THE EFFECT OF EDUCATION ON POLITICAL ENGAGEMENT IN WEAKLY INSTITUTIONALIZED COUNTRIES: EVIDENCE FROM NIGERIA

Transcription

1 THE EFFECT OF EDUCATION ON POLITICAL ENGAGEMENT IN WEAKLY INSTITUTIONALIZED COUNTRIES: EVIDENCE FROM NIGERIA HORACIO A. LARREGUY JOHN MARSHALL FIRST DRAFT: SEPTEMBER 2013 Abstract Developing countries are currently experiencing unprecedented increases in primary schooling. While there is strong evidence that education increases civic and political participation in developed democracies, such pro-democratic results may not occur in weakly institutionalized countries. Using variation across gender and local governments areas in the intensity of Nigeria s 1976 Universal Primary Education reform to instrument for primary schooling, we identify large and long-term political effects: up to 33 years after starting primary school, survey respondents that benefited from the program are considerably more interested in politics, more likely to vote and contact local government officials, participate in community associations, and support democracy in the abstract. Contrary to the concern that political engagement might reflect institutional capture, our effects are strongest among minority religious groups and in the most fractionalized areas. JEL: D72, I25. Key words: primary education, political engagement, civic engagement, Nigeria. This paper benefited from helpful conversations with and suggestions from Nahomi Ichino, Olayinka Idowu and Ayodele Iretiayo. Participants at the MIT Political Economy Workshop provided essential feedback. We are greatly indebted with Jonathan Phillips and Musiliu Adeolu Adewole who facilitated the public school census data, and the Nigeria National Bureau of Statistics who provided the Harmonized Nigeria Living Standard Survey. Alejandra Menchaca provided support and patience throughout the project. All errors are our own. Department of Government, Harvard University, Cambridge, MA (hlarreguy@fas.harvard.edu). Department of Government, Harvard University, Cambridge, MA (jlmarsh@fas.harvard.edu). 1

2 1 Introduction Seeking to meet the Millennium Development Goal (MDG) of universal primary education, many developing countries are investing heavily in early education. Although in 1999 primary school enrollment in sub-saharan Africa was only 57%, the proportion has been rising rapidly in the last decade (UNESCO 2008). This newly educated generation will become increasingly politically relevant, especially in the face of evidence that many of sub-saharan Africa s nascent democracies are not consolidating (Opalo 2012) and are failing to hold governments to account beyond the voting booth (Bratton and Logan 2006). Consequently, it is crucial to understand how increased primary schooling will affect the existence and quality of democracy in currently weakly institutionalized settings. We do this by identifying the effects of publicly-provided primary education in Nigeria on civic and political engagement. To shed light on this important issue, we examine the relationship between primary education and civic and political attitudes and participation in Nigeria. In 1976, Nigeria s military government ambitiously implemented its Universal Primary Education (UPE) program, providing six tuition-free years of primary education to all six year-olds. The impact of UPE differed hugely across Nigeria s regions and by gender: while the Western region already had relatively advanced educational provision, predominantly Muslim Northern areas required considerable financial support and investment in capacity to meet the target. Combined with Afrobarometer survey data, , we use such variation in the intensity of the UPE program to identify the political effects of primary schooling using a difference-in-difference strategy to instrument for primary schooling. How education affects democracy remains fiercely debated. An early and optimistic modernization literature suggested education encourages democratic norms of tolerance (e.g. Lipset 1959) and lays the groundwork for successful democratic consolidation (Dahl 1971). However, recent cross-national statistical analyses dispute the causal relationship underpinning the clear positive 2

3 correlation between education and democracy (Acemoglu, Johnson and Robinson 2005; Barro 1999; Castelló-Climent 2008; Glaeser, Ponzetto and Shleifer 2007; Przeworski et al. 2000; Papaioannou and Siourounis 2008). Research in developed democracies has consistently found that education increases turnout (see Sondheimer and Green 2010) and other forms of political engagement (e.g. Almond and Verba 1963; Putnam, Leonardi and Nanetti 1994). However, it is far from obvious that education similarly cultivates democratic and pro-social behaviors in developing democracies. Rather, recent research in developing contexts has suggested education breeds greater disenchantment with democracy and support for anti-democratic methods (Friedman et al. 2011), institutional capture (Gugerty and Kremer 2008), and extremist behaviors (Berrebi 2007; Krueger and Maleckova 2003). Despite considerable attention at the macro-level, there is relatively little micro-level evidence on the effect of education on political and community engagement and support for democracy outside Western democracies. While the Afrobarometer and other surveys have permitted unprecedented opportunities to examine the political behavior and preferences of citizens in Africa s nascent democracies (e.g. Bratton and Mattes 2001; Mattes and Bratton 2007), the causal effect of formal schooling at the individual-level has proved hard to identify. The few studies focusing on estimating causal effects have focused on foreign donor-supported programs providing civic education (Finkel and Smith 2011; Finkel, Horowitz and Rojo-Mendoza 2012; Gottlieb 2012), exam performance incentives (Friedman et al. 2011) and informal adult education (Kuenzi 2006) rather than publicly-provided education. Illuminating the causal relationship between primary education and civic and political engagement in Nigeria, a new and highly relevant policy context, can provide valuable information to educational reformers in developing countries, as well as help to disentangle the relationship between education and democracy. By using surveys up to 33 years after the initial reform, unlike previous studies we are able to estimate the enduring effects of primary schooling. Our study also addresses the concern that increased participation reflects capture by a dominant ethnic/religious 3

4 group. Reduced form and IV estimates show that Nigeria s UPE program had considerable political implications. First, primary schooling significantly increased interest in politics and basic forms of political participation including registering to vote, voting and contacting local politicians. Primary schooling is no panacea though: perhaps unsurprisingly, it fails to cultivate more costly forms of political activity like contacting MPs or political demonstration. Second, primary schooling has large positive effects on the likelihood of respondents attending community meetings, joining associations and actively participating in associations. Third, primary schooling increases support for democracy in the abstract without affecting satisfaction with democracy in practice or trust in politicians, indicating that education increases intrinsic support for democracy. We provide a variety of checks to support the identifying parallel trends and exclusion restriction assumptions. Finally, we show that the effects of primary schooling are strongest in more religiously fragmented areas and among minority religion respondents, while there is no indication that primary schooling increases ethnic identity or segregated participation. Together, these results provide strong evidence that primary education fosters civic and political engagement, even in weakly institutionalized contexts like Nigeria. The effects are larger and more robust than studies examining the short-term effects of civic education. Encouragingly, the concern that the skills education provides may lead to elite capture or exacerbate existing cleavages finds no support. This fits with the observation that educated citizens have become more aware and supportive of democratic institutions, but remain unimpressed with Nigeria s partly free democracy (Freedom House 2013). We suggest that such greater awareness and willingness to exercise political rights can support greater accountability (Besley 2006) and reduce susceptibility to vote buying (Kramon 2009), thereby supporting democratic consolidation. Given the religious and ethnic diversity of Nigeria s large population, low levels of democratic accountability and its struggles with democratic consolidation, it represents an important case in its own right, but also provides considerable insight into the prospects of many developing countries as they seek to 4

5 achieve universal primary education. This paper proceeds as follows. Section 2 generates hypotheses linking schooling and civic and political engagement in developing countries. Section 3 provides an overview of education policies in Nigeria, focusing particularly on the 1976 UPE reform. Section 4 details the research design. Section 5 shows our main results and robustness checks. Section 6 concludes. 2 Theory: education in weakly institutionalized environments 2.1 Civic and political engagement and support for democracy Education is generally viewed as unequivocally beneficial for civic and political interest and participation in developed countries (Almond and Verba 1963; Converse 1972; Lipset 1959; Putnam 2000). Converse (1972:324) even argues education is everywhere the universal solvent, and the relationship is always in the same direction. Explanatory mechanisms for this association include: education imparting the skills to engage and understand politics (Almond and Verba 1963; Rosenstone and Hansen 1993; Verba, Schlozman and Brady 1995), especially where the curriculum is civically-oriented (Niemi and Junn 1998); breeding greater tolerance and rationalism (Lipset 1959); bringing individuals closer to decision-making power, and thus incentivizing participation (Nie, Junn and Stehlik-Barry 1996) and mobilization by political elites (Rosenstone and Hansen 1993); and education raising the benefits of engagement within an individual s social network (Abrams, Iversen and Soskice 2010; Glaeser, Ponzetto and Shleifer 2007). Education may also have significant downstream effects through an individual s income or social network. However, an influential literature has challenging these theoretical foundations, arguing that education reflects family background (Jencks et al. 1972), early life experiences (Kam and Palmer 2008), the social hierarchy (Campbell 2009; Nie, Junn and Stehlik-Barry 1996), and cognitive ability (Luskin 1990). Since education is a proxy for other characteristics, these authors believe 5

6 education s correlation with participation is spurious. This highlights the important concern regarding selection into additional education. Many empirical studies in advanced democracies document positive correlations between education levels and civic and political engagement outcomes, especially turnout and political information. Focusing primarily on political participation outcomes in the US, studies have utilized innovative research designs to obviate selection problems and identify large positive causal effects for high school and university (e.g. Berinsky and Lenz 2011; Dee 2004; Marshall 2013; Milligan, Moretti and Oreopoulos 2004; Sondheimer and Green 2010). While the evidence from developed democracies generally finds education increases civic and political engagement, it is far from clear that the same logics apply in weakly institutionalized developing countries where support for democracy itself is not guaranteed. Cross-national correlative evidence using Afrobarometer surveys provides tentative favorable evidence: access to public goods measured by schooling and health clinics increases turnout, vote registration, protest and contacting political leaders (MacLean 2011), while Mattes and Bratton (2007) show that demand for democracy across sub-saharan Africa increases in procedural understanding, information and awareness, and education increases tolerance (Bratton, Mattes and Gyimah-Boadi 2005). However, we discuss evidence below suggesting education could increase political capture and extremism. Moreover, many voters express dissatisfaction with the performance of democracy (Bratton and Mattes 2001; Mattes and Bratton 2007), but simultaneously fail to appreciate how governments can be held to account by citizens and opposition parties beyond polling day (Bratton and Logan 2006). As noted above, these are particularly salient issues as consolidation of Africa s third wave democratization stalls. Studies over the last decade have examined education-oriented donor-sponsored interventions in developing democracies, finding consistently large increases in political knowledge and local participation but mixed behavioral outcomes like voting and weak effects on democratic attitudes. Civic education programs are positively correlated with local political participation in Argentina, 6

7 the Dominican Republic, Poland and South Africa (Morduchowicz et al. 1996; Finkel 2002) but increase citizen knowledge, values and voter registration in Zambia without affecting voting in national elections (Bratton et al. 1999). Researchers assessing participatory civic education argue only high-quality participatory provision produces large effects: phases I and II of Kenya s National Civic Education Programme increased political knowledge, tolerance and local-level participation without robustly affecting democratic values (Finkel and Smith 2011; Finkel, Horowitz and Rojo-Mendoza 2012), finding large social network externalities (Finkel and Smith 2011), while high-frequency high school civics instruction in South Africa increased political knowledge and helped students navigate how democratic values and skills relate (Finkel and Ernst 2005). Experimental studies address the concern that such observational data remains vulnerable to selection biases, but provide similar but less sanguine results. Friedman et al. (2011) use experimental variation to show school performance incentives provided by Kenya s Girls Scholarship Program generated interest in and knowledge of democracy, but did not affect support for democracy, voting or participation among young, poor and rural women from weak ethnic groups in a male-dominated Kenyan society. Similarly, a randomized community civic education program discussing decentralization and democracy using an encouragement design in Democratic Republic of Congo finds increases in political knowledge across a range of questions, but did not increase feelings of political efficacy or change in attitudes toward democracy or produce network spillovers (Finkel and Rojo-Mendoza 2012). Gottlieb (2012) randomly assigned information about government capacity to villages in Mali, and found increases in civic information and a greater likelihood of voting according to government performance in hypothetical simulations. There is also evidence that informal education can increase political participation: Kuenzi (2006) finds providing informal basic numeracy and literacy education to Senegalese adults who missed school as children considerably increases their propensity to vote, contact public officials and participation in the community. However, there are reasons to doubt such optimism in divided societies and remain cautious 7

8 regarding results identified barely a year after a program was administered. Linz and Stepan (1996) argue that democratic participation and norms cannot be fostered until an effective state bureaucracy has been established, while Almond and Verba (1963) speculate that democratic norms are part of a long-term process. Furthermore, even if such effects are robust, such participatory behavior may not be as benign as in the developed world. In weakly institutionalized environments education has ambiguous effects on political capture. Nichter (2008), Blaydes (2006) and Kramon (2009) respectively find evidence that vote buyers target the least educated in Argentina, Egypt and Kenya. Education could thus reduce political capture as educated voters choose to cast their preferences rather than sell their vote. However, education could also itself engender institutional capture: in a Kenyan field experiment providing funding and skills to disadvantaged women s groups, Gugerty and Kremer (2008) find that rather than increase productivity the intervention caused treated groups to be taken over by bettereducated and richer women. Surveys results from Zambia also show that civic education affects citizen knowledge and values among privileged members of society (Bratton et al. 1999). Whether education cultivates extreme political views and behavior remains contentious. Finkel, Horowitz and Rojo-Mendoza (2012) show that civic education reduced support for ethnic and political violence following Kenya s disputed 2007 election. On the other hand, Friedman et al. (2011) find participation in the Kenya s Girls Scholarship Program did not affect traditional modes of political participation, but increased a desire for autonomy, not necessarily through democratic means, as well as sympathy for the use of violence in politics. More directly, Krueger and Maleckova (2003) find education secondary education is correlated with becoming a Hezbollah fighter, while Berrebi (2007) finds Palestinian suicide bombers are more likely to have secondary or university education. 8

9 2.2 Moving forward Although there is a now a burgeoning literature examining political behavior in Africa, causal identification is hard to come by. The few studies emphasizing causal estimates have focused on short-term effects of NGO-provided programs beyond formal public education, and generally rely upon regression and matching techniques to address selection bias concerns. Even in the experimental case of Friedman et al. (2011), the effects they identify are highly localized to a particular set of young marginalized female compliers. Furthermore, unlike results for political knowledge, effects on behavioral outcomes have not proved decisive. We fill these important gaps in the literature by estimating the causal effects of publicly-provided primary education among Nigerian adults, for a large and highly empirically relevant set of compliers who would not have been educated without UPE, by exploiting regional variation across genders in the intensity of the UPE program. Given the fact that Nigeria is characterized by low demand for vertical accountability (Bratton and Logan 2006), and had early experience with universal primary education, it represents an ideal testing ground for these hypotheses. Moreover, as discussed below, Nigeria s primary school curriculum is academically and civically oriented (Asagwara 1997; Csapo 1983), and thus represents the type of case where education could enhance individuals civic and democratic character. This paper focuses on identifying the direction of the effect of primary schooling on civic and political engagement, in addition to examining whether any positive effects have anti-democratic implications. Since voting is a low-cost activity, it is important to examine more time-consuming and advanced forms of engagement to assess education s effects. Building on the extant literature, we test the following hypotheses: H1 Primary schooling increases interest in politics, political participation, community participation and support for democracy. H2 The positive political effects of primary schooling are largest among those who stand to gain 9

10 from political capture. H1 assesses the optimistic findings from advanced democracies that have not received decisive support in developing countries, while H2 represents the concern that the participatory benefits of education may be subverted in weakly institutionalized settings. We next overview primary education in Nigeria before testing these hypotheses. 3 Primary education in Nigeria Nigeria is Africa s most populous nation containing 162.5m people in 2011, and is one of the continent s most ethnically, linguistically and religiously diverse. Nigeria is also a major oil and gas producer, but ranks poor by GDP per capita terms. Despite this, Nigeria has experimented with some of the largest-scale and most progressive education policies in the developing world. For all these reasons, Nigeria represents an important case with implications for many other countries. This section provides a brief historical overview of education in Nigeria, before detailing the 1976 educational reforms that underpin our identification strategy. 3.1 Background Prior to independence in 1960, Great Britain had divided the country into three semi-autonomous administrative regions: the predominantly Muslim North, Christian East, and mixed West regions. 1 European-style education was introduced to Nigeria under British colonial rule in the 1840s, but was provided by Christian missionaries seeking to civilize and convert the local population as the British government preferred to provide missionaries with grants than establish formal education (Fafunwa 1974). The distribution of missions in Figure 1 shows that while the Western and especially Eastern regions were relatively densely populated with missions, large parts in the North were poorly served. Western-style education was often prohibited in most of the North so as to 1 Former federal capital Lagos was semi-autonomous. 10

11 avoid conflict with local religious leaders, the colonial government did not want missionaries interfering with Islamic practices. Up until the 1950s, these missions served as the primary source of education, and thus entrenched early disadvantages among Northerners. Figure 1: Christian missionary stations in Nigeria 1925 (source: Streit 1929) Universal government-supported education, a response to the widespread inability of parents to pay schools fees, started in the mid-1950s. In 1955, the Western region implemented a program of free six-year universal primary education. After doubling enrollment within a year (Csapo 1983), this was extended to Lagos and the East in 1957 with similarly dramatic enrollment increases (Abernethy 1969), and the North in 1958 (Bray 1981; Fafunwa 1974). However, these programs varied considerably in the length of education provided, how they were financed, and their success in enrolling students (Bray 1981). From its inception, the East experienced severe financial problems and lacked trained teachers, as well as considerable reluctance by the local-majority Catholic church (Achor 1977); consequently the program was quickly scaled back. Enrollment was particularly low in the North (Adewole 2012), with less than 5% of children attending primary school 11

12 in 1970 and especially low rates for girls (Csapo 1983), which reflected the colonial government s continued unwillingness to interfere with Muslim practices (Achor 1977; Csapo 1983; Fafunwa 1974; Osili and Long 2008), traditional attitudes towards women (Csapo 1983; Niles 1989) and lack of funding (Achor 1977). After independence most primary education programs were cutback, with Nigeria s newly-designed regions differing in their willingness to fund education (Osili and Long 2008; Oyelere 2010) Universal Primary Education program By 1976, Nigeria had 19 states, which for historical reasons varied considerably in their primary education policies, schooling capacity and enrollment. Against this backdrop and buoyed by their oil revenue boom, Nigeria s post-independence military government led by aspiring nation-builder Olusegun Obasanjo announced in 1974 one of the most ambitious education projects in African history (Bray 1981:1). 2 Starting 1st September 1976, the government implemented its nationwide Universal Primary Education (UPE) program. The program provided six years of free primary education starting from six years of age for all students, and aimed for 100% primary enrollment by 1981 (Csapo 1983), while enrollment would become compulsory by This reflected the goals of rapid economic development and unifying the nation by reducing biases in educational and economic development across states and genders (Csapo 1983). In order to implement UPE, significant federal government investment was required, especially in classroom capacity, teacher training and teaching equipment. Due to the different educational histories across different parts of the country, investments varied considerably across the country. Osili and Long (2008) show that federal budgetary allocations for primary school construction 2 Although this policy was clearly related to the oil revenue boom the country benefited from in the 1970s (Asagwara 1997; Csapo 1983; Osili and Long 2008), when Obasanjo returned to power in the 1990s he re-implemented UPE in 1999 (extending it to nine free and compulsory years of education) as Universal Basic Education without such an oil boom, while the Western region had previously supported UPE without oil revenues (Oyelere 2010). 12

13 across states, totaling 700m Nairu, reflect this differential, with per capita funding disproportionately distributed to Eastern and particularly Northern states. This funding intended to construct 150,995 new classrooms by 1980, of which 106,505 were to be built in the North (Csapo 1983), in addition to 80,000 new teachers and 6,699 new classrooms for teacher training (Nwachukwu 1985). Given the prevalence of corruption in Nigeria, translating budgetary allocations into school accessibility outcomes is not straight-forward. Accordingly, to verify the system s expansion we turn to school construction data. Figure 2 shows public school construction spiked around 1976 to accommodate UPE. Despite free provision in the 1960s in some parts of the country, only the significant expansion around the UPE reform ensured that most students could attend school. After 1976 school construction was comparatively low. Figure 2 also shows that the rise is almost entirely due to public, not private, school construction Founded Public primary schools Private primary schools Figure 2: Number of public and private primary schools founded since independence (source: Nigerian Primary School Census 2008) 13

14 There are two main channels through which the UPE reforms encouraged primary school attendance. First, free provision lowered the price in most areas except the remaining pockets of free schooling in the West. Second, the dramatic rise in school availability meant that schools were less capacity-constrained and students were required to travel much shorter distances to attend school. The importance of such availability differed greatly across the country. Since both cost and availability incentives were strongest in the North, and given initial enrollment rates were much lower, we expect to find larger effects on individual enrollment. This pre-reform gap between actual and universal primary education is key to our identification strategy. UPE dramatically increased enrollment nationwide. The number of students in primary school rose, much faster than population growth, by 124% from 4.4m in 1974 to 13.8m in 1981 (Osili and Long 2008; Oyelere 2010) this exceeded government expectations, based on the 1963 Census, of 11.5m in 1980 (Bray 1981; Csapo 1983). In time for the new school year starting September 1976, 2.5m students registered to attend school for the first time (Achor 1977). The gross male primary enrollment rate increased from 60.3 in 1974 to in 1981, while the gross female enrollment rate increased from 40.3 in 1974 to in 1981 (Osili and Long 2008). Given huge disparities between their initial levels, differences were also pronounced by region. As Figure 3 using the Afrobarometer samples used in our empirical analysis illustrates, enrollment increases were particularly marked in local governments areas (LGAs) with below-median initial enrollment by gender ( high intensity cases). Nevertheless, it is important to note that as many as 25% of students in Benue and Plateau dropped out before completing primary school after UPE s introduction (Csapo 1983); such incomplete primary schooling is reflected in our empirical analysis. As well as literacy and numeracy, the Nigerian primary curriculum emphasized national unity, citizenship rights and obligations and effective community at an early age. The military government set seven grand objectives for the UPE curriculum in its National Policy on Education 1977: inculcating literacy, numeracy and communication; sound basis for effective thinking; citizenship education; character and moral training; developing adaptability; skills to function in the local 14

15 Proportion completed primary school Year of birth Low intensity High intensity Figure 3: Trends in primary school enrollment by UPE intensity (source: Afrobarometer) community; and preparation for further educational enhancement (Achor 1977). Although in practice such objectives did not deviate from the bookish pre-upe culture of schooling (Asagwara 1997:200; Csapo 1983) and were more aspirational than realistic, Nigerian schooling aimed to instill the cognitive skills required by students to participate effectively in political and community affairs. The school day started and ended each with a national pledge to eradicate illiteracy and make education a right (Csapo 1983). Despite increased enrollment, primary schooling did not become more vocationally-oriented (Asagwara 1997; Csapo 1983). The 1976 program had mistakenly assumed oil revenues would persist (Csapo 1983) alongside economic growth of 5-10%, of which 25% could be captured as tax revenues (Achor 1977). Once the civilian government (handed power in 1979) was forced to finance an already under-funded UPE program, the policy ended in 1981 after an initial period where local governments were made to bear some of the burden, and primary education was not made compulsory as planned in 1979 (Csapo 1983). Following this, most states reintroduced school fees excluding the West- 15

16 ern states dominated by the United Nigeria Party -as the federal government no longer provided grants for teacher salaries and training (Osili and Long 2008). Nevertheless, Figure 3 shows that enrollment barely changed, and with time continued to increase (see also Asagwara 1997). Importantly, this implies that school availability and better inputs (such as trained teachers) rather than fees principally drove enrollment decisions. Supporting this claim, Ozigi and Ocho (1981) find that the experience of UPE raised Northern parents willingness to pay for schooling. 4 Research design Nigeria holds federal elections for the offices of President, Senate and House every four years. The 1999, 2003 and 2007 elections which followed a period of military rule correspond to our sampling period. The People s Democratic Party (PDP) has retained the presidency and legislative majorities throughout this period. Former military ruler Olusegun Obasanjo won the presidency in 1999 and 2003, while Umaru Yar Adua won the 2007 election. However, rather than free and fair, such elections regularly experienced vote buying, polling irregularities and violence during the campaign and on election day (e.g. Bratton 2008). This section first describes the Nigerian survey data containing political responses. We then explain how we identify the effect of primary schooling on civic and political engagement by using difference-in-differences and instrumental variable strategies, leveraging differences, by gender, in the impact of UPE across today s 774 LGAs. 4.1 Survey data The main dataset used in this paper is the Afrobarometer, which is a nationally-representative sample of voting-age citizens. 3 The Afrobarometer surveys economic, political and social attitudes 3 Surveys are random samples stratified (first by region) by population sizes. Individuals in institutionalized settings are excluded. 16

17 across up to 35 African countries. We use all four rounds for Nigeria, which cover with samples conducted every two years. Our main dependent and independent variables, described below, draw from this dataset. To construct our variables measuring the differing intensity of the 1976 UPE reform we draw upon several Nigeria-specific datasets; see following subsection Dependent variables In order to address our hypotheses, we examine a range of dependent variables under four main categories: interest in politics, political participation, community participation and support for democracy. These outcomes extend beyond low-cost activities like voting to consider fundamental ideological building blocks for establishing effective democratic accountability such as support for democratic institutions and social capital. Detailed variable definitions and summary statistics are provided in the Appendix. Interest in politics is measured by two variables. First, Discuss politics often is a dummy for the 19% of the sample that responded that they frequently discussed politics with friends or family. 4 Second, we created News scale, a summative rating scale averaging ordinal five-point scales asking how frequently respondents follow the news on television, by radio or in newspapers; the scale has a high Cronbach s alpha inter-item reliability score of We measured political participation using five behavioral indicators. Registered voter and Voted are dummies for the 78% and 62% of the sample that are registered voters and voted at the last federal election. Attend demonstration is a dummy capturing active protest in the last year. Participation is also measured by contacting political figures; accordingly, Contact local councilor and Contact MP are respectively dummy variables for respondents who contacted the relevant political figure in the last year. Contacting elected politicians, especially national MPs, is far less prevalent. Community participation is measured by group membership and attendance. We code dummy 4 Since 67% of individuals stated that they discuss politics occasionally, we lack the variation to examine an alternative frequency. 17

18 variables for Attend community meeting in the last year and current membership of and active participation in local associations Association member and Active association member over the last year. Following the existing literature, support for democracy is differentiated by support in the abstract and support for its practice. Intrinsic support implies support for democracy as the best form of government in the abstract, regardless of socioeconomic circumstances. To approximate this, we measure Support democratic institutions using a summative rating scale combining eight items abstractly characterizing liberal democracy with separation of powers: three-point ordinal scales of support for checks and balances and support for term limits, as well as dummies for opposition to one-party rule, military rule and rule by one man, opposition to presidential discretion, opposition to governments banning organizations, and support for freedom of press. The Cronbach s alpha reliability coefficient is 0.60, indicating the variables fit together coherently. Support for the existing practice of democracy in Nigeria is measured by a dummy for Satisfied with democracy and a Trust politicians scale averaging trust in six government institutions: the President, National Assembly, Independent Electoral Commission, local government council, ruling party, and opposition party (Cronbach s alpha of 0.84) Primary education The key explanatory variable in this analysis is primary school education. The Afrobarometer asks respondents about their education, providing six responses: no schooling, incomplete and complete primary school, incomplete and complete secondary school, and some college. The complete sample of responses in Nigeria is shown in Figure 4. Figure 4 shows that 20% of the sample has no education at all, while a further 16% have at most completed primary school. Given Nigeria s relatively progressive education policies in Africa, the proportion attending secondary school is high, although progression to university is low. Since this paper will be focusing on Nigeria s UPE program, we are most interested in education up to 18

19 Low UPE intensity High UPE Intensity No education (19.5%) Incomplete primary school (5.0%) Complete primary school (10.6%) Incomplete secondary school (14.5%) Complete secondary school (44.5%) University (6.0%) Figure 4: Education distribution by UPE intensity primary level. Consequently, the analysis will utilize a three-category ordinal variable, Primary schooling, coded 0 for no education, 1 for incomplete primary school and 2 for complete secondary school. 5 Although imposing linearity on this relationship is not ideal, using a dummy for completed primary education can seriously upwardly bias instrumental variable estimates (Imbens and Angrist 1994). Since by 1976, most students had at least some primary schooling, our results are primarily identifying the effect of moving from incomplete to complete primary school. 4.2 Identification strategy To examine the effects of education on political behavior and their implications for democracy, we exploit gender and regional differences in the impact of Nigeria s UPE program to instrument for primary school education. We now detail how differences in the intensity of UPE program can be 5 The first-stage (below) shows that our instrument does not affects education above primary level. Unsurprisingly, using a 6-category scale instead does not substantively affect the results. 19

20 used to identify the effects of primary schooling Reduced form Our identification strategy which is similar to Duflo (2001) and Bleakley (2010) exploits temporal and spatial variation. 6 The temporal dimension distinguishes the periods before and after the 1976 UPE reform. Although UPE was abandoned in 1981, it had powerful persistent effects. The exact reason enrollment did not revert to pre-1976 levels is hard to discern school availability, input quality and information about the value of education are plausible explanations but it is sufficient for our purpose to note that primary school enrollment remained relatively steady after 1981 before continuing to increase (as Figure 3 shows). Since UPE affects all students of eligible age, a second dimension of variation is required to distinguish the introduction of UPE from cohort differences. As noted above, there is considerable variation in the enrollment effects of UPE across LGAs and by gender. This second dimension defines the intensity of the UPE reform. Such spatial variation permits a difference-in-differences (DD) strategy, where low-intensity areas serve as control units able to differentiate trends in education from the impact of UPE in high-intensity areas. This ameliorates the concern that the introduction of UPE simply reflects underlying trends across Nigeria. Given we are interesting in individual survey responses, we must map the intensity of UPE to individuals. To operationalize this, we count any individual born after 1969 who is thus eligible to benefit from UPE s educational expansion as impacted by the UPE program; 7 this defines the reform dummy Post-UPE. Like Bleakley (2010), we measure the differential intensity of UPE s 6 Similar identification strategies have been adopted across developing countries. 7 Note that students born after 1964 were eligible for at least some level of free schooling. However, Figure 3 shows that there was no sharp deviation before 1970, suggesting that this nuance is unimportant. Our results are robust to coding cohorts born after 1964 as UPE-eligible and removing the partially-eligible cohorts. Oyelere (2010) notes that grade skipping and over and under-age entry are also uncommon. 20

21 impact across LGAs by using the gap between actual and potential capacity to approximate the scope of the program s effect. Using representative local government survey data from the Harmonized Nigeria Living Standards Survey (HNLSS), 8 we define UPE Intensity as the male or female proportion of the LGA population for those born between 1960 and 1969 that had not completed primary school before the treatment. 9 Thus, UPE intensity varies by gender across LGAs, but does not vary over time. The kernel density distribution for our sample shown in Figure 5 shows that while many respondents lived in parts of the country with near-universal primary education for both genders, a large proportion of LGAs especially among female students did not. Density Intensity Male Female Figure 5: UPE intensity distribution across LGAs and by gender Individuals are mapped to LGAs based on their current LGA of residence. We assume that respondents were educated in the same LGA they currently reside in. This is unproblematic in 8 This survey was jointly carried out by the World Bank and Nigerian National Bureau of Statistics. Ten households from ten enumeration areas were surveyed in all 774 local governments of Nigeria, adding up to a total of 77,400 households surveyed in the country. 9 See robustness checks for alternative measures of intensity. 21

22 Nigeria, where migration is very rare: only 2.4% of HNLSS respondents had not always lived in their current town or village, while Osili and Long (2008) find no differences between movers and non-movers in levels of schooling. By interacting Post-UPE and Intensity, we can estimate the reduced form effect of UPE exposure in the following DD regression by including state and survey year fixed-effects: Y i,g,c,l,s,t = β 1 Intensity g,l + β 2 ( Post-UPE c Intensity g,l ) + X i γ + µ g + κ c + η s + ζ t + ε i,g,c,l,s,t, (1) where Y i,g,c,l,s,t is a political outcome variable, X i is a vector of individual-specific covariates (religion dummies and a rural-urban dummy 10 ), and µ g, κ c, η s and ζ t are respectively gender, cohort, state and survey fixed-effects. κ c subsumes the linear Post-UPE c term. Throughout we conservatively cluster standard errors by state. 11 Our DD specification enables us to focus on within-state, within-cohort and within-gender changes over time. The key identifying assumption in DD analyses is parallel trends. Parallel trends requires that without UPE changes in Y i,g,c,l,s,t would not have differed across high and lowintensity UPE areas. Figure 3 is consistent with this assumption, showing very similar trends across intensities before the reform. Furthermore, we show below that our results are robust to including state-specific time trends and performing placebo tests for cohorts unaffected by the reforms. 10 Given both variables are unlikely to have changed over the course of a respondent s life, they are counted as pre-treatment. The results are robust to their exclusion. 11 Although our variation in intensity is for the current 36 states (and Lagos), only 19 states existed in Our standard errors are slightly higher using fewer clusters, but remain highly statistically significant. 22

23 4.2.2 Instrumental variables In order to estimate the effects of primary schooling, we would ideally estimate the following equation using OLS: Y i,g,c,l,s,t = βprimary schooling i + X i γ + µ g + κ c + η s + ζ t + ε i,g,c,l,s,t. (2) Although we estimate such naive regressions for comparison purposes, these estimates are uninformative for two principal reasons. First, since which individuals receive longer schooling is unlikely to be (conditionally) random, such estimates may be substantially biased (Kam and Palmer 2008). Second, the effect of primary schooling is likely to be heterogeneous across individuals. While individuals that would have attended primary school anyway might not be expected to experience large effects, schooling could make a big difference for those with the lowest propensity to attend school because they are unlikely to benefit from other stimuli encouraging political engagement. In order to obtain unbiased estimates for the population of students that comply with UPE incentives, we use an instrumental variable (IV) strategy. IV estimation builds upon the reduced form estimation strategy by estimating a DD specification for our first-stage: Primary schooling i,g,c,l,s,t = α 1 Intensity g,l + α 2 ( Post-UPE c Intensity g,l ) +X i γ + µ g + κ c + η s + ζ t + ε i,g,c,l,s,t, (3) using OLS, where Post-UPE c Intensity g,c,l is the excluded instrument. Given our endogenous variable, primary schooling, takes three values, we cannot interpret our results as local average treatment effects (Imbens and Angrist 1994). However, with discrete treatments we can estimate the local average causal response (LACR) for UPE intensity-compliers by weighting the causal effect at each value of primary schooling by the proportion of people affected by the instrument at that value (Angrist and Imbens 1995). We use 2SLS to estimate equation 23

24 (2) using the predicted values from equation (3) to approximate this weighting (Angrist and Imbens 1995). 12 In addition to the DD parallel trends assumption, identifying the LACR of primary school on political outcomes requires a strong first-stage, monotonicity, and an exclusion restriction requiring that UPE intensity has no effect on political outcomes except through increasing primary schooling. Unlike OLS, which averages across the entire sample, IV estimates apply only to instrument-compliers. The first-stage for our excluded instrument can be easily verified. Column (1) in Table 1 shows a large positive effect for this interaction. Substantively, this confirms that the reform was most effective in raising primary schooling in LGAs which experienced greatest school construction and/or had the lowest initial rates of primary enrollment. The relationship is strong, yielding an F statistic of 23.6 for the inclusion of our excluded instrument; this comfortably exceeds the recommended minimum of 10 (Staiger and Stock 1997). Checking the parallel trends assumption, columns (2) and (3) shows the first-stage results are robust to including state-specific survey year and birth year (cohort) time trends, although unsurprisingly the coefficient magnitude declines somewhat. Restricting the sample to those born before 1970 and using 1959 as a placebo reform, as expected column (4) finds no clear effect for the interaction. The remaining regressions examine dummies for attaining each level of schooling. Columns (5) and (6) show that the interaction has a large effect on incomplete primary schooling, but there remains a strong first-stage for completing primary schooling; supporting our three-point linear measure of primary school, the coefficients are broadly consistent with a linear relationship. Justifying our decision to restrict attention to primary schooling, columns (7) and (8) show that the UPE intensity does not affect secondary school attendance. The key IV assumption is the exclusion restriction. We discuss this assumption in detail below, and provide strong evidence to dismiss plausible violations of IV s key identifying assumption. 12 With covariates, 2SLS requires re-weighting by covariate values (Abadie 2003). We ignore this subtlety because 2SLS provides a good approximation. 24

25 Table 1: First-stage effect of UPE intensity on schooling Primary Primary Primary Primary Incomplete Complete Incomplete Complete schooling schooling schooling schooling primary primary secondary secondary (1) (2) (3) (4) (5) (6) (7) (8) OLS OLS OLS OLS OLS OLS OLS OLS Post-UPE Intensity 0.333*** 0.281*** 0.221*** 0.193*** 0.140*** (0.069) (0.071) (0.066) (0.033) (0.038) (0.042) (0.035) Placebo Intensity (0.089) Linear state trends No Survey Cohort No No No No No Observations 15,145 15,145 15,145 5,218 15,145 15,145 15,145 15,145 F statistic Mean Outcome St. Dev. Outcome Mean Treatment St. Dev. Treatment Notes: standard errors clustered by state in parentheses; * denotes p < 0.1, ** denotes p < 0.05, *** denotes p <

26 5 Results 5.1 UPE, primary schooling and political engagement Tables 2-5 report OLS, reduced form and 2SLS results estimating population-average correlations, the effect of exposure to high-intensities of UPE, and the IV estimate for primary schooling respectively. The reduced form estimates capture the full effect of UPE intensity, while the 2SLS estimates identify primary schooling s effect on political behavior for UPE program-compliers. Both the reduced form and 2SLS estimates suggest that UPE had considerable political implications. The UPE-compliers, which our 2SLS estimates identify the effects for, are individuals that only attended primary school because of UPE. They come from relatively disadvantaged backgrounds, given education is an important status symbol in Nigeria, unlike the high-status types Nie, Junn and Stehlik-Barry (1996) analyze. Given such compliers are less likely to have benefited from other stimuli promoting civic and political engagement, it is not surprising to find large IV effects for this group, as is common for such groups (e.g. Finkel and Smith 2011; Friedman et al. 2011; Zaller 1992). Furthermore, since we estimate effects up to 33 years after respondents benefited from UPE, we pick up many downstream effects of education (e.g. increased income, social status), which also account for larger effects than studies surveying participants shortly after any treatment Interest in politics The UPE intensity reduced form and 2SLS primary schooling estimates in Table 2 strongly support the hypothesis that forced schooling increases measures of interest in politics. In particular, the IV estimate in column (3) reinforces the 5% percentage point reduced form effect by showing a unitincrease in primary school increases a respondent s propensity to frequently discuss politics by 16 26