THE EFFECT OF EDUCATION ON POLITICAL PARTICIPATION IN ELECTORAL AUTHORITARIAN REGIMES: EVIDENCE FROM ZIMBABWE

Transcription

1 THE EFFECT OF EDUCATION ON POLITICAL PARTICIPATION IN ELECTORAL AUTHORITARIAN REGIMES: EVIDENCE FROM ZIMBABWE KEVIN CROKE GUY GROSSMAN HORACIO A. LARREGUY JOHN MARSHALL NOVEMBER 2014 A large and influential literature suggests that education increases political participation. However, this relationship may not hold outside advanced and consolidating democracies. In electoral authoritarian regimes, educated voters may instead deliberately disengage. If education increases critical capacities, political awareness and support for democracy, educated citizens may believe that participation is futile or falsely legitimizes autocrats. We test this argument in Zimbabwe, a paradigmatic electoral authoritarian regime governed by Robert Mugabe since Exploiting a major education reform that induced variation in access to secondary education across cohorts, we find clear evidence that education decreases individual political participation: surveys conducted since 1999 show that citizens induced to attain more education by the reform are substantially less likely to vote, contact local Councillors or attend community meetings. Supporting the disengagement hypothesis, we find that such educated citizens experience better economic outcomes, are more interested in politics, and are more supportive of democracy, but are also more likely to be critical of the government and support opposition parties. Consistent with our conclusion that education s participatory effects are conditional upon sufficiently democratic institutions, education s negative effect on participation has dissipated following the 2008 election, which for the first time gave the opposition a share of executive power. This article benefited from helpful conversations with and suggestions from Karen Grépin. Department of Global Health and Population, Harvard School of Public Health, Boston, MA (kcroke@hsph.harvard.edu). Department of Political Science, University of Pennsylvania, Philadelphia, PA (ggros@sas.upenn.edu). Department of Government, Harvard University, Cambridge, MA (hlarreguy@fas.harvard.edu). Department of Government, Harvard University, Cambridge, MA (jlmarsh@fas.harvard.edu). 1

2 Introduction The question of what motivates political participation is central to political science. Participation is essential for holding governments to account, and for influencing incumbents to adopt some actions and policies and not others. A vast literature points to education as a major driver of voting, as well as many other forms of civic action (Almond and Verba, 1963). In fact, it has been argued that the positive relationship between education and political participation is one of the most reliable results in empirical social science (La Due Lake and Huckfeldt, 1998, 567). In a similar vein, Hillygus (2005, 25) states that the idea that education is a primary driver of increased political participation is largely uncontested, while Putnam (1995) posits that education is the best individual level predictor of participation. There are, however, several problems with these law-like assertions. First, isolating the effect of education as distinct from innate ability (Dawes et al., 2014), socioeconomic status (Jennings and Niemi, 1968), or family background (Nie, Junn and Stehlik-Barry, 1996) on political participation, is a formidable challenge (Rosenzweig and Wolpin, 2000). Debates between education as cause versus education as proxy remain far from settled (Kam and Palmer, 2008; Henderson and Chatfield, 2011). Second, and with few exceptions, the existing empirical literature investigating the causal link between education and political participation has generally focused on a small set of rich advanced democracies (Sondheimer and Green, 2010). Analyses in developing countries have primarily focused on aggregate correlations between education and the likelihood that autocratic regimes transition to democracy, 1 or the association between education and violent forms of protest (Samii and West, 2014; Campante and Chor, 2012a). The lack of individuallevel analyses of the determinants of political participation during normal times is particularly acute in non-democratic and semi-democratic settings. The study of politics in such contexts has only recently experienced a revival as the nature of dictatorial regimes has changed (Schedler, 2013). This paper examines the causal relationship between education and political participation under electoral authoritarianism. Electoral authoritarian regimes are a hybrid: while they permit some popular participation and elite contestation by holding periodic elections, they fall a long way short of genuine democracy. In such regimes, elections while not purely pro forma are far from fair, the government is almost assured of 1 See, Acemoglu et al. (2005), Barro (1999), Kamens (1988), and Murtin and Wacziarg (2014). See Glaeser, Ponzetto and Shleifer (2007) for the role education plays in consolidating democracy. 2

3 remaining in power, and many other forms of political action are closely monitored and often limited by the regime (Levitsky and Way, 2010; Schedler, 2002, 2013). What political role do the more-educated citizens play in such settings? Improving our understanding of the effect of education on political participation in electoral authoritarian regimes is important for both theories of regime stability and change. If education is in fact a critical driver of political participation, as the extant literature suggests, then investment in education in these contexts could be a key tool to improving citizen welfare by more effectively holding governments to account (Manin, Przeworski and Stokes, 1999) and by generating pressure for greater citizen inclusion (Lerner, 1958; Lipset, 1959). If, however, education is associated with decreased political participation in electoral authoritarian regimes as we suggest below then this might shed light on one reason why, contrary to the expectation of the democratic transition literature (Linz and Stepan, 1996), many countries that took initial steps towards liberalization got stuck in electoral authoritarian equilibria (Carothers, 2002). There are some good theoretical reasons to suspect that the positive relationship between education and political participation does not necessarily apply in electoral authoritarian settings. Political participation in mature and some consolidating democracies can be understood as the embodiment of the liberal notion of free will (Rousseau, 1997). The act of voting, in particular, has been viewed as a manifestation of the democratic ideals of political equality and individual agency. 2 Yet, by contrast, in many non-democratic settings, quintessential political actions like voting cannot necessarily be viewed as the exercise of agency and free will. Indeed, various forms of political participation in electoral authoritarian settings are designed to mobilize citizens to demonstrate allegiance to the regime, and not to aggregate societal preferences or enable citizen voice (Hermet, 1978). Elections in many electoral authoritarian regimes are not designed to allow citizens to freely elect their leaders, but rather to provide incumbents with a facade of legitimacy, appease the international community, and demonstrate the omnipresence of the regime (Levitsky and Way, 2002). Under such conditions, political participation is likely to lose both its normative and instrumental appeals for many. 2 For this reason, many see participation measures, such as voter turnout, as a key indicator of democratic quality (see, Lijphart (1997)). A normative commitment to political equality can explain why one of the main areas of emphasis for the vast voter turnout literature is on increasing the political participation of the poor and uneducated. This focus, however, may have contributed to ignoring the issue of low levels of participation among more educated and wealthier members of society (Kasara and Suryanarayan, 2014). 3

4 When participation cannot achieve its goals of providing genuine input into the political process, or when participation merely serves to buttress the regime, then refraining from political participation can serve as a powerful form of dissent (Hermet, 1978; Karklins, 1986) or may reflect the recognition that costly political action is futile (Downs, 1957). This study s key insight is that more educated citizens are more likely to exercise deliberate political disengagement. First, education imbues citizens with cognitive abilities that allow for more critical thinking, which may then translate into lower levels of support for the incumbent regime, and thus less interest in legitimizing it with their participation. 3 Second, education may similarly lead to value change, with more educated individuals placing a higher premium on democratic values such as self-expression and individual voice, in contrast to social conformity and solidarity (Inglehart and Welzel, 2005). Third, if education also increase knowledge and understanding of politics, educated voters may be more aware that political participation will not affect political outcomes. Fourth, educated individuals may feel a higher level of disillusionment with autocratic politics and economic mismanagement, given their greater economic potential (Campante and Chor, 2012b, 2014). We test this disengagement argument using the case of Zimbabwe. First, Zimbabwe under President Mugabe is a paradigmatic electoral authoritarian regime, ruled by a civilian-military junta (Levitsky and Way, 2010; Sithole, 2001). 4 While elections have been held regularly since 1980, when the opposition has mounted a credible threat to win power (most notably in 2008), the regime ratcheted up violence and engaged in massive vote rigging to ensure that the opposition could not take power. Thus, as in many electoral authoritarian regimes, elections in Zimbabwe provide some restrictive opportunity for public opinion to be registered. They have yet to offer voters, however, a genuine ability to determine the ultimate distribution of power. Second, Zimbabwe is an excellent case study because we are able to leverage a major educational reform to identify the causal relationship between education and participation. Immediately after majority rule was achieved in 1980, Zimbabwe implemented a far-reaching education reform that created a discontinuity in access to education between cohorts. The reform, which removed examination score, geographic and finan- 3 For example, educated citizens may be able to critically evaluate regime propaganda, or may have the interest or financial means to access more critical foreign media. 4 Though it allows for elite contestation, members of the ruling party, the African National Union Patriotic Front (ZANU-PF), believe that ZANU-PF has a right to rule in perpetuity and thus will never recognize an opposition victory at the polls (Bratton and Masunungure, 2008, 43). 4

5 cial constraints, substantially increased access to secondary education for black students, but also induced some students to remedially attend primary school. We exploit this natural experiment to analyze the effect of education on political participation by comparing cohorts that were just young enough to enjoy greater access to secondary education to cohorts that were just too old to take advantage of the reform. To allow for non-compliance across cohorts, we also estimate instrumental variable regressions. Our primary result is that, at least in Zimbabwe, education reduces levels of political participation. Contrary to the conventional wisdom, a higher level of education reduces not only voting, but also other forms of non-contentious participation such as contacting one s local Councillor and attending community meetings. Consistent with our argument that non-participation may be an informed choice by relatively cognizant and politically aware citizens, we find that education significantly increases interest in politics, support for democracy, and economic well-being. Furthermore, consistent with our claim that such citizens become disenchanted with authoritarian politics, we also find that increased education decreases support for the ruling party, reduces perceptions of government performance, and increases support for the main opposition party. We interpret this combination of reduced political participation coupled with improved political information, better living standards, greater support for democratic institutions and reduced support of the incumbent autocrat, as evidence that educated citizens are consciously choosing to withdraw from the political sphere. To increase confidence in our interpretation of results we examine two plausible alternative explanations for our findings. First, we rule out the possibility that more educated constituents are less integrated into patronage or vote buying mobilization networks, which could be another channel leading to lower turnout. Second, we show that there is no evidence in our data that educated constituents are more likely to face violent repression, designed to suppress participation. Furthermore, we show that the large negative relationship between education and participation weakens significantly after the 2008 election, which (temporarily) resulted in a power sharing agreement between ZANU-PF and MDC, the main opposition party. This finding suggests that more educated constituents are willing to re-engage with politics when there is an opening of the political sphere allowing for more meaningful contestation. This paper s main contribution is to the vast literature on the relationship between education and political participation. Notwithstanding the centrality of education for political theories of democratic citizenship, 5

6 much is still unknown about the nature of the relationship between educational attainment and political attitudes and behavior. If education reduces political participation in electoral authoritarian regimes, this provides an important qualification to the literature that commonly assumes a law-like positive relationship. We show that education increases the ability of citizens to participate in politics, and leads to greater interest in politics. However, our findings also suggest that whether citizens decide to use these facilities or deliberately withdraw from the political arena is likely to depend on systemic attributes; specifically whether the regime is sufficiently democratic and receptive to bottom-up input. The paper also contributes to the literature on political participation in developing countries, especially that which focuses on attitudes rather than resources as the key determinant for participation. For example, Cox (2003) and Norris (2002) find a positive association between institutional trust and voter turnout, and Kasara and Suryanarayan (2014) show that voting turnout for higher-resource constituents depends on the congruence of their political preferences with those of the poor and on the state s taxation capacity. In addition, our results also challenge a body of work that argues that education serves as a tool of indoctrination and social control, instilling obedience to authority (Lott, 1999; Kremer and Holla, 2009). Our findings suggest instead that education supports critical thinking, which can lead citizens, in political settings such as Zimbabwe s, to oppose authority deemed illegitimate. Finally, our results speaks to a growing cross-country literature seeking to isolate the impact of education on democracy (Acemoglu et al., 2005; Glaeser, Ponzetto and Shleifer, 2007; Murtin and Wacziarg, 2014; Woodberry, 2012). Our paper speaks to this literature by demonstrating that the impact of education might be conditional on a country s level of contestation. Related literature The relationship between education and political participation in advanced democracies is the subject of a vast literature. Whereas early work suffered from insufficient attention to causal identification, a number of recent studies have been able to credibly identify a positive effect of education on political participation in the developed world. 5 By contrast, the study of the effect of education in developing countries is less 5 Notable examples that offer causal evidence on this relationship in developed countries include Sondheimer and Green (2010); Milligan, Moretti and Oreopoulos (2004); Dee (2004); Berinsky and Lenz (2011); Persson (2011); Tenn (2007); Kam and Palmer (2008); and Henderson and Chatfield (2011). 6

7 developed and still suffers, to the most part, from insufficient attention to causality. 6 Importantly, none of the recent studies that seek to identify a causal relationships between education and participation in a developing country has examined this question in the context of a repressive electoral authoritarian regime. Friedman et al. (2011) use a field experiment in Western Kenya to study the effect of an increase in education induced by a secondary school girls scholarship program. They find that secondary education made young women more politically informed, less deferential to political authority, and more likely to reject gender biased violence. They do not find, however, that secondary education increases support for democracy, political efficacy, or voting. A clear limitation of this paper is that it examines the effect of secondary education over a relatively short time-frame (4-5 years after initial enrollment), and on a small number of beneficiaries that are not representatives in terms of sex and age. Our research design instead examines the long-term effects of mass public education for a representative sample of the population. More so, given the comprehensiveness of Zimbabwe s education reform, we are able to identify education s downstream effects for a wide range of compliers that small scale field experiments cannot reach. Our paper speaks most directly to two recent working papers examining the long-term political effect of education. Wantchekon, Novta and Klašnja (2013) use placement of the first missionary schools in Benin as a plausible source of exogenous variation in access to education. They find that both the first generation of formally educated Beninois and their descendants are more likely to join political parties and to campaign for political parties. Wantchekon, Novta and Klašnja (2013) do not, however, report results regarding voting behavior and political attitudes, perhaps since their main focus is education s effect on well-being. Larreguy and Marshall (2014) exploit Nigeria s 1976 Universal Primary Education reform to show that increased educational attainment causes more political participation in the form of voting, contacting local officials, attending community meetings, and devoting attention to and gaining knowledge about political events. While the above papers arguably identify the impact of education on political participation, they all examine this relationship in the context of genuine political contestation. While Benin, Kenya, and Nigeria cannot be classified as consolidated democracies, they have experienced competitive elections and turnovers of executive power since adopting far-reaching liberalization reforms. 7 In Zimbabwe, by contrast, there has 6 See, for example, Finkel and Ernst (2005), Finkel and Smith (2011), Finkel, Horowitz and Rojo- Mendoza (2012), Kuenzi (2006), Kuenzi and Lambright (2005) and MacLean (2011) who find that education is positively associated with greater turnout using survey data from ten African countries. 7 Kenya and Benin have both had electoral transitions between the former ruling party and opposition 7

8 Polity 2 Score by Country Year Polity countryname Benin Nigeria Zimbabwe Figure 1: Polity-2 scores by year. The vertical line (at 1996) represents the first presidential elections turnout captured by the Afrobarometer, which run its first round of surveys in been no alternation in executive power since majority rule was achieved in 1980, and widespread election rigging took place in every election, at least since These differences are reflected, for example, in Zimbabwe s Polity-2 score, which is significantly lower than that of Benin and Nigeria throughout the period of elections covered by Afrobarometer (see Figure 1). Nevertheless, the level of contestation in Zimbabwe is not constant: the increase in the Polity score following the 2009 election reflects an important period of power-sharing where genuine change appeared possible. We discuss this in greater detail below as this change in electoral competition informs an important test of our disengagement argument. Notwithstanding the growing attention to the relationship between education and political participation in developing countries, much is still unknown, especially regarding the nature of this relationship in electoral authoritarian countries. Our paper offers the first causal evidence on the education-participation nexus in a repressive non-democratic setting that does not offer citizens genuine contestation. parties. In Nigeria, presidential power has alternated between northerners and southerners within the ruling PDP since the return to democracy in

9 Politics and secondary education in Zimbabwe Zimbabwe (then known as Rhodesia) was a British colony for much of the 20th century, with a small white settler elite, a large black African majority, and an apartheid-like set of institutions that ensured white dominance of political and economic life. In 1965, as the wave of independence swept through the African continent, the white settler-led government declared independence from Britain in order to prolong their domination of the country. Armed resistance to white rule began in the mid-1960s and intensified after 1972, finally resulting in free elections and black majority rule in 1980 (Bratton, 2014). Robert Mugabe, Zimbabwe s first post-independence head of state, still serves as Zimbabwe s president, and was most recently re-elected in Economic growth in the first years of majority rule was relatively strong, as Mugabe refrained from making radical economic changes, and donors made large commitments to support the new nation. As a consequence, health, education and other public services were dramatically expanded to the previously disenfranchised black population. Political conflict, however, emerged before long, based on divisions between the two main independence movements. Despite their temporary Patriotic Front alliance, Mugabe s Zimbabwe African National Union (ZANU) did not accept the legitimacy of the rival Zimbabwe African People s Union (ZAPU), led by Joshua Nkomo (Bratton, 2014). Mugabe dropped Nkomo from the cabinet in 1983 and soon after ZANU led a campaign of bloody repression, known as the Gukurahundi, against the ZAPU areas of Matebeleland, causing an estimated 20,000 deaths (Bourne, 2011). By 1987, an uneasy truce was brokered and the two parties were merged into ZANU-PF, and the constitution was changed from a parliamentary to a presidential system with a strong executive office, held by Mugabe. The merger left the new ZANU-PF party with 99 percent of parliamentary seats. Politics in the 1990s was thoroughly dominated by ZANU-PF, which grew increasingly repressive. In 1990, Mugabe won 78 percent of the presidential vote. The opposition largely boycotted the 1995 parliamentary and 1996 presidential elections, and ZANU-PF won 117 out of 120 seats (in 1995) and Mugabe won 93 percent of the presidential vote in (Levitsky and Way, 2010). Opposition to Mugabe s increasingly autocratic rule began to crystallize in the late 1990s, when labor unions, religious, and civil society groups mobilized around the goal of introducing a new constitution with term limits. These groups even- 9

10 tually formed Zimbabwe s key opposition party, the Movement for Democratic Change (MDC). Mugabe responded by introducing his own version of a new constitution, which was put to referendum and defeated in February Later that year, the MDC won an unprecedented 58 out of 120 seats in parliament. 8 The surprising outcome of the referendum coupled with the opposition s success in the 2000 parliamentary election mark a clear turning point. Mugabe responded by sharply increasing political repression, and by launching a campaign of land redistribution; white farmers were dispossessed via land invasions and farms were given to war veterans and ZANU-PF allies. Agricultural output collapsed, and macroeconomic stability, already in jeopardy through poor governance and overspending, was further eroded. Ahead of the 2002 parliamentary elections, ZANU-PF and state security organs organized a campaign of targeted beatings and harassment of opposition supporters. Restrictive laws requiring approval for meetings of more than five people and registration of journalists were passed and enforced. The regime created no go zones in rural areas in which opposition campaigns could not operate, conducted nighttime political reeducation campaigns, known as pungwes, and turned the national youth national service program into a violent militia known as the Green Bombers (Bratton, 2014). In the run-up to the 2002 election, Morgan Tsvangirai, the MDC presidential candidate, was charged and tried with treason, which undoubtedly helped Robert Mugabe to win the elections decisively. Political mobilization and protest against the regime continued after the 2002 presidential election, as well as the targeted repression of MDC supporters. ZANU-PF continued to dominate parliament following the 2005 elections in which it won 65 percent of parliamentary constituencies. This continued domination was partly achieved due to MDC s indecisive position regarding a possible boycott of the 2005 elections as well as due to the formation of a splinter party, known as MDC-M (Bratton, 2014). Immediately after the 2005 parliamentary elections, Mugabe launched Operation Murambatsvina ( Drive Out the Rubbish ), which was a campaign of destruction of informal urban settlements (often MDC strongholds) that displaced over 700,000 people. Shortly after, hyperinflation emerged reaching 1,281% in 2006, and 231,000,000% in 2008 (Bratton, 2014, 84). 8 At the time Zimbabwe s Parliament had 150 seats, out of which 30 were directly appointed by Mugabe, leaving ZANU-PF with a large working majority despite the close election result. However, as Bratton (2014) notes, this was the first time since independence that Zimbabwe had a meaningful opposition with enough parliamentary seats to block constitutional amendments. 10

11 The opposition was more unified in the lead up to the 2008 elections. Benefiting from what were initially somewhat lower levels of election-related violence than in 2002 (Levitsky and Way, 2010), the MDC won an outright parliamentary majority in the 2008 election, as well as the majority of municipal councils. MDC presidential candidate Morgan Tsvangvirai outpolled Mugabe in the first round of the presidential election, but the electoral commission stated that he did not clear the 50 percent threshold (the reported margin was 48% to 43%), forcing a second round. Prior to the second round, ZANU-PF and the military launched a brutal campaign of intimidation and beatings against MDC supporters and candidates, known as Operation Mavhotera Papi ( How Did You Vote? ) and Tsvangvirai withdrew from the race. Mugabe won the resulting sham election with a landslide, but international pressure forced a government of national unity, with Mugabe as president and Tsvangvirai as prime minister (LeBas, 2014). The formation of the national unity government and the end of hyperinflation allowed the economy to rebound somewhat after 2009, and political violence declined. With economic recovery, some basic public services were restored. Political outcomes, however, fell short of meeting the opposition expectations. Though the MDC chose the National Assembly speaker and gained control of several ministries, ZANU-PF was unwilling to share power in meaningful ways. In particular it would not loosen control of coercive institutions such as the police and military. The MDC was politically weakened by growing internal divisions, and by the loss of moral authority implied by sharing power with ZANU-PF. This has set the stage for the return of ZANU-PF dominance (LeBas, 2014). In the July 2013 elections, Mugabe won a large victory over Tsvangvirai, gaining 70% of the seats in parliament and putting an end to the power-sharing agreement. The 2013 election is reported to have been less violent than previous elections, although it may be that, given accumulated legacy of political violence, less overt repression was required to secure victory. The education reform of 1980 Prior to black majority rule, access to education for the black community was deliberately restricted. While schooling was compulsory and free for whites (until age 15), black Zimbabweans, who were not required to attend school, had to pay relatively high school fees. In addition, black Zimbabweans were required to pass a series of increasingly difficult exams in order to continue past primary school, while continuation to the first cycle of secondary school was automatic for whites. The education budget for black Zimbabweans was 11

12 tightly controlled at 2 percent of GDP, while out of pocket secondary school tuition and boarding fees cost almost two months of the average African wage 9. Available data suggests that the government spent about 12 times more per capita on primary schooling of whites (Dorsey, 1989). Planning documents from the period further reveal that the desired projection was for only 10% of black Zimbabweans to attend any secondary school, and for only 3% to complete secondary education (Nhundu, 1992). 10 King (2013) documents that such divide and rule policies were implemented in many African colonies, which had come to associate education with greater unrest. Starting in April 1980, the new post-colonial government implemented a wide-ranging set of educational reforms. Primary education was made free and compulsory for all Zimbabweans, regardless of color. While some fees were applied for secondary school, automatic progression from primary to secondary school was decreed. Furthermore, age barriers were removed for older children, allowing those who did not start school on time to attend. The government also undertook a large-scale school building campaign and reopened schools that had been closed during the independence war. The number of primary schools almost doubled (from 2,401 to 4,291) between 1980 and 1986, while the number of secondary schools increased dramatically, from 177 to 1,276 (Bourne, 2011). This massive infrastructure investment was financed by both a sharp increase in the budget share devoted to education and through communities voluntary in-kind labor contributions. Zimbabwe s education reform had an immediate effect: overall student enrollment literally doubled in one year (Narman, 2003). Key to our empirical strategy, this increase was principally driven by secondary enrollment, which rose from 66,215 students in 1979 to 537,427 in 1986 (Dorsey, 1989). The change is also apparent in the primary-secondary progression statistics: while in 1979 only 25% of primary school leavers continued to secondary schools, by 1986, 78% did (Figure 2). University enrollment also increased: while only 300 Africans were allowed to enroll per annum before 1980, the number of tertiary students reached almost 10,000 within a few years from independence (Chung, 2006). Based on the nature of Zimbabwe s education reform, we focus on the expansion of secondary, rather than primary education. For primary school students, the formal abolition of school fees was the main 9 Authors calculation based on 1979 school fee data and 1977 wage data from Riddell (1980) 10 See Atkinson (1972) for an illuminating account of Rhodesia s education system and the policies dictating the level of schooling Africans received. 12

13 Secondary Education Expansion 500 Enrollment Form 1 Secondary (total) 400 N. enrolled students (in 1,000) Figure 2: Number of enrolled students in Form 1 (first year of secondary education) and in all Secondary education grades by years ( ). Source: Government of Zimbabwe, Annual Report of the Secretary of Education,

14 constraint that was relaxed. The data at hand, however, do not suggest that the reform significantly affected primary education attainment. This is likely because 80% of black Zimbabweans already enrolled in primary school even under white rule. (There is also evidence that some primary schools continued charging informal fees (Nhundu, 1992).) Had the 1980 reforms significantly affected primary enrollment, we would to observe a sharp discontinuity in education for the cohorts of primary school starting age in 1980, i.e. those born in However, we find no evidence of a discontinuity around those birth years (see Figure 4 below). 11 By contrast, the education reform had a large, significant effect on secondary enrollment, which stood at about 6% pre For secondary school students, three key constraints were reduced: a) the exam system that was designed to restrict the number of blacks in secondary schools was removed, b) geographic access was improved, given the massive infrastructure investment, and c) financial barriers were reduced, given that many secondary students could continue living at home instead of paying high boarding fees to distant schools. 12 Similarly to Agüero and Bharadwaj (2014) and Grépin and Bharadwaj (2014), we therefore restrict our focus to secondary school access. 13 Analysts of Zimbabwe s education reforms note that given the rapid expansion, qualified teachers could not be hired quickly enough, instructional quality often suffered, and school construction lagged behind enrollment leading to overcrowded classrooms. 14 While some slippage in quality was perhaps inevitable, given the speed of the reform, quality clearly remained high enough to deliver substantial material benefits to its beneficiaries, as we demonstrate below. Notwithstanding the above challenges, the universal and 11 Importantly, there is also no discontinuity in education levels for primary-school aged cohorts for the cohorts in the DHS data; see, for example, Grépin and Bharadwaj (2014). 12 Riddell (1980, 46) estimates boarding fees to be twice as large as tuition fees for secondary schooling circa There is some increase in primary attainment for those at the secondary age threshold. This is due to the fact that expanded secondary access has induced a return to primary school, and due to the implementation of a program allowing teenagers to return to complete primary school on an abbreviated three year schedule (Chung, 2006, 228). 14 A construction lag could potentially violate our identification assumption if, for example the lag was correlated with unobserved features of areas that are also correlated with support of the regime. Agüero and Ramachandran (2014) who use a similar identification strategy for health outcome, have tested whether the education reform created a spatial-based rollout imbalance. They find that while some districts indeed opened schools sooner than others, by 1983 all disparities were eliminated. For example, comparing the educational attainment of those born in districts where secondary schools opened earlier to those born in districts that opened schools later shows no difference. Furthermore, Chung (2006, 272) provides a first hand account from within the Ministry of Education, arguing that Ministry technocrats were able to successfully resist pressure from politicians to build schools based on political considerations. 14

15 ambitious nature of its education reforms made Zimbabwe a widely recognized leader in expanded access to education in Africa during the 1980s and early 1990s (Dorsey, 1989). In sum, Zimbabwe went through two dramatic social and political processes in the past 30 years. On one hand, Zimbabwe has undergone an expansive education reform that reached its peak in the early 1990s. On the other hand, the country has undergone a political transformation that is characterized by an increasingly repressive political order, especially since the late 1990s. These transformations provide a unique setting to examine the role education plays in electoral authoritarian regimes that allow some restricted political contestation, but that are also characterized by social and economic vulnerability, opposition intimidation and political violence. Research design In this section we discuss the data sources, identification strategy and estimation approach that we use to identify the effects of the education reforms discussed above on individual-level political participation. Data To explore the long-term effects of education on political participation in Zimbabwe we combine into a single dataset all available rounds of the Afrobarometer surveys conducted in 1999, 2004, 2005, 2009, 2010, and Since the Afrobarometer drops questions and adds new questions across survey rounds, different survey years may be used to test different outcome measures. We focus exclusively on black respondents, who comprise the overwhelming majority of Zimbabwe s population and who were the education reform s target group The Afrobarometer initiative conducts national represenatative surveys on the attitudes of citizens in selected African countries towards various aspects of political and economic development. The Afrobarometer is a collaborative enterprise of Michigan State University, the Institute for Democracy in South Africa and the Ghanaian Centre for Democratic Development. 16 Unfortunately, there were not sufficient white voters to conduct a difference-in-difference analysis utilizing the fact that the education reform was specifically targeted at Zimbabwe s black population. However, Agüero and Ramachandran (2014) report that, using 2002 census data, there is no discontinuity in education for white Zimbabweans at the 1980 threshold. 15

16 Education Attainment Distribution count No schooling Incomplete primary Complete primary Incomplete high Complete high Incomplete college Complete college Figure 3: Sample Distribution of Education Attainment. N = 8,020. Education is this paper s key explanatory variable, and is measured using a seven-point ordinal scale with the following categories: no schooling, incomplete primary, complete primary, incomplete secondary, complete secondary, incomplete college, complete college. A one category increase in the education measure is equivalent to about 2-4 years of education given the discrete nature of the variable. Figure 3 shows the distribution of this education measure in our data, indicating that the modal level of schooling is incomplete or complete high school. Our study principally examines the causal effect of educational attainment on political participation. Political participation is operationalized using four binary indicators. Voted indicates whether the respondent reported voting in the most recent legislative or presidential election. In addition, we examine indicators for directly contacting one s local government councillor (Contacted local Councillor), attending a community meeting (Attend community meeting), or joining other community members in raising an issue (Raise issue at meeting) within the past 12 months. Respectively, 73%, 41%, 68% and 67% of respondents engaged in such activities. 17 We then combine these four variables, which are positively correlated with a Cronbach s 17 We focus on relatively undemanding forms of participation that relatively low-level education might reasonably affect. Contacting one s MP and attending a demonstration represent additional forms of partic- 16

17 Free schooling Figure 4: Operationalization of Secondary access, the study s key treatment variable. alpha of 0.67, into a summary index Participation scale. 18 The scale s relatively high inter-item correlation suggests that combining our binary measures captures a coherent latent dimension of political participation. Although we also present results for each component separately, we place greatest weight on our scale measure, which averages over the noise contained in the binary indicators. Variation in access to secondary schooling In order to identify the causal effect of educational attainment, we exploit cross-cohort variation in access to secondary schooling arising from Zimbabwe s 1980 education reform. Specifically, we compare black citizens from cohorts that were affected by the reform to cohorts that were too old to benefit from the educational expansion. Accordingly, we define those born in 1967 or later, who were 13 or younger when the reform got ipation. Consistent with their comparatively higher costs, both are rare in our sample: only 19% contacted their MP or attended a demonstration. Given the low levels of such activities, it is unlikely that median levels of education would induce change in such behavior. Unreported results confirm this, finding no no effect of education on either activity. 18 All summary (scaled) indices are constructed using the alpha command in Stata, which does not use casewise deletion and therefore maximizes the available information from the constituent variables. Thus a score is created for every observation for which there is a response to at least one item. The summative score is then divided by the number of items over which the sum is calculated. 17

18 implemented, as fully treated (treatment value equals to one). Those born in 1963 or earlier, and thus aged 17 or older in 1980, are defined as a control group not affected by the reform (treatment value of zero). Finally, black Zimbabweans aged between 14 and 16 at the time of the reform onset are considered partially treated. Such students (or more accurately, potential students) are coded according to the number of additional years of schooling available to them; for example, a black Zimbabwean aged 15 in 1980 is coded as receiving a dosage of half treatment, while (potential) students aged 14 and 16 are coded as receiving quarter and three-quarter dosages. 19 This coding scheme, shown graphically in Figure 4, defines our key exogenous variable, Secondary access. It is important to emphasize that access to public education does not necessarily entail enrollment. Rather, the mapping from educational access a treatment variable that is applied equally to all cohort members to a constituent s educational outcome is probabilistic rather than deterministic. This is because not all primary students continue to secondary school, but also because some older individuals whose education had been interrupted by the war returned to school in later years (Narman, 2003). Nevertheless, our measure of access identifies the reduced form effects of the reform under the key assumption that black Zimbabweans are, in expectation, identical with the exception that only some cohorts were eligible to enjoy access to secondary education. To ensure that we are comparing similar treated and untreated respondents, we only compare respondents from cohorts born just before the first cohort to be partially eligible to respondents from cohorts born just after the first cohort to be fully treated. Figure 5 shows trends in education across cohorts, providing cohort means in educational attainment. The mean education scale (in the top left) demonstrates clearly that cohorts fully treated by the reform (i.e. born in 1967 or later) exhibit substantially higher levels of education compared to cohorts born in 1963 or earlier. The increase is large and almost equivalent to moving from complete primary to incomplete secondary or from incomplete to complete secondary education. The figure also demonstrates that the reform increased the education levels of partially treated cohorts, but by less than fully treated cohorts. We now turn to examine the plausibility of our main identification assumption. Although there is a slight long-run uptick in schooling for cohorts born too early to be of secondary school age at the time of the reform, Figure 5 indicates that the pre-treatment trend in education is relatively flat once we focus on the five 19 This approach to partial treatment follows Bleakley (2010) and Johnson (2014) who employ step functions in similar settings. 18

19 Education Incomplete primary school Complete primary school Scale Proportion Proportion Incomplete secondary school Complete secondary school Incomplete college Proportion Proportion Proportion Figure 5: Trends in educational attainment Notes: Each grey dot represents average education for a given cohort (birth year). Large dots reflect larger samples sizes. Black line is a local polynomial fitted either side of the reform (indicated by the vertical dashed line). cohorts on both sides of the reform s cutoff point. 20 Below we also document relatively flat pre-treatment trends in our political outcomes for cohorts either side of the reform. Furthermore, Figure 6 plots pre-treatment characteristics and district-level variables, comparing cohorts on both sides of the reform. The results suggest that our sample is balanced across treatment groups with respect to gender, age at the date of the survey and district-level political aggregates. There is, however, a slight imbalance with respect to tribe: the first two plots point to a small difference in the proportions of Shona and Ndebele tribe members, Zimbabwe s two largest tribes. We note that the findings of this paper are robust to the inclusion of any and all pre-treatment covariates. Note also that an increase in Shona 20 In addition to the graphical representation, we conduct more formal checks that pre-trends are not driving our results. Specifically we use a placebo reform and polynomial trends as robustness checks. We describe these checks and their results below. 19

20 respondents the tribe most closely aligned to Mugabe likely underestimate any decline in participation or support for ZANU, since Shona respondents are less likely than other tribes to withdraw support from the regime. Proportion Shona Proportion Ndebele Proportion Male Proportion Age Proportion District incumbent vote share Proportion District turnout Figure 6: Trends in educational attainment Notes: Each grey dot represents average education for a given cohort (birth year). Large dots reflect larger samples sizes. Black line is a local polynomial fitted either side of the reform (indicated by the vertical dashed line). We conclude this section by acknowledging that many changes occurred in Zimbabwe around independence. Our identification strategy, however, relies on the fact that the education reform disproportionately affected cohorts that are close in age. Thus, while independence has undoubtedly brought about many other social, economic and political reforms, our underlying assumption is that these other changes did not affect children differentially around the age cutoffs defined above (see also Agüero and Bharadwaj 2014, 490). 20

21 Estimation strategy Building on the claim that access to secondary schooling is exogenous across cohorts, we utilize two main approaches to identifying the long-run political effects of secondary education. Our first approach includes partially treated respondents. A key advantage of this approach is that by using all cohorts and leveraging differences in treatment intensity, it is hard to see how common shocks associated with independence would differentially affect different cohorts receiving different levels of access to education. We first estimate the reduced form effects of increasing the availability of secondary education which is equivalent to an intent-to-treat (ITT) analysis using the following OLS regression: Y icdt = γsecondary access c + η t + ε icdt, (1) where Y icdt is an outcome measure, and Secondary access c our key treatment variable allows the effect of the education reform to vary across partially treated adolescents. We include survey fixed effects, η t, to account for time varying shocks that impact respondents differently across survey rounds. Throughout we cluster standard errors by district. 21 To maximize the comparability of treated and untreated respondents, we focus only on the last cohorts not to be affected by the reform and the first cohorts to be affected by the reform. In particular, we consider a bandwidth of five cohorts on either side of the reform cutoff years of birth (1963 and 1967). This allows us to better control for political and cultural shocks that may have lasting effects. Since socialization processes operate differently at different stages of life (Alwin and Krosnick, 1991; Sears and Valentino, 1997), substantially older and younger voters could vary in their response to such shocks, thereby violating our identification assumption. We show that our results are robust to a variety of specification tests. In order to identify the political and economic effects of actual schooling, as opposed to the availability of secondary education, we turn to an instrumental variables (IV) analysis. Our IV analysis identifies the (local average treatment) effect of schooling for black voters that received additional schooling because of the 1980 reform, but would not have done so otherwise (Angrist, Imbens and Rubin, 1996). In the first stage, 21 It turns out that this is more conservative than clustering by cohort. Our results are also robust to simultaneously double clustering by both district and cohort (see Cameron, Gelbach and Miller 2011). 21