Foreign Language Learning and Trade

• Author: Victor Ginsburgh,Jacques Melitz,Farid Toubal
• Published date: 01 May 2017
• DOI: http://doi.org/10.1111/roie.12268
• Date: 01 May 2017

Foreign Language Learning and Trade

Victor Ginsburgh, Jacques Melitz, and Farid Toubal*

Abstract

The paper is devoted to an econometric analysis of learning foreign languages in allparts of the world. Our

sample covers 193 countries and 13 important languages. Four factors significantly explain learning: world

population of speakers of home language, trade with speakers of foreign language, linguistic distance

between home and foreign language and literacy. Trade may well deserve more emphasis than the other

three factors, not only for its significance, but also because its direction can change faster and by a larger

order of magnitude. Controlling for any of the 13 target languages, including English, is of no particular

importance.

1. Introduction

There is wide awareness of the role of foreign trade in stimulating learning of foreign

languages.

1

Yet despite much research to date on the influence of common languages

on foreign trade,

2

the influence of trade on language learning has been studied very lit-

tle.

3

In general, econometric work on language learning has been largely confined to

the decision of immigrants and linguistic minorities to learn the primary language in

their country of residence in order to increase their work possibilities and wages.

4

In this paper, we concentrate instead on learning foreign languages, meaning lan-

guages in common use abroad but not at home. Quite explicitly, we do not treat the

learning of the home language by immigrants and by linguistic minorities, possibly

concentrated in certain regions, like Basque speakers in Spain or Gujarati speakers in

India. This learning is largely a different subject, we think. Most nationals acquire their

language as children, almost without effort, while for residents who do not know the

home language, learning it is likely to dominate learning of any foreign language since

the person may well need the home one for daily living.

We also take a world perspective on learning foreign languages. Thus, we study learn-

ing of 13 languages in 193 countries. These languages are Chinese, English, Spanish, Ara-

bic, Russian, French, Portuguese, German, Malay, Japanese, Turkish, Italian and Dutch,

in descending order of number of speakers. The languages are chosen because of their

importance in either of two senses or both: sheer size or international spread. One of our

basic concerns was increasing the sample variance. Specifically, we wanted to include

* Ginsburgh: European Center for Advanced Research in Economics and Statistics (ECARES),

Universit

e libre de Bruxelles, 50 Av. F. D. Roosevelt, CP 114/04, 1050 Bruxelles, Belgium. Tel: 132-2-

650-3846; Fax: 132-2-650-4012; Also affiliated to CORE, Universit

e catholique de Louvain, Belgium. E-

mail: vginsbur@ulb.ac.be. Melitz: Ecole Nationale de la Statistique et de l’Administration Economique

(ENSAE), 3 Avenue Pierre Larousse, 92240 Malakoff, France. Also affiated to Centre d’ 

Etudes

Prospectives et d’informations Internationales (CEPII), Paris, France and Centre for Economic Policy

Research (CEPR), London, UK. Toubal: Ecole Normale Sup

erieure de Cachan, Cachan 94230, France.

Also affiliated to Paris School of Economics and CEPII, Paris France. The authors would like to thank

the participants in seminars at CREST, Paris, Ekaterinburg, Russia, University of Jaume 1, Castelon,

Spain, and ECARES, Belgium, for valuable comments. They are especially indebted to Julien Martin

and Olivier Loisel for their comments. Farid Toubal wishes to acknowledge the support provided by the

iCODE Institute (Idex Paris-Saclay).

V

C2016 John Wiley & Sons Ltd

Review of International Economics, 25(2), 320–361, 2017

DOI:10.1111/roie.12268

some languages that are large but not widely spread. Chinese and Japanese fit the bill,

especially the latter. We also wanted to have some languages with only a moderate popu-

lation of speakers in any one country but notable international spread. Malay and Dutch

are our best examples. Our data is cross-sectional and centers around 2005. We examine

five separate influences on learning based on elementary theoretical considerations and

the availability of data, and all five of them emerge as important with the expected sign.

Four of these influences entered in earlier theoretical discussion of language learning in a

game-theoretical context (see Selten and Pool, 1991; Church and King, 1993; Shy, 2001).

The first of these four influences is the size of the world population of speakers of a for-

eign language; it encourages learning. The second is literacy (as a reflection of the ability

to learn), which encourages learning too. Third, a large world population of speakers of

the home language discourages learning. People who possess a large language have less

incentive to learn any other one. Fourth, linguistic distances also discourage learning.

When the distance between languages increases, learning decreases. The fifth influence, at

the heart of this study, is trade. Trade with speakers of a foreign language raises learning

of the speakers’ language. Of major note, trade is well correlated with geographical prox-

imity to speakers, common borders with them, past historical ties to them, and current

economic and political associations with them. Therefore its influence could reflect many

factors besides commerce. This is part of the interest of the variable. Separating out these

various sources of the variable’s influence would be a fitting subject for later research.

There are obvious differences between the operations of the five influences. Literacy

encourages learning without regard to the target language. The size of the home lan-

guage discourages learning generally too. By contrast, the size of the population of the

target language, and trade with this particular population, both promote learning of

the particular language. Finally, the linguistic distance between home and target lan-

guage affects the choice of foreign language to learn.

The various motives for language learning that are reflected in trade emerge as the

most important factor in our empirical findings. Conditional on the presence of learners

of a language in a country, a 1 percentage-point increase in the trade share with speak-

ers of the language will increase learners of the language (as a percentage of the total

population) by around 1.4 percentage points. This is a large effect. It emerges after con-

trolling for the reciprocal effect of learning on the trade share; that is, after instrument-

ing the trade share. Without conditioning on positive learning, a 1% increase in the

trade share with speakers will also increase the probability of some positive learning

after controlling for endogeneity. A doubling of the trade share causes a 13% probabil-

ity of some positive learning. This next effect is smaller than the aforementioned one

but as precisely estimated. These are also estimates without separate individual controls

for national specificities. If we control for such specificities by introducing a separate

fixed effect per country, the influence of trade on learning even goes up. It rises from

1.4 to 2.1 percentage points for learning when there is some and from a 13% to a 16%

probability of learning overall.

The paper proceeds as follows. Section 2 provides some theoretical discussion of the

influences on learning that we choose to investigate in the empirical study. Section 3

discusses the econometric model. Section 4 turns to the data and section 5 describes

the estimation method. Sections 6–8 are devoted to results.

2. The Theoretical Background

In explaining language learning on a world basis, the game-theoretical literature pro-

vides a useful point of reference. This literature would suggest treating the total

FOREIGN LANGUAGE LEARNING AND TRADE 321

V

C2016 John Wiley & Sons Ltd

numbers of speakers of the different languages in the world and the costs of learning

as outstanding factors. To explain, assume that there is a single home language per

country and that everyone in each country possesses this language. In each country,

therefore, all people can already communicate with the whole world population of

speakers of their home language, including those who live abroad. The larger this total

world population is, the better able they are to communicate and the lower are their

marginal benefits of learning another language. Therefore, the total world population

of speakers of the home language exerts a negative influence on learning at home.

This theoretical result accords well with common experience. We encounter more

monolingualism in large-language countries, such as the USA or the UK, than in

small-language ones, such as the Baltic countries.

In the case of each country, consider next the total world number of speakers of a

foreign language. The larger this number is, the larger are the benefits of learning the

language. On this ground, Chinese and English should attract more learning than

other languages since they are the two largest in the world. Arabic and Spanish should

rank high too.

However, as the game-theoretical literature stresses (see Gabszewicz et al., 2011;

Selten and Pool, 1991), there are costs of learning, if only the time and effort required

to learn. One international measure of this cost is the distance between the home and

the foreign language. The specific measure of this distance we shall use rests on expert

judgments by ethnolinguists. This should clearly help to reconcile the evidence with

the fact, for example, that there is less learning of Chinese, Japanese and probably

Arabic in countries using Indo-European languages than sheer numbers of speakers

would lead us to expect. Another indicator of the cost of learning is the literacy rate.

Higher literacy, implying higher education, should make learning easier and thereby

promote learning of foreign languages.

If we limit ourselves to the preceding influences on learning, however—namely,

world numbers of respective speakers of home and target language, linguistic distance

and literacy—we miss one important dimension, if not several. Consider the relative

interest of learning English, French and German in a Spanish-speaking country in

South America as opposed to Spain. For both countries, world speakers of these lan-

guages and the linguistic distances between them are exactly identical

5

and differences

in the literacy rate cannot explain a preference for learning one language rather than

another. Yet it is evident that French and German are of greater interest relative to

English in Spain than South America because of higher levels of interaction with

French- and German-speaking countries in Spain. Introducing bilateral trade ties

between countries is one important means of repairing this difficulty. It is not the only

one. We tried others that proved of less interest, as we will see. Also, two of the other

possible influences—geographical distance to speakers and common border with

speakers—could not be combined with trade, though they mattered, since they serve

as instruments for trade. They are thus only reflected through trade. As already

emphasized, bilateral trade ties reflects many things, not only commercial ties but also

geography, politics and possibly history, for example, a common ex-colonial past. In

addition, the impact of this variable is multi-faceted. From the standpoint of people as

consumers, trade ties are an additional reflection of the benefits of learning foreign

languages besides the world number of speakers. Indeed, in the presence of trade ties,

the total world population of speakers of a target language is possibly best interpreted

as reflecting non-market advantages of learning this language, stemming from the abil-

ity to interact socially with foreign speakers and to benefit from their cultures and cul-

tural heritages (though trade ties will cover these advantages to a non-negligible

322 Victor Ginsburgh, Jacques Melitz, and Farid Toubal

V

C2016 John Wiley & Sons Ltd