Nevio Moreschi 492346
Title: The nature of the education cleavage in Democratic Eroding countries
Abstract
The current third wave of autocratization has assumed global proportion. Many transitioning
countries have arrested their process of democratization. In some instances weaker and
consolidated democracies alike have also undertaken a process of democratic erosion or
democratic hollowing. In these latter cases, however, the elected elite needs time to
substantially undermine those liberal-democratic institutions that constrained their power.
The process of democratic erosion often overlaps with multiple electoral cycles, which gives
those parties committed to liberal democracy a chance to challenge the elite at the ballots and
interrupt the autocratization process. This paper investigates whether this political division
between incumbent parties and pro-democratic opposition coincides with a specific cleavage
as defined by Baiern and Maier.
Keywords: Autocratization, Democratic Erosion, Educational cleavages, Political Cleavages.
Theoretical and methodological framework
Introduction and starting hypotheses
In the context of transitional countries, the emergence in the political arena of a
pro-democratic camp against an authoritarian one has been observed in several instances
(Saxonberg, 2001; Selcuk and Hekimci, 2020). A reconfiguration of a political cleavage
along similar lines is taking place also in those countries transitioning from a democratic
regime to an authoritarian one. In this paper, the concept of autocratization is declined
according to Luhrmann and Lindberg’s definition. Autocratization is thus seen as the loss of
democratic traits of a regime, regardless of whether the country affected by it is a fully
consolidated democracy, and whether or not this process results in a full democratic
breakdown (Luhrmann and Lindberg, 2019). In their systematic analysis of the phenomenon
the two scholars have also observed that in the majority of cases, the democratic countries
that transitioned to an authoritarian system have done so as a result of democratic erosion
(ibidem). In other words, in the countries in question, some of the fundamental institutions of
liberal democracy have been intentionally undermined by a political elite that rose to power
through regular elections.
1
As aforementioned, when this process of erosion takes place, a united political opposition
tends to form against the ruling elite and under the banner of democracy (Selcuk and
Hekimci, 2020). Since this usually happens while free elections are still in place, often this
pro-democratic camp challenged the non-democratically committed elite at the ballots. As an
example, it is possible to mention Venezuela and the formation of the Democratic Unity
Roundtable (Valenzuela, 2014), or Poland and the creation of the Civic Coalition for the
general elections of 2019. In other cases, such as in Turkey, these alliances failed to formalize
into a proper electoral coalition. This latter outcome is mainly due to the persistence among
the opposition parties of different inter-cutting cleavages that run even deeper than the
authoritarian-democratic one (Selcuk and Hekimci, 2020).
These differences in political opposition practices to halt the process of democratic erosion
call for another clarification of a key concept, that of cleavage. When the term cleavage has
been used to indicate the re-positioning of parties in the context of democratic erosion, the
concept encapsulates two specific dimensions: a political one (opposition coalition vs
incumbent party) and a value one (pro-democracy vs pro-transition). This translates into an
abandonment of the demanding definition of the term developed by Bartolini and Mair, a
definition that after decades of debates was finally able to bring some clarity to the literature
on the subject (Robert, 2002; Kriesi, 2010). Bartolini and Mair's conceptualization of the
term, following its original interpretation by Rokkan, includes a third dimension: a social one.
Summarizing their argument: a cleavage is composed of a socio-structural element, a
normative element that comprises all those values and beliefs that provide a sense of in-group
identification within the social group, and the organizational element of the social one in
institutions, of which a party is its most political expression (Bartolini and Mair, 1990, p.
215). Abandoning the social element means getting rid of a fundamental component of
cleavage as intended by those who theorized it.
Yet this approach characterized the studies of the European party systems of the early 2000s.
Political parties' divisions were found to be along post-materialistic values vs more strictly
materialistic ones, but these cleavages were also believed to be characterized by a lack of any
specific social group of references. This until Stubager found in the two societal segments of:
“more formally educated” and “less formally educated” the missing groups of reference of
those new values and political cleavages that other authors had previously observed
dominating the European political landscape (Stubager, 2010). Stubager reached the
conclusion that the particular features of the Danish education system tend to expose the
Danish students to liberal values. Consequently, the longer the amount of time spent in the
2
education system, the more likely it is to form a sense of belonging with other students on the
ground of shared liberal values. The two social and normative elements then find expression
in the organizational realm, with the social group of more formally educated students voting
for parties representing liberal values, and the parties reframing their stances and priorities in
an attempt to capture this electoral constituency. The Danish scholar was however skeptical
about the validity of his findings for other regions. Nonetheless, a series of research that
followed Stubager’s breakthrough showed a link between obtaining a higher level of
education, adopting liberal values, and voting for a specific political party even in other
European regions (Pavlovic et al, 2019; Ford & Jennings, 2020).1
From these premises, the present study aims at finding and analyzing similar links between
social segments with lower education and not just political parties with authoritarian
tendencies but political parties whose elite is behind a clear process of autocratization.
Doing so requires also finding an actual correlation between lower education and
authoritarian tendencies. Because even though education can be found as a significant
predictor of specific political support, what we are trying to assess here is the existence of
cleavage, and thus the value element and its role as connecting link becomes fundamental.
As such the first set of hypotheses can be framed as follows:
H1: Different education groups are more likely to vote for different political parties
(Link between the social-structural dimension the and organizational one)
H1b: Less educated citizens will be more authoritarian oriented
(Identification of the group’s shared values)
Hypothesis 1c: Liberal-authoritarian values cause a difference in voting behavior among
different educated social segments in democratic-eroding countries.
(Assessment of the influence of the group’s values in connecting the structural dimension to
the organizational one)
Given the existential perils of liberal democracy in the chosen case studies, other sets of
values are introduced into the equation in order to observe their influence on party support,
starting from the absence or presence of a pro-democratic commitment. This is a sentiment
1
Zoran Pavlović, Bojan Todosijević & Olivera KomarIt (2019) have found a correlation between
education and specific party preferences in most of the Balkan countries.
3
that is likely to be still acquired by the citizens through their time spent in the education
system. For instance, a majority of Cross-national studies have shown that the educational
level is the dominant social structural factor conditioning support for democracy, regardless
of the democratic quality of the country taken into consideration or the region covered by the
study (Evans and Rose, 2012; Chong and Grandstein, 2015, p.364). However, more recent
studies present more mixed results about the universality of education’s positive influence on
democratic behaviors and values. Some research confirms the existence of the phenomenon
while downplaying its significance for those citizens educated in authoritarian regimes
(Diwan and Wartanova, 2020). Other studies completely rule out the positive correlation
between education and democratic values when education is attained in authoritarian
countries with a long authoritarian legacy (Osterman and Robinson, 2022).
The four case studies of the current paper differ in regards to geographical region, levels of
democratic quality at the start of the process of erosion, and length of their democratic
history. So in the light of the more recent literature, the correlation between preferences for
autocracy-democracy and education level will also be tested.
Hypothesis 2b: Lesser educated citizens will prefer an autocratic regime, especially in those
countries with a longer democratic history
Moreover, the results of the elections in the democratic eroding countries can be a turning
point between a fully authoritarian transition and a democratic recovery, thus democratic
preferences might also account for the difference in voting behavior between different
educated social segments.
Hypothesis 2c: Democratic-autocratic preferences cause a difference in voting behavior
among different educated social segments in democratic-eroding countries.
However, it is important to note that the concept of democracy is not usually totally rejected
by the incumbent of democratic eroding countries. Democracy is rather re-framed to
downplay all those institutional elements established to limit the leader’s power, which are at
the base of the liberal declination of the term. For instance, after Hugo Chavez was elected
president of Venezuela in 1999 one of his first acts was drafting a new constitution. The new
constitution would highlight the participatory dimension of democracy at the expense of the
liberal institutions set up by the Punto Fijo Pact of 1968, which were until that point the
4
pillars upon which the democratic political system of the country was built (Canache, 2012;
Coker, 2014, p, 88; Garcia-Guadilla and Mallen, 2018). According to the new constitution the
“People”, as a unitary group, became the sole and ultimate source of authority. This newly
established communal model of democracy soon translated into the prevarication of civil and
individual rights by the will of the majority (Garcia-Guadilla and Mallen, 2018), and allowed
the leader chosen through popular election to raise himself above all the other political
institutions.
Hungary's process of democratic erosion has also been accompanied by a reframing of the
concept of democracy both at a constitutional level and in the public discourse. The features
of “democracy” have not been formally altered by the constitutional amendments. However,
Fides’s supermajority in the legislative assembly paved the way to the substantial
undermining of core liberal-democratic principles. For instance, the separation of powers
characterizing a liberal-democratic institutional setting has been hampered by the executive
extension of its control over the central bank and the appointments for the judiciary offices.
Check and balances assured by some prerogatives of the parliamentary opposition have also
been progressively scrapped (Sajò, 2019; Drinóczi and Bień-Kacała, 2019). This new
reconfiguration of the Hungarian political system has been made explicit by Orban himself
when in 2014 he declared that “We [the Hungarian people] have to abandon liberal methods
and principles of organizing a society. The new state that we are building is an illiberal state,
a non-liberal state” (Buzogany, 2017, p. 2). Even in this case the legitimacy of the incumbent
still derives from the electoral process and the coincidence between the will of the people and
the will of the majority (Sajo, 2019). These two cases fit the main model of autocratization
through democratic erosion developed by Lhurmann and Lindberg (2019) and parallel also
the main patterns of autocratization exhibited by Turkey and Poland (Drinóczi and
Bień-Kacała, 2019). Bolivia’s history of democratic erosion is quite peculiar. Evo Morales
was elected for the first time in 2005, and despite the drafting of a new constitution that
strengthened the idea of popular and indigenous participation, the erosion of liberal
democratic institutions never reached the intensity observed in the aforementioned cases and
the incumbent was always re-elected through fair and free elections. Nonetheless, the
phenomenon was still present and gained speed after the attempt by Morales to remove the
presidential terms limit in 2017. (Velasco Guachalla et al, 2021). This brief excursus on the
characteristics of the process of democratic erosion was necessary for mainly two reasons.
First, in the eventuality that Hypothesis 2c is refused, the reason can now be traced back to
the fact that the incumbents generally do not present themselves as anti-democratic. Thus the
5
democratic preferences of citizens might not be playing a role in explaining the divergence in
support for the incumbent party. Second, in the case of Hypothesis 1a being accepted, it
brings up the possibility that behind the incumbent’s support by the lesser-educated lies their
broader and non-strictly liberal conception of democracy rather than a stronger preference for
a fully autocratic regime.
Hypothesis 3b: Lesser educated citizens are more likely to have an illiberal conception of
democracy
Hypothesis 3c: Illiberal-liberal conceptions of democracies cause a difference in voting
behavior among different educated social segments in democratic-eroding countries.
Methodology
Case studies
The methodology chosen to test these hypotheses closely follows the one illustrated by
Pavlovic et al (2019) in their article “Education, Authoritarianism, and Party Preference in
the Balkans”. The case studies, of course, differ since their aim is to explore the educational
cleavage in the Balkan region, while in this case, the focus is on countries undergoing a
process of autocratization. The democracies chosen as subjects of this study are Bolivia,
Turkey, Hungary, and Poland. Their choice is based on the fact that in all these countries the
quality of liberal democracy has declined steadily in recent years, as both the literature (Agh,
2016; Gerschewski, 2020; Velasco Guachalla et al, 2021) and the V-Dem index indicate
(Figure 1). The index also shows the date of the beginning of the phenomenon in each of the
case studies: 2005 in Turkey, 2009 in Hungary, 2015 in Poland, and since 2015 the quality of
democracy worsened sharply in Bolivia as well.
Figure 1
6
The data that will be used in this study are taken from the EVS-WVS Joint dataset
2017-2020. In all four countries, the surveys were taken in 2017. The year, with perhaps the
exception of Turkey, is time-wise optimal. In fact, a survey taken too close to the incumbent’s
first election runs the risk of not being fully revealing, since the process of democratic
erosion is protracted in time, and since also the political opposition needs time to mobilize.
At the same time, data about voting intentions from surveys conducted too long after the
beginning of the phenomenon of autocratization would risk being invalidated by the citizens'
disillusions with the electoral process, since it has to be expected that it progressively
become less free and less fair. Even the citizens' fear of an autocratic elite that has already
solidified its power might lead to untrustworthy responses.
Concepts operationalization
A three-level categorical variable on the highest education level attained by the respondents
is often used in the analysis section as the independent variable. The variable was already
codified in its three levels (primary, secondary, and tertiary education) by the WVS and the
EVS. Regarding the political support for each political party, in Hungary, Bolivia, and Turkey
the respondents were asked to answer the following question: “If the elections for the
Parliament were held tomorrow/next week, would you vote? Which party or coalition would
you vote for?”, while in Poland the question was framed in a slightly different way “Which
party or coalition do you feel the closest to?”. Not all the parties of the 4 countries are going
to be observed in their relation to respondents' education and their values. In this study, only
the party of the incumbent and the main opposition party/coalition are taken into
consideration.
The most complex part regards the construction of the values dimensions. For the
authoritarian-liberal one, Pavlovic et al. utilized a conventional scale of authoritarianism
composed of nine items obtainable from the EVS and WVS questionnaires: (1) whether
homosexuality, (2) abortion, (3) divorce, (4) euthanasia, and (5) having casual sex are
acceptable; (6) whether people prefer a strong leader who does not have to bother with
parliaments and elections; (7) whether obedience is an important quality child should learn at
home; (8) whether one should always love and respect one’s parents regardless of their
qualities and faults; (9) and whether people believe that greater respect for authority in the
future is a good thing (Pavlovic et al, 2019). This scale has been adopted already by Regt et
7
Al. (2011, p, 302), which in turn re-elaborated it from Altemeyer’s Right-wing
authoritarianism (RWA) scale (ibidem).
In Altemeyer’s original work authoritarianism was conceptualized as the covariance of a (1)
strict adherence to conventional norms and values (conventionalism), (2) an uncritical
subjection to authority (authoritarian submission), and (3) feelings of aggression toward
violators of norms (authoritarian aggression) (Altemeyer, 1988).2
In the RWA scale the first 5
items measure the latent variable “conventionalism,” while the later 4 the “authoritarian
submission and aggression” (ibidem).
In this paper, the model’s fit for each one of the case studies singularly will be asserted
through Confirmatory Factor Analysis of the scale’s items (Ariely and Davidov, 2011).3
An
important difference is the exclusion of one of these nine original items, namely “whether
one should always love and respect one’s parents regardless of their qualities and faults”
since the question was not present in all the four questionnaires.
The dimension of the democracy-autocracy preference (DAP) scale is operationalized using
the WVS and EVS items resulting from the following questions: “For each, would you say it
is a very good, fairly good, fairly bad, or a very bad way of governing this country?”: (1)
Having a strong leader who does not have to bother with parliament and elections. (2) Having
experts, not governments, make decisions according to what they think is best for the country
(3) Having the army rule (4) Having a democratic political system (Ariely and Davidov,
2010). After having conducted a CAF in all the single case studies on cross-national
democratic attitudes, Ariely and Davidov (ibidem) found that the DAP model was not
acceptable for many of those countries. The two authors believed that the main cause lays in
the 4th item “having a democratic political system”, and in the broad arrange of meanings
that can be attributed to the term. Since this weakness of the model could just be
context-dependent, at first the D.A.P. will be tested maintaining all of its original items.
Finally, the third value dimension investigates the respondents’ illiberal-liberal conception of
democracy through the following questions: “How much is this feature essential for a
democracy?”: (1) Democracy: Religious authorities interpret the laws (2) Democracy: People
choose their leaders in free elections. (3) Democracy: The army takes over when the
government is incompetent. (4) Democracy: Civil rights protect people’s liberty against
3
The CFA estimates the relations between observed indicators (our 9 items) and the hypothesized
latent construct (authoritarianism), and provides fit indices that report whether the hypothesized
structure of associations between a latent construct and its proposed indicators fits the data.
2
The reader can see how the three dimensions parallel those of “the authoritarian personality”
developed by Adorno in his homonymous book.
8
oppression. (5) Democracy: Women have the same rights as men. Higher results will be
associated with the respondent's conception of democracy which is in line with the classic
liberal interpretation of democracy, with items 1 and 3 re-coded in a way to fit this overall
logic of the model. The attempt to construct this model, referred to from now on as “LID
(Liberal-Illiberal Democracy)”, is original in nature but it will be justified in the light of the
existing literature.
As mentioned in the previous section, the term “democracy” has come to assume different
things in the public discourse of democratic eroding countries.
The classic liberal-democratic conception entails, alongside the democratic component of the
popular will, a separation of powers among different institutional bodies, a system of checks
and balances that effectively maintain and reinforce this separation (Canache, 2012, p.98), the
respect of civil and minority rights (Dahl, 1989, 318; Canache, 2012, p.98), the separation
between religion and state, the rule of law and the citizens' equality before it (Dahl, 1989, p.
88-114), and the non-intervention of the army in political matters (Huntighton, 1957; 1996).
The LID scale unfortunately does not cover all these aspects. Items such as “Having a strong
leader who is not bothered by the legislative” are linked to the idea of the separation of
powers but do not ask the respondents what is its relation to democracy. The item regarding
religious authorities is included because of the autocratization features observable in some of
the case studies. For instance, in Turkey, the reframing of “democracy” as “the people’s will
represented by the president Erdogan” was paralleled by the reframing of “Turkish people” as
“the religious Turkish people”(Sozen, 2020, p.19). While In Poland scholars have observed
an alliance between segments of the national Catholic religious hierarchies and the ruling PiS
(Zug and Zug, 2020; Meyer-Resende and Hennig, 2021).
Analysis
Chapter 1
Testing hypothesis H1: Different education groups are more likely to vote for different
political parties
In order to test Hypothesis H1, it has been conducted a series of binomial logistic regression
for each case study with the voting for the incumbent political party versus voting for any
other party as a dependent variable. A second logistic regression was carried out this time
with “voting for the main opposition party versus voting for any other party” as the
dependent variable. In this first phase education was the only predictor of party orientation
9
used, no other control variables were added. As expected, education and voting for the
incumbent party are negatively correlated in all the countries observed, while the relation
between being more educated and voting for the main opposition party is not equally linear.
For instance, in the case of Hungary, higher educated citizens are also less likely to support
the coalition composed of “Dialogue For Hungary” and the moderate “Hungarian Socialist
Party” (MSZP), which since 2014 run together in different coalitions to challenge “Fidesz”
and Orban.
The results of the binary logistic regressions are reported in the Annex, while Figure 2 shows
the decrease in the predicted probability of voting for the Law and Justice (PiS), Fidesz,
Movement for Socialism (MAS), and the Turkish Justice and Development Party (AKP), at
the increase in education level.
Figure 2 Mean predicted probabilities of different voting outcomes for educational level
10
Before continuing with the testing of the main hypothesis, a preliminary step consists in
identifying the parties’ stance on the main values and concepts that the EVS, DPA, and LID
scales try to capture. If it appears that the parties more strongly supported by the
less-educated citizens are also characterized by strong support for traditional beliefs, a
stronger emphasis on political authority, and a more negative view of constitutionalism and
democracy, then this would be an ulterior hint toward a connection of the socio-structural
side of the clevage with the organizational side through the sharing of the aforementioned
values. This connection could just be a case of correlation without causation, and therefore it
will require more rigorously testings later on. Nonetheless, results in line with the starting
hypothesis would further justify the direction of this study.
Through data obtained from the Manifesto Project Database, Table 1 highlight the parties'
stances on selected issues as expressed by their electoral manifesto of 2015 and 2018, with
higher scores being synonyms of a higher emphasis on a specific issue.
Table 1: Parties’ manifestos scores on selected issues
Country Constitutionalism:
Positive
Democracy:
positive
Traditional
Morality:
Positive
Political Authority Strong
government
Poland 2014 - 2018 2014 - 2018 2014 - 2018 2014 - 2018
PiS 0.146 - 0.548 1.987 - 1.371 2.338 - 2.955 0.205 - 0.426
Civic Platform (PO) 0.091 - 0.236 0.912 - 5.813 0-000 - 0.079 0.000 - 0.000
Hungary 2014 - 2018 2014 - 2018 2014 - 2018 2014 - 2018
Fidesz 1.133 - 0.000 1.133- 0.673 0.567 - 8.52 5.666 - 0.448
Hungarian Socialist
Party
0.000 - 0.000 4.727 - 4.173 0.169 - 0.298 0.000 - 0.000
Dialogue 0.000 - 0.055 3.459 - 3.77 0 - 0.328 0.364 - 0.055
Bolivia 2014 2014 2014 2014
MAS 1.514 0.336 0.336 0.589
National Unity (UN) 0.000 5.057 0.000 0.126
Turkey 2014 - 2018 2014 - 2018 2014 - 2018 2014 - 2018
11
AKP 0.000 5.972-1.795 1.529-0.649 0.049-0.115
CHB 5.972 - 0.134 9.524- 8.011 0.348- 0.000 0.000- 0.000
Perhaps counterintuitively, it is possible to see that some of the parties leading the process of
democratic erosions are also emphasizing their support for constitutionalism. This could be
explained by their efforts at legitimizing their rule and increasing their power through
constitutional emendation which was mentioned in the introductory section.
The other scores seem more in line with the study’s hypotheses. In the case of Poland
between 2014 and 2018, the PiS manifestos decreased their emphasis on democracy to stress
the importance of greater political authority and traditional values (the main dimensions
behind the EVS authoritarian scale). Hungary is perhaps the case study in which the cleavage
around democracy is the most formalized at the organizational-political level: with Fidesz
refocusing its platform on conventional values and political authority, and the opposition
keeping their ground as the main proponent of democracy. A similar trend can be seen also in
the non-European autocratizing countries.
Testing Hypothesis H1b: Less educated citizens will be more authoritarian oriented
In both the works mentioned in the methodology, the latent variable “authoritarianism” was
constructed by first subjecting the nine items to a multiple-groups confirmatory factor
analysis with a maximum likelihood method of estimation (Regt et al, 2011; Pavlovic et al,
2019). Similarly in this study, a series of Confirmatory Factor Analyses (CFAs) were
conducted in each of the case studies. The software employed for the tests, AMOS.16, is not
capable of computing variables presenting missing values, so an expectation-maximization
algorithm has been used to bypass the issue. The model analyzed is the one depicted in
Figure 3. The 8 items have been divided among those constructing the “Adherence to
conventional norms” and those composing the “Submission to Authority” dimensions of an
authoritarian personality, in accordance with Altemeyer's reconceptualization of the term.
Finally, the variable “Important child quality: obedience” has been re-coded in order for it to
follow the same direction as the others (Higher score = Lower authoritarian tendencies).
12
Figure 3: Confirmatory factor analysis model of EVS Authoritarianism Model
The alpha value of the 9 items scale found by Regt et al, (2010) was around 0.60. Similar
values have been observed in the country by country analysis by Pavlovic et al. The overall
alpha value of the modified EVS authoritarian model used in the present study does not differ
despite the elimination of one of the items. The alpha reported in Table 2 is the standardized
one because some of the items in the survey were operationalized on different scales4
.
Table 2: EVS Authoritarianism Scale: Goodness of Fit Statistics of CFA5
Country Standardized Alfa6
RMSEA CFI
Poland ,703 0,070 0,961
Bolivia ,580 0,031 0,980
Hungary ,660 0,066 0,954
Turkey ,755
6
A generally accepted rule is that an α of 0.6-0.7 indicates an acceptable level of reliability, and
0.8-1.00 is a very good level (Cho; 2016).
5
The goodness of the statistics of the Confirmatory Factor Analysis are interpreted according to the
guidelines set by Schreiber, James B. et al. in “Reporting Structural Equation Modeling and
Confirmatory Factor Analysis Results: A Review” (2010).
4
For instance, the question “In which circumstance is Homosexuality Justifiable?” is a Likert scale
with 10 possible replies from “never justifiable” to “always justifiable”, while the variable“Having a
strong leader who does not bother with parliament and elections” is coded on a scale of just 4 values.
13
The model’s Goodness of Fit for Turkey, like for all the others, is acceptable according to the
literature. However, the standardized factor loading of the items associated with the
“submission to authority” dimension is non-existent.7
The only reason why the model is
statistically acceptable appears then to be the strong loading of the items of the sole
“conventionalism” dimension.
This misfit of the model in only this singular case study might be due to cultural peculiarities
of the country, the non-equivalence of particular items across countries, or a combination of
both (Ariely and Davidov, 2011), it can not be ruled out also the human error during the CFA
or in the interpretation of its outputs. Whatever the reason is, the EVS authoritarian scale has
not been used in the analysis of the Turkey case study.
In all the other instances the one-way ANOVA shows that the authoritarian scale is positively
correlated with the level of formal education. Higher values in the authoritarian scale, or
greater liberal values, are associated with higher levels of education attained. This is in line
with Stubagher’s findings on the Danish education cleavage (Stubager, 2010), Ford and
Jennings’s ones about the Western European region (2020), and those of Pavlovic et al’s
research on the Balkan countries (2019). It appears that even in countries of Central Europe
such as Hungary and Poland, and even as far as Latin America with the Bolivian case, there
is a correlation between education and the adoption of liberal values.
Stubagher’s in his research went a step further, following more closely Bartolini and Maier’s
conceptualization of cleavage. He also demonstrated that for the social segment of the more
educated, liberal values function as identity providers allowing them to strengthen their
self-identification as a different group. The one-way ANOVA does not allow us to test this
dimension, but its results reported in Table 1 are nonetheless very telling regarding the
connection between the two variables. A connection which is moderate in Central Europe η2
> 0.06 and weaker but still statistically significant in Bolivia with η2 > 0.03.8
8
The interpretation of the η2 has been made following the SPSS guide available at the following link
https://www.spss-tutorials.com/effect-size/. Rule of thumbs in statistics are usually frowned upon.
However, at least in regards to the Central European case-studies, these results seem in line with the
theoretical literature.
7
The Factor loading of Turkey’s “EVS authoritarian model” can be found in the annex.
14
Figure 4: Relationship between authoritarianism and education
F(2, 2058)=38,26 p<0.001 η2= 0,036 F(2, 1504)=49,06 p<0,001 η2=0,061
F(2, 1340)=58,285 p<0.001 η2=0.080
Hypothesis 1c: Liberal-authoritarian values cause a difference in voting behavior among
different educated social segments in democratic-eroding countries.
Finally, it is time to observe the phenomenon of the distribution of political support by
values and level of education in order to assess its nature as a proper cleavage. To do so the
role of authoritarian-liberal values in party support will be estimated by introducing the EVS
authoritarian scale as a control value alongside the independent variable “level of education”.
At first, the Kappa index of the sole variable “level of education” was estimated by summing
15
the standard deviations of the three coefficients of each of its three levels. This “gross” Kappa
(Pavlovic et al, 2019) was later confronted with the net Kappa, or in other words the overall
Kappa once the regression coefficient of the authoritarian scale is also introduced. The
greater the change between the two Kappa indexes, the greater the influence of authoritarian
values in explaining the differences in political preferences among the education groups. The
results of this operation are summarized in Table 3.
Table 3
Case study Gross Kappa Net Kappa Change Relative change
Poland
PiS 0,35 0,22 -0,13 -37%
PO 0,45 0,31 -0,14 -31%
Hungary
Fidesz 0,37 0,28 -0,09 -24%
Coalition 0,70 0.48 -0,22
Bolivia
MAS 0,36 0,22 -0,14 -38%
UN 0,54 0,33 -0,21 -39%
The changes in Kappa value show the influence of authoritarian and liberal values in
explaining the differences in political preferences by the different educational groups. This is
true for all the case studies and respective parties, except for the anti-Fidesz coalition in
Hungary, for which the two variables were not significantly correlated.
Interpretation of the findings
The analysis conducted thus far demonstrates the existence of a full-fledged cleavage in all
the four case studies observed. The lines of this cleavage separate the lesser and more
educated segments of society and, politically, those parties that have undertaken a process of
democratic erosion and those that are trying to stop them through the electoral competition.
What partially separates these social segments from each other, and links them to specific
political parties, is their different stance on the liberal-authoritarian axis. In fact, in line with
16
the mainstream literature, a strong correlation has been found between higher levels of
education and stronger liberal values. This correlation was more significant in the Central
European countries, and only moderately significant in Bolivia.
Stubager’s theory about the education system being a carrier of liberal values only in
Northern Europe and not necessarily in other European regions has been disproved in
following studies on the subject (Pavlovic et al, 2019; Lujan, 2020). Different education
levels have also been found to account for the divergent political support of specific parties in
other Latin American countries (Barrera et al, 2021). Coherently with this literature, The
results of the one-way ANOVA seem to indicate that indeed the education system of Bolivia
exposes the students to liberal values, but to a lesser extent than the European ones.
The results of the kappa comparison present us with the reverse picture, with liberal values
accounting to a lesser degree for the political divergences among educational groups in
Poland and Hungary. This might be due to the presence in the two countries of reinforcing
value cleavages that characterize the education groups and their political choices (Pavlovic et
al., 2019). Yet this is also the case in Bolivia with the ethnic and educational cleavages being
reinforcing cleavages in the Bolivian society (Reimao and Tas, 2017, p.236). Another
reasonable hypothesis is linked to the supply-side dimension of politics. The clearer the
party’s stance on an issue the more accurately voters may decide whether to accept the party’s
pleas for support (Pavlovic et al. 2019). Adopting this perspective then MAS would just be
more able to promote those authoritarian-oriented values that characterize specifically the
lesser-educated Bolivians, yet the scores in Table 1 also run against this theory.
Chapter 2
Hypothesis 2b: Lesser educated citizens will prefer an autocratic regime
The analysis of the previous chapter provided some evidence in favor of the existence of a
cleavage that has as its social groups of references educated and less educated citizens, with
the latter mostly supporting those parties responsible for the declining democratic quality of
their country. Following the most recent findings in the literature on cleavages it has been
hypothesized that the set of values linking these social groups to these political parties was to
be searched within the authoritarian-liberal values dimension. Indeed, it appears that some of
the divergence in political preferences by different education groups can be explained by the
different degrees of authoritarianism characterizing their average member (Figure 4, Table 3).
17
However, this research aims to go one step further, investigating if the political support given
by the less educated to the incumbents’ parties, is also purposefully given because of a
weaker democratic commitment. This implies that voters are aware of the process of
democratic dismantling and yet they are not necessarily against it or even go as far as to see it
as a more favorable option than the democratic status quo. The democratic commitment is
going to be tested by adopting the DAP model as an independent variable. This model, like
the previous authoritarian scale, is also constructed through EVS and WVS items.
Figure 5: Confirmatory factor analysis model of Democracy-Autocracy Preference (D.A.P.)
Similar to the testing of the EVS authoritarian model, also in this case an
expectation-maximization algorithm has been used to account for the missing variables so
that the Confirmatory Factor Analysis could run. The item “Having a democratic political
system” has been re-coded in order to follow the other items, so that higher scores would be
synonymous with stronger preferences for a democratic system. Finally, the standardization
of the items was not necessary since all of them were ordered on the same ordinal scale with
1 and 4 as the extreme values.
Table 4: D.A.P. : Goodness of Fit Statistics of CFA
Country Alfa RMSEA CFI
Poland 0,454 0,149 0,890
Bolivia 0,287 0,052 0,973
Hungary 0,235 0,127 0,872
Turkey - - -
18
As it appears from the results of the CFAs reported in table 4, the measurement model is not
acceptable in two of four of the case studies. In the case of Poland the RMSEA is greater
than 0.1, and the factor loading for the item “Having a democratic political system” is null. In
the case of Hungary, alongside the RMSEA greater than 0,1, the alpha value is also very low.
For Bolivia, the model is acceptable but despite being re-coded the democratic variable is still
negatively loaded. Ariely and Davidov's (2011) research would suggest that the problem lies
in the open interpretation of the term “democratic”. These findings could indicate that the
DAP model as it is, might not be an expression of any latent variable in these specific
countries, let alone the latent variable that it was in our interest to observe: democratic
commitment. Having run into the same exact problem as Ariely and Davidov (2011), it can
be expected that no matter which modification will be made to the model, it is unlikely that
its outcomes will be acceptable. The elimination tout-court of what the two authors believed
to be the problematic item is also not an option: with only three items a CFA model has zero
degrees of freedom and is therefore not testable.
A recent argument made by Sokolov (2021) is relevant for understanding the cause behind
the model's weakness.9
According to him, such weakness goes deeper than the general
misunderstanding of the term democracy by the respondents, and it rather concerns the nature
of the items chosen. A non-democratically committed citizen might believe that democracy is
not the best form of government and that the rule of a strong leader not bothered by the
legislative is something of an improvement. Simultaneously, the same respondent might be
particularly skeptical of the army’s involvement in politics or it might be characterized by a
general mistrust for technocrats. This logic can be brought to a higher level: more religious
social segments might tend to prefer a theocratic regime change, and they would still be
characterized by a higher preference for autocracy over democracy despite being completely
against the eventuality of the army’s rule over society. What Sokolov’s argument implies in
this context is the following: it would be theoretically unjustified to require strong
intercorrelations and, thus, interchangeability, between these specific indicators used to
measure authoritarianism (Sokolov, 2021). Thus it would be inappropriate to use
correlation-based methods like SEM or CFAs to assess the validity of this kind of measure.
The course of action adopted was then to construct and operationalize a scale of autocratic
preferences despite the worse CFAs results, and repeat the analysis of the previous section.
9
Sokolov’s reasoning was developed for another set of other items of the EVS, some of
which will be composing the LID scale, but mutatis mutandis it covers also the subject of the
current analysis of the DAP model
19
As figure 6 shows, the results of the ANOVAs are strongly context-dependent. For the
Central European democracies, there is indeed a significant correlation between higher
education and higher commitment to democracy over authoritarian alternatives. In Turkey,
while the nature of the correlation between the two variables fits the hypothesis, such
correlations are non-significant. The findings relative to the case study of Bolivia diverge
from those of the rest. The Bolivian citizens who have attained a middle-level education, are
on average more skeptical about democracy than those less educated. The one-way ANOVA
between education and each of the single items composing the DAP scale (annex) indicates
that this is mostly due to the technocratic stances characterizing the more educated segments
of Bolivian society. Another partial explanation for this non-linear correlation is the
preference expressed by the mid-educated for a strong leader, which as expected is greater
than that expressed by higher educated Bolivians, but surprisingly is also greater than those
of the less educated.
Figure 6: Relationship between democratic preferences and education
F(2, 1299)= 41,686 p<0.001 η2= 0,060 F(2, 1467)=12,417 p<0.001 η2=0,017
20
F(2, 2030)= 14,090 p<0.001 η2=0,014 F(2, 2287)=1,054 p<0.349 η2=0,001
Hypothesis 3c: Autocratic preferences cause a difference in voting behavior among
different educated social segments in democratic-eroding countries.
With the introduction of the DAP scale as a control variable, the Gross-Net kappas gap is in
some instances larger than that resulting from the employment of the Authoritarian-Liberal
scale. However, the binomial regressions tell us that the DAP scale is not a significant
predictor of party preferences for most of the parties observed, with the exception of Fidesz
and the Turkish ones. The hypothesis that the preference for a regime change can explain the
differences in political preferences by the different educational groups can not be accepted or
rejected in toto, but it is instead context-dependent.
Table 5
Case study Gross Kappa Net Kappa Change Relative change
Poland
PiS 0,35 0,20 -0,15
PO 0,45 0,34 -0,11
Hungary
Fidesz 0,37 0,23 -0,14 -38%
Coalition 0,70 0,48 0,22
Bolivia
MAS 0,36 0,21 -0,15
UN 0,54 0,31 -0,24
Turkey
AKP 0,24 0,16 -0,08 -0,33%
CHP 0,22 0,19 -0,03 -12,5%
Interpretation of the findings
It has been found that in Poland, and to a lesser degree in Bolivia and Hungary the level of
education can account for the differences in autocratic preferences within the countries’
respective population. However, it is important to remember that Poland is also the country
21
with the highest level of preferences for democracy in absolute terms, despite this strong
correlation between the two variables among the case studies.
While the results regarding the Central European countries fell in line with the literature on
the education’s role in the strengthening of democratic support and democratic-related
political values, Bolivia offers us once again a fertile field for speculation. To start, a higher
than average preference for a strong leader has been found. A cross-national study conducted
by Sprong et al. in 2019, identified a significant link between economic inequality and the
citizens’ desire for a strong leader. When MAS achieved its first victory, the Gini index of the
country was close to 0.60, among the highest in the world (World Bank, 2022). To this day it
remains incredibly high, but it has been brought down to a historically low of 0.40 by the
time of Morales ousting from power. In the context of Bolivia, the strong leader's ability in
tackling the issue might have reinforced the societal belief that only by increasing the
executive powers income inequality can be reduced.
Another peculiarity of Bolivia is the strong preference for a technocratic system of
decision-making characterizing the more educated strata of society. MAS rose as a
counter-movement to those parties of the establishment which embraced the neo-liberal
agenda. Once in power, one of the first moves of the new government was the reduction of
the power wielded by the technocrats of the Ministry of Finance (Dargent, 2015, p.159).
Given the more educated opposition to Morales, pro-technocratic stances may be adopted as a
self-identification mechanism, to differentiate them further from the less educated and the
anti-technocratic and more populistic values of their party of reference (Teik, 2014).
Not much can be said about the data reported in Table 4, it appears that despite the existence
of a correlation between a higher level of education and a stronger democratic commitment to
autocratic alternatives, the former variable is not the main reason behind the respondents'
choice for supporting or not the incumbent party or its opposition. A Notable exception is
Hungary, in which the DAP scale is significant, and when used as a control variable produces
a larger Kappas divergence than the EVS authoritarian scale. It seems that in the country the
liberal vs authoritarian values is not the only set of beliefs differentiating the social segments
of the educational cleavage. This might even explain the country’s lower Kappas difference
observed in Table 3.
From a formal standpoint, Fidesz is not the only incumbent party that has abandoned any
mention of democracy in a positive light, while its opposition started to be characterized by a
specular trend (Table 1). This peculiarity might concern the empirical level and the intensity
of the process of autocratization affecting the country, but further research is needed.
22
Chapter 3
Hypothesis 3b: Lesser educated citizens are more likely to have an illiberal conception of
democracy
Sokolov's model was based on 6 items: “People choose their leaders in free and fair
elections”, “Civil rights protect people from state oppression”, “Women have the same rights
as men.”, “Religious authorities ultimately interpret the laws.”, “The army takes over when
the government is incompetent”, and “People obey their rulers.” (Sokolov, 2021). When a
CFA was conducted for the last three items that construct the dimension named authoritarian
notion of democracy (AND), the results were not significant. This is when the scholar
formulated the theory introduced in the previous chapter (ibidem). Given the overlapping of
items between the AND and LID scale10
, it is even more justified to put aside the CFAs
analysis and directly operationalize the latter scale.
F(2, 1485)=25,288 p<0.001 η2= 0,033 F(2, 2044)=8,888 p<0.001 η2= 0,009
10
The reader can see that of those 6 items, 5 are the same used to construct the current LID scale.
The decision to omit the item “People obey their rulers.” is linked to the fact that abiding by the ruler's
decision is not a sign of an illiberal interpretation of democracy per-se , unless some specifications are
added such as “people obey their rulers even if their decisions go against the spirit of the constitution,
or translate in an infringement of minorities’ rights”.
23
F(2, 1303)=36,243 p<0.001 η2= 0,053 F(2, 2363)=3,962 p=0.019 η2= 0,003
Hypothesis 3c: Illiberal-liberal conceptions of democracies cause a difference in voting
behavior among different educated social segments in democratic-eroding countries.
While at least in Hungary and Poland the one-way ANOVA showed that there is a strong and
significant correlation between higher formal education and an interpretation of democracy
closer to its liberal configuration and to its mainstream academic definition. The Gross-Net
Kappas comparison indicates that this is not the reason for the differences in political
preferences among the different educational groups. The LID scale is not a statistically
significant predictor for the majority of the parties under analysis. Only in two instances, did
we observe a divergent result: in regards to the political support of the PiS party in Poland
and for the CHB in Turkey. In the first case, the sign is coherent with the hypothesis,
although its impact is way weaker than that of the EVS authoritarian scale. In the second
case, the magnitude of change in Kappa is so minimal that it can be excluded from the
analysis despite its significance.
Table 6
Case study Gross Kappa Net Kappa Change Relative change
Poland
PiS 0,35 0,30 0,05 -14%
PO 0,45 0,41 0,05
24
Hungary
Fidesz 0,37 0,34 0,03
Coalition 0,70 0,68 0,02
Bolivia
MAS 0,36 0,29 -0,07
UN 0,54 0,42 -0,12
Turkey
AKP 0,24 0,20 0,04
CHP 0,22 0,23 +0,01
Conclusion
That a political competition between non-democratic committed elites and an organized
opposition emerges in democratic eroding countries was already evident without the need for
extensive research. That these two political camps draw their support from specific social
segments could have also been expected, and in fact, the first chapter of this study confirms
it. Even though their constituencies are not exhausted by these two social categories, it
appears that more educated citizens are less likely to feel closer to those parties responsible
for the hollowing of their democracy, while less-educated citizens are one of their electorate
of reference. Some authors would have been satisfied with these results and call for the
existence of an educational cleavage in the democratic countries undergoing a process of
autocratization. However, attempting to remain faithful to Bayern and Maier's
conceptualization of the term this study wanted to investigate the cleavage in all three of its
classic elements.
Identified the socio-structural element (the group of citizens who have attained only a lower
level of education) and its organizational element (the party of the incumbent leader), what
was lacking at the end of the first part of the present study was the psychological-normative
elements (Stubager, 2010). What is the set of ideas and beliefs that provide a sense of
in-group identification within the social category, and that works as a linkage between the
social category and the political party of reference? Following the findings of research that
have already covered educational cleavages in other contexts, we started from the hypothesis
that these values were those associated with an authoritarian personality. Indeed, from the
25
Kappas analysis, it appears that authoritarian and liberal values account substantially for the
divergence in political support by the different educational groups.
Nonetheless, the particular political situation of the case studies leads to hypothesizing other
values that could explain the support given to the incumbent party by the less educated
section of the observed societies. For instance, the literature found a link between education
and democratic support. Not only this link is confirmed to exist in three of the four case
studies, but in Hungary, these different degrees of democratic commitment are what explain
the support, or lack of it, for Fidesz.
To double-check that this democratic commitment was a commitment to the liberal
conception of democracy and not to its illiberal version, a new values scale was introduced
into the equation. What emerged is that the more educated citizens of the Central European
countries have indeed an idea of democracy more closely associated with the common
academic understanding of it, and for this reason, they might be more skeptical of their
national incumbent and its attempts at re-framing the concept. However, the different
understanding of democracy alone bears no influence on the citizens' political choices.
This study will now end on two final notes. The first is that overall education, for different
reasons according to the context, remains an effective method to reduce people's allure for
those politicians endangering our democracies. The second is to remember that this research
hasn’t examined those other cleavages that might exist in the autocratizing democracies
alongside the educational one. While recognizing the importance of education and the role
played by the more educated in the defense of democracy, we should then refrain from falling
into an over-simplistic and elitist view of this phenomenon.
Bibliography
Ágh, A. (2016) “The Decline of Democracy in East-Central Europe, Problems of PostCommunism,
63(5-6), p.277-287.
Altemeyer, B. (1998) “The Other “Authoritarian” Personality”, Advances in Experimental
Social Psychology, 30, p.47-91.
Ariely, G. and Davidov, E. (2011) “Can we Rate Public Support for Democracy in a Comparable
Way? Cross-National Equivalence of Democratic Attitudes in the World Value Survey”,
Social Indicators Research.
Barrera, O., Leiva, A., Martínez-Toledano, C. and Zúñiga-Cordero, Á. (2022) “Social
Inequalities, Identity, and the Structure of Political Cleavages in Argentina, Chile, Costa
Rica, Colombia, Mexico, and Peru, 1952–2019” In: Gethin, A., Martínez-Toledano, C.
26
and Piketty, T. ed. Political Cleavages and Social Inequalities: A Study of Fifty
Democracies, 1948–2020. London, Harvard University Press, p. 458-507.
Bartolini, S. & Mair, P. (1990) Identity, competition, and electoral availability. The stabilization
of European electorates 1885-1985. Cambridge, Cambridge University Press.
Buzogány, A. (2017) “Illiberal democracy in Hungary: authoritarian diffusion or domestic
causation?”, Democratization, 24(7), p.1307-1325.
Canache, D. (2012) “The Meanings of Democracy in Venezuela: Citizen Perceptions and
Structural Change”, Latin American Politics and Society, 54, p.95-122.
Chong, A. and Gradstein, M. (2015) “On Education and Democratic Preferences”, Economics
and Politics, 27, p. 362-388.
Coker, T. (2014). Dimensions of Democracy in Contemporary Venezuela. Social Justice, 40(4
(134)), 88–105.
Dahl, R. A. (1989). Democracy and its critics. New Haven , Yale University Press.
Dargent, E. (2015). Technocracy and democracy in Latin America: The experts running
government. New York, Cambridge University Press.
Diwan, I, Vartanova, I (2020) “Does Education Indoctrinate?” International Journal of
Educational Development, 78.
Drinóczi, T. and Bień-Kacała, A. (2019) “Illiberal Constitutionalism: The Case of Hungary and
Poland,” German Law Journal, 20(8), p.1140–1166.
Evans, G. and Rose, P. (2012) "Understanding Education's Influence on Support for Democracy
in Sub-Saharan Africa," Journal of Development Studies, 48(4), p.498-515.
Ford, R. and Jennings, W. (2020) “The Changing Cleavage Politics of Western Europe”. Annual
Review of Political Science, 23, p.295-314.
García-Guadilla, M. P. and Mallen, A. (2019) “Polarization, Participatory Democracy, and
Democratic Erosion in Venezuela’s Twenty-First Century Socialism”, The ANNALS of
the American Academy of Political and Social Science, 681(1), p.62–77.
Gerschewski, J. (2021) “Erosion or decay? Conceptualizing causes and mechanisms of
democratic regression”, Democratization, 28(1), p.43-62.
Huntington, S. P. (1957). The Soldier and the State: The Theory and Politics of Civil–Military
Relations. Harvard University Press.
Huntington, S. P. (1996). Political Order in Changing Societies. Yale University Press.
Kriesi, H. (2010) “Restructuration of Partisan Politics and the Emergence of a New Cleavage
Based on Values”, West European Politics, 33, p.673-685.
27
Lührmann, A. and Lindberg, I. S. (2019) “A third wave of autocratization is here: what is new
about it?”, Democratization, 26(7), p.1095-1113.
Luján, D. (2020) “Ideological Vote and Electoral Performance of the Bolivian MAS,
2002–2014”, Latin American Politics and Society, 62(4), p.75–97.
Meyer,-Resende M, and Hennig, (2021) “A. Polish Catholic Bishops, Nationalism and Liberal
Democracy”, Religions, 12(2), p.94.
Österman, M. and Robinson, D. (2022) “Educating Democrats or Autocrats? The
Regime-Conditional Effect of Education on Support for Democracy”, Political Studies.
Pavlović, Z.; Todosijević, B. and Komar, O. (2019) “Education, Authoritarianism, and Party
Preference in the Balkans”, Problems of Post-Communism, 66(6), p.402-417.
Regt, S., Mortelmans, D. and Smits, T. (2011) “Left-wing authoritarianism is not a myth, but a
worrisome reality. Evidence from 13 Eastern European countries”, Communist and
Post-Communist Studies, 44, p.299-308.
Reimão, M.E. and Taş, E.O. (2017) “Gender Education Gaps among Indigenous and
Non-Indigenous Groups in Bolivia”, Development and Change, 48, p.228-262.
Roberts, K. (2002) “Social Inequalities Without Class Cleavages in Latin America's Neoliberal
Era”, Studies in Comparative International Development, 36, p.3-33.
Saxonberg, S. (2001). The fall: a comparative study of the end of communism in
Czechoslovakia, East Germany, Hungary, and Poland. Amsterdam, Harwood Academic.
Selçuk, O. and Hekimci, D. (2020) “The rise of the democracy – authoritarianism cleavage and
opposition coordination in Turkey (2014–2019)”, Democratization, 27(8), p.1496-1514.
Sözen, Y (2020) “Popular Will against Democracy: Populist Autocratization in Turkey”,
Journal of Social Sciences, 1(1), 9-29.
Sokolov, B. (2021) “ Measurement Invariance of Liberal and Authoritarian Notions of
Democracy: Evidence From the World Values Survey and Additional Methodological
Considerations”, Frontiers in Political Science, 3.
Sprong, S. et al. (2019) ‘“Our Country Needs a Strong Leader Right Now”: Economic
Inequality Enhances the Wish for a Strong Leader’, Psychological Science, 30(11),
p.1625–1637.
Stubager, R. (2010) “The Development of the Education Cleavage: Denmark as a Critical Case”,
West European Politics, 33(3), p.505-533.
Teik, K.B. (2014) “Technocracy and Politics in a Trajectory of Conflict”, Southeast Asian
Studies, 3(2), p.415–438.
28
Valenzuela, P. A. (2014). “Caída y resurgimiento: La evolución de la oposición política
venezolana durante el gobierno de Hugo Chávez”, Política y gobierno, 21(2), p.379-408.
Velasco, X. ., Hummel, C., Handlin, S., and Smith, A. E. (2021) “Latin America Erupts: When
Does Competitive Authoritarianism Take Root?”, Journal of Democracy, 32(3), p.63-77.
World Bank (2022) “GINI Index for the Plurinational State of Bolivia 1999-2020”.
29