An Economics Approach to Language Policy and Linguistic Justice

For an overview, see Vaillancourt (1983) and Grin (2003).

For example, the analogy between currency and language is employed by Calvet (2002), the notion of “linguistic market”, among others, by Ehlich (2007) and Inoue (2007), and the term “linguistic protectionism” by Schulzke (2014). For a critical discussion of these metaphors from an economics point of view, see, among others, Grin (2005a) and Gazzola (2014b).

See, for example, Van Parijs (2003) and Robichaud (2011, 2017).

Cf. Till Burckhardt’s contribution in this book (chapter “Linguistic Disenfranchisement and Labour Mobility in Europe”).

Academic literature in language economics currently includes some 500 titles. A recent bibliography is provided by Gazzola et al. (2016).

The papers of Thorburn (1971) and Jernudd (1971) are illuminating examples. For a brief history of LPP, see Jernudd and Nekvapil (2012).

E.g. Beacco (2016), Calvet (1996), Dell’Aquila and Iannàccaro (2004), Spolsky (2004, 2012), Johnson (2013), Marten (20016), or Tollefson and Pérez-Milans (2017). The books edited by Ricento (2006a) and by Hult and Johnson (2015) are partial exceptions.

See Gazzola and Grin (2017) for an overview.

The literature in discourse analysis is vast and cannot be summarized here. For an introduction, see, among others, Gee and Handford (2012).

The term allocation refers to how goods are used in individual consumption and in the production of individual firms. The goal of efficient allocation basically means that no resources are wasted; they are on the margin employed where they provide the highest benefits. This is independent of who is reaping the benefits. Distribution, on the other hand, is concerned with who benefits from economic activities in comparison to other individuals. We return to this distinction in Sect. 1.1 where the concept of allocative efficiency is discussed in some detail. Distributive issues are discussed in Sect. 6.

See, for instance, Kymlicka (1995), Kymlicka and Patten (2003), Van Parijs (2011), Ricento et al. (2015), or De Schutter and Robichaud (2016).

Kymlicka (1995) and Patten (2009) reflect this tradition very well. Patten (2009), for instance, lists five basic approaches. First, toleration basically implies that prohibitions on language use are absent; individuals in their private lives can use whatever language they desire. Second, accommodation is a minimal right; an individual should in certain situations receive assistance, if he/she is severely disadvantaged because of language. Third, context of choice signifies that the individual should be able to live a “full” life in his or her “own” culture. Fourth, the end-state argument attributes an intrinsic value to linguistic diversity, drawing parallels to biological diversity. Fifth, the fairness argument is used to argue for equality of opportunities of members of all groups; a minority should not be disadvantaged in comparison to a dominant majority. Our approach in this essay comes close to the fairness argument. See also Patten (2014) and the extensive discussions by Alcalde (chapter “Linguistic Justice: An Interdisciplinary Overview of the Literature”) and Shorten (chapter “Justice in the Linguistic Environment: Narrow or Wide?”) in this book.

See De Schutter (2007), Peled (2010), Ricento (2014), and Schmidt (2014).

See De Schutter (2007), Mowbray (2012), Peled (2010), and Van Parijs (2011).

Two representative collections of papers in this area are Skutnabb-Kangas et al. (2009) as well as May and Hornberger (2008).

That is, a situation where the spontaneous interaction in the market leads to undesirable consequences. Common examples are environmental problems or monopolistic structures. The various concepts used will be explained in detail, as we go along.

The separation of the normative analysis of distribution (fairness or justice) and of allocation (efficiency) generally goes back the work of Richard Musgrave, see Musgrave (1956/1957).

This includes actions taken by any public authority in order to influence the functioning of society with the goal of increasing efficiency or improving justice, for instance by providing equal opportunities in the access to various functions of society to people speaking different languages.

It is important to understand that optimality or maximization only makes sense with respect to well-defined goals. The definition of the goals is basically a political issue outside the realm of rational analysis. The analysis can only try to find and compare the ways leading toward the given goal.

There are numerous introductory texts giving an overview of the field. Hindriks and Myles (2006) is a good example, theoretically stringent without losing the contact to empirical reality.

See, for instance, Gazzola (2014b), Gazzola and Grin (2017), Grin (2003), or Wickström (2016b).

For a notable exception, see the work of François Vaillancourt and coauthors, for instance, Vaillancourt (1997), Desgagné and Vaillancourt (2016), or chapter “Linguistic Justice in a Federal Setting: The Case of Canadian Provinces” in this book.

Generally, it is assumed that people only care about their own well-being and ignore inter-dependencies such as altruism and envy. Conceptually, there is no problem introducing such elements into the preferences. However, this might make an efficiency analysis very opaque and render it useless. A more fruitful approach might be to introduce distributional issues exogenously, see below.

Consider the trivial example of two cookie monsters dividing a cake. All divisions of the cake giving a bigger piece to one monster when the other one receives less is (Pareto) efficient.

In our cake example, we could choose an equal division of the cake between the two cake-eaters.

A drastic example can be found in the Talmud, see Hillman (2009): two men are in the desert far away from a water source and have only enough water for one of them reaching the source. There are three possible outcomes: one equitable one, both die; and two efficient ones, only one of them dies.

The concepts, known as the Kaldor-Hicks criteria, go back to Kaldor (1939) and Hicks (1939). See also Ng (2004).

Basically, what the policy is worth to the individual. The concept and its drawbacks are further discussed in some detail in Sect. 4.

The underlying idea is that if transfers of resources between individuals were freely feasible, one should ask the question whether the gain of the winners due to some policy measure is high enough for the winners to fully compensate the losers and still have a net gain. If the policy issue changes the utility distribution (expressed in money terms and net of costs for the policy measure) between two individuals from (5,  4) to (7,  3), the two individuals would not agree upon which is the better one; both are Pareto efficient. However, if income transfers were possible, the first individual could transfer 1.5 money units to the second person if the policy is enacted. This would lead to the income distribution (5.5,  4.5) which is Pareto superior to (5,  4). Hence the policy measure leads to a potential Pareto improvement. In a cost-benefit analysis, one would simply compare the sums 5 + 4 = 9 and 7 + 3 = 10 and conclude that the aggregated net benefits increase due to the policy.

This is a standard problem in economic policy analysis. See, for instance, the theory of optimal taxation. We can apply this to our cake example from footnote 24, if the original distribution of the cake is very unequal. Assume that the monster on the losing side has to struggle harder and harder, that is, using up more and more of his slice of cake, in order to introduce a policy measure forcing the monster with the bigger slice to transfer some more cake to him. Part of the cake is then lost in the process of dividing it more equally—its size decreases with increasingly just distributions. Is it better for the monster on the losing side to have a small slice of a big cake or a big slice of a small cake, and should we only consider the size of the slice of the losing monster, or also the size of the slice of the stronger monster in making an evaluation?

This has, of course, been done by a number of authors. See the bibliography by Gazzola et al. (2016) for references.

We are only considering a few of all logically possible situations.

No policy intervention means that the school system is totally privatized and each set of parents freely chooses how to educate their own offspring.

The dynamics, involving language shift, is not being considered. For a short discussion of language dynamics, see Sect. 4.3.

Compare the situation involving traffic congestion in a city with some people owning cars and others only bicycles. In laisser-faire the cars might dominate and bicyclists will suffer many accidents. A government policy taxing driving and constructing bicycle paths will lead to a totally different transport equilibrium.

For a more detailed discussion of the concept of a network externality, see Church and King (1993) or Dalmazzone (1999) as well as the discussion in Sect. 3.2 below.

We stress that we have assumed that the costs in the intervention cases are covered over the general government budget, to which everyone contributes in a fashion that is independent of language use. If compensation payments were possible, the unfair allocations could be made fair(er) with the help of such compensations.

Of course, one could look at language as a piece of art just like a beautiful painting, a mathematical theorem, or a poem and admire its structure sui generis. This seems, at least partially, to be one of the main motivations of the online “conlang” community.

The ability to learn a language is probably an attribute with which the individual is born. This is, however, controversially discussed among some linguists.

The concept of human capital from knowing languages plays a very central rôle in chapter “Languages, Human Capital and Well-Being in Sub-Saharan Africa” of Katalin Buzási and Péter Földvári in this book.

Economists talk about the propensity—or willingness—to pay; see Sect. 4.

Social capital is made up of norms and trust between individuals, which can be fostered through interactions. Here, the connection to the linguistic environment is close. In a seminal article on the concept, James S. Coleman writes:

Social capital, however, comes about through changes in the relations among persons that facilitate action. If physical capital is wholly tangible, being embodied in observable material form, and human capital is less tangible, being embodied in the skills and knowledge acquired by an individual, social capital is less tangible yet, for it exists in the relation among persons. Just as physical capital and human capital facilitate productive activity, social capital does as well. For example, a group within which there is extensive trustworthiness and extensive trust is able to accomplish much more than a comparable group without that trustworthiness and trust. (Coleman 1988, pp. S100–S101, italics in the original)

Compare the situation in the EU labor market with its free movement of labor between the member states. The possibility of realizing this freedom depends to a large extent on the linguistic repertoire of the workers in different countries, see also Till Burckhardt’s paper, chapter “Linguistic Disenfranchisement and Labour Mobility in Europe”, in this book or Aparicio-Fenoll and Kuehn (2016). For a critical discussion of the actual language skills in the EU, see also chapter “The Myth of English as a Common Language in the European Union (EU) and Some of Its Political Consequences” by Barbier in this book. One may speculate over the extent to which acquisition planning—the teaching of English in almost all European schools—contributed to Brexit: Due to the facts that English is spoken in the United Kingdom and that English is the best-known foreign language in other EU countries, the transaction costs for most workers, who want to take advantage of the free movement of labor, are the lowest if they go to Britain. The considerable presence of workers from other EU countries in Great Britain was one of the main arguments for Brexit used by the Brexit proponants.

We assume that a large part of individual preferences is determined by chance, just like many talents.

Compare the analysis in Stiegler and Becker (1977).

It could, however, influence the rate of innovation, thereby altering the state of knowledge.

We could also talk about direct and indirect values of language instead of use and nonuse value. In communication, language plays a crucial rôle but in the provision of cultural identity, many other factors are important such as religion, traditions, etc.

The distinction between autocentric and ecocentric values is not very clear-cut. There is no fundamental difference between “egoistic” and “altruistic” preferences. If I derive the same pleasure from making my friend enjoying my bottle of the 1961 Château Palmer as if I had drunk it myself my pleasure is equally “egoistic” in both cases. The distinction has more to do with who has the control over the action and if the action produces positive externalities (presuming my friend enjoys the wine).

A third type of good, a common, can also be found in the literature. A common is a rival non-excludable good. Common examples are fish in the sea or our environment in general.

For a further principal discussion of these issues, the reader is referred to any of the many standard texts on public economics, for instance Hindriks and Myles (2006).

If we are dealing with a group of people learning a given language, the situation is more complicated. Take the case of providing education in an immigrant language in a school district. After a class in the language in question has been set up, the quality of education can vary with the number of pupils; our hypothesis (which can be tested) is that if an additional pupil comes to a class with only one or two pupils, the quality of the education of the first pupils probably increases, that is, the new pupil constitutes a positive externality. However, if there are already 15 pupils in the class, an additional one could reduce the attention the teacher can give to the other pupils and, hence, cause a reduction in the quality of education; we have a negative externality due to the new pupil.

This argument seems to apply to the distribution of health care in most societies, basically an individual good, but not to the distribution of Porsche cars, another individual good.

We are here neglecting the network-externality property, see below. There are, of course, also arguments of the type that universal education increases the social capital, hence creating a collective good.

Generally, externality also has to do with non-shielding, the impossibility to exclude oneself, in the case of a negative externality and with non-exclusion, the impossibility to exclude someone, in the case of a positive externality.

We are implicitly assuming that the individuals are not altruistic. Altruism could internalize the network externality.

See Church and King (1993) as well as Dalmazzone (1999) for interesting analyses of the consequences for language policy of this property. In de Swaan (2001) a good displaying network externality is referred to as a “hypercollective” good.

Cf. Wickström (2016a).

One might speculate about which policy would be the best to meet this goal. Given that it is illusory to teach all major languages in the schools of all countries, a second-best policy might take its point of departure in the fact that, with few exceptions, the major European languages belong to three big families and are relatively closely related within each family. A sensible policy could then be to make all pupils competent in a Germanic, Romance, and Slavic language. A migrant worker, having learned German, say, would relatively fast acquire a receptive knowledge of Swedish if working in Sweden. The same would hold for someone having learned Polish and working in Croatia, or for someone having learned Spanish and working in Romania. High costs of mobility would then remain if working in countries like Hungary, Finland, Greece, or the Baltic States. Of course, there are many alternative approaches that might reduce the language problem in connection with mobility. A more direct one is combined investments in language skills pre- and post-mobility, see Grin et al. (2002) as well as Chiswick and Miller (2015), or Gazzola (2016a).

This, of course, does not imply that government provision in all such cases is more efficient—or egalitarian—than leaving the issue to individual initiatives. Public choice scholars have pointed out that in many instances public action can be inefficient or even detrimental; see, for instance, Buchanan and Tullock (1962) or Buchanan (1987). This shows the need for evaluation of language policy.

See, for instance, Grin (1994) or Grin and Vaillancourt (1997).

See, for instance, de Briey and Van Parijs (2002) or de Swaan (2001).

See, for instance, Fidrmuc and Ginsburgh (2007).

The closer is one to the given capacity (the number of available physicians, say), the more detracts an additional individual from the consumption of the other individuals. Note that the good here is the availability of a physician with certain medical skills; it is not the availability of the one and only Dr. Smith. The availability of the dermatologist Dr. Smith is an individual good characterized by full rivalry, but the availability of a competent dermatologist, who can be Dr. Smith or any of 15 other physicians, is a good with a certain degree of non-rivalry. Also, since there are set-up or fixed costs that are independent of the number of individuals using the service as well as variable costs directly dependent on the number of users, average costs will in general exceed marginal costs. One often talks about “natural monopolies”. See also Sect. 5 below on the structure of costs.

Cf. Gazzola (2014a, 2016b), who shows that language knowledge in the EU, and as a consequence access to various documents and services of the EU, are strongly correlated with citizens’ education and income. Note that the distributional consequences also depend on the individual’s evaluation of this access. This is an additional dimension to consider in analyzing the distributional consequences of the language policy in the EU.

See Fidrmuc and Fidrmuc (2016).

See Desmet et al. (2009).

See Fidrmuc et al. (2005).

Van Parijs (2008, p. 27).

Recall that we use the word “distribution” as a terminus technicus referring to the distribution of resources (or welfare) between individuals. See Sect. 1.1.

First-best reaction basically means that one looks for the optimal response under the assumption that there are no binding institutional restrictions. A second-best reaction would be the optimal response given that institutional constraints would have to be respected.

Compare the situation in the Spanish Basque area, as described by Cenoz (2008).

Our definition, albeit analytically rather specific, is also quite “narrow”. That this can lead to serious drawbacks is pointed out in chapter “Justice in the Linguistic Environment: Narrow or Wide?” by Andrew Shorten in this book. On the other hand, in the evaluation of the outcomes we are very close to Shorten’s “wide” definition of justice.

Its practical usefulness is also put in doubt by many economists; see, for instance, Ginsburgh’s (2017) very critical assessment.

The implicit income also include intangibles available to the individual and to which he or she attaches a certain value.

See de Scitovszky (1941). For a lucid discussion of the concept of consumer surplus, see Morey (1984). Intuitively, the provision of a certain good or service alters the propensities to pay for this and other goods due to a perceived change in implicit income. In that way, the evaluations ex post and ex ante differ and the evaluation of a policy measure providing a certain good can be different before and after it is implemented.

For example, in court one might have to wait a long time for a trial in the chosen language when the number of cases increase. One observes the same effect in a swim club or golf club when it becomes crowded.

Cf. also François Vaillancourt’s contribution to this book (chapter “Linguistic Justice in a Federal Setting: The Case of Canadian Provinces”), where he, among other things, discusses the ex ante/ex post problematic.

This is a frequent argument in dynamic models. See Templin et al. (2016) for a review of the literature and an example.

This “paradox” has the same structure as the original Scitovsky paradox (de Scitovszky 1941). Their income effects in the different states of the world influence individual preferences, here the “status effects” in different states of the world do the same thing.

The concept of merit good was introduced by Musgrave (1956/1957) in order to justify public intervention when evaluation is not strictly based on individual preferences. The departure from individual preferences could be justified by individuals’ uncertainty or limited access to information, as well as differences between ex post and ex ante evaluation. An example of the latter could be education. After having received an education, I value it more than before I received it.

Compare, however, the efforts to revive languages like Cornish or Manx, which seems to contradict this conclusion.

The end-state argument can be found in Patten (2009), among others.

See, for instance, Abrams and Strogatz (2003).

See Wickström (2005).

See Minett and Wang (2008), Fernando et al. (2010), or Templin et al. (2016).

See, for instance, Cenoz (2008).

Compare chapter “Linguistic Justice in a Federal Setting: The Case of Canadian Provinces” by François Vaillancourt in this book. He finds a lower limit for the propensity to pay by considering the necessary costs that are imposed on an individual in the absence of a given right (for instance, the cost of translation). See also the critical assessment of Ginsburgh (2017).

For a detailed presentation of cost-effectiveness analysis, see Levin and MacEwan (2001).

The policy makers’ decisions are part of the political process and the political pressure from the voters presumably has an influence. In this way, one could say that the propensity to provide language planning measures is indirectly based on the voters politically expressed propensities to pay.

Also intermediate cases are, of course, possible: a fixed sum per measure plus a certain sum per individual beneficiary.

Imagine that there are two minority languages in a country, one spoken by a fairly big community and the other one by a rather small group. If the budget for social services is the same for both language groups, the quality of the services for the speakers of the smaller language would be higher than for the speakers of the larger one.

Opportunity costs is the value of the best alternative we give up in order to realize the policy adopted.

One might think of the example of public signs in a certain language in a given region. Here all costs are fixed and independent of the number of users but vary with the size of the territory. Less clear-cut examples are services of a public office or public education in a given language. Here one part of the costs, like those for producing printed documents or textbooks, are more or less fixed and one part, like the time of the public servants or school teachers, are almost proportional to the number of users of the language. In each case, the cost structure can be assumed to be concave, see below.

We are assuming that no translations from the majority language into the minority language occur.

Again, we are here talking of additional costs of a parallel system. See footnote 98.

For a more detailed discussion, see Wickström (2016b).

See, for instance, Grin and Vaillancourt (1999) for a comparison of the effects of different policies.

Patten’s (2009) accommodation argument is hence extended to cover all individuals who want to take advantage of the right.

See Grin and Vaillancourt (1999), Grin (2003, 2005b), and Grin and Gazzola (2013).

Compare Wickström (2005), Fernando et al. (2010), Wickström (2014), as well as Templin et al. (2016).

To us equality means that speakers of different languages are treated equally. In the literature other definitions of equality can be found, for instance, equality of languages, that is, treating all languages equally independently of the number of speakers. This comes close to Patten’s (2009) end-state argument. See Pool (1987) for a further discussion of this issue.

This is related to Patten’s (2009) context-of-choice and fairness criteria.

One could define the benchmark in a diametrically opposite way. In this case, no rights are inherently in effect, and all provisions of specific rights have to be consciously decided upon. With our chosen benchmark, all individuals are provided with all possible rights to use their language of choice in any social situation, and restrictions have to be motivated. We call this approach “liberalism”. The opposite one could then be labeled “absolutism”. Compare also the basic legal philosophy behind the Anglo-Saxon common-law tradition and the continental one associated with the Code Napoléon. Of course, the two benchmarks are distributionally very different. See also Wickström (2007).

For a more detailed discussion, see Wickström (2016b) and the references therein.

Using the difference between aggregated benefits and implementation costs is an example of potential Pareto efficiency, see Sect. 1.1. Were all types of compensation payments possible, this could bring us to Pareto efficiency.

Compare this situation with the optimal-taxation problem; see, for instance, Hindriks and Myles (2006).

Cf. Haselhuber (2012).

This trade-off is present in this book both in the contribution of Till Burckhardt (chapter “Linguistic Disenfranchisement and Labour Mobility in Europe”), who looks at the unequal “mobility disenfranchisement” that could be overcome with more resources invested in language-learning in European schools at a certain cost, and in chapter “Choosing Working Languages in a Multilingual Organization: A Statistical Analysis with a Particular View on the European Union” by Dietrich Voslamber.

This type of argument can be found in the work of Van Parijs, who argues for the introduction of English as the sole official language in the European Union. He argues that this could be fair if the speakers of other languages are adequately compensated for learning English; see, for instance, Van Parijs (2011). See also Pool (1987, 1991) for the original discussion of the topic.

In the context of language policy, a person who because of insufficient language knowledge cannot communicate with a public institution in one of its official languages. See, for instance, Ginsburgh et al. (2005).

Consider the calculations in Fidrmuc and Ginsburgh (2007). One can look at the analysis here as a cost-effectiveness analysis. The effect is that a given individual has the ability to communicate with the institutions of the European Union (the authors use the expression “disenfranchisement” for the inability to communicate with the EU; other values of language use for the individual, like boosting his or her identity, are ignored in the analysis). This can be achieved by giving various languages an official status. The data are based on EU25 with then 20 official languages. By adding Maltese to a policy with only English, French, and German as official languages, Maltese speakers who do not master English, French, or German will be enabled to communicate directly with the institutions in Brussels. This would, according to the authors, come at an annual cost of 831.30 euros per disenfranchised speaker of Maltese in the absence of official status for Maltese. If we consider a situation where the alternatives are 19 languages (excluding Maltese) or 20 (including Maltese), this sum would increase somewhat, but probably not very much. Assuming that the language planner attaches a value of less than 800 euros to enable the direct communication with Brussels of an average European citizen, it would be efficient not to give Maltese an official status. This situation is, of course, not just, and we have the trade-off between efficiency (no official status for Maltese) and justice (official status for Maltese). The fact that Maltese is an official language in the EU can be interpreted in such a way that the rational planner gives a weight greater than zero to justice in the trade-off between efficiency and justice. (Whether the language policy in the EU is based on rational arguments or not is a different question.)

In the European Union, the knowledge of languages other than the mother tongue, especially the knowledge of English, is as a rule positively correlated with income and education, see Gazzola (2016b). Note that this does not necessarily imply that propensities to pay for an official status of the non-English mother tongues are negatively correlated with income and education; it might be much more important for the rich to communicate with Brussels than for the poor, and, hence, the rich might have a higher propensity to pay for the right to use their mother tongue in such communication in spite of the fact that they could use English at a sufficiently high level. The rich might also have higher propensities to pay in general, because their income is higher and the demand for communication ease is a “normal” good (a good with a positive income elasticity of demand; that is, the demand for the good increases with increasing income).

The intuitive argument that one should “put oneself in the shoes of others” to make fair and impartial decisions—to decide behind a veil of ignorance—goes back at least to Plato (1888, 1980) and has its most prominent modern exponent in Rawls (1971). For a critique of the veil-of-ignorance approach, see chapter “Towards an Adaptive Approach to Linguistic Justice: Three Paradoxes” by Yael Peled in this book.

In chapter “Linguistic Justice in a Federal Setting: The Case of Canadian Provinces” in this book, François Vaillancourt in his conclusions discusses the limits of the possible input of the economist. See also Cardinal and Sonntag (2015).

The combination is found in Finland, for example. There, a critical mass of 3000 people or a fraction of at least 8% of the population in the community is required for giving rights to the local language minority. Romania and Slovakia, in comparison, require 20% of the population for local language rights.

See, for instance, Wickström (2015).

See, for instance, Boadway and Shah (2009).

Currently discussed reforms of the jurisdictions in Finland and Norway seem to distance these countries from this ideal.