Overstimation of The Level of Democracy

Article
Comparative Political Studies

2023, Vol. 56(2) 228–266
Overestimation of the © The Author(s) 2022
Article reuse guidelines:
Level of Democracy sagepub.com/journals-permissions
DOI: 10.1177/00104140221089647
journals.sagepub.com/home/cps
Among Citizens in
Nondemocracies
Eddy S. F. Yeung 
Abstract
Overestimation of the level of democracy is prevalent among citizens in
nondemocracies. Despite such prevalence, no research to date has sys-
tematically documented this phenomenon and examined its determinants. Yet
given the renewed interest in the role of legitimacy in authoritarian survival,
studying whether and why this phenomenon arises is important to our
understanding of authoritarian resilience. I argue that, even in the absence of
democratic institutions in nondemocracies, autocrats exercise media control
in order to boost their democratic legitimacy. This façade of democracy, in
turn, benefits their survival. Combining media freedom data with individual
survey response data that include over 30,000 observations from 22 non-
democracies, I find that overestimation of the level of democracy is greater in
countries with stronger media control. But highly educated citizens over-
estimate less. These findings shed light on media control as a strategy for
authoritarian survival, and have important implications for modernization
theory.
Keywords
non-democratic regimes, comparative public opinion, media control,
democratic legitimacy, authoritarian resilience
Emory University, Atlanta, GA, USA
Corresponding Author:
Eddy S. F. Yeung, Department of Political Science, Emory University, Tarbutton Hall, 1555 Dickey
Drive, Atlanta, GA 30322-1007, USA.
Email: shing.fung.yeung@emory.edu
Yeung 229
“Authoritarian equilibrium rests mainly on lies, fear, or economic

prosperity.”
— Adam Przeworski (1991, p. 58)
Introduction
In recent years, many citizens in democracies have expressed substantial
dissatisfaction with the functioning of democracy in their country, as signaled
by the rising trend of antiestablishment political movements in established
democracies.1 This contrasts with the curious observation that many citizens
in nondemocracies2 have displayed considerable satisfaction with the au-
thoritarian governance in their country, as indicated by their widespread
acknowledgment of autocratic leaders (Frye et al., 2017, Guriev & Treisman,
2020b).
This suggests a possible misalignment between the perceived level of
democracy and the measured level of democracy. That is, people may un-
derestimate the level of democracy in democracies, but overestimate it in
nondemocracies.3 Data from the World Values Survey (WVS) reveal this
interesting pattern (Figure 1). While underestimation of the level of de-
mocracy is prevalent in established democracies such as the U.S. and Japan,
overestimation is evident in nondemocracies such as China and Singapore.4
Similar phenomena can also be observed in other countries. However, this is
unlikely due to how individuals in democracies and nondemocracies define
democracy differently, since their conceptions of democracy are, in fact, fairly
similar (Figures S1–S3). This leads to an unsolved puzzle: Why do citizens in
nondemocracies overestimate the level of democracy in their country?
While underestimation of the level of democracy in democracies is closely
linked to the abundant literature on democratic deficit,5 surprisingly no re-
search to date has focused on the overestimation side. Yet, studying over-
estimation is also important: If we study democratic deficit due to its
implications for democratic survival, then we should also study why citizens
overestimate regime democraticness given its potential implications for au-
thoritarian resilience.
In nondemocracies, citizens’ overestimation of the level of democracy
contributes to authoritarian resilience through regime legitimation. This is
premised on congruence theory in the democratization literature: Autocracies
gain legitimacy when their citizens perceive them to be democratic; such le-
gitimacy, in turn, empowers autocrats to not “supply” democracy as citi-
zens believe that their “demand” for democracy is already satisfied (Inglehart &
Welzel, 2005; Qi & Shin, 2011; Rose et al., 1998). Consequently, auto-
crats are incentivized to create a façade of democracy to stabilize their re-
gime (Gerschewski, 2013; 2018; Tannenberg et al., 2021). But how is a
nondemocracy able to make this façade of democracy so credible that its
230 Comparative Political Studies 56(2)
Figure 1. Perceived and Measured Levels of Democracy—All Sampled Countries in

the World Values Survey (Wave 6: 2010–16).
Note: Variables are rescaled to range from 0 to 1. Data on perceived levels of democracy are
obtained from the WVS (Wave 6: 2010–16). They are country-average data after dropping
nonrespondents, based on the variable V141 in the WVS. Data on measured levels of democracy
are obtained from V-Dem’s Electoral Democracy Index (corresponding years). Details on
these variables, as well as a discussion of their comparability, are provided in Section 4.
citizens are willing to believe that their country is already democratic when it
is not?
I argue that media control is an important strategy.6 By controlling the
media, autocrats can selectively disseminate information that could mislead
citizens into believing that the regime is already democratic. This could
include, for example, the occasional instances where popular policy demands
are aptly responded to by the government (Truex, 2017), as well as the “rule of
law” development within the state (Stockmann & Gallagher, 2011; Whiting,
2017). Crucially, media censorship can cover up common autocratic practices
of the government, including violent repression and electoral fraud. Unable to
obtain information about such regime-revealing practices, ordinary citizens in
nondemocracies may in turn overestimate the democraticness of their country.
Yeung 231
Thus, media control provides a powerful tool for autocrats to create a

democratic façade.
But not all citizens are gullible: Those who are highly educated are not
easily manipulated. They tend to be more resistant to political messages
disseminated by the media (Geddes & Zaller, 1989; Kennedy, 2009; Moehler &
Singh, 2011; Zaller, 1992) and more resilient to media censorship than the
general populace (Roberts, 2018, 2020). Therefore, I additionally argue that
highly educated citizens overestimate the democraticness of their country to a
lesser degree.
Combining media freedom data with individual survey response data that
include over 30,000 observations from 22 nondemocracies, I find suggestive
evidence for my arguments. The cross-national analysis reveals that over-
estimation of the level of democracy is greater in autocracies with stronger
media control. Highly educated citizens, however, overestimate less than their
less educated peers do. Although causal relationships are harder to prove, the
results withstand a battery of theoretically and empirically motivated ro-
bustness checks.
To my knowledge, this paper is the first to systematically document the
prevalence of overestimation of the level of democracy and examine its
determinants. In doing so, I make three distinct contributions. First, I enrich
the emerging literature on informational autocracies (Guriev & Treisman,
2019). Nowadays, autocrats rely much less on violent repression and official
ideologies than their predecessors did; instead, they demonstrate strong ea-
gerness to manipulate information by controlling the media (Gunitsky, 2015;
Guriev & Treisman, 2019). This empirical observation calls for scholarly
attention: Why is information manipulation so attractive to autocrats? How
does it benefit their survival? This study offers a new perspective by sug-
gesting that media control can increase citizens’ tendency to overrate the
democraticness of their country. This helps autocrats build democratic le-
gitimacy that relieves their pressure to democratize.
Second, I contribute to the revitalized literature that brings back legitimacy
to discussions of authoritarian resilience. Scholars have invested considerable
but rewarding effort in the study of strategic cooptation and repression in
autocracies, yet less is known about how legitimation may contribute to
authoritarian survival. The question of legitimacy, therefore, becomes “a
known unknown for autocracies” (Gerschewski, 2018, p. 661). Here, I relate
two strands of literature—legitimacy and authoritarian resilience—by sug-
gesting that part of the reason autocracies survive is that their citizens are
misled into believing that they are already living in a democracy; hence their
low demand for democracy. Thus, democratic legitimacy is important not only
for democratic survival (Lipset, 1959; Chu et al., 2008) but also for au-
thoritarian survival.
Third, I expand the literature on media control. Scholars of authoritarian

politics often argue that the media play an important role in autocracies.
However, this body of work—predominantly and understandably—focuses
on China (e.g., Chen & Yang, 2019; Huang & Yeh, 2019; King et al., 2013;
2017; Lorentzen, 2014; Roberts, 2018; Stockmann and Gallagher, 2011;
Wang & Huang, 2021) and Russia (e.g., Enikolopov et al., 2011; Reuter &
Szakonyi, 2015; Rozenas & Stukal, 2019).7 This calls into question the
relevance of their findings to other autocracies. I fill this gap by studying
media control cross-nationally, with a dataset that includes not only China and
Russia but also 20 other nondemocracies from all parts of the world. My
extended analysis shows that the media finding still holds even after dropping
China and Russia from the sample.
The remainder of this paper is structured as follows. Section 2 reviews the
relevant literature and argues that autocrats are incentivized to use media
control to create a façade of democracy. Section 3 introduces the theoretical
framework and hypotheses. Section 4 defines the variables and describes their
sources, followed by the results (Section 5) and robustness checks (Section 6).
Section 7 concludes.
Perceived Democraticness, Legitimacy, and

Authoritarian Survival
Misalignment between perceived and measured levels of democracy
Perceived levels of democracy often fall short of public expectations in
democracies. Democratic deficits are found not only in the European Union
(EU) (Majone, 1998) but also in other parts of the world (Norris, 2011). Many
EU citizens perceive themselves to be underrepresented under the EU gov-
ernance and are therefore skeptical about the democratic representation of the
EU (Rohrschneider, 2002). More generally, Dalton (2004) documents the
erosion of political support in advanced industrial democracies, where public
trust in politicians and governments were declining substantially. Norris
(2011) further suggests that this phenomenon is not only visible in estab-
lished democracies but also in younger democracies, especially in post-
communist states such as Bulgaria and Serbia. Recent data from the 2020
Democracy Perception Index reveal that two-fifths of people living in de-
mocracies believe their country to be undemocratic (Dalia, 2020).
If misalignment between the perceived and measured levels of democracy
implies that people can underestimate the level of democracy in democracies,
then it is also plausible that people can overestimate it in nondemocracies.
Although the existing literature does not pay close attention to the overes-
timation side, this phenomenon is indeed prevalent in nondemocracies
(Figure 1). For instance, the WVS data suggest that the average perceived
Yeung 233
levels of democracy in China and Singapore were 6.44 (in 2013) and 6.87 (in
2012) on a ten-point scale, which were comparable to the average perceived
levels of democracy in the U.S. (6.40 in 2011) and Japan (6.72 in 2010). This
phenomenon is even more noticeable in Kazakhstan and Slovenia. In 2011,
Kazakhstan was rated by V-Dem’s Electoral Democracy Index with a score of
0.24 on a unit interval, while Slovenia was given a score as high as 0.86. Yet,
the average perceived level of democracy in Kazakhstan (6.84) was about 2.2
points higher than that in Slovenia (4.65) on a ten-point scale. The mis-
alignment between the perceived and measured levels of democracy is even
more pronounced when we focus squarely on nondemocracies (see Figure 2,
which shows a negative relationship within a sample of nondemocracies). In
short, there are huge discrepancies between the perceived and measured levels
Figure 2. Perceived and Measured Levels of Democracy—Sampled Nondemocracies

in the WVS.
Note: Variables are rescaled to range from 0 to 1. Data on perceived levels of democracy are
obtained from the WVS (Wave 6: 2010–16). They are country-average data after dropping
nonrespondents, based on the variable V141 in the WVS. Data on measured levels of democracy
are obtained from V-Dem’s Electoral Democracy Index (corresponding years). Details on
these variables, as well as a discussion of their comparability, are provided in Section 4.
Nondemocracies are identified using V-Dem’s Regimes of the World (v2x_regime).
of democracy across different types of regimes and different regimes of the

same type.
Past research suggests that perceived levels of democracy can be affected
by multiple factors. A cross-national survey conducted by the Pew Re-
search Center suggests that they are positively correlated with perceived
economic performance (Wike et al., 2017). While the national economic
performance may be important in terms of influencing perceived demo-
craticness in less affluent countries, political factors—in the form of gov-
ernance effectiveness—appear to be more relevant in more affluent countries
(Rohrschneider & Loveless, 2010). Perceived democraticness also depends on
people’s conceptions of democracy: Individuals have various interpretations
of the meaning of democracy, and these varying interpretations, in turn, shape
their perceptions of how democratically governed their country is (Miller
et al., 1997). This may partially explain why they have contrasting views on
the level of democracy in their own country even under the same governance.
Mattes and Bratton (2007) also find that people appear to form such con-
ceptions by learning about what democracy is and does. They argue that
people learn about “the content of democracy through cognitive awareness of
public affairs” and “the consequences of democracy through direct experience
of the performance of governments and (to a lesser extent) the economy”
(p. 192). This suggests the manipulability of people’s perceptions on the
democraticness of their country.
Legitimacy and authoritarian survival

But why do autocrats want to mislead their citizens into thinking that their
regime is democratic when it is, in fact, undemocratic? The reason probably
lies in the importance of legitimacy in their survival. On one hand, popular
support for the regime is conducive to regime survival as it reduces mass
threats; on the other, legitimation complements cooptation and repression by
reducing the persuasion costs for coopting elites and by reducing potential
opposition (Gerschewski, 2013). Indeed, recent work has shown that regime
legitimation strategies are frequently adopted by autocrats, who emulate
democratic procedures to generate legitimacy of their rule (Tannenberg et al.,
2021; Von Soest & Grauvogel, 2017).
Autocrats also use the language of democracy to legitimate their rule.
Analyzing the political speeches of world leaders, Maerz (2019) finds that
leaders from nondemocracies are as likely as leaders from democracies to
speak of democratic procedures in public discourses. For example, Islam
Karimov, the former president of Uzbekistan, “constantly stresses in his
speeches the regime’s apparent aim of building a democratic state, intensi-
fying the democratic transformation and ‘consolidating democratic values
in the minds of the people’” (p. 13). Similarly, his successor, Shavkat
Yeung 235
Mirziyoyev, claimed that he strove to build “a free, democratic, humane state,”

and that his electoral victory was a result of “the elections [which] were
conducted in the atmosphere of sound competition and struggle among
political parties” (p. 14). Most recently, Sadyr Japarov, the newly elected
president of Kyrgyzstan, described himself as a “democratic person” upon his
electoral victory (The Economist, 2021). What these examples show is that
autocrats often value and pursue democratic legitimacy.
Congruence theory further elucidates why legitimacy relieves democra-
tization pressure. The theory is a simple demand–supply framework, which
holds that democratization takes place when public demand for democracy in
a state outstrips its institutional supply of democracy (Dalton & Shin, 2006;
Inglehart & Welzel, 2005; Mattes & Bratton, 2007; Qi & Shin, 2011; Rose
et al., 1998). If congruence between public demand and institutional supply is
attained, then regime change is highly unlikely. Yet on the supply side,
autocracies—by definition—lack democratic institutions. The differentiation
then lies on the demand side: Only autocracies with sufficiently high public
demand for democracy will face democratization pressure. When autocrats are
able to mislead their citizens into thinking that their regime is already
democratic, democratic legitimacy is achieved even in the absence of dem-
ocratic institutions. Under such circumstances, even if citizens support de-
mocracy in principle,8 their demand—understood as the discrepancy between
support for democracy-in-principle and dissatisfaction with democracy-in-
practice (Qi & Shin, 2011)—is still low because they have strong “practical
support for democracy-in-practice” (p. 247). Congruence theory, therefore,
illustrates why some autocrats are less pressured into democratization and
how legitimacy can be conducive to authoritarian survival.
This theory is especially relevant in electoral autocracies. In these regimes,
citizens’ cost of driving democratization is relatively low: They can simply
vote out the autocrats in favor of pro-democracy candidates in general
elections. Nonetheless, congruence theory may be less applicable to autoc-
racies without electoral institutions. In these regimes, citizens’ demand for
democracy is not easily translated into democratization efforts, since social
movements are often costly to citizens and ineffective due to collective action
problems. Thus, instead of suggesting that all types of autocrats are incen-
tivized to mislead their citizens into thinking that their regime is democratic, I
posit that such incentives are particularly pronounced among autocrats in
electoral autocracies.
Media control as a strategy in autocracies

Autocrats may be incentivized to create a façade of democracy, but how do
they mislead their citizens? One way to do so, I argue, is through media
control. The argument is premised on the strand of research that uses media
control to explain the high level of public support or regime legitimacy in

autocracies, particularly in China. Li (2004), for example, argues that the lack
of free media in China enhances people’s trust in the Chinese Communist
Party (CCP). Kennedy (2009) also suggests that the Chinese government’s
strict control over the media plays a vital role in maintaining a high level of
popular support in China. Studying how the Chinese government uses dif-
ferent means of media control to promote its “rule of law” reforms, Stockmann
and Gallagher (2011) argue that the Chinese media strongly shape public
opinion on the CCP’s legitimacy, thereby contributing to its regime stability.
In particular, they suggest that the private Chinese media, which “fully
conforms to state censorship demands” (p. 459), played an important role in
making citizens believe that the “rule of law” reforms were well implemented
in China (see also Whiting, 2017). Overall, these studies show that China
effectively uses media control to consolidate regime support, even in the
absence of democratic institutions and the presence of citizens’ support for
democracy-in-principle.9
In their recent contribution, Guriev and Treisman (2019) introduced a new
concept: informational autocracy. They argue that, instead of relying on
ideologies and repression to survive, these regimes survive by manipulating
information. Informational autocrats lead citizens to believe that their regimes
are legitimate, and have a high tendency to conceal rather than publicize cases
of state brutality. They have become increasingly common in recent years
(Guriev & Treisman, 2019). In another important contribution, Guriev and
Treisman (2020b) find that autocracies with greater media censorship enjoy
stronger popular support than those with less censorship. Formalizing with a
game-theoretic model, they further show that information manipulation
prevails over overt repression when the informed elite is sufficiently large,
thereby underscoring autocrats’ incentives to censor and coopt media for their
survival (Guriev & Treisman, 2020a).10 Given the suggested importance of
media control for authoritarian survival (see also Gunitsky, 2015), perhaps it is
unsurprising that media freedom is higher in democracies than in autocracies
(Stier, 2015).
Theoretical Framework and Hypotheses

I propose two hypotheses to explain why citizens in nondemocracies over-
estimate the level of democracy in their country. They concern media control
and education.
As argued in the preceding section, media control helps autocrats mislead
their citizens into believing that their country is democratic. On one hand,
suppression of media freedom makes the public poorly informed of the
undemocratic practices engaged in by the autocrat (Xu et al., forthcoming).
With such public ignorance resulting from a lack of information, citizens may
Yeung 237
then overrate the democraticness of their country. On the other hand, if the
media were free, citizens would likely learn from the independent media
about, for example, the procedural irregularities of their country’s elections,
thereby realizing that their country is not truly democratic with free and fair
elections (Kerr & Lührmann, 2017). This was the case for the democratization
experiences of Georgia, Serbia, and Ukraine, where the independent media
played a key role in shaping public opinion and strengthening the opposition
force by disseminating information that revealed the undemocratic practices
of the regime (McFaul, 2005, pp. 11–13).
The revelation of—and citizens’ realization of—electoral manipulation can
have profound implications for authoritarian survival. For instance, Lesta
(2019) finds from electoral autocracies in Africa that citizens living in op-
position strongholds are much less likely to perceive their country to be
democratic and legitimate. She attributes this finding to the dissemination of
“information about the authoritarian nature of the government” by opposition
parties (p. 1).11 In addition, Reuter and Szakonyi (2021) find experimental
evidence from Russia that many citizens, who genuinely believe that elections
in Russia are free and fair, would withdraw their support for the ruling party if
they were informed that its candidates committed electoral fraud. These
studies show a clear incentive for autocrats to exercise media control: By
restricting citizens’ access to regime-revealing information, they fortify the
democratic façade and thus consolidate their regime.
Zaller’s (1992) exposure–acceptance model furthers our understanding of
how media control helps autocrats create a democratic façade. Stockmann and
Gallagher (2011) succinctly state the model: “a person’s likelihood to be
persuaded by a piece of information depends on two factors: first, his or her
likelihood to be exposed and comprehend the message (reception) and,
second, his or her likelihood to accept the message (acceptance)” (p. 451).
In nondemocracies where the government exerts tighter control over the
media for political messaging, citizens are more exposed to (dis)infor-
mation that is intended to make them believe that their country is already
democratic or prevent them from believing that their country is undem-
ocratic. For instance, the Chinese government fabricates hundreds of
millions of social media posts a year to distract public attention from
skeptics of the CCP (King et al., 2017).12 Such media control enables the
autocrat to choose what kind of information to be—and not to be—exposed
to citizens. This effectively manipulates the reception side and, conse-
quently, prohibits citizens from having sufficient information to detect the
undemocratic nature of the regime. Thus, citizens living in non-
democracies with highly state-controlled media should overestimate the
level of democracy in their country more than those living in non-
democracies with freer media:
H1 (media control): A higher level of media control in nondemocracies in-

creases the extent to which citizens overestimate the level of democracy in their
country.
Yet, the exposure–acceptance model suggests that acceptance of the message

also matters. It is unreasonable to assume that all citizens in nondemocracies are
susceptible to political messaging. Naturally, we would expect that highly
educated citizens are less gullible, since they tend to be more well-informed
about the current political conditions in their country (Grönlund & Milner,
2006). They also tend to have a higher demand for political information and a
greater capacity to overcome media censorship (Chen & Yang, 2019; Roberts,
2018, 2020). Indeed, past research has suggested that education “increase[s] the
quality of civic knowledge” (Dee, 2004, p. 1697) and that “higher education
imparts the knowledge, skills, and political familiarity that help in navigating the
political world” (Hillygus, 2005, p. 27). Such knowledge and skills also allow
the highly educated to “follow international events more closely and place their
own situation within the context of world events” (Sanborn & Thyne, 2014, p.
781). Consequently, they tend to be more critical of state-controlled media
(Moehler & Singh, 2011), especially in the shadow of globalization that fa-
cilitates international flow of information (Sanborn & Thyne, 2014).
Relatedly, scholars have applied the exposure–acceptance model to explain
why highly educated citizens tend to be least supportive of their government in
nondemocracies. The key is their resistance to political messages: They are highly
exposed to political messages disseminated by the state-controlled media as they
are politically aware, yet they do not easily accept such messages—unlike their
less educated counterparts. Such dynamics between media control and higher
education have been found not only in contemporary China (Kennedy, 2009) but
also in authoritarian Brazil (Geddes & Zaller, 1989). These findings further
suggest that highly educated citizens are more able to reject state propaganda.
Indeed, many democratic movements have been organized by college-
educated students. One example is the 1989 Tiananmen Square protests in
China, where the severe political corruption and public aspirations for Western
democracy (among other factors) triggered a sizable group of Chinese stu-
dents to call for democracy in China (Wong, 1990; Zhu & Rosen, 1993).
Another example is the Carnation Revolution in Lisbon, which was catalyzed
by college students and led to Portugal’s democratization in 1974 (Accornero,
2016). A more recent example is the Arab Spring of 2011, where the ex-
pansion of education played an important role (Campante & Chor, 2012).
Explaining the rigid opposition to the Soviet regime among highly educated
citizens, Roeder (1989) demonstrates that the most educated were also the
most resistant to state propaganda: They expressed the highest disagreement
with “official interpretations of events reported in the paper” and were most
critical of “the informational content of official publications” (p. 871).
Yeung 239
Higher education additionally promotes political participation and expands

social networks, which are also conducive to individual access to genuine
information about the regime. Dahlum and Wig (2021) demonstrate that
university-related protests “are more likely to emerge in dictatorships and that
protests in university locations are more likely to concern democracy and
human rights” (p. 3, emphasis added). In another contribution, they show that
“education increases the frequency of mass protest, by alleviating collective-
action problems and motivating mass opposition, particularly in autocracies”
(Dahlum & Wig, 2019, p. 3). By actively participating in politics, highly
educated citizens can expand their social networks and acquire new infor-
mation through exchanging discourses with new social contacts (De Micheli,
2021). In turn, they may update their preexisting beliefs accordingly (Mische,
2008; Munson, 2010). Thus, the highly educated—who are also more active
and networked in politics (see also Campbell, 2013; McClurg, 2003)—should
have better access to genuine information from their peers about the electoral
process in their country.
In short, I argue that education produces the fundamental knowledge and
social networks that make highly educated citizens less susceptible to media
control. This leads to the following hypothesis:
H2 (education): Highly educated citizens in nondemocracies overestimate the

level of democracy in their country to a lesser degree.
Given H1 and H2, one may raise the following concern: What if the
public—including less educated citizens—observed that the government
exercised media control?13 Wouldn’t such observations make them view their
government as less democratic? This concern is valid, which is why autocrats
do not exercise media control overtly nowadays (Roberts, 2018). Guriev and
Treisman (2019) argue that informational autocrats who exercise media
control conceal censorship from the public by adopting less obvious tech-
niques. Here are some examples:
· In Singapore, Lee Kuan Yew used to coopt corporate boards of key

media outlets, which engaged in censorship for him.
· In Hungary, Viktor Orbán imposed an advertisement revenue tax in
order to make critical media outlets less resourced and more vulnerable
to takeovers by government allies.
· In Russia, Vladimir Putin uses surrogates and economic pressure to
control the media.
· In Peru, Alberto Fujimori bribed most private media.
Given such discreet means of media control, it is often difficult for citizens
in nondemocracies to detect censorship. In the rare case that citizens are able
to observe media censorship, they are usually the highly educated ones
(Guriev & Treisman, 2019). These individuals, in turn, become less sup-
portive of the incumbent regime as they dislike being deceived (Guriev &
Treisman, 2020b). This therefore renders further support to H2, while leaving
H1 generally untarnished.
A bigger threat to H1 and H2 is political orientation.14 Robertson (2017)
shows that, due to motivated reasoning, “citizens’ underlying political ori-
entations affect both the kind of information they gather and how they process
that information” (p. 589). Shirikov (2021) further argues that confirmation
bias, in addition to motivated reasoning, can make existing supporters of the
political leader more susceptible to media manipulation. Reuter and Szakonyi
(2021) also demonstrate that partisanship plays an important role in shaping
individual perceptions about the electoral process in their country: Self-
identified supporters of the ruling party tend to believe that their country
has fairer elections and is more democratic. Therefore, citizens’ access to
regime-revealing information—be it facilitated by freer media or better
education—does not unconditionally change their perceptions of the dem-
ocraticness of their country. Their political predispositions may limit their
ability to correct misconceptions about the state of their government. This
additional nuance poses a threat to H1 and H2, leaving them an empirical
question that remains to be tested.
Data and Methods

To test the hypotheses, I construct a unique dataset that includes the following
variables:
· Dependent variable: an ordinal variable indicating the degree to which

an individual in a nondemocracy overestimates the level of democracy
in their own country.
· Key independent variables: media freedom of various nondemocracies;
and individual education levels.
· Control variables: an array of individual characteristics, including sex,
age, marital status, employment status, income, and social class; and
economic performance, rule of law, and government effectiveness of
various nondemocracies.
To construct the dependent variable, Overestimation, I first obtain indi-

vidual perceived levels of democracy from the WVS (Wave 6: 2010–16),15
followed by obtaining the measured levels of democracy of different regimes
from V-Dem over the period 2010–16 (see next subsection for more details).
To construct the variable Media Freedom, I obtain data from Freedom House’s
(2017) Freedom of the Press, and separately from Solis and Waggoner’s (2021)
Yeung 241
Media System Freedom, over the period 2010–16. I construct the variable
University Degree and control variables on individual characteristics by
obtaining the demographic data from the WVS.16
In constructing the dataset, I first exclude individuals who did not provide
their responses or provided a response of “Don’t know” regarding the per-
ceived level of democracy in the WVS. I subsequently exclude individuals
whose countries are classified as democracies by V-Dem’s Regimes of the
World (v2x_regime).17 Additionally, I exclude individuals from Hong Kong
and Palestine due to their data unavailability in Freedom of the Press. This
leaves us with a total of 31,742 individual observations from 22 non-
democracies. The 22 nondemocracies are Algeria, Armenia, Azerbaijan,
Belarus, China, Egypt, Haiti, Iraq, Jordan, Kazakhstan, Libya, Malaysia,
Morocco, Nigeria, Pakistan, Russia, Rwanda, Singapore, Thailand, Ukraine,
Yemen, and Zimbabwe.18 With the exception of China, these are all electoral
autocracies. This is particularly useful for my theory testing, since I posit that
the theory is especially relevant to electoral autocracies.
Dependent variable
The dependent variable, Overestimation, is an ordinal variable indicating the
degree to which an individual overestimates the level of democracy in their
country. Data on perceived levels of democracy are retrieved from the variable
V141 in the WVS, which asks: “How democratically is this country being
governed today?” The options range from 1 (“Not at all democratic”) to 10
(“Completely democratic”) and include “Don’t know.” On the other hand,
data on measured levels of democracy are retrieved from V-Dem’s Electoral
Democracy Index (v2x_polyarchy) (Coppedge et al., 2021; Teorell et al.,
2019). V-Dem rates the democraticness of each regime by year on a scale of 0
(least democratic) to 1 (most democratic), based on the extent to which the
ideal of electoral democracy in its fullest sense is achieved. It is largely a
procedural measure of democracy that captures five components, namely
“Elected Officials, Clean Elections, Associational Autonomy, Inclusive
Citizenship, and Freedom of Expression and Alternative Sources of Infor-
mation” (Teorell et al., 2019, p. 71).
To achieve numerical comparability between perceived and measured
levels of democracy, I rescale v2x_polyarchy so that it also ranges from 1 to
10, followed by rounding it off to the nearest integer. The dependent variable,
Overestimation, is then constructed by subtracting the rescaled v2x_polyarchy
of a country from an individual’s perceived level of democracy in that country.
For example, a Russian respondent who rated the democraticness of their
country at 6 would overestimate the level of democracy by 2 points (i.e.,
Overestimation = 2), since the rescaled v2x_polyarchy for Russia was 4 in the
corresponding year. If a Singaporean respondent rated the democraticness of
their country at 2, they would underestimate the level of democracy by 3

points (i.e., Overestimation = 3), since the rescaled v2x_polyarchy for
Singapore was 5 in the corresponding year. In the dataset, respondents
overestimated the level of democracy in their country by 1.55 points on
average (Table 1). Figure 3 shows the distribution of Overestimation.
Some potential measurement problems must be noted, however. The first
problem concerns the meaningfulness of directly comparing the WVS item
with V-Dem’s index: Citizens may differ from expert coders in their con-
ceptions of democracy. In the appendix, I show that most citizens in non-
democracies actually view free elections and civil rights protection as essential
characteristics of democracy (Figure S1), which is compatible with the
conceptions of democracy by V-Dem. In Section 6, I also conduct several
important robustness checks to address the concern more directly.
Another problem is differential scaling. While V-Dem’s coders could
follow a coding scheme and would give the most undemocratic regimes very
Table 1. Summary Statistics of Variables.
Number of Standard
Variables Observations Mean Deviation Minimum Maximum
Individual-Level
Variables
Overestimation 31,742 1.55 2.86 4.00 8.00
University degree 31,678 0.18 0.38 0.00 1.00
Female 31,713 0.53 0.50 0.00 1.00
Age 31,704 39.82 15.47 18.00 102.00
Age2/100 31,704 18.25 14.02 3.24 104.04
Married 31,710 0.59 0.49 0.00 1.00
Unemployed 31,649 0.09 0.29 0.00 1.00
Income 30,948 3.80 2.05 0.00 9.00
Subjective social 30,939 1.64 1.01 0.00 4.00
class
Country-Level Variables
Freedom of the 22 29.27 9.78 7.00 48.00
press
Media system 22 40.38 15.36 12.53 65.61
freedom
Log GDP per capita 22 9.12 0.97 7.33 11.24
Growth rate of 22 2.19 6.91 24.10 14.10
GDP per capita
Rule of law 22 38.05 15.26 17.40 84.60
Government 22 40.23 18.16 8.40 93.40
effectiveness
Yeung 243
Figure 3. Distribution of the Dependent Variable.
low scores, citizens may be more reserved in giving extremely low scores to
their country in survey responses—even if they believe their country to be
highly undemocratic. Such a tendency among citizens may even be am-
plified due to preference falsification (but see Shen & Truex, 2021). A
remaining problem pertains to V-Dem’s index per se: Media freedom is part
of V-Dem’s Electoral Democracy Index, while this study regards media
freedom as an explanatory variable. Each of these concerns constitutes
important limitations, and is separately addressed—and mitigated—by the
additional tests in Section 6.
Independent variables
The first key independent variable is media freedom. The data source,
Freedom of the Press, rates the media freedom of each regime based on field
research, reports of international NGOs, and interviews with domestic and
international news media. It has been widely used by previous research on
media control (e.g., Brunetti & Weder, 2003; Egorov et al., 2009; Guriev &
Treisman, 2020b; Kellam & Stein, 2016). It focuses on citizens’ ability to
provide and access news and information through both traditional mass media
and emerging informal social media in legal, political, and economic aspects
(Freedom House, 2017). Legally, it examines how the laws and regulations of
a state influence media content, and how they are used by the state to enable or
restrict the media’s operational ability. Politically, it examines how official
censorship and self-censorship are practiced, as well as the editorial inde-
pendence of both state-owned and privately owned media outlets. Eco-
nomically, it examines how media ownership is structured within the state,
whether it is transparent and highly concentrated, whether private outlets’

establishment and operation are costly (e.g., high taxation, prohibition on
foreign investment), and whether the state allocates advertising or subsidies
disproportionately to pro-government private outlets. Therefore, it impor-
tantly captures the covert strategies of media control adopted by informational
autocrats as discussed by Guriev and Treisman (2019). I reverse-code this
variable such that 0 indicates “least free” and 100 indicates “most free.” In the
dataset, Freedom of the Press varies from 7 to 48 with a mean of 29, indicating
varied levels but a general lack of media freedom across the analyzed
countries (Table 1).
In addition to Freedom of the Press, I use a new media freedom measure,
Media System Freedom (Solis & Waggoner, 2021). Solis and Waggoner
(2021) argue that this latent measure of media freedom is more reliable than
other existing measures of media freedom. They use it to replicate Egorov
et al.’s (2009) seminal study—which uses Freedom of the Press to obtain its
influential media finding—and show that the results do not hold when Media
System Freedom is instead used. This points to a need for using Media System
Freedom as an alternative measure of media freedom. In my analysis, I
multiply the Media System Freedom scores by 100 such that they range from 0
(least free) to 100 (most free). In the dataset, Media System Freedom varies
from 13 to 66 with a mean of 40 (Table 1).
Another key independent variable is university education. Data are re-
trieved from the variable V248 in the WVS, which asks the respondents their
highest education level attained. Those who choose “University-level edu-
cation, with degree” are treated as individuals with a university degree. A
dummy variable University Degree is then compiled, which assigns the value
of 1 to these respondents and 0 otherwise. 18 percent of respondents had a
university degree at the time when the survey was conducted (Table 1).
Control variables
A number of individual characteristics—sex, age, marital status, employment
status, income, and social class—are controlled for. Controlling for these
variables is important because they may be related to both an individual’s
education level and their likelihood of overestimating the democraticness of
their country. These individual control variables are constructed using the
WVS.
I additionally control for country-level variables. First, I control for na-
tional economic performance by using GDP per capita and its one-year growth
rate.19 It is important to control for economic performance because it affects
citizens’ perception and evaluation of the regime on one hand (Guriev &
Treisman, 2020b; Wike et al., 2017), and may be correlated to media freedom
on the other. I additionally control for Rule of Law and Government
Yeung 245
Effectiveness. Data are obtained from the Worldwide Governance Indicators

(Kaufmann et al., 2010), with both variables originally ranging from 2.5
(worst) to 2.5 (best). To ease interpretation, I rescale them such that they range
from 0 (worst) to 100 (best). Controlling for them is important: On one hand,
citizens’ evaluation of the regime may be inflated by its state capacity
(Stockmann & Gallagher, 2011; Truex, 2017); on the other, strong states are
also likely to be more capable of exercising media control.
Estimation strategy
I use multilevel modeling to test the hypotheses. Specifically, I fit a series of
hierarchical linear models that treat individuals as the first level and countries
as the second.20 At the individual level, the model is specified as follows:

Overestimationij ¼ β0j þ β1j þ r1j University Degreeij
(1)
þ β2j Individual Controlsij þ εij
where i indexes an individual and j indexes a country. The residual term r1j is
included to allow the education effect to vary across countries,21 and β0j is the
average intercept plus some country-dependent deviation:
β0j ¼ γ00 þ γ01 Media Freedomj þ γ02 Country Controlsj þ u0j (2)
Under this framework, the coefficients of interest are γ01 and β1j. Spe-
cifically, a negative γ01 would indicate that media control is associated with a
higher degree of individual overestimation of the level of democracy in their
country. A negative β1j, on the other hand, would indicate that university
education is associated with a lesser degree of overestimation. Thus, negative
γ01 and β1j would be evidence in favor of H1 and H2.
Results
Main analysis
Table 2 summarizes the estimation results.22 All models, which vary in the
proxies for media control, show that higher levels of media freedom are
negatively and statistically significantly associated with overestimation of the
level of democracy. The effect is also large. Model 2 predicts that a twenty-
point decline in Freedom of the Press (around two-standard deviations in the
dataset) is associated with a two-point increase in the degree of overesti-
mation. One way to contextualize this finding is to refer to the Freedom of the
Press scores in 2017. According to the 2017 data, a twenty-point decline is
equivalent to deterioration of the level of media freedom from that in Sin-
gapore to that in China: If the level of media freedom in Singapore
Table 2. Overestimation of the Level of Democracy, Media Control, and Education.
Degree of Overestimation
Model 1 Model 2 Model 3 Model 4
Freedom of the press 0.0783** 0.1061***

(0.0320) (0.0251)
Media system freedom 0.0706*** 0.0701***
(0.0189) (0.0120)
University degree 0.3991*** 0.3990*** 0.3994*** 0.3994***
(0.0585) (0.0585) (0.0585) (0.0585)
Female 0.1632*** 0.1632*** 0.1631*** 0.1631***
(0.0546) (0.0546) (0.0546) (0.0545)
Age 0.0216*** 0.0216*** 0.0216*** 0.0216***
(0.0044) (0.0044) (0.0044) (0.0044)
Age2/100 0.0262*** 0.0262*** 0.0262*** 0.0262***
(0.0052) (0.0052) (0.0052) (0.0052)
Married 0.0299 0.0299 0.0299 0.0297
(0.0594) (0.0593) (0.0593) (0.0594)
Unemployed 0.1383 0.1378 0.1386 0.1383
(0.0861) (0.0860) (0.0860) (0.0859)
Income 0.1630*** 0.1630*** 0.1631*** 0.1630***
(0.0309) (0.0309) (0.0309) (0.0309)
Subjective social class 0.0104 0.0106 0.0105 0.0105
(0.0381) (0.0381) (0.0381) (0.0380)
Log GDP per capita 0.1094 0.6011** 0.2446 0.7900***
(0.2780) (0.2548) (0.2889) (0.2943)
Growth rate of GDP per capita 0.0418* 0.0120 0.0042 0.0459
(0.0246) (0.0226) (0.0244) (0.0285)
Rule of law 0.0229 0.0293
(0.0405) (0.0386)
Government effectiveness 0.0415 0.0682**
(0.0325) (0.0315)
Constant 2.4983 7.3656*** 6.3660** 9.7995***
(2.8197) (2.0897) (3.0888) (2.5702)
Number of observations 30,414 30,414 30,414 30,414
Number of countries 22 22 22 22
Log-likelihood 68,636.79 68,631.43 68,634.69 68,631.22
BIC 137,428.4 137,438.3 137,424.2 137,437.9
Note: Both Freedom of the Press and Media System Freedom range from 0 (least free) to 100 (most
free). Income takes a ten-point scale ranging from 0 (lowest) to 9 (highest). Subjective Social Class
takes a five-point scale ranging from 0 (lowest) to 4 (highest). Both Rule of Law and Government
Effectiveness range from 0 (worst) to 100 (best). Robust standard errors are clustered at the
country level and reported in parentheses: *p < 0.10,**p < 0.05,***p < 0.01 (two-tailed).
Yeung 247
retrogressed to that in China, then based on the estimate, an average Sin-

gaporean citizen would overestimate Singapore’s level of democracy by two
additional points in 2017.
How much do the results change if we instead use Media System Freedom
as a proxy for media control? Not much. Models 3 and 4 predict that a two-
standard deviation decline in Media System Freedom is also associated with a
two-point increase in the degree of overestimation. This result is reassuring
and by no means trivial, especially when latest research has failed to use
Media System Freedom to replicate seminal findings based on Freedom of the
Press (Solis & Waggoner, 2021).
So far, the results provide empirical support for H1. What about H2? All
columns in Table 2 show that university education is negatively associated
with overestimation of the level of democracy. The coefficient is statistically
significant at the one-percent level. On average, a university graduate
overestimates the level of democracy in their country by 0.4 points less than
their less educated peers do.
Given the body of literature that views education as propaganda (Cantoni
et al., 2017; Testa, 2018; Voigtländer & Voth, 2015), the education effect
documented here is also nontrivial. There are two possible explanations. First,
previous studies focus on the indoctrination effects of education at the primary
or secondary level, but not at the tertiary level. Second, even if university
education serves as propaganda in nondemocracies, the state’s purpose may
not be in indoctrination, but in costly signaling (Huang, 2015). Thus, the
education finding is not incompatible with past research, and is a novel finding
on its own.
In short, the results provide empirical support for H1 and H2. The fol-
lowing subsection probes the mechanism.
Probing the mechanism

I probe the mechanism for media control in two ways. First, I test if citizens
who live in electoral autocracies are more likely to believe that the elections in
their country are fair when their country exercises a high level of media
control. This is a relevant test, since one proposed mechanism for media
control is that electoral autocracies, which often engage in electoral fraud,
could censor the media to mislead their citizens into believing that the regime
is democratic with free and fair elections. To test this mechanism, I leverage
the same wave of the WVS and compare cross-nationally how an average
citizen from each electoral autocracy rates the electoral process in their
country. My analysis shows that citizens living in electoral autocracies with
highly state-controlled media are less likely to think that their country’s
elections are unfair and manipulated, as compared to citizens living in
electoral autocracies with freer media (Figure 4; see also Figure S8). This is
Figure 4. Media Freedom and Citizens’ Views on the Electoral Process in Their
Country.
Note: Data on citizens’ views on the electoral process in their country are obtained from the
WVS (Wave 6: 2010–16). They are country-average data after dropping nonrespondents, based
on variables V228A, V228B, V228D, and V228H in the WVS. Armenian, Belarusian, Chinese,
Moroccan, and Russian respondents were not asked these questions. Data on media freedom
are obtained from Freedom House’s Freedom of the Press (corresponding years).
consistent with my argument: Media control helps to create a façade of

democracy by denying citizens’ access to information that potentially reveals
the undemocratic nature of the regime. This analysis also echoes recent work
which suggests that media control is important for the survival of electoral
autocracies, where many citizens may genuinely—and wrongly—believe that
the manipulated elections in their country are free and fair (Reuter &
Szakonyi, 2021).
To probe further, I analyze how individual Internet consumption interacts
with media freedom. In countries with weaker media control, citizens would
have better access to regime-revealing information through conventional
Yeung 249
sources (i.e., newspapers, TV, and radio), so the Internet would not have much
additional value in disseminating politically sensitive information to citizens.
In countries with stronger media control, however, citizens’ access to regime-
revealing information through conventional sources would be limited, so the
Internet—as an information source which is harder to monitor—would open
up opportunities for citizens to acquire information that would lower their
assessment of the democraticness of their country.23 If this argument is
correct, we should observe that individuals with more Internet consumption
are generally less likely than their counterparts to overestimate the level of
democracy. Additionally, this effect should vary between countries that ex-
ercise different levels of media control: It should be more pronounced in
countries with less media freedom.
I therefore leverage the individual Internet consumption data from the
WVS (V223), with daily users of the Internet coded as 4 and never-users coded
as 0 on a five-point scale. I then fit a new set of hierarchical linear models to
incorporate a cross-level interaction term between Internet consumption and
media freedom, while allowing the coefficient estimates of education and
Internet consumption to randomly vary between countries (Heisig &
Schaeffer, 2019). The results presented in Figure 5 support my conjecture.
Where the media are more controlled by the state, frequent users of the
Figure 5. Interaction Effect between Media Freedom and Individual Internet

Consumption. Panel A: Freedom of the Press Panel B: Media System Freedom.
Note: Full sets of controls are included in the estimation. Fixed marginal effects are computed by
assuming that all other covariates are at their mean values. The number of observations at the
individual level is 29,051. Twenty-one (rather than 22) nondemocracies are sampled, as
Moroccan respondents were not asked about their Internet consumption in the WVS. While
the estimated interaction effect in Panel A is statistically insignificant, the effect in Panel B is
statistically significant at the five percent level. See Table S1 for the full results.
Internet overestimate the level of democracy less; where the media are freer,
users and non-users of the Internet overestimate with a similar degree. The
interaction effect is positive and statistically significant where Media System
Freedom is used as a proxy for media control (Table S1).24 This further
underscores the importance of media control in making people overrate the
democraticness of their country.
To probe the role of Internet access further, I test if Internet filtering and
social media monitoring at the country level are associated with citizens’
views on the electoral process in their country. Figures S9 and S10 document
strong correlations: Citizens in nondemocracies tend to overrate the electoral
process in their country as Internet filtering and social media monitoring are
more widespread (see also Bailard, 2012; Miner, 2015; but see Reuter and
Szakonyi 2015 for a more nuanced account).
Finally, I probe the mechanism for university education. I argued that
highly educated citizens overestimate the level of democracy in their country
to a lesser degree, since they are more well-informed and thus more able to
resist state propaganda and detect electoral fraud. If this is true, then con-
ditional on living in the same nondemocracy, university graduates should view
the electoral process in their country as less fair. To test this claim, I deploy a
series of questions on electoral fairness in the WVS and regress the resulting
variable—perceived electoral fairness—on university education and Internet
consumption, while using country fixed effects and a rich set of individual
controls to sweep out country- and individual-level confounders (see
Appendix B.2). In line with my theoretical predictions, I find that the co-
efficient estimates for both University Degree and Internet Consumption are
negative and statistically significant (see Appendix B.2). This suggests that
highly educated citizens evaluate the electoral process in their country more
negatively, as compared to their less educated peers in the same country. It also
underscores the role of Internet access, since citizens who use the Internet
more frequently are more negative toward the electoral process in their
country, as compared to the less frequent users in the same country.
Robustness Checks
I conduct additional analyses to check the robustness of my findings.
Addressing measurement problems with the dependent variable

One major concern is that there is a misalignment between citizens’ and expert
coders’ conceptions of democracy. If this is true, then the measured levels of
democracy by V-Dem are conceptually incomparable with citizens’ perceived
levels of democracy. To address this issue, I check the results by constructing
the dependent variable using democracy indices that adopt different
Yeung 251
conceptions of democracy. For this purpose, I use Polity5’s Revised Combined

Polity Score (Center for Systemic Peace, 2020), Freedom House’s (2020a,
2020b) Freedom in the World, and four other indices of V-Dem—v2x_libdem,
v2x_partipdem, v2x_delibdem, and v2x_egaldem (Coppedge et al., 2016;
Sigman & Lindberg, 2019)—to reconstruct the dependent variable. Figure 6
shows that substantive results for media control and university education do
not change.
To further address the concern over a potential misalignment between
citizens’ and expert coders’ conceptions of democracy, I drop respondents
whose conceptions of democracy are apparently incompatible with those of
V-Dem’s expert coders. I define these respondents as those who give any one
of the following responses in the WVS on a scale of 1 (“Not an essential
characteristic of democracy”) to 10 (“An essential characteristic of
democracy”):
1. Governments tax the rich and subsidize the poor: 10.

2. Religious authorities interpret the laws: 10.
3. People receive state aid for unemployment: 10.
4. The army takes over when government is incompetent: 10.
5. The state makes people’s incomes equal: 10.
6. People obey their rulers: 10.
7. People choose their leaders in free elections: 5 or below, or
nonresponse.
Figure 6. Robustness Check—Reconstructing the Dependent Variable Using

Alternative Democracy Indices.
Note: The estimation strategy follows the model specification characterized by Equations (1)
and (2). Coefficient estimates for university education are based on models that use Freedom of the
Press as a proxy for media control. Size and precision of the estimates do not meaningfully change
where models based on Media System Freedom are instead used. Vertical lines represent 95%
confidence intervals.
After dropping these respondents, the sample size greatly decreases from
30,414 to 11,517. Yet, the new results presented in Figures S11 are strikingly
similar to those presented in Table 2. I imposed further restrictions by
lowering the thresholds for items (1) to (6), additionally dropping respondents
who give an “8” or “9” to these items. The results remained largely unchanged
(Figure S11). To deal with individuals’ inherent tendency to use Likert scales
very differently, I conduct additional tests by only including those who rate the
conventional features of democracy as strictly higher than items (1) to (6).25
The results, once again, remain robust (Figure S12). In sum, there is no
evidence that the results reported in Table 2 can be explained by the different
conceptions of democracy between citizens and expert coders.26
To address the issue of differential scaling (v2x_polyarchy vs. the WVS
item on perceived levels of democracy), I reconstruct the dependent variable
in two ways. First, I standardize v2x_polyarchy and perceived levels of
democracy in the full sample, followed by subtracting the former from the
latter for each individual. Second, I collapse the original dependent variable
into a binary one, with 1 indicating that an individual overestimates the level
of democracy in their country (and 0 otherwise). These new measures ar-
guably demand less direct comparability between v2x_polyarchy and the
WVS item on perceived democraticness. As documented in Appendix C.3,
they produce substantively similar results (Tables 2 and S4).
Another important concern is that the media freedom measures could
simply pick up the effect of free and fair elections on citizens’ perceptions of
regime democraticness. This is particularly concerning (1) when media
freedom is part of V-Dem’s index and (2) when nondemocracies with the most
restricted media often have the most unfree and unfair elections. To mitigate
this concern, I conduct three tests.27 First, I construct another dependent
variable by dropping the media freedom component from V-Dem’s index and
redo the main analyses based on this variable (Appendix C.3). Second, I
conduct placebo tests using the two electoral components of V-Dem’s index
(Appendix C.4). Third, I conduct horse race tests using these two variables
(Appendix C.4). While results from the first test highlight the robustness of the
media finding, results from the second and third tests further underscore the
uniqueness of this finding.
Excluding countries that are most likely to bias the results

It is possible that the results are biased as they are strongly driven by the results
from one particular country. One such potential candidate is China, which is
branded by Ringen (2016) as a “controlocracy” and the “perfect dictatorship.”
Moreover, China is the only nondemocracy that is not an electoral autocracy in
my sample. After dropping Chinese respondents from the dataset, I conduct
the same multilevel analyses, and substantive results remain unchanged
Yeung 253
Figure 7. Robustness Check—Dropping Respondents from Different Countries.

Note: The estimation strategy follows the model specification characterized by Equations (1)
and (2). Coefficient estimates for university education are based on models that use Freedom of the
Press as a proxy for media control. Size and precision of the estimates do not meaningfully change
where models based on Media System Freedom are instead used. Vertical lines represent 95%
confidence intervals.
(Figure 7). Additionally, I exclude Russian respondents from the analyses and
obtain similar results (Figure 7). I also exclude Pakistani, Thai, and Ukrainian
respondents as these countries are democracies under Boix et al.’s (2013)
dichotomous coding of democracy. Reassuringly, both media and education
findings remain robust (Figure 7).
Addressing concerns about preference falsification

Another potential problem is preference falsification. The concern is twofold:
At the country level, citizens living in repressive regimes may conceal their
true evaluation of their country; at the individual level, high-income citizens
living in nondemocracies may be systematically more likely to conceal their
true preferences (Jiang & Yang, 2016). I conduct two tests to address the
concern at the country level. First, I exclude the most repressive regimes from
the analysis. If preference falsification in nondemocracies were a serious
problem in the WVS, we would expect it to be most prevalent in highly
repressive regimes. After dropping respondents from the most repressive
countries from my analysis,28 I find that the substantive results remain un-
changed (Figure S17). Second, I control for human rights protection. The idea
is to treat human rights protection as a proxy for preference falsification at the
country level: Citizens living in countries with greater protection of human
rights should be less likely to feel the need to falsify their preferences in
opinion surveys. Table S5 shows that the results are robust to this additional
control.
I conduct four tests to address concerns about preference falsification at the
individual level. In the first two tests, I drop high-income and privileged
respondents from the analysis. These tests are important because these in-
dividuals may be less likely to suffer reprisals for their opinions—or un-
derstand that survey responses will not be shared with the government—while
they are likely to be more educated at the same time.29 My third test is to drop
respondents who are worried about state monitoring,30 and the fourth is to
drop respondents who had other people around who could follow the in-
terview.31 All tests converge to the same conclusion: The results remain robust
to the exclusion of these respondents (Figure S17). I additionally examine the
nonresponses for the question on perceived democraticness in the WVS, and
find no evidence that media control is correlated with nonresponses (Table
S6).
Conducting additional tests for the education finding

I conduct additional robustness checks for the education finding. One possible
concern is that highly educated citizens may simply have more restrictive
operational definitions of democracy, regardless of the regime type. That is,
these citizens may underrate the level of democracy even in actual de-
mocracies. I rule out this alternative explanation in three ways. First, I show
that highly educated citizens and their less educated peers, in fact, have similar
conceptions of democracy (Figures S4–S6). Second, I find that highly edu-
cated citizens overestimate less than their less educated peers do, even after
controlling for citizens’ confidence in the courts, government, and civil
service (Table S8). Third, I demonstrate that, in democracies, highly educated
citizens do not perceive their country to be less democratic compared to their
less educated peers (Table S7). These analyses further highlight the
uniqueness of the education finding to nondemocracies.
Other important tests

Finally, I conduct a series of less theoretically driven—yet empirically
important—robustness checks. First, I redo the main analyses by using OLS
with cluster-robust standard errors based on wild bootstrap (Cameron et al.,
2008) (Table S9). Second, I check if the assumed linear relationship between
media control and overestimation is reasonable,32 and find no evidence for a
concaving effect of media control on overestimation (Figure S18). Sensitivity
analyses additionally suggest that the results are highly robust to unobserved
confounders (Figure S19–S21), which could include preference falsification.
Yeung 255
In sum, the main results reported in this paper withstand a battery of theo-
retically and empirically motivated robustness checks.
Conclusion
Overestimation of the level of democracy is prevalent among citizens in
nondemocracies. Despite such prevalence, no research to date has system-
atically documented this phenomenon and examined its determinants.
Autocrats—especially those in electoral autocracies—are incentivized to
create a façade of democracy because legitimacy often plays an important role
in their survival. Based on this premise, I argued that media control helps to
create this façade, but not all citizens are gullible: Highly educated ones are
not easily manipulated. I combined media freedom data with individual
survey response data to test this argument, and found suggestive evidence for
it.
This study has implications for media control in autocracies. Previous
research suggests that autocrats sometimes prefer freer media because greater
media freedom may help improve the quality of government, while it also
emphasizes the potential drawback associated with the collective action
problems in organizing revolutions (Egorov et al., 2009; Lorentzen, 2014).
Indeed, autocrats often censor the media as they fear collective action (King
et al., 2013), but one added benefit from media control that the extant
scholarship has neglected is its boost of democratic legitimacy. With this
added benefit, autocrats are even more incentivized to suppress media
freedom. This, along with the fear of collective action, may explain why we
observe unprecedentedly high levels of media control in many enduring,
resilient informational autocracies nowadays (Guriev & Treisman, 2019).
This study may also have implications for modernization theory (Glaeser
et al., 2007; Lipset 1959). As the results show, highly educated citizens are less
likely to overestimate the level of democracy in their country. To the extent
that autocracies require democratic legitimacy to survive, an ever-increasing
level of education in a country implies that more and more citizens would
eventually realize that their country is in fact being undemocratically gov-
erned. The resulting loss of democratic legitimacy of the incumbent regime
may, in turn, threaten its survival. This threat is particularly credible when the
regime has electoral institutions, since under a competitive autocracy citizens
have a genuine chance to oust the authoritarian rulers in elections. If this
speculation is true, then we may expect more electoral autocracies, whose
citizens are becoming increasingly educated, to be vulnerable to collapse in
the future.33
This study is not without limitations. As noted, the evidence presented here
is only suggestive. It cannot, and should not, be seen as proof of causal
relationships.34 This is because not only does the empirical analysis rely on
cross-national data that are often noisy and insufficiently fine-grained,35 it also
lacks a credible identification strategy: While reverse causation may be less
likely, omitted variable bias—especially at the country level—may be harder
to rule out. The mechanism behind the education finding is also not entirely
clear, since the effect could be driven by people’s self-selection into university
education. The aim of this article, consequently, is more modest: That is, to
systematically document the prevalence of overestimation of the level of
democracy in nondemocracies, generate theoretically motivated hypotheses to
account for this phenomenon, and show whether the existing data can provide
prima facie evidence for these hypotheses.
Hence, this paper opens up exciting avenues for future research. First and
foremost, it invites more rigorous tests of the relationships explored in this
paper. For example, there are novel ways to gauge media freedom subna-
tionally (Guriev et al., 2021), and future scholarship may leverage this type of
fine-grained data to tease out the causality between media control and
overestimation of the level of democracy. Future research should also follow
up with the following questions: Is media control unconditionally useful for
autocrats? Does it play a key role in enabling autocrats to create a façade of
democracy? If not, what are the other important factors? Addressing these
questions likely involves hard work, but it will certainly pay off by con-
tributing to our understanding of media control, authoritarian resilience, and
democratization.
Acknowledgements
I am indebted to Jennifer Gandhi, Natália Bueno, James Kung, Kai Quek, Pearce
Edwards, and Hsu Yumin Wang for their continued guidance and support. I am also
grateful to Peter Carroll, Tara Farkouh, Joey Glasgow, Yuequan Guo, Nahomi Ichino,
Guoer Liu, Siniša Mirić, Eitan Paul, Davon Thurman, Htet Thiha Zaw, and participants
at ECAPS for their comments on previous drafts of the paper. I also thank the editors
and three anonymous reviewers at CPS for their valuable suggestions and advice. All
errors are my own.
Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research,
authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or
publication of this article.
Yeung 257
ORCID iD
Eddy S. F. Yeung  https://orcid.org/0000-0003-2843-6810
Supplemental Material
Supplementary material for this article is available online at the CPS website http://
journals.sagepub.com/doi/suppl/10.1177/00104140221089647
Notes
1. Examples include the Occupy Wall Street movement in the U.S. and the Yellow
Vests movement in France.
2. This article uses “nondemocracy” and “autocracy” interchangeably.
3. By overestimation, I refer to the situation in which the perceived level of de-
mocracy exceeds the measured level of democracy in a given country.
4. The Electoral Democracy Index by V-Dem rated the U.S. and Japan at 0.90 and
0.85 in 2011 and 2010, respectively, while the average perceived levels of de-
mocracy in these two countries were only 0.60 and 0.64 (on a unit interval) in
corresponding years. China and Singapore were rated 0.09 and 0.41, respectively,
in 2013 and 2012, while the average perceived levels of democracy in these two
countries were 0.60 and 0.65 in corresponding years—nearly identical to those in
the U.S. and Japan.
5. The traditional usage of this term refers to the European Union’s lack of legiti-
macy, but scholars who study the legitimacy crisis in democracies have applied
this concept more broadly to reflect any situation where the perceived democratic
performance falls short of public expectations (Norris, 2011, Ch. 1).
6. By media control, I refer to government control of the media in terms of its
influence on media content, either by practicing media censorship (which includes
both official censorship and self-censorship), misreporting news in a self-serving
manner (i.e., media bias), or dominating media ownership by the state. This
understanding of media control follows Gehlbach and Sonin (2014).
7. There are, however, exceptions. See, e.g., Bleck and Michelitch (2017) for media
control in Mali; Gläßel and Paula (2020) in East Germany.
8. Citizens in nondemocracies indeed value democracy. When asked about the
importance of democracy in the WVS, most respondents in nondemocracies said
democracy is “absolutely important” to them (Figure S7).
9. In the 2013 WVS, among the 2300 Chinese respondents who were asked about
how they viewed the importance of democracy on a ten-point scale (1 = Not at all
important; 10 = Absolutely important), three-tenths gave a score of 10 and nearly
two-fifths gave a score of 8–9. Only 42 respondents (less than 2%) gave a score of
1–4.
10. In their model, citizens do not observe the leader’s type while the informed elite
does. Citizens overthrow the leader if they conclude that the leader is incompetent.
The dictator’s decision is to “invest in making convincing state propaganda,
censoring independent media, coopting the elite, or equipping police to repress

attempted uprisings” (Guriev & Treisman, 2020a, p. 1).
11. Lesta does not test the mechanism concerning information control in oppositional
strongholds, however.
12. Skeptics of the CCP on the basis of its lack of democracy is definitely not rare in
China. In the context of Chinese social media, these skeptics are called the
“American Cent Party,” who “express Western democratic values and criticize the
Chinese communist regime online” (King et al., 2017, fn. 2).
13. For the backfiring dynamics of media control, see Gläßel and Katrin (2020), Guriev
and Treisman (2020b), Roberts (2018), and Shadmehr and Bernhardt (2015).
14. I thank an anonymous reviewer for pointing this out.
15. I use the version released on September 12, 2018, which includes Haiti’s data in
2016. Wave 7 (2017–20) of the WVS is not used because the main media freedom
dataset used in this study has not been updated since 2016.
16. World Value Surveys are nationally representative surveys conducted in countries
from different parts of the world by using a common questionnaire in the native
language. See Lechler and Sunde (2019) for a recent publication that uses the
WVS.
17. In V-Dem’s coding scheme, regimes are classified as democracies only if they
score above 0.5 on the Electoral Democracy Index.
18. Three other nondemocracies in the WVS sample—Kuwait, Qatar, and
Uzbekistan—are excluded because respondents in these countries were not asked
to evaluate the democraticness of their country.
19. Instead of the one-year growth rate, I have also used other growth rates as controls
(e.g., five-year growth rate). Results remain unchanged as these alternative proxies
are used.
20. Compared to an alternative approach which ignores the hierarchical structure of
the dataset by pooling all data and subsequently adjusting for country clusters, the
current approach tends to produce more conservative estimates of standard errors
associated with the country-level variables (Gelman & Hill, 2006; Steenbergen &
Jones, 2002).
21. This is a conservative practice compared to simply fitting a random intercept
model (Heisig & Schaeffer, 2019).
22. Replication materials and code can be found at Yeung (2022).
23. Although online censorship is also common in nondemocracies, access to po-
litically sensitive information via the Internet is much harder to prevent. For
example, while conventional media are “tightly controlled” in Saudi Arabia,
citizens’ use of Twitter creates trouble to the regime given Twitter’s noncom-
pliance with the Saudi government’s requests (Pan & Siegel, 2020, p. 109). In-
ternet control in China is carefully executed, but Internet censorship circumvention
is not uncommon among highly educated and tech-savvy Chinese citizens, who
use proxying and Virtual Private Networks to leap over the Great Firewall of China
(Roberts, 2018).
Yeung 259
24. To confirm that the heterogeneous effects are unique to Internet consumption, I
have also estimated the cross-level interaction effects between media freedom and
individual consumption of conventional information sources (newspapers, TV,
and radio). The estimates are all statistically insignificant (p > 0.20).
25. I thank an anonymous reviewer for suggesting this test.
26. Relatedly, I also drop respondents who do not believe that democracy is important.
Substantive results remain unchanged (Figure S13).
27. I thank two anonymous reviewers for raising this concern and suggesting relevant
tests.
28. The countries are Iraq, Pakistan, and Yemen—identified by using the Political
Terror Scale (Gibney et al., 2019).
29. I thank an anonymous reviewer for highlighting this possibility.
30. I use the variable V186 in the WVS for this purpose.
31. I use the variable V252 in the WVS for this purpose.
32. See Egorov et al. (2009), Huang et al. (2019), and Lorentzen (2014) on het-
erogeneous media effects that are sometimes found in nondemocracies.
33. Speculatively, the recent electoral defeat of the United Malays National
Organisation—the dominant party in Malaysia for six decades—might be an
example. The near record low of electoral support for the People’s Action Party of
Singapore in 2020 may hint at this possibility, too.
34. However, it is also worth noting that the robust correlation between media control
and overestimation of the level of democracy, as found in this study, could be an
underestimate. This is because citizens living in opposition strongholds may not be
subject to state-wide media control (Lesta, 2019), while the media finding
documented here is obtained after including these individuals in the sample.
35. For instance, while the media freedom data used in this study are at the country
level, citizens living in different regions of the country may face different levels of
media control.
References
Accornero, G. (2016). The revolution before the revolution: Late authoritarianism and
student protest in Portugal. Berghahn Books. https://doi.org/10.2307/j.ctvpj7h76
Bailard, C. S. (2012). A field experiment on the Internet’s effect in an African election:
Savvier citizens, disaffected voters, or both? Journal of Communication, 62(2),
330–344. https://doi.org/10.1111/j.1460-2466.2012.01632.x
Bleck, J., & Michelitch, K. (2017). Capturing the airwaves, capturing the nation? A
field experiment on state-run media effects in the wake of a coup. Journal of
Politics, 79(3), 873–889. http://dx.doi.org/10.1086/690616
Boix, C., Miller, M., & Rosato, S. (2013). A complete data set of political regimes,
1800–2007. Comparative Political Studies, 46(12), 1523–1554. https://doi.org/
10.1177/0010414012463905
Brunetti, A., & Weder, B. (2003). A free press is bad news for corruption. Journal of Public
Economics, 87(7), 1801–1824. https://doi.org/10.1016/S0047-2727(01)00186-4
Cameron, C., Gelbach, J. B., & Miller, D. L. (2008). Bootstrap-based improvements

for inference with clustered errors. Review of Economics and Statistics, 90(3),
414–427. https://doi.org/10.1162/rest.90.3.414
Campante, F. R., & Chor, D. (2012). Why was the Arab world poised for revolution?
Schooling, economic opportunities, and the Arab spring. Journal of Economic
Perspectives, 26(2), 167–187. https://doi.org/10.1257/jep.26.2.167
Campbell, D. E. (2013). Social networks and political participation. Annual Review of
Political Science, 16(1), 33–48. https://doi.org/10.1146/annurev-polisci-033011-
201728
Cantoni, D., Chen, Y., Yang, D. Y., Yuchtman, N., & Zhang, Y. J. (2017). Curriculum
and ideology. Journal of Political Economy, 125(2), 338–392. https://doi.org/10.
1086/690951
Center for Systemic Peace. (2020). Polity5 annual time-series, 1946–2018 (Version
2018) [Data set and code book]. http://www.systemicpeace.org/inscr/p5v2018.xls
Chen, Y., & Yang, D. Y. (2019). The impact of media censorship: 1984 or brave new
world? American Economic Review, 109(6), 2294–2332. https://doi.org/10.1257/
aer.20171765
Chu, Y. H., Bratton, M., Lagos, M., Shastri, S., & Tessler, M. (2008). Public opinion
and democratic legitimacy. Journal of Democracy, 19(2), 74–87. https://doi.org/
10.1353/jod.2008.0032
Coppedge, M., Gerring, J., Knutsen, C. H., Lindberg, S. I., Teorell, J., Altman, D.,
Bernhard, M., Fish, M. S., Glynn, A., Hicken, A., Lührmann, A., Marquardt,
K. L., Mcmann, K., Paxton, P., Pemstein, D., Seim, B., Sigman, R., Skaaning,
S.-E., Staton, J., & Ziblatt, D. (2021). V-dem codebook v11.1, Varieties of De-
mocracy (V-Dem) project.
Coppedge, M., Lindberg, S., Skaaning, S.-E., & Teorell, J. (2016). Measuring high
level democratic principles using the V-Dem data. International Political Science
Review, 37(5), 580–593. https://doi.org/10.1177/0192512115622046
Dahlum, S., & Wig, T. (2019). Educating demonstrators: Education and mass protest in
Africa. Journal of Conflict Resolution, 63(1), 3–30. https://doi.org/10.1177/
0022002717721394
Dahlum, S., & Wig, T. (2021). Chaos on campus: Universities and mass political
protest. Comparative Political Studies, 54(1), 3–32. https://doi.org/10.1177/
0010414020919902
Dalia (2020). Democracy perception index – 2020 [Data set]. https://daliaresearch.
com/blog/democracy-perception-index-2020
Dalton, R. J. (2004). Democratic challenges, democratic choices: The erosion of
political support in advanced industrial democracies. Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780199268436.003.0009
Dalton, R. J., & Shin, D. C. (2006). Citizens, democracy, and markets around the
Pacific Rim: Congruence theory and political culture. Oxford University Press.
Dee, T. S. (2004). Are there civic returns to education? Journal of Public Economics,
88(9), 1697–1720. https://doi.org/10.1016/j.jpubeco.2003.11.002
Yeung 261
De Micheli, D. (2021). Racial reclassification and political identity formation. World

Politics, 73(1), 1–51. https://doi.org/10.1017/S0043887120000179
Dukalskis, A., & Gerschewski, J. (2017). What autocracies say (and what citizens
hear): Proposing four mechanisms of autocratic legitimation. Contemporary
Politics, 23(3), 251–268. https://doi.org/10.1080/13569775.2017.1304320
Egorov, G., Guriev, S., & Sonin, K. (2009). Why resource-poor dictators allow freer
media: A theory and evidence from panel data. American Political Science
Review, 103(4), 645–668. https://doi.org/10.1017/S0003055409990219
Enikolopov, R., Petrova, M., & Zhuravskaya, E. (2011). Media and political per-
suasion: Evidence from Russia. American Economic Review, 101(7), 3253–3285,
https://doi.org/10.1257/aer.101.7.3253
Freedom House. (2017). Freedom of the Press research methodology. https://
freedomhouse.org/freedom-press-research-methodology
Freedom House. (2020a). Freedom of the press [data set]. https://freedomhouse.org/
sites/default/files/2020-02/FOTP1980-FOTP2017_Public-Data.xlsx
Freedom House. (2020b). Freedom in the World (Version 2020) [Data set]. https://
freedomhouse.org/sites/default/files/2020-02/2020_Country_and_Territory_
Ratings_and_Statuses_FIW1973-2020.xlsx
Frye, T., Gehlbach, S., Marquardt, K. L., & Reuter, O. J. (2017). Is Putin's popularity
real? Post-Soviet Affairs, 33(1), 1–15, https://dx.doi.org/10.1080/1060586X.
2016.1144334
Geddes, B., & Zaller, J. (1989). Sources of popular support for authoritarian regimes.
American Journal of Political Science, 33(2), 319–347. https://doi.org/10.2307/2111150
Gehlbach, S., & Sonin, K. (2014). Government control of the media. Journal of Public
Economics, 118(1), 163–171. https://doi.org/10.1016/j.jpubeco.2014.06.004
Gelman, A., & Hill, J. (2006). Data analysis using regression and multilevel/
hierarchical models. Cambridge University Press. https://doi.org/10.1017/
cbo9780511790942
Gerschewski, J. (2013). The three pillars of stability: Legitimation, repression, and co-
optation in autocratic regimes. Democratization, 20(1), 13–38. https://doi.org/10.
1080/13510347.2013.738860
Gerschewski, J. (2018). Legitimacy in autocracies: Oxymoron or essential feature?
Perspectives on Politics, 16(3), 652–665. https://doi.org/10.1017/
S1537592717002183
Gibney, M., Cornett, L., Wood, R., Haschke, P., Arnon, D., Pisanò, A., & Barrett, G.
(2019). The Political Terror Scale 1976-2018 [Data set and code book]. http://
www.politicalterrorscale.org
Gläßel, C., & Paula, K. (2020). Sometimes less is more: Censorship, news falsification,
and disapproval in 1989 East Germany. African Journal of Paediatric Surgery:
AJPS, 64(3), 682–698. https://doi.org/10.1111/ajps.12501
Glaeser, E. L., Ponzetto, G. A., & Shleifer, A. (2007). Why does democracy need
education?. Journal of economic growth, 12(2), 77–99, https://doi.org/10.1007/
s10887-007-9015-1
Grönlund, K., & Milner, H. (2006). The determinants of political knowledge in

comparative perspective. Scandinavian Political Studies, 29(4), 386–406. https://
doi.org/10.1111/j.1467-9477.2006.00157.x
Gunitsky, S. (2015). Corrupting the cyber-commons: Social media as a tool of au-
tocratic stability. Perspectives on Politics, 13(1), 42–54. https://doi.org/10.1017/
S1537592714003120
Guriev, S., Melnikov, N., & Zhuravskaya, E. (2021). 3G Internet and confidence in
government. Quarterly Journal of Economics, 136(4), 2533–2613. https://doi.
org/10.1093/qje/qjaa040
Guriev, S., & Treisman, D. (2019). Informational autocrats. Journal of Economic
Perspectives, 33(4), 100–127. https://doi.org/10.1257/jep.33.4.100
Guriev, S., & Treisman, D. (2020a). A theory of informational autocracy. Journal of
Public Economics, 186, 104158. https://doi.org/10.1016/j.jpubeco.2020.104158
Guriev, S., & Treisman, D. (2020b). The popularity of authoritarian leaders: A cross-
national investigation. World Politics, 72(4), 601–638. https://doi.org/10.1017/
S0043887120000167
Heisig, J. P., & Schaeffer, M. (2019). Why you should always include a random slope
for the lower-level variable involved in a cross-level interaction. European So-
ciological Review, 35(2), 258–279. https://doi.org/10.1093/esr/jcy053
Hillygus, D. S. (2005). The missing link: Exploring the relationship between higher
education and political engagement. Political Behavior, 27(1), 25–47. https://doi.
org/10.1007/s11109-005-3075-8
Huang, H. (2015). Propaganda as signaling. Comparative Politics, 47(4), 419–437.
https://doi.org/10.5129/001041515816103220
Huang, H., Boranbay-Akan, S., & Huang, L. (2019). Media, protest diffusion, and
authoritarian resilience. Political Science Research and Methods, 7(1), 23–42.
https://doi.org/10.1017/psrm.2016.25
Huang, H., & Yeh, Y. Y. (2019). Information from abroad: Foreign media, selective
exposure, and political support in China. British Journal of Political Science,
49(2), 611–636. https://doi.org/10.1017/S0007123416000739
Inglehart, R., Haerpfer, C., Moreno, A., Welzel, C., Kizilova, K., Diez-Medrano, J.,
Lagos, M., Norris, P., Ponarin, E., & Puranen, B. (Eds.). (2014). World Values
Survey: Round six country-pooled datafile version (version 2018) [Data set and
code book]. :Madrid: JD Systems Institute. http://www.worldvaluessurvey.org/
WVSDocumentationWV6.jsp
Inglehart, R., & Welzel, W. (2005). Modernization, cultural change, and democracy:
The human development sequence. Cambridge University Press. https://doi.org/
10.1017/CBO9780511790881
Jiang, J., & Yang, D. L. (2016). Lying or believing? Measuring preference falsification
from a political purge in China. Comparative Political Studies, 49(5), 600–634.
https://doi.org/10.1177/0010414015626450
Yeung 263
Kaufmann, D., Kraay, A., & Mastruzzi, M. (2010). The Worldwide governance In-
dicators: A summary of methodology, data and analytical issues. (World Bank
Policy Research Working Paper No. 5431)
Kellam, M., & Stein, E. A. (2016). Silencing critics: Why and how presidents restrict
media freedom in democracies. Comparative Political Studies, 49(1), 36–77.
https://doi.org/10.1177/0010414015592644
Kennedy, J. J. (2009). Maintaining popular support for the Chinese Communist Party:
The influence of education and the state-controlled media. Political Studies,
57(3), 517–536. https://doi.org/10.1111/j.1467-9248.2008.00740.x
Kerr, N., & Lührmann, A. (2017). Public trust in manipulated elections: The role of
election administration and media freedom. Electoral Studies, 50(1), 50–67.
https://doi.org/10.1016/j.electstud.2017.08.003
King, G., Pan, J., & Roberts, M. E. (2013). How censorship in China allows gov-
ernment criticism but silences collective expression. American Political Science
Review, 107(2), 326–343. https://doi.org/10.1017/S0003055413000014
King, G., Pan, J., & Roberts, M. E. (2017). How the Chinese government fabricates
social media posts for strategic distraction, not engaged argument. American
Political Science Review, 111(3), 484–501. https://doi.org/10.1017/
S0003055417000144
Lechler, M., & Sunde, U. (2019). Individual life horizon influences attitudes toward
democracy. American Political Science Review, 113(3), 860–867. https://doi.org/
10.1017/S0003055419000200
Lesta, N. W. (2019). The political geography of electoral autocracies: The influence of
party strongholds on political beliefs in Africa. Electoral Studies, 60(1), 102047.
https://doi.org/10.1016/j.electstud.2019.05.001
Li, L. (2004). Political trust in rural China. Modern China, 30(2), 228–258. https://doi.
org/10.1177/0097700403261824
Lipset, S. M. (1959). Some social requisites of democracy: Economic development and
political legitimacy. American Political Science Review, 53(1), 69–105. https://
doi.org/10.2307/1951731
Lorentzen, P. (2014). China’s strategic censorship. American Journal of Political
Science, 58(2), 402–414. https://doi.org/10.1111/ajps.12065
Maerz, S. F. (2019). Simulating pluralism: The language of democracy in hegemonic
authoritarianism. Political Research Exchange, 1(1), 1–23. https://doi.org/10.
1080/2474736X.2019.1605834
Majone, G. (1998). Europe’s “democratic deficit”: The question of standards. Euro-
pean Law Journal, 4(1), 5–28. https://doi.org/10.1111/1468-0386.00040
Mattes, R., & Bratton, M. (2007). Learning about democracy in Africa: Awareness,
performance, and experience. American Journal of Political Science, 51(1),
192–217. https://doi.org/10.1111/j.1540-5907.2007.00245.x
McClurg, S. D. (2003). Social networks and political participation: The role of social
interaction in explaining political participation. Political Research Quarterly,
56(4), 449–464. https://doi.org/10.1177/106591290305600407
McFaul, M. (2005). Transitions from postcommunism. Journal of Democracy, 16(3),

5–19. https://doi.org/10.1353/jod.2005.0049
Miller, A. H., Hesli, V. L., & Reisinger, W. M. (1997). Conceptions of democracy
among mass and elite in post-Soviet societies. British Journal of Political Science,
27(2), 157–190. https://doi.org/10.1017/S0007123497000100
Miner, L. (2015). The unintended consequences of Internet diffusion: Evidence from
Malaysia. Journal of Public Economics, 132(1), 66–78. http://dx.doi.org/10.
1016/j.jpubeco.2015.10.002
Mische, A. (2008). Partisan publics: Communication and contention across Brazilian
youth activist networks. Princeton University Press. http://dx.doi.org/10.1515/
9781400830817
Moehler, D. C., & Singh, N. (2011). Whose news do you trust? Explaining trust in
private versus public media in Africa. Political Research Quarterly, 64(2),
276–292. https://doi.org/10.1177/1065912909349624
Munson, Z. W. (2010). The making of pro-life activists. University of Chicago Press.
http://dx.doi.org/10.7208/chicago/9780226551210.001.0001
Norris, P. (2011). Democratic deficit: Critical citizens revisited. Cambridge University
Press. https://doi.org/10.1017/CBO9780511973383
Pan, J., & Siegel, A. A. (2020). How Saudi crackdowns fail to silence online dissent.
American Political Science Review, 114(1), 109–125. https://doi.org/10.1017/
S0003055419000650
Przeworski, A. (1991). Democracy and the market: Political and economic reforms in
eastern europe and Latin America. Cambridge University Press. https://doi.org/
10.1017/CBO9781139172493
Qi, L., & Shin, D. C. (2011). How mass political attitudes affect democratization:
Exploring the facilitating role critical democrats play in the process. International
Political Science Review, 32(3), 245–262. https://doi.org/10.1177/
0192512110382029
Reuter, O. J., & Szakonyi, D. (2015). Online social media and political awareness in
authoritarian regimes. British Journal of Political Science, 45(1), 29–51. https://
doi.org/10.1017/S0007123413000203
Reuter, O. J., & Szakonyi, D. (2021). Electoral manipulation and regime support.
World Politics, 73(2), 275–314. https://doi.org/10.1017/S0043887120000234
Ringen, S. (2016). The perfect dictatorship: China in the 21st century. Hong Kong
University Press. http://dx.doi.org/10.5790/hongkong/9789888208937.001.0001
Roberts, M. E. (2018). Censored: Distraction and diversion inside China’s Great
Firewall. Princeton University Press. https://doi.org/10.23943/9781400890057
Roberts, M. E. (2020). Resilience to online censorship. Annual Review of Political
Science, 23(1), 401–419. https://doi.org/10.1146/annurev-polisci-050718-032837
Roeder, P. G. (1989). Modernization and participation in the Leninist developmental
strategy. American Political Science Review, 83(3), 859–884. https://doi.org/10.
2307/1962064
Yeung 265
Rohrschneider, R. (2002). The democracy deficit and mass support for an EU-wide
government. American Journal of Political Science, 46(2), 463–475. https://doi.
org/10.2307/3088389
Rohrschneider, R., & Loveless, M. (2010). Macro salience: How economic and po-
litical contexts mediate popular evaluations of the democracy deficit in the
European Union. Journal of Politics, 72(4), 1029–1045. https://doi.org/10.1017/
S0022381610000514
Rose, R., Mishler, W., & Haerpfer, C. (1998). Democracy and its alternatives: Un-
derstanding post-communist societies. Johns Hopkins University Press.
Rozenas, A., & Stukal, D. (2019). How autocrats manipulate economic news: Evi-
dence from Russia’s state-controlled television. Journal of Politics, 81(3),
982–996. http://dx.doi.org/10.1086/703208
Sanborn, H., & Thyne, C. L. (2014). Learning democracy: Education and the fall of
authoritarian regimes. British Journal of Political Science, 44(4), 773–797.
https://doi.org/10.1017/S0007123413000082
Shadmehr, M., & Bernhardt, D. (2015). State censorship. American Economic Journal:
Microeconomics, 7(2), 280–307. http://dx.doi.org/10.1257/mic.20130221
Shen, X., & Truex, R. (2021). In search of self-censorship. British Journal of Political
Science, 51(4), 1672–1684. https://doi.org/10.1017/S0007123419000735
Shirikov, A. (2021). Who trusts state-run media? Polarized perceptions of news
credibility in non-democracies [Unpublished manuscript]. https://doi.org/10.
2139/ssrn.3686299
Sigman, R., & Lindberg, S. I. (2019). Democracy for all: Conceptualizing and
measuring egalitarian democracy. Political Science Research and Methods, 7(3),
595–612. https://doi.org/10.1017/psrm.2018.6
Solis, J. A., & Waggoner, P. D. (2021). Measuring media freedom: An Item Response
Theory analysis of existing indicators. British Journal of Political Science, 51(4),
1685–1704. https://doi.org/10.1017/S0007123420000101
Steenbergen, M., & Jones, B. S. (2002). Modeling multilevel data structures. American
Journal of Political Science, 46(1), 218–237. https://doi.org/10.2307/3088424
Stier, S. (2015). Democracy, autocracy and the news: The impact of regime type on
media freedom. Democratization, 22(7), 1273–1295. https://doi.org/10.1080/
13510347.2014.964643
Stockmann, D., & Gallagher, M. E. (2011). Remote control: How the media sustain
authoritarian rule in China. Comparative Political Studies, 44(4), 436–467.
https://doi.org/10.1177/0010414010394773
Tannenberg, M., Bernhard, M., Gerschewski, J., Lührmann, A., & Von Soest, C.
(2021). Claiming the right to rule: Regime legitimation strategies from 1900 to
2019. European Political Science Review, 13(1), 77–94. https://doi.org/10.1017/
S1755773920000363
Teorell, J., Coppedge, M., Lindberg, S., & Skaaning, S.-E. (2019). Measuring pol-
yarchy across the globe. Studies in Comparative International Development,
54(1), 71–95. https://doi.org/10.1007/s12116-018-9268-z
Testa, P. A. (2018). Education and propaganda: Tradeoffs to public education provision

in nondemocracies. Journal of Public Economics, 160(1), 66–81. https://doi.org/
10.1016/j.jpubeco.2018.03.001
The Economist. (2021, January 11). Sadyr Japarov is elected president of Kyrgyzstan
in a landslide. http://econ.st/3qjUusw
Truex, R. (2017). Consultative authoritarianism and its limits. Comparative Political
Studies, 50(3), 329–361. https://doi.org/10.1177/0010414014534196
Voigtländer, N., & Voth, H.-J. (2015). Nazi indoctrination and anti-Semitic beliefs in
Germany. Proceedings of the National Academy of Sciences, 112(26),
7931–7936. https://doi.org/10.1073/pnas.1414822112
Von Soest, C., & Grauvogel, J. (2017). Identity, procedures and performance: How
authoritarian regimes legitimize their rule. Contemporary Politics, 23(3),
287–305. https://doi.org/10.1080/13569775.2017.1304319
Wang, C., & Huang, H. (2021). When “fake news” becomes real: The consequences of
false government denials in an authoritarian country. Comparative Political
Studies, 54(5), 753–778. https://doi.org/10.1177/0010414020957672
Whiting, S. H. (2017). Authoritarian “rule of law” and regime legitimacy. Comparative
P o l i t i c a l S t u d i e s , 5 0 ( 1 4 ) , 1 9 0 7 – 1 9 4 0 . h t t p s : / / d o i . o r g / 1 0 . 11 7 7 /
0010414016688008
Wike, R., Simmons, K., Stokes, B., & Fetterolf, J. (2017). Globally, broad support for
representative and direct democracy. Pew Research Center. http://assets.
pewresearch.org/wp-content/uploads/sites/2/2017/10/17102729/Pew-Research-
Center_Democracy-Report_2017.10.16.pdf
Wong, G. W. (1990). Tiananmen: Causes and consequences. Washington Quarterly,
13(1), 79–95. https://doi.org/10.1080/01636609009477533
Xu, X., Kostka, G., & Cao, X. (forthcoming). Information control and public support
for social credit systems in China. Journal of Politics. https://doi.org/10.1086/
718358
Yeung, E. S. F. (2022). Replication data for: Overestimation of the level of democracy
among citizens in nondemocracies. https://doi.org/10.7910/DVN/HUAADF
Zaller, J. (1992). The nature and origins of mass opinion. Cambridge University Press.
https://doi.org/10.1017/CBO9780511818691
Zhu, J.-H., & Rosen, S. (1993). From discontent to protest: Individual-level causes of
the 1989 pro-democracy movement in China. International Journal of Public
Opinion Research, 5(3), 234–249. https://doi.org/10.1093/ijpor/5.3.234
Author Biography
Eddy S. F. Yeung is a PhD student in the Department of Political Science at
Emory University. His research focuses on authoritarian politics and com-
parative public opinion, particularly the role of political communication in
shaping regime attitudes and policy preferences. He can be reached at
shing.fung.yeung@emory.edu.

Overstimation of The Level of Democracy

Uploaded by

Copyright:

Available Formats

Overstimation of The Level of Democracy

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Overstimation of The Level of Democracy

Uploaded by

Copyright:

Available Formats

Article

Comparative Political Studies

Emory University, Atlanta, GA, USA

“Authoritarian equilibrium rests mainly on lies, fear, or economic

Figure 1. Perceived and Measured Levels of Democracy—All Sampled Countries in

Thus, media control provides a powerful tool for autocrats to create a

Third, I expand the literature on media control. Scholars of authoritarian

Perceived Democraticness, Legitimacy, and

Figure 2. Perceived and Measured Levels of Democracy—Sampled Nondemocracies

of democracy across different types of regimes and different regimes of the

Legitimacy and authoritarian survival

Mirziyoyev, claimed that he strove to build “a free, democratic, humane state,”

Media control as a strategy in autocracies

control to explain the high level of public support or regime legitimacy in

Theoretical Framework and Hypotheses

H1 (media control): A higher level of media control in nondemocracies in-

Yet, the exposure–acceptance model suggests that acceptance of the message

Higher education additionally promotes political participation and expands

H2 (education): Highly educated citizens in nondemocracies overestimate the

· In Singapore, Lee Kuan Yew used to coopt corporate boards of key

Data and Methods

· Dependent variable: an ordinal variable indicating the degree to which

To construct the dependent variable, Overestimation, I ﬁrst obtain indi-

their country at 2, they would underestimate the level of democracy by 3

Table 1. Summary Statistics of Variables.

Figure 3. Distribution of the Dependent Variable.

whether it is transparent and highly concentrated, whether private outlets’

Effectiveness. Data are obtained from the Worldwide Governance Indicators

Table 2. Overestimation of the Level of Democracy, Media Control, and Education.

Model 1 Model 2 Model 3 Model 4

Freedom of the press 0.0783** 0.1061***

retrogressed to that in China, then based on the estimate, an average Sin-

Probing the mechanism

consistent with my argument: Media control helps to create a façade of

Figure 5. Interaction Effect between Media Freedom and Individual Internet

Addressing measurement problems with the dependent variable

conceptions of democracy. For this purpose, I use Polity5’s Revised Combined

1. Governments tax the rich and subsidize the poor: 10.

Figure 6. Robustness Check—Reconstructing the Dependent Variable Using

Excluding countries that are most likely to bias the results

Figure 7. Robustness Check—Dropping Respondents from Different Countries.

Addressing concerns about preference falsiﬁcation

Conducting additional tests for the education ﬁnding

Other important tests

Declaration of Conﬂicting Interests

censoring independent media, coopting the elite, or equipping police to repress

Cameron, C., Gelbach, J. B., & Miller, D. L. (2008). Bootstrap-based improvements

De Micheli, D. (2021). Racial reclassiﬁcation and political identity formation. World

Grönlund, K., & Milner, H. (2006). The determinants of political knowledge in

McFaul, M. (2005). Transitions from postcommunism. Journal of Democracy, 16(3),

Testa, P. A. (2018). Education and propaganda: Tradeoffs to public education provision

You might also like

Freedom of the press 0.0783 0.1061*