The Social Science Journal
ISSN: 0362-3319 (Print) 1873-5355 (Online) Journal homepage: https://www.tandfonline.com/loi/ussj20
Political freedom, education, and value
liberalization and deliberalization: A cross-national
analysis of the world values survey, 1981-2014
Tony Huiquan Zhang
To cite this article: Tony Huiquan Zhang (2020): Political freedom, education, and value
liberalization and deliberalization: A cross-national analysis of the world values survey, 1981-2014,
The Social Science Journal, DOI: 10.1080/03623319.2020.1727221
To link to this article: https://doi.org/10.1080/03623319.2020.1727221
Published online: 26 Feb 2020.
Submit your article to this journal
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=ussj20
THE SOCIAL SCIENCE JOURNAL
https://doi.org/10.1080/03623319.2020.1727221
Political freedom, education, and value liberalization and
deliberalization: A cross-national analysis of the world values
survey, 1981-2014
Tony Huiquan Zhang
Faculty of Social Sciences, University of Macau, Taipa, Macau, China
ABSTRACT
ARTICLE HISTORY
Since World War II, value liberalization has been a worldwide trend, but more
recently, there has been a resurgence of conservativism. Modernization and
cultural theories have difficulty explaining the shift to liberal or illiberal
values, but the political environment, an underestimated contextual factor,
could shed light on the mechanisms driving it. This study used hierarchical
linear modeling on all six waves of the World Values Survey data (1981–2014)
to demonstrate that political freedom helps to explain the rise and fall of
liberal attitudes when controlling for societal affluence, inequality, and cultural backgrounds. It finds that political freedom conditions the effects of
education: although education is usually considered a liberalizing force, its
effect is much weaker in non-free than free societies. The findings remind us
of the importance of a free political environment for a robust democracy and
point to the complex nature of educational effects, namely the ability of
education to socialize people in liberal or illiberal directions depending on
the political context and the regime’s agenda.
Received 10 April 2019
Revised 27 December 2019
Accepted 30 December 2019
KEYWORDS
Political Freedom;
education; liberal Values;
democratization;
modernization Theory
Introduction
Public opinion is a long-standing topic of discussion in the social sciences, as it has myriad
implications for political and social life in such wide-ranging areas as voting (Dalton, 2009; Norris
& Inglehart, 2019), social policies (Gerhards, 2010; Kwon & Hughes, 2018), trust of others and social
capital (Welzel, 2018), contentious politics (Kensicki, 2001; Torcal, Rodon, & Hierro, 2016), and
democratization (Welzel, 2018). Changes in public opinion, therefore, are a central concern of social
scientists, as they will inevitably have social and political implications. Generally speaking, we have
seen an overall trend towards value liberalization in the post-war world. The establishment of what
Ronald Inglehart and his associates call a “broad syndrome of interrelated values” (Welzel &
Inglehart, 2009, p. 216) has rendered people more participatory, secular, trusting, and tolerant –
in a word, more liberal (Adamczyk, 2017; Andersen & Fetner, 2008; Solt, 2011; Treas, 2002).
Elsewhere, Inglehart uses the term “emancipative values” to describe the contemporary increase in
citizens’ overall support for democracy, human rights, civic participation, and self-expression
(Inglehart & Welzel, 2010; Welzel, 2018). At the same time, and more worryingly, however,
a counter-movement of “value deliberalization” has been taking place, expressed in a global resurgence of conservativism, intolerance, xenophobia, and authoritarianism (Grasso, Farrall, Gray, Hay,
& Jennings, 2019; Inglehart & Norris, 2017; Norris & Inglehart, 2019).
Two main perspectives have been used to explain the shift to liberal attitudes. Modernization
theory argues that economic development and affluence encourage liberalism (Inglehart & Baker,
2000; Welzel, 2018). Meanwhile, cultural theorists believe cultural contexts create change
CONTACT Tony Huiquan Zhang
huiquanzhang@um.edu.mo
Universidade, Taipa, Macau E21-3009, China
© 2020 Western Social Science Association
Faculty of Social Sciences, University of Macau, Avenida da
2
T. H. ZHANG
(Flanagan & Lee, 2003). However, modernization theory and cultural theory face challenges in new
market economies and advanced industrial countries alike (Zhang & Brym, 2019). In the former,
economic growth has failed to yield value liberalization and democratization; in the latter, the once
liberalized attitudes have been shifting back to extremist conservativism in recent years (Koehler,
2016).
An untapped area of inquiry that could shed light on this phenomenon is the context of political
freedom and its effect on attitude liberalization. Public opinion and democratization studies focus
mainly on how political liberalism at the individual level is conducive to democratization and
freedom (Fukuyama, 2006; Lipset, 1959; Saha, 2000; Welzel, 2018), not the other way around. This
study took a new angle and argued that political freedom might influence a number of liberal
attitudes. By doing so, it followed an increasing number of researchers who are paying attention to
the joint effects of institutional and individual characteristics in shaping political beliefs and
behaviors (Andersen & Fetner, 2008; Dalton, Van Sickle, & Weldon, 2010; Kwon & Hughes, 2018;
Zhang & Brym, 2019). Adopting a cross-level perspective, it proposed that political freedom not only
serves as the main effect but also moderates other individual-level predictors’ effects on liberal
attitudes. To probe its hypothesis, it analyzed data from all six waves of the World Values Survey
(hereafter WVS) data (1981–2014), looking at the role of political freedom in attitude liberalization
and asking whether it moderates other factors, notably education.
The study contributes to the literature in two ways. First, it posits political freedom as
a contextual-level explanation of individual-level liberal attitudes, thereby complementing and
adding to economic and cultural explanations. Second, it considers how political freedom could
moderate the educational effects driving attitude change. Previous studies find education is positively
associated with liberal attitudes (Campbell & Horowitz, 2016; Easterbrook, Kuppens, & Manstead,
2016; Ohlander, Batalova, & Treas, 2005; Treas, 2002; Weil, 1985; Zhang, Brym, & Andersen, 2017),
but this study found that more education sometimes means less liberal attitudes, specifically in nonfree societies. In other words, the educational effect is not universal, and this variation could be
partially explained by the political environment. Importantly, the study warns us that in certain
political environments, value deliberalization could occur through agencies of socialization, such as
education systems.
Value liberalization: why political freedom matters
Since the Second World War, despite some fluctuations and exceptions, most societies have become
more liberal in their attitudes to a broad range of topics (Inglehart & Baker, 2000; Welzel, 2018).
Previous studies have identified a number of economic and cultural factors underlying this evolution
(Flanagan & Lee, 2003; Huntington, 1993a; Inglehart & Welzel, 2010). As noted above, modernization theorists emphasize the role of economic development and argue that economic security frees
individuals from worrying about basic needs and stimulates higher-level needs (Inglehart & Baker,
2000; Inglehart & Welzel, 2010; Welzel, 2018). Meanwhile, cultural theorists say cultural backgrounds set the basis for society’s preferences and influence subsequent attitude shifts through pathdependent effects (Flanagan & Lee, 2003; Huntington, 1993a; Inglehart & Baker, 2000; Schwartz,
2006).
Modernization theory and cultural theory are equally plausible, but recent empirical analyses
of many developing countries and emerging market economies challenge the validity of both. For
example, contrary to the economic development theory, researchers have found China and Russia
are not becoming more liberal despite rapid economic growth (Brym, 2016; Zhang, 2018).
Furthermore, China’s upper class is less liberal than its working-class (Zhang et al., 2017).
Similarly, in Turkey, the new middle class benefitting from economic growth has not embraced
democratization, aligning instead with the authoritarian regime (Sarfati, 2017). In short, modernization theory seems to fail to account for the resistance to attitude liberalization in those cases.
THE SOCIAL SCIENCE JOURNAL
3
Cultural explanations face serious challenges from exceptional cases too. Unlike its Confucian
neighbors (Japan, South Korea, Taiwan, Hong Kong), mainland China has not softened its attitudes
to liberalism (Zhang et al., 2017; Zhao, 1998; Zheng, 2015). We see a similar gap between Russia and
its Orthodox neighbors: while many Eastern European and former communist societies have
successfully democratized and embraced a more liberal-democratic system, Putin’s Russia is showing
a resurgence of nationalism and authoritarianism (Brym, 2016; Makarychev & Medvedev, 2015;
Shaykhutdinov, Konitzer, Pacek, & Zvonovksii, 2010). These outliers defy what cultural theory says
about attitude shifts. The recent rise of extremist right-wing ideologies and politics in West Europe
and North America has raised even more questions about existing explanations (Grasso et al., 2019;
Norris & Inglehart, 2019). Although the revival of extremist conservativism could be partially
explained by the grievances of rising inequality, unemployment, surging crime, and the refugee
crisis in the West, these grievances do not completely explain the global trends in attitude
deliberalization.
Arguably, previous work underestimates how political freedom, as an important environmental
factor, could be independently associated with the spread of liberal attitudes. Political freedom has
wide-ranging implications for social mobility, the distribution of power and wealth, and the media’s
messages (Acemoglu, Verdier, & Robinson, 2004; Jowett & O’Donnell, 2014). Public opinion studies
touch on political freedom but typically treat it as a consequence of liberal attitudes (Fukuyama,
2006; Lipset, 1959; Saha, 2000). For example, some researchers have analyzed the way political
contexts affect public opinion on certain policies or issues, such as how political cleavages and voting
systems affect policy preferences (Nir & McClurg, 2015), paying relatively less attention to political
freedom as a factor affecting the attitude shift directly or indirectly.
Yet free and non-free societies may be fundamentally different in attitude formation. To be sure,
freedom of speech is not absolute in relatively free societies; some minorities are comparatively
voiceless, and some media messages dominate. Nonetheless, various ideologies and opinions are
allowed to compete for support in the public sphere, fostering a relatively diverse and tolerant
political culture. By the same token, there may be a more homogeneous and obedient political
culture in non-free societies, and, if so, the agentic role of authoritarian regimes and ruling elites
cannot be ignored. To retain power, authoritarian governments or dictators seek to control information and the mass media (Jiang, 2010; King, Pan, & Roberts, 2013). Voices advocating citizen
participation, individual and collective autonomy, and political reform are muted; voices advocating
respect for established authority and traditional values are encouraged. Instead of promoting
tolerance, authoritarian regimes tend to incite nationalism and xenophobia, in part, to divert
domestic discontent from their own actions (Solt, 2011; Weiss, 2014). They reward citizens handsomely for compliance and punish them harshly for disobedience (Acemoglu et al., 2004). Liberal
attitudes, even those not directly concerning regime legitimacy or public life, such as freedom of
abortion, international adoption (Makarychev & Medvedev, 2015), gender equality (Zheng, 2015),
and sexual minority rights (Zhang & Brym, 2019), are discouraged, as those in power may consider
them potentially dissident, or at least inclining to a mindset of individualism and disobedience.
To sum up, there are many reasons to expect a non-free political environment will discourage
liberal attitudes. These considerations motivate the first hypothesis:
Hypothesis 1: A society’s level of political freedom will be positively correlated with its liberal
attitudes.
Political contexts, educational effects, and value change
As discussed above, a society’s level of political freedom could make a difference in public opinion
and attitude formation. In relatively free countries, people tend to be exposed to diverse opinions
and lifestyles and to be able to speak, assemble, and engage openly in religious and ethnic practices
4
T. H. ZHANG
of their choosing. The political environment can also affect individuals’ attitudes and preferences
through two important socializing tools: mass media and the education system. Scholars from
various disciplines have pointed to the interactions between the political environment and media
(Avraham, 2002; Kensicki, 2001; Skoric, Zhu, & Pang, 2016) and, more relevant for present purposes,
between political context and educational systems. As education is consistently found to be positively
associated with liberal attitudes (Campbell & Horowitz, 2016; Easterbrook et al., 2016; Nir &
McClurg, 2015; Ohlander et al., 2005; Treas, 2002; Weil, 1985), this suggests the possibility of an
interaction between the level of political freedom and individual educational attainment.
Most modern governments have the capacity and possibly the motivation to influence education
policies, school curricula, and investment in education and research (Jowett & O’Donnell, 2014).
Researchers of nationalism and political culture emphasize that modern states strive to foster
a collective mindset among their citizens (Anderson, 2006; Gellner, 1994). Gellner argues that the
complexity of industrialized societies demands they create and maintain a common cultural infrastructure shared by their citizens. State-endorsed formal education may be such an infrastructure.
Formal education institutionalizes the socialization of norms and consensus and is compulsory in
most modern societies (Barber, 2012; Meyer, Ramirez, Frank, & Schofer, 2007).
All else the same, non-democratic regimes have relatively more capacity to exercise this type of
influence than democratic ones because they monopolize state power. And, arguably, they are more
motivated to exercise such influence since they must deal with an inherently higher level of underlying discontent than is typical in democratic regimes. In non-free societies, education is thus an
especially important institution for legitimation, mass persuasion, and regime stability (Brady, 2009;
Jiang, 2010; Saha, 2000; Weiss, 2014).
This possibility is backed up by examples of regime interference with curricula in a number of
comparatively unfree societies. In the former Yugoslavia, for example, the personality cult surrounding Josip Broz Tito was evident in history textbooks glorifying his accomplishments and denigrating
his opponents. Glaring biases were removed only after democratization (Pavasović Trošt, 2014).
Similarly, in reaction to the 1989 Tiananmen Student Protests, post-1989 China initiated a “patriotic
education campaign” to regain legitimacy (Brady, 2009; Zhao, 2004). A major component of the
campaign was the manipulation of higher education curricula. Today, Chinese university students
are required to take courses in political propaganda, including Marxism-Leninism, Maoism, and
Deng Xiaoping Theory (Wang, 2008), and attitudes encouraging protests are strictly suppressed on
Chinese campuses and in the media (King et al., 2013; Radio Free Asia, 2019). Liñán observes similar
attempts by the current Russian regime: “There is a close relationship between the implementation of
centralized (state) education systems and the construction of identities … . Textbooks have been,
and still are, a very valuable vehicle for political propaganda” (2010, p. 270).
These examples suggest that societal-level political freedom not only influences individual-level
liberal tendencies directly; it also interacts with the effect of education. Education has a positive effect
on liberal attitudes (Weil, 1985), but in non-free societies, this effect might be less influential. Hence
this paper’s second hypothesis:
Hypothesis 2: Education’s liberalizing effect on attitudes is greater in free societies than in non-free
societies.
Data and method
WVS data, 1981–2014
Since the research hypotheses concern contextual effects, especially the role of political freedom,
testing them requires cross-national data. This study drew on all six waves of the WVS, spanning the
period 1981 to 2014. The WVS project provides individual-level data on demographic variables and
THE SOCIAL SCIENCE JOURNAL
5
political attitudes and behavior. The six waves cover 99 societies. Since many societies were surveyed
for multiple waves, the data contain 242 country by survey year (hereafter, “country-year”) observations. These societies vary widely in their stages of economic development, levels of inequality,
cultural backgrounds, and political environments, thus offering an excellent opportunity to identify
the effects of political freedom by controlling for other contextual factors.
To analyze the contextual-level effects, the study merged contextual-level data from authoritative
secondary sources with the individual-level WVS data. To deal with concerns about missing
information, it excluded country-year observations with incomplete data on one or more key
country-level variables, i.e., GDP per capita, Gini, or Freedom House Index, as these data should
not be imputed. This procedure left 88 countries with 225 valid country-year observations. The 88
countries and their levels of political freedom are displayed in Figure 1. As the figure shows, the
societies cover a wide range of cases, ranging from less free societies like China and Iran to free
societies such as Finland and Norway.
Though the contextual level data were now complete, there was missing information at the
individual level. The study generated multiple imputations (N = 5) using the chained equation
method1 to minimize this problem. After data cleaning and multiple imputation procedures, around
90% of the original data were retained in the final regression analyses. There was still a small amount
of missing data for certain dependent variables, but this was negligible.2 Statistical estimates were
based on pooled data from the imputed datasets.
Contextual-level information
Previous studies find the level of economic development (Inglehart & Baker, 2000), inequality
(Andersen & Fetner, 2008), and culture zone (Flanagan & Lee, 2003; Huntington, 1993a; Zhang &
Brym, 2019) are significantly associated with liberalism, so this study controlled for these contextual
variables at the level of country-year. For each country-year observation, it collected the following
information: GDP per capita, Gini coefficient, and Freedom House Index for that society in the
corresponding year. This research design, together with the multi-level modeling, was able to capture
both the continuity and the changes within a country over time.3
GDP per capita served as the measure of economic development, with purchasing power parity
figures converted into 2005 US dollars (World Bank, 2015). The study modeled the logged term of
GDP to respond to the skewness of the original distribution. The Gini coefficient of inequality came
from the Standardized World Income Inequality Database (Solt, 2009), which is based on household
disposable income (post-tax, post-transfer), using data from the Luxembourg Income Study. The
Gini coefficients of the selected countries ranged from 17.6 to 60.8, covering a wide range of societies
from very equal to highly unequal. Culture zone was a set of dummy variables that took into account
the constructed cultural traditions and dominant religions suggested by Huntington (1993a) and
Schwartz (2006): (1) Western/West Europe and North America (the reference group), (2) Catholic/
Latin America, (3) Orthodox/East Europe and Russia, (4) Islamic/Middle East and North Africa, (5)
Sub-Saharan Africa, (6) Indian/South Asia, and (7) Confucian/East and South East Asia.
1
Multiple imputation was executed with the R statistical package Amelia II (Honaker, King, & Blackwell, 2011). The Amelia II
package employs EMB (expectation maximization with bootstrapping) methods in estimation. The literature disagrees on the
optimal number of imputations. Hershberger and Fisher (2003) believe five multiple imputations are not enough, and more
(potentially hundreds) is ideal. Yet Von Hippel (2005) argues that using five to ten is more than sufficient and does not cause
considerable loss in precision. The study adopted m = 5 as the final number for multiple imputations. More (m = 10, 20) were
tested, but the differences in their estimates were trivial.
2
After multiple imputations, there was no missing information for the individual-level predictors, such as age, gender, marital
status, occupation, and educational attainment. However, there was some missing information for the dependent variables and
national-level statistics, and these data should not be imputed. Fortunately, the missing rate was not high for the dependent
variables. More details on the numbers of observations and descriptive statistics are in Table 2.
3
For instance, the UK was surveyed in 1998 and 2005. In 1998, the UK had a GDP per capita of 33344.01 dollars (measured in
constant 2005 US dollars), a Gini coefficient of 34.36%, and a Freedom House Index of 1.5. In 2005, the three numbers were
39934.78 dollars, 34.88%, and 1.0, respectively.
6
T. H. ZHANG
Figure 1. Political freedom in the 88 WVS-surveyed societies (data from freedom house, average values used for cases surveyed in
multiple waves).
The focal predictor at the contextual level was political freedom as provided by Freedom House
(2014). Freedom House investigates many aspects of the political environment in each society of
interest, such as free and fair elections, competitive party systems with real opposition, freedom of
association and speech, and so on. After collecting the relevant information, Freedom House
generates two indicators for freedom in a society, one for civil liberties and the other for political
rights. This study took the average of the two indicators and treated the mean value as the Freedom
House Index (hereafter FHI). The FHI ranged from 7 (least free) to 1 (most free), with intervals at
a distance of 0.5 points,4 and was used here as a continuous measure.
Dependent variables
The following four items were selected from WVS to measure liberal attitudes: attitudes to abortion, to
homosexuality, to gender equality in the job market, and to collective action. Attitudes to abortion (Asal,
Brown, & Figueroa, 2008), homosexuality (Adamczyk, 2017; Andersen & Fetner, 2008; Gerhards, 2010;
Ohlander et al., 2005; Zhang & Brym, 2019), gender equality (Bolzendahl & Myers, 2004; Kaufman,
Bernhardt, & Goldscheider, 2017; Schoon, Cheng, Gale, Batty, & Deary, 2010; Van Egmond, Baxter,
Buchler, & Western, 2010), and social movements (Bernburg, 2015; Brym, Godbout, Hoffbauer, Menard,
& Zhang, 2014; Dalton et al., 2010) are common foci for public opinion researchers.
They also have good availability in the WVS dataset. All the variables employed here were
surveyed in all six waves of WVS; other variables of possible interest were either not measured or
were measured differently in one or more waves. As these items concern different aspects of liberal
attitudes, using them allowed the hypotheses to be tested from different angles, thus allaying
concerns about tautology, social desirability biases, or cultural biases. For instance, in some cultures,
4
The coding by Freedom House needs some attention in the interpretation of results, as it uses smaller values to represent more
freedom and larger values to represent less freedom. However, the study did not change the direction of coding to respect the
original coding and to ensure consistency with other studies using the same dataset.
THE SOCIAL SCIENCE JOURNAL
7
respondents could give false answers to questions on their attitudes to homosexuality or abortion, or
in non-democracies, people might give deceptive responses when asked their political opinions.
However, this study’s choice to analyze multiple attitude items should alleviate these concerns. If the
findings emerging from multiple items share the same pattern, they could partially alleviate our
concerns of respondents’ biases, contextual variations, and circular reasoning. Therefore, the discovered pattern could be seen as more convincing and robust.
In terms of the measurement and question wording, the selected variables show consistency
across countries and surveys. In the WVS, questions on attitudes to abortion and homosexuality are
worded similarly. Interviewers ask the respondents, “Please tell me for each of the following
statements whether you think it can always be justified, never be justified, or something in between,
using this card.” The respondents are expected to rate how much they see the relevant behavior as
justifiable on a 1–10 scale, where 1 represents “never justifiable” and 10 “always justifiable.” For
gender equality, however, respondents are asked, “Do you agree or disagree with the following
statements: When jobs are scarce, men should have more right to a job than women”. To follow the
previous coding and ensure comparability, the study recoded their answers as follows: “agree” was 1,
“disagree” was 10, and “neither agree nor disagree” was 5.5.5
A series of questions in the WVS measure support for collective action. The questions ask
whether respondents have “actually done,” “might do,” or “would never, under any circumstances,
do” any of the following: (1) sign a petition; (2) join in a boycott; (3) attend a lawful demonstration;
(4) join an unofficial strike; (5) occupy a building or factory. The study assigned a score of 1 to each
“will never do” response and a score of 10 to each “have done” answer; “might do” received
a midpoint score of 5.5. Based on the five items, the principal component analysis identified only
one eigenvalue greater than 1, suggesting the scale was unidimensional. Cronbach’s alpha for the five
items was 0.75, indicating they were internally consistent.6 Thus, they can be considered as having
high face validity. They also had good predictive validity: they correlated highly and in the expected
direction with measures of other dimensions of liberalism, including questions tapping tolerance of
minority groups, attitudes to freedom of speech, and so on.
The study constructed a 1–10 scale by taking the average value of all five items, with 1 the least
supportive of collective action and 10 the most supportive. This comprehensive measure allowed the
assessment of the degree to which respondents supported freedom of expression and assembly by
inclination and action, something recognized as a key aspect of liberalism (Flanagan & Lee, 2003;
Inglehart & Baker, 2000; Inglehart & Welzel, 2010). Table 1 displays the dependent variables and shows
their details in the WVS data, question wordings, value codings, and means and standard deviations.
After the four dependent variables were properly constructed, the next step was to examine how they
were associated with political freedom at the aggregate level. Figure 2 shows the scatterplots for each item
and political freedom. The x-axes in the figure are the same for all, i.e., the Freedom House Index ranging
from the least free (FHI = 7) to the most free (FHI = 1). The y-axes are the four dependent variables. The
study calculated the average values for each country-year observation. All four scatterplots show
a positive correlation between political freedom and liberal attitudes, thus offering support for
Hypothesis 1 that national freedom and individual support for liberal attitudes are related.
5
For this variable, a multi-level multinomial logistic regression model would be more appropriate. However, to simply the
discussion and table presentation, the study treated it like the other three dependent variables. When fitted with both types
of models, the main findings, especially the interaction effects addressed by this study, were consistent. Details of codes and
results can be requested from the author.
6
To ensure comparability across countries, the study calculated Cronbach’s alpha for the scale of each country-year (Alemán &
Woods, 2016). Of the 225 country-year observations, 178 (79.1 percent) had alpha values at or above the conventional 0.7 cut-off
point (Nunnally, 1978). Another 45 country-year observations had alpha values between 0.6 and 0.69. Only two cases (0.9 percent
of country-years) – Venezuela in 2001 and Tanzania in 2000 – had alphas below 0.6. The study retained the country-year
observations with a low Cronbach’s alpha for the following reasons: first, they represented a small proportion of the
observations; second, a comparison of the regression modelling results before and after the removal of cases revealed only
trivial differences; third, psychometricians consider 0.6 a permissible cut-off for scales with fewer than ten items and items with
fewer than seven response options if the scales are valid and theoretically justified, as was the case here (Loewenthal, 2001).
8
T. H. ZHANG
Table 1. Items and recoding of dependent variables from WVS data (1 = least liberal, 10 = most liberal).
Variable
DV1
Justifiable:
Abortion
DV2
Justifiable:
Homosexuality
DV3
Gender Equality
In Hiring
DV4
Willingness to
Support Collective
Action
Question Wording
Please tell me for each of
the following statements
whether you think it can
always be justified, never
be justified, or something
in between, using this
card.
Do you agree or disagree
with the following
statements:
When jobs are scarce, men
should have more right to
a job than women
Now I’d like you to look at
this card. I’m going to read
out some different forms
of political action that
people can take, and I’d
like you to tell me, for
each one, whether you
have done any of these
things, whether you might
do it or would never,
under any circumstances,
do it.
Items
F120
F118
C001
E025: Petition
E026: Boycott
E027:
Demonstration
E028: Strike
E029: Occupy
Buildings/
Factories
Value Coding
Rate on a 1–10 scale:
Always
Justifiable = 10
Never Justifiable = 1
Rate on a 1–10 scale:
Always
Justifiable = 10
Never Justifiable = 1
Disagree = 10
Neither = 5.5
Agree = 1
Have done = 10
Might do = 5.5
Would never do =
Have done = 10
Might do = 5.5
Would never do =
Have done = 10
Might do = 5.5
Would never do =
Have done = 10
Might do = 5.5
Would never do =
Have done = 10
Might do = 5.5
Would never do =
Mean (SD)
3.42 (2.86)
3.12 (2.99)
5.67 (4.09)
3.04 (2.89)
1
4.80 (3.63)
Averaged
Index
3.52
(2.46)
1
3.90 (3.27)
1
2.48 (2.63)
1
1.71 (1.86)
1
Individual-level variables
On the individual level, studies show that age, gender, marital status, and occupation are significant
predictors of liberalism (Andersen & Fetner, 2008; Bolzendahl & Myers, 2004; Treas, 2002; Zhang,
2018; Zhang et al., 2017). Therefore, these served as the study’s control variables. Gender was
a dummy variable (female = 0, male = 1); age range was 18 to 99 years; marital status was collapsed
into three categories: single/never married (the reference group), married/cohabiting, and widowed/
separated/divorced. The variable of occupation had eight categories: (1) unemployed (the reference
group); (2) student; (3) retired; (4) unskilled manual labor; (5) skilled manual labor; (6) non-manual
office worker; (7) professional; (8) manager/owner. For all the categorical variables, the first category
was used as the reference, and dummy variables were created for the rest.
The focal predictor at the individual level was educational attainment. Education levels were
recoded into five categories: (1) no or little formal education (the reference group); (2) elementary
school completed; (3) middle school completed; (4) high school completed; (5) college degree and
above. In several countries, education was coded as years of formal education. For those countries,
the study recoded years of education as follows: 0–5 (no or little education), 6–8 (elementary school
completed), 9–11 years (middle school completed), 12–14 years (high school completed), 15 years+
(college/university/higher levels).
See Table 2 for a summary of all independent and dependent variables.
Modeling and robustness checks
The study used multilevel models also known as hierarchical linear models (HLM) to analyze both
the individual- and contextual-level effects and to predict respondents’ support for the liberal
THE SOCIAL SCIENCE JOURNAL
9
Table 2. Summary of individual and contextual variables (imputed data).
Variables
N
Independent Variables
Percentage or Mean(SD)
Individual-Level
Male (Female = 0)
Age (18–99)
Marital Status
Never Married (= 0)
Married or Cohabiting
Divorced, Separated or Widowed
Job
Not Employed (= 0)
Student
Retired
Unskilled Manual Worker
Skilled Manual Worker
Non-Manual Office Worker
Professional
Managerial/Owner
Level of Education
None/Little (= 0)
Elementary
Middle School
High School
College or Above
Contextual-Level (by country-year)
GDP per capita (in constant 2005 USD)
Gini Coefficient (0–100)
Culture Zone
Western/West Europe North America (= 0)
Catholic/Latin American
Orthodox/East Europe and Russia
Islamic/Middle East North Africa
African
Indian/South Asia
Confucian/East Asia
Freedom House Index
Dependent Variables (1–10, 10
41
54
34
36
21
7
32
225
= most supportive)
Support
Support
Support
Support
299563
287490
298237
296951
for
for
for
for
Free Abortion
Homosexuality
Gender Equality in Employment
Collective Action
166099
344255
48.25%
40.80 (16.17)
86383
219161
38698
25.09%
63.66%
11.24%
78691
29964
24726
40022
56957
34846
31770
47266
22.86%
8.70%
7.18%
11.63%
16.55%
10.12%
9.23%
13.73%
45174
70846
89306
84422
54494
13.12%
20.58%
25.94%
24.52%
15.83%
225
225
10803.56 (13905.42)
38.45 (9.56)
18.22%
24.00%
15.11%
16.00%
9.33%
3.11%
14.22%
3.13 (1.77)
3.42
3.12
5.67
3.52
(2.86)
(2.99)
(4.09)
(2.46)
attitudes discussed above. In all models, individuals (Level 1) were nested in country-year (Level 2),
and country-year was nested in countries (Level 3). All models included the fixed effects of
individual-level predictors and the intercepts of the waves. They also included random terms for
country intercepts and country-year intercepts to allow across-context variations at both Level 2 and
Level 3. More importantly, all models had a random term for education across contexts, allowing the
effect of education to vary across country-year.
This three-level modeling design can be formally specified as follows: let Yijk be the outcome
variable for the Level 1 unit i (i = 1, …, I), which stands for an individual who lives in a country-year
context, or Level 2 unit j (j = 1, …, J). The Level 2 units are clustered under the Level 3 units, the
country observations k (k = 1, …, K). Then, the outcome variable Yijk, in this case, the support for
liberal attitude, can be expressed as:
Yijk ¼ β0jk þ δXijk þ eijk
The Level 2 model is a decomposition of β0jk, which can be stated as:
(1)
10
T. H. ZHANG
Figure 2. Scatterplots of freedom and liberal attitudes at the aggregate level.
β0jk ¼ β0k þ ejk
(2)
where β0k represents the random intercept denoting the overall group mean in the kth unit in Level 3,
i.e., the countries. β0k can be further decomposed to obtain the following Level 3 model:
β0k ¼β0 þ ek
(3)
where β0 is the overall control group means, and ek is the corresponding residual, assumed to be
independent of Xijk and other residuals.
Table 3 shows the numbers of observations for each model. There were 87 to 88 countries for the
various models at Level 3; at Level 2, there were 212 to 218 country-year observations depending on
how complete the data were for each case. Within each country-year observation, around 1,000
individual respondents were surveyed, meeting the requirements for 30 minimum observations at
higher levels demanded by multilevel modeling for robust estimation (Bryan & Jenkins, 2015;
Stegmueller, 2013).
The statistical models were motivated by the study’s hypotheses. For all four dependent variables,
the study used the following sequence. The first model tested whether political freedom was
a positive predictor of liberal attitudes, as Hypothesis 1 predicted. The next model tested
Hypothesis 2 by including the interaction of freedom and education. The final model included all
contextual effects. By adding GDP per capita, the Gini coefficient, and culture zone, it was possible
to see whether the previous models held up when political freedom was controlled. Modeling results
are displayed in Table 3. Due to space constraints, the table only displays the results from the four
final models for all the dependent variables (Models 1–4).
To ensure the robustness of the analysis, the study took the following steps. First, multicollinearity
could be a concern, as a few variables were correlated (e.g., GDP per capita and Freedom House Index).
Therefore, the study checked the variance inflation factor (VIF) and GVIF, as advised by Fox and
Monette (1992). No high VIF or GVIF was detected in the diagnosis.7 Second, as the WVS data contain
7
Details and codes available upon request.
11
THE SOCIAL SCIENCE JOURNAL
Table 3. Estimates of hierarchical linear models predicting support for liberal attitudes based on WVS data (1981–2014).
Intercept
Individual-level predictors
Age in years (18–99)
Male (Female = 0)
Marital Status (Single = 0)
Married/cohabiting
Divorced/separated/widowed
Occupation (Unemployed = 0)
Student
Retired
Unskilled manual
Skilled manual
Non-manual office
Professional
Manager/owner
Levels of education (None = 0)
Elementary school completed
Secondary school completed
High school completed
College or above
Contextual-level predictors
Freedom House index
GDP per capita (logged)
Gini coefficient
Culture Zones (Protestant/West Europe and
North America as reference)
Catholic/Latin America
Orthodox/Eastern Europe & Russia
Islamic/Middle East & North Africa
African
India and South Asia
Confucian/East Asia
Freedom * education interactions
Elementary school completed
Secondary school completed
High school completed
College and above
Random Components
Level 3: Country intercept
Level 2: Country-year intercept
Educational level
Elementary school completed
Middle school completed
High school completed
College and above
Num. Groups: Country
Num. Groups: Country-Survey Year
Num. Observations
AIC
BIC
Log Likelihood
Model 1 Free
Abortion
Model 2 Tolerance of
Homosexuality
Model 3 Gender
Equality In Hiring
Model 4
Protests
4.77***
0.71
4.59***
2.27*
−0.01***
−0.01
−0.02***
−0.31***
−0.02***
−1.04***
−0.01***
0.36***
−0.04**
0.08***
−0.16***
−0.05**
−0.27***
−0.12***
−0.02
−0.06***
0.10***
−0.03
0.05**
0.13***
0.24***
0.28***
0.24***
0.16***
−0.01
0.09***
0.16***
0.33***
0.37***
0.25***
0.36***
0.19***
0.15***
0.24***
0.40***
0.64***
0.39***
0.23***
0.02
0.14***
0.25***
0.33***
0.48***
0.26***
0.33***
0.75***
1.06***
1.42***
0.42***
0.89***
1.46***
2.14***
0.57***
1.22***
1.60***
2.04***
0.43***
0.91***
1.36***
1.95***
−0.01
0.17
−0.05***
−0.06
0.49***
0.00
0.05
0.34**
−0.01
−0.00
0.22**
−0.01
−0.83**
−0.33
−1.28***
−0.60
−0.46
−0.82*
−0.71*
−1.23***
−1.55***
−0.80
−0.91
−1.21**
−0.31
−1.01*
−2.84***
−0.24
−2.08*
−2.10***
−0.74**
−1.04***
−1.20***
0.07
0.13
−1.22***
−0.06***
−0.12***
−0.17***
−0.19***
−0.11***
−0.21***
−0.32***
−0.42***
−0.09***
−0.16***
−0.20***
−0.20***
−0.06***
−0.12***
−0.18***
−0.21***
0.39
0.30
0.45
0.47
0.72
0.70
0.19
0.34
0.03
0.14
0.21
0.38
88
218
299563
1397529.13
1398017.20
−698718.57
0.09
0.22
0.53
0.88
88
214
287490
1322522.64
1323008.81
−661215.32
0.01
0.23
0.45
0.61
87
212
298237
1606827.49
1607315.35
−803367.75
0.07
0.17
0.36
0.50
87
214
296951
1287229.04
1287716.70
−643568.52
*p < 0.05, **p < 0.01, ***p < 0.001.
a relatively large number of respondents, the significance of some findings may be exaggerated (Lin,
Lucas, & Shmueli, 2013). To ensure the estimated coefficients, confidence intervals, and significance
levels were robust at smaller sample sizes, the study applied the bootstrap resampling method to the
original dataset (5% resample of original data, N = 1000). No evidence countered the main findings.
Third, the study tested the models using alternative measurements of political freedom, such as Polity
IV project’s Combined Polity Score (Marshall, Gurr, & Jaggers, 2017), to replace the Freedom House
Index. Similar patterns appeared, boosting confidence in the findings.
12
T. H. ZHANG
Results
Table 3 displays the estimates from all four models, and Figure 3 visualizes the effects of the focal
term, i.e., the interaction between freedom and education for each item. Model 1 predicts whether
and how much the respondent believes abortion is justifiable. The model shows older, married,
lower-class, and less-educated populations have more negative attitudes to abortion than younger,
unmarried professionals and better-educated people, something also noted in previous empirical
work. Males and females do not show a significant difference in attitudes to abortion. At the
aggregate level, GDP per capita does not show a significant effect, but the Gini coefficient is
negatively associated with acceptance of abortion. Each extra point of increase in the Gini leads to
a drop in acceptance of abortion by 0.05 on a 1–10 scale. The difference in acceptance in a relatively
equal society (assuming Gini = 25, e.g., Finland 2005, Japan 1990) vs. an unequal society (Gini = 45,
e.g., Mexico 2005, Philippines 1996) is approximately 1 point. In other words, inequality is associated
with intolerance of abortion. Culture zones matter greatly for attitudes to abortion. Compared to
West Europe and North America, the Catholic world (−0.83, p < .01), Islamic world (−1.28,
p < .001), and Confucian societies (−0.82, p < .05) all feel negatively about abortion.
For the focal variables, political freedom, and individual educational attainment, higher educational attainment means higher level of acceptance of abortion, as also found in previous work (Asal
et al., 2008). The interaction effects between freedom and education are associated. Note: FHI is
coded in such a way that higher numbers mean less freedom, from 1 (most free) to 7 (least free).
Interestingly, although education is generally considered to be a liberalizing force, it is less so in nonfree societies. Furthermore, the moderating effect is even more salient among those with higher
education: the effect size steadily increases, starting from no education to elementary school (−0.06,
Figure 3. Interaction effect of political freedom and education on liberal attitudes.
Note: Fitted values are from final models. All variables except education and freedom are set to typical values (mean
values for quantitative variables and proportions for categorical variables).
THE SOCIAL SCIENCE JOURNAL
13
p < .001) and moving up through the levels, from middle school (−0.12, p < .001) to high school
(−0.17, p < .001) and university (−0.19, p < .001). The interaction can be seen in Figure 3’s first plot
(upper left), as well. The dash-dot line connecting the black squares stands for the predicted values
for free societies; the regular line in the middle stands for mid-level freedom; the triangles connected
by the dotted line represent the non-free societies. As the figure shows, in the free world, education
brings the acceptance level up from 3.25 (no or little education) to 4.5 (college and above); the effect
is weaker for mid-level freedom, and it is the weakest for non-free societies, where those who go to
college do not differ from those with no education.
Model 2 predicts tolerance of homosexuality. Most findings are similar to those in Model 1 for
abortion: older, married, and unemployed people are more conservative. One noticeable difference at
the individual level is gender, which is no longer indifferent. Males are, on average, 0.31 points lower
than females in their tolerance of homosexuality. Another difference appears for GDP per capita.
A higher GDP per capita is significantly associated with a higher level of tolerance. For example, logged
GDP per capita ranges from 5.25 to 11.14 (difference = 5.89), representing an original value ranging
from 190 USD to 69,095 USD. A 1-unit increase in the logged term is associated with an increase in
liberalism of 0.49 points on a 1–10 scale. That is, controlling for other variables, the richest society is
approximately 2.89 (0.49 * 5.89) points higher on tolerance of homosexuality on a 1–10 scale. The
findings for culture zones are similar to previous findings as well: Catholic, Orthodox, Islamic, and
Confucian societies are more conservative than Protestant ones. African and South Asian societies
show no significant differences from the reference group.
For the focal effect of the interaction between political freedom and education, Model 2 tells a more
impressive story. As Table 3 shows, Model 2’s interaction effects are greater than those found in Model 1,
as also shown in Figure 3 (lower left). Education still plays a liberalizing role in free countries, but in the
non-free countries, represented by the triangles on the dotted line, more education results in less
tolerance of homosexuality. At least in this regard, education plays a deliberalizing role in non-free
societies, contradicting those who claim education liberalizes people across contexts (Andersen & Fetner,
2008; Easterbrook et al., 2016; Meyer et al., 2007; Nir & McClurg, 2015).
Model 3 examines people’s attitudes to gender equality in the job market. By and large, males are
more conservative than females. Seniors, the married, the lower classes, and the less educated are more
conservative as well. GDP per capita is positively related to a gender-equal inclination. In this case, the
role of culture is slightly different: the Catholic world no longer shows significantly lower support than
the Protestant world. However, Orthodox, Islamic, Confucian, and South Asian societies are not big
fans of gender equality in employment. Again, there is an interaction effect, and the pattern shown in
Figure 3 (upper right) is consistent with that found in Model 1 and Model 2.
Finally, Model 4 shows support for collective action. On this item, males are more supportive
than females, which could be interpreted as males’ higher engagement in politics (Verba, Burns, &
Schlozman, 1997) and gendered political socialization (Sherkat & Blocker, 1994). Other predictors
are somewhat similar to previous findings: youth, more education, and better jobs are associated
with more support for collective action. National affluence encourages support of collective action,
but certain culture zones like Latin America, Eastern Europe, the Middle East, and East Asia show
a reluctance to support protests. Figure 3 (bottom right) shows an interaction effect, as well.
Overall, the results point to a consistent pattern: political freedom conditions the effect of
education on liberal attitudes. Only in free societies does education have a strong liberalizing impact
on individuals.
Conclusion and discussion
The major contribution of this study is its systematic investigation of how political freedom matters in
attitude change. Many studies have examined how economic and cultural contexts affect liberal attitudes,
but the political environment is less thoroughly studied. This study’s analysis of cross-national WVS
data suggests political freedom is an important contextual variable closely related to attitude shift. First,
14
T. H. ZHANG
political freedom, as a main effect, is associated with liberal attitudes in and of itself. Second, as
a moderator, political freedom conditions individual-level attitude changes, especially the educational
effects.
These findings have several important implications for the sociology of education, public opinion
studies, and research on democratization. For one thing, they draw attention to the importance of political
freedom, or more generally speaking, political context, in influencing public opinion. Although previous
theories provide useful perspectives, no single theory can fully explain the cross-national variations in
liberal attitudes. For example, both modernization theory and cultural theory fail to adequately deal with
some new market economies, where despite economic success, people are reluctant to embrace liberalism.
In this sense, the theories need to be qualified. This study sheds light on the reluctance: the economy and
culture may no longer be obstacles to liberal attitudes, but the political environment could be.
The findings have methodological implications as well: many previous works are limited in their
case selection. Focusing solely on free societies may prevent scholars from seeing how things work
when there is a lack of freedom. Benefitting from its wide coverage of more than 80 countries and
over 200 country-year observations, this study provides a more thorough answer to the variations in
attitude changes. More work remains to be done, of course, and future comparative analyses should
broaden the scope of examination.
The study also explains deliberalization, something modernization theory cannot do. The finding
of the impact of political freedom partially explains why both developed and developing countries
are shifting towards a less liberal and tolerant political culture. At the aggregate level, shifts in the
regime and leadership could hurt political freedom and this, in turn, could generate a political
environment hostile to liberal attitudes; at the individual level, a regime’s intentions could be realized
through education’s socializing process and foster intolerance in a younger generation. This is
a timely alert, given the resurgence of right-wing extremism in West Europe and North America.
The finding that unfree political contexts hinder the liberalizing effect of education suggests the
importance of political change before attitude change. The apparent success of the “third wave” of
democratization (Huntington, 1993b) encouraged the optimistic, if not naïve, belief that the democratic world should welcome non-democracies such as Russia and China into the World Trade
Organization, nurture their affluence, and expect liberal democracy to follow in due course. This
belief has motivated Western countries, international organizations, and corporations not to push
too hard on the human rights front (Menon, 2015; Ratuva, 2014). However, this study challenges the
view that economic affluence, education, the free market, and exposure to diverse opinions and
lifestyles are sufficient to foster the attitude and behavioral basis required for liberal democracy
(Glaeser, Ponzetto, & Shleifer, 2007; Inglehart & Welzel, 2010; Lipset, 1959). Although these factors
correlate positively with democratization, political context also makes a difference.
Some regimes may energetically resist political reform and social change. As de Mesquita and
Downs note, “Although development theorists are right in assuming that increases in per capita
income lead to increases in popular demand for political power, they have consistently underestimated the ability of oppressive governments to thwart those demands” (2005, p. 78). Similarly,
education will not inexorably bring about liberalization. Political reform is needed to maximize the
liberalizing effect of education and lay the foundation for liberal democracy. Furthermore, only with
actual political participation can liberal attitudes be reinforced (Quintelier & Van Deth, 2014);
democracies provide people with opportunities for participation, while non-democracies try their
best to ensure the opposite (King et al., 2013). In the long run, then, different political contexts may
generate different mind-sets. This possibility suggests the importance of institutional context to
liberal attitudes (Zhang & Brym, 2019) and challenges modernization theory’s optimistic assumption
that we can expect a rising middle class to demand democracy (Fukuyama, 2006).
Admittedly, the study has some limitations. One problem of the WVS data is that they are based
on longitudinal cross-sectional surveys instead of panel data. As a result, it is difficult to argue for
causal links, and there may be concerns about the direction of the proposed mechanisms. After all, as
previous scholars have argued, liberal individuals, liberal educational systems, and liberal regimes
THE SOCIAL SCIENCE JOURNAL
15
have a reciprocal influence (Meyer et al., 2007). A second possible limitation is that the study
proposes a “top-down” mechanism: from the regime to educational systems and then to individuals
(Zhao, 1998). A “bottom-up” mechanism is equally reasonable: liberal citizens may establish
a democratic government and demand a more liberal educational system (Welzel, 2018). In other
words, the study has found an association, but there is a possibility of endogeneity and spuriousness
if it argues for causality.
The researchers acknowledge these limitations and ask future scholars to examine the topic more
closely, using better data and methods. At the present time, however, this may be difficult. Although
panel data are increasingly available, they are usually within a single society or a few societies that are
geographically adjacent. These samples are often similar in economic, cultural, and political contexts;
therefore, they cannot provide enough information for comparative analysis when the research goal is
to reveal contextual effects. In this sense, using the WVS dataset with its a longitudinal and crosssectional data was a trade-off: the study prioritized the need for a wide range of societies over the need
to search for causality. Future scholars equipped with better data sources could do things differently.
The present paper argues for a “top-down” mechanism that political contexts influence individual
values through educational systems. However, the author acknowledged that the reverse causal link of
a “bottom-up” mechanism is possible as well. The risk of reciprocal causation is another limitation in
this paper. However, in the twentieth century, there were quite a few cases that are in favor of the topdown mechanism instead of the counter argument. These are mostly political divisions resulting from
military conflicts, invasions, and occupations, especially during the Cold War. To some extent, these
political divisions can be considered “natural experiments” and examining them can partially alleviate
the concerns of endogeneity. Cases include the divisions between Taiwan and mainland China, North
and South Korea, South and North Vietnam in the 1950–1960s, West and East Germany, and so on.
Before the political divide, these societies were similar in their economic, cultural, and political
environments; at the individual level, people were not that different. However, after the division, the
once similar individuals took different paths simply because they lived on opposite sides of a border –
a border that was often arbitrarily decided by international superpowers. In these cases, changes in
political freedom preceded the educational reforms and individual attitude shifts. A “top-down”
mechanism emerged, whereby once similar people were gradually channeled into diverging routes.
To argue for a “top-down” mechanism is not to refute the “bottom-up” mechanism addressed by
scholars like Christian Welzel (2018); both are valid and important. Whereas Welzel and other
modernization theorists focus on how liberal values bring democracy, this paper notices how nondemocracies prevent value liberalization from happening. The “top-down” and “bottom-up”
mechanisms complement each other. On the one hand, authoritarian governments engage in
propaganda and mass education because they believe liberalized people will seek democracy; this
showcases the “bottom-up” mechanism. On the other hand, this study helps explain why democratization has not yet taken place in many countries: their regimes carry out purposeful efforts to
prevent this. Institutional change is required for value change; this is a “top-down” mechanism.
Finally, this study is timely. Scholars are noting a trend towards value de-liberalization (Zhang,
2018; Zhang & Brym, 2019) and democracy deconsolidation in established democracies (Foa, 2018).
If this trend persists, we might see a cycle of positive feedback, where illiberal tendencies within both
the political system and individuals reinforce each other (Crawford, 2014). We need to be aware of
this danger and take action if our goal is a tolerant, inclusive and liberal world that values gender
equality, the rights of sexual minorities, and freedom of expression.
Highlights
● Political freedom is associated with liberal attitudes;
● Political freedom moderates education’s ostensibly liberalizing effect on attitudes;
● Reduced political freedom and regimes’ agentic control of education systems could lead to
value deliberalization.
16
T. H. ZHANG
ORCID
Tony Huiquan Zhang
http://orcid.org/0000-0002-3587-5910
References
Acemoglu, D., Verdier, T., & Robinson, J. A. (2004). Kleptocracy and divide-and-rule: A model of personal rule.
Journal of the European Economic Association, 2(2–3), 162–192.
Adamczyk, A. (2017). Cross-national public opinion about homosexuality: Examining attitudes across the globe.
Berkeley, CA: University of California Press.
Alemán, J., & Woods, D. (2016). Value orientations from the world values survey: How comparable are they
cross-nationally? Comparative Political Studies, 49(8), 1039–1067.
Andersen, R., & Fetner, T. (2008). Economic inequality and intolerance: Attitudes toward homosexuality in 35
democracies. American Journal of Political Science, 52(4), 942–958.
Anderson, B. (2006). Imagined communities: Reflections on the origin and spread of nationalism. London, UK: Verso
Books.
Asal, V., Brown, M., & Figueroa, R. G. (2008). Structure, empowerment, and the liberalization of cross-national
abortion rights. Politics and Gender, 4(2), 265–284.
Avraham, E. (2002). Social-political environment, journalism practice and coverage of minorities: The case of the
marginal cities in Israel. Media, Culture & Society, 24(1), 69–86.
Barber, B. (2012). Aristocracy of everyone. New York, NY: Ballantine Books.
Bernburg, J. G. (2015). Economic crisis and popular protest in Iceland, January 2009: The role of perceived economic
loss and political attitudes in protest participation and support. Mobilization: An International Quarterly, 20(2),
231–252.
Bolzendahl, C. I., & Myers, D. J. (2004). Feminist attitudes and support for gender equality: Opinion change in women
and men, 1974–1998. Social Forces, 83(2), 759–789.
Brady, A. M. (2009). Marketing dictatorship: Propaganda and thought work in contemporary China. Lanham, MD:
Rowman and Littlefield.
Bryan, M. L., & Jenkins, S. P. (2015). Multilevel modelling of country effects: A cautionary tale. European Sociological
Review, 32(1), 3–22.
Brym, R. (2016). After postmaterialism: An essay on China, Russia and the United States. Canadian Journal of
Sociology, 41(2), 195–212.
Brym, R., Godbout, M., Hoffbauer, A., Menard, G., & Zhang, T. H. (2014). Social media in the 2011 Egyptian uprising.
The British Journal of Sociology, 65(2), 266–292.
Campbell, C., & Horowitz, J. (2016). Does college influence sociopolitical attitudes? Sociology of Education, 89(1),
40–58.
Crawford, N. C. (2014). Institutionalizing passion in world politics: Fear and empathy. International Theory, 6(3),
535–557.
Dalton, R. J. (2009). Economics, environmentalism and party alignments: A note on partisan change in advanced
industrial democracies. European Journal of Political Research, 48(2), 161–175.
Dalton, R. J., Van Sickle, A., & Weldon, S. (2010). The individual–institutional nexus of protest behaviour. British
Journal of Political Science, 40(1), 51–73.
de Mesquita, B. B., & Downs, G. W. (2005). Development and democracy. Foreign Affairs, 84(5), 77–86.
Easterbrook, M. J., Kuppens, T., & Manstead, A. S. (2016). The education effect: Higher educational qualifications are
robustly associated with beneficial personal and socio-political outcomes. Social Indicators Research, 126(3),
1261–1298.
Flanagan, S. C., & Lee, A. R. (2003). The new politics, culture wars, and the authoritarian-libertarian value change in
advanced industrial democracies. Comparative Political Studies, 36(3), 235–270.
Foa, R. S. (2018). Modernization and authoritarianism. Journal of Democracy, 29(3), 129–140.
Fox, J., & Monette, G. (1992). Generalized collinearity diagnostics. Journal of the American Statistical Association, 87
(417), 178–183.
Freedom House. (2014). Freedom in the world 2014. Retrieved from http://www.freedomhouse.org/report/freedomword/freedom-word-2014
Fukuyama, F. (2006). The end of history and the last man. New York, NY: Simon and Schuster.
Gellner, E. (1994). Encounters with nationalism. Hoboken, NJ: Wiley-Blackwell.
Gerhards, J. (2010). Non-discrimination towards homosexuality: The European Union’s policy and citizens’ attitudes
towards homosexuality in 27 European countries. International Sociology, 25(1), 5–28.
Glaeser, E. L., Ponzetto, G., & Shleifer, A. (2007). Why does democracy need education? Journal of Economic Growth,
12(2), 77–99.
THE SOCIAL SCIENCE JOURNAL
17
Grasso, M. T., Farrall, S., Gray, E., Hay, C., & Jennings, W. (2019). Thatcher’s children, Blair’s babies, political
socialization and trickle-down value change: An age, period and cohort analysis. British Journal of Political Science,
49(1), 17–36.
Hershberger, S. L., & Fisher, D. G. (2003). A note on determining the number of imputations for missing data.
Structural Equation Modeling, 10(4), 648–650.
Honaker, J., King, G., & Blackwell, M. (2011). Amelia II: A program for missing data. Journal of Statistical Software, 45
(7), 1–47.
Huntington, S. P. (1993a). The clash of civilizations? Foreign Affairs, 72(3), 22–49.
Huntington, S. P. (1993b). The third wave: Democratization in the late twentieth century (Vol. 4). Norman: University
of Oklahoma Press.
Inglehart, R., & Baker, W. E. (2000). Modernization, cultural change, and the persistence of traditional values.
American Sociological Review, 65(1), 19–51.
Inglehart, R., & Norris, P. (2017). Trump and the xenophobic populist parties: The silent revolution in reverse.
Perspectives on Politics, 15(2), 443–454.
Inglehart, R., & Welzel, C. (2010). Changing mass priorities: The link between modernization and democracy.
Perspectives on Politics, 8(2), 551–567.
Jiang, M. (2010). Authoritarian informationalism: China’s approach to internet sovereignty. SAIS Review of
International Affairs, 30(2), 71–89.
Jowett, G. S., & O’Donnell, V. (2014). Propaganda and persuasion. Newbury Park, CA: Sage Publications.
Kaufman, G., Bernhardt, E., & Goldscheider, F. (2017). Enduring egalitarianism? Family transitions and attitudes
toward gender equality in Sweden. Journal of Family Issues, 38(13), 1878–1898.
Kensicki, L. J. (2001). Deaf president now! Positive media framing of a social movement within a hegemonic political
environment. Journal of Communication Inquiry, 25(2), 147–166.
King, G., Pan, J., & Roberts, M. E. (2013). How censorship in China allows government criticism but silences collective
expression. American Political Science Review, 107(2), 326–343.
Koehler, D. (2016). Right-wing extremism and terrorism in Europe. Prism, 6(2), 84–105.
Kwon, R., & Hughes, E. (2018). Multiculturalist policies in an age of immigration: Do multiculturalist policies
influence negative immigrant attitudes toward homosexuality? Ethnicities, 18(5), 655–691.
Lin, M., Lucas, H. C., Jr, & Shmueli, G. (2013). Research commentary—too big to fail: Large samples and the p-value
problem. Information Systems Research, 24(4), 906–917.
Liñán, M. V. (2010). History as a propaganda tool in Putin’s Russia. Communist and Post-communist Studies, 43(2),
167–178.
Lipset, S. M. (1959). Some social requisites of democracy: Economic development and political legitimacy. American
Political Science Review, 53(1), 69–105.
Loewenthal, K. (2001). An introduction to psychological tests and scales. London, UK: Psychology Press.
Makarychev, A., & Medvedev, S. (2015). Biopolitics and power in Putin’s Russia. Problems of Post-Communism, 62(1),
45–54.
Marshall, M. G., Gurr, T. R., & Jaggers, K. (2017). Polity IV project: Political regime characteristics and transitions,
1800–2016. Center for Systemic Peace.
Menon, A. (2015). Littler England: The United Kingdom’s retreat from global leadership. Foreign Affairs, 94(6),
93–100.
Meyer, J. W., Ramirez, F. O., Frank, D. J., & Schofer, E. (2007). Higher education as an institution. In P. J. Gumport
(Ed.), Sociology of higher education: Contributions and their contexts (pp. 187–221). Baltimore, MD: Johns Hopkins
University Press.
Nir, L., & McClurg, S. D. (2015). How institutions affect gender gaps in public opinion expression. Public Opinion
Quarterly, 79(2), 544–567.
Norris, P., & Inglehart, R. (2019). Cultural backlash: Trump, brexit, and authoritarian populism. Cambridge, UK:
Cambridge University Press.
Nunnally, J. C. (1978). Psychometric theory. New York, NY: McGraw-Hill.
Ohlander, J., Batalova, J., & Treas, J. (2005). Explaining educational influences on attitudes toward homosexual
relations. Social Science Research, 34(4), 781–799.
Pavasović Trošt, T. (2014). A personality cult transformed: The evolution of Tito’s image in Serbian and Croatian
textbooks, 1974–2010. Studies in Ethnicity and Nationalism, 14(1), 146–170.
Quintelier, E., & Van Deth, J. W. (2014). Supporting democracy: Political participation and political attitudes.
Exploring causality using panel data. Political Studies, 62, 153–171.
Radio Free Asia. (2019, January 25). Seven Maoist students detained in Beijing after talking to foreign media. Retrieved
from https://www.rfa.org/english/news/china/maoists-01252019110014.html
Ratuva, S. (2014). A new regional cold war? American and Chinese posturing in the Pacific. Asia and the Pacific Policy
Studies, 1(2), 409–422.
Saha, L. J. (2000). Education and active citizenship: Prospects and issues. Educational Practice and Theory, 22(1), 9–20.
18
T. H. ZHANG
Sarfati, Y. (2017). How Turkey’s slide to authoritarianism defies modernization theory. Turkish Studies, 18(3),
395–415.
Schoon, I., Cheng, H., Gale, C. R., Batty, G. D., & Deary, I. J. (2010). Social status, cognitive ability, and educational
attainment as predictors of liberal social attitudes and political trust. Intelligence, 38(1), 144–150.
Schwartz, S. H. (2006). A theory of cultural value orientations: Explication and applications. Comparative Sociology, 5
(2), 137–182.
Shaykhutdinov, R., Konitzer, A., Pacek, A., & Zvonovksii, V. (2010). Education, the economy, and post-materialist
values in Russia: Evidence from the Chechnya conflict. European Journal of Economic and Political Studies, 3(1),
101–120.
Sherkat, D. E., & Blocker, T. J. (1994). The political development of sixties’ activists: Identifying the influence of class,
gender, and socialization on protest participation. Social Forces, 72(3), 821–842.
Skoric, M. M., Zhu, Q., & Pang, N. (2016). Social media, political expression, and participation in Confucian Asia.
Chinese Journal of Communication, 9(4), 331–347.
Solt, F. (2009). Standardizing the world income inequality database. Social Science Quarterly, 90(2), 231–242.
Solt, F. (2011). Diversionary nationalism: Economic inequality and the formation of national pride. The Journal of
Politics, 73(3), 821–830.
Stegmueller, D. (2013). How many countries for multilevel modeling? A comparison of frequentist and Bayesian
approaches. American Journal of Political Science, 57(3), 748–761.
Torcal, M., Rodon, T., & Hierro, M. J. (2016). Word on the street: The persistence of leftist-dominated protest in
Europe. West European Politics, 39(2), 326–350.
Treas, J. (2002). How cohorts, education, and ideology shaped a new sexual revolution on American attitudes toward
nonmarital sex, 1972–1998. Sociological Perspectives, 45(3), 267–283.
Van Egmond, M., Baxter, J., Buchler, S., & Western, M. (2010). A stalled revolution? Gender role attitudes in Australia,
1986–2005. Journal of Population Research, 27(3), 147–168.
Verba, S., Burns, N., & Schlozman, K. L. (1997). Knowing and caring about politics: Gender and political engagement.
The Journal of Politics, 59(4), 1051–1072.
Von Hippel, P. T. (2005). Teacher’s corner: How many imputations are needed? A comment on Hershberger & Fisher
(2003). Structural Equation Modeling, 12(2), 334–335.
Wang, Z. (2008). National humiliation, history education, and the politics of historical memory: Patriotic education
campaign in China. International Studies Quarterly, 52(4), 783–806.
Weil, F. D. (1985). The variable effects of education on liberal attitudes: A comparative-historical analysis of
anti-semitism using public opinion survey data. American Sociological Review, 50(4), 458–474.
Weiss, J. C. (2014). Powerful patriots: Nationalist protest in China’s foreign relations. Oxford, UK: Oxford University
Press.
Welzel, C., & Inglehart, R. (2009). Political culture, mass beliefs, and value change. In C. W. Haerpfer, P. Bernhagen,
R. F. Inglehart, & C. Welzel (Eds.), Democratization (pp. 126–143). New York, NY: Oxford University Press.
Welzel, C. (2018). Culture and regimes: The democratizing force of emancipative values. In R. A. Scott &
S. M. Kosslyn (Eds.), Emerging trends in the social and behavioral sciences (pp. 1–17). Hoboken, NJ: Wiley.
World Bank. (2015). GDP per capita in constant US dollars. Retrieved from http://data.worldbank.org/indicator/NY.
GDP.PCAP.KD
Zhang, T. H. (2018). Contextual effects and support for liberalism: A comparative analysis. (Doctoral dissertation).
University of Toronto, Toronto, Canada.
Zhang, T. H., & Brym, R. (2019). Tolerance of homosexuality in 88 countries: Education, political freedom, and
liberalism. Sociological Forum, 34(2), 501–521.
Zhang, T. H., Brym, R., & Andersen, R. (2017). Liberalism and postmaterialism in China: The role of social class and
inequality. Chinese Sociological Review, 49(1), 65–87.
Zhao, D. (2004). The power of Tiananmen: State-society relations and the 1989 Beijing student movement. Chicago,
IL: University of Chicago Press.
Zhao, S. (1998). A state-led nationalism: The patriotic education campaign in post-Tiananmen China. Communist and
Post-communist Studies, 31(3), 287–302.
Zheng, W. (2015). Detention of the feminist five in China. Feminist Studies, 41(2), 476–482.