2. Literature and theory

Modernization theory and the political socialization literature argue that education leads to political participation and democratization: as citizens become more educated, they expand their interests, better understand politics, and develop democratic values (Lipset, Reference Lipset1959; Deutsch, Reference Deutsch1961). Recent research emphasizing causality (Milligan et al., Reference Milligan, Moretti and Oreopoulos2004; Dee, Reference Dee2004; Sondheimer and Green, Reference Sondheimer and Green2010) indeed find a positive effect of education on civic attitudes, political interest, and participation—including turnout.

These findings, however, do not comport with historical trends (Putnam, Reference Putnam1995): while education in the United States has increased steadily over time, we have not witnessed a similar trend in voter turnout. To reconcile this inconsistency, scholars have pointed to omissions in research design: after adjustments to causal identification, the effect of education on political participation disappears or significantly weakens (Kam and Palmer, Reference Kam and Palmer2008; Henderson and Chatfield, Reference Henderson and Chatfield2011). Most relevantly, Marshall (Reference Marshall2019) argues that the results in Milligan et al. (Reference Milligan, Moretti and Oreopoulos2004) and Dee (Reference Dee2004) have not adequately isolated the causal effect of education, and that there is no effect of education on turnout in the United States upon correction. Rather than participation, education primarily plays the role of shaping democratic values (Marshall, Reference Marshall2016, Reference Marshall2019) and interest in politics (Siedler, Reference Siedler2010). In a wider sample of 15 European countries, Borgonovi et al. (Reference Borgonovi, d'Hombres and Hoskins2010) finds that the relationship between education and turnout disappears once they instrument education with compulsory schooling. Articles using alternative identification strategies generally also find no significant effect of education on voter turnout in the United States, Sweden, and Norway—either in the short or long run (Tenn, Reference Tenn2007; Berinsky and Lenz, Reference Berinsky and Lenz2011; Pelkonen, Reference Pelkonen2012; Ahlskog, Reference Ahlskog2021).Footnote ¹

Yet, there is strong evidence that education positively affects political participation through formal channels in developing contexts. Larreguy and Marshall (Reference Larreguy and Marshall2017) find that education increased political participation in Nigeria: more educated citizens are more likely to be interested in and informed about politics, to contact local government officials, to participate in community associations and meetings, and to vote in elections. Formal education in Mali (Bleck, Reference Bleck2015) and nonformal education in Senegal (Kuenzi, Reference Kuenzi2006) have also increased political knowledge and participation. The relationship between and education and participation also extends across generations: Wantchekon et al. (Reference Wantchekon, Klašnja and Novta2015) demonstrate that educated individuals in colonial Benin, as well as their descendants in post-colonial Benin today, are more likely to be politically active.

What explains these different results between Western and developing contexts? We argue that the strength of a country's democratic institutions might play an important role in dictating the extent to which educated citizens choose to participate in politics. When deciding whether to participate, individuals weigh the clear costs against the more uncertain benefits of political participation (Riker and Ordeshook, Reference Riker and Ordeshook1968). While individuals may vote or engage in other forms of political participation for expressive reasons (Schuessler, Reference Schuessler2000), this is particularly unlikely in developing contexts—and even more so for more educated individuals—since political parties often have indistinguishable political platforms (Manning, Reference Manning2005).

Thus, political participation when democratic institutions are established and well-functioning imposes greater cost than benefit. However, as the expectation of democratic improvement increases, so do the benefits of political participation. Unlike results from Western democracies, findings from developing contexts primarily consider countries and time periods where democratic institutions are fragile. In these cases, political participation provides an avenue for change for citizens who seek to bolster democracy and combat poor governance. This argument accords with the findings from (Croke et al., Reference Croke, Grossman, Larreguy and Marshall2016), which investigates formal political participation in an authoritarian context and finds a reduction in political participation among more educated citizens, suggesting disengagement where there is no avenue for change.Footnote ²

A key mechanism linking education to greater political participation in developing contexts is that the benefits of political participation are more identifiable to individuals who are more interested in, and knowledgeable of, political affairs. As citizens receive more education, they are more likely to critically evaluate government performance and make determinations about whether political participation may lead to improvements in governance and democratic quality (Lipset, Reference Lipset1959; Deutsch, Reference Deutsch1961). Educated citizens therefore recognize that the returns to participation depend significantly on the strength of democratic institutions. While participation in autocracies is unlikely to lead to change and is likely to validate the regime, the basic democratic institutions that characterize emerging democracies, such as relatively free and fair elections, can be a venue for citizens to oust poorly-performing, entrenched incumbents. Ultimately, when democratic institutions are accountable and democracy is functioning well, political participation is unlikely to have much impact. This is especially so when there is limited policy differentiation across political alternatives, as in the context of Africa (Manning, Reference Manning2005). Thus, the effect of education on political participation should vary with the strength of democratic institutions in a country.

We examine Senegal, a stable but developing democracy that has scored highly in terms of the strength of its democratic institutions. Senegal is an ideal case to test our argument: despite being one of Africa's oldest and strongest democracies, the strength of its democratic institutions was tested when presidential over-reach threatened its democratic institutions in the mid-2000s. This internal change in Senegal's democratic quality allows us to test for the differential effect of increased education on political participation as a function of democratic institutional strength, and the mechanisms through which increased education affects political participation.

Specifically, if education affects political participation through increased political awareness, we should expect no differential effect of education on participation during periods of strong democratic institutions, but a positive differential effect during periods of when these institutions are threatened. On the contrary, if education affects democracy through other mechanisms such as greater democratic values or connectedness to national political life (Almond and Verba, Reference Almond and Verba1963; Evans and Rose, Reference Evans and Rose2007), expressive voting (Schuessler, Reference Schuessler2000), or bureaucratic language proficiency (Bleck, Reference Bleck2015), education should increase political participation regardless of whether democratic institutions are being challenged.

We formulate four testable hypotheses. First, we test the core channel through which our argument runs: that formal education in Senegal does indeed increase an individual's understanding of, and interest in, politics.

Second, when democratic institutions are strong and there is little policy differentiation between political parties—as in the context of many African countries such as Senegal—educated citizens should not be more likely to participate in politics.

However, education should increase participation if democratic institutions are threatened. Educated citizens—because they have greater knowledge and political awareness—should be more willing to participate in politics since the benefits of participation increase during these periods.

Finally, more educated citizens in stable but developing democracies should demand constraints on aspects of their government that most threaten democracy when these threats emerge. In the case of Senegal, more educated citizens should be more aware of, and critical of, the hyperpresidentialism in the country (Beck, Reference Beck2012). They should therefore exhibit greater support for checks on presidential power when democratic institutions are weak.

3. Democracy in Senegal

3.3 Education and perceptions of threats to democracy

The president's political manipulation of the law was noticed earlier by individuals with higher education. While Senegalese democracy faced generally little scrutiny during Wade's initial years as president,Footnote ³ enthusiasm quickly began “ebbing away, first among Senegalese intellectuals and then among foreign analysts” Mbow (Reference Mbow2008, 157).

We emphasize the difference in interpreting political events between those with greater educational attainment, and those with less. Specifically, education led people to care more about democracy and discern threats to democracy more quickly. Accordingly, we examine how participants with and without secondary school education evaluated democracy in Senegal between 2002 and 2014. We use Afrobarometer surveys from 2002 to 2014 provide a suggestive correlation between education and the ability to perceive declines in Senegal's democratic quality. Based on the sequence of political events, threats to democracy began in 2003 and continued until the 2012 elections. Due to the highly publicized elections in 2007, it became clearer to everyone by the 2008 survey that democratic institutions had eroded in Senegal. However, before that, threats to Senegal's democracy should have primarily been noticed by the educated citizenry.

First, in Figure 1, survey responses validate qualitative accounts that democracy was threatened: citizen views of Senegalese democracy sharply declined between 2002 and 2008 across the board. However, educated citizens became critical much sooner, and held opinions that were more in line with ongoing Senegalese politics by 2005. Respondents were asked whether they were satisfied with Senegal's democracy and whether they believed that Senegal is a functioning democracy. While more than 50% of the population answered affirmatively, this is largely driven by those with less education. Among those who had attended some secondary school, opinions already dropped substantially in 2005—when threats to democracy had just begun. On the other hand, respondents with primary education or lower reported roughly similar levels of satisfaction between 2002 and 2005, and were even more positive that Senegal was a functioning democracy in 2005. By 2008, all citizens reported lower satisfaction; but, satisfaction was lower for those who had attended secondary school.

Figure 1. Citizen Views on Senegalese Democracy. Note: How democratic is Senegal is coded 1 if respondents feel that Senegal is “A democracy, but with minor problems” or “A full democracy”, and 0 if the respondent believes that Senegal has “major problems” or is “not a democracy.” Satisfaction with Senegal's democracy is coded 1 if respondents feel “fairly satisfied” or “very satisfied” with Senegal's democracy, and 0 if the respondent feels “not very satisfied” or “not at all satisfied.”.

A similar pattern emerges in citizens’ views on Senegalese electoral institutions (Fig. 2). In 2005, respondents were asked to reflect upon whether the 2000 elections—when an opposition party won for the first time in the history of Senegal—were free and fair. All respondents converge in their positive opinions. In 2008 however, the views of more educated respondents drop far more than those of less educated respondents, suggesting a greater awareness about the election problems that took place during the 2007 elections. Education again seems to be correlated with lower trust in the Electoral Commission, and particularly so in 2005 than in 2008. Altogether, these descriptive data lend support to our argument that education leads to higher levels of political awareness and knowledge.

Figure 2. Citizen Views on Elections. Note: Elections are free and fair is not asked in 2003. The variable is coded 1 if the respondent agrees that the last national election is “completely free and fair” or has only “minor problems,” and 0 if the respondent felt that there were “major problems” or that elections were “not free and fair.” In the 2005 survey, the previous election indicated is the 2000 election. In the 2008 survey, the previous election is 2007, while in the 2013 and 2014 surveys, the previous election is 2012. Trust electoral commission is coded as 1 if the respondent trusts the EC “somewhat” or “a lot,” and 0 if they trust the EC “just a little” or “not at all.”.

Notably, Figures 1 and 2 show convergence in positive opinions among the Afrobarometer respondents by 2013, reflecting events from the 2012 election and the power transition from Wade to Sall. In 2012, Wade aimed to run for a third term, claiming that the two-term limit law shouldn't apply to his first term as it was established after the 2000 election. The loyalist-packed Constitutional Court permitted his candidacy. During this period however, important pop culture figures in the Y'en a Marre movement helped to mobilize broad opposition and widespread pressures for change expanded significantly. In a strong demonstration of democratic strength, Wade lost the 2012 election by a landslide as Senegalese citizens voted against legitimizing a third-term presidential campaign. Correspondingly, views about democracy and democratic institutions increased sharply in 2013 (Figs. 1 and 2). Citizens with different levels of education converged again in their views about the strength of the Senegalese democracy.

4. Background on educational access in Senegal

Until the early 2000s, education in Senegal lagged far behind other countries with respect to primary school enrollment and gender parity (UNESCO, 1995, 2015). Post-primary education was worse: middle school capacity was only around 25%, meaning that a large percentage of those who completed primary school could not attend middle school. In 2000, Dakar hosted the World Education Forum, where 73 participating countries pledged to uphold the goals of UNESCO's Education for All movement—to provide and mandate education, improve literacy rates, and correct gender imbalances in school enrollment. The Senegalese government then released a multi-phase plan in September 2000 to improve basic education (primary and middle school) for all children. The education reform curriculum emphasized improving reading skills within a broader social sciences and math curriculum (Clasby, Reference Clasby2012; USAID, 2017), which is particularly important for increasing political interest and awareness (Hillygus, Reference Hillygus2005).

In 2001, Senegal took its first steps toward meeting these goals by adopting a new Constitution that pledged to significantly expand educational infrastructure. A large-scale school construction program quickly followed in mid-2001; by 2011, 666 new middle schools had been built.Footnote ⁴ Second, the government introduced free and mandatory middle schooling for all of its citizens in 2004. The positive benefits of these large-scale changes are reflected in Senegal's education statistics, where the greatest gains were seen at the middle school level. Between 2000 and 2007, middle school enrollment rates increased by 95%, while secondary school enrollment increased by 78% and primary school enrollment increased by 25% (DeStefano et al., Reference DeStefano, Lynd and Thornton2009). Transition from primary school completion to lower secondary school increased from 35% in 1999 to around 88% by 2011; by 2012, secondary school attendance had increased to 41%. (UNESCO, 2015).

5. Research design

To identify the effects of increased education on political participation in Senegal, we use a DiD identification strategy. This strategy exploits differences in the potential impact of the Senegalese school reform that started in the early 2000s to capture plausibly exogenous variation in individual school attainment. We use this identification strategy rather than simply using survey respondents’ education measure as it allows us to account for confounding unobservable variables. Education is endogenous to a variety of factors, including socio-economic factors, individuals’ taste for learning (both in school and outside), and political factors that inhibit both education and political access (Borgonovi et al., Reference Borgonovi, d'Hombres and Hoskins2010). These omitted variables may significantly bias estimates (see Appendix A.10).Footnote ⁵

The DiD strategy relies on the impact of the reform varying along two dimensions: first, the age of the individual at the time of the reform (individuals already past middle school age were unaffected) and second, where the individual lives (some areas had far fewer middle schools than others prior to the reform, and these areas therefore had lower baseline levels of education). The variation along these two dimensions allows us to estimate educational attainment increases for individuals who were young enough to be affected by the reform, and compare these effects between areas with previously high education levels and areas with previously low education levels.

We find that our DiD strategy effectively captures plausible exogenous variation in educational attainment (Appendix A.7). We then use the DiD strategy to estimate the effects of the reform on political participation and attitudes. We present the average effects over the entire time period covered by the Afrobarometer surveys, and then introduce heterogeneous effects to compare effects between the period when democracy is threatened and periods of democratic strength.Footnote ⁶

Our DiD strategy captures the effect of the reform on individual educational attainment by comparing differences in education of cohorts young enough and cohorts too old to benefit from the reform, across areas with varying pre-reform middle-school completion rates. First, since middle school begins when children are 13 years old and comprises four years of schooling, we compare educational attainment of those who were 16 or younger by the year 2002—and thus were just young enough to benefit from the reforms—with those who are over 16 years of age and thus were not affected. Second, different areas in Senegal should have been affected differently by middle school construction depending on how educated the areas already were before the education reforms. Specifically, areas with high levels of schooling prior to the reform should have seen relatively smaller increases in education compared to previously under-served areas.Footnote ⁷

We define the intensity of the reform as the potential effect of reforms based on prior education levels. We calculate intensity across different areas by computing the middle-school attainment of the individuals that were too old to have benefitted from the education reform. We use the communauté rurale (CR), the lowest administrative unit above individual villages, as our unit of area. We then find the area's average educational attainment rates in the 2002 Senegalese census for individuals between the ages of 17–26 (i.e., the ten cohorts immediately prior to the first cohort exposed to the reform) to determine the pre-reform middle school attainment within a CR. We further break this down by gender because females tend to have lower education levels than males in Senegal. Reform intensity for gender g in CR l is:

$${\text intensity}_{l, g} = 1-{{\text total} \ g \ {\text of \ ages \ 17-26 \ who \ completed\ middle\ school\ in }\ l\over {\text total} \ g\ {\text of\ ages\ 17-26\ in\ }l}$$

We treat those who were 16 or younger when the education reforms came into effect as affected by the reform. However, we do not expect 16-year-olds in 2002 to have the same level of exposure as 13-year-olds, as the latter had more years to benefit. In order to capture the fact that individuals who were of middle school age were only partially affected by the reforms, we define our exposure variable, post, as follows. We assign a post value of 1 to those who were ages 13 or younger and were thus fully treated. We assign a value of 0.75–14-year-olds, a value of 0.5–15-year-olds, a value of 0.25–16-year-olds, and a value of 0 to those 17 years old and older.Footnote ⁸

5.1 Estimation strategy

We estimate the effects of education on political participation and attitudes using Senegal's Afrobarometer surveys (r2–6). We matched Afrobarometer respondents to school reform intensities based on their CR of residence and gender. We restrict our sample to those born after 1970 for comparability. Our DiD specification is:

(1)$$y_{i, g, c, l, t} = \tau \left(\,post_t \times intensity_{g, l}\right) + X_{i, g, c, l, t} \gamma + \kappa_c + \eta_{g, l} + \zeta_t + \epsilon_{i, g, c, l, t},\; $$

where y _i,g,c,l,t is an outcome for individual i of gender g in cohort c, CR location l, and survey year t. We further include fixed effects for survey year, ζ _t; cohort, κ _c; and CR by gender, η _g,l. In this equation, post _t and intensity _g,l are omitted since they are colinear with ζ _t and η _g,l, respectively. Finally, to increase power, we also include ethnicity and urbanity fixed effects as control variables, X _i,c,l,t. All regressions include CR-level clustered standard errors.

While we mostly focus on estimates of the reform on political outcomes, we compute school—a three-point scale of educational attainment for each individual respondent in the Afrobarometer surveys—to validate that the education reform effectively induced a plausibly exogenous variation in educational attainment. We code respondents as “0” if they did not receive any formal schooling, “1” if they had some primary school education, and “2” if they completed primary school or above.Footnote ⁹ In the appendix, Figure A1 and Table A10 confirm that increased access to middle schools led to an overall increase in schooling from the primary school-level and above.Footnote ¹⁰ We provide all IV estimates as robustness checks in Appendix Section A.12.

To capture how the effects vary when democratic institutions are threatened, our heterogeneous-effects equation estimates:

(2)$$\eqalign{y_{i, g, c, l, t} & = \tau_1 \left({\text \,post}_i \times\ {\text intensity}_{i, g, c, l, t}\right) + \tau_2 \left({\text \,post}_t \times\ {\text intensity}_{i, g, c, l, t} \times\ {\text threatened}_t \right)\cr & \quad + X_{i, g, c, l, t} \gamma + \kappa_c + \eta_{g, l} + \zeta_t + \epsilon_{i, g, c, l, t},\; }$$

where threatened _t captures the period under which presidential over-reach threatened democracy in Senegal—which overlaps with the 2005 and 2008 Afrobarometer surveys. As with Equation (1), lower-order terms are omitted since they are colinear with the fixed effects. Throughout our analysis, we standardize post × intensity to allow for easier interpretation of the coefficients.

We first confirm that we observe parallel trends for our main outcome of interest, school, in Appendix Figure A1. We further demonstrate the validity of our estimates by running the same regression on two placebo cutoffs—1969 and 1974—and produce parallel trend plots for our main outcomes. Appendix Tables A19 and A20 show no overall treatment effect for placebo birth-year cutoffs prior to school reform year, and Appendix Figure A2 shows parallel pre-treatment trends for our main outcomes. Second, we show that the extent to which an individual was exposed to the reform is orthogonal to the likelihood that she is surveyed in the Afrobarometer surveys. In Table A8, we show balance on a variety of respondent attributes determined prior to the treatment.