Consequences of Linguistic Distance for Economic Growth
| Author | Erkan Gören |
| Published date | 01 June 2018 |
| DOI | http://doi.org/10.1111/obes.12205 |
| Date | 01 June 2018 |
625
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 80, 3 (2018) 0305–9049
doi: 10.1111/obes.12205
Consequences of Linguistic Distance for Economic
Growth*
Erkan G ¨
oren†
†School of Computing Science, Business Administration, Economics, and Law (Faculty II),
Institute of Economics, Carl von Ossietzky University Oldenburg, Building A5, 26111
Oldenburg, Germany (e-mail: erkan.goeren@uni-oldenburg.de)
Abstract
This paper advances a new country-level measure of ethno-linguistic diversity, making use
of Greenberg’s definition of diversity by synthesizing information on the share of different
ethno-linguistic groups in a country’spopulation and, more importantly, information on in-
tergroup linguistic distances derived from a recently developed lexicostatistical approach.
I show that this measure captures ethno-linguistic diversity at lower levels of linguistic
aggregation. However, unlike the commonly used phylogenetic language tree approach,
I found that these distance-weighted diversity measures continue to have a strong neg-
ative statistical association with economic growth that is not sensitive to the underlying
resemblance function between ethno-linguistic groups.
I. Introduction
Measures of ethno-linguistic diversity appear to have strong predictive power in empir-
ical economics and have become standard explanatory variables in the field of growth
economics. One crucial point in the measurement and construction of diversity is the multi-
dimensional concept of intergroup distance between groups. In economics, the lack of ad-
equate quantitative measures of intergroup distances has frequently compelled researchers
to rely on established definitions of ethno-linguistic groups, although some have carefully
constructed their own ethno-linguistic classifications (Alesina et al., 2003; Fearon, 2003).
Furthermore, many researchers have generallyassumed that interg roup distance is constant
and identical across groups precisely because of the difficulty of defining and measuring
this concept.1
JEL Classification numbers: O11, O5.
*I thank J¨urgen Bitzer, Jonathan Temple, and two anonymous referees for their useful comments and suggestions
that substantially improvedthe paper. I am also grateful to session participants at the 28th Cong ress of the European
Economic Association 2013, University of Gothenburg, the 15th annual Conference of the European Trade Study
Group 2013, University of Birmingham, the Annual Conference of the Royal Economic Society 2014, University
of Manchester, as well as seminar participants at the Carl von Ossietzky University Oldenburg 2012 for helpful
comments and suggestions. All remaining errors are my own.
1One argument for treating intergroup distance across groups as constant and equal was put forward by Garc´ıa-
Montalvo and Reynal-Querol (2005), who argued that ‘[
…] the dynamics of the “we” vs. “you” distinction is more
powerful than the antagonism generated by the distance betweenthem’.
626 Bulletin
Considerable research has been done on intergroup distances (Esteban and Ray, 1994;
Ginsburgh, Ortu˜no-Ort´ın and Weber, 2005; Desmet, Ortu˜no-Ort´ın and Weber, 2009), yet
their quantitative measurement still remains a major challenge for empirical economics. It
is a question of considerable importance: without a reliable measure of intergroup distance
it is difficult to understand how intergroup distance affects socioeconomic outcomes (e.g.,
economic growth, civil conflicts, and redistribution) or even to predict the direction of the
effect.
There are currently two main approaches for measuring intergroup distances based on
methods from the field of comparative linguistics. The first uses the concept of ‘phylo-
genetic language trees’ (Fearon, 2003; Desmet et al., 2009) to capture the genealogy and
hence the relationships among languages (e.g. the percentage of common nodes between
two languages in a global phylogenetic language tree) as a proxy for intergroup linguistic
distances. Although this approach has some merits in regard to the measurement of ethno-
linguistic diversity at various levels of linguistic aggregation (Desmet, Ortu˜no-Ort´ın and
Wacziarg, 2012), it has some conceptual shortcomings when it comes to the measure-
ment of intergroup linguistic distances. The second approach is based on the limited and
arguably outdated lexicostatistical measures of the percentage of shared cognates between
language pairs (e.g. meanings from two different languages considered to share a common
etymological origin, based on the expert opinions of comparative linguists) presented in
the article by Dyen, Kruskal and Black (1992).
In this paper, I borrow from an alternative, objective and easy-to-implement statistical
method that calculates intergroup linguistic distances based on recent advances in the field
of comparative lexicostatistical analysis. The Automated Similarity Judgement Program
(hereafter referred to by its abbreviation, ASJP) has developed a statistical method that is
capable of computing intergroup linguistic distances for a large number of languages and
dialects across the world based on a standardized list of meanings. This method makes use
of the concept of the Levenshtein (1966) distance, whichindicates the number of inser tions,
deletions, or substitutions of a single character to transform one word into another. The
larger this number, the greater the linguistic distance. Comparing the linguistic distance data
derived from the proposed ASJP methodology with the commonly employed phylogenetic
language tree approach reveals quite significant differences. In contrast to the phylogenetic
language tree approach, the results show that theASJP methodology is capable of capturing
subtle linguistic differences between ethno-linguistic groups across language families.
I highlight the importance of incorporating intergroup linguistic distances in the mea-
surement of ethno-linguistic diversity across countries. To accomplish this task, I apply
the Greenberg (1956) index of diversity to population share data from the Ethnologue
database (World Language Mapping System, 2009) using linguistic distance data derived
from the proposed ASJP lexicostatistical approach. In addition, I construct two additional
measures of diversity to examine the influence of linguistic distances on the calculation
of distance-weighted diversity measures. The first index is based on the same population
share data from the Ethnologue database, but also employs linguistic distance data derived
from the Ethnologue phylogenetic language tree approach. Again, using the same popu-
lation share data from the Ethnologue database, the second index is related to measuring
ethno-linguistic diversity at various levels of aggregation based on the same methodology
as in Desmet et al. (2012). Contrasting both distance-weighted diversity measures with the
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
Linguistic distance and economic growth 627
measure of ethno-linguistic diversity calculated at various levels of linguistic aggregation
would provide information on the relative importance of incorporating intergroup linguis-
tic distances into the measurement of ethno-linguistic diversity. The comparative analysis
reveals that distance-weighted diversity measures based on linguistic distance data from
the ASJP methodology are consistent with the idea of measuring ethno-linguistic diversity
at relatively lower levels of linguistic aggregation.
After incorporating distances, I revisit the link between ethno-linguistic diversity and
economic growth using a panel of developed and developing countries during the period
from 1960 to 2009. The empirical analysis illustrates that distance-weighted diversity
measures continue to havethe typical negative association with economic growth. However,
unlike the usually employed phylogenetic language tree approach, this association is not
sensitiveto the functional specification underlying the measurement of intergroup linguistic
distances. This finding is robust to a wide range of additional controls and across different
model specifications.
This paper is part of a vast literature debating the impacts of ethno-linguistic diversity
on socioeconomic outcomes such as economic growth and the provision of public goods
(Easterly and Levine, 1997; Alesina et al., 2003; Alesina and La Ferrara, 2005). The im-
portance of incorporating linguistic distance into the measurement of diversity has been
the subject of recent research. For example, Desmet et al. (2009) investigate the impact
of various measures of ethno-linguistic diversity on redistribution in a cross-section of
countries incorporating linguistic distance data derived from the Ethnologue phylogenetic
language tree approach. Moreover, Desmet et al. (2012) examine the relationship between
ethno-linguistic diversity calculated at various levels of linguistic aggregation and a wide
range of socioeconomic outcomes (e.g. civil conflicts, redistribution, provision of public
goods, and economic growth).
To the best of the author’s knowledge, this is among the first papers in economics to use
linguistic distance data derivedfrom the ASJP approach to quantify the impact of intergroup
linguistic distance on economic growth. Other papers that have used this method focused
specifically on the role of intergroup linguistic distance in shaping international migration
patterns (Adser`a and Pytlikova, 2015), on the effectiveness of language acquisition of
immigrants (Isphording and Otten, 2013), and on international trade flows (Isphording
and Otten, 2013; Melitz and Toubal, 2014).
The paper is organized as follows. Section II discusses some important conceptual
differences between phylogenetic language trees and lexicostatistical methods with
respect to the measurement of intergroup linguistic distances. The preferred ASJP method
for quantifying intergroup linguistic distances is discussed in Section III. Section IV
describes the linguistic distance data and presents the new distance-weighted measures
of ethno-linguistic diversity. Section V provides an empirical illustration when incorporat-
ing linguistic distances in the relationship between ethno-linguistic diversityand economic
growth. Finally, Section VI concludes by summarizing the main findings of the paper.
II. Measurement issues affecting linguistic distance
Until recently, diversity (be it ethnic, linguistic, or religious) has typically been measured
using a variation of the Herfindahl–Hirschman-based index without accounting for resem-
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial