Posted on Leave a comment



Economic Approaches to
Language and Bilingualism

Richard G. Harris
Department of Economics
Simon Fraser University

Excert from

3.1 Language as a Human Capital
3.2 Languages and Network Externalities
3.3 Migration and Language Dynamics


In this section the economic theories of language are reviewed and their implications for the changes that ICT technology may bring to the supply and demand of languages is taken up. The economics literature on language is surprisingly small, and much of that is empirical.16 In this section the focus is on alternative conceptual frameworks that economists have used in examining language. To the non-economist the difference between these theories may not appear evident. Each uses the common language of economics and there is a great deal of intersection between the approaches in that all are based on traditional neoclassical microeconomic theory emphasizing individuals’ maximizing behaviour and the role of markets in setting prices and rates of return. The differences among them are really a matter of emphasis. The first (and oldest) theory is the human capital model, which stresses that language acquisition by individuals is an investment decision with up front costs and future (and uncertain) benefits (as in any form of education or training). The human capital approach brings to bear the large literature in economics dealing with the determinants of investment decisions and their consequent effects on the supply of capital goods in the economy. The second set of theories deals with social interaction aspects of language that are treated in the economics literature as network externalities. In common with many communications networks a language is useful because others use it, and this gives rise to a natural form of social economy of scale in the use of a particular language. In this literature there is a presumption that the benefits of speaking a particular language do not accrue entirely to the individual learning that language. There is a prima facie case that a free market in language choice will not result in a socially optimal outcome. The third set of theories discussed emphasize the interaction between geography and language. Language use is naturally concentrated geographically. Over time however as a consequence of trade and the migration of individuals, a region’s pattern of economic specialization and linguistic composition can change. It is the long run dynamics of this process that are emphasized here.

16.Some recent reviews include Breton (1997), Coulmas (1992) and Grin (1996) and much of the recent Canadian relevant material is covered in Department of Canadian Heritage(1997).

3.1 Language as a Human Capital

Virtually all economic models of language depend in some way on the basic supply and demand framework derived from the human capital framework.17 The acquisition of a second language is a general skill which has value by virtues of its ability to provide individuals the ability to interact with a larger group of individuals. Its great strength is the ability to make predictions about the wages of those with different linguistic capabilities. The theoretical framework was originally set-out by Breton (1964) and extended to a general equilibrium framework by Breton and Mieszkowski (1979). In its simplest form it assumes that the benefits and usefulness of a second language are entirely appropriable by the individual undertaking to learn the language, and that languages are learned only for their potential economic benefit, rather than for personal or cultural reasons. The basic theory is built upon the traditional Marshallian model of a competitive industry. Assume that language training is provided by a competitive industry that provides learning at marginal and average cost. Under these conditions in a market equilibrium prices will be set in labour markets and in the language training market such that individuals learning a second language will be compensated by a favourable wage differential, such that the present value of the additional earnings just compensates for the cost of learning that language. Included in “cost” are the foregone earnings due to time spent learning the second language. The value of the second language, in whatever use it is put, reflects the additional productivity that skill brings to the production, marketing and trading activities in which it is used. Language is almost always thought of as a general as opposed to a firm specific skill, and therefore individuals are the primary agent responsible for investment in the acquisition of additional language training.

17.The literature is fairly extensive and much of it by Canadians. A non-exhaustive list includes Breton (1964), (1978), Chiswick and Miller (1988), Grenier (1982), Migue (1970), Raynauld and Marion (1972), and Vaillancourt (1985).

A common application of the theory is to the case of a minority language user learning a majority language, or a lingua franca, which is widely used for business purposes. The additional earnings are assumed to reflect the productivity benefits that communication in the lingua franca bring. The origin of language used as the lingua franca is not necessarily explained in the analysis. The distributional implication of the theory is that individuals whose mother language is the lingua franca earn an economic rent, reflected in their wages, which exist as it is not necessary for them to acquire the lingua franca. Minority language individuals must learn the lingua franca, and therefore must bear the cost of that learning.18 As Breton and Mieszkowski emphasize the analysis is not complete unless one asks what determines the prices of the goods and services in the economy. In an economy in which all prices are set internationally there can be no price adjustment effect and all differences must be reflected in wages and the cost of training. If some goods and services are provided locally, and not traded internationally, then there is question of what language is used in production and distribution of those goods and how the price of those goods is determined. If language training is necessary for the production and distribution of non-traded goods then consumers will bear part of the cost of that training in the form of higher prices.

18.Breton and Mieszkowski (1979) and Breton (1978) pursue the distributional implications of this asymmetry under a variety of assumptions.

It is commonly argued there are failures in the market for human capital which may lead to an underinvestment by individuals in the appropriate level of education and training if all educational markets were privately organized without government intervention. These failures may occur for a variety of reasons: inability to legally create transferable property rights over the future wages of individuals; non-diversifiable income risk that occupational specialization entails; and informational asymmetries in the market for educational loans which leads to inefficient credit rationing of those borrowing for educational-training purposes. Government intervention in the form of subsidies to education are usually viewed as the appropriate policy response to these problems. All of these market failures may exist in the case of language acquisition but it is difficult to rationalize any additional subsidies specific to language training, rather than more generally for education, unless one can identify additional marginal social benefits due to market failures specific to language-learning.

Language of course is not only learned, it is used, and as such the use to which it is put may be subject to subsidy, taxation and/or regulation. Within the human capital framework regulatory interventions on language use shift the demand curve for language, and thus impact on the supply of those learning the language by affecting the returns to the language. A general principle of optimal, or first-best, policy design is that a market failure should be corrected as close as possible to its source. Thus if there is a failure in the language learning market the correction should be applied at the source. For a variety of practical and political reasons however this dictum on policy design is widely disregarded in practice. In the case of languages if it is deemed that not enough individuals are using, or learning, a language, then regulations regarding the necessary use of that language may be a second-best policy response. These regulations may raise the cost of doing business and thus shift the burden from the government (or taxpayer), who would otherwise be providing subsidized language training, to the consumer in the form of higher prices. A deeper explanation of why such policies are used requires an appropriate positive theory of politics.

i) Human Capital Theory and ICT

How can we think of the recent developments in ICT in light of the human capital approach to language? The new communications technologies can be thought of as impacting upon the supply and demand for languages, and language learning, in a number of ways. It is helpful to elaborate on the concept of a commodity or service characteristic labeled “language”. It is useful when thinking about ICT developments to think of the primary economic activity as “communications” – a produced intermediate service used in virtually all organizational aspects of economic activity. Communication cost reflects the technology of providing communication services, the inputs to the production of communication services, and the price of those inputs. Language skills, the ability to read and write and/or translate between languages is one such input. We can think of ICT as providing new and different ways of providing communication services. Because it is a technology which substantially reduces the cost of communication, especially over longer distances, we expect that its widespread adoption will give rise to increased demands for the business input “communication services”, not the least of which is because of its ability to expand effective market size and allow greater specialization over larger geographic areas.19

19.In our discussion of international markets in section 4, this aspect of ICT will be dealt with at greater length.

As an essential input to communications services the demand for language skills should rise as consequence of the increased availability of ICT. However language skills are for the most part, excluding computer translation software, embodied in human beings, and these individuals have other characteristics which contribute to their bundle of human capital. ICT technology is only useful to those capable of using it effectively. Therefore the increased demand for language skills due to ICT should fall primarily on those workers who have both the technical and language skills. Consider the Breton-Mieszkowski case of a minority language group of a small open economy having to acquire the lingua franca in order to engage in international trade. The introduction of ICT raises the degree of internationalization and increases the demand for those who jointly supply the ability to work in the lingua franca and to use (and supply) ICT services. Holding international prices constant this will raise the returns to those having both sets of skills and reduce the returns to minority language monolinguals. The wages of bilinguals who are not capable of working with ICT may or may not increase. At the margin non-computer literate bilinguals will be pushed into those activities where the ability to work in the lingua franca is important, but ICT technology is not. As ICT technology becomes more pervasive those type of jobs become more scarce, and therefore eventually the wages of the group competing for those jobs should fall.

On the supply side there are two sorts of effects. One, the cost of language learning will be reduced to some extent by computer-aided learning technologies. How important this effect might be is yet to be determined, but it should lead to a reduction in the wage premium to those acquiring the lingua franca, other things being held equal. Second, and more importantly, within any given geographic labour market, the introduction of ICT will increase the availability of language services to business. These services will be made available via technologies such as the Internet and video-conferencing. Imagine for example someone in a minority linguistic region who wishes to market a particular product in a foreign market. These services can be purchased by specialized multilingual international marketing firms via the Internet, rather than produced by the firm itself. The increase in the international supply of language services reduces the scarcity of those capable of working in both the home language and the lingua franca. Holding demand for these services constant, this effect would otherwise reduce the bilingual wage premium relative to home language monolinguals.

The net impact on bilingual wages depends on the relative magnitude of the increased demand for workers who are bilingual, and on the effective increase in the international supply of language services. The argument made above was simplified by focusing on the impact of ICT on one regional economy in isolation. In reality ICT is being introduced around the world and similar effects will be felt everywhere. If the global demand for multilingual communications services rises sufficiently then the prices of foreign language service providers will also rise, mitigating the negative effect on the wage premium to language acquisition noted in the discussion of supply.

The policy response of governments to perceived market failures in language markets will also be affected by ICT innovations. An obvious implication of course is that, to the extent the cost of language acquisition has been reduced by this technology, governments may be more prone to subsidize the use of modern computer aided instruction if they perceive a need to increase the number of bilinguals in the economy. They may also perceive that the bottleneck in the labour market is not with the number of bilinguals, but rather with bilinguals capable of using ICT. In this case the response would be to target computer training at bilinguals. ICT may also lower the direct cost to governments, or increase the effectiveness, of regulating the use of language. For example a government run software search engine could check Web sites on the World Wide Web for the language used. This information in turn could be used for enforcement of language quotas on the Internet, analogous to the current discussions about the regulation of pornography on the Internet. In general the potentially invasive nature of this technology when used by governments so inclined raises the prospects of greater use of specific regulations to enforce or promote the use of a language, rather than the use of more general supply oriented policies. How societies will reconcile the use of such technology to promote a specific language use in the face of other social objectives such as privacy, and freedom of speech will undoubtedly be one of the more difficult policy issues in the near future.

3.2 Languages and Network Externalities

i) Language as a Single Communications Network

Languages, as in the case of other technologies for communication, exhibit what economists refer to as network externalities.20 That is the value to an individual of the use of a particular communication device, for e.g. fax machines, is useful only to the extent that others have a compatible device. In these circumstances the adoption by one individual of a specific communication device confers a social benefit on all those using a compatible technology. In the case of language, an individual that chooses to learn a particular language confers a benefit upon all those currently using the same language. When network externalities exist private investment decisions on technological adoption will not result in an optimal allocation of resources. This is because the balancing of private costs against private benefits does not take into account consequences of that decision for the size of the network and thus the benefits to others on the network. When the individual’s choice is between two competing and mutually exclusive communication technologies, a switch from one technology to another confers a cost on all those using the discarded technology, as the size of their compatible group has been diminished. The case of the Betamax versus VHS video format is a common example of the rivalry between two competing devices, the use of which both exhibit an economy of network size. When an individual switched from Beta to VHS they conferred a benefit on existing VHS users and a cost on Beta users. Neither of these costs or benefits are incorporated in individuals’ decision criteria. When individuals make a choice to use, in a relatively permanent sense, one language or another, a similar set of costs and benefits arise and thus a discrepancy arises between privately determined language choices and those that would maximize social efficiency.

20.The economics literature on networks is quite large. Some recent surveys are Economides (1996) and Katz and Shapiro (1986) (1994). Specific applications of network models to language include Church and King (1993), Grin (1989), and Laitin (1994). Closely related are the earlier models of Carr (1976) and Hocevar (1975).

It has been argued, by Carr (1976) for example, that it would be most efficient if all communication were conducted in one language, thus minimizing aggregate communication costs. Network externalities give rise naturally to social increasing returns associated with single language use – in effect language use is like a natural monopoly. If there are two languages, both of which are equally efficient at communication, then the existence of a network externality dictates that the efficient outcome would be the use of one language-monopoly. It is possible, as described by Church and King (1991), that a free market in language choice with network externalities would result in bilingual outcomes even though this is not efficient. In this case an appropriate policy response is to subsidize the use of one language or the other. With different sized language groups the cost minimizing policy in these circumstances is for the minority group to invest in learning the majority language – a limited form of biligualism.

ii) Network Systems

The network literature also examines what is the now commonly recognized problem of competition among network systems. A communication network for example consists of both hardware, (e.g. the telephone lines and computers) and the software, which allows different computers to communicate. The overall system consists of these complementary software products, the purchase decision which in principle is made by many different individuals, together with the hardware infrastructure linking them. There are many examples of network type systems competing with each other. VHS versus Betamax, PC’s versus Mac’s and so on. In the case of languages the natural analogue is multiple languages competing with one another, and the obvious questions arise regarding the benefits to standardization and compatibility between competing systems. Software competition is one factor which affects the prospects for system survival. In choosing to purchase a particular piece of hardware the size of the installed hardware base of the system is another critical factor. The larger the installed base the more reason consumers have to believe the system will be used in the future. When new hardware appears, and the installed base is low, the risk that the system will not be available in the future is high and this conditions hardware purchase decisions. A similar effect is evident when multiple languages are (potentially) in use. More popular languages have a reduced risk attached to their future use, and this sets up positive feedback effects on their adoption – consequently languages that are expected to be popular will be. Competition among languages is not unlike competition between network systems.

As Katz and Shapiro (1994) emphasize, competition among alternative network systems leads to three essential problems – expectations, coordination, and compatibility. Private and public institutions affect outcomes in a systems competition in each of these dimensions. The potential market failures in a network system is quite different than that which occurs in a pure communications network, and hinges on the distinction between the hardware and software of the system.

The expectations of both consumers and suppliers are important as the development of the system and the construction of its components is distributed over time. An individual for example choosing to learn a particular software language, or a seller choosing to produce a particular computer component such as a disk drive, must form expectations about the quality, reliability and extent of the rest of the system in the future. Once such investments are made they tend to be system specific.

The need for coordination on hardware choice is motivated by the presence of scale economies in the provision of the system software. A market failure can occur due to the linkage over time between the investment decisions on hardware, subsequent software development, and the feedback from one to the other. If average costs to users fall as the number of users adopting a specific piece of software rises, for example due to the large fixed costs of software development, then we would say there are economies of scale in the software provision stage. In this case it would be desirable if system purchasers could all ex ante agree on a single hardware type. The purchase of the hardware depends upon what software consumers think will become available. If there are economies of scale in the provision of the software, the ultimate availability of that software will depend upon what other consumers choose. This gives rise to positive feedback effects and the desirability for a coordination mechanism amongst consumers. Conventional market allocation mechanisms in which only prices are used to provide signals cannot solve the coordination problem in these circumstances. More direct mechanisms are necessary.

To get an efficient outcome overall requires a common set of expectations by both buyers and sellers, a coordination mechanism and ultimately, in the event of multiple systems, system compatibility amongst components. Obviously government regulation is one mechanism by which coordination can be achieved. But as emphasized in the network literature firms themselves often recognize the problem and form self-regulating industry or standards associations whose precise role is to perform the coordination function.

To what extent is the “hardware-software” paradigm of network systems theory relevant to languages? It is to this issue we now turn.

iii) Multiple Languages as Competing Network Systems

Language systems have many of the characteristics of the “hardware-software” paradigm in network systems described above, and competition between languages can be analyzed in a similar manner. In this instance the language, or more precisely the ability of individuals to use it, is the hardware of the language system. Individuals must make investments in learning a language; these investments once made are sunk costs, i.e. they are irreversible and highly durable. Their ultimate value depends upon the future availability of social and market institutions in which the language is useful – the institutions in this case are the “software” of the language system. As language is the basis for most communication almost any institution or activity which revolves around communication can be regarded as the software of the language system. This would include the social infrastructure such as the education system, the libraries, newspapers, television and radio, public and private signs, and so on. Many of these are linguistically specific. Individuals and firms give value to the system by either learning a language (investing in “hardware”), or contributing to the provision of “software”. A free market in languages has all the characteristics of a network systems competition. In this instance free market outcomes need not be efficient. Specifically coordination on a particular language (the hardware) is necessary because of the economies of scale in the provision of the “software” side of a language system.21

21.It is important to emphasize that the market failure occurs even in the absence of network externalities in the choice of language. The market failure will not occur if all software is provided under conditions of constant cost – in this case there are no downstream economic benefits to use of a single language.

Market structure in the “software” sector can affect also the outcome of the competition and conditions the necessary policy response to the “hardware-software” problem. The existence of scale economies in a particular software sector may be such that perfect competition cannot be sustained within that sector and an imperfect market structure is a more natural outcome. Suppose for example the book industry is monopolistically competitive – firms produce imperfect substitutes under conditions of decreasing cost, but any firm is free to enter (or leave) the industry. Books are quite obviously an important part of the “software” of a language system. With the free entry and exit of firms to the industry, books would be priced at average cost but above the marginal cost of production. Books are a good example where consumers value diversity in the range of goods offered. One policy response would be to offer a subsidy to the acquisition of a specific language which would increase the size of the “installed base” of a particular language group. In turn this would give rise to a greater variety of books published in that language as it would help to condition publishers expectations that a market for books in this language would exist. As in the usual analysis of monopolistic competition such a subsidy can be on balance welfare improving. A partial solution in this case is to use policy to coordinate expectations at the hardware end, even though the ultimate price-social cost divergence is in the software end.22 If a subsidy were offered to book production price could be set equal to the social marginal cost of book production, but there is no guarantee the coordination problem on language would be solved – there might either be too many languages or too few.

22.Thus a first best, or fully efficient, outcome cannot be achieved by this policy alone.

Most of the devices available to firms to coordinate expectations in the case of systems competition, such as an industry standards association, are not available in the case of languages. Language adoption decisions are made by thousands of unrelated individuals and firms. The commitment of the state to provision of the necessary “social software” of the language systems can provide a focal point for expectations and thus coordinate individual investment decisions. This is an instance where the public sector may lead the private sector, and in so doing solve the coordination problem.

One issue which arises under a regime of system competition is whether a component designed to work one system functions in another system – referred to as system compatibility. When people within a given geographic-economic region use different languages, this might be referred to as a case of systems compatibility failure. Bilingual systems are however precisely examples of overlapping language systems which offer a limited degree of compatibility. Both private and public institutions affect the degree of compatibility achieved and it is natural to inquire as to whether there is too little or too much compatibility.

Obviously the existence of bilinguals and translation services gives rise to a certain degree of compatibility.23 The degree of compatibility is clearly however subject to choice, both by individuals and by governments. Defining compatibility in an analogous sense to the network literature is a bit difficult. In the case of face-to-face communication it might be literally defined as the ability of any two people to converse in the same language. This would imply that everyone is bilingual, or that there exist a lingua franca that everyone speak. In a pure communications network benefits arise from the consumption externalities that increased size brings, with “size” being determined by the degree of compatibility achieved. In the instance of competition between two language systems the benefits of compatibility arise from better scale economies on the software side – that is the same software can be used for both languages. An example would be public signs, for which the marginal cost of providing the same information in a second language is very low; i.e. most of the cost of the sign are fixed costs not associated with the actual labeling. Another example would be libraries, again whose costs are largely fixed costs. If libraries provide material in both languages (i.e. the software is compatible) this benefits users of both language systems.

23.A formal model of translation with network effects is developed by Tamura (1996).

The cost of compatibility depends upon the technological requirements necessary, and the mechanism by which compatibility is achieved. The network literature provides two broad methods by which compatibility is achieved between systems – i.e. make software components of the systems interchangeable. Adapters are technologies which make a component designed for one system functional on the other. Standardization is defined as a constraint on systems design such that all components are expressly designed to be interchangeable. Within language systems translators (either human or computerized) are analogous to adapters. The cost of compatibility can thus be measured by the cost of providing translation services and the degradation in the quality of communications which translation gives rise to. Standardization in a language system is analogous to the requirement that all software be compatible with both languages so that the ultimate communication services which language is used to produce do not depend upon the language used. In some cases the cost of standardization may not be high. Libraries once built can carry books in any language. In other situations standardization is either not feasible or very costly. If one thinks for example of electronic voice media – such as radio and television, or the printing of catalogues by retail stores – the services must actually be duplicated in both languages to give both language groups the same benefits.

Now consider the possibility of a purely market-determined language outcome. With all software components privately provided there is no general guarantee that the market will provide the socially efficient degree of compatibility. The “tipping” phenomena, in which network scale gives rise to strong lock in effects, may result in multiple linguistically specialized services with a substantial degree of incompatibility. While this might be efficient for cost reasons the market structure effects of this scale and specialization may be negative. Monopoly or oligopoly in the supply of downstream language goods can produce an inefficient degree of bilingual service. This is particularly true if a firm providing a particular component can preserve market power, or protect its monopoly rents, by maintaining incompatibility with one of the languages. Markets might be segmented linguistically thus reducing demand elasticities and allowing prices to rise without attracting entry. From a social point of view the correct outcome may be to provide a bilingual (linguistic compatible) service, which would give rise to larger scale economies and thus lower marginal cost to both groups.

In modern economies language systems are rarely, if ever, the outcome of a free market mechanism. Because a language system involves the provision of software much of which is publicly provided, it is not really meaningful to speak of a purely market-based competition amongst languages and the theory must be amended to take this into account. Examples of language system software that are publicly provided include schools, libraries, printing of laws, traffic signs, and weather broadcasts. In many of these types of quasi-public goods there are economies of scale in production and thus the basic intertemporal coordination problem linking language choices to software provision exists. If governments supply bilingual public goods (software infrastructure) this would increase the likelihood of language compatibility in other privately provided parts of the system. It might in some cases be efficient to provide bilingual public software goods, even if ex post it appears that duplication of services is occurring. The reason is that the provision of such goods is in part a solution to coordinating expectations on future language use. The alternative, of providing only unilingual public goods, would lead to limited compatibility of the two languages, and possibly hasten the demise of the smaller of the two language systems.

In system markets the most conventional method for achieving compatibility of systems components would be for an industry standards group to decide on some form of standardization. In linguistic markets there is no good analogue to the “industry” whose commonality of interests is sufficiently well defined. Government may therefore have a role to play in providing a coordination function, either compulsory (as in the case of legislated solutions), or voluntary by promoting certain social conventions (such as bilingualism) in the provision of public services. The case for forcing linguistic compatibility of the overall language system, a fully comprehensive bilingual system, would seem to be particularly important where significant scale economies are possible in the provision of bilingual services. These must be weighed against the cost savings possible if similar services are provided in only one language or perhaps not at all to one linguistic group.

iv) Network View of Language and ICT Innovations

Summarizing, the main characteristics of the network externalities approach to language are:

  • economies of size exist in the use of a single language which give rise to externalities in language choice decisions by individuals
  • the hardware-software paradigm of network systems suggests that with economies of scale in the provision of activities or institutions which are language intensive there is a need for an expectations coordination on the language (or languages) to be used in the future.

The impact of ICT innovation within this framework is similar to that of the human capital approach but with the additional effects that follow from the presence of network effects. ICT increases the intensity and spatial scope of communication and thus increases the geographic span of the network externality in language use. The agglomeration and lock-in effects associated with these externalities are now extended geographically. The private return to use of the lingua franca increases and this raises the probability that the lingua franca will be used in any communication. With the positive feedback and lock in effects, as more people use the dominant language on a global scale local language use patterns change. Thus the initial pattern of language use will be upset by the ICT in favor of larger linguistic groupings spread over greater areas. An immediate policy implication is that provincial borders are less relevant to extent of the market for services provided in French than they were previously.

From the perspective of the hardware-software paradigm, ICT is itself a major new form of language systems “software”, as it is a primary means of communication in which written and spoken language is used. In some aspects of ICT provision there are economies of scale which are dependent on the use of single language. Many examples could be given which hinge on the necessity of various components of the system to talk to one another. Standardization of communication protocols is necessary and this is greatly facilitated by the use of a single language. This is not unlike the requirement in international aviation that virtually all pilots and air traffic controllers use a single language which happens to be English. The existence of these economies might strengthen the case for coordinating on a single lingua franca outcome. On the other hand a great deal of ICT is not language specific – both hardware and software. E-mail programs can be easily adapted for use in any language; distance learning via ICT means that specialized courses can be made available to a much larger group and thus the scale economy restriction that otherwise forces language compatibility is reduced. In short it is possible that ICT rather than reinforcing single language outcomes actually reduces the significance of scale economies in the provision of language software. Thus the focus shifts to the possibility of greater competition between languages and use of multiple languages within single regional/ economic markets.

ICT also changes the role of some aspects of the provision of other language systems “software”. Book publishing provides an important example. There are significant economies of scale of book publishing via traditional (pre ICT technology). The existence of these scale economies gives rise to the desirability for a single language coordinated outcome. ICT by reducing the fixed costs of setting type, editing and other aspects of publishing make shorter runs much more economical – that is economies of scale in publishing have been reduced by ICT. The impact of ICT on language and books is that with the reduction of the scale economies in book publishing the need for ex ante coordination on language is not as great. This simultaneously reduces the role for governments in attempting to coordinate on particular language outcomes, and increases the prospects for a multilingual publishing industry.

In summary the network point of view on language suggests that there are potential gains to social coordination mechanisms on language which markets cannot be expected to perform. As in the provision of public goods, the government may have a role to play in providing the lead in language market outcomes. This conclusion is complicated by the fact that the scope of the externalities may go beyond the borders of most national governments, and this problem is exacerbated with the widespread diffusion of Internet type networks.

3.3 Migration and Language Dynamics

Some more recent economic theories of language take a longer time perspective and emphasize the interaction between economic specialization, linguistic concentration, and the mobility of people on regional incomes and growth. Mobility of people in response to persistence income differences between regions is a general theme of the literature on regional economics. An important empirical observation has been that linguistic assimilation occurs much faster when linguistic groups become dispersed geographically24. Clearly the time horizons considered here are quite long – decades at least. While there are benefits to trade between regions there are also transactions costs imposed on that trade by linguistic differences between regions. If these costs are high enough and the linguistic groups are highly concentrated regionally it is possible that little trade between regions will result; the outcome is both geographic and linguistic isolation. In a review of the sociolinguistic literature on language use John and Yi (1996) identify a common pattern in the evolution of minority languages over time. There are three identifiable stages:

  • 1. Two language groups co-exist geographically with limited economic interaction.
  • 2. One group experiences a secular boom; industrialization would be an example.
  • 3. The second language group assimilates into the economically larger group and few monolingual speakers of the second language remain.

It is argued that a number of cases fit this pattern including the Ainu in Japan, the Welsh in Britain and the Breton in France.

24.John and Yi (1996) summerize this literature.See also Dunkin (1996)

An economic model pertinent to this type of situation is provided by John and Yi (1996); with two regions and two languages. They assume that a common language is necessary for both trade and production. Individuals initially only know one language but can learn at a cost a second language in order to produce with others using that language. They also add a dynamic migration decision to the analysis by assuming that over time people may choose to move between regions if income differences are sufficiently large enough. With endogenous language learning there are some assumed economic benefits to both economic and linguistic agglomeration. Network externalities are assumed to exist both with language learning and with language use. An individual who speaks language A and moves to another region confers an external benefit on all speakers of A within the host region and a loss on all speakers of A in the source region. The theoretical prediction of this analysis is that complete assimilation of the smaller language group is more likely if the dominant language region is also the growing region, with immigration to the expanding region and adoption by immigrants of the dominant language. There are also partial-assimilation outcomes if both language groups are sufficiently similar in size. The gains to learning the other language in this case are sufficiently small while the cost is high. Production in this case occurs with linguistic specialization within each region. Full assimilation is less likely the higher the mobility costs and the more even the other (exogenous) determinants of economic growth across regions. In general the theory emphasizes the important role that migration plays in the longer term evolution of language use through its impact on language group size. Their analysis hinges on the presence of a positive feedback effect with the A language network externality, and the interaction with migration in response to economic opportunities which are specific to one region. To provide a benchmark, the theoretical prediction of their analysis with only one region, and therefore no migration opportunities, is full assimilation of the minority language due to the economies of scale the network externality introduces.

Immigration policy and regional migration therefore clearly will play an important role in the long term determinants of language use if this theory is correct. John and Yi also provide an explicit model of bilingualism in that they assume there is an initial group of individuals who can use both languages. This captures the important fact that languages are learned for a number of reasons other than economic – cultural, parental, educational and peer influences for example. In this instance they show that with initial bilinguals there is a possibility that monolingual coexistence can be sustained within the same region, as the bilinguals provide the necessary transaction services for interregional trade, and thus income growth can be sustained in both regions without migration.

They argue that Canada relative to the United States provide useful benchmarks on these alternative linguistic outcomes. In the United States immigrant minority language groups are assimilated with internal migration between regions of the U.S. contributing to the speed of the assimilation process. In Canada they argue that French versus English is increasingly based on geography, resulting in linguistic isolation of the two groups.

i) Mobility and ICT

Over the longer term trade in goods between regions and the movement of people are substitutes. The geographic pattern of language use and other location specific factors provide a great deal of inertia in individuals’ location decisions. ICT technology by promoting virtual mobility of labour services reduces the cost of trading, relative to the costs of migration. This has some fairly predictable consequences for the interaction of language use and learning with migration.

a) The benefits to migrating from a region in which one’s native language is not used extensively to a region in which the language is used extensively may be reduced due to the consumption benefits conferred by virtue of the creation of larger virtual networks of specific language users. This would tend to result in greater persistence in location decisions and thus a tendency to lock-in the historical geographic distribution of language users when the new technology is introduced.

b) The benefits to migration for purposes of participating in an economic boom may be reduced by the availability of ICT technology. This will be particularly true for those who are involved in the production of goods and services which are in demand by the region experiencing the boom. Integration of production across space by use of ICT has the potential effect of transferring an initial boom in one region to similar producers in other regions. This will reduce the flow of migrants that might otherwise occur for reasons of employment, but to increase the amount of language learning necessary for international production. To the extent these migrants would have been part of a linguistic minority this would reduce the rate of assimilation of the minority language.