Introduction
When the Gambia, a small British colony surrounded by French settlements, obtained independence in 1963, it chose English as its official language. In 2014, President Yahya Jammeh adopted Arabic as the new language for official purposes and education, stating, ‘We no longer believe that for you to be a government you should speak a foreign language; we are going to speak our own language’.Footnote 1 However, Arabic is not the language of any Gambian ethnic group.Footnote 2 The decision was politically and religiously motivated since Jammeh was pivoting the country towards Arabic allies and, in 2015, proclaimed the Islamic Republic of Gambia.Footnote 3 This decision was reversed in 2017 by the new President Adama Barrow.Footnote 4
This paper explores the relationship between the official language of a country and accountability, arguing that the former can be one of the factors that help explain variations in the latter. Typically, a country’s official language is used in government, judiciary, and administration, while accountability refers to the constraints imposed on political power by the ruling elite.Footnote 5 Effective accountability requires that these elites justify their actions and face potential sanctions from the citizenry (Luhrmann et al., Reference Luhrmann, Marquardt and Mechkova2020). The Gambian case exemplifies several key elements we consider and puts forward the political role of languages.
In this study, we emphasise the relational nature of accountability, where the obligation of elites to inform, justify, and face the consequences of their actions requires frequent interaction with the population (Schedler et al., Reference Schedler, Diamond and Plattner1999). We argue that the official language is crucial for the functioning of accountable institutions, as it serves as the primary channel for communication between the elites and the people. To explain the role of the official language in holding elites accountable, we draw on two interpretations of the function of language in social interactions. First, language is a tool for communication. Second, it shapes the cultural patterns that form the basis of interactions between individuals. A language distant from those spoken by the majority complicates the understanding of elite behaviour and the formal processes for sanctioning that behaviour. In addition, it hampers the formation of shared cultural patterns that impose constraints on elite power.
To test our hypothesis that linguistic distance decreases accountability, we have compiled a cross-sectional dataset covering 147 countries. We used various accountability measures from multiple sources and quantified the linguistic distance using the Average Distance from the Official Language (ADOL) proposed by Laitin and Ramachandran (Reference Laitin and Ramachandran2016). We examined this relationship within a model that incorporates several factors explaining cross-country variation in accountability. Our analysis reveals that greater linguistic distance is consistently associated with lower levels of accountability across all measures examined, and these findings remain robust after controlling for other variables. Furthermore, we investigated the role of secondary and tertiary education in mediating the impact of linguistic distance on accountability, demonstrating that higher education levels may mitigate the negative effects of a foreign official language.
The novelty of the paper lies in the link between language and accountability. Economists have addressed language-related topics in various ways (Ginsburgh and Weber, Reference Ginsburgh and Weber2020). First, in international trade, language similarity facilitates exchanges (Melitz, Reference Melitz2008). Second, in healthcare, government guidance in the common language is more effective (Djité, Reference Djité2008; Gomes, Reference Gomes2014). Third, instruction in a non-native language impairs student performance (Laitin and Ramachandran, Reference Laitin and Ramachandran2022). Fourth, assimilation policies that require education in a foreign language can lead to in-group behaviour (Bisin et al., Reference Bisin, Patacchini, Verdier and Zenou2011; Fouka, Reference Fouka2020). Fifth, ethnolinguistic fractionalisation, which involves the number, size, and location of cultural groups, has been associated with civil conflict and low economic growth (Alesina et al., Reference Alesina, Devleeschauwer, Easterly, Kurlat and Wacziarg2003; Easterly and Levine, Reference Easterly and Levine1997). Sixth, studies that examine the effect of culture on economic outcomes often use linguistic variables as proxies for cultural values (Galor et al., Reference Galor, Özak and Sarid2018; Licht et al., Reference Licht, Goldschmidt and Schwartz2007; Tabellini, Reference Tabellini2010). Our work adds a new perspective by considering how the official language shapes the relationship between elites and citizens.
Linguistic anthropology offers a radical view of the link between language and power. Although most linguists follow Saussure (Reference Saussure1986) and Chomsky (Reference Chomsky1965, Reference Chomsky1986) in studying language as formal structures, linguistic anthropology sees it as a social action, a cultural resource, and a set of sociocultural practices (Schieffelin, Reference Schieffelin1990, p.16). People do things with words (Austin, Reference Austin1962; Searle, Reference Searle1969). Linguistic anthropologists (Brenneis and Macaulay, Reference Brenneis and Macaulay1996; Duranti, Reference Duranti1997; Hanks, Reference Hanks1996) view language, spoken or written, as embedded in sociocultural relations. Words are non-neutral because ‘All words have the taste of a profession, a genre, a tendency, a party, a particular work, a particular person, a generation, an age group, the day and hour’ (Bakhtin, Reference Bakhtin1981, p.238). Unequal power relations can lead to symbolic violence, as seen when one dialect is considered superior, giving social significance to arbitrary differences (Bourdieu, Reference Bourdieu1991, p.170). Thus, language and power are intertwined, with language reflecting and potentially altering social reality (Gumperz and Levinson, Reference Gumperz and Levinson1996; Hill and Mannheim, Reference Hill and Mannheim1992; Kaschula and Wolff, Reference Kaschula and Ekkehard Wolff2020; Spender, Reference Spender1980).
The rest of the paper is structured as follows: Section ‘Official language and accountability’ explains how language impacts accountability, reviewing the relevant literature. Section ‘Historical background: official language, colonization, and decolonisation’ provides a historical background. Section ‘Data’ presents the data. Section ‘Empirical strategy & results’ shows the empirical analysis, and Section ‘Robustness checks’ provides robustness checks. Section ‘Potential mechanisms: education’ discusses the role of education, and Section ‘Conclusions’ concludes. Supplementary Material, available in an online repositoryFootnote 6 , contains further robustness checks and information.
Official language and accountability
Historically, the adoption of an official language has been conceived as a relevant part of the homogenisation and standardisation process that allows the control of a territory and the people who live on it (Blanc and Kubo, Reference Blanc and Kubo2024). This language becomes the means of communication between citizens and between citizens and the state.
However, countries face significant linguistic diversity within their territory. According to UNESCO (2009), 118 of the 209 countries have a linguistic diversity score of at least 0.33, and 88 are above 0.5.Footnote 7 To address this diversity, countries may adopt different models (Cabrera, Reference Cabrera2024), ranging from monolingual models that use a single shared language for institutions to multilingual models that allow the use of more languages in public administration. However, multilingual models face relevant transaction costsFootnote 8 but have the advantage of representing the majority of the population (Ginsburgh and Weber, Reference Ginsburgh and Weber2005).
Furthermore, extensive linguistic diversity, along with indigenous languages lacking written traditions (Laitin and Ramachandran Reference Laitin and Ramachandran2025), can hinder the use of multilingual models. In India, 22 indigenous languages are co-official, yet English remains dominant in education, the judiciary, and administration. In Algeria, most people speak a dialect different from classical Arabic, which has led to resistance against the use of classical Arabic in schools; French remained the dominant language for years despite efforts to replace it. These cases show that, despite linguistic diversity and multilingual policies, one official language often dominates state-citizen interactions.
Our argument unfolds in two parts to explain how an unfamiliar official language weakens constraints on elite power. First, it limits citizens’ understanding of key information needed for accountability. Second, it hampers the development of shared cultural values that underpin accountable institutions.
Information
Regarding the first line of argument, we start our reasoning by stressing that information is an important component of accountability (Goetz, Reference Goetz2008; Schedler et al., Reference Schedler, Diamond and Plattner1999; Williams, Reference Williams2015).Footnote 9 The public’s ability to scrutinise and hold the elite accountable depends on the accessibility of information regarding their conduct.Footnote 10 Citizens’ access to information is essential for accountability, from elections to interactions with state administrative bodies. While elections theoretically allow voters to replace governments not following mandates, accountability requires an informed electorate aware of government actions. Informed voters use available information to decide whether to support the incumbent or the opposition. However, without insights into the consequences of prevailing policies, voters may not penalise the incumbent (Kolstad and Wiig, Reference Kolstad and Wiig2009). Ferraz and Finan (Reference Ferraz and Finan2008) showed that disseminating information on corruption through randomised municipal audits in Brazil significantly reduced the re-election prospects of corrupt mayors.
Furthermore, the relevance of comprehension of information for accountability extends beyond the electoral process. The widespread diffusion of information acts as an incentivising mechanism for elites to pursue policies beyond narrow self-interest (Gavazza and Lizzeri, Reference Gavazza and Lizzeri2009; Reinikka and Svensson, Reference Reinikka and Svensson2004, Reference Reinikka and Svensson2005). An empirical investigation in Uganda by Reinikka and Svensson (Reference Reinikka and Svensson2011) underscores that furnishing systematic information to monitor local officials’ administration of a substantial education grant programme curtails the misappropriation of public funds and increases public expenditure in education.
Building on these arguments, we argue that a limited understanding of the official language reduces citizens’ ability to grasp information about elite behaviour, weakening their monitoring capacity. Prior work has shown that language is a major barrier to accessing key information and delivering public services. For example, health programmes often fail when people cannot understand medical instructions (Djité, Reference Djité2008; Gomes, Reference Gomes2014).
Second, limited proficiency in the official language hinders access to and fair treatment within the court system. Zhang and Lee (Reference Zhang and Lee2020) found that in 18th-century France, improving citizens’ command of standard French enhanced their ability to navigate the judicial system without intermediaries. Language barriers have also been linked to under-reporting of crime (Pogrebin and Poole, Reference Pogrebin and Poole1990). In Senegal, Dissake (Reference Dissake2022) notes that, despite hearings being in French or English, many litigants lack proficiency in either. Similarly, studies of the U.S. court system show that many immigrants struggle with English, and inadequate translation services limit access to a fair trial (Davis, Reference Davis1985; Urbina, Reference Urbina2004).
Third, linguistic barriers can hinder democratic participation. Individuals from linguistic minorities may be discouraged from voting if voter information, ballots, and assistance are not available in a language they understand. In the U.S., about eight million citizens are estimated to have limited English proficiency for voting (US Commission on Civil Rights, 2018). Studying South Africa, McLaughlin (Reference McLaughlin2015) finds that voter turnout is higher in municipalities with inclusive language policies.
It is worth noting that language can be an obstacle to human capital formation. The literature consistently finds that receiving education in a language other than one’s mother tongue negatively affects student performance and educational attainment, as it introduces additional cognitive challenges (Laitin and Ramachandran, Reference Laitin and Ramachandran2022). Jain (Reference Jain2017) shows that Indian districts where the language of instruction differed from the local language had lower literacy and college graduation rates compared to districts where it matched the mother tongue.
Several studies have examined the impact of language-related education reforms. Eriksson (Reference Eriksson2014) finds that the Bantu Education Act, which mandated eight years of mother tongue instruction for all Black students in South Africa, improved literacy, educational outcomes, and ultimately wages among the Black population. Ramachandran (Reference Ramachandran2017) shows that introducing mother tongue instruction in primary schools for Ethiopia’s largest ethnic group increased reading skills and the likelihood of completing primary education. Similarly, Jain (Reference Jain2017) finds that aligning the medium of instruction with students’ home language led to long-term improvements in educational attainment.
Culture
In linking official language to accountability, we view language not only as a communication tool but also as central to shaping and transmitting cultural norms. This connection is supported by a long-standing literature, beginning with Whorf (Reference Whorf1956) and Sapir (Reference Sapir1985), which argues that language reflects the worldview of its culture. Language thus shapes the cultural categories that influence identity and behaviour. Nisbet (Reference Nisbet2003) illustrates this by showing that multilingual individuals respond differently to the same question depending on the language used. At a macro level, Kashima and Kashima (Reference Kashima and Kashima1998) find that grammatical rules, such as pronoun dropping, are correlated with cultural traits like individualism, indicating that language structure can shape self-other perceptions.
Two seminal papers in the economics of culture, Licht et al. (Reference Licht, Goldschmidt and Schwartz2007) and Tabellini (Reference Tabellini2010), use the language spoken in a country as a key explanatory variable for the culture of the country. In motivating this choice, Licht et al. (Reference Licht, Goldschmidt and Schwartz2007, p.672) argue that the ‘grammar of a language may transmit and reproduce culture and social categories’. These papers gave stimulus to a growing literature that identifies language as an important determinant of culture (Chen, Reference Chen2013; Gay et al., Reference Gay, Hicks, Santacreu-Vasut and Shoham2018; Hicks et al., Reference Hicks, Santacreu-Vasut and Shoham2015; Santacreu-Vasut et al., Reference Santacreu-Vasut, Shenkar and Shoham2017, Reference Santacreu-Vasut, Shoham and Gay2013).
Based on this literature, we argue that the official language of a country is one of the factors that leads to the formation of shared cultural values among the people who live there. Conversely, the inability of the majority of the population to speak the official language creates a cultural gap between the people and the elite who are educated and imbued in the culture conveyed by the official language.
In linking culture to accountability, we highlight how a lack of shared linguistic and cultural background can affect institutions. Drawing on North (Reference North1990), we view institutions as ‘human-devised constraints that shape human interaction’ (North, Reference North1990, p. 3), comprising both formal and informal rules. Formal institutions include codified laws and regulations enforced by governments and legal systems. Informal institutions are unwritten norms and conventions embedded in social and cultural practices. Crucially, North argues that formal institutions often rely on informal ones: for example, a legal system may fail if cultural norms do not support compliance with the rule of law.
We argue that the lack of a shared cultural background between elites and citizens weakens the informal foundations of accountable institutions. These institutions rely on common norms that constrain elite behaviour and help citizens interpret their relationship with power. When such norms are shared, people intuitively grasp what is expected from elites and the consequences of challenging misconduct. In contrast, when values are unfamiliar, citizens cannot discern the principles guiding elite actions. This gap between formal and informal institutions reduces electoral participation and weakens democratic mechanisms to check power. Even where constitutions mandate accountability, their impact is limited if expressed in a language distant from everyday communication.
Previous studies support the link between culture and institutions, as well as the connection between culture and accountability (Alesina and Giuliano, Reference Alesina and Giuliano2015; Bisin and Verdier, Reference Bisin and Verdier2017; Guiso et al., Reference Guiso, Sapienza and Zingales2015; Kaasa and Andriani, Reference Kaasa and Andriani2022). A common point in all these studies is that institutions reflect the culture, and that to be effective, institutional reforms must consider the cultural context.
Therefore, based on these considerations, our hypothesis is that:
H1: Linguistic distance between the official language and the language(s) spoken by the majority of the population reduces accountability.
Historical background: official language, colonization, and decolonisation
This section provides a historical overview of colonial and post-colonial language policies to illustrate how linguistic separation and hierarchies were established and maintained. Understanding this background is pivotal for our argument that linguistic distance influences accountability. The enduring effects of these policies created social and political divisions by privileging certain languages and groups while excluding others, thereby shaping patterns of access to power and participation in governance. By tracing these developments, we shed light on how linguistic exclusion has been historically embedded and continues to affect the capacity for accountable governance today.
Missions, colonies, and languages
Colonial governments co-opted educated indigenous people to translate and help run the local government and chose small, educated elites they were able to control and tie to the interests of the colonial state. In addition, colonial governments had little interest in mass education since it was expensive, reduced revenues, and potentially fostered rebellion. If education was encouraged, government officials primarily backed a practical variety, such as carpentry, masonry, and horticulture (Furley and Watson, Reference Furley and Watson1978; Kelly, Reference Kelly and Kelly2000a,c; Manning, Reference Manning1998; Sundkler and Steed, Reference Sundkler and Steed2000).Footnote 11
For example, in Vietnam, the French colonial government closed indigenous schools, pressured the Japanese government to prevent the Vietnamese from acquiring education in Japan, and educated only as many Vietnamese after primary school as the colonial government could hire and control (Kelly, Reference Kelly and Kelly2000a,c). The French focused education on practical skills (particularly farming) and did not provide the Vietnamese with skills to let them compete with French settlers for senior positions in the colonial administration. Throughout Africa, the French educated only a tiny elite, purposely trained to be separate from other Africans in language and culture (Grier, Reference Grier1999; Kelly, Reference Kelly and Kelly2000b,d). Similarly, the Italians, Portuguese, and Spanish also educated only a small portion of the non-European population in their colonies (Isichei, Reference Isichei1995).
The British made little effort to educate colonial subjects since they tried to run their colonies as cheaply as possible (Ferguson, Reference Ferguson2002; Frankema, Reference Frankema2012). Enslavers and those who used unskilled and forced labour were especially averse to education, often even refusing to teach enslaved people how to read (Blouet, Reference Blouet1990). For example, before 1813, the British East India Company schools in India trained a total of only a few hundred students, almost exclusively elite Muslims and Hindus from the highest castes, and lower-caste Hindus were explicitly excluded from company schools (Ingham, Reference Ingham1956).
Protestant missionaries were instrumental in administering colonies cheaply. They promoted education aligned with British interests, offering incentives to missionaries who educated a few students intensively in English with a utilitarian curriculum. Mass education efforts were steered toward practical skills (Furley and Watson, Reference Furley and Watson1978; Ingham, Reference Ingham1956). This system produced a small elite and a broader class of English-speaking labourers trained for roles needed by British firms, such as planters and tradespeople.
Patterns of adoption during decolonisation
Decolonisation created room for a new language policy once the imperial powers left their colonial possessions. Patterns of adoption of new languages have been different around the world. In Indochina, French was quickly replaced as an official language by the local national languages. In North Africa, post-independence policies attempted to impose Arabisation, obstructed by the educational heritage of the French system, particularly in the sciences and the universities, and the loyalty of part of the population to Berber, a pre-Arabic language spoken by populations historically hostile to Arabisation. The result of independence was to replace the hegemony of colonial rule with partial success in imposing a new national dominant language (see Spolsky (Reference Spolsky2018) for an extensive review).
Things were more complicated in Sub-Saharan Africa (SSA) and East Asia. In SSA, around eight hundred distinct languages are spoken, none by more than 8% of the population, as with Swahili and Hausa (Ornstein, Reference Ornstein1964). In this context, revolutionary leaders needed to communicate both domestically and internationally, choosing French and English (the languages of the former colonisers) as the ‘national’ languages, despite only 1 to 10% of the population speaking these Western tongues. A similar pattern occurred with Portuguese in SSA and Latin America (Spolsky, Reference Spolsky2017).
In India, there are a hundred major languages or dialects, and the new Republic of India chose Hindi as the overall medium, together with a subsystem of twelve regional co-official tongues.Footnote 12 English was kept as an unofficial junior partner with Hindi, but there was strong resistance to abandoning it since the reliance upon English in government, industry, research, and higher education was significant. Besides, in southern India, where Telugu and other Dravidian (non-Indic) tongues are spoken, there was strong resentment against ‘Hindi imperialism’, and English was favoured (Ornstein, Reference Ornstein1964).
Indeed, a major reason for choosing an alien language is its neutrality. In a multilingual nation, unless one language group is numerically and/or politically dominant, selecting an outside language as official may be a fair and workable solution. All indigenous groups ‘start at zero’, since learning the alien language is equally onerous for all groups. This can even reduce the risk of internal conflict. However, an equal ‘beginning’ is unlikely since some group members already know the new official language, with all the benefits (for example, in terms of education) that this entails.
Language and elite closure
Over time, the continued dominance of colonial languages not only shaped language policy but also became a powerful tool for social differentiation. The continued attractive status of the colonial language and the privileges it conferred decreased the acceptance of the need to develop language education policies that recognise existing diversity, so that a large proportion of the population continued to be blocked from access to economic and political power. In many circumstances, local languages were stigmatised and banned from school use, reducing their status even in the eyes of their speakers. These features were described by Siachitema (Reference Siachitema and Cheshire1991) on language attitudes in Zambia in institutional participation, where English is the official language. Spolsky (Reference Spolsky2018) provides several examples of language policy during and after the French colonisation.
The elite were distinguished by their use of the colonial language, which served to set them apart. Myers-Scotton (Reference Myers-Scotton1993), in her model of elite closure, identifies three ways in which elites are distinguished from others:
1. If the elites are fluent in a language different from those that others know, this fluency is an identifying feature of elites.
2. If elites use, particularly in public situations, a language different from the one others use, their use patterns are a means of identifying elite members. This gives a utilitarian value to the language and its use.
3. If the way elites speak is judged positively and therefore has psychological value, they are incentivised to retain this evaluation by ensuring that their linguistic abilities and use patterns remain distinctive.
In Kenya, English serves as the elite language in government and education, with English–Swahili switching used by civil servants to exclude less-schooled speakers (Myers-Scotton, Reference Myers-Scotton1993). In Tanzania, the shift from Swahili in primary to English in secondary school limits rural advancement and concentrates public jobs among Anglophones (Roy-Campbell and Qorro, Reference Roy-Campbell and Qorro1997). Under apartheid, South Africa enforced Afrikaans and English for university access and top jobs; English still dominates law, finance, and education post-1994 (Webb, Reference Webb2002). In Zambia’s Copperbelt, Town Bemba spread as an urban lingua franca, but Standard English remains key for white-collar jobs, reinforcing labour divides (Spitulnik, Reference Spitulnik1996).
In Senegal, French dominates administration and post-primary education, limiting rural Wolof speakers’ advancement and reserving civil-service jobs for urban francophones (Albaugh, Reference Albaugh2014). Côte d’Ivoire shows a similar pattern: capital-city French signals elite status, while Mandé and Kru speakers are channelled into the informal sector (Calvet, Reference Calvet1999; Djité, Reference Djité2008). In Mali, French-only legislation deepens the gap between the state and citizens, allowing a schooled elite to capture rents (Djité, Reference Djité2008). In the Democratic Republic of Congo, the colonial legacy of the ‘évolués’ persists: French fluency remains key to accessing state jobs and the formal mining sector, while Lingala and Swahili speakers are largely confined to subsistence roles (Albaugh, Reference Albaugh2014). In Cameroon, unequal access to francophone schools and French-dominated civil-service exams reinforce regional inequalities and limit upward mobility (Calvet, Reference Calvet1999; Djité, Reference Djité2008).
Data
This section introduces the dataset used in the empirical analysis. Section ‘Dependent variable: accountability’ details the dependent variables, while Section ‘Independent variable: Linguistic distance’ discusses the independent variable. The control variables are reported in Table 1 with further details in Section A1 of the Supplementary Material.
Control variables overview
Dependent variable: accountability
Measuring accountability is challenging, particularly when examining it across multiple countries. To capture various dimensions of accountability, we use several measures from different sources.
The first variable is Executive Constraints (xconst) from the Polity 5 dataset.Footnote 13 This variable is widely used in the literature (Chandra and Rudra, Reference Chandra and Rudra2015), and it captures the institutionalised constraints on the decision-making powers of chief executives, whether individuals or collectives.
The second and third variables are drawn from the V-Dem dataset.Footnote 14 These variables are the Vertical Accountability index (v2x_veracc) and the Diagonal Accountability index (v2x_diagacc). Both capture different aspects of accountability related to Executive Constraints. The Vertical Accountability index measures the extent to which citizens can hold the government accountable, such as through the ability to organise political parties freely or participate in free and fair elections. In contrast, the Diagonal Accountability index emphasises civil society rather than political institutions, assessing how citizens, civil society organisations, and independent media can hold the government accountable.
Lastly, we extract a fourth variable from the V-Dem dataset, called Engaged Society (v2dlengage). This variable is primarily concerned with public engagement and assesses the level of public involvement in the decision-making process, especially when significant policy changes are being considered. It measures how actively citizens and various societal groups shape policy outcomes and influence governmental decisions. This includes evaluating mechanisms for public consultation, the extent of civil society’s input, and the effectiveness of platforms that allow for citizen feedback and participation in critical policy discussions.
All measures are at the country level, averaged over 2000-2018. This time frame was chosen for two main reasons. First, averaging helps address missing data by broadening coverage and improving reliability. Second, we avoid periods of major disruption, such as the COVID-19 pandemic or the Soviet Union’s collapse, that could distort institutional dynamics. Excluding such events helps ensure a more stable and representative view of accountability. Other measures are listed in Table 2, with further details in Section ‘Different measures of accountability’.
Summary statistics
Independent variable: Linguistic distance
To measure the distance from the official language, we employ the ADOL indicator, proposed by Laitin and Ramachandran (Reference Laitin and Ramachandran2016). This indicator measures the linguistic distance between language groups within a country and the official language. ADOL for a given country
$ i$
is calculated as:
$$ADO{L_i} = \mathop \sum \limits_{j = 1}^n {P_{ij}}{d_{jo}}$$
where
$ n$
represents the number of linguistic groups in the country,
$ {P}_{ij}$
is the population share of the group
$ j$
in country
$ i$
. Finally,
$ {d}_{jo}$
measures the distance of the language of group
$ j$
from the official language
$ o$
.
To assess the distance between two languages, the index, in line with existing literature (Fearon and Laitin, Reference Fearon and Laitin2000; Fearon, Reference Fearon2003; Fearon and Laitin, Reference Fearon and Laitin1999; Laitin, Reference Laitin2000), relies on linguistic trees. These tree diagrams classify the structural relationship between languages using nodes, which signify the evolutionary points of language development and differentiation. Formally, the linguistic distance between two languages is computed as:
$${d_{ij}} = 1 - {\left({{no.\ common \ nodes\ ij} \over {1/2(no.\ nodes\ language\ i + no.\ nodes\ language\ j)}}\right)^\lambda }$$
Here,
$ {d}_{ij}$
equals one when there are no common nodes between the two languages, signifying maximum difference. Conversely, the measure approaches zero as the Number of common nodes increases, indicating minimal difference. A value of
$ \lambda $
equal to 0.5 is selected following Fearon (Reference Fearon2003) and other studies.
Figure 1 shows the correlation between our accountability measures and linguistic distance at the country level. The scatterplots indicate that, despite considerable variation across countries, there is, on average, a negative correlation between accountability and ADOL.
Relationship between accountability measures and ADOL.
It’s worth noting that ADOL differs conceptually from the commonly used indices of linguistic diversity (Alesina et al., Reference Alesina, Devleeschauwer, Easterly, Kurlat and Wacziarg2003; Desmet et al., Reference Desmet, Ortuño-Ortín and Weber2009; Esteban et al., Reference Esteban, Mayoral and Ray2012; Greenberg, Reference Greenberg1956). ADOL focuses on the top-down relationship between the elite and the citizens. Ethnolinguistic fractionalisation (ELF) indices measure linguistic heterogeneity within the total population. To put it differently, while the first is concerned with the vertical relationship between the people and the ruling class, the second is about diversity among the different groups that form the citizens of a country. Although these aspects may sometimes lead to similar dynamics, they also reflect distinct underlying forces.
For example, Argentina exhibits low values in both ADOL and linguistic fractionalisation due to the predominance of Spanish as the official language spoken by the majority. In contrast, Chad scores high on both indices, as it has two official languages (French and Modern Standard Arabic) and over 120 indigenous languages. Chadian Arabic, a vernacular variant, functions as a lingua franca and commercial language spoken by 40–60% of the population. French is widely spoken in major cities and among most men in the southern region, serving as the primary language of instruction (Ethnologue, 2023).
Conversely, Angola presents low linguistic fractionalisation but high ADOL, as Portuguese is the sole official language, while 46 other languages, mostly Bantu, are spoken. Six of these – Portuguese, Chokwe, Kikongo, Kimbundu, Oshiwambo, and Umbundu – hold institutional status (Ethnologue). A contrasting case is Lesotho, where the Constitution recognises two official languages (Sesotho and English). Sesotho is the first language of more than 90% of the population and is widely used for communication, whereas English is reserved for official interactions. Minority languages include Zulu, Phuthi, and Xhosa (Baker and Prys Jones, Reference Baker and Prys Jones1998).
Therefore, although ADOL captures a distinct concept from ELF indices, we also control for different measures of linguistic fractionalisation to understand how it may impact the relationship between ADOL and accountability.
In Table 1, we present all the control variables. For a more detailed description, please see Section A1 in the Supplementary Material.
Table 2 presents the summary statistics for our main variables. Some additional variables are also in this table, but they will be discussed in Sections ‘Robustness checks’ and ‘Potential mechanisms: education’.
Empirical strategy & results
To empirically investigate the relationship between the distance between the official language and the languages spoken within a country and accountability, we use the following equation:
$$ {A}_{i}={\alpha }_{0}+{\alpha }_{1}ADO{L}_{i}+{\alpha }_{2}{X}_{i}+{\epsilon }_{i}$$
where the index
$ i$
represents different countries.
$ {A}_{i}$
is the chosen measure of accountability (as shown in Section ‘Dependent variable: accountability’),
$ ADO{L}_{i}$
refers to our proxy for linguistic distance, and
$ {X}_{i}$
is a vector of controls, as discussed in Table 1. In Table 3, we present our main findings.Footnote
15
Main results
Notes: robust standard errors in parentheses and t-statistics in brackets. * p < 0.1, ** p < 0.05, *** p < 0.01. Controls includes: Pop (log), GDP per capita at independence (log), GDP growth (%), Arable Land area (log), Oil rents (%. gdp), External confl., Unitary, Nr. languages (log), British Legal origin, French Legal origin, Socialist Legal origin, Previous Colony, Absolute latitude (log). All observations are at the country level.
The results show a negative and statistically significant relationship between distance from the official language and accountability. Specifically, greater linguistic distance is associated with lower levels of accountability across all the measures we examine. This finding also holds after accounting for control variables.Footnote 16
In terms of the magnitude of the effect, the results indicate that a 1% increase in the ADOL corresponds to a 1.28% decrease in Executive Constraints (Column (2)). Similarly, a 1% increase in ADOL leads to a 0.61% decrease in the Vertical Accountability Index (Column (4)). Additionally, the Diagonal Accountability Index (Column (6)) shows a 0.51% decrease for each 1% increase in ADOL, and the Engaged Society index (Column (8)) decreases by 0.81% with a 1% increase in ADOL.
Robustness checks
To check the robustness of our results, we replicate the analysis in Eq. (1) using different proxies for accountability, our dependent variable (Section ‘Different measures of accountability’), and different measures of linguistic diversity, a key control in our analysis (Section ‘Different measures of linguistic diversity’). Further robustness checks are reported in the Supplementary Material.
Different measures of accountability
Measuring accountability is complex, and as discussed in Section ‘Dependent variable: accountability’, various indicators capture its different dimensions. To ensure the consistency of our results, we replicate the analysis using alternative dependent variables that represent distinct aspects of accountability. Specifically, we utilise three measures: Voice & Accountability, Accountability Index, and Reasoned Justification.
Voice & Accountability is taken from the World Governance IndicatorsFootnote 17 and measures perceptions of citizens’ ability to participate in government, along with freedoms of expression, association, and the media. Averaged at the country level from 2000 to 2018, the index ranges from -2.5 (low accountability, e.g., dictatorships) to +2.5 (high accountability, e.g., democracies).
Accountability Index and reasoned justification are average measures from V-Dem between 2000 and 2018. The first variable generalises the Vertical Accountability Index and Diagonal Accountability Index, interpreting accountability as constraints on government power through justification and sanctions. The second variable, reasoned justification, measures how political elites justify their positions on major political changes. The findings are presented in Table 4.Footnote 18
Robustness checks using alternative measures of accountability
Notes: robust standard errors in parentheses and t-statistics in brackets. * p < 0.1, ** p < 0.05, *** p < 0.01. Controls includes: Pop (log), GDP per capita at independence (log), GDP growth (%), Arable Land area (log), Oil rents (% gdp), External confl., Unitary, Nr. languages (log), British Legal origin, French Legal origin, Socialist Legal origin, Previous Colony, Absolute latitude (log). All observations are at the country level.
Overall, the results in Table 4 are always negative and statistically significant and in line with Table 3. In terms of magnitude, an increase of 1% in ADOL generates a decrease of 0.9% in the (average) Voice & Accountability (Column (2)), of 0.67% in the (average) Accountability Index (Column (4)), and of 0.57% in the (average) Reasoned Justification (Column (6)).
Different measures of linguistic diversity
In our main model in Equation (1), we used the number of languages spoken in a country as our control for a country’s linguistic diversity.Footnote 19 In this section, we substituted this proxy with an index of Ethno-Linguistic Fractionalisation (ELF with distance) collected from Laitin and Ramachandran (Reference Laitin and Ramachandran2016). This index measures the probability of two randomly chosen individuals speaking different languages in a country, including the linguistic distance between all ethnic group dyads in that country.Footnote 20 Results reported in Table 5 show that the relationship between the chosen measures of accountability and ADOL remains negative, in line with the main findings. Moreover, they are all statistically significant, except in Column (6), where the significance drops.Footnote 21
Robustness checks using ELF with distance
Notes: robust standard errors in parentheses and t-statistics in brackets. * p < 0.1, ** p < 0.05, *** p < 0.01. Controls includes: Pop (log), GDP per capita at independence (log), GDP growth (%), Arable Land area (log), Oil rents (% gdp), External confl., Unitary, Nr. languages (log), British Legal origin, French Legal origin, Socialist Legal origin, Previous Colony, Absolute latitude (log). All observations are at the country level.
In Section A4, we further check the robustness of our results using two alternative proxies for linguistic diversity: a standard ELF and the index of peripheral diversity by Desmet et al. (Reference Desmet, Ortuño Ortín and Weber2005). The results, provided in Tables A4 and A5, are coherent with our hypothesis.
Potential mechanisms: education
Here, we explore the role of education in mediating the relationship between linguistic distance and accountability. The connection between accountability and education is well-documented (e.g., Botero et al. (Reference Botero, Ponce and Shleifer2013)). Numerous studies link education to critical elements of the accountability process, including political participation (Verba and Nie, Reference Verba and Nie1987; Verba et al., Reference Verba, Schlozman and Brady1995), voting (Dee, Reference Dee2004; Milligan et al., Reference Milligan, Moretti and Oreopoulos2004), government oversight (Reinikka and Svensson, Reference Reinikka and Svensson2004), and civic engagement (Glaeser et al., Reference Glaeser, Ponzetto and Shleifer2007). We highlight that secondary and tertiary education can mitigate the negative effects of adopting an official language distant from the local languages.
This hypothesis is supported by evidence showing that in many multilingual countries, students begin receiving instruction in the official foreign language at the secondary level. Col (2024) reports that while children in some African countries initially receive instruction in their local languages, all education from the secondary level onwards is conducted in the official foreign language. This pattern is also observed in former colonies in Asia, where historically, higher education was intended for a small elite group destined to become the country’s ruling class (Grier, Reference Grier1999; Ingham, Reference Ingham1956; Isichei, Reference Isichei1995; Kelly, Reference Kelly and Kelly2000b,d).
Given this evidence, we argue that educating students in the official language at the secondary and tertiary levels may influence both channels through which a distant official language harms accountability. First, as more individuals are educated in the official language, they gain greater access to information about elite behaviour, making it more usable. Second, learning in the official language implicitly transmits the culture embedded in it. This link has been explored in the literature on the effect of education in the official language on national identity. Blanc and Kubo (Reference Blanc and Kubo2024) show that in provinces where schools teaching in standard French were established in the 19th century, there is a stronger sense of national identity today than in other provinces. According to their study, education in standard French was crucial for shaping national identity.
To test this hypothesis, we include a series of educational measures from the dataset by Barro and Lee (Reference Barro and Lee2013) as controls. Specifically, we use the averages between 2000, 2005, and 2015 of the percentage of students attending primary, secondary, and tertiary education. Descriptive statistics for these variables are presented in Table 2, with data available only for 127 countries. In Table 6, we replicate the main analysis incorporating the various education measures.Footnote 22
Possible mechanisms: education
Notes: robust standard errors in parentheses and t-statistics in brackets. * p < 0.1, ** p < 0.05, *** p < 0.01. Controls includes: Pop (log), GDP per capita at independence (log), GDP growth (%), Arable Land area (log), Oil rents (% gdp), External confl., Unitary, Nr. languages (log), British Legal origin, French Legal origin, Socialist Legal origin, Previous Colony, Absolute latitude (log). All observations are at the country level.
Overall, the results suggest that education plays a role in the relationship between linguistic distance and accountability, but this influence is not uniform across all levels of education. Specifically, when we include primary education in our models, the coefficients for ADOL remain statistically significant, indicating that primary education alone does not fully mitigate the negative effects of linguistic distance on accountability.
However, when we include the percentage of students attending secondary and tertiary education, the statistical significance of the ADOL coefficients diminishes. This loss of significance suggests that secondary and tertiary education might act as important mediators in the relationship between linguistic distance and accountability. In other words, as more individuals receive education at these higher levels, the negative impact of linguistic distance on accountability appears to weaken.
Conclusions
This paper argues that the linguistic distance between a country’s official language and the languages spoken by its citizens is one of the factors that affect accountability. A more distant language makes it harder to understand elite behaviour and the formal mechanisms for sanctioning it, while also hindering the development of shared cultural norms that constrain elite power. This relationship has historical roots, particularly in the processes of colonisation and decolonisation, during which the languages of former colonisers were often adopted as official languages in newly independent countries. Using a cross-sectional dataset of 147 countries, we document a negative association between linguistic distance and accountability which is robust across several model specifications. We suggest that secondary and tertiary education can help mitigate the negative effects of a linguistically distant official language.
Although the cross-sectional nature of the data leaves some unobservable factors unmodelled, the paper provides clear evidence of the relationship between official language and accountability, laying the groundwork for future research. Building on our findings, future work should explore whether the effect of a distant official language is driven by comprehension difficulties or by a cultural gap between the elite who speak the official language and the broader population. This distinction is particularly relevant given recent advances in AI-driven translation technologies. To this end, examining the effects of changes in language policy in specific contexts could shed light on the mechanisms underpinning the link between official language and accountability. Another promising research avenue involves investigating the factors that influence the choice of an official language, specifically whether linguistic distance reflects elite rent-seeking behaviour or represents a neutral compromise in linguistically diverse countries. The interplay between linguistic diversity and linguistic distance warrants further attention. In this regard, future research could examine more accurately the role of education in mediating the effects of these two dimensions.
Supplementary material
Baronchelli, Adelaide; Foresta, Alessandra; Ricciuti, Roberto (2025), “Supplementary Material for: “The Words That Keep People Apart: Official Language and Accountability””, Mendeley Data, V2, doi: 10.17632/tr97hh863m.2
Acknowledgements
The author are grateful to the five anonymous referees for their valuable comments and suggestions.
Funding
Roberto Ricciuti acknowledges support by the European Union – Next Generation EU within the framework of PNRR Mission 4 – Component 2 – Investment 1.1 under the Italian Ministry of University and Research (MUR) programme “PRIN 2022”, Ref. Project Code: 20227NN492 – The Economics of Beliefs, Culture and Institutions – CUP: B53D23009720006.
Open access
