Democratic and autocratic imprinting and its long-term effects

We assume that all political regimes attempt to influence citizens’ thinking to maintain political order (Easton Reference Easton1965; Paglayan Reference Paglayan2022). We propose that education is a crucial channel of imprinting, as it is an important policy area controlled by the state, through setting national curricula and regulating teacher recruitment. Teaching pro-regime ideas and prescribing supportive curricula in schools and universities are common features of both autocracies and democracies (Lott Reference Lott1990, Reference Lott1999; Eibl-Eibesfeldt and Salter Reference Eibl-Eibesfeldt and Salter1998).Footnote ¹

It is almost impossible for ordinary citizens to escape exposure to education while growing up. Most states have compulsory education, meaning all students are exposed to state-controlled socialization efforts, unlike other forms of socialization via the media or the arts, which can be circumvented. Schools shape citizens’ beliefs from an early age. Social psychology indicates that the most intense imprinting occurs when individuals are young (Alwin and Krosnick Reference Alwin and Krosnick1991; Bartels and Jackman Reference Bartels and Jackman2014). During early childhood and adolescence, citizens are easily influenced by these regimes’ imprinting attempts, which have significant consequences for the crystallization of political values and attitudes in later life. Both democratic and autocratic imprinting are particularly strong when people live in a democracy or an autocracy during their impressionable years, which coincide with their school years.

Imprinting through education can hence be perceived as an instrument ‘to promote long-term social order by indoctrinating young children to accept the status quo, behave as “good citizens,” and respect the state and its laws’ (Paglayan Reference Paglayan2022: 11). However, democracies and autocracies define ‘good citizens’ differently, impacting school curricula and instructional methods. Democracies emphasize civic competence and internalization of democratic norms like tolerance and pluralism (Finkel and Smith Reference Finkel and Smith2011). The ‘good citizen’ in democracies learns the importance of political participation, such as viewing voting as a civic duty (Westheimer and Kahne Reference Westheimer and Kahne2004). In contrast, authoritarian education promotes obedience, loyalty to the regime, and uncritical thinking (Mirshak Reference Mirshak2020), often using rote memorization (Kipnis Reference Kipnis2011).

Given the goal of building citizen support and obedience, it is unsurprising that education is a key policy area for many autocracies. Paglayan (Reference Paglayan2021) documents that non-democracies invest in mass schooling to increase human capital and indoctrinate citizens. Autocracies validate their rule through political ideologies, nationalistic slogans, ethnic exclusions, or religious ideas (Eibl-Eibesfeldt and Salter Reference Eibl-Eibesfeldt and Salter1998; Paglayan Reference Paglayan2022; Neundorf, Nazrullaeva, Ksenia Northmore-Ball et al. Reference Neundorf, Nazrullaeva, Northmore-Ball, Tertytchnaya and Kim2024). These regimes provide a narrative to legitimize the incumbent (Gerschewski Reference Gerschewski2018, Reference Gerschewski2013) and portray them as competent (Guriev and Treisman Reference Guriev and Treisman2020).

We argue that these indoctrination efforts by autocracies result in a persistent mental imprint, even after the regime collapses. The norms and values formed during authoritarian rule become reference points for political attitudes and behavior in later life. Diwan and Vartanova (Reference Diwan and Vartanova2020) show that education in an autocracy negatively affects democratic behavior (e.g., voting and civic engagement), but their aggregate data do not identify the mechanisms driving this relationship, especially as they rely on coarse educational categories.

Research on the impact of education content is limited. Cantoni, Chen, Yang et al. (Reference Cantoni, Chen, Yang, Yuchtman and Jane Zhang2017) demonstrate that new pro-regime content in China’s school curriculum led to skepticism about free markets, increased trust in government officials, and alignment with the regime’s views on political participation and democracy. However, Kao (Reference Kao2021) finds that this effect is significant only among students whose parents worked for the government, raising doubts about the overall effectiveness of authoritarian education.

Regarding democratic education, evidence shows that civic education in schools can instil support for democracy and promote democratic knowledge in both new (Finkel and Ernst Reference Finkel and Ernst2005; Tobin Reference Tobin2010; Fesnic Reference Fesnic2016) and established democracies (Evans and Rose Reference Evans and Rose2012; Campbell and Niemi Reference Campbell and Niemi2016). Different approaches to civics instruction mediate the effect (Galston Reference Galston2004; Finkel and Ernst Reference Finkel and Ernst2005). Democratic education is most effective among disadvantaged children lacking parental socialization, with a ceiling effect among children from more educated and higher socio-economic status families, suggesting alternative channels of political socialization (Campbell and Niemi Reference Campbell and Niemi2016; Neundorf, Niemi, and Smets Reference Neundorf, Niemi and Smets2016).

Existing research highlights two gaps: the long-term effects of democratic versus autocratic education have not been directly compared, and the impact of authoritarian education on political attitudes and behavior remains understudied and contradictory. This paper addresses these gaps, aiming to uncover the mechanism between education (democratic and autocratic) and its long-term effects. We focus on two dimensions: education content and regime efforts to promote specific political education content. We will introduce the concept and measures of these dimensions below.

We formulate three theoretical expectations and raise one open question. First, we expect a long-term effect of authoritarian education on support for democracy, with more authoritarian education during impressionable years resulting in lower pro-democratic values later in life. Second, we expect democratic education to leave a lasting pro-democratic mark on citizens. Third, we anticipate that the impact of autocratic and democratic education depends on the curriculum content and the regime’s efforts to promote these values. The open question concerns the relationship between these effects: to what extent can the observable gap in democratic support between those who experienced authoritarian versus democratic education be attributed to the dampening effect of autocratic education or the boosting effect of democratic education?

We begin by establishing that the documented gap in support for democracy between those socialized in autocracies and democracies (e.g., Neundorf Reference Neundorf2010; Pop-Eleches and Tucker Reference Pop-Eleches and Tucker2017) is causally linked to education. Using the German Democratic Republic (GDR) as a case study, we test our first two theoretical expectations. We then broaden our perspective with a global analysis of democratic and autocratic education and its long-term effects, aiming to generalize our findings beyond the GDR case and address the open question.

The legacy of socialist education in East and West Germany

Focusing on the case of the German Democratic Republic (1949–1990), we first establish the causal impact of authoritarian versus democratic schooling on support for democracy. To estimate this effect, we leverage the fall of the Berlin Wall as an external shock and the birth-month cut-off rule for the school start year, examining how an extra year of exposure to socialist (rather than democratic) education impacts individuals’ support for democracy today. This empirical setup has been previously employed to study the effect of socialist education on labor market outcomes (Fuchs-Schündeln and Masella Reference Fuchs-Schündeln and Masella2016) and trust in the constitutional court (Cheruvu Reference Cheruvu2023).

The East German education system was designed to indoctrinate pupils with socialist ideology, encouraging activities aligned with the regime’s beliefs (Westheimer and Kahne Reference Westheimer and Kahne2004; Fulbrook Reference Fulbrook2014). The stated goals of civics education in the GDR were to convey knowledge about the ‘theoretical foundation of the Party of the working class, the positioning of dialectic and historical materialism in Marxism-Leninism, the program of the Socialist Unity Party of Germany (SED) as the major document that provides information about the ways and the goals of further societal development in the German Democratic Republic’ (Ministerium für Volksbildung 1980: 5; own translation). As such, it is fair to conclude that the aim of civics education in the GDR was ‘an indoctrination of Communist values, masqueraded as “state citizens education” and with an overall objective that can only be described as the very opposite of the independent and critical citizen’ (Buck and Geissel Reference Buck and Geissel2009: 233).

Until its abrupt ending in 1989/1990, the education system remained remarkably constant and was under tight control by the Ministry of People’s Education.Footnote ² Grammes, Henning, and Hans-Joachim (Reference Grammes, Schluss and Vogler2006) describe, for example, the various control mechanisms by the state that encompassed classroom visits, inspections, surveys, and interrogations of school principals, teachers, and pupils, as well as performance tests. These measures were implemented in so-called ‘operative Einsätze’ (operative missions) for which both the Central Committee of the SED and the Ministry of People’s Education were responsible.

Until 1989, the profession of schoolteachers was heavily regulated by the state. Being overly critical of the GDR or applying for permission to leave the country (‘Ausreiseantrag’) led to severe sanctions, including occupational bans, making teachers a particularly vulnerable group. Moreover, ‘being a teacher in the GDR meant not just having a function at school, it also meant compulsory visits to parents and drawing up reports on these visits’ (Den Hertog and De Jong Reference Den Hertog, De Jong, De Jong, Lalenis and Mamadouh2000: 125). Teachers were obliged to enforce participation in state-led youth organizations (Free German Youth, FDJ). For higher positions, active membership in the regime party SED ‘was as good as indispensable’ (ibid.).

There is anecdotal evidence that, particularly at the end of the GDR, some teachers and pupils became more critical against the official doctrine and looked for ways to circumvent the official education policies (Bathke et al. Reference Bathke, Chalupsky and Hoffmann1995). Yet, this effect should be taken with a grain of salt and should not be overestimated. In qualitative interviews, even critical teachers put forward that they did not dare to openly criticize the GDR and promote democratic values, but rather that they tried to avoid glorifications of past GDR achievements (Farouk and Camia Reference Farouk and Camia2022: 52–59).Footnote ³

While education systems are usually policy areas that are very hard and slow to reform (Ansell and Lindval Reference Ansell and Lindvall2013), the transformation of the GDR education system is described in the relevant literature as ‘sweeping’ with political education as being the ‘most radical’ area (Bruen Reference Bruen2014: 317). It soon became clear that a fundamental and rapid transformation of the education sector was of critical importance for the German unification process. As such, changes were of high political priority and were funded with large sums (Pritchard Reference Pritchard1998).

If we have a closer look at the timeline of events after the fall of the Berlin Wall, this becomes evident. Already in October 1989, Margot Honecker stepped down as Minister for People’s Education. At the end of October 1989, school curricula for civics education (‘Staatsbürgerkunde’) were declared no longer binding for teachers. From November 1989 onwards, teachers were themselves responsible for the content of their teaching. During the first half of 1990, East Germany saw a massive introduction of new textbooks and school materials. In 1990, 26.5 million new textbooks were printed by the Volk und Wissen publishing house, while a further 2.46 million textbooks were supplied from West Germany (Fuchs Reference Fuchs1997: 95–96). In March 1990, a new binding framework plan for civics education was introduced (Fuchs Reference Fuchs1997: 95–99). The reformed curriculum aimed at fostering democratic competences and values in line with the free democratic basic order (‘freiheitliche-demokratische Grundordnung’, fdGO), students’ ability for critical reflection and engagement with societal topics, as well as coming to terms with the GDR past, including its socialist ideology (Ministerium für Bildung und Wissenschaft 1990).

Given these sharp qualitative differences in educational experiences, which dramatically changed within a short time, and in line with our theory, we expect that respondents with a longer time spent under the socialist education system will have lower support for democracy compared to their peers. The sweeping changes in educational practice and policy, we described above, allow us to leverage the German reunification and the strict birthdate cut-off used for school start dates in the GDR to isolate the causal effect of socialist education on support for democracy.

We employ a difference-in-difference-in-difference (DiDiD) research design that leverages the school enrollment cut-off date to ensure causal identification.Footnote ⁴ In this DiDiD setup, we compare three groups: (1) respondents born before June 1st, (2) cohorts still in school during reunification (born between 1973 and 1982), and (3) East and West Germans.

Firstly, in the GDR, children born before June 1st started school the same year they turned six years old, while those born on or after June 1st started school the following year. We create a binary variable (born before June) with a value of 1 for individuals born between January and May and 0 for those born between June and December. Secondly, cohorts of students born between 1973 and 1982 potentially experienced between 1 and 12 years of socialist education, with the difference spent under democratic education. Within these cohorts (born between 1973 and 1982), each respondent potentially experienced one additional year of socialist schooling if they were born before June 1st (compared to their peers who were born after June 1st), as they were all still in school after reunification in 1990. Cohorts born before 1973 had completed their education by the time the Berlin Wall fell, and those born after 1982 only experienced the reformed education system (and are not included in the analysis).Footnote ⁵ We generate a binary variable (Cohort) that takes a value of 1 for individuals born between 1973 and 1982 and 0 otherwise. Finally, we use information on whether individuals spent most of their youth in East or West Germany to capture their exposure to socialist schooling, creating a dummy variable (East) with a value of 1 for those from East Germany and 0 for those from West Germany.Footnote ⁶

The identification of the causal effect of socialist schooling relies on two main assumptions. First, we assume that a child’s month of birth is random and thus determines the year they started school without parental manipulation. Given the rigorous implementation of GDR school enrollment policies, we find no evidence that parents could influence the year their children started school.Footnote ⁷ Second, we assume (based on the qualitative evidence provided above) that the transition of school systems and educational practices in East Germany was swift, beginning with the 1990/91 school year. As discussed above, schools quickly adopted West German textbooks and introduced new curricula. Even if some teachers did not implement these changes promptly, this should only make our estimates more conservative.

Our analysis uses data from the biannual German General Social Survey (ALLBUS), which includes measures of democratic support for surveys conducted in 2008 and 2018.Footnote ⁸ The ALLBUS began incorporating respondents from the former GDR in 1991, right after reunification, providing a comprehensive dataset of individuals who were born in Germany after 1943 – as this is the first cohort to have started school after the German division in 1949Footnote ⁹ – and completed Germany’s tripartite secondary school system. Our dependent variable comes from the ALLBUS survey data that measures respondents’ support for the idea of democracy, on a scale from 1 (very in favor) to 6 (very opposed). We invert this variable so that higher values mean more support for democracy and lower values weaker support. We estimate the following OLS model to analyze the data:

(1) $$\hskip-21pt{Y_{ics}} = {\it {\beta _0} + {\beta _1}Ag{e_i} + {\beta _2}Eas{t_i} + {\beta _3}Cohor{t_i} + {\beta _4}BeforeJun{e_i} + {\beta _5}Eas{t_i} \times Cohor{t_i} + {\beta _6}Eas{t_i}} \\ \hskip16pt\times \it{BeforeJun{e_i} + {\beta _7}Cohor{t_i} \times BeforeJun{e_i} + {\beta _8}Eas{t_i} \times Cohor{t_i} \times BeforeJun{e_i}{ + _l} + {\rm{ }}{\gamma _s}}$$

where Y is the predicted level of support for democracy reported by a respondent. The β ₈ coefficient of the triple interaction between East, Cohort, and BeforeJune captures the effect of an extra year of exposure to socialist education on an individual’s support for democracy within their own birth-year cohort. More simply, this captures the within-treated cohort (born between 1973 and 1982) variation where those in the treatment condition (born before the 1^st of June) have experienced one additional year of socialist schooling compared to those in the control condition (born on or after the 1^st of June).Footnote ¹⁰ Our theoretical argument receives support if β ₈ is negative and statistically significant. It would show that within each treated cohort, those born before June 1st (and who had spent one extra year in socialist education) have lower levels of democratic support compared to their birth-year peers born after June 1st. The γ_s are survey-year fixed effectsFootnote ¹¹ , and θ _l are federal state (Länder) fixed effects (to account for any unobserved regional effects), while the standard errors are clustered at the BirthYear-BeforeJune-East level. We also control for respondents’ ageFootnote ¹² and gender to account for the potential life-cycle effects on the outcomes and any confounding effects within the cohort dummy.

Table 1 reports the results of the DiDiD estimation, highlighting the central coefficient β ₈ of Equation 1. As expected, the coefficient of the triple interaction is negative and statistically significant. Substantively, this means that – compared to their contemporaries born in the same year – the extra year of socialist schooling in East Germany for those born before the 1^st of June between 1973 and 1982 caused a decrease of 0.319 points on the original scale (at the 95% level) in democratic support. This captures the within treated cohort effect and is equivalent in size to about half of a standard deviation in democratic support for the entire sample. This result provides support for our first theoretical expectation that authoritarian indoctrination drives the negative impact of growing up in a non-democratic regime on later-life support for democracy.

Table 1.

DiDiD coefficients on democratic support and satisfaction (Model 1)

Table 1. Long description

A table with 12 rows and 4 columns. The columns are labeled ‘Outcome’, ‘Coef.’, ‘s.e.’, and ‘Democratic support’. The rows list various factors such as ‘Born before June’, ‘Cohort (1973-1982)’, ‘Before June x Cohort’, ‘East Germany’, ‘Before June x East’, ‘Cohort x East’, ‘Before June x Cohort x East’, ‘Age’, ‘Sex’, ‘Constant’, ‘Survey-year fixed effects’, and ‘Survey weights’. Each row provides the coefficient and standard error for the corresponding factor. Notable values include a coefficient of -0.055 for ‘Born before June’ with a standard error of 0.031, and a coefficient of -0.319 for ‘Before June x Cohort x East’ with a standard error of 0.128. The table also includes the number of observations (3,863) and the R-squared value (0.057).

Significance: ***p < 0.01, **p < 0.05, *p < 0.1.

Note: Included survey years for democratic support: 2008 and 2018. The dependent variable was inverted so that higher values represent more support for democracy. Data: ALLBUS.

Since we cannot test the parallel trends assumption given the lack of pre-treatment values for democratic support, we run several robustness tests for our analysis. First, we run models in which we remove the land fixed effects (Model A2.1) or remove university graduates (Model A2.2). The results remain substantively identical when land fixed effects are removed, while the effect is no longer statistically significant when removing university graduates from the sample. In the conclusion, we present potential reasons as to why our effect disappears when removing university graduates and provide avenues for future research. Second, we iteratively shift the treatment window in terms of months away from January to see how sensitive our treatment is to the inclusion or exclusion of treatment/control months (see Tables A4 and A5). We manage to capture the negative effect of socialist education with as little as two treatment months included in the treatment condition, and the effect similarly disappears once two control months (or more) are included in the treatment condition. Third, we run a placebo test (see Table A6) by using only months from the control condition such that each month – starting with June – is incrementally included in the treatment condition and compared with the remaining months (until December, while all the treatment months – January to May – are excluded from the analysis). These coefficients do not exert systematic or consistent effects on the outcome, do not reach standard levels of statistical significance, or, when they do, are opposite to our results, which is consistent with the expected effects of placebo tests.

Fourth, we change the treatment window of the treated cohorts to fully Soviet education years (born between 1952 and 1962) (pre-fall of the Berlin Wall) and to fully democratic education (born between 1988 and 1999) (post-fall of the Berlin Wall). These act as placebo cohorts, as the individuals within them would have been exposed to similar lengths of socialist or democratic education. There are no statistically significant effects for the DiDiD coefficient when estimating our main models with these two placebo cohorts (see Table A7). Fifth, we also combine the two approaches described above, but the DiDiD coefficient is mostly statistically insignificant when combining the placebo cohorts with the incremental placebo birth-month windows (see Tables A8 and A9). Finally, we re-estimate the main model and exclude each federal state iteratively from the estimation sample (Tables A10–12). Our results remain largely consistent under this robustness test.Footnote ¹³ Overall, these findings reinforce our confidence that our DiDiD setup appropriately captures the negative, long-term effect of authoritarian education on support for democracy.

Global analysis: research design

While the GDR case provides compelling causal evidence of the negative imprint of authoritarian schooling, it also raises the question of whether the effect we identify reflects primarily the presence of authoritarian content or the absence of democratic education for the treated cohorts (born January to May). To disentangle these effects and assess their broader relevance, we now turn to a cross-national analysis that explicitly measures the content and effort of political education across regimes.

Measuring democratic and authoritarian school education

We distinguish between two dimensions of political education: its content and the effort devoted to political socialization. Content refers to the political values and norms conveyed through schooling, ranging from authoritarian messages that emphasize obedience and loyalty to authority to democratic messages that promote pluralism, political rights, democratic institutions, and critical thinking. Effort captures the extent to which regimes prioritize the transmission of political values through the education system, regardless of the ideological direction of those values.

To measure educational content, we construct a democratic content index based on four expert-coded indicators from the V-Indoc dataset:

1. 1. Teaching democratic institutions and values: ‘How would you characterize the dominant societal model(s) or ideology(ies) promoted through the history curriculum?’Footnote ¹⁸ Selected options: Democratic norms, e.g., liberalism or pluralism, and democratic institutions, e.g., elections.

2. 2. Teaching political rights: ‘To what extent does the curriculum of subjects that include the teaching of political values cover topics related to individuals’ political rights and duties?’

3. 3. Teaching pluralism: ‘When historical events are taught, to what extent are students exposed to diverse views and/or interpretations of these events?’

4. 4. Teaching critical thinking: ‘To what extent do students have opportunities to discuss what they are taught in history classes?’

Each indicator is also analyzed separately in the empirical section. Responses are measured on four-point ordinal scales and rescaled to range from 0 (most authoritarian) to 1 (most democratic). Higher values indicate a stronger emphasis on democratic norms and citizenship.

Political education effort is measured using an index capturing (1) whether primary school students are required to study a subject focused on political values,Footnote ¹⁹ (2) whether such instruction is required in secondary school, and (3) the extent to which the history curriculum promotes a specific societal model or ideology. Together, these indicators capture the degree to which political value transmission is institutionalized within the education system.

Following research on political socialization, we assume that educational imprinting is most likely to occur during respondents’ impressionable years. We therefore assign each respondent the average value of the educational measures observed in their country between ages 5 and 15, corresponding to the period of compulsory schooling for most cohorts. This procedure generates a unique measure of educational exposure for each country-cohort born from 1940 onwards.Footnote ²⁰

For some respondents, these formative years overlap with a regime transition. Rather than assigning such cohorts to a single regime type, we average educational exposure across the full impressionable period. This approach reflects our interest in cumulative educational experiences and captures the mixed political messages that transitional cohorts may encounter before, during, and after regime change. Because regime transitions are relatively rare, most respondents were exposed exclusively to either democratic or authoritarian educational systems during their school years.

Exploring political education content and effort

Before turning to the main analyses, we briefly examine the relationship between the two educational dimensions introduced above. Figure 2 plots the democratic content and political education effort indices separately for democracies and autocracies. As expected, democracies and autocracies differ substantially in the content of political education. Democracies are considerably more likely to promote democratic values, institutions, rights, and pluralism through the curriculum. The average value of the democratic content index is 0.67 in democracies, compared to 0.33 in autocracies.

Figure 2.

Relationship between education content and effort by regime type (1945–2020).

Note: Indicators range from low to high. Regime type is measured using the V-Dem regime classification (Coppedge et al. Reference Coppedge, Gerring, Knutsen, Lindberg, Teorell, Alizada, Altman, Bernhard, Cornell, Fish, Gastaldi, Gjerløw, Glynn, Hicken, Hindle, Ilchenko, Krusell, Lührmann, Maerz, Marquardt, McMann, Miri, Paxton, Pemstein, Seim, Sigman, Skaaning, Staton, Sundström, Tzelgov, Wang, Wig, Wilson and Ziblatt2021). The results are based on annual data of the 106 countries included in the analyses below. Pearson’s R correlation coefficients are reported between the two indicators by regime type. Data: V-Indoc (Neundorf et al. Reference Neundorf, Nazrullaeva, Northmore-Ball, Tertytchnaya, Kim, Benavot, Bromley, Knutsen, Lutscher, Marquardt, Paglayan, Pemstein, Seim and Rydén2023).

Differences between regime types are less pronounced when considering political education effort. Autocracies devote slightly more attention to political education overall, with an average effort score of 0.62 compared to 0.48 in democracies. At the same time, democracies exhibit considerably greater variation in effort, ranging from countries that place little emphasis on political education to those where civic education occupies a central place in the curriculum.

The relationship between content and effort also differs across regime types.Footnote ²¹ In democracies, the two dimensions are only weakly related, suggesting that the degree of emphasis placed on political education varies largely independently of its democratic content. In contrast, autocracies display a somewhat stronger positive association between content and effort, indicating that regimes that place greater emphasis on political education are also more likely to promote a coherent set of political values through the curriculum.

The model

We test our theoretical proposition on the effect of education on support for democracy using the methodological approach of cohort analysis (Neundorf Reference Neundorf2010; Neundorf and Niemi Reference Neundorf and Niemi2014). This approach allows the identification of distinct characteristics of those generations that were socialized under different political and educational contexts. We expect to find clear generational differences in terms of democratic attitudes of people who experienced different educational contexts within the same country and between countries. This aggregate pattern of varying socialization experiences across countries and within the same country over time allows us to study the lasting imprint of political education as implemented by different political regimes.

For this, we need to distinguish three co-linear time trends: age, period, and cohort (APC) effects. A person could have positive or negative views of democracy because she is young – the so-called life cycle or ageing effect – or because she lives in a country that faces a big political corruption scandal – the so-called period effect that affects everyone, no matter their age or birth year – or because she was socialized at a certain point in history – the cohort effect. Here, we are mainly interested in the cohort effect, which we argue contains the imprinting effect of the education system in place during a person’s impressionable years. We therefore estimate an age, period, and cohort model. The most important covariates are, firstly, the age of the respondents, which we include as age in years. Secondly, cohorts are typically distinguished by their year of birth. Lastly, we include the survey year to capture the period effect. The problem of estimating these three time effects simultaneously is the so-called fundamental APC identification problem:

(2) $${Birth \;Year}\,\left( C \right) = {Survey \;Year}\;\left( P \right)-{Age}\;\left( A \right)$$

To break the linearity between these three factors, we refrain from including the birth year of respondents and instead (as introduced above) use the educational context during each country-cohort’s impressionable years as a proxy (Rodgers Reference Rodgers1982).Footnote ²⁷ We then simply estimate a linear regression model, using clustered standard errors on the country-birth cohort level as well as year dummies.

Results: the long-term effect of school education on democratic attitudes

We present our results in three steps. First, we investigate the unconditional long-term effect of the education content a person was exposed to on her democratic attitudes today. Second, we present the results of an interaction between the education content and the regime type at the time of someone’s schooling. Third, we include the effort the regime puts into teaching politics in school and how this interacts with the content.

Turning to our first analysis, Table 2 reports the regression coefficients and their standard errors, estimating the effect of our five measures of education content – one index and its four sub-components – on support for democracy (upper panel) and strong leaders (lower panel). But before discussing these findings, we turn to Model 0 for both outcomes, which excludes the education measures. Here, we replicate the findings of the existing literature that demonstrates that people who grew up in an autocracy are significantly less supportive of democracy (b = −1.157) and more supportive of a system with a strong leader (b = 4.977). The goal of our analysis is to uncover the mechanism behind this observational finding by explaining away this effect or at least reducing its size.

Table 2.

Linear regression predicting support for democracy and strong leaders

Table 2. Long description

A table comparing the effects of different education content measures on support for democracy and strong leaders. The table has 13 rows and 12 columns. Column headers include M1.0, M1.1, M1.2, M1.3, M1.4, M1.5, M1.6, M2.0, M2.1, M2.2, M2.3, M2.4, M2.5, and M2.6. Row labels include Autocratic regime (aged 5-15), Education environment (aged 5-15): Democratic education index, Teaching democratic institutions & values, Teaching political rights, Teaching pluralism, Teaching critical thinking, Autocratic x democratic education, Constant, Observations, and R-squared. Each cell contains regression coefficients and standard errors. Notable trends include significant coefficients for autocratic regime and democratic education index across different models.

Significance: *p < 0.05; **p < 0.01. Clustered standard errors by country-birth cohort.

Note: Entries are regression coefficients and their standard errors of a linear regression model. The dependent variables are measured as 0 = ‘no support at all’ and 100 = ‘complete support’. All models control for the following: individual controls: age, gender, education, and working. Macro controls – impressionable years: Age of political regime, (log) GDP per Capita, modernization index, political corruption, public goods provision. Macro controls – survey year: (log) GDP per Capita, political corruption, level of electoral democracy. Study FE, including dummy variables for the studies that were harmonized (with the WVS as the reference category). For full results, see Appendix 6.

In Models M1.1 to M1.5 in Table 2, we add the educational context measures one by one. Except for M1.5 predicting support for democracy, the effect of the regime type becomes insignificant and is reduced by 25% (M1.4) to 62% (M1.2) in size when controlling for the educational content a person was exposed to (compared to M1.0). The results confirm that respondents’ support for democracy today is higher the more democratic the school education was when they were between 5 and 15 years old (b = 2.098). Turning to the sub-components of this index, we see that this positive effect is mainly driven by the inclusion of democratic institutions and values in the curricula (b = 4.327) and the teaching of pluralist views (b = 3.162). However, teaching critical thinking does not affect support for democracy.

Next, we turn to the results predicting support for a strong leader. First, we note that the main socialization effect of the regime type is reduced by 11% (M2.5) to 32% (M2.1) when controlling for the education context at the time (compared to M2.0), even though the effect remains significant. Table 2 confirms that the educational content someone was exposed to, especially the inclusion of more diverse, pluralist views in the curriculum (b = −10.445), has a strong impact on her support for personalist rule. Overall, being exposed to the most democratic school curriculum compared to the most autocratic one reduces support for a strong leader by 7.4 percentage points.Footnote ²⁸

In Appendix 7, we present ten robustness tests of the findings reported in Table 2. These tests address alternative operationalizations, estimation strategies, sample restrictions, and model specifications. Specifically, we (1) distinguish regime age in democracies and autocracies, (2) restrict the analysis to country-years with at least three V-Indoc expert coders, (3) introduce country fixed effects to rely exclusively on within-country variation across cohorts, (4) estimate Hierarchical Age-Period-Cohort (HAPC) models, (5) estimate multilevel random-effects models with respondents nested in country-cohorts, (6) replicate the analyses using only the World Values Survey and European Values Study, (7) additionally control for household income, (8) include continent fixed effects, (9) exclude respondents without formal secondary education, and (10) exclude the contemporaneous democracy measure from the model.

Overall, the substantive conclusions remain highly robust. Across most specifications, the estimated effects of democratic educational content remain similar in direction and magnitude to those reported in Table 2 for both support for democracy and support for strong leaders. The strongest challenge arises from the country fixed-effects specification. Because this model relies solely on within-country variation across cohorts and removes all cross-national differences in educational content, three of the five democratic-content coefficients predicting support for democracy become statistically insignificant, although they remain positive in sign. In contrast, the corresponding effects on support for strong leaders remain consistently negative and statistically significant. Taken together, these findings suggest that part of the association between democratic educational content and democratic support is driven by cross-national differences in educational environments, whereas the relationship between democratic education and lower support for authoritarian alternatives is particularly robust across specifications. The remaining robustness tests confirm the main findings and indicate that they are not driven by coding decisions, data harmonization, sample composition, continent-specific patterns, or the inclusion of contemporary democracy as a control variable.

The results so far confirm that education content matters in shaping citizens’ democratic attitudes, whether strengthening their commitment to democracy or weakening their support for authoritarian alternatives such as personalistic rule. Table 2 estimates the average effect of our education measures using the theoretically possible range of these indicators. However, as we have shown in Figure 2 and Appendix 4, there is a considerable difference in how much democratic content autocracies and democracies implement in their education. In the next step, we therefore investigate whether the content of schooling has a different impact depending on the regime type. To assess this, we simply interact the regime in power during respondents’ youth (1 = autocratic; 0 = democratic) and the democratic content index. The results are included in Models 1.6 and 2.6 in Table 2. Figure 3 further plots the predicted values of this interaction, which reveal different patterns for our two outcomes.

Figure 3.

Predicted values (and 95% c.i.) of the interaction between pro-democratic education content and the regime type (aged 5–15).

Note: Estimates of the interaction effects are reported in Table 2, M1.6. and M2.6..

The positive effect of democratic education on support for democracy reported in Table 2 is driven by democracies only (b = 5.413), while in autocracies we find a weak negative effect (b = −1.891). We find a similar pattern when looking at the sub-components of the index, as reported in Appendix 7. However, we find no difference between autocracies and democracies in terms of how the education context affects long-term support for personalistic rule. In both regimes, a more democratic curriculum leads to significantly lower levels of support. There are no differences in the slope between democracies and autocracies when looking at the sub-components of the index.

Does this analysis help us to understand the gap in democratic attitudes, documented in Figure 1? Yes, to some extent, given that the level of democratic education differs sharply between democracies and autocracies. For example, taking the average values of democratic education, the predicted level of support for personalistic rule is 42 percentage points in autocracies and 34 percentage points in democracies. Looking at support for democracy, the predicted levels are 69 in autocracies and 71 percentage points in democracies. This is a very small difference and does not reflect the big difference we documented in Figure 1, where levels of democratic support were significantly lower in young compared to old democracies.

In the last step, we therefore explore whether the sharp differences in democratic attitudes are not just a product of the content of education but also the effort a regime puts into teaching politics in school. Figure 4 plots the predicted values of the interaction between pro-democratic education content and the level of political education efforts for autocracies and democracies separately. The two predictive lines distinguish between low and high efforts using the bottom and top-10% value cut-off points. As the figure reveals, there are no conditional effects of the effort of authoritarian regimes in teaching more or less pro-democratic educational content. However, when turning to democracies, the effort put into promoting politics in the curriculum (e.g., having compulsory civics courses) has a strong moderating effect on the educational content. At the most pro-democratic end of the scale, we find a 9-percentage point difference between democracies that strongly emphasize politics in their curricula versus those that do not.

A similar picture emerges when looking at support for personalist rule, where we find a staggering 12-point difference between high-and low-effort democracies that both teach very pro-democratic curricula. If we compare the high-effort democracies that include average pro-democratic content in their education to high-effort autocracies with average content, the estimated difference in support for democracy is a 7-point difference, with a 13-point difference in support for a personalist ruler. However, if we compare low-effort regimes, the differences are merely 1 point for democratic support and 8 points in support for strong rulers.

Figure 4.

Predicted values (and 95% c.i.) of the interaction between pro-democratic education content and level of political education efforts by regime type.

Note:Predicted values and their 95% confidence intervals were estimated based on the same regression as specified in M1.1 and M2.1 in Table 2, including country-birth cohort clustered standard errors. The content index ranges from 0 (most authoritarian) to 1 (most democratic). The political education effort was divided into the bottom-10% and top-10% values. The full results are reported in Appendix 9.