently encountered belief, the closest relativeof language economicsis not culturaleconomics, but environmental economics.There aretwo main reasons for this. The firstis that environmental economicsis largely concerned withweighing the advantagesand drawbacks of different policy options,and that the conceptualand methodological problems thatarise in thisendeavour are strikingly similarto those encountered in language policy evaluation.The secondis that language, justlike the biophysical environment,is much morepervasive in humanexperience thanthe type of goods and services on which most of ‘cultural economics’ concentrates. Welfare becomesmore problematic once we movefrom the individualto the collectivelevel. Becausepolicies affect societiesas a whole,the problem is whether society’s welfare as a whole will be higher asa result ofthe policy ; this is why the definition indicatesthat language policy takesplace ‘atthe societal level’.33 Apartfrom the fact,discussed below, that this gives akey roleto the state, itraises the questionof the criteriaaccording to whichthe authoritiesconclude thata policy will indeed increase aggregate welfare, given thatpolicies typically createwinners and losers – oratleast drive awedge between big winnersand smallwinners. Economists will, however, readily concede thatwelfare analysis takesplace withina politicaland moral framework reflected in institutional arrangements,and agree thatpolicy choicemust abide by ademocraticprocess basedon majorityrule withappropriate safeguards for the interestsof the minority – or, most probably, minorities.

‘The state and its surrogates’ Anyone, downto the individual,can have a language policy. 34 A multina- tionalcompany can havea language policy.However, using the notionof policy forsuch a broadrange ofsituations would expose usto the riskof stripping the conceptof much ofitsmeaning, which iswhy, in line withmost ofthe literature,I stickto the notionof a society takingsteps to influence itslinguistic environment, justas itmaydevelop anenvironmentalpolicy affecting landuse, wastedisposal orgas emissions. In otherwords, what I meanhere isactual or potential public policy. Typically,the organthrough which societytakes such action is the state,and we shallsee below thatthere areintrinsic reasons linked tothe nature oflinguistic environmentswhich require the stateitself toact,because in the absence ofstate intervention,the private(market) sectorwould give riseto behaviour which would not resultin the linguisticenvironment desired. This does not imply, however,that the statecannot delegate the implementationof someparts of language policy toother bodies. For example, privateoperators can run minoritylanguage televisionchannels. The pointhere isthat in the absence of state 30 Current Issues in Language Planning interventionthrough deliberate language policy,private operators may not provide the service.

Policy v. politics Finally, readerswill observethat ‘ policy’is understood here notin termsof politicalpower play between actorsholding opposing views,but in termsof a comparisonbetween different solutions,each of which has a priori advantages anddrawbacks, which can usually beexpressed as‘ benefits’and ‘ costs’. This emphasisharks back to a definition in policy analysisdiscussed for example by Scharpf (1997)between ‘interaction-oriented’and ‘ results-oriented’policy analysis.The definition used here clearlyfocuses on results-orientedanalysis, while interaction-orientedanalysis would be consideredcloser to the politics of language, onwhich a substantialliterature exists, 35 andwhich in general doesnot translate into economic questions as defined at the beginning of this paper. Thisdistinction between whatis part of ‘politics’and what is part of ‘ policy’ directlyties into the questionof involvementby the stateand the reasonsfor the latter.

Language policy and state involvement Itis generally acceptedthat in moderndemocratic societies, a certainrange of dutiesis among the responsibilitiesof the state.Why thisshould be soisan issue thathas occupied politicalphilosophers since antiquity.From the standpointof mainstream(neoclassical) economics, however, the questionmay be approached withoutresorting to moral or philosophical considerations, referring instead, justlike the restof fundamentaltheory, to the problem ofscarceresources that havealternative use. Thisis not to say thateconomic analysis necessarily rejects moralor philosophicalconsiderations. These mayoffer the mostconvenient way toprovide alogicalbasis to someforms of stateintervention – forexample, in distributiveissues. However, it istechnical, rather than moral or philosophical, argumentsthat establish, within the systemicrationality of economicanalysis, the necessity of state involvement in certain cases. Wehaveseen that languagepolicy aimsat modifying alinguisticenvironment in orderto increase welfare. The questionthen arisesof why the linguistic envi- ronmentthat exists independently ofdeliberate stateintervention in language is not the best,andon whatgrounds we canpresume toimproveit through deliber- ateintervention. One couldindeed haveargued, following a standard lais- sez-faire ideology,that government should not intervene, andthat maximum welfare will automaticallyflow from the decentralisedactions of people (individuals,firms, third-sector organisations 36),allowingmarket mechanisms to regulate the linguisticenvironment. This will be expected toproduce,as itwere, the rightamount of each of the featuresof this environment, such as levels of skillsin different languages,patterns of use ofdifferent languages,types ofatti- tudestowards different languages,extent ofvisibility ofdifferent languages,etc. If the free marketcould be seen asanappropriatemechanism whereby allof society’s goalscan always be reachedand maximum aggregate welfare deliv- ered, there wouldbe nogrounds,in mainstreameconomic theory, for state intervention,and everything shouldbe left toprivate initiative. The underlying adjustmentmechanism goes as follows:if notenough ofsomethingis produced, Language Planning and Economics 31 itsprice rises,and producers will respondby increasingoutput to the desired level (while the risein price simultaneouslydiscourages some consumers); and if toomuch ofsomething is being produced,its price will drop,meaning thatmore ofthe goodwill be absorbedby consumers,while someproducers will simply turnaway from a productionline thatno longer generates sufficient profits.Let me repeat thatthere isno exogenous notionof ‘ toomuch’ or ‘ toolittle’ . ‘Too much’ is revealed by the presence ofexcesssupply, and‘ toolittle’by the presence of excess demand. While allthis constitutes a fairly credible line ofargumentfor simple goods suchas tomatoes, television sets or cartyres,it becomes more difficult toaccept in the caseof complexcommodities such as education,health, or the environment– and,of course,language. Even fromthe standpointof mainstreameconomics, there are strong analytical reasons for state intervention. Accordingto economic theory, there aresome cases where the marketis not enough. These casesare known as ‘marketfailure’ . When there is‘ marketfailure’, the unregulated interplay ofsupply anddemand results in aninappropriate level ofproductionof some commodity, where ‘inappropriate’can mean ‘ too little’or ‘ toomuch’ . If the marketdoes not work, the statehas to step in. 37 In theory, there are essentially six sources of market failure: 38 (1) insufficient information,which prevents economicagents (producers and consumers)from making the rightdecisions in termsof output, purchases, lending andborrowing; (2) high transactioncosts, which move agents not todo something that would eventually have been economically beneficial; (3) the factthat some markets cannot exist (for example, yet-unborn generations cannotbe present ontoday’ s oilmarket to express their valuation of this non-renewable resource); (4) the existence of‘ marketstructure imperfections’ such as monopolies and oligopolies; (5) the presence of‘ externalities’, thatis, of a situationwhere the behaviourof one agentaffects (positively ornegatively) the positionof another agent, without the gainor loss so created giving riseto a correspondingcompensation; (6) the existence ofso-called ‘ public goods’, whichin the ‘pure’textbook case displaytwo main characteristics described below:‘ non-rivalconsumption’ and‘ impossibilityof exclusion’ . If somelinguistic environmentsare socially preferable tootherlinguistic environments,and if atleastone formof marketfailure occursin the productionof at leastone feature oflinguistic environments,then stateintervention is justified from the standpoint of economic welfare theory. Simple asatis,this result is far from innocent. Contrary to awidely held belief, language policy isnot an ‘ un-economic’endeavour; quite the contrary,it may be the economicthing todo,if itpromisesto deliver anincreasein welfare. Different reasonscan be invoked toexplain resistanceto this interpretation. Experience suggeststhat one sometimesencounters, among non-economists, an automatic assumptionthat economic analysis necessarily views language policy asacostly indulgence. Thisperception may,of course,have been encouraged by the very narrowvision of benefits andcostsadopted by someeconomists themselves (see 32 Current Issues in Language Planning e.g. Jones,2000) not to mention omission by the sameof the distributive (as distinctfrom allocative)dimensionsof linguistic environments;many economists instinctivelyview alllanguage policy asill-advised tampering with the free interplay ofmarket forces. From an analytical standpoint, however, all this is plain wrong,and language policy canbe justified economicallyfor reasons very similar to those brought to bear on environmental policy. Thispoint, however, deserves a closerlook. In the caseof linguistic diversity, marketfailure certainlyemerges throughmore than one ofthe sixchannels just mentioned.In fact,a strongcase could be madethat all sixsources of marketfailure arepresent, providing asmany economic reasons to engage in language policy.Furthermore, these manifestationsof market failure areoften interconnected. Let us briefly consider them in turn:

(1) The presence of insufficientinformation canbe invoked,but itmayprove a weakargument. Such aclaimwould imply, forexample, thatactors do not realisehow dependent the qualityof their linguistic environmentis on linguistic diversity.Consequently, outof their ignorance,they fail toallo- cateby themselvessufficient resourcesto the maintenanceof linguistic diversity,with the resultthat minority languages (whether these areseen as‘ indigenous languages’spoken by ‘indigenous peoples’or as ‘autochthonouslanguages’ spoken by ‘linguistic minorities’39)suffer attri- tionand decline. Touse anotherexample, somecommentators claim that socialactors in non-anglophone countriesfail torealise that their linguistic environmentwill notjust be affected, but adversely affected, by encroach- mentsof English, andtherefore fail tobehave in awaythat would secure a certainquality of their linguisticenvironment. 40 Thisline ofargument, however,suffers fromone majorflaw, namely, the implicationthat people don’t knowwhat they aredoing andwhat is good for them – andits corol- larythat the enlightened academicor politician knows better. Outside of addictivegoods (which maytherefore be seen notas ‘goods’, but as‘bads’), thisline isunconvincing (atleast in democraticsocieties), and economists will usually defer topeople’ s expressionof preferences asrevealed by their behaviour, whatever these preferences may be. (2) The caseof hightransaction costs ismore interesting, though paradoxically little-explored in the caseof language policy.The issueis analytically not quite the sameas that mentioned in the preceding section,namely, that transactioncosts may be lowerwhen people adopta commonlanguage. The issuehere isthat owing to the existence ofcostsattaching to the practical conditionsneeded tomake exchange possible,people mayabstain from a potentiallyprofitable exchange. Transposingthis problem toour case meansthat people mustbe shownto behave in awaythat results in the emergenceofasub-optimallinguisticenvironment.Ifthenotionof transactioncosts is broadened toinclude the costsof transition fromone type oflinguistic environmenttoanother,this line ofargumentcan certainly hold.For example, one mightargue thatthe interestsa supra-national organisationsuch as the European Union,which currently uses11 official languages,41 wouldbe better servedif internalcommunication took place in Esperanto,without translation and interpretation (Grin, 1997c; Pool, 1996). Language Planning and Economics 33

Thiswould require civilservants and elected parliamentariansto learn the language. Until the timewhen allparticipants have adequate command of the language, extracosts will emerge (for example, language-learning costs),and these will only be worthshouldering ifallparticipants adopt the samestrategy; until they committhemselves to it,transaction costs will be deemed excessiveand the status quo (orsome alternative, non-optimal solution) will prevail. (3) Non-existent markets represent aperfectly relevantconcern in the caseof language. If languagesare allowed to die out now orin the nearfuture, they will nolonger be availablein the moredistant future foryet-unborn genera- tionsto express, eventhrough market mechanisms ,their preference forhaving these languagesaround. Sticking toa pure marketrationality at a given pointin timeis tantamount to assumingthat future (market) demandfor variousthreatened languages will be zero,and nothing bearsout this assumption(or, more precisely, nothing bearsit out more than it bears out the reverse assumption,namely, that future generationswill highly value linguisticdiversity). Furthermore, linguistic environments do not lend themselvesto apopular logicalfallacy regarding non-renewable resources, accordingto which reserves of non-renewable resources increase (despite the burning offossilfuels) because new reservesare regularly discovered, or because risingenergy prices wouldmake known resources, previously consideredtoo costly to exploit, suddenly economical.First, despite the re-creationof expressionsof diversitythrough the processof globalisation, itis unlikely thatyet untapped repositoriesof linguistic diversitywill replace languagesnow on the verge ofextinction;second, even if thiswere the case,one tonof crude oilmay be aperfect substitutefor any other, whereasa manifestationof diversity, once lost, cannot be simply replaced by another.Hence, the ‘non-existentmarkets’ source of market failure is arguablypresent in the caseof linguistic environments. However, it is exceedingly difficult tomovefrom this general observationto any kind of policy implications.It is not possible toclaimthat future generationsdo not careabout the characteristicsof their linguisticenvironment(s); but itis no morepossible thatthey do,thereby considerablylessening the strengthof the missing markets argument as a basis for state intervention. (4) The issueof market imperfections needs tobe understoodagainst the background ofcompetitionas a‘perfect’market. ‘ Perfection’involves no moral judgement –rather,it refers tothe ideathat owing to itsspecific characteristics,42 acompetitivemarket is one where the uncoordinatedbehaviour of actorswill resultin anoptimallevel ofproductionof various commodities, fromfoodstuffs and energy tolive artsperformances, all of thisat the lowest possible price thatcovers production costs. To the extent thatthe ‘good’ whichlanguage policy mayhelp to‘produce’is a certainlinguistic environment,it should be clearthat the exceedingly complexnature of the latter resistsanalogy with the moresimple goodsthat may be tradedon aneatly demarcatedmarket. Nonetheless, addressing the issueat a highly aggregate level, one mayargue thatlinguistic environmentsare the resultof behav- iourby innumerable actorsin highly contrastedpositions of power,which violatesthe ‘freedom ofentry’ condition. 43 In arelatedmanner, one may 34 Current Issues in Language Planning

claimthat when the behaviourof actorsis analysed not in individual,but in group terms,some groups of actors (for example usersof some major

language LY) can influence the relative‘ value’of different linguistic environments,thereby artificiallylowering the ‘price’of somelinguistic environ- mentswhile ‘pricing’others out of the market.This violates the ‘atomicity’ condition,with the resultthat actors will be tempted todemand the ‘lower-price’(and presumably lessdiverse) linguisticenvironment, whereas their choiceswould have been different in the absence ofundue influence by

speakers of LY.The currentvogue ofEnglish in advertisingand youth culture providesan example. Thisline ofinquiry, however,still awaits closerexamination than the literaturenow has to offer. In particular,it is importantto note that it is predicatedon afairly directapplication of the standardmarket model to linguistic environments, and considerable conceptualclarification is still necessary for such applications to come to fruition.Hence, itremains difficult, atthis time, to give precise policy substanceto the a priori defensible idea,on the analyticalplane, thatlinguistic environments should be ‘produced’ in ‘more competitive’ conditions. (5) Externalities areacruciallyimportant dimension of linguistic environments, andalso one where promisingavenues forresearch have been opened in recent yearsby Church andKing (1993);(see alsoDalmazzone, 1999, for an accessibleintroduction). A positiveexternality ariseswhen the behaviourof some actor h gives risenot only tothe resultthat she anticipated(and which initiallymoved her toaction),but alsoto positiveeffects accruingto other actors.Such positiveeffects canbe seen aspure windfallprofits, because these lucky ‘otheractors’ do not have to compensate actor h.Considerthe caseof alanguage communitycomprising N individualsspeaking language

LX.Suppose nowthat the outsider h,whospeaks language LY, decides to learntheir language, andthat none ofthe N members ofthe LX language communityspeaks LY. Actor h therefore standsto gain the possibilityof communicatingwith N persons,and will be readyto investtime and money in language learning aslong asthe value ofher communicativegain exceeds the value ofher investment.At the sametime, she providesto N individuals, atno cost to them, the possibilityof interacting directly with one more person– herself. All externalitiesare not positive, however, and language learning cangive riseto negative externalitiestoo: suppose nowthat some

members ofthe LX language communitydo speak LY,andmake a living as interpreters between LX and LY,charginga certainrate for their services.The fact that h learns LX amountsto new competition,which will tend topush downwardsthe ratethey cancharge. This is a negative externality,for whichthey arenot compensated, because actor h will haveno reason to give them a subsidy that would leave their income unchanged. The key pointis that because ofthe presence ofpositive and negative externalities,decentralised decisions made by actorsare unlikely toresultin the bestpossible linguistic environment.State intervention is required to modify behaviourthrough policy, which may rely onincentives oron mandatoryregulation. Typically, the externalityargument is used in the literatureto justify supportfor the learning ofa commonlanguage –in practice,to push formore learning ofEnglish. Thisrecommendation, of course, Language Planning and Economics 35

ispredicated on analytical decisions made regarding what countsas abene- fit orasacost.More precisely, one needs toassume jointly: (a) thatlanguage learning isjustified by communicationneeds, andthat the learning of a commonlanguage will consequently be beneficial because itwill make communicationeasier, and will in particularreduce tradingcosts; (b) that the generalised learning ofacommonlanguage doesnot harm linguistic diversity or thatdiversity counts for nothing. Remove either ofthese two assumptions,and the externalityargument, while stillanalytically relevant, maybe used in supportof rather different recommendations,such as massiveinvestment in the learning ofEsperanto,or even, in given areas,of regional or minority languages. (6) Marketfailure relatedto the public good natureof language maybethe most fundamentalway to justify stateintervention. Linguistic environments havesome features in commonwith biodiversity, which is generally recognised asa type ofpublic good.Without engaging in debatable biological metaphors,it is useful toobserve that the sameanalytical reasons that justify interventionto preserve andmaintain our natural environment (which, for the analyticalreasons already explained, simply cannotbe left overto market forces) also apply to linguistic environments . Justlike otheramenities that surround us, such as streetlighting orthe qualityof airand water, languages, as well asthe greateror lesserdiversity ofthese languages,constitute an environmentwhich presentsthe corechar- acteristicsof ‘public goods’. Toexplain thispoint, it is useful todefine the conceptsof ‘ non-rivalconsumption’ and ‘ impossibilityof exclusion’. Apples provide anexample of‘rivalconsumption’ , because the factthat I eat anapple meansthat no one else caneatit. Languages are cases of ‘non-rival consumption’because the use ofalanguage by one persondoes not reduce the ‘amount’of language availablefor use by anotherperson – quite the contrary,as hasjust been shown.Returning tothe caseof apples, there isa ‘possibilityof exclusion’(from consumption) because whoeverdoes not pay forthe apple (ordoes not make the effort togrow it) will notbe able to eatit. In the caseof language, by contrast,there is‘ impossibilityof exclusion’since there isnopractical mechanism (particularly a price-based one) forkeeping apersonfrom experiencing aparticularlinguistic environment. Hence, there isabsolutely no guarantee that the free market(that is, decen- traliseddecisions made by socialactors) generates the behaviourthat will resultin the establishmentand maintenance of the sociallyoptimal linguisticenvironment. Decentralised behaviourmay fail,for example, toguaran- tee anadequate degree of overalllinguistic diversity or an appropriate presence, use, learning, etc. of minority languages in our environment. Nevertheless,some dimensionsof the linguisticenvironment can in principle be left toprivateinitiative. This may be the case,for example, when one considersonly the acquisitionof languagesof wider communication, because privateactors can be directly rewardedfor learning orteaching them.Hence, people willnormallyinvest in language learning (demand) inproportiontothebenefits they derive fromit; and the language teaching (supply) required canbe provided andpaid for as a resultof this demand.This, however, does not hold for several dimensions of our 36 Current Issues in Language Planning

linguistic environment,such as the visibility orpresence ofregional or minoritylanguages. To the extent thatthe preservationof thoselanguages hasbeen, atthe outset,recognised as a‘good’, itfollows that only the stateor itssurrogates can be countedon to takethe measuresthat will resultin an appropriatepresence andvisibility of regionalor minoritylanguages in our linguistic environment.

Letus repeat whatmay be the mostimportant point in thisdiscussion, namely,that the casefor state support may very well be made not onthe basisof politicalarguments or of anappeal tohuman rights or minorityrights, but onthe basisof economicwelfare theory,and taking account of somespecific featuresof diverse linguisticenvironments as a valuablecommodity. All thisraises the question of the value of language, to which I now turn.

The ‘value’ of language Aswe haveseen in the preceding paragraphs,the rationalefor policy inter- ventionhinges cruciallyon the identificationand measurement of advantages anddrawbacks, which it is often convenient totranslateinto ‘ benefits’and ‘costs’. However,these termshave no meaning unless they areunderpinned by aclear conceptof ‘value’. Itis therefore unsurprising thatat somestage or other,much of the politicaldebate aboutlanguage raisesthe questionof the ‘value of language’, withsome participants in thisdebate referring tolanguage asa‘trea- sure’, aformof wealth, etc. It is easyto forgetthat most of the time,such state- mentsamount to littlemore than metaphor; some seem toconsiderthe notion thatlanguage is‘ valuable’as aforegone conclusion,and go on to assume that as a consequence ,languagesare obviously valuable inan economic sense . This is not necessarilyso. Limitations of time and space prevent usfrom entering afull discussionof economic concepts of value andtheir implicationsfor language (Grin,1997a). Let us simply observethat when somethingis valuablein the eyes ofsocialactors, this is usually reflected intheir behaviour.For example, if learning Sámi were obviouslyvaluable, people wouldtake notice without being told, they wouldlearn and use it,the language probably wouldnot be in suchdire straits,and would not be, asitis, in need ofstrongsupport measures. This is not tosaythat Sá mi (orany other minority language) is not valuable.However, to make this point clear, it is important to consider different forms of value. Letus firstconsider the questionof value fromthe standpointof the individ- ualactor, who experiences whatI shallcall ‘ privatevalue’ . Within it,a distinction mustbe madebetween ‘market’and ‘ non-market’values. Market values are reflected in prices orsomesuch indicator. Suppose forexample thatspeaking language LX makesit easier to sell goodsto the LX-speaking public andthereby gives riseto higher profits,or thatan LX-speaking employee earnsmore, all other thingsbeing equal, because heknowslanguage LX:in thiscase, LX has market value.

However, non-market value alsoexists. For example, knowing language LX gives accessto the associatedcultural sphere, facilitatessocial contact with members ofthe LX-speaking community,etc. Such value istypically not reflected in marketprices, but itwill be experienced by individualactors if their tastes,or preferences, include contactwith LX-ish culture andcommunities. The argument Language Planning and Economics 37 maybe extended fromthe caseof one specific language ( LX) to the diversity of languages;non-market value attachesto linguistic diversity for people who value diversityin their linguisticenvironment ,inthe sameway asthey mayvalue acertain(natural) environmental quality. Advocates of diversitymust not forget, however, that some people may simply not care.

The marketand non-market value thatattaches to language LX for a given individual canserve toexplain hischoice to learnor notto learn language LX, and toagree ornotto agree tothe spending oftax resources for programmes in favour of language LX.Yetan additional step isrequired toanalyse choices at asocial level – and hence to tackle policy questions. So far,market and non-market value havebeen described atthe privatelevel. Atthe sociallevel, which refers tovalue forsociety as awhole,the distinction between ‘market’and ‘ non-market’is still relevant, but socialvalue isdifferent fromprivate value andhas to be computedsomehow. Most economists would agree thatsocial values should be seen asthe aggregate ofprivatevalues. As afirst approximation,aggregation amounts to a simple sum;this means that the sumof privatemarket valuations, over all individuals in agiven society,yields social marketvalue, while the sumof privatenon-market valuations yields social non-marketvalue. In ourexample, the value oflanguage LX orofaparticular linguistic environmentdefined, amongother traits, by the statusof language LX in it,is therefore the sumof social market value andsocial non-market value. This is summarised in Table 1.

Table 1 Elements of value Private Social Market A C Non-market B D

A represents privatemarket value, B privatenon-market value, C social marketvalue, and D socialnon-market value. Total social value SV istherefore equal to C+D.Howshould this be computed?Generally, fora societywith N persons (1,2,. ..,i,. ..,N) whoseindividual market valuation of language LX or of agiven linguisticenvironment is mvi,andnon-market valuation is nmvi, total social value SV is given by:

N SV = mv + nmv å i i i =1

Asimple yet general decisionrule canbe derived fromthisway of formulating language policy choices:the policy thatought to be selected andimplemented, allother things being equal, isthe one whichmaximises SV, minus the corre- sponding costsof the policy,because thismeans adopting the policy fromwhich maximumnet welfare canbe expected. Thismay sound like statingthe obvious; unfortunately,macro-level language policy recommendationsare routinely madewith only the mosttenuous attempt at checking thatwelfare wouldindeed 38 Current Issues in Language Planning be increased.However, while thisprovides a general structurefor estimating value, several conceptual and empirical questions arise:

(1) First,simple summationcan be aninappropriate mode ofaggregation, because ofthe possibilityof externalities, as discussed earlier. If person h (a native

speaker oflanguage LY)learnslanguage LX,anticipatinga money return on thisinvestment, she increasesthe poolof speakerswho know LX. This will affect the situationof anotherperson, say j,in different ways.Suppose that j

alreadyknows language LX.Onthe one hand,the factthat h has learned LX increasesthe relevance of language LX andhence the value of j’s language skills.At the sametime, one more LX-speaking personon the labourmarket mayerode the wagepremiums accruingto LX speakers,and j may see his wagesituation deteriorate. The socialmarket value ofteachinglanguage LX tothose who do not speak ittherefore cannotbe computedas the sum of the

gainsthat each non-speaker standsto makeby learning LX,if only because, amongother reasons, (a) existingspeakers may gain or lose;(b) the potential

gainsto non-speakers of LX whodecide tolearn it will be affected by the numbers ofothernon-speakers whosubsequently decide tolearn it once

they observethat some fellow LY-speakers havegained fromthis move. At thistime, the problem ofaggregation (which istypically a network externalityone) doesnot seem tohave found asatisfactorysolution in the theoretical literature. (2) The componentsof non-market value arevery difficult toidentify theoretically, andno less difficult tomeasure empirically; furthermore, they arelikely alsoto be subject tothe networkexternality problem justdescribed in the caseof marketvalue. (3) Policyis about moving from an existing linguistic environment to another linguisticenvironment giving riseto higher welfare.This move entails benefits, but alsocosts. While someelements ofcost have been mentioned (in the formof lossesfor some members ofsociety), the precise amountof presumably straightforwarddirect costs, based on the expenditure required toachieve the benefits expected froma policy,is often very difficult toevaluate.

Itis therefore notpossible, at this time, to really computethe ‘value’of a language, andthis expression must be handled withcaution; unfortunately, not allcommentators do. It isequally delicateto estimate the ‘value’of a linguistic environmentas comparedto another. 44 Aconsiderableinvestment in researchon the identificationand measurement of the benefits (marketand non-market) that canbe expected froma particularpolicy andof most of the costs,direct and indi- rect,associated with such a policy move,now needs tobe made.This means that in practice,analysts need tosettlefor more modest goals, and to concentrate on parts of the problem of value. Evaluation work usually proceeds as follows: · networkor externalityeffects areignored, largely onthe groundsthat some arepositive while othersare negative, and are thus likely tocancel each other out to a significant extent; · the emphasisis placed onprivate market value, mainly by estimatingits mostimportant component, namely, the statisticalrelationship between language skills and wage rates; Language Planning and Economics 39

· these estimatesof (average) privatemarket benefits associatedwith language skillscan then be combined withaverage (per person) public spending forimparting those skills in orderto compute social market value; · non-marketvalues are kept outof the calculation,largely onthe groundsthat forlack of data, there ispractically no alternative (although some useful evaluationmethods could be importedfrom environmental economics). In practice,therefore, the questionof value isreduced toa subsetwhich essen - tiallycoincides with cell Ain Table 1.In the sectionof this paper onlanguage educationpolicy, we shalltherefore mostlyfocus on thissubset of the general problem of the value of language.

Language policy, diversity, and the ‘fundamental rule’ If one wishesto informpolicy choicesabout language withreference tothe conceptof ‘value’, itisimportant to steer clear of metaphors,and to parse the problem in orderto identify andmeasure components of value;yet wehavejust seen thatconceptual and empirical difficulties aresuch that a full estimationof the respective socialvalue ofalternative linguistic environments remains largely outof reach,forcing usto concentrateon the evaluationof privatemarket values. The general inference isthat an economicapproach cannot provide readily applicable recipes for solving language policy problems. Nonetheless,it offers the advantageof providing astructurewhich is often lacking frommany policy discussions,and one general theoreticalresult can be ventured regarding the relativevalue ofdifferent linguistic environments.In orderto present thispoint, let usassume that these linguisticenvironments are differentiated fromone anotherby the extent of diversity they accommodate.The applicationof basic economic concepts then suggeststhat society is likely tobe bestoff notwhen ittries to eliminate diversity,nor when itattempts to embrace limitless diversity. The argumentrests on a ‘fundamentalrule’ of policy choice(Stokey & Zeckhauser,1978), which requires selecting thatoption which promises to deliver the highest net benefit –thatis, ‘ benefit minuscost’ . Asageneral starting point,it is probably easyto agree thatdiversity, like mostthings, carries both advantages and drawbacks,which forsimplicity we maycall benefits andcosts. Itis understood that these arenot confined tomonetary ones and that non-marketelements, asamatterof soundprinciple, mustbe takeninto consid- eration.More diversitywill entailmore benefits and morecosts. However, bene- fitstend torise at a decreasing rate,while coststend toriseat an increasing rate.45 Thisimplies thatbenefit andcost curves will generally behave asin Figure 2.This means,in turn,that there isa pointwhere the difference between benefits and costsis largest, indicating the maximumlevel of net benefit. Thiscorresponds to an optimal level of diversity at d*, which is neither zero nor infinite. Though apparentlyinnocuous, the resultthat socially optimal diversity is likely to be positive and finite hasmajor political implications, because itindicates thatfrom an economic standpoint, policies striving to preserve orimpose linguistic homogeneity –or,in otherwords, ‘ zerodiversity’ – areill-advised, since they underestimatethe benefits andoverestimate the costsof diversity. 40 Current Issues in Language Planning

Figure 2 Optimal diversity

Conversely,generous calls(often motivatedby humanrights concerns) to embrace boundless linguisticdiversity and to set up policiesfor the integral recognitionof alllanguages in society,including thoseof immigrantgroups, howeversmall, tend tomakethe symmetricalmistake. These basicprinciples of scaleand optimisation are surprisingly under-used in economicanalysis, as pointed outsome 20 yearsago by Gilman(1983: 40). 46 Itis also worth noting that onthe basisof the very general assumptionsadopted here, one wouldbe led to questionthe widely held belief thatdiversity is negatively correlatedwith macro-economic welfare (for a considered discussion, see Nettle, 2000). Asa shortcutto the estimationof benefits andcosts, three strategiescan be adopted.The firstis simply toleave the comparisonprocedure tothe political debate,possibly after presenting, tothe extent possible,detailed information to votersabout the practicalconsequences ofthe variousoptions considered. In theory,preferences expressed throughthe votewill embody voters’valuation of marketand non-market benefits andcosts, and even if majorityvoting offers no solutionto the problem ofexternalities, particularly network externalities as discussedabove, the procedure atleast presents commendable democratic credentials. 47 Anotherstrategy, which yields aclearerpicture ofthe non-marketcompo- nentsof value involved,is theoretically possible, though, to my knowledge, untried asyet. Evaluationmethods derived fromenvironmental economics can be applied tolanguage choices(Grin, 1994c): a representativesample of residents canbe askedin asurvey how much they wouldbe willing topay (for example throughextra taxes, or asapercentage oftheir totaltax burden) forsociety to movefrom the currentlinguistic environment to another, presumably more desirable,linguistic environment which a proposedpolicy wouldaim at achieving. Conversely,respondents may be askedhow much they wouldbe willing to pay towardsa policy whoseaim would be to prevent apresumably undesirable change in their linguistic environmentfrom happening. An example ofthe first situationis people’ s willingness topay toachieve asubstantialincrease in average secondlanguage skillsin the population– implicitly defined asabenefit – perhaps througha generalisationof bilingual classesin the educationsystem. An example ofthe secondsituation is people’ s willingness topay tostop the decline Language Planning and Economics 41 ofthreatened language –whosefinal demise,implicitly defined asaloss,would become unavoidable in the absence of such a policy. Thirdly,estimates of the aggregatecost of a given language policy canbe presented tosurvey respondents,who are then askedif they considerthis cost acceptableor not. This implies thatcosts have first been estimated,which, most ofthe time,is not done. It is important to pointout that in caseswhere suchcalcu- lationshave been carriedout, the costsentailed by language policy often turnout tobe much lowerthan is commonly assumed (see ‘Fromunilingual tobilingual education’below) andfall withina range thata majorityof votersare probably quite willing to pay in order to secure the linguistic environment they desire.

On cost-effectiveness evaluation The relevance ofeconomicsto language planning lies chiefly in itscapacity to drawon a robustconceptual framework that helps guaranteelogical consistency,as well asthe general characterof analytical work. The latterproceeds froma constanteffort touse asfew explanatoryvariables as possible in orderto explain agiven phenomenon. Thisguards against the temptationto explain a situationby case-specific, ad hoc reasons– whichwould of course strip the expla- nationof anyrelevance asa(general) theoryor asanillustrationof atheory;it is noaccident that ‘ adhockery’is one ofthe mostscathing insults that economists canhurl atone another.The downsideof the approachis thatthe applicationof toolssuch as ‘ benefits’and ‘ costs’quickly runsinto daunting conceptual and methodologicalproblems –unless one isprepared touse these termsin amuch loosersense, with the unsurprising resultthat propositions formulated on their basismay not hold up tocloser scrutiny. Hence, economicscan be ofservicein language policy workand offer insightsand information that other approaches usually donotprovide, but only withinlimits. Accordingly, economic contribu- tionsover recent yearshave tended toemphasisevarious forms of policy evalua- tionusing alimitedrange ofwell-established concepts,such as ‘ effectiveness’ and ‘cost-effectiveness’. Policy evaluationis a field ofspecialisationin itsown right. It has its basis in politicalscience but itusually displaysan essentiallyeconomic concern for the effective use ofscarceresources (Dunn, 1994;Stokey &Zeckhauser,1978). In this framework,policy goalsare exogenously set,and will normallyhave been adoptedas aresultof politicaldebate. The roleof the policy analystwill then be to comparepossible coursesof actionfor the transpositionof political choices into a setof policy options.The comparisonmust be basedon informationregarding the relativeadvantages and drawbacks of each policy option,and recommenda- tionsregarding implementation.This focus on the selection,design andevalua- tionof policy optionscannot be exclusive, norcan it be completely insulated from political debate; moreover, it is not intended to replace it. 48 Despite these limitations,this type ofexercise will require alogicallyrigorous analysis,possibly referring toa policy-to-outcomepath linking the policy deci- sionat one end withits expected outcomesat the otherend, parsingthe wayin between intosuccessive cause-and-effect steps(Grin, 2003; Grin & Vaillancourt, 1999).This type ofpolicy evaluationexercise entailsthe followingstages, which canbest be explained withthe use ofan example. Letus therefore considerthe caseof apolicy whosegoal is to ensure the long-term maintenanceof aminority 42 Current Issues in Language Planning language LX,andwork backwards from the ultimatepolicy outcometo the policy measure: · First,policy outcomesmust be expressed in termsof the ultimategoals of a policy.In ourexample, one woulduse someindicator of language use or language vitality.Thisis obviously more relevant to evaluating success thaninformation about the directoutputs of specific measures,such as the number ofteachers undergoing trainingin orderto be able toteach the minoritylanguage in schools.While anincreasein the number ofqualified teachersshould per se be conduciveto the successof the policy,it is only an intermediate goal, and merely a means to an end. · Second, the desired policy outcomemust be modelled,even ifinformally, aspartof asetof causalrelationships. The policy outcomewill be castas a dependent or‘ explained’variable, while the policy measurewill be castas an independent or‘ explanatory’variable. Any contextualelements deemed relevantby the analystmay of coursebe included in the model,as long as suchinclusion is not confined tovague statementsto the effect that‘ the successof the measurewill alsodepend onfactors x, y, and z’,but formu- latesan explicit directionalityabout this dependence. Nomajor insight is gained bysimply mentioning the fairly obviousfact that reality is complex andthat the density ofinterconnectionsbetween processesis boundless (Pool, 1991a). · Third(assuming that the policy being evaluatedis concerned with language maintenanceas reflected in the extentof its use),these cause- and-effect relationshipsmust be grounded inarelevantrepresentation of language use whichshould, explicitly ornot,be anchoredin atheoryof behaviour.This will makeit possible toassessthe difference thata policy hasmade, and ensure thatcertain policy measuresare notcredited, say, for increasesin language use thatmight have occurred anyway, even in the absence of those measures. · Fourth,once this general frameworkis in place,the workingsof apolicy measurecan be followedin proper sequence allthe wayfrom its adoption by the authoritiesthrough its direct effects andultimately to its more distant effects on relevant indicators of language maintenance. Effectiveness canbe defined in different ways.Unfortunately, some of them obliteratethe crucialdistinction between ‘effectiveness’and ‘ cost-effectiveness’. Letus therefore adoptthe followingconvention: ‘ effectiveness’will simply mean ‘havingan effect’or ‘makinga difference’, while ‘cost-effectiveness’will imply thata given effect isachieved withthe lowest possibleuse ofresources or,alternatively,that given acertainuse ofresources, the best (highest)possible effect is achieved.In thissense, assessing effectiveness will require analyststo establish whether the implementationof aparticularpolicy measureshould be expected tohave(or has had, in the caseof ex-post evaluations)an effect onthe indicatorof language maintenance chosen at the outset. In orderto evaluate cost-effectiveness, it is of course necessary to evaluate costs.The questionof costevaluation would require afull-fledged paper ofits own,and the readeris referred tomaterialsquoted in thissection for details. Let us simply point out two essential principles. Language Planning and Economics 43

· First,cost must not be confused withexpenditure. Expenditure isan accountingconcept that attaches to the resourcesinvested in apolicy (for example, the salariesof the teachershired toteachthe minoritylanguage). Cost,by contrast,attaches to the outcomesof a policy (for example: ‘how much doesit costto increase by 5%the percentage oftheir leisure timethat school-agechildren spend using the minoritylanguage instead ofthe major- itylanguage?’ ). Moving fromexpenditure tocost figures requires approxi- mations(sometimes rough ones) in orderto apportionitems of expenditure tounits of outcome.Though often intellectually frustrating,this exercise cannotbe side-stepped, given, precisely, the very complexityof the causal relationships involved. · Second, whatmatters is not total cost, but the difference between the cost societyincurs if agiven policy isimplemented, andthe costit would have incurred if the policy hadnot been implemented. One classicalexample is thatof potentiallydamaging tensions or even conflictthat may arise between amajorityand a minorityif the latteris denied recognition– possibly throughnegation of itslanguage. Grantingofficial status to the minoritylanguage maybe costly,but these costsshould be weighed againstthose thatwould have occurred in caseof worseninginter-group conflict.This problem canbe defined asthat of identifying the proper ‘counterfactual’. 49 Once anevaluationof costhas been produced,it becomes possible toestimate, foreach proposed policy, an indicator of cost-effectiveness. This can be done either by dividing totaloutcome by totalcost (which yields anindex ofwhatcan be achieved per euro,mark, franc or dollar spent) orby dividing totalcost by total outcome(which yields anindex ofthe costto be facedin orderto obtain one unit of output). The procedure can,in theory,be applied acrosscompletely different policy areas,as long asthey havea commongoal that can be expressed in termsof a commonunit ofmeasurement (Grin &Vaillancourt,1999). Such generalised comparisons,however, may be replaced by morecircumscribed procedures within policy domains,such as education or the media,thereby reducing the need toresortto a stringof assumptions in orderto expressthe outcomesof very different types of intervention along the same scale. Before closingthis section, it seems necessary to stress once more that the instrumentsthat economic analysis can provide arejust that – instruments– and thatthey cannotserve asasubstituteto the inputs andinsights offered by other approaches.The specific contributionof economicsoften operatesby recasting language problemsin anew light,and redefining themin termsof alternatives whoseimplications must be spelled out,and between whichchoices have to be made.The makingof suchchoices, however, must remain a politicalprocess, which policy analysis is not intended to replace. 50

An Application: Language Education Policy Educationplays a key partin language policies,because itis the single most importantvehicle oflanguage policy.This is why thisclosing section, devoted to anempiricalapplication of some of the toolspresented earlier,focuses on educa- tionalissues. The following pointswill be examined:the estimationof private 44 Current Issues in Language Planning returnson investmentin second(or foreign) language learning; the estimationof socialrates of return onlanguage teaching;and the bilingualisationof education systems.

The market value of language skills

Private benefits Forthe sakeof brevity, thisexample focuseson second or foreign language skills(two notions between whichno major distinction needs tobe madefrom the standpointof economicanalysis). Ultimately, the analyticalprinciples arethe same for all language skills,including proficiency inone’s firstor mainlanguage (usually the mothertongue), but in the courseof anaccount concerning first languagesas well, interpretationand commentary would need tobe expanded accordingly and considerably lengthen the discussion. Letus therefore considerthe commonsituation where the level oflanguage skillsacquired by actorsis, to alargeextent, the outcomeof language education policy.This requires abrief excursusin the field of educationeconomics , which will be confined tosome essentials; for a textbooktreatment, see e.g. Lemelin (1998), and for an extensive survey of its subfields, Psacharopoulos (1987). Educationeconomics as a recognisedfield of specialisationin economics emerged in the early1960s. One ofits main pillars is human capital theory, whose thrustis as follows.An actorwith specific skillswill tend tobe moreproductive thananother without those skills. Since wagerates generally reflect productivity, amoreskilled personwill tend toearn more. If skillsare acquired through the educationprocess, education yields benefits represented by wagepremiums. Even if educationis costly,both for individuals and for society, it maytherefore be seen asaninvestmentwhose rate of return canbe estimated.Typically, most ofthe investmenttakes place in the early years,while the returnson the invest- mentappear during workinglife. Letus compare two education scenarios: the ‘long’(A) and‘ short’(B) ones,which are associated with different earnings levels. These scenarios,which start diverging forindividuals reaching the age of 15, are represented with age-earnings profiles as shown in Figure 3. AreaFE standsfor ‘ forgone earnings’, while areaSE standsfor ‘ schooling expenditure’(by the individual orthe parents),and surface GH represents the gainsresulting fromlonger education.At first sight, the investmentis profitable if GH > FE + SE.However,time must be takeninto account. Let us assume, for simplicity,that we areconsidering the choicesof ayoung personcontemplating either scenario,and that all the extraexpenditure entailed by choosinglong studies wouldtake place in the current period;the currentvalues of FEandSE therefore represent costsreasonably accurately. However, the gainswill only appear oversuccessive years in thatperson’ s workinglife. Thismeans that the present value of future earnings needs to be compared with the sum of FE and SE. Assumethat working life stretchesover 40 years,from 25 to65. Let us define the earningsprofiles asAandB respectively, stretchingover those 40 successive years (1,2,. ..,j,. ..40). The rate ofreturn onthe investmentis the value i which verifies the equation: Language Planning and Economics 45

Figure 3 Age earnings profiles

40 A - B j j - FE - SE = 0 å j j =1 (1+ r)

Hence, calculatingprivate market value comesdown to estimatingthe value of r in the aboveequation, using appropriatedata and standard econometric tools.A similarlogic can be applied when comparingthe profiles ofactors with or without secondlanguage skills,yielding estimatesof benefits (which takethe form of language-based earnings differentials), usually in percentage terms. Private investment costs Evaluatingthe costsof investment in secondlanguages raises a number of difficulties. Costsborne by the individual acquiringan educationinclude twoparts: direct expenditure onbooks, tuition, etc., and forgone earnings.For our purposes, these costscan be omittedfor the followingreasons. In mostcountries where educationis publicly provided andlanguages taught as partof thiseducation, students’direct expenditure onschool materials, though notnil, isrelatively minorand can be assumedaway (if itwere takeninto account, it would not result in amajordifference in the estimatedrates of return). Asregards forgone earnings,they tend towardszero for learners under legal workingage, because they wouldnot be allowedto sell, on the labourmarket, time not spent atschool. Furthermore,even beyond the legal minimumworking age, it would be imprac- ticalto trade for a wagethe amountof time specifically takenaway from language classesscattered over the weekly schedule. Hence, FEandSE canbe 46 Current Issues in Language Planning assumedto tend towardszero. Finally, there arealso other elements ofprivate expenditure, suchas fees paidfor evening classesas part of adult education. Usually,however, adult or continuing educationis a privatelymade decision, nota state-imposedpolicy, and hence notpart of the costsof apublic education policy. The maincomponent of expenditure istherefore madeup ofstatespending on language teaching.These costs,however, are borne indirectly throughtaxation, which meansthat they areunrelated toayoung learner’s actualconsumption of language education.They aretherefore notrelevant elements of private invest- mentcost, and treatment of thispart of costis postponed to the examinationof social rates of return below. Private earnings differentials One ofthe moststrongly established models for the selectionand design of language educationpolicies isderived fromthe combinationof the language economicsperspective withhuman capital theory developed in education economics.The fundamentalset of conceptsused in thiscombination has been described in the preceding paragraphs.I nowturn to the applicationof this modelto the statisticalestimation of whatis often referred toas the ‘privaterates of return’ on second language skills. The term‘ rateof return’, in thiscontext, is not fully appropriate.As explained above,the conceptof rateof return presupposes thathuman capital istreatedas aninvestmententailing (mostly)current expenditure in orderto generate future benefits. In otherwords, the passageof timeshould be explicitly takeninto account.However, the overwhelming majorityof existing statis- ticalworkonthe private valueofsecondlanguage skillseschews the questionof time.Typically, informationwill begathered onthe current value ofarange of variablesfor individual observations,and these valuesrelated to one another throughmultivariate analysis; hence, itis more appropriate to speak about‘ earnings differentials’. Hence, mostof the literaturein thisarea therefore presents differentials, not rates of return. Theexample usedbelow isthatofEnglish asaforeign language inthecase ofSwitzerland.The reasonfor this choice is that it is,at thistime, the only one forwhich datamake it possible toassess the value ofEnglish asan internationallanguage, thatis, in acountrywhere itis neither anofficialnor a demolinguisticallydominant language. Atthe sametime, the dataset used includes unusually detailedinformation about the levels andtypes oflanguage skills.The datacover a representativesample of 2400 observations over three of Switzerland’s four language regions.A specific methodologywas developed in thissurvey forthe collectionof language competence variablesthrough tele- phone interviews,in orderto ensure the stabilityand comparability of the infor- mationgathered. All thisprobably makesthis data base one ofthe richest internationally in this field. It contains questions on the following items: · respondents’L2 skills, differentiating between the ‘four skills’(under- standing,speaking, reading andwriting) andwithin each skill, four skill levels (fluent, good, basic, none); · non-school channels of L2 acquisition; · L2 use on the workplace; Language Planning and Economics 47

· standardsocio-economic characteristics (education, etc.), including labour income. The computationof averagelabour income by level ofEnglish language skills revealsa strongpositive correlation. However, English-language skillsare also correlatedto otherdeterminants of income, particularly education, and higher earningsaccruing to thosewho speak English maysimply reflect the factthat, havinga higher education,they canhold better-paying jobs.Hence, wagepremi- umsmay have nothing todowithcompetence in English.The sameproblem can be illustratedby the followingexample: better-educated people whoearn more aremore likely tohavehad more Latin at school than others, but one wouldnot necessarilyconclude thatthey arerewarded for some (residual) abilityto translate Cicero. In orderto circumventthis problem, the classicalsolution is to use astatistical instrumentcalled ‘ ordinaryleast squares regression’ (OLS forshort). OLS meth- odologywill notbe presented here, andreaders interested in explanationson technicaltreatment are referred tosuccessivestudies on Quebec andCanada by Vaillancourt(in particular1988, 1991)and, as regardsthe dataused here, toGrin (1999c).OLS areparticularlyuseful toseparatethe effect ofvariousdeterminants of income. Typically, an equation with the following form will be estimated:

2 lnY = a + b1 E + b 2 X + b 3 X + b 4 L + b 5 F + e where lnY standsfor the logarithmof labourincome, E foreducational level (measured inyears), X forexperience (alsomeasured in years), L forsome indicatorof language skills,and F (depending onthe dataavailable) for other factorsconsidered relevant in thedeterminationof labourincome (for example, arespondent’s type ofemployment). Finally, e isa randomterm whose expected value iszero. Using, in theexample below, notjust English language skills,but alsoeducation and experience asregressors, we obtainthe setof resultspresented inTable 2.In thistable, the coefficient foreachlevel ofcompetence in English reports,in percentage points,by howmuch the earningsof an individual displaying thatlevel ofcompetence will exceed thoseof someone whohas no competence in English, but hasequivalent educationand work experience. 51 The resultsin Table 2indicatethat English language skillsare highly rewardedon the Swisslabour market. Even controllingfor education and experi-

Table 2 Net earnings differentials (%) forEnglish language skills Switzerland 1994/ 95 (language coefficients only)

Men (n = 1141) Women (n = 803) Fluent 24.09 25.19 Good 18.03 39.52 Basic 8.93 18.09

Adj. R2 0.360 0.095

Source: Grin (1999c: 164). All reported coefficients are significant at the 99% level 48 Current Issues in Language Planning ence asisdone here, the premiums clearlyrise along with the level ofcompetence in English.Extended analysiswith additional control variables confirms the econometricrobustness of the results;it also shows, however, that reality is more complexthan these simple figures suggest.First, there aresignificant differences between Switzerland’s language regions.When the analysisis carried out for the three regionsseparately (German, French andItalian-speaking) much higher ratesof return forcompetence in English appearin German-speaking Switzer- land.In French-speaking Switzerland,by contrast,knowledge ofGermanas a secondlanguage ismore highly rewardedthan knowledge ofEnglish.Second, returnsare sector-dependent: in someeconomic sectors (typically, thosethat displaya strongorientation towards international trade), English ishighly rewarded; in other sectors, the earnings differentials are low. Estimationsof earnings differentials accordingto foreign orsecond language skillsconstitute relevant information for individuals who contemplate learning anotherlanguage, but froma language policy perspective, socialrates of return (see below) wouldin principle mattermore. However, social rates of return on language teachingare usually notavailable, and private earnings differentials alreadyprovide ample materialfor public debate.The type ofuse towhichthese figures canbe put will,however, be case-specific.In the Swisscontext discussed here, these figures arequoted in the ongoing debate onthe relativerelevance of teachingnational languages (as opposed to English) aspartof the curriculum (see e.g. Lüdi, 1998; Watts & Murray,2001), because they showthat the return on teachingthem is high (and,in the caseof French- andItalian-speaking Switzer- land,that knowledge ofanothernational language, German,is more financially profitable than knowledge of English). If comparabledata were availablefor other European countries,it would be possible tocarry out similar calculations and to assess whether the currenttrend towardsincreased investment in the teachingof English andthe concomitant neglect ofotherlanguages amounts to a financiallywell-advised attitude or not. Such resultswould be particularlyinteresting forBritain and Ireland, where investmentin foreign language learning isreputedly very low(Graddol, 1997), andwhere foreign languagesare no longer partof the subjects required for‘ A levels’at the end of secondaryeducation. In the caseof Qué bec, estimatesof (private) language-based earningsdifferentials confirmthe value ofEnglish asa secondlanguage, andhave major political significance astoolsfor monitoring overtime the evolutionof the respective socio-economicstatus of anglophones andfrancophones – withor without knowledge ofthe otherlanguage (Vaillancourt,1996). If figures onprovincialspending forthe teachingof English were combined withthis information, in orderto yield estimatesof socialrates of return,it would become possible toaddress the issueof social over- or under-investment in the teachingof English –or,for that matter, of other languages like German and Spanish.

Social rates of return on foreign language teaching The estimationof socialrates of return tendsto be technicallymore demanding, because itdoes take the timedimension into account through actualisation, andestimates therefore do deserve the label of‘rateof return’ . Furthermore,data Language Planning and Economics 49 ongovernment expenditure forlanguage teachingalso need tobe gathered and processed. However,data on this component of education spending arepractically non-existent,or at least extremely rare,because currenteducation accounting practicesgenerally fall shortof ‘analyticalaccounting’ , andhence donot yield figures onexpenditure by subject (OECD, 1998).The costsof specific school programmesoffered throughthe medium ofaspecific language haveoccasion- allybeen estimated(e.g. Patrinos& Velez, 1996);to my knowledge, however,the only example ofsuch cost figures havingbeen systematicallyderived from aggregatespending figures isa studyon language teachingin Switzerland(Grin &Sfreddo, 1997).They suggesta totalspending of CHF1500(approximately 1000or US$ 1000)per studentand per yearfor the teachingof all second languages.On average,10% of totaleducation spending isdevoted to secondlanguage teaching.This figure excludes post-secondaryeducation. It can never- thelessbe used asareference point,taking account of the factthat it reflects an educationsystem in whichlearners in shortstreams typically learn one foreign language forthree years,while learnersin long streamslearn one foreign language forseven yearsandanother for four years. 52 Thisfigure of10%isproba- bly notmarkedly different fromshares observable in otherWestern European countries,so thata range of5%to15%oftotaleducation spending canbe seen as anacceptable a prioriapproximation of public spending onsecond language teaching for those countries. The estimationprocedure isthe following(for adetailedexplanation, see Grin, 1999b, Chapter 9): · twodistinct age-earnings profiles areestimated, for unilinguals andbilinguals respectively – or for some similar alternative; · respective figures for labour income in each period are estimated; · the difference between bothprofiles ateach period isobtained by subtract- ing the lower from the higher profile; · if resulting figures reflect (asmay be the casedepending onthe natureof the dataused) underlying monthlyearnings, they will be multiplied by 12to obtain yearly amounts; · anassumptionis made regarding the period oftimefor which estimations are carried out – usually, the outset of the typical agent’s working life; · future earningsdifferentials, estimated for one typicalagent, are discounted overthe timehorizon from the startingpoint of the period oftimeconsid- ered; · figures onper-capitaspending onthe teachingof the language concerned are then brought into the calculation; · earningsdifferentials onthe one hand,and per-capita spending onthe otherhand, are entered intoa rate-of-return equationas presented abovein the sub-section on ‘Private benefits’; · the discountrate which annuls the left-hand termof the equationrepre- sentsthe socialrate of return onthe teachingof the language concerned, given estimated earnings differentials and teaching expenditure. Aselectionof social rates of return onforeign language teachingin Switzerlandis 50 Current Issues in Language Planning

Table 3 Socialrates ofreturn onforeignlanguage teaching (%), men, Switzerland 1994a

L2 Language region French German Italian German 6.5 n.a. 21.5 French n.a. 10.0 11.7 English 4.7 12.6 n.s.

Source: Grin (1999c: 194) a:Expenditure figures for1993– 94 school year. Earnings information from late1994 (96% ofweighted sample) orearly 1995(remaining 4%).Control variables in underlying earnings equations: age,age squared, education(years). n.a.:not applicable.n.s.: not significant. Reported values are based on coefficients significant at the 99% level

Table 4 Social rates of return on foreign language teaching (%) Switzerland 1994 a

L2 Language region French German Italian German 7.4 n.a. n.s. French n.a. 7.8 n.s. English 6.4 13.1 n.s.

Source: Grin (1999c: 194) a:Expenditure figures for1993/ 94school year. Earnings information from late1994 (96% ofweighted sample) orearly 1995(remaining 4%).Control variables in underlying earnings equations: age,age squared, education(years). n.a.:not applicable.n.s.: not significant provided in Table 3.Only resultsfor the malepart of the sample,which are all statisticallyrobust, are reported in the table,because resultsfor women do not display the same degree of statistical reliability. 53 If one iswilling toadmit results for women, where one ofthe underlying coeffi- cientsis not statistically significant at the 99%(but where significance,for the coefficients concerned,is usually in the 80%to 95% range, making them worthy of attention),it becomes possible tocombine ratesfor men andfor women and to computea weighted averageof social rates of return forthe entire sample.The correspondingfigures arereported in Table 4. Socialrates of return forthe teachingof foreign languagestherefore vary, accordingto L2 andto language region,from 6% to13%.Clearly, the mostprofit- able ofthe socialinvestments is for English in German-speaking Switzerland. However,the teachingof English in French-speaking Switzerlandis a valuable investmentas well. All these figures comparefavourably with the averagerate of return onfinancial capital, since realrates of return (net ofinflation) on riskless assetsnormally lie in the 3–5% range, while the opportunitycost of capital conven- tionallyused in cost-benefit analysesis 10%. In the Swisscase at least, therefore, the teachingof foreign languagesis a very valuableinvestment, even onthe basisof Language Planning and Economics 51 strict market value. If non-market value were alsotaken into account (to reflect in particularthe politicaland cultural reasons that may be calledupon forteaching foreign languages), the social rates of return would, of course, be higher. Figures areunlikely tobe markedlydifferent in otherWestern European countries.However, it would be hazardousto attempt to generalise directlyfrom the Swisscase. In orderto gainknowledge aboutthe ratesof return onpublic investmentin language teachingin othercountries, it is essential to gatherappro- priate data. Twolimitations of the socialrate-of-return approachmust be recalledhere. The firstone isthat it combines information on benefits currently observedwith informationon currentcost. In reality,however, current cost is incurred in order togenerate future benefits. Even if the estimationprocedure implies aprojection ofbenefits intofuture periods(and then the actualisationof benefits), the causal relationshipbetween costand benefit thatthe calculationimplicitly assumes amountsto anactof faithin the factthat benefits will remainin the samerange in the future. Thisassumption is acceptable for planning overrelatively short hori- zons(certainly up tofive years,possibly a littlemore), since the marketvalue of foreign language skillsis unlikely tochange markedlyover a relativelyshort period.However, longer-term assumptionsabout the value ofspecific skills (fromforeign languagesto mathematics or computerliteracy) can only be tentative. Similaracts of faith are routinely madein the well-established field of educationeconomics for the estimationsof returns on education as a whole;unfortunately,it is far from certain that significantly better-performing strategiesexist. First,analytical refinements aimingto replace currentfigures by hypotheticalfuture figures mayentail considerable effort forprecious little gain in reliability.Second, incomplete asit is, the type ofknowledge providedby rate-of-returncalculations remainsbetter thantotal ignorance – orthan the purely arbitrarypronouncements thatmay otherwise replace themin the mediaand political space. The secondlimitation has already been mentioned,but itbears repeating. Even if the problem oftimediscrepancy were magicallysolved thanks to perfect foresight,social rates of return,based as they areon privatemarket benefits , can only give partof the picture. The informationthey provide mustbe seen in the broadercontext of politicaldebate oflanguage policies,in whichnon-market values(usually under the label of‘culture’, ‘history’or ‘ identity’) havea leading role to play.

From unilingual to bilingual education The issueof minoritylanguage revitalisationhas in recent yearsbeen receiv- ing increasingattention in sociolinguisticdebate, largely asaresultof agrowing concernfor minority rights, language rightsand linguistic human rights (Kontra et al.,1999;May, 2001; Phillipson, 2000;Skutnabb-Kangas &Phillipson, 1994). The literatureextends to law (e.g. de Varennes, 1996;Henrard,2000) andpolitical philosophy (e.g. Kymlicka,1995). Beyond the considerablevariety of moreor lesssuccessful experiences todate with revitalisation efforts, specialists’ assess- mentof their feasibility rangesfrom the reserved (e.g. Edwards,1994) to the cautiously optimistic (e.g. Fishman, 1991). Generally, whether basedon linguistics,sociology, law or political philoso- 52 Current Issues in Language Planning phy, the literaturestresses either practicalfeasibility ordesirabilityin termsof certainlegal, moral or political standards. Economic analysis shifts attention to another plane. Asregards (legal) ‘standards’(e.g., forminority protection), economists tend tolookat themas afundamentally extra-economicconcern, because they are,in essence, normative.Economists generally view normsas acontextwithin which agents’behaviour may be studied,but the studyof norms themselves would then fall withinthe purview ofothersocial scientists, or oflawyers.The economic approachto ‘ rights’as areflection ofnorms, however, starts out by notingthat rightsare social and political constructs. Hence, there isnosuch thing asan abso- lute right.Putting itdifferently, the weakmay availthemselves of a‘right’only to the extent thatthe stronghave an interestin grantingand upholding thisright (Simonnot, 1998). Asregards feasibility, economistswill usually contendthat it cannot be the issue:provided enough resourcesare invested intoit, there isno reason why language revitalisationshould be radicallyimpossible. The problem isthat it maybe anextraordinarilyexpensive goalto achieve. Hence, the realquestion, froman economic standpoint, is whether the outcomeis worth the resources devotedto achieving it.This, of course,brings usback to the problem ofadvan- tagesand drawbacks, their identification,and their measurement.It is no surprisethat in politicaldebate, attacks on revitalisationpolicies often criticise themas anegregious wasteof resources(on Welsh,see e.g. Davies,quoted by May, 2001: 266; on Scottish Gaelic, see McLeod, 2001). Hence, anallocativeargument (that is, a casemade in termsof efficient allocationof resources)in favourof minoritylanguages must logically offer proofthat resourcesspent onminoritylanguages – oron‘diversity’, which isa consequence ofminoritylanguage preservation– arewell-spent. Thismeans, as aminimum, that the benefits mustbe higher thanthe costs;applying amoredemanding stan- dardof proof,the net difference between benefits andcostsmust be higher under the policy pursued (which aimsat revitalisation through a given setof measures) than under any alternative policy. Asshown in the sectionsof thispaper on‘ The Economicsof Language Policy’ and‘ An Application:Language EducationPolicy’ , difficulties ofidentification andmeasurement make benefit evaluationparticularly speculative: benefits ultimatelyrest on individualvaluations of different linguistic environments,and these valuationsmay be extremely different; andthe presence of network externalityeffects canturn the aggregationof individualvaluations into an intractableproblem. Generally, however,it stands to reason that a positive difference between benefits andcosts is more likely toarise if benefits arehigh, or if costsare low. In thiscase, economic analysis can be used todevise ashortcut towardsinformed decisionmaking, which will againbe illustratedby the caseof language revitalisation policy. Letus start out from a currentlinguistic environment in whichthe survivalof minoritylanguage LX isseriously impaired. Suppose nowthat the authoritiesare contemplatinga policy forthe long-term maintenanceof LX.If successfully implemented, thispolicy wouldyield another,presumably morediverse, linguisticenvironment characterised by the continuing presence of LX. For social actors,both linguistic environments have a certainvalue, made up ofthe usual Language Planning and Economics 53 marketand non-market components. Even if thisvalue cannotbe expressed in euros,rupees ordollars,actors will generally be able tosay if they prefer one or the other,or if they areabsolutely indifferent between them.Against this background,however, actors can be asked howmuch theywould be willingto pay to movefrom the currentto the morediverse linguistic environment.Alternatively, anestimationof the costof doing socanbe computed,and people (the general public, orvoters in areferendum, orparliamentarians,as the casemay be) canbe askedif they considerthe amountjustified ornot. In thiscase, the whole problem ofbenefit evaluationis simply sidestepped, andcosts alone move to the fore- ground,giving costevaluation a majorstrategicimportance in policy debates.As itturns out, estimates of language policy costsare quite low– much lower,at any rate,than many commentators appear to assume, usually onthe basisof little evidence. Toclinch thispoint, I shallconfine myself toone example, namely,that ofmoving from a unilingual educationsystem (dominant language only: LY) to a bilingual educationsystem ( LX+LY)in whichboth languages are used asa medium of instruction. Asageneral startingpoint, let usremember thatstates have a general respon- sibility toprovide andfinance compulsoryeducation. Given thisresponsibility, there isa certain cost attachingto it. Hence, the realcost of bilingual educationis the costit entails overand above the counterfactual– thatis, a unilingual education system.What deserves inclusion as arelevantcost figure constitutesa potentially complexand still little-explored domainof inquiry, andfor the sakeof brevity, I shall move on directly to some simple figures. Calculationshave been madein the caseof the teachingthrough Basque in the Basquecountry in Spain (Grin &Vaillancourt,1999), or through indigenous languages(particularly Maya) in Guatemala(Patrinos & Velez, 1996).These independently produced studiesarrive at very closeestimates, all in the 4%to5% range.Such figures probably represent upper-bound estimates,because ofthe evolution,in the long run,of some cost components, which havemainly transi- toryrelevance. In the caseof the Basquecountry, the figures for1997 (later converted into euros) are as follows:

Training for Basque-medium teachers: 20.947m Production of Basque teaching materials: 0.950m Institutional overhead: 1.160m Total 23.0570m

Given thatin the sameyear, some 180,000 pupils were schooledin whollyor partlyBasque-medium streams,this amounts to approximately 133per pupil andper year. 54 Thisis an amountthat a majorityof residentsof the Basquecoun- tryare presumably quite willing topay throughtheir taxes– withmany residents possiblyprepared topay more.One shouldalso point out that the largestitem in the abovesum (training for Basque-medium teachers)is one thatshould be considereda temporarycost. The reasonis that in the long run,there isno reason why itshould cost significantly more to train teachers to teach through the medium ofBasquerather than Spanish. Therefore, suchcosts cease to be relevant 54 Current Issues in Language Planning elements in the calculation,and the long-term costof bilingualisation, consequently, is even lower than the figure just estimated. Further developments onsuch questions are possible, but they largely remainat the theoreticalstage (Vaillancourt & Grin,2000). For example, in certaincontexts, moving from a unilingual toabilingual schoolsystem means thatpupils will be able toget educationin alanguage thatthey understand well, insteadof alanguage thatthey understandpoorly. This has the following effects: · adecline in the repetition rate(children takingthe sameclass twice because of failing grades), which entails a reduction in costs; · adecline in the dropoutrate (children leaving the systembecause offailing grades), which entails an increase in costs; · better resultsin termsof cognitiveacquisition, entailing higher productivity and ultimately a more prosperous economy and higher tax revenue. Forlack of data,it is not possible, at this time, to estimate these effects. However,simulations can be carriedout. They consistentlysuggest, even with conservativeassumptions, that the benefits (lower costsand higher productivity) largely offsetthe costsof lowerdropout rates plus the extraexpenditure. In otherwords, the decisionto move from a unilingual toa bilingual education systemvery probably paysfor itself, even when non-marketbenefits arenot brought into the calculation. More generally, availableevidence indicatesthat the costsof language policy, acrosspolicy domains,are quite modest.The costof Québec’ s language policy overthe 1974–1984 period, using minimumand maximum estimates and including notonly directexpenditure but alsothe value ofoutputloss, ranges from C$2.770mto C$4.785m.This is from 0.28% to 0.48%of provincialGDP, thatis, lessthan half apercentage point(Vaillancourt, 1992: 212). As to the costof Canada’s federal policy ofofficial bilingualism, anestimateby the Office ofthe Commissionerof Official Languages(1991: 23) standsat C$25.8per residentper year. Thisconclusion can only be atemporaryone, since aconsiderableamount of theoreticaland empirical work remains to be done in orderto increase our knowledge ofthe costsof minoritylanguage promotion;even morework remainsto evaluate benefits. Yeton the basisof available evidence, policies aimingat the protectionand promotion of threatenedlanguages appear to be well worth their cost.

Concluding Remarks In thisessay, I haveattempted to provide acomplete overviewof economic approachesto language andlanguage planning. Followingthe introductionand adiscussionin the secondsection of the reasonsthat make the endeavouruseful, Ihaveexamined someepistemological and methodological issues in the third section,‘ Language:Too Much Alive forthe DismalScience?’ . ‘In The Economics ofLanguage:History and Main Strands of Research’, anextensive review ofthe literaturehas been presented. The fifth section,‘ The Economicsof Language Policy’, hasfocused on the wayin whichlanguage policy issuescan be handled in economicperspective. ‘An Application:Language EducationPolicy’ has presented Language Planning and Economics 55 in moredetail one ofthe mostimportant areas of studyin language economics, namely,the returnson investment in language learning, andlinked thisup with language policy questions. Thispanorama highlights the strengthsand limitations of economicsas an analyticalvehicle toapproachlanguage matters.The strengthslie in economics’ capacityto apply arobustanalytical methodology to the questionsat hand, stressinglogical consistency and anchoring language processes(particularly at the ‘status’rather than ‘ corpus’level) in well-structuredtheories of agency and public policy.These aspects,as well aseconomists’ relative familiarity with problemsof evaluation, make economics a very useful toolin language policy analysis.The limitations,largely derived fromthe methodologyitself, have to do withthe factthat many qualitative aspects, which are highly relevantto the study oflanguage issues,are difficult tocombine withthe fundamentally quantitative orientationthat underpins mosteconomic analysis, in particularits chief tool, modelling. In otherwords, it is not easy to make the subject matter(namely, the reciprocallinks between language andeconomics)amenable toeconomic analysis. The general conclusion,therefore, shouldbe anuanced one:the economicsof language canprovide useful insightson certain language-related issues(in particular,it can shed light onprocesses that other analytical approaches would largely leave unexplored), but itcertainly does not replace them.Clearly, there is complementarity between the contributionsthat various disciplines canmake in the studyof language processes,including thosein whichthe latterinfluence, or are influenced by, economic processes. The economicapproach to language isalso useful in thatit helps usconsider criticallyseveral hasty analogies. We haveseen, forexample, thatthe rhetoricof ‘language asatreasure’, justlike itsconverse, ‘ linguisticdiversity as aburden’, generally donothold up tocloser scrutiny. The fallacyof the formeranalogy can be shownby recalling thatwhen somethingis indeed atreasure,social actors usually knowthis, and do not need tobe told.At the sametime, it would be wrongto conclude thatactors’ apparent disregard for diversity, as reflected in non-diversity-maintainingpatterns of behaviour, proves that diversity is not valuablein aneconomicsense. Given the ‘public good’character of language, decentralisedchoices by socialactors may fail toensure the sociallymost desirable level oflinguistic diversity.Moreover, language doesnot serve asimple communicationfunction, and non-market values, even in anessentiallyortho- dox economic analytical framework, must be taken into account. The studyof the links between language andeconomics is, at times,danger- ouslyexposed tooversimplifications. Some seem reasonableenough atfirst sight,but closerexamination reveals our ignorance of the specific processesat play.Consider for example the widespreadnotion that the factthat the United States(with allits economic weight), isEnglish-speaking ‘explains’the current dominanceof English. Economic weight certainlyplays a majorrole, but our understandingof the implied cause-and-effect relationshipsremains incomplete. Such relationshipscertainly bring intoplay aplurality offactors,which are alsologically interconnected; and these interconnectionsare not, at this time, fully identified, let aloneexplained, whether by economistsof language, socio- linguists,or otherspecialists. This is but one example, but ithelps toillustrate the 56 Current Issues in Language Planning need fora significantresearch effort todevelop ourunderstanding of issues aboutwhich Pool(1991a) once observed that people often hold‘ extraordinarily stubborn beliefs’. Asnoted in ‘The Economicsof Language: History and Main Strands of Research’section of thispaper, aswell asinotherpapers included in thisissue, manytopics deserve closerscrutiny. They range fromthe roleof language in tradeto the economicdimensions of translation and interpretation or the proper identificationof network externalities. One ofthe mostpressing issues,however, isthat of the distributiveimpacts of language policies,which are only beginning tobe explored (Durand, 2001;Pool, 1991c; Vaillancourt & Grin,2000; Van Parijs, 2001).Yet it is one withgrowing political relevance –towit, discontent over the treatmentof different languagesin multilingual structuressuch as the European Union.55 Generally, there isa risingneed foreconomic expertise in policy evaluation, whether onthe plane ofallocativeor onthat of distributive issues, as evidenced, forexample, by requestsfrom international organisations and governments. Thisraises the questionof whether academicinstitutions provide the context withinwhich this necessary research can be developed. Onthiscount, there area few promisingsigns, marginal as they maybe. Forexample, the economics professionis beginning tonotice the relevance oflanguage mattersas anobject of study.This is evidenced by the recent publicationof a readerin language economics,in aseriesdevoted to emerging areasof economics(Lamberton, 2002).More importantly,the intellectual climateis progressively becoming more favourableto taking interdisciplinarity seriously, also (modestly) in economics departments. Wecantherefore closethis paper ona noteof cautious optimism. There is sociallyuseful andintellectually challenging workto be done,and the frame- workconditions for embarking onfurther workon the manifoldlinks between economicand linguistic processes are progressively improving.It is important, however,not to losesight of one absolutenecessity of the endeavour:its spirit mustremain truly interdisciplinary,and strive for the combinationof tools, concepts and methods from a broad range of disciplines.

Correspondence Any correspondenceshould be directedto François Grin ETI, University of Genevaand Education Research Unit, 40, bd. du Pont.d’avre, CH-1211 Geneva, Switzerland.

Notes 1.The first version ofthis textwaswritten in early 2001,and the authorexpresses thanks toFranç ois Vaillancourt,Jean-Jacques Ducret and MonicaHeller forhelpful sugges- tions and comments. This revised version (2003)has alsobenefited from the papers written in the meantime byBruthiaux and Ozolins, which offercomment onthe origi- nal version and are included in this issue of CILP. The usual disclaimer applies. 2.The Commonwealth ofIndependent Statescomprises the former SovietUnion minus the three Baltic Republics. 3.The adjective ‘classical’is used, byand large, formost economists from the time of Adam Smith (whose Wealthof Nations waspublished in 1776)to the time ofLé on Walras,best known forhis ‘Principes d’é conomie politique’published in 1874.This Language Planning and Economics 57

includes Marx,who is generally regarded asa non-mainstream classicaleconomist. Walras’s writings, which mark the ‘marginalist revolution’and the overwhelming adoptionof the ‘utility theory ofvalue’, open the era ofmainstream, alsocalled ‘neoclassical’economics. Modern mainstream economics remains steeped in neo-classical orthodoxy, butit has regularly been expanded toaccommodate more complexaspects ofhuman behaviourwhile retaining the core tenets ofthe theory, such asthe notion thatsocial actors act rationally in order tomaximise their satisfaction.For anoverview ofthe development ofeconomic thought, the reader is referred toBlaug (1985),or more succinct treatments in Wolff (1993) or Simonnot (1998). 4.Interestingly, anon-line provider offree economic datacalls itself ‘the dismal scien- tist’ (see http://www.dismal.com). 5.This assumption does not imply awholesale rejection ofalternative philosophical principles, in particularthose thatwould inspire one tolive alife ofrenunciation. However, economists consider this philosophy tobenon-representative ofthe major- ity of humankind. 6.The term ‘agency’is used here in its standard sociologicalmeaning torefer tothe behaviourof actors enjoying acertain degree ofautonomyin their action.Interest- ingly, economists routinely use the term ‘agent’where sociologists talk of‘ actor’. However, in some specialist subfields ofeconomic theory, the ‘agent’acts on behalfof a ‘principal’, and his autonomy tends to be constrained as a result. 7.In later writings (e.g.Becker, 1981),Becker insisted thatthe relevance ofeconomics is not limited to human choices,but that it applies to any type ofchoice.I shall not enter this debate, which is beyond the scope of this paper. 8.Assigning behaviouralpatterns tosocial ‘ norms’, forexample, may involve rather reductionist assumptions aboutthe contents ofthese norms, the way in which they are perceived and possibly internalised bysocialactors, and the latters’reaction tothese norms; ultimately, the analyst will end upnot with atheory ofagency, butwith a collection of ad hoc accounts of behaviour for different specific cases. 9.This characterisation ofeconomic modelling, which applies almost systematically in micro-economics, may bequalified in the caseof macro-economics. For example, a national economy may berepresented bya set ofequations (which constitute the formalexpression of relationships between variables),and this set ofequations will be regarded asamodel even ifeach of them represents anaccounting identity rather than a causal link. 10.A particularlygood example, in the field oflanguage economics, is Lang’s elegant analysis (1986)of the effectof communicationcosts on wage rate differentials between speakers of different languages. 11.For example, his reference (2000:140) to an earlier finding (Lazear,1999) that immigrants who live in ‘ghettos with many others from their native land are much less likely tobe fluent in English than those who live in communities in which they are a smaller minority’is, atbest, unsurprising. Basic sociolinguistic research yields the same result, along with considerably more phenomenological substance. 12.For general introductions tothe economics oflanguage, see e.g. Grin (1994a,1996a, 1996b, 1999a); Vaillancourt (1985a); Grin and Vaillancourt (1997). 13.The first paperthat specifically focuseson this question, however, is anearlier one by Vaillancourt (1978). 14.Some commentators use, somewhat loosely, the term ‘ethnicity’as aquasi-synonym fora person’s linguistic ‘belongingness’. Insuch cases,it is not clearif language is ipso facto seen as a proxy for ethnicity, or the other way around. 15.Statistical work typically resorts tomultivariate analysis using ordinary least squares (OLS)regressions. The applicationof this instrument is discussed under ‘An Applica- tion: Language Education Policy’ in this paper. 16.See information on the ELP on the Councilof Europe website (http:// culture2.coe.int/portfolio/). 17.I amreferring here to external language dynamics, thatis, the position oflanguages vis-à-vis each other, not to patterns of internal language change. 18.Unless, ofcourse, the appleis cutup in smaller pieces; butthen different consumers 58 Current Issues in Language Planning

will not get the apple,but a small partof it; furthermore, it is difficultto see how a language couldbe cut up into small bits,for different bits tobe used bydifferent people.The reader should note, however, thatthe network effectis not unambigu- ously positive; see the sub-sectionon ‘Aworking definition oflanguage policy’in this paper. 19.Alternatively, firms may be said to buy the flow of services from capital. 20.Although the ownership ofcapital,as aproduction factor,is much more concentrated than the ownership of labour, households remain, ultimately, the owners of capital. 21.However, this effectis strong in the caseof English, noticeablein the caseof German, and non-significant in the caseof French (the same result holds forSpanish when the sample is broadened toinclude developing countries, asufficient number ofwhich are Spanish-speaking). There seems tobe no clearinterpretation ofthese findings, other than the obviousconclusion thatit is not the commonalityof languageper se that matters, butother variables which may becorrelated with language –or which may also,considering the relatively small number ofcountries considered, bepurely idiosyncratic. Helliwell (1999:16) also points outthat in the specificcase of Canada, language commonality through French does have apositive and significant impacton foreign trade patterns. 22.On the relative lackof attention tolanguage matters in the advertising literature, see Holden (1987) or Grin (1995). 23.On the relevance ofdevelopment asanareain which economics and language disciplines could fruitfully cooperate, see Bruthiaux (2000). 24.However, the question ofeconomies ofscale in the translation and distribution of booksis currently being investigated byMé litz (2000); the issue ofthe valueof language services is addressed by Ozolins (this issue). 25.Often in the caseof Quebec; see e.g. Breton (1964,1978); Breton and Mieszkowski (1977); Vaillancourt (1978). 26.Some economists might objectthat Pool’ s work is less partof ‘ economics’than of ‘rational choice’, afield ofspecialisation in politicalscience. However, this objection cannotbe sustained, since the boundaries between ‘rational choice’(particularly applications of it) and parts of ‘public economics’ are not very meaningful. 27.‘ Aggregate’in thatit sums upthe publicexpenditure oflocal,regional and national authorities. 28.Recent work byRubinstein (1999)is probablythe first instance ofthe same issue tobe taken upbyan economist, and one ofthe very few economic contributions thatcanbe seen asaddressing language corpus.Rubinstein proposes amathematicalanalysis of language seen as the product of an ‘optimiser’ developing a communication code. 29.Some commentators have noted thatmuch ofRossi-Landi’ s earlier exposécan be found,in remarkably similar wording, in Bourdieu, without any credit being given by the latter to the former. 30.The importance oflanguage skills may beapositive, butprobably not alinear, function ofthe monetary valueof exchange. While language plays amajor partin negotiat- ing afirst deal between buyer and seller, renewals ofa similar deal afterwards may require nothing butthe most elementary foreign language skills. Furthermore, language is likely tohave relatively greater importance when market exchange concerns goods and services whose abstractcomponents are more important–and need tobeformulated linguistically. Yet it is often scientifically perilous toventure generalisations aboutthis type ofgoods. All this suggests thata much closer analysis ofthe language ofvarious types ofmarket exchange should beundertaken, from business deals between companies to retail trade on the internet. 31.For example, Singapore’s ‘SpeakMandarin’ campaign which aimed atreplacing southern Chinese dialects by Mandarin (Chee, 1990; Gopinathan & Pakir, 1994). 32.A collection ofcasestudies recently edited byMackey (2000)provides agood descrip- tion oflinguistic environments in multilingual urbansettings, where these environments are jointly characterised by‘ source factors’grounded in history, politics, demolinguistics, and reflected in indicators oflanguage visibility, patterns of language use, etc. Language Planning and Economics 59

33.In this perspective, the resistance ofthe Macedonian government (summer 2001)to granting co-officialstatus to Albania illustrates its opinion thatthe nation’s welfare is higher ifMacedonian is its sole officiallanguage. Most Albanian-speaking citizens of Macedoniahold the oppositeview. Inpolicy analysis terms, it couldbe reformulated as a claim that aggregate welfare will be higher if both languages are official. 34.For example, one may have apersonal policy,for ideological reasons, todeliberately go against convention and not touse certain dominant languages in certain contexts. 35.The conceptualdistinction between ‘politics’and ‘policy’is not always clearly made, with the result thatthe literature pays much more attention tomatters thatshould be considered partof ‘political’, not ‘policy’analysis. On recent developments bridging language politics and language policy,see e.g. Ricento (2000).On the culturalembed- ding of language policies, see Schiffman (1996). 36.Third-sector organisations are not-for-profitand non-government suppliers ofvari- ouskinds ofservices, and caninclude volunteering activities organised bycivil society associations. 37.Readers will observe thatthe autonomyof actorsoperating on free markets is therefore viewed asastarting point and implicitly ‘normal’state of affairs,and thatstate intervention needs tobe justified becauseit consecrates adeparture from it.A reverse stand couldbe adopted,rejecting the liberal position fromthe outset ,in line forexample with Marxian theory. 38.For a detailed treatment, see Cornes and Sandler (1996). 39.Many commentators would mention ‘ethnic minorities’at this stage. For reasons explained in the concluding section, references toethnicity are deliberately avoided in this paper. 40.At the time ofwriting, this debateis gaining momentum in Germany and is no longer confined tonarrow academiccircles, with publiccalls to protect the German language from anglicisims (see WeltAm Sonntag ,11February 2001).France has adopted legisla- tion onthis some time ago(the well-known ‘Loi Toubon’of 4 August 1994)which regulates language use in the publicsphere (forregular updateson the debatein France, see http://www.langue-francaise.org). There is broadconsensus among professionals tothe effectthat English is particularlylikely to(adversely) affect linguistic environments, not so much through any kind ofdeleterious influence on major languages like German and French, butmainly becauseit displaces small languages, even in contexts where English is not,a priori, the most directly threaten- ing language ( TheEconomist ,6June 1998,quoting the founder ofthe Foundationfor Endangered Languages ). 41.Danish, Dutch,English, Finnish, French, German, Greek, Italian,Portuguese, Span- ish, and Swedish. This totalis due toincrease to20 with the current round ofEU enlargement. 42.Competitive markets require fourconditions: ‘atomicity’(meaning thatthere is such a high number ofbuyers and sellers thatno buyer orseller caninfluence the price at which the good or service is traded); freedom ofentry and exit (anyone canbuy or sell onthe market, which implies, in the caseof high fixed production costs,that anyone will have accessto enough capitalto finance the necessary investment); transparency (buyers and sellers are in aposition toknow –orinfer –other participants’willingness tobuy or sell); and homogeneity (the good orservice exchanged is fully defined, and each unit of a good or service, on its market, is identical with any other unit). 43.This interpretation is borne outby the observation thatprocesses oflanguage spread canusually betraced backto a typically non-market relationship between groups of actors,such asmilitary conquest; this point is acknowledged even byauthors most hostile to language policy intervention (e.g. Jones, 2000). 44.Although one might argue thatif the same errors and omissions are made in the evaluation of two linguistic environments, these errors and omissions on bothsides ofthe ledger will canceleach other out,making comparison paradoxicallysafer than isolated evaluation. 45.The economically minded reader will recognise the conceptsof decreasing marginal utility (fromconsumption theory) and rising marginal cost(fromproduction theory). 60 Current Issues in Language Planning

46.Of course, the benefit and costcurves do not necessarily behave asshown in Figure 2. Even under the assumption ofrising marginal costand decreasing marginal benefits, it is possible tomap a casewhere d* is zeroor very close tozero;alternatively, very high estimates ofbenefits together with very low estimates ofcosts will generally push d* ever further tothe right. Suchextreme representations, however, are much less plausible. 47.However, the results ofmajority voting cannotbe interpreted asnecessarily yielding the best solution in the sense ofeconomic theory, assuggested byArrow’s ‘impossibil- ity theorem’. 48.Policy is not the work of‘asingle rational and disinterested mind [who]analyses the situation,develops ...optimalpolicy options, chooses between them, embodies parts ofthem in the law,embodies others in administrative mechanisms, and closes the cycle bymonitoring results. [Rather,]the politicalfunction of policy analysis is to contribute to the continuous debate between coalitions . . . ’ (Wimberley, 2000). 49.The coststhat do not arise thanks tothe adoptionof a policy,instead ofbeing computedas elements ofthe costof the counterfactual,may becomputedas benefits of the policy. The net value of the policy remains unchanged. 50.It is often difficultfor this simple message tobe heard, despite detailed explanation (alsoin this paper)and constantepistemological precautions;to wit, the disagreement byone authorin this issue with my reference (in the section on ‘The Economics of Language Policy’in this paper)to the Macedonian situation.My experience ofvarious instances ofethnic or linguistic conflict(particularly over three years spent atthe European Centre forMinority Issues) suggests thatthe assessment ofthis type of conflictsituation canbenefit from aconceptualrecasting ofthe issue, including in policy analysis terms. Suchconceptual recasting, however, is merely aninstrument to help us think about the issue in a fresh way, not a ready-made recipe. 51.Technically, this estimation procedure yields results in log points; these have, however, been converted to percentage points in Table 2 for easier interpretation. 52.These figures do not include hours ofsecond language instruction atprimary level (usuallyin grades 4to6), since instruction starts afresh atsecondary school (which usually means grade 7),without formally banking onskills thatlearners have presumably had the opportunity to acquire in the preceding years. 53.Each figure is based onacombinationof two rates ofreturn estimated with twodiffer- ent earnings equations (one formen who ‘know’and one formen who ‘don’t know’ the foreign language considered, taking asacut-offpoint knowledge atthe ‘fluent’or ‘good’level on the one hand, atthe ‘basic’or ‘nil’level on the other hand). The rates reported in Table 3 are based on coefficients that are all significant at the 99% level. 54.Given anaverage expenditure in Spainof 2800per pupilper year, the extra costof having a bilingual education system can be estimated at approximately 4.75%. 55.A written question byEuropean MPB.Staes(10 January 2001) raises the question of linguistic discrimination in European institutions following aspateof jobadvertise- ments looking fornative speakers ofEnglish (http://europa.eu.int/eur-lex/pri/en/ oj/dat/2001/ce174/ce17420010619en02330233.pdf).

References Ammon, U.(1994)The present dominance ofEnglish in Europe, with anoutlook on possible solutions to the European language problem. Sociolinguistica 8, 1–14. Appel, R.and Muysken, P.(1987) LanguageContact and Bilingualism .London: Edward Arnold. Arcand, J.-L.(1996) Development economics and language: The earnest search fora mirage? International Journal of the Sociology of Language 121, 119–57. Australian Language and Literacy Council(ALLC) (1994) Speakingof Business. The Needs of Businessand Industry for Language Skills. Conference Proceedings. Canberra:National Board of Employment, Education and Training. Becker, G. (1957) The Economics of Discrimination . Chicago: Chicago University Press. Becker, G.(1976) TheEconomic Approach to Human Behavior .Chicago:Chicago University Press. Language Planning and Economics 61

Becker, G. (1981) A Treatise on the Family . Cambridge: Harvard University Press. Berset, A.,Weygold, S.-A.,Crevoisier, O.and Hainard, F. (2000) Main-d’Œ uvreÉ trangère et Diversitédes Compé tences [ForeignWorkforce and Diversity of Skills ].Paris:L’ Harmattan. Blaug, M.(1985) EconomicTheory in Retrospect (4thedn). Cambridge: Cambridge University Press. Bourdieu, P.(1982) CeQue ParlerVeut Dire.L’ É conomiedes É changesLinguistiques [Language and Symbolic Power ]. Paris: Fayard. Breton, A. (1964) The economics of nationalism. Journal of Political Economy 62, 376–86. Breton, A.(1978)Nationalism and language policies. CanadianJournal of Economics 11, 656–68. Breton, A. (ed.) (1999a) Exploring the Economics of Language . Ottawa: Canadian Heritage. Breton, A.(1999b)Une analyse économique de lalangue (An economic analysis of language).In A.Breton (ed.) EconomicApproaches to Language and Bilingualism . Ottawa: Canadian Heritage. Breton, A.and Mieszkowski, P.(1977)The economics ofbilingualism. In W.Oates(ed.) The Political Economy of Fiscal Federalism (pp. 261–73). Lexington: Lexington Books. Bruthiaux,P. (2000)Supping with the dismal scientists: Practicalinterdisciplinarity in language educationand development economics. Journalof Multili ngualand Multicultural Development 21, 269–91. Calvet,L.-J. (1996) LesPolitiqu esLinguisti ques [LanguagePolicies ].Paris:Presses Universitaires de France. Cancian,F. (1966)Maximiza tion asnorm, strategy and theory: Acomment on programmatic statements in economic anthropology. AmericanAnthropologist 68 (2) (Part 1), 465–70. Carr,J. (1985)Le bilinguisme auCanada: L’ usage consacre-t-ill’ anglais monopole naturel? [Bilingualism in Canada:Is the use ofthe English language anatural monopoly?]In F.Vaillancourt(ed.) Économieet Langue [Economicsand Language ] (pp. 27–37). Québec: Conseil de la langue française. Chee, T.S.(1990) Multilingualismin Singapore:Two Decadesof Development .Singapore: National University of Singapore. Chiswick, B.and Miller, P.(1985)Immigrant generation and income in Australia. Economic Record 61, 540–53. Chiswick, B.and Miller, P.(1995)The endogeneity between language and earnings: International analyses. Journal of Labor Economics 13, 246–88. Chiswick, B.and Miller, P.(2000)The complementarity oflanguage and other human capital: Immigrant earnings in Canada. RIIM Working Paper 00-08, Vancouver. Chiswick, B.and Miller, P.(2002)Immigrant earnings: Language skills, linguistic concentrations and the business cycle. Journal of Population Economics 15, 31–57. Chiswick, B.,Patrinos, H.A.and Hurst, M.(2000)Indigenous language skills and the labor market in adeveloping economy: Bolivia. EconomicDevelopment and Cultural Change 48, 349–67. Chiswick, B.and Repetto, G.(2001)Immigrant adjustment in Israel:The determinants of literacy and fluency in Hebrew and their effectson earnings. In S.Djajic (ed.) International Migration:Trends, Policies and Economic Impact (pp.204– 88). London: Routledge. Choi, E.K. (2002)Trade and the adoptionof auniversal language. InternationalReview of Economics and Finance 11, 265–75. Chorney, H.(1998)Valeur du bilinguisme pourles employeurs et rôle des manipulateurs de symboles dans de grandes entreprises exportatrices [Bilingualism in employee recruitment and the role ofsymbolic analysts in leading export-oriented firms]. InA. Breton (ed.) Explorerl’ ÉconomieLinguistique [EconomicApproaches to Language and Bilingualism ] (pp. 207–48). Ottawa: Patrimoine canadien. Christofides, L.and Swidinsky, R.(1998)Bilinguisme et revenu [Bilingualism and earnings]. In A.Breton (ed.) Explorerl’ É conomieLinguistique [EconomicApproaches to Language and Bilingualism ].(pp. 135–205). Ottawa: Patrimoine canadien. Church, J.and King, I.(1993)Bilingualism and network externalities. CanadianJournal of Economics 26, 337–45. Colomer, J. (1991) The utility of bilingualism. Rationality and Society 2, 310–34. 62 Current Issues in Language Planning

Colomer, J.(1996a)To translate ortolearn languages? An evaluationof social efficiency. International Journal of the Sociology of Language 121, 181–97. Colomer, J.(1996b) La Utilitatdel Bilingü isme. Una Propostade Plurilisme Lingü ístic a Catalunya,Espanya i Europa [TheUtility of Bilingualism.A Proposalfor Linguistic Pluralism in Catalonia, Spain and Europe ].Barcelona: Llibres a l’Abast. Cooper,R. (1989) LanguagePlanning andSocial Change .Cambridge: Cambridge University Press. Cornes, R.and Sandler, T.(1996) TheTheory of Externalities,Public Goodsand Club Goods (2nd edn). Cambridge: Cambridge University Press. Cremer, R.and Willes, M.(1991)Overcoming language barriers tointernational trade: A text-basedstudy ofthe language ofdeals. Journalof Asian-Pacific Communication 5, 147–61. Crystal, D. (1997) English as a Global Language . Cambridge: Cambridge University Press. Crystal, D. (2000) Language Death . Cambridge: Cambridge University Press. Dalmazzone,S. (1999)Economics oflanguage: Anetwork externalities approach.In A. Breton (ed.) Exploringthe Economics of Language (pp.63– 87). Ottawa: Canadian Heritage. Daly, H. (1977) Steady-State Economics . San Francisco: Freeman. Dávila, A. and Mora,M. (2000)English fluency ofrecent Hispanic immigrants tothe United Statesin 1980and 1990. EconomicDevelopm entand Cultural Change 48, 369–89. Dunn, W.(1994) Public PolicyAnalysis. An Introduction .Englewood Cliffs:Simon and Schuster. Durand, C.(2001) La Miseen Placedes Monopoles du Savoir [TheSetting up of Knowledge Monopolies]. Paris: L’Harmattan. Dustmann, C.(1994)Speaking fluency, writing fluency and earnings ofmigrants. Journal of Population Economics 7, 133–56. Dustmann, C.and van Soest,A. (2001)Language fluency and earnings: Estimation with misclassified language indicators. Review of Economics and Statistics 83, 663–74. Edwards, J. (1994) Multilingualism . London: Routledge. European CulturalFoundation (1998) Reportof the Conference ‘ WhichLanguages for Europe?’. N. Bos (ed.). Amsterdam: European Cultural Foundation. Fishman, J. (1991) Reversing Language Shift . Clevedon: Multilingual Matters. Fogel, W.(1966)The effectsof low educationalattainment onincomes: Acomparative study of selected ethnic groups. Journal of Human Resources 1, 22–40. Generalitat de Catalunya(ed.) (1997) Actesdel Congré s EuropeuSobre Planificació Lingü ística [Proceedingsof theEuropean Congress on LanguagePlanning ].Barcelona:Departament de Cultura. Generalitat de Catalunya(ed.) (1999) Polítiques Lingü ístiques a Països Plurilingü es [Language Policy in Multilingual Countries ]. Barcelona: Departament de Cultura. Gilman, R. (1983) From dismal science to joyful art. In Context 2, 28–48. Gopinathan,S. and Pakir,A. (1994) Language,Society and Education in Singapore .Singapore: Times Academic Press. Grenier, G.(1984)The effects oflanguage characteristicsonthe wagesof Hispanic-American males. Journal of Human Resources 19, 22–40. Grimes, B.(ed.)(1998) Ethnologue:Languages of the World (11thedn). Dallas:Summer Institute of Linguistics. Grin, F.(1990)The economic approachto minority languages. Journalof Multilingualand Multicultural Development 11, 153–73. Grin, F. (1992)Towards athreshold theory ofminority language survival. Kyklos 45,69–97. Reprinted in D.Lamberton (ed.)(2002) TheEconomics of Language (pp.49– 76). Cheltenham: Edward Elgar. Grin, F. (1994a)The economics oflanguage: Matchor mismatch? InternationalPolitical Science Review 15, 27–44. Grin, F. (1994b)The bilingual advertising decision. Journalof Multilingual and Multicultural Development 15, 269–92. Grin, F. (1994c)L’ identification des béné fices de l’amé nagement linguistique: Lalangue Language Planning and Economics 63

comme actifnaturel [Identifying the benefits oflanguage planning: Language asan environmental asset].In C.Phlipponneau and A.Boudreau (eds) Sociolinguistiqueet Aménagement des Langues [Sociolinguisticsand Language Planning ](pp.67– 101). Moncton, NB: Centre de recherche en linguistique appliquée. Grin, F.(1995)Plurilinguisme en matière de publicité[Plurilingualism in advertising]. In H.Goebl,P. Nelde, Z.Starýand W.Wölck (eds) Kontaktlinguistik (pp.438– 44). Berlin: Walter de Gruyter. Grin, F.(1996a)Economic approaches tolanguage and language planning: An introduction. International Journal of the Sociology of Language 121, 1–16. Grin, F.(1996b)The economics oflanguage: Survey, assessment and prospects. International Journal of the Sociology of Language 121, 17–44. Grin, F.(1997a)Diversité linguistique et théorie économique de lavaleur [Linguistic diversity and economic theory ofvalue]. In J.Hatem (ed.) Lieuxde l’ Intersubjectivité [Locus of Intersubjectivity ] (pp. 155–74). Paris: L’Harmattan. Grin, F.(1997b) Langueet Diffé rentiels de StatutSocio-é conomique en Suisse [Languageand Socio-economic Status Differentials in Switzerland ]. Berne: Federal Statistical Office. Grin, F.(1997c)Gé rer le plurilinguisme européen: Approche économique auproblè me de choix[Managing Europe’s plurilingualism: An economic perspective on the problem of choice]. Sociolinguistica 11, 1–15. Grin, F.(1999a)Economics. In J.Fishman (ed.) Handbookof Languageand Ethnic Identity (pp. 9–24). Oxford: Oxford University Press. Grin, F.(1999b)The notions ofsupply and demand in the economic analysis oflanguage. In A.Breton (ed.) New CanadianPerspectives. Exploring the Economics of Language (pp. 31–61). Ottawa: Canadian Heritage. Grin, F. (1999c) Compétences et Ré compenses. La Valeurdes Langues en Suisse [Skills and Rewards:The Value of Languages in Switzerland ].Fribourg: Éditions Universitaires de Fribourg. Grin, F. (2001) English as economic value: Facts and fallacies. World Englishes 20, 65–78. Grin, F.(2003) LanguagePolicy Evaluation and the European Charter for Regional or Minority Languages. London: Palgrave. Grin, F. and Rossiaud, J.(1999)Mondialisation, processus marchands et dynamique des langues [Globalisation,market processes and language dynamics]. InS.Abou(ed.) Uniformisationou Différenciation des Modè les Culturels [TowardsUnity or Difference in CulturalModels ](pp.113– 42). Paris: AUPELF-UREF (‘L’Actualitéscientifique’ ) and Beirut: Presses de l’Université Saint-Joseph. Grin, F.,Rossiaud, J.and Kaya,B. (2002)Langues de l’immigration et intégration professionnelle en Suisse [Immigrant languages and labourmarket integration in Switzerland]. In H.-R.Wicker, R.Fibbiand W.Haug(eds) LesMigrations et la Suisse [Switzerland and Migrations ] (pp. 404–33). Zürich: Seismo Verlag. Grin, F.and Sfreddo, C.(1997) Dépenses Publiques pour l’ Enseignementdes Langues Secondes en Suisse [Public Spendingfor Second Language Teaching in Switzerland ]. Aarau: CSRE-SKBF. Grin, F.and Sfreddo, C.(1998)Language-based earnings differentials on the Swiss labour market: Is Italian a liability? International Journal of Manpower 19 (7), 520–32. Grin, F.and Vaillancourt,F. (1997)The economics ofmultilingualism: Overview ofthe literature and analyticalframework. InW.Grabe(ed.) Multilingualismand Multilingual Communities (pp. 43–65). Cambridge, MA: Cambridge University Press. Grin, F.and Vaillancourt,F. (1999) Thecost-effectiveness evaluation of minority language policies:Case studies on Wales,Ireland and the Basque Country .Monograph series, no 2. Flensburg: European Centre for Minority Issues. Grin, F.and Vaillancourt,F. (2000)On the financing oflanguage policies and distributive justice. In R.Phillipson (ed.) Rightsto Language. Equity, Power, and Education (pp. 102–10). New York: Lawrence Erlbaum. Helliwell, J.(1999)Language and trade. In A.Breton (ed.) Exploringthe Economics of Language (pp. 5–30). Ottawa: Canadian Heritage. Henderson, W.,Dudley-Evans, T.and Backhouse, R.(eds) (1993) Economicsand Language . London: Routledge. 64 Current Issues in Language Planning

Henrard, K.(2000) Devisingan AdequateSystem of Minority Protection . The Hague: Martinus Nijhoff. Hocevar, T. (1975) Equilibria on linguistic minority markets. Kyklos 28, 337–57. Hocevar,T. (1983)Les aspectsé conomiques de ladynamique fonctionnelle des langues [The economics offunctional language dynamics]. LanguageProblems and Language Planning 7, 135–47. Holden, N.(1987)Language barriers asdifferential constraints on the international behavior offirms. In H.Tonkin and K.Johnson-Weiner (eds) TheEconomics of Language Use (pp.119– 39). New York: Center forResearch and Documentation on World Language Problems. Jernudd, B.(1971)Notes on economic analysis forsolving language problems. In B. Jernudd and J.Rubin (eds) Can LanguageBe Planned? (pp.263– 76). Honolulu: University Press of Hawaii. Jones, E.(2000)The casefor a shared world language. InM.Casson and A.Godley (eds) Cultural Factors in Economic Growth (pp. 210–35). Berlin: Springer. Kaplan,R. and Baldauf,R. (1997) LanguagePlanning. From Practice to Theory . Clevedon: Multilingual Matters. Kontra,M., Phillipson, R.,Skutnabb-Kangas, T. and Varády, T.(eds) (2000) Language: A Right and a Resource . Budapest: Central European University Press. Kymlicka, W.(ed.)(1995) TheRights of MinorityCultures .Oxford:Oxford University Press. Labrie, N.(1993) La ConstructionLinguistique de la Communauté Europé enne [TheLinguistic Construction of the European Community ].Paris: Honoré Champion. Lambert, R.and Moore, S.(1990) ForeignLanguage in theWorkplace .Annals ofthe American Academy of Political and Social Science, no 511. Newbury Park: Sage. Lamberton, D. (ed.) (2002) The Economics of Language . Cheltenham: Edward Elgar. Lang, K.(1986)A language theory ofdiscrimination. QuarterlyJournal of Economics 101, 363–82. Lazear, E. (1999) Language and culture. Journal of Political Economy 107, S95–S126. Lazear, E. (2000). Economic imperialism. Quarterly Journal of Economics 116 (1), 99–146. Lemelin, C.(1998) L’Économisteet l’ Éducation [TheEconomist and Education ].Montréal: Presses de l’Université du Québec. Linder, S. (1970) The Harried Leisure Class . New York: Columbia University Press. Lüdi, G.(1998) Gesamtsprachenkonzept.Esquisse de Nouvelles Recommandations . Berne: Conférence des directeurs cantonaux de l’instruction publique (CDIP/EDK). Mackey,W. (ed.)(2000) EspacesUrbains et Coexistence des Langues [UrbanSpaces and Language Coexistence ]. Theme issue of Terminogramme 93–94. Marschak, J. (1965) Economics of language. Behavioral Science 10, 135–40. Maurais,J. (1987) Politiqueet Aménagement Linguistiques [LanguagePolicy and Planning ]. Québec: Conseil de la langue française and Paris: Robert. May, S. (2001) Language and Minority Rights . Harlow: Pearson. Mayer, T. (1993) Truth Versus Precision in Economics . Aldershot: Edward Elgar. McCloskey, D.(1990) TheRhetoric of Economics .Madison: University ofWisconsin Press. McLeod, Wilson (2001)Gaelic in the New Scotland:Politics, rhetoric and publicdiscourse. Journalof Ethnopolitics and Minority Issues in Europe(JEMIE) .On WWWathttp:/ / www.ecmi.de/jemie/. Mélitz, J. (2000)English-language dominance, literature and welfare. Unpublished manuscript. Institut d’é tudes politiques, Paris. Migué, J.-L.(1970) Le nationalisme, l’uniténationale et lathé orie économique de l’information [Nationalism, national unity and the economic theory ofinformation]. Revue Canadienne d’Économique , 3, 183–98. Mühlhä usler, P.(1996) LinguisticEcology: Language Change and Linguistic Imperialism in the Pacific Region . London: Routledge. Mühlhä usler, P.(2000)Language planning and language ecology. CurrentIssues in Language Planning 3, 306–67. Nelde, P.(1997)Language conflict.In F.Coulmas(ed.) TheHandbook of Sociolinguistics (pp. 285–300). London: Basil Blackwell. Language Planning and Economics 65

Nettle, D.(2000)Linguistic fragmentation and the wealth ofnations: The Fishman-Pool hypothesis reexamined. Economic Development and Cultural Change 48, 335–48. Oaxaca,R. (1973)Male-Female wage differentials in urbanlabor markets. International Economic Review 14, 693–709. ÓCinnéide, M.and Keane, M.(1988)Local socio-economic impacts associated with the Galway Gaeltacht. Gaillimh, Galway: Coláiste na hOllscoile Gaillimhe. OECD (1998) L’Investissementdans le Capital Humain. Une ComparaisonInternationale [Human Capital Investment.International Comparisons ]. Paris: OECD. Officeof the Commissioner ofOfficial Languages (1991) OfficialLanguages. Some Basic Facts. Ottawa: Office of the Commissioner of Official Languages. Patrinos, H.and Velez, E.(1996)Costs and benefits ofbilingual educationin Guatemala.A partial analysis. Human Capital Development Working Paper , no. 74. World Bank. Pendakur, K.and Pendakur, R.(1998)Parlez et vousrecevrez: Laconnaissance linguistique comme capitalhumain [Speakand ye shall receive: Language knowledge ashuman capital].In A.Breton (ed.) Explorerl’ ÉconomieLinguistique [Economic Approaches to Language and Bilingualism ] (pp. 97–134). Ottawa: Patrimoine canadien. Pennycock, A.(1994) TheCultural Politics of English as an InternationalLanguage . London: Longman. Phillipson, R. (1992) Linguistic Imperialism . Oxford: Oxford University Press. Phillipson, R.(ed.)(2000) Rightsto Language. Equity, Power, and Education . New York: Lawrence Erlbaum. Pool,J. (1991a)A taleof twotongues. Unpublished manuscript.Department ofPolitical Science, University of Washington. Pool, J. (1991b) The world language problem. Rationality and Society 3, 21–31. Pool, J. (1991c) The official language problem. American Political Science Review 85, 495–114. Pool,J. (1996)Optimal language regimes forthe European Union. InternationalJournal of the Sociology of Language 121, 159–79. Price,A. (1994) TheDiversity Dividend .Brussels/Dublin: European Bureaufor Lesser Used Languages. Psacharopoulos,G. (ed.)(1987) Economicsof Education: Research and Studies . Oxford: Pergamon Press. Raynauld, A.and Marion, P.(1972)Une analyse économiquede ladispari té inter-ethnique des revenus [An economic analysis ofthe interethnic earnings differentials]. Revue Économique 23, 1–19. Raynauld, A.,Marion, P.and Béland, R.(1969)La ré partition des revenus selon les groupes ethniques auCanada [Income distribution byethnic group in Canada]. Unpublished manuscript.Commission royale d’enquê te sur le bilinguisme et le biculturalisme, Ottawa. Ricento, T.(ed.)(2000) Ideology,Politics and Language Policies: Focus on English .Amsterdam: John Benjamins. Robbins,L. (1935) An Essayon theNature and Significance of Economic Science (2nd edn). London: Macmillan. Robins, R. and Uhlenbeck, E. (eds) (1991) Endangered Languages . Oxford: Berg. Rossi-Landi, F.(1968) IlLinguaggio come Lavoro e ComeMercato [Languageas Work and Market]. Milan: Bompiani. Rubinstein, A.(1999)Five essays oneconomics and language. Unpublished manuscript. Department of Economics, Princeton University. Sabourin, C.(1985)La thé o rie des environnements linguistiques [Language environments theory]. In F.Vaillancourt(ed.) Economieet Langue [Economicsand Language] (pp. 59–82). Québec: Conseil de la langue française. Scharpf,F. (1997) GamesReal Actors Play. Actor-Centered Institutionalism in PolicyResearch . Boulder, CO: Westview Press. Schenk, R.(1987)Altruism asa source ofself-interested behavior. Public Choice 53 (2), 187–92. Schiffman, H. (1996) Linguistic Culture and Language Policy . London: Routledge. Selten, R.and Pool,J. (1991)The distribution offoreign language skills asa game 66 Current Issues in Language Planning

equilibrium. In R.Selten (ed.) GameEquilibrium Models, vol. 4, Social and Political Interaction (pp. 64–87). Berlin: Springer. Sen, A. (1985) Commodities and Capabilities . Amsterdam: North Holland. Simonnot, P.(1998) Trente-neufLeç ons d’ ÉconomieContemporaine [Thirty-NineLessons in Contemporary Economics ]. Paris: Gallimard. Skutnabb-Kangas,T. (2000) LinguisticGenocide in Educationor Worldwide Diversity and Human Rights? Mahwah, NJ: Lawrence Erlbaum. Skutnabb-Kangas,T. and Phillipson, R.(eds) (1994) LinguisticHuman Rights.Overcoming Linguistic Discrimination. Berlin: Mouton de Gruyter. Sproull,A. (1996)Regional development and minority language use: The caseof Gaelic Scotland. International Journal of the Sociology of Language 121, 93–117. Stanton,P.J. and Lee, J.(1995)Australian culturaldiversity and export growth. Journal of Multilingual and Multicultural Development 16, 497–511. Stokey, E. and Zeckhauser, R. (1978) A Primer for Policy Analysis . New York: Norton. Swedberg, R.(1990) Economicsand Sociology Redefining their Boundaries: Conversations with Economists and Sociologists . Princeton: Princeton University Press. Taddei, X.and Antomarchi, F.(1997) Économieet Identité – Ecunumia Identitaria [Economics and Identity – Identity Economics ]. Ajaccio: Albiana. Thorburn, T.(1971)Cost-benefit analysis in language planning. InB.Jernudd and J.Rubin (eds) Can Language Be Planned? (pp. 283–305). Honolulu: University Press of Hawaii. Vaillancourt,F. (1978)La Charte de lalangue française auQué bec. Un essai d’analyse [The French Language Charter in Québec. A tentative analysis]. CanadianPublic Policy/ Analyse de politiques 4, 284–308. Vaillancourt,F. (1980) Differencein Earningsby Language Groups in Quebec,1970. An EconomicAnalysis .Québec: Centre international de recherche sur le bilinguisme (publication B-90). Vaillancourt,F. (ed.)(1985a) Économieet Langue [Economicsand Language ].Québec: Conseil de la langue française. Vaillancourt,F. (1985b)Le choixde lalangue de consommation [Choosing the language of consumption].In F.Vaillancourt(ed.) Économieet Langue [Economicsand Language ] (pp. 209–20). Québec: Conseil de la langue française. Vaillancourt,F. (1988) Langueet Disparité s deStatut É conomiqueau Québec: 1970 et 1980 [Languageand Economic Differentials in Quebec:1970 and 1980 ].Québec: Conseil de la langue française. Vaillancourt,F. (1991) Langueet Statut É conomiqueau Québec: 1980– 1985 [Languageand economic status in Quebec: 1980–1985 ]. Québec: Conseil de la langue française. Vaillancourt,F. (1992)Language and publicpolicy in Canadaand the United States:An economic perspective. InB.Chiswick (ed.) Immigration,Language and Ethnicity. Canada and the United States (pp. 179–228). Washington: AEI Press. Vaillancourt,F. (1996)Language and socioeconomicstatus in Quebec:Measurement, findings, determinants, and policy costs. InternationalJournal of theSociology of Language 121, 69–92. Vaillancourt,F. and Grin, F.(2000)The choiceof a language ofinstruction: The economic aspects. Unpublished manuscript. Report to the World Bank Institute, Washington. Vaillancourt,F. and Leblanc,M. (1993) La Propriété de l’Économiedu Québec Selon le Groupe d’AppartenanceLinguistique [TheOwnership of theQuebec Economy by Language Group ]. Langue et société series. Québec/ Montréal: Office de la langue française. Van Parijs, P. (2001) Linguistic Justice. Politics, Philosophy and Economics 1, 59–74. de Varennes, F.(1996) Language,Minorities and Human Rights .The Hague:Martinus Nijhoff. Watts,R. and Murray, H.(2001) Diefü nfte Landessprache? Englisch in derSchweiz [The Fifth NationalLanguage? English in Switzerland ].Berne: Publikationder Akademischen Kommission der Universität Bern. Wimberley, J.(2000).Is rational policy compatiblewith democracy?Unpublished manuscript, D.G. IV, Council of Europe, Strasbourg. Wolff,J. (1993)Les Pensées Économiques des Origines ànos Jours[Economic Thought from its Origins to the Present]. Paris: Economica.