INTRODUCTION
Ethnicity remains an important factor in shaping politics and everyday life on the African continent, not in the least in the context of Kenya (e.g. Ahlerup & Isaksson Reference Ahlerup and Isaksson2015; De Luca et al. Reference De Luca, Hodler, Raschky and Valsecchi2018; Harris & Posner Reference Harris and Posner2019; Lonsdale Reference Lonsdale, Hino, Langer, Lonsdale and Stewart2019; Beiser-McGrath et al., Reference Beiser-McGrath, Müller-Crepon and Pengl2021). Research shows that voting patterns in Kenya largely hew to ethnic lines (Bratton & Kimenyi Reference Bratton and Kimenyi2008), that rent in Nairobi's Kibera slums varies by tenants’ background compared with that of the local chief (Marx et al. Reference Marx, Stoker and Suri2019), and that public investment, for example in road construction (Burgess et al. Reference Burgess, Jedwab, Miguel, Morjaria and i Miquel2015) and educational infrastructure (Kramon & Posner Reference Kramon and Posner2016; Li Reference Li2018), as well as project aid and local funds (Briggs Reference Briggs2014) have often been disproportionally allocated to the home region of the president in power as well as to regions where the president's ethnic group was demographically dominant.Footnote 1 Ethnic favouritism within Kenya's educational sector arguably has had particularly nefarious consequences (Alwy & Schech Reference Alwy and Schech2004). Not only because educational expenses constitute a substantial share of government expenditure (e.g. Berman et al. Reference Berman, Cottrell and Ghai2009; Li Reference Li2018; Simson & Green Reference Simson and Green2020), but also because of the long-term and often persistent benefits that accrue to unfairly favoured ethnic groups. Indeed, unfair advantage with respect to the allocation of educational resources (both in terms of infrastructure and availability of qualified teachers) may lead to substantial differences between different ethnic groups’ returns to education and consequently their future economic opportunities and status in society (Brown Reference Brown2011; Kramon & Posner Reference Kramon and Posner2016; Simson Reference Simson2021; see also Ricart-Huguet Reference Ricart-Huguet2021).
Interestingly, ethnic favouritism in public goods provision appears to have attenuated somewhat since the advent of democracy, as leaders have attempted not only to appeal to coethnics for electoral support, but to non-coethnic citizens as well (see e.g. Horowitz Reference Horowitz2016; Simson & Green Reference Simson and Green2020). This is illustrated by the observation that political parties increasingly campaign for programmatic policymaking instead of clientelist targeting. In Kenya, for instance, Burgess et al. (Reference Burgess, Jedwab, Miguel, Morjaria and i Miquel2015) no longer find evidence of disproportionally skewed road investments since the reintroduction of multiparty politics in the early 1990s. Moreover, Simson (Reference Simson2019) shows that, as a result of recent policies promoting regional equity and affirmative action measures, candidates from less developed districts are currently even more likely to work in the public sector than others when controlled for education.Footnote 2 And while some studies have argued that ethnic favouritism is still quite prevalent in the education sector in Kenya (see e.g. Franck & Rainer Reference Franck and Rainer2012; Kramon & Posner Reference Kramon and Posner2016), recent studies have shown that this is much less the case (Li Reference Li2018; Simson & Green Reference Simson and Green2020).
Yet, notwithstanding shifting realities, clientelist practices or ethnic politics seemingly remain common in weak bureaucratic states as they continue to operate largely in personalised and informal ways (see Abdulai & Hickey Reference Abdulai and Hickey2016 for a case study of the education sector in Ghana; Opalo Reference Opalo2022 for a case study of the community development fund in Kenya; or Verwimp Reference Verwimp2020 for a case study of test scores on the national exams in Burundi). It may therefore not surprise that the perception that politicians and state officials use their positions of power to favour their own kin has persisted as well. Survey data show that Kenyans who do not belong to the ethnic group of the president in power are more likely to feel treated unfairly by the government than those who share the same ethnic background as the president (Bratton & Kimenyi Reference Bratton and Kimenyi2008; Ahlerup & Isaksson Reference Ahlerup and Isaksson2015; Simson Reference Simson2019). Vice versa, 83% of Kenyans believe that government leaders ‘almost always’ or ‘some of the time’ favour their own ethnic groups (Horowitz Reference Horowitz2019).Footnote 3 With respect to public sector appointments too, a whopping 80% of Kenyans believe they are at least partly dependent on one's ethnic background (Mwabu et al. Reference Mwabu, Munga, Nyanjom and Onsomu2013). Adolescents share these perceptions: secondary school students in Nairobi consider that being from the ‘right tribe’ or having a ‘godfather’ (connections into the right circles) increases their likelihood of obtaining a job more than having a degree would (King et al. Reference King, Harel, Burde, Hill and Grinsted2020). Such perceptions are harmful, as they affect institutional and inter-ethnic trust, and, often, raise electoral stakes (e.g. Franck & Rainer Reference Franck and Rainer2012; Burgess et al. Reference Burgess, Jedwab, Miguel, Morjaria and i Miquel2015; Kramon & Posner Reference Kramon and Posner2016).
In a context where ethnic loyalties and favouritism appear to have become embedded in the political system, it is crucial to investigate how the education system exacerbates, mitigates or countervails a political culture of ethnicisation and ethnic favouritism. While recent research has carefully investigated and assessed the extent and impact of ethnic favouritism in the allocation of educational resources (Kramon & Posner Reference Kramon and Posner2016; Li Reference Li2018; Simson & Green Reference Simson and Green2020), so far, very little systematic research has been conducted on the ways in which ethnicity and ethnic diversity are dealt with inside the classroom. In this respect, it is important to note that besides teaching essential knowledge and skills that prepare young people for gaining access to the labour market, schools and teachers also impart societal norms and values. As such, they contribute in important ways towards shaping and influencing young people's social and political attitudes (e.g. Jennings & Niemi Reference Jennings and Niemi1974; Torney-Purta Reference Torney-Purta2002). Teachers and school administrators who favour, or give undue benefits to, a pupil who belongs to their own ethnic group may do so either consciously or unconsciously, motivated by ethnic solidarity, frustrated by (perceived) discrimination, or in an attempt to keep a positive image of the in-group – or just to be nice. Notwithstanding the underlying motivations, such actions risk being perceived as ethnic favouritism or risk conveying the message that ethnic favouritism is normal and socially acceptable, within, but by extension also outside, of a school context. Moreover, when experienced or perceived as such, unfair treatment may evoke feelings of frustration that may affect pupils’ academic motivation and/or outcomes, as research in Western contexts has convincingly shown (see e.g. van den Bergh et al. Reference van den Bergh, Denessen, Hornstra, Voeten and Holland2010).Footnote 4 Conversely, teachers and school administrators who refrain from providing ethnic favours and openly and consistently speak-out against this type of behaviour convey important messages about fairness and merit, while simultaneously delegitimising the practice of ethnic favouritism and, as a consequence, ethnic politics.
By systematically investigating and analysing how Kenyan teachers are dealing with issues of ethnicity and ethnic favouritism in their classrooms, this article addresses an important academic lacuna. Our study is explorative in nature and combines results from a large-scale survey with insights from in-depth follow-up interviews. Our survey was conducted between April and June 2016 among 894 secondary school teachers in the county of Nairobi and contained a range of questions concerning teachers’ experiences with ethnic favouritism and stereotyping as well as their own behaviour vis-à-vis pupils with different ethnic backgrounds as themselves. In order to determine the prevalence of ethnic favouritism among our sample of surveyed teachers, we included a list experiment, which is an extremely useful methodological technique to elicit people's true attitudes concerning sensitive issues (such as having unfairly favoured a coethnic), which may be susceptible to social desirability biases if asked directly.
The article will proceed as follows. In the following sections, we briefly describe the Kenyan context, and review the current literature on education and ethnic favouritism in the country. Next, we turn to our methodological framework, before presenting and discussing the results of the study. We conclude by outlining directions for future research.
THE KENYAN CONTEXT
There are more than 42 ethno-linguistic groups living together in Kenya, among whom the Kikuyu (18.8%), Luo (13.4%), Luhya (12.7%), Kalenjin (10.8%) and Kamba (10.5%) are the largest (Branch Reference Branch2011; Burgess et al. Reference Burgess, Jedwab, Miguel, Morjaria and i Miquel2015).Footnote 5 Although present prior to colonisation, ethnic identities were made salient in the way the British divided the country into tentatively ethnically homogeneous districts that facilitated governance – with the exception of three districts, a single ethnic group constituted the absolute majority in every district (Burgess et al. Reference Burgess, Jedwab, Miguel, Morjaria and i Miquel2015: 1825; Li Reference Li2018: 194; for a more extensive overview, see Berman et al. Reference Berman, Cottrell and Ghai2009). When the colonial government subsequently allowed the formation of African political associations in the years prior to independence, it only permitted associations to be formed along district lines, further shaping Kenyans’ understandings of interests, rights and mobilisation strategies in local, and hence largely ethnic terms – it was members of these associations who would form Kenya's ‘national’ parties in the last years of the colonial regime (Lynch Reference Lynch2006: 252; quotation marks in original). Importantly, the former British administrative divisions were maintained after independence – and even granted less freedom than before (Berman et al. Reference Berman, Cottrell and Ghai2009: 474; italics added); and local political dynamics further exploited through the harambee system (KiSwahili for ‘pulling together’). Against a background of weak fiscal and bureaucratic capacity constraints, these self-help projects emerged in the 1960s encouraging politicians to invest in their local communities to boost development and ensure re-election, further reinforcing ethnic saliency and clientelism (Lynch Reference Lynch2006: 252; Berman et al. Reference Berman, Cottrell and Ghai2009: 476; D'Arcy & Cornell Reference D'Arcy and Cornell2016; Opalo Reference Opalo2022: 2). Meanwhile, those in power nationally were increasingly perceived to favour their own ethnic group as well. Kenyatta (Kikuyu; 1963–1978), Moi (Kalenjin; 1979–2002), as well as Kibaki (Kikuyu; 2002–2013) are said to have favoured their own communities, among others through land redistribution policies and public appointments (e.g. Lynch Reference Lynch2006; Berman et al. Reference Berman, Cottrell and Ghai2009).
Long time a (de facto) one-party state (1963–1991), the advent of democracy (1991–1992) did force leaders seeking public office to appeal to members outside of their own ethnic group as coalition building is unavoidable in a highly diverse society such as Kenya (e.g. Horowitz Reference Horowitz2016; Kramon & Posner Reference Kramon and Posner2016). Presidential rallies therefore became more about pursuing potential swing voters (mainly those without a coethnic leader in the race) rather than only about mobilising voters in parties’ ethnic strongholds (Horowitz Reference Horowitz2016). At more local levels, however, ethnic favouritism persists. Created to replace the harambee community self-help system and reduce the costs for candidates in legislative elections, the Community Development Fund, for instance, continues to function as the primary means for Kenyans to evaluate politicians’ contributions to local, instead of national, development (Opalo Reference Opalo2022).Footnote 6
Despite the emergence of democratic elections, divisive policies and the fear of political exclusion has thus remained an important feature of Kenya's politics (Lonsdale Reference Lonsdale, Hino, Langer, Lonsdale and Stewart2019). Accordingly, many Kenyans believe it remains in their best interest to elect a strong ethnic leader in order to get a share of the national cake, even if they themselves do not consider ethnicity central to their identity (Bratton & Kimenyi Reference Bratton and Kimenyi2008; for a general discussion on the instrumentality of ethnic identities in African elections, see Eifert et al. Reference Eifert, Miguel and Posner2010). By the same logic, politicians who seek office, at least locally, maximise their votes by ‘playing the ethnic card’, and hence mobilising support along ethnic lines in exchange for material benefits (Eifert et al. Reference Eifert, Miguel and Posner2010; Horowitz Reference Horowitz2016). Kenya has, consequently, become trapped in a ‘vicious circle’ of perception and action (Lynch Reference Lynch2006), in which ethnic favouritism has become self-sustaining (see e.g. Eifert et al. Reference Eifert, Miguel and Posner2010; Burgess et al. Reference Burgess, Jedwab, Miguel, Morjaria and i Miquel2015; Horowitz Reference Horowitz2016). Once an expression of civic virtue and social obligation towards one's local community – i.e. moral ethnicity or ethnic solidarity – ethnic favouritism has thus become the driving force behind ‘political tribalism’ (see Berman et al. Reference Berman, Cottrell and Ghai2009: 469; Lonsdale Reference Lonsdale, Hino, Langer, Lonsdale and Stewart2019). The ensuing high stakes of winning an election have led to increased ethnic tensions and occasional violence around election times (Branch Reference Branch2011; Hornsby Reference Hornsby2013; Kramon & Posner Reference Kramon and Posner2016). In the aftermath of the disputed 2007–2008 presidential elections, most notably, ethnic clashes erupted throughout Kenya, causing the death of at least 1,000 people and displacing more than 350,000 Kenyans internally (e.g. Berman et al. Reference Berman, Cottrell and Ghai2009: 500; Horowitz Reference Horowitz2016: 335). Large-scale violence was avoided during the 2017 and 2022 presidential elections, in spite of controversies.
The Kenyan Education System
The education system in Kenya is currently undergoing reform. By 2026 the former 8-4-4 structure (whereby children attended 8 years of primary, 4 years of secondary and 4 years of higher education) will have been replaced by 2 years of pre-primary, 6 years of primary, 6 years of secondary (divided in junior and senior levels) and 3 years of higher education (2-6-6-3 structure). Education is provided in English, although pupils can attend the first three years of primary school in their local language. To progress from primary to secondary and later on to tertiary education, students have to pass the Kenyan Certificate of Primary Education (KCPE) and the Kenyan Certificate of Secondary Education (KCSE), respectively.
While access to private secondary schools is largely conditional on pupils’ financial situation – even though secondary education is supposedly free of costs since 2008 (Lewin et al. Reference Lewin, Wasanga, Wanderi and Somerset2011: 19; Smith et al. Reference Smith, Marks, Novelli, Valiente and Scandurra2016: 56)Footnote 7 – admittance to the highly stratified public system depends on pupils’ score on the national exams. Moreover, public schools apply regional quota. At the top of the four-tier hierarchy, there are the national schools, which admit the best-performing pupils from all over the country. District schools, the most common type of school (70% of public schools), are at the bottom and accept lesser performing pupils from the local area. County and extra-county schools, lastly, form the middle tier. They attract respectively middle-range to high-performing students coming from within the district, but also from outside students’ own districts and/or provinces (Lewin et al. Reference Lewin, Wasanga, Wanderi and Somerset2011; Makori & Onderi Reference Makori and Onderi2014; Nyatuka & Bota Reference Nyatuka and Bota2014).
Education and Ethnic Favouritism in Kenya
In the early days of colonialism, missionaries introduced a highly segregated education system, with different schools for White, Coloured and African pupils (Li Reference Li2018: 194). At the time, only few African students attended school as many could not afford for school fees, had to herd cattle or lived too far away (Hornsby Reference Hornsby2013: 139). School proximity was to a large extent determined by communities’ openness to missionaries (Nyatuka & Bota Reference Nyatuka and Bota2014: 49). When Kenya gained independence, President Kenyatta pledged to improve access to education for African children and invested large amounts of government funds in the education sector. With annual budget increases of 17%, Kenya had one of the highest proportions of government expenditure on education in the world (Berman et al. Reference Berman, Cottrell and Ghai2009: 472).
Although enrolment in primary and secondary schools rose two- and five-fold, respectively (Berman et al. Reference Berman, Cottrell and Ghai2009), enrolment rates remained rather low outside of Kenyatta's home region, the Central province (Alwy & Schech Reference Alwy and Schech2004: 270; Hornsby Reference Hornsby2013: 140). And, whereas nearly all teachers in Central province and Nairobi, where Kenyatta's ethnic group predominated as well, were professionally qualified by the end of Kenyatta's term, only 70% or fewer of teachers outside of those areas were (Alwy & Schech Reference Alwy and Schech2004: 270). As a result, Kikuyu children of school-going age in the 1960s on average completed respectively 47% and 39% more years of primary education, compared with Kalenjin and Luo children (Kramon & Posner Reference Kramon and Posner2016: 7). Inequalities were even more obvious at the secondary level. Where Central and Nyanza provinces were fairly well provided for, there were virtually no secondary schools in the Coast and North-Eastern provinces, nor among nomadic groups (Hornsby Reference Hornsby2013: 141; Nyatuka & Bota Reference Nyatuka and Bota2014: 50). Regional inequalities further deepened with the building of harambee schools – nearly 58% of all harambee projects concerned education (Opalo Reference Opalo2022: 5) – which were primarily constructed by Kikuyu. A legacy of colonial times, Kikuyu were at that time more educated compared with other groups, and as a result valued education more (Hornsby Reference Hornsby2013: 140; Li Reference Li2018: 195; Simson & Green Reference Simson and Green2020).
When Moi, a Kalenjin from the Rift Valley, took over the presidency, the focus of educational policies shifted towards constructing secondary schools in underserved provinces (Simson & Green Reference Simson and Green2020). The newly built schools were, however, not only attended by local pupils, but also by pupils from privileged provinces such as Central. In an attempt to further enhance regional equity for pupils from underprivileged marginal areas, Moi then introduced regional quota stipulating that 85% of admitted students should come from the local area (Hornsby Reference Hornsby2013: 447; Nyatuka & Bota Reference Nyatuka and Bota2014: 51). Whereas primary enrolment rates improved significantly among all groups (Hornsby Reference Hornsby2013: 446), there was a 54% increase in secondary school attainment among the Kalenjin relative to the national average under his administration (Kramon & Posner Reference Kramon and Posner2016: 7).
In 2003, Moi was succeeded by Mwai Kibaki, once again a Kikuyu (2003–2013). Kibaki pledged to make primary education free, thereby mainly targeting urban slums and (semi) arid regions where school density was low (Simson & Green Reference Simson and Green2020). Still, the strongest increase in primary school attainment relative to the national average was observed among the Kikuyu, even though their primary school completion was already near the maximum (ceiling effect). Likewise, while secondary school rates were overall declining under Kibaki, the smallest reduction was among the Kikuyu – along with the Kalenjin and Luo (Kramon & Posner Reference Kramon and Posner2016: 7).
It seems that, since independence, having a coethnic president in office during one's school-aged years has translated into educational benefits. Indeed, Kramon & Posner (Reference Kramon and Posner2016) estimate that those benefits amount to increases of respectively 0.36 and 0.12 years of primary and secondary schooling completed. Similarly, Li (Reference Li2018) finds that residents of districts where coethnics of the president are the majority gain an increase of around 4% in the likelihood to complete primary education (meaning that coethnics living outside of these districts do not benefit). Belonging to the ethnic group of the minister of education would have resulted in additional years of schooling too: for every year of primary (secondary) school attendance with a coethnic minister of education, a learner gains 0.06 (0.09) more years of primary (secondary) education – resulting in an additional third (tenth) of a year if the minister served throughout one's primary (secondary) schooling (Kramon & Posner Reference Kramon and Posner2016).
The evidence is contested, however. First of all, the studies of Kramon & Posner (Reference Kramon and Posner2016) and Li (Reference Li2018) diverge on the persistence of ethnic favouritism in the multiparty era. Whereas the former find evidence in favour, the latter refutes those findings. At university-level too, there is no impact of coming from the home region of the president in power in recent times (Simson Reference Simson2021).Footnote 8 Second, Simson & Green (Reference Simson and Green2020) argue that both studies rest on problematic assumptions. Rapid educational attainment across, and convergence among, all larger ethnic groups, they contend, overshadows any marginal advantage of having a coethnic president: rather than conclusive evidence of ethnic favouritism, the slow-down of the rate of Kikuyu compared with Kalenjin attainment under Moi's presidency, for instance, is due to the former reaching the ceiling of primary education earlier. And, the head start of Kikuyu children would be a legacy of the colonial period more than the result of disproportionate spending in Kikuyu regions under Kenyatta (Simson & Green Reference Simson and Green2020; see also Ricart-Huguet Reference Ricart-Huguet2021); more generally, there is an important effect of intergenerational transmission of educational attainment, as children of parents with higher levels of education are more likely to attain higher levels too (Simson Reference Simson2021).
Stark regional differences in school infrastructure and teacher deployment have remained, nevertheless. Today, access to education is most precarious in Northeastern province, largely because of security threats in the region (Smith et al. Reference Smith, Marks, Novelli, Valiente and Scandurra2016: 49). Reminiscent of the experience of the 2007–2008 post-electoral crisis, the Kenyan authorities have sought to address these differences, among others, by vesting each county with at least two national schools (see the final report of the Truth, Justice and Reconciliation Commission 2013; Makori & Onderi Reference Makori and Onderi2014; Nyatuka & Bota Reference Nyatuka and Bota2014: 48). The distribution of national schools remains unequal however: in (2010) 2014, there were (6) 9 national schools in former Central province; (6) 7 in Nairobi; (5) 11 in Rift Valley; (0) 4 in Coast Province; (0) 5 in Eastern Province; (1) 6 in Nyanza; (0) 1 in Northeastern; and (0) 5 in Western (Nyatuka & Bota Reference Nyatuka and Bota2014: 52).
It follows that, historically, ethnic favouritism has been an important factor in explaining the persistence of regional inequalities in education. One question which is directly linked to the educational inequalities, but has not yet been sufficiently empirically investigated in Kenya is the extent to which ethnic favouritism has affected educational practices and teaching behaviour. In the remainder of this paper, we will specifically focus on this important academic and policy void.
CONCEPTUALISING ETHNIC FAVOURITISM AMONG TEACHERS
Before turning to our analysis, it is important to understand why teachers would give favours to students from their own ethnic group because of their shared group belonging. Whereas politicians usually provide material goods or services to citizens because they expect their vote in return (for a discussion of ethnic-based clientelism, see Isaksson & Bigsten Reference Isaksson and Bigsten2017),Footnote 9 there is no material return, or increased likelihood of promotion, for teachers who favour students belonging to their own ethnic group – not taking corrupt practices into consideration (e.g. receiving financial or material benefits in return for, for instance, a good grade or the leaking of exams).Footnote 10 Yet, in the political realm too, there is more to favouring one's own kin than a quid pro quo alone. Notably, Adida et al. (Reference Adida, Gottlieb, Kramon and McClendon2017)'s concept of ‘ethnically motivated reasoning’ builds on Tajfel's (Reference Tajfel1974) social identity theory: their field experiment on ethnic voting in Benin shows that citizens do not only vote for coethnic politicians because they expect their region to be favoured by policies or redistribution, but also because citizens derive self-esteem and pride from seeing their group do well compared with others (see Adida et al. Reference Adida, Gottlieb, Kramon and McClendon2017). Likewise, research on education and discrimination in Western contexts suggests that teachers, often unconsciously, favour coethnic pupils to protect the positive image of their in-group (Schuchart et al. Reference Schuchart, Glock and Dunkake2021; see also Kleen et al. Reference Kleen, Bonefeld, Glock and Dickhäuser2019). ‘Ethnically motivated reasoning’ would moreover allow citizens – here, teachers – to affirm their status as ‘good’ members of their group (Adida et al. Reference Adida, Gottlieb, Kramon and McClendon2017). This is reminiscent of Kenya's traditional ‘moral ethnicity’ (Berman et al. Reference Berman, Cottrell and Ghai2009), or the expectation that you have to act generously towards and maintain links with your rural home in order to gain social status and nurture ‘civic virtue’ (Lonsdale 1994, in Lynch Reference Lynch2006). Although such forms of ethnic favouritism are arguably well-intentioned, they can be harmful. Indeed, showing positive affect, such as sympathy and trust, or limiting prosocial behaviour (e.g. helping) to the in-group can be considered discriminatory by the out-group (Brewer & Miller Reference Brewer and Miller1996). We argue, furthermore, that these socialisation experiences of children and adolescents in the classroom contribute to the normalisation and acceptance of ethnic favouritism beyond the school context and may even contribute to legitimising ethnics politics.
Teachers who want to act as ‘good’ members of their in-group will deliberately give undue benefits to coethnic students. Teachers may, however, also favour pupils from their own kin without being aware of it. Such unconscious behaviour (often non-verbal communication) can cause ethnic disparities too in, for instance, penalising misconduct (see van den Bergh et al. Reference van den Bergh, Denessen, Hornstra, Voeten and Holland2010; Kleen et al. Reference Kleen, Bonefeld, Glock and Dickhäuser2019; Schuchart et al. Reference Schuchart, Glock and Dunkake2021). It is important to note in this respect that, like anybody else, teachers have a tendency to exhibit implicit preferences for the in-group (for a case study among Kenyan teachers, see Kuppens et al. Reference Kuppens, Langer and Ibrahim2018). These preferences guide behaviour, unless people are aware of them and motivated not to act upon them (Dasgupta Reference Dasgupta2004).
DATA AND METHODOLOGY
To explore ethnic favouritism within the classroom, we conducted a large-scale survey among secondary school teachers in Nairobi, Kenya, between April and June 2016.Footnote 11 In total, 925 teachers from 64 schools across all sub-counties (existing at the time of the study) participated – we only use a subset of 894 teachers, excluding teachers who did not respond to the list experiment.Footnote 12 The questionnaire was self-administered on a tablet (Samsung S5), using the software package ‘Qualtrics’ (Qualtrics LLC, Provo, Utah), after being piloted both as a pen-and-paper and digital survey in respectively three and two schools. To complement the survey data, we conducted in-depth follow-up interviews with 18 secondary school teachers.Footnote 13 The interviews were nested, i.e. interviewees were selected based on their responses to the survey questions (and having consented to be contacted for follow-up interviews), ensuring variation on background characteristics (gender, subject taught and sub-county).
The participating 64 schools were randomly selected after stratification by district and type of school based on a list of 258 schools provided by the Ministry of Education (176 private and 82 public schools). All selected schools agreed to participate and effectively took part in the study – two schools no longer existed and were replaced with the most similar school. Table I provides an overview of the school characteristics. Teachers who were present at school at the time of our survey participated – only a few refused participation, resulting in a cooperation rate of nearly 100%. In cases when less than half of the schools’ teachers participated, schools were visited a second time (response rate = 69%). Table II gives a summary of teacher characteristics.
Table I. Descriptive Statistics of school characteristics (N = 64)
Table II. Descriptive Statistics of teacher characteristics (N = 894)
Whereas the findings of the study may not be generalisable to all parts of the country, Nairobi was deemed an interesting case study given that the county is highly diverse. Still, levels of ethnic favouritism could be significantly lower in the county because of that diversity: through repeated contact with fellow teachers and students from various backgrounds, teachers may have developed more positive out-group perceptions and attitudes than teachers teaching outside Nairobi (see Allport Reference Allport1954). It is also important to note that the majority of schools in Nairobi county (and our sample) are private, in contrast to only 12% of schools nationwide. Yet, private schools are, on average, 3.5 times smaller than public schools (MoE 2016), which means that the majority of teachers in Nairobi, and the sample (56.6%), are active in a public school.
Measures
In terms of measures, we make use of two direct and one indirect measure to explore ethnic favouritism within schools. The direct measures focused on preferential and unfair treatment of coethnic learners. First, we asked teachers to what extent they prefer to have pupils that belong to their own ethnic group, using a four-point scale (1 = I have no preference; 2 = I have a slight preference; 3 = I have a moderate preference; 4 = I have a strong preference). Second, we asked teachers how often they think that pupils in their school are treated unfairly because of their ethnic background (1 = Every day; 2 = A few times a week; 3 = A few times a month; 4 = A few times a year; 5 = Never), and, if so, by whom (i.e. by other pupils, teachers or the principal).
As an indirect measure, we included a list experiment, also called the Item Count Technique (ICT). List experiments are used to estimate the prevalence of attitudes or behaviour that are subject to a social desirability bias, such as drug use, abortion or racist attitudes, which are commonly underreported in direct measurements; as well as voting and church attendance, which are frequently overreported (see e.g. Blair & Imai Reference Blair and Imai2012; Comşa & Postelnicu Reference Comşa and Postelnicu2013 for references). List experiments present respondents with a list of items concerning respondents’ attitudes or past behaviour. A first group, the control group, receives a list that consists of non-sensitive items alone (usually three or four) – i.e. items about attitudes or behaviour that are not susceptible to a social desirability bias. The second, experimental, group receives the same list with the exception that a sensitive item has been added – in our case favouring a coethnic student.Footnote 14 Respondents are randomly assigned to either group – though it is also possible to present both lists to every respondent. Next, instead of asking respondents to identify which precise items are applicable to them, they are asked to indicate how many items apply. In this way list experiments offer respondents full anonymity, as they no longer have the incentive to lie about their attitudes or behaviour (e.g. Blair & Imai Reference Blair and Imai2012; Comşa & Postelnicu Reference Comşa and Postelnicu2013). When all items apply (a so-called ceiling effect) however, list experiments no longer guarantee anonymity. Hence, it is recommended to include at least one item that is relatively uncommon, meaning that it should apply only to a handful of respondents. Respondents’ privacy is also compromised when none of the items is applicable – a floor effect. Hereto, it is important to include an item that is highly likely to apply to all respondents (see e.g. Holbrook & Krosnick Reference Holbrook and Krosnick2010; Blair & Imai Reference Blair and Imai2012; Comşa & Postelnicu Reference Comşa and Postelnicu2013). Taking these conditions into account, we presented teachers with the following list experiment (the order of the items in the list was not randomised):
Below we list different actions that you may have taken or scenarios, which you may have encountered as a teacher. We are not interested in finding out which specific things may have happened to you, but only how many of these scenarios apply to you:
Please write down a number between 1 and 4 (Experimental group: 5) that corresponds to the number of scenarios that apply to you. Thus, if all four (Experimental group: 5) statements are true in your case, you would write “4” (Experimental group: “5”), if however only one scenario applies to you, you would record “1”:
• I have skipped a small part of the curriculum due to time concerns;
• Experimental group only: I have favoured a student from my own ethnic group;
• I have used video material in class;
• I have made a spelling error on the blackboard;
• I have invited an external speaker to my classroom;
Although discussed in the reverse order, the list experiment was put at the beginning of the survey, while the direct questions were put at the end in order to avoid a contagion effect. It is important to note that list experiments in general, and this experiment in particular, assess the extent to which the sensitive item is present within the sample population, here ethnic favouritism among secondary school teachers. The list experiment therefore does not explain the frequency with which this sensitive behaviour is conducted. In other words, it does not tell us anything about the number of students that have received undue benefits, or about how many times they received favours.
We estimate the proportion of people to whom the sensitive item applies by aggregation. By subtracting the mean sum of positive responses in the control group from the mean sum of positive responses in the experimental group, the difference-in-means test yields the proportion of people that have engaged in the sensitive behaviour (e.g. Holbrook & Krosnick Reference Holbrook and Krosnick2010: 44). Using the count provided by the respondents as the dependent variable and group belonging (treatment/control) as an independent variable, it is also possible to estimate multivariate analyses and discern which teacher characteristics help to explain who engages in ethnic favouritism (see e.g. Holbrook & Krosnick Reference Holbrook and Krosnick2010; Blair & Imai Reference Blair and Imai2012). Teachers’ level of ethnic identification was deemed of interest in this respect. Therefore, we also asked teachers to what extent they identify as a Kenyan and/or as a member of their ethnic group (1 = I feel only Kenyan; 2 = I feel more Kenyan than from my ethnic group; 3 = I feel equally Kenyan as from my ethnic group; 4 = I feel more from my ethnic group than Kenyan; 5 = I feel only from my ethnic group). We expect that teachers who identify stronger with their ethnic group are more likely to have favoured a coethnic student in the past.
ETHNIC FAVOURITISM AMONG KENYAN SECONDARY SCHOOL TEACHERS
Direct measures
The teachers in our sample, first of all, seem to identify more as Kenyan than as a members of their ethnic group. Whereas 22.3% of teachers only identify as Kenyan, 31.8% feel more Kenyan than part of their ethnic group. 39.3% of teachers identify equally as Kenyans and members of their ethnic groups. Only few teachers identify more (5.9%) or only (0.8%) with their ethnic group.Footnote 15 Nationwide, Afrobarometer data (2015) shows that fewer than 10% of Kenyans identify more strongly with their ethnic group than with the national identity, positioning Kenya on the lower end of professed ethnic identification in sub-Saharan Africa (also see Isaksson & Bigsten Reference Isaksson and Bigsten2017). Arguably, these proportions are affected by social desirability in light of the country's legacy of ethnic politicisation and conflict. Differences in teachers’ feelings of identity cannot be explained in terms of sex or ethnic group belonging – including when ethnicity is recoded to a binary variable indicating whether or not a teacher belongs to an ethnic group associated with the president in office at the time of the data collection (i.e. Kikuyu, Kalenjin, Embu or Meru). Older teachers, however, appear to be marginally more likely to identify more as Kenyan (rs = 0.077, p <0.05). Also, teachers of social sciences are proportionally less likely to identify solely with the Kenyan identity than others, while generally they do identify more with the Kenyan identity than with their ethnic group.Footnote 16
Interestingly, 24.2% of the teachers express a preference for coethnic pupils (slight preference = 10.4%; moderate preference = 9.5%; strong preference = 4.3%). Contrary to expectations, there are no differences in teachers’ preferences among teachers who identify more with their ethnic group or more with the Kenyan nationality. Nor did we find any significant differences in preferences according to teachers’ sex, subject taught or ethnic group belonging.Footnote 17 Older teachers do express fewer preferences to have pupils from their own ethnic group (rs = −0.171, p <0.000).
Given the particular school system in Kenya, we also test whether there are significant associations between type of school (i.e. national, extra-county, county, district or private; and day or boarding school) where the teacher works and their preferences and feelings of identity: it is possible that teachers working at a national (boarding) school, supposedly much more diverse, have fewer preferences and identify more with the Kenyan nationality than teachers working in district (day) schools. Yet, we do not find any significant associations.Footnote 18 Arguably, even district (day) schools might be very diverse in an urban setting such as Nairobi – studies show that national identity usually is stronger than ethnic identity among urban than rural Kenyans (Kramon et al. Reference Kramon, Hamory, Baird and Miguel2022; see also King et al. Reference King, Harel, Burde, Hill and Grinsted2020 who made similar observations among secondary school students in Nairobi).
Before turning to teachers’ own behaviour, we examine to what extent teachers have the perception that pupils are treated unfairly at school (see Table III). Only few teachers think that unfair treatment occurs daily (3.1%) or weekly (4.3%) at school, compared with 57.5% of teachers who believe it never occurs at all. Among those who believe pupils are treated unfairly at school, most think it happens among pupils (68.4%).
Table III. Teachers’ perceptions of the occurrence of unfair treatment and main responsible (in per cent):
Indirect measures
Table IV summarises the results of the list experiment. From the difference-in-means test (2.57–2.31 = 0.26), we deduce that at least 26% (SE = 0.08) of teachers have already favoured a student from their own ethnic group (95% confidence interval = 10.32–41.68%) – to assess the reliability of the difference-in-means test, we also ran it using the R package ‘List’ (Version 8.2; Blair et al. Reference Blair, Imai, Park, Coppock and Chou2016) obtaining a result of 25.7% (SE = 0.08; 95% confidence interval = 10.02–41.38%).Footnote 11 The assumptions (1 = Random groups; 2 = No design effects; 3 = No liars) are not significantly violated.Footnote 19 Notably, this proportion is very close to the proportion of Kenyans that supports clientelism: almost 25% (strongly) agreed that ‘once in office, elected leaders are obliged to help their home community or group first’ (Afrobarometer data 2015; see Isaksson & Bigsten Reference Isaksson and Bigsten2017). Note that support for clientelism is towards the higher end in Kenya compared with other sub-Saharan African countries (although far below Sierra Leone's score of about 50%).
Table IV. Observed Data from List Experiment
To explore whether teachers’ background characteristics or attitudes can predict whether or not they will give benefits to their kin, we conducted two multivariate regressions, including teachers’ sex, age, course taught, ethnic group belonging (group in power or not), as well as their feelings of identity and their preference to have coethnic pupils (see Table A II in Appendix for results and explanations). None of these variables could, however, explain ethnic favouritism.Footnote 20 There was, nevertheless, a significant effect of identifying more or only with the ethnic group. This effect should be interpreted with caution, however, as the sample included only few teachers who identified as such.
Insights from interviews
In the interviews, nearly all teachers unequivocally dismissed ethnic favouritism. One teacher argued that it ‘kills professionalism’ (Male teacher, 24 years old, English). Instead, interviewees affirmed that teachers ‘just look at your potential, your ability’ (Female teacher, 25 years old, Economy and Biology), and that they ‘just award what he (cf. the pupil) has done … When a child has gotten an 80, it's 80. If it's 50, it's 50’ (Male teacher, 22 years old, English). Moreover, some of them stressed their role as role models: ‘As a teacher, you should be able to bring equality, equity between students so that you can eradicate the issue of tribalism’ (Male teacher, 23 years old, KiSwahili). What is more, they identified schools as the ideal setting to combat ethnic favouritism: ‘we can teach them (cf. pupils) how to choose leaders, and to teach people not just to choose people because they come from their tribe, but because they can work’ (Male teacher, 25 years old, Mathematics and Biology). It is noteworthy in this regard that the English curriculum includes course material reflecting on motivations to vote for a good leader (see Ibrahim Reference Ibrahim2021).
Yet, some teachers did admit that ethnic favouritism occurs, but stressed its subtlety: ‘(Teachers would) not give preference, because that one is seen, that one will be out. But the way they behave, they are not committed to other students, that's where the problem comes’ (Male, 41 years old, Economy, Geography and Christian Religious Education). Another teacher explained: ‘You find that he (cf. a Luo teacher) has a lot of patience with the students from the Luo community, but no patience with the ones from the Kikuyu community or from other communities’ (Female teacher, 42 years old, English, Life Skills Education and Christian Religious Education). Likewise, pupils would be more likely to turn to teachers from their own community as well:
If I'm a Kikuyu, you find that all the Kikuyu students will come to me whenever they have a problem. Or … the students from the Luyha community will be going to that person, and whenever they are together, they talk in their native language. (Male teacher, 25 years old, Mathematics and Biology)
Other examples of ethnic favouritism included giving transport fares to coethnic pupils: ‘Let us say you come from really far, then that student asks you fare, you can give fare’ (Male teacher, 27 years old, KiSwahili and History and Government). Note how these teachers talk about ethnic favouritism using hypothetical terms or speak in third person (‘they’), suggesting they are not susceptible to such behaviour. Importantly, teachers seemed aware that showing positive affect to coethnic pupils could harm others: ‘If a teacher does that (cf. giving ethnic favours), most of the students will not perform well, the students of the other communities’ (Male teacher, 25 years old, Mathematics and Biology).
Teachers who do give benefits would therefore not willingly disadvantage learners from other groups. Rather, they argued that some teachers would act out on the ‘wrong’ belief that ‘when you help your own, you have done something good’ (Male teacher, 27 years old, KiSwahili and History and Government). Another teacher confirmed: ‘they say: this one belongs to us; this one belongs to us. They (cf. teachers) take them (cf. pupils) as they come from the same family!’ (Female teacher, 42 years old, English, Life Skills Education and Christian Religious Education). These views are a good expression of moral ethnicity and in line with the social identity theory and ethnically motivated reasoning (see Adida et al. Reference Adida, Gottlieb, Kramon and McClendon2017).
DISCUSSION AND CONCLUSION
Historically, ethnic loyalties and favouritism have skewed the allocation of public resources across different regions and/or ethnic groups in Kenya, affecting inter-ethnic trust and, often, raising the stakes of elections. While ethnic favouritism has been observed in different sectors and policy areas, favours in the domain of education have attracted particular attention as educational expenses make up a substantial share of government expenditures and because of the substantial benefits accrued by unfairly favoured groups in terms of returns to education – subsequently determining their future economic opportunities and status in society. Yet, besides human capital accumulation, schools and teachers play a crucial role in shaping and influencing adolescents’ social and political attitudes. While they can delegitimise ethnic favouritism by refraining from ethnic favouritism and/or by speaking out against this practice, teachers who – consciously or unconsciously – give undue benefits and preferential treatment themselves risk conveying the message that ethnic favouritism is normal and socially acceptable, which could render students more susceptible to ethnic politics. Moreover, feelings of discrimination may generate mistrust towards non-coethnic teachers.
The extent to which the socialisation experiences of children and adolescents at schools contribute to the normalisation and acceptance of ethnic favouritism, and particularly teachers’ role therein, is thus an important area of research, which has been insufficiently investigated and researched. The current article is therefore addressing a very important academic lacuna. Based on a large-scale survey among 894 secondary school teachers in Nairobi and 18 in-depth follow-up interviews, we have systematically assessed and examined teachers’ in-group preferences, their perceptions of unfair treatment of pupils based on ethnic background, and most notably whether teachers have already given undue benefits to pupils from their own ethnic group themselves. In order to accurately assess the latter issue, we employed a list experiment, which is an extremely useful methodological technique to elicit respondents’ true attitudes and responses concerning sensitive behaviour.
Crucially, we found that one fourth of secondary school teachers in Nairobi admitted to having a preference for teaching coethnic pupils and that about 25% of teachers admitted to have already favoured a student belonging to his/her own ethnic group. It is notable that these proportions are of the same magnitude as those of support for clientelism among the larger Kenyan population. This finding is worrying for at least two reasons. First, we hypothesised that ethnic favouritism would be lower among teachers in Nairobi because of repeated inter-group contact. Second, teachers work with adolescents, shaping and influencing their social and political attitudes. The interviews indicated however that such favours were often not blatant in nature, which could explain why only a minority of teachers believed that pupils were at times discriminated against at school on the basis of their ethnic background. Rather than discriminating against out-groups, it seems that teachers favour coethnic pupils to show solidarity with one's kin – as tradition dictates – and/or to act as a ‘good’ member of the in-group – in line with Adida et al.'s (2017) concept of ‘ethnically motivated reasoning’. No matter the nature of the intentions, however, ethnic favouritism may be detrimental to inter-group trust, may shape adolescents’ attitudes towards ethnicity and ethnic favouritism, and, in the longer term, impact their susceptibility to ethnic politics. It is positive in this respect that many teachers we interviewed were aware of their potential role in exacerbating the politicisation of ethnicity and therefore consciously tried to put ethnic considerations aside in school.
More research is needed to understand who engages in ethnic favouritism. Notably, teachers who preferred to teach coethnic students were not more likely to have given undue benefits to students from their own group in the past. Neither could we determine any other background characteristic indicative of such behaviour, although our evidence suggests that identifying more with one's ethnic group than with the national identity plays a role. Future research could therefore look more carefully at this and other characteristics, including for example altruism. Past research has already shown in this respect that slum dwellers’ level of altruism informs their collaborative behaviour more than in-group bias or ethnic favouritism in experimental games (Berge et al. Reference Berge, Galle, Posner, Zhang, Bjorvatn, Miguel and Tungodden2019). Other than examining teacher characteristics, future research could expand this study to include rural areas in Kenya and examine if ethnic favouritism, like ethnic identification (Kramon et al. Reference Kramon, Hamory, Baird and Miguel2022; see also King et al. Reference King, Harel, Burde, Hill and Grinsted2020), is more prevalent in those areas. Third, future research could look into the different types of ethnic favouritism. Since the list experiment did not specify the type of favour, we do not know what favours or undue benefits teachers were (not) thinking of. Neither did the experiment evaluate the frequency of giving favours. Whereas the interviews provided some insights into both issues, observational studies could shed more light and help to circumvent any outstanding effects of social desirability. Alternatively, it would be of interest to interview third parties, such as school administrators or pupils, to reflect on teachers’ behaviour. Research among pupils could also assess the extent to which students are affected by ethnic favouritism, controlling for a range of factors, including the position of the student's ethnic group vis-à-vis that of the teacher, previous experiences of unfair treatment and the timing and setting of such behaviour – a Ugandan study showed in this regard how even small cues can trigger stereotype threat among female students who attended co-ed (mixed gender) schools (Picho & Stephens Reference Picho and Stephens2012). Finally, it would be of interest to examine whether teachers in countries with high (low) levels of support for clientelism, or with high (low) levels of ethnic identification, engage more (less) in giving undue benefits to coethnic students. Afrobarometer data show that both professed ethnic identification and open support for clientelism is neither moderate nor extreme in Kenya compared with other Sub-Saharan African nations. The unequal allocation of public resources for education was found to be moderate in Kenya too (Franck & Rainer Reference Franck and Rainer2012). We thus argue that ethnic favouritism in the classroom is an important area where more research is needed, including in-depth case studies, comparative research and experimental studies in a range of ethnically divided countries both with and without a history of political tensions.
Appendix
Table A I. Tests of randomisation of the list experiment
Table A II. Effects of demographic variables, coethnic preferences and feelings of identification on ethnic favouritism (N = 984)
Open access
