Exploring popular conceptions of democracy through media discourse: analysing dimensions of democracy from online media data in 93 countries using a distributional semantic model

ABSTRACT

Survey studies show that popular support for democracy is strong in democratic and non-democratic countries. Naturally, the question is if democracy actually means the same thing in different linguistic, cultural, and political contexts. Mass media is often mentioned as decisive in forming citizens' understandings of democracy, but the media discourse is rarely in focus in comparative studies on popular conceptions of democracy. This article contributes to the debate by analysing data collected from online media in 93 countries. By utilizing tools from natural language processing, we provide new insights based on methods that are both extensive, flexible and cost-efficient. Our analysis shows that the media discourse revolves around democracy as governance, as outcomes and as values, but that these abstract understandings have additional dimensions. Our main contributions are three: (i) we show that the media discourse is related to popular understandings of democracy; (ii) our results indicate that there are common denominators of how the D-word is discussed in media across the globe, but when analysing the dimensions in more detail, common denominators are few and (iii) by relating democracy to everyday politics, media seems to legitimize any regime as democratic rather than being a beacon for liberal democracy.

KEYWORDS:

• The meaning of democracy
• distributional semantics
• word2vec
• editorial media
• social media
• dimensions of democracy

1. Introduction

Survey studies over the last decades show that a vast majority of people all around the world can offer a definition of democracy, and that popular support for democracy is very strong both in democratic and non-democratic countries.¹ Moreover, democracy is today the only regime form with positive connotations and political leaders in authoritarian or semi-authoritarian countries describe their specific regimes as democratic, although adding different adjectives, and democratic rule is asserted in the constitutions of 184 countries in the world.² Democracy today, it would seem, is not only “a universal value […] that people anywhere may have reason to see […] as valuable”,³ they actually do value democracy. A growing number of cross-cultural survey studies challenge, however, the idea that “democracy has emerged as a universal value”⁴ and that widespread support and desire for liberal democracy all over the world would (potentially) lead to democratic transition in authoritarian states and to consolidation of democracy in newly established democracies.⁵ The main critique is that democracy may mean different things in different linguistic, cultural, and political contexts.⁶ Knowing “how strongly people desire democracy” is, for instance, “meaningless unless we also know how people understand democracy”.⁷ We thus need to ask what citizens are expressing support for (desire or are satisfied with) when asked about an essentially contested concept such as democracy;⁸ we “need to find out what concept people have in mind when they rate democracy’s importance”.⁹

To explain how and why popular understandings of democracy vary significantly both within a single country and between countries, individual-level factors as well as the impact of the economic, political and cultural contexts are put forward.¹⁰ In these studies, mass media is often mentioned as decisive in forming citizens understandings, consistent with media studies showing the significant role of media in shaping public opinion.¹¹ Mass media is conceived of “as one of the primary agencies for learning about democracy […]”,¹² not least in authoritarian states, where media is “the main, if not unique, source of information about democracy, whatever the nature of this information”.¹³ A recent study has also shown that the media have a key role in democratization in a long-term macro-level perspective, i.e. for democratic transition,¹⁴ and during such transitions, citizens turn to media for information and especially so “in countries that are further away from democratic consolidation”.¹⁵ However, the media discourse and its content are, for practical reasons, rarely in focus in studies explaining differences in popular conceptions of democracy.

The state of knowledge on both individual and macro level factors affecting popular conceptions of democracy is well-advanced through the vast number of previous survey research. This article contributes to the contemporary debate on popular conceptions of democracy by analysing the media context in 93 countries based on geo-coded language data collected from online editorial and social media. We use distributional semantics – a statistical approach for quantifying semantic similarities based on co-occurrence information collected from large amounts of online text data – to account for language differences between lexical realisations of the concept democracy across languages.¹⁶ New technological advancements make these kinds of studies comparably cheap, in comparison to survey studies, and we may also access the media discourse in otherwise closed authoritarian regimes in real time. The data enables us to analyse the meanings of words in normal, unsolicited and contemporary language use. Compared to other methodological approaches with the objective of identifying and measuring cross-cultural and cross-lingual discrepancies, this approach has the advantage of enabling us to analyse how concepts are used in their “natural habitat”,¹⁷ i.e. to analyse specific language contexts in which democracy is used. Instead of asking people questions about their conceptions of democracy, we “listen” to the societal debate (through online media). Thereby, we capture how “democracy is discussed in public discourse”, which “likely has a strong impact on public understanding and endorsement of democracy”.¹⁸

The overarching aim of this article is descriptive as we explore how distributional semantics applied to the media discourse can be used to provide a background context to an analysis of the meaning of democracy in different social, cultural and political contexts across languages and countries. Our analysis shows that the media discourse across the globe on a more abstract level revolves around democracy as a system of governance, democracy as outcomes and democracy as values. Within each of these abstract understandings of democracy, however, the media discourse may be categorized further. Democracy as a system of governance, for instance, is discussed in terms related to politics in general, to institutions and to democracy as regime form (also with additional layers). Together, these different understandings, or dimensions, of democracy constitute what we call a “democracy tree”. In relation to previous research our main contributions and findings are three: First, we show that the media context is connected to popular understandings of democracy, although our data includes these additional dimensions to the more abstract categories compared. Second, democracy is discussed in a similar way in the media discourse in all or most cultural regions of the world and in all different regime types, i.e. in terms of democracy as a system of governance, as outcomes and as values. On this more abstract level there are thus common denominators of the D-word in the media across the globe. When analysing the different dimensions of democracy in more detail, however, the thought of finding common denominators vanishes. Third, by relating democracy to everyday politics, media seems to legitimize any regime as democratic rather than being a beacon for liberal democracy, something that could help explain why there is such little cross-cultural variation in support for democracy across the globe.

In the following section two, we present our data. Thereafter, we briefly review the rich literature on popular conceptions of democracy. In section four, we present our results, i.e. our dimensions of democracy in form of a “democracy tree”. In part five we discuss our findings in relation to previous research and in the concluding part six, we sum up our findings.

2. Data: the LES online distributional semantic lexicon

Our analysis is based on the methods and data cumulatively developed within the research project “Linguistic Explorations of Societies” (LES).¹⁹ Using the online distributional semantic lexicon developed within the LES-project,²⁰ this article is a theoretical and analytical continuation of previous work within the project, explicitly situating the study within the contemporary debate on the meaning of democracy. The lexicon is based on word embedding models developed from distributional semantics, which is a sub-field of natural language processing (NLP). In essence, distributional semantics is concerned with the quantification of semantic similarity between various linguistic items, and its basic idea is that semantic similarity can be defined as linguistic distributions.²¹ The by now famous phrase “know a word by the company it keeps”, uttered by the British linguist, John Rupert Firth, highlights the key assumption within distributional semantics: words that share similar distributions also share similar meanings.²² Alessandro Lenci defines the distributional hypothesis as follows: “The degree of semantic similarity between two linguistic expressions A and B is a function of the similarity of the linguistic contexts in which A and B appear”.²³ As such, if we observe that word A and word B continuously occur in the same contexts, we are correct in assuming that A and B are semantically similar.²⁴

Distributional semantic models thus produce word embeddings, which are numeric representations of words as vectors in a multidimensional vector space.²⁵ A higher degree of closeness between vectors (measured as cosine similarity that ranges from −1, no similarity, to +1, perfect similarity) thus signifies a higher degree of semantic similarity between the words the vectors represent. A distributional semantic model can compute different co-occurrence relations, and we are looking to capture so-called paradigmatic similarity relations. By modelling paradigmatic relations in text data from different languages and countries, we can uncover similarities and differences in terms of both usage and meaning of words across cultures. If we recall the notion of “meaning as use” famously coined by Ludwig Wittgenstein,²⁶ claiming that words are meaningless outside the context in which they are used, paradigmatic word embeddings thus enable us to explore the meaning and usage of the word democracy in its “natural habitat”.

The LES-lexicon, whose computational implementation is based on a word2vec variant known as the Continuous Bag of Words (CBOW) model, accesses online text data from both editorial content and social media. This data is provided by large data vendors for a substantial number of languages and countries. Moreover, each model is uniquely based on data from a specific language, country, and source type, whether editorial or social media.²⁷ The text data is delivered in various web document formats – such as editorial news articles, blog entries, forum posts, and the like – and is tagged with metadata that includes the language, country, and source type.²⁸

In this study we have decided to combine the media types for two reasons. First, our aim is not to tell anything meaningful about differences between editorial and social media, although it is possible to analyse such differences across countries (a venue for future research). Second, and more importantly, the two media types are not very different on the aggregate in terms of democracy connotations.

In Figure 1, we have correlated the top 50 most frequently occurring terms that are being used in the same way as the word democracy in online editorial and social media.²⁹ As shown, the correlation between the two sources is impressively high (b .71, R² 73). Without preceding our analysis, sovereignty is clearly the most frequently mentioned term related to democracy, along with terms such as nation and politics, which evidently have validity as they all relate to the term democracy. A word such as dictatorship also appears to be very common and a bit more so in social media, while another regime form, monarchy, is more common in editorial media. Even though such terms must be considered as quite distant from democracy, their presence is not surprising given that the algorithms provide us with terms used in a similar manner. This means that antonyms for democracy can also be captured. In this particular case dictatorship and monarchy can be seen as alternative systems of rule – or regime forms – to democracy.

Figure 1. Top 50 most frequent democracy topics across 93 language-country units.

Note: Relative frequency is calculated as the absolute frequency divided by the number of existing language-country units from editorial media and social media.

We focus on countries as the units of analysis. This means that we have aggregated our results on the basis of countries even though some countries in our data contains several languages.³⁰ For this study, the neighbour terms of the target term democracy and its appropriate translations from 117 language-country combinations (15 from editorial data and 15 from social data) was exported in January 2018. From this, a total of 3510 terms were extracted and subsequently translated. Because translation requires an intimate knowledge of the languages coupled with a good understanding of the country context in which the languages are spoken (which cannot be assured using machine translation), a group of international annotators were tasked with translating the neighbour terms. Considering the potential risks of mistranslation, potentially leading to worsened translational discrepancies, this approach may seem precarious, particularly considering the language “agnostic” approach of a study like this.³¹ However, the translation procedure was conducted in a supervised setting where each annotator was tasked with translating the terms and provide several translation suggestions where applicable. They were also required to provide descriptions for each term, explaining its meaning in greater detail including the general use of the term in the specific country context. To ensure inter-translation reliability, we assigned two native speakers for most language-country combinations. All translations were carefully examined with the support of machine translation where only one annotator was employed or, in very few cases, where two translators disagreed. After the translation procedure, a few language-country combinations were omitted from the study as their data quality was deemed too poor, which left us with 93 language-country combinations in total and 3414 neighbour terms to democracy.³²

One major caveat with the data is that all corpora the models have sampled texts from are cumulative in nature given that they consist of texts ranging from early 2015 (when the lexicon was first built) to early 2018 when all neighbour terms for this study were exported. We currently have no means of controlling for over-time variation, which remains a limitation of the study. However, as most models are based on very large amounts of text – the average number of words is approximately 150 million for all editorial media corpora and 140 million for all social media corpora (see Table 1) – this effectively makes most models less susceptible to dramatic changes over time even with subsequent model updates.³³

Table 1. Summary statistics of lexicon data size in number of words by online source type (2019)^a.

Source	Mean	SD	Min	Max	Total
News	150 200 000	101 644 149	1941	351 100 000	21 175 792 758
Social	143 100 000	102 835 088	17 000	378 700 000	20 181 014 528

^a Table taken from Dahlberg et al. “Distributional Semantic Lexicon”.

3. Popular conceptions of democracy revisited

Although our main aim is to describe how democracy is discussed in online media across different social, cultural, linguistic, and political contexts, of importance is, of course, how the media debate relates to popular understandings of democracy. In general, popular conceptions of democracy have been analysed by using two different types of methods, survey research and qualitative ethnographic studies.³⁴ In survey research, studies are either based on close- or open-ended questions. Studies based on the former use questions from either single- or cross-country surveys, like regional barometers or the World Values Surveys (WVS). By having a predefined set of questions, this method reconstructs popular conceptions of democracy indirectly. The responses to the questions asked are analysed by sorting them into different conceptions of democracy using taxonomies, most often derived and predefined beforehand from democratic theory.³⁵ There is no well-recognized and common way of constructing these taxonomies, since they vary depending on the aim of the study and the data used.³⁶ This is evident in a brief overview of the literature within the field where scholars, for instance, use a distinction between an electoral, liberal and redistributive democracy, a procedural, instrumental and authoritarian democracy, a liberal, electoral and instrumental democracy, a liberal, electoral, social and direct democracy, a liberal democracy discourse vs. a guardianship discourse, liberty-procedural democracy vs. performance-distributive democracy, a procedural vs. a substantive conception, a political and an economic conception, or between an ideal of democracy as political or social rights.³⁷

Open-ended surveys, in contrast, “allows (and requires) respondents to define democracy in their own words”.³⁸ This approach thus avoids forcing people to conceptualize democracy “in terms of a few elements” predefined by a theoretical understanding of democracy,³⁹ simultaneously as it confirms whether people actually have an understanding of democracy or not.⁴⁰ In, for instance, a study conducted in Eastern Europe, Asia, Africa, and Latin America by Russel Dalton et al.,⁴¹ the different open-ended responses were categorized into three broadly defined conceptions of democracy based on previous surveys and scholarly definitions: (i) democracy as freedom, civil liberties and citizen rights; (ii) as democratic institutions and the democratic political process; and (iii) as social benefits, including references to economic and social development, as well as equality, justice, peace or stability, i.e. different forms of outcomes of the political system. In another study, on how Latin American citizens’ conceptions of democracy affect political behaviour and attitudes, Damarys Canache categorized the open-ended survey responses into eight different categories – no meaning, liberty and freedoms, political equality, participation, rule of law, economic and social outcomes, negative meaning and other meanings – in order to stay close to how citizens formulate democracy in their own minds.⁴²

There are, however, several problems with studies based on surveys. First, they delimit the multidimensional character of the concept democracy through the restricted battery of questions in studies based on close-ended questions, and through the limited number of possible answers in surveys using open-ended questions.⁴³ Second, close-ended questions enforce an answer although the question may not be related to the respondent’s understanding of democracy at all,⁴⁴ and in open-ended surveys the coding scheme becomes topical for the possible conclusions. Third, in cross-cultural and cross-language comparisons survey studies rarely analyse if the meaning of the different conceptions found is the same.

These limitations are avoided in studies based on ethnographic methods tracing the meaning of democracy through how the concept is used in everyday language and how it fits into a “semantic field” of related concepts, i.e. its associated meanings, ideals and standards. The main idea here is thus similar to distributional semantics: the meaning of a concept is its usage.⁴⁵ The ethnographic approach allows for findings where a person may express both multivalent and contradictory conceptions of democracy and where seemingly alternative conceptions of democracy (like the taxonomies in the survey studies above) may be complementary and parts of a larger whole. The main concrete method to uncover the meaning of democracy within a specific social, cultural or language context is through different forms of interviews or focus groups to capture respondents’ subjective meanings of democracy.⁴⁶ Ethnographic studies show that democracy, like in open-ended surveys, most often is related to civil liberties and values like freedom, liberty and rights, as well as to political participation and outcomes such as peace or socio-economic benefits.⁴⁷ In addition to these familiar conceptions, however, the ethnographic approach also uncovers meanings rarely mentioned in survey studies, such as personal virtues like “loyalty” or “honesty”.⁴⁸ Moreover, the in-depth analyses clearly show that popular conceptions of democracy have a particular meaning in a specific context. The same words used in describing or defining democracy in different languages, like freedom and liberty, may not have the same meaning in all languages, or in different cultural or political contexts, and are therefore not directly comparable.⁴⁹ However, the ethnographic method is obviously hard to apply in order to establish cross-national comparability, which explains why such studies rarely lay claim to generalizations above a single country case, or beyond the comparison between a few countries.

In survey studies, variance within, as well as between, countries is most often explained by traditional theories in political science explaining political behaviour and attitudes on an individual level,⁵⁰ focusing on an individual’s resources and motivations.⁵¹ Differences between countries or regions are further explained by context, commonly taking its starting-point in theories on the importance of culture,⁵² for instance, if a country is characterized by a democratic political culture and values,⁵³ or the importance of religion.⁵⁴ Other important contextual factors frequently used in different studies are economic development and economic performance, the level and the quality of democracy, as well as the democratic history of a country.⁵⁵ However, these explanatory factors are mainly applied to explain variation in conceptions of democracy already taken for granted. As Canache argues: the “survey measures reflect assumptions about what political principles, values, and practices define democracy and there implicitly presume that citizens elsewhere share such assumptions”.⁵⁶

A growing number of cross-cultural survey studies today take this latter challenge head on by showing that the meaning of democracy varies with the cultural and political context.⁵⁷ In a recent study, Doh Chull Shin found that “most people in the authoritarian and post-authoritarian world of today”, although expressing a strong support for democracy, “appear to prefer a hybrid system of mixing democratic and authoritarian politics to a liberal democracy”.⁵⁸ In many parts of the world, authoritarian notions of democracy are mixed up with liberal notions of democracy. Widespread support for democracy may in that way paradoxically coexist with a lack or an “outright absence” of democracy without undermining the legitimacy of authoritarian rule.⁵⁹ One explanation for this confusion is a lack of emancipative values, i.e. values that emphasizes freedom of choice and equality of opportunities.⁶⁰ In liberal cultures these values are widespread and well-established, while in authoritarian cultures these values are often absent. “In the absence of emancipative values, authoritarian notions of democracy blossom and lend a principled form of legitimacy to autocratic rule, disguised under lip service to democracy”.⁶¹

In addition to emancipative values, scholars have suggested that the immediate political context, the political regime, determines the meaning of democracy on an individual level.⁶² Tom Ulbricht distinguishes, for instance, between four types of political regimes (democracies, hybrid democracies, hybrid autocracies and autocracies) and shows that the less democratic a political regime is, the more authoritarian the term democracy is understood to be. “[A]ny political regime operates as its own normative benchmark” and its definition of democracy “trickles down in society” affecting the citizens’ conceptions of democracy making them mistake authoritarian rule for democracy.⁶³

The importance of the cultural and political context brings the role of mass media to the fore, as an important source of information on politics, even in a changing media landscape.⁶⁴ Media contributes to shaping public opinion, not least in terms of agenda setting and framing.⁶⁵ In relation to regime type, for instance, an important question is if media are able and willing to critically analyse the prevailing political regime and report on the government’s actions, something that under authoritarian regimes could be harder.⁶⁶ Leaders in authoritarian states are not seldom conceived of as “keen on using the mass media and education system to indoctrinate their citizens with such manipulated discourse on democracy that work to their advantage”,⁶⁷ in contrast to parts of the world where emancipative values are established and functions as “the prime evaluation standard of the critical media […]”.⁶⁸

Despite being so important, the media discourse and its content are rarely or never in focus in cross-cultural and cross-linguistic studies on popular conceptions of democracy. The ambition of this study is to remedy that neglect. What is the meaning of democracy in normal uncontrolled contemporary language use? How is the media discussion related to popular understandings of democracy? And how are different connotations of democracy travelling around the world?

4. Results: the “democracy tree”

In comparison to other cross-cultural and cross-linguistical studies on the meaning of democracy, the data from the LES-lexicon has a clear resemblance with data from open-ended surveys, since there are no predefined democratic models structuring the gathering of the data as in most close-ended survey studies.⁶⁹ “Listening” to the debate on democracy in the media context, however, widens what democracy may be all about in contrast to asking explicit survey questions like “To you, what does ‘democracy’ mean?”,⁷⁰ and with limited space for the open-ended answer. In this way, our analysis is more like ethnographic studies where democracy in day-to-day discourse may be related to, for instance, virtues such as honesty and loyalty.⁷¹ Another resemblance to ethnographic studies is that the LES-data show the meaning of democracy in specific cultural and linguistic contexts through how the concept is used in everyday language and how it fits into a semantic field of related concepts. Naturally, the data are much less ambitious in depth in comparison to ethnographic studies but instead they allow for cross-cultural comparisons between an indefinite number of language-country combinations.

Having 3414 neighbour terms to democracy from 93 language-country combinations, the coding of the data is decisive for the results. The data was coded by the same two persons (the authors of this article) at three different points in time with six months in between them.⁷² In the initial stage, the coding was conducted inductively in the sense that we created our first categories exclusively based on the data and the native annotators’ comments. Previous studies have shown that theoretical understandings of democracy often are far from “ordinary citizens in their views and understandings of democracy”,⁷³ something also reflected in our data on the media discourse. Each term was categorized into what became eleven categories: democracy as belief system and ideology; as values; a system of governance; unity; nation-building; system-capacity; actors; action; redistribution and social democracy; religion; and corruption.

In the second step of the coding process, our aim was to reduce and structure the data so that it could serve as an analytical framework comparable with and useful to cross-cultural survey studies on the meaning of democracy with the explicit starting-point in previous research. As shown in section three, there are no well-defined meanings of democracy or canonical sets of popular conceptions to be found, no matter if the studies are based on open- or close-ended surveys, or ethnographic methods. There is, however, a common core in most studies related to either a procedural or a substantive understanding of democracy.⁷⁴ In the procedural understanding, there is an obvious core stemming from Joseph Schumpeter’s minimal definition of democracy related to the electoral process,⁷⁵ and Robert Dahl’s work in democratic theory on democracy as institutions, rules and procedures for governance, like free elections, majority rule, multiparty systems, and the rule of law.⁷⁶ In Dahl’s procedural democracy, there are also different liberal values included as indispensable preconditions for democracy, like civil liberties and individual freedoms,⁷⁷ which is a more demanding (liberal) procedural understanding.⁷⁸ Values may, however, also be interpreted as goals of democracy in terms of freedom, social justice or equality and thus as a substantive understanding,⁷⁹ together with ideas on democracy as delivering concrete social and economic outcomes such as redistribution, development, peace or legitimacy and stability (good governance), thus emphasizing the instrumental value of democracy.⁸⁰

This distinction between a procedural and a substantive understanding of democracy served as our foundation in creating two abstract categories, democracy as a system of governance and democracy as outcomes. However, we added a third abstract category: democracy as values. There are both empirical and theoretical reasons for this third category. First, when democracy is related to terms like, for instance, equality or freedom in our data, we cannot discern if they are discussed as preconditional liberal values for democracy as a system of governance (and thus as part of a procedural understanding of democracy), or as an outcome of a democratic regime (as a substantive understanding). Second, studies based on open-ended surveys often differentiate values from both a procedural and a substantive understanding of democracy.⁸¹ Third, values are topical in the contemporary debate for explaining different popular conceptions of democracy across the globe.⁸² Moreover, adding terms like equality and freedom, or specific civil rights, into a procedural understanding of democracy would bias our coding towards a liberal democratic understanding of democracy.

These three abstract categories were then used to differentiate further dimensions of democracy, together constituting what we call a “democracy tree” (see Figure 2).⁸³ These categories correspond to results from a previous study using the LES-lexicon finding five abstract categories (community, belief systems, principles, governance, and outcome) theoretically derived from the work of David Easton and Pippa Norris.⁸⁴ That these categories are included in our “democracy tree” signals that our coding is reliable and that our data have validity. We have, however, gone several steps further in our differentiation of various dimensions of democracy, making our results comparable to other cross-cultural survey studies.

Figure 2. The democracy tree (percentages within each category in parentheses).

Note: Percentages are calculated within each category for the different branches except for those marked with *, where the number indicates the proportion of terms in UC3 relating to actors or conflict.

Our first main category, democracy as a system of governance (or, in short, democracy as governance), is clearly related to a procedural understanding of democracy. One dimension of democracy as governance thus consists of terms with reference to institutions – formal institutions, rules and procedures – like national assembly, constitution and meritocracy. However, in a democracy these institutions create a “unique system for organizing relations between rulers and the ruled”⁸⁵ – a democratic regime form – comparable to other systems of government, other regimes.⁸⁶ In our data, as mentioned above (Figure 1), both dictatorship and monarchy, two alternative regime forms to democracy, are among the most frequently occurring terms but there are also other terms referring to autocracies, as well as to religious and bureaucratic regime forms. Democracy as regime form thus constitutes another dimension of democracy as governance. Democracy is also analogous to politics in general within the media discourse closely related to the role of the media, and like how citizens conceive of democracy in ethnographic studies.⁸⁷ This third dimension of democracy as governance, relating to the daily practices of the ruling system of governance, includes terms like politics, and diplomacy as well as terms referring to actors within the political system.

When democracy as outcome is discussed in the literature (substantial democracy) there is a distinction made between, on the one hand, outcomes related to expectations on what democracy might lead to or counteract, like peace, economic stability, or corruption, i.e. in terms related to the capacity of the political system to deliver some desired outcomes,⁸⁸ and, on the other, outcomes directly related to the social dimension of democracy, i.e. to different policies on social reforms and redistribution, ensuring general welfare and social equality.⁸⁹ In our data, the expectations on democracy span over this wide field of different desired outcomes mentioned in the literature. For instance, we found terms related to system capacity in different ways, for instance, to stability, legitimacy and accountability, but also terms relating to its opposite, corruption. Development is another dimension of the more abstract category outcome, discussed in the media in terms of democratization, modernization and globalization. Several terms relate to peace, or to its antonym conflict, while terms relating to redistribution or social democracy were few. However, we also found terms like sovereignty (the most frequent term of all in our data as shown in Figure 1), self-determination, and community. These, and similar terms, have been coded as a specific dimension of the abstract category outcome called nation-building. It is a dimension of democratic outcomes that is rarely mentioned in previous studies, excluded among the questions in close-ended surveys and in open-ended surveys such terms are often conceived of as miscellaneous and included in a category “other”.⁹⁰ However, an understanding of democracy in these terms is not uncommon in ethnographic studies.⁹¹

Democracy is, however, also clearly related to values in different ways in the literature: as “essential goals” of democracy (freedom and liberty), as “core democratic values” (political liberties or gender equality) or to “emancipative values” serving as prerequisites for a liberal democratic understanding of democracy.⁹² Many terms in our data relates to democracy as values, and we coded these terms into four different dimensions. From previous studies, we recognize both the dimension of democracy including terms referring to universal values such as freedom, tolerance or equality, and the dimension including explicit references to civil liberties and freedoms. However, like in ethnographic studies,⁹³ we also found terms related to personal characteristics or virtues that are value-laden like dignity and integrity, constituting a third dimension of democracy as values. The fourth dimension, rarely analysed in other studies, includes terms referring to systems of ideas based on values, or what we have called belief systems. One dominant dimension here includes terms relating to political ideologies, including terms such as ideology, capitalism, and socialism. However, we also found terms relating to other systems of ideas based on values: to religions, to belief systems in general (what we have called democracy as doctrine), and to ethics.

On an aggregate level, democracy as outcome dominates in the media debate: 1473 terms were coded into this category (43 percent of all terms). This is not surprising, since the main role of the media is to inform their citizens about politics, and to monitor and critically scrutinize the actions and performance by the political elite, although the latter – “watchdog journalism” – may be harder or impossible to perform in non-democratic countries with limited press freedom.⁹⁴ 1041 terms were coded into democracy as governance, and 900 terms into democracy as values (constituting 30 and 26 percent of all terms respectively).

5. Discussion: to “listen” instead of asking questions

Of interest here is, of course, how well our data corresponds to popular conceptions of democracy derived from asking close- or open-ended questions, i.e. from indirect or direct definitions based on answers from citizens themselves. Could our analysis help “investigate factors that shape how individuals receive and react to the information they get about democracy and democratic rule”?⁹⁵ To compare our results with studies based on close-ended questions on democracy is, however, not an easy task. In the WVS/EVS surveys there are, for instance, only two out of nine survey questions on democracy that are remotely comparable with any of our three abstract categories: “Civil rights protect the people’s liberty against oppression” and “Women have the same rights as men”. When we aggregate and correlate these questions with our conception democracy as values, we find a moderate but significant correlation (Figure 3).

Figure 3. LES values vs values (WVS/EVS 2017–2021).

Note: Malaysia was excluded in the correlation since it turned out as a severe outlier with a Cook’s distance of 1. The correlation is, however, still significantly positive with Malaysia included.

Based on open-ended questions in 49 countries (established democracies, new democracies and non-democracies), the 2007 study by Dalton et al. is one of the most ambitious and well-cited.⁹⁶ As mentioned above, they found three dominant conceptions of democracy: democracy as freedom/civil liberties/citizen rights, as political process, and as social benefits. Conceptions of democracy as freedom/civil liberties/citizen rights dominate in their material, while democracy as social benefits are the least frequent.⁹⁷ In their coding, terms related to civil liberties, personal freedoms, group rights, and group freedoms were included in the category freedom/civil liberties/citizen rights; terms like voting, electoral choice, multiparty competition, government by the people, majority rule, and government effectiveness and accountability in the category the political process; and terms related to socio-economic development, personal security, equality and justice, peace, and unity to the category social benefits. They also created a residual (“other”) category consisting of “other positive terms” and “other negative terms” related to democracy, but also of other “miscellaneous responses”, such as “national independence”, “change government”, and references to individual politicians or political parties.⁹⁸

In relation to our study, these conceptions correspond roughly to our categories democracy as values, as a system of governance and as outcome. There are, however, several differences of importance. We have, for instance, categorized terms related to values – like equality and justice – in the general conception of democracy as values besides terms related to civil liberties and freedoms. Furthermore, instead of creating a residual other category we have included terms such as independence, sovereignty and self-determination (democracy as nation-building), as well as names of politicians and political parties (democracy as politics). In spite of the differences in coding, there is a strong and significant correlation between our category democracy as values and their conception of democracy as freedom/civil liberties/citizen rights, as well as between our democracy as outcome and their social benefits when we aggregate and correlate the different conceptions.⁹⁹ The correlation between democracy as a system of governance (ours) and the democratic political process (theirs) is less strong but still significant (Figures 4–6).¹⁰⁰

Figure 4. LES values vs freedom/rights/liberty (Dalton, Shin, and Jou, “Popular Conceptions of the Meaning of Democracy”).

Note: In the data from Dalton, Shin, and Jou, “Popular Conceptions,” we have in Figures 4–6 removed the “don’t know” category and recalculated the proportions based on substantial answers. The bivariate models are based on OLS with robust regression weights.

Figure 5. LES Outcome vs social benefits (Dalton, Shin, and Jou, “Popular Conceptions of the Meaning of Democracy”).

Figure 6. LES Governance vs political process (Dalton, Shin, and Jou, “Popular Conceptions of the Meaning of Democracy”).

The relatively strong correlation between our analysis of online media and studies based on open-ended questions suggests that there is a close relation between how democracy is discussed in media and how citizens define democracy in their own words. In contrast to these studies, however, we have added new and more fine-grained layers to the more abstract categories, for instance, the dimensions of democracy as nation-building (outcome), as regime form (governance) and as belief system (values). With our democracy tree, we can therefore dig deeper into the broader concepts of democracy found in previous research and discern how the concept differs across cultural, linguistic, and political contexts, and we are able to track the least common denominators of the D-word across the world. The “democracy tree” may thus be used as a framework for analyses of the social, cultural and political context in which democracy is discussed across languages and countries, something we will briefly illustrate by using Christian Welzel’s categorization of the world’s countries into regional clusters.¹⁰¹

On a general level, analysing our three abstract categories, democracy seems to be discussed in similar ways in the media discourse across the globe, although there are small but substantial differences on which aspect of democracy is emphasized (Figure 7). Democracy as outcome is, for instance, predominant in the media context in Sub-Saharan Africa, Latin America, North Africa and the Middle East but also in Catholic Europe and Ex-Communist East. Democracy as values dominates in the English West, Protestant Europe, and South Asia, while democracy as governance is dominant in Ex-Communist West and East Asia.

Figure 7. Conceptions of democracy in different cultural regions (relative frequency).

When we probe deeper into these abstract categories, however, the variance between different cultural regions increases in line with findings in previous studies.¹⁰² If we, for instance, compare the dimensions of democracy within the media discourse in Ex-Communist West, Protestant Europe and South Asia – three cultural regions where democracy on this more abstract level seems to be discussed in a similar way – we find several important differences: politics in general is more important in Ex-Communist West than in the other two regions, civil liberties and universal values are more important in Protestant Europe, and while system capacity is the single most important outcome in Ex-Communist West (and almost exclusively mentioned in terms of corruption), redistribution is more important in Protestant Europe than in the other regions and peace is more important in South Asia (Figure 8).

Figure 8. Dimensions of democracy in three different cultural regions (relative frequency within regions).

Probing further into these dimensions, our analysis shows that regime form is the dominant dimension of democracy as governance in all three regions, however, in Protestant West and South Asia terms relating to a democratic regime form dominate, while terms relating to autocratic regime forms dominate in Ex-Communist West (see Appendix 3). Universal values are more frequently related to democracy in the media debate in Protestant Europe and among those values, equality is the dominant one (and there are explicit references to gender equality, absent in the other two regions). Moreover, in Ex-Communist West, terms relating to tolerance as a universal value and to multiculturalism as an ideology are completely absent, a finding that corresponds well with current survey research showing a decline in ethnic tolerance in this region.¹⁰³ In the margin of the debate on “Asian values”,¹⁰⁴ it is worth noticing that terms referring to human rights are missing in the South Asian media discourse and that there are no explicit references to civil liberties.

A similar pattern is found when applying our democracy tree on different regime types using V-dem data: liberal democracies, electoral democracies, and autocracies (electoral and closed).¹⁰⁵ Previous survey studies have shown that the immediate political context affects popular conceptions of democracy,¹⁰⁶ but – once again – the differences in the media discourse between different regime types are rather negligible on an abstract level (Figure 9).¹⁰⁷

Figure 9. Conceptions of democracy in different regime types (relative frequency).

However, when we probe further into the different dimensions of democracy, we find clear differences (see Appendix 3). On a more general level, democracy as values is more important in liberal democracies and one of the corner stones in the theoretical construct of liberal democracy, civil liberties are, not surprisingly, much more prominent in the media discourse in liberal democracies in comparison to electoral democracies and autocratic regimes. Democracy as governance is in liberal and electoral democracies most often discussed in terms of regime form, while in autocracies politics in general is the dominating dimension. Moreover, democracy as governance in terms of regime form is in liberal democracies discussed primarily in terms of a democratic regime form in contrast to both electoral democracies and autocracies, where autocratic regime forms dominate the media debate. In liberal democracies, institutions are primarily understood in terms of rules, in contrast to electoral democracies and autocracies where governance as procedures and formal institutions respectively dominate. Redistribution as outcome seems to have salience mainly in the media discourse in liberal democracies. If we continue our analysis even further, probing all the way down to term level, the differences in the media discourse between democracies and autocracies become even more apparent.

6. Concluding remarks

In an inspiring article on conceptions of democracy based on survey data, Ulbricht is asking for more qualitative approaches towards anchoring the D-word,¹⁰⁸ and in a recent article Chapman et al. ask for more research on the role of the public discourse in the meaning-making processes of popular understandings of democracy.¹⁰⁹ We are offering an alternative in both these respects that is thick in its description, broad in its coverage and relatively cheap in economic terms. Instead of asking people about their conceptions of democracy in a survey, we “listen” to the public discourse through online media. By focusing on both editorial and social media, we do not only capture how established media is framing democracy but also how people are talking about democracy online in general in their daily lives. Although not all people are represented or active in online communities, the online debate might still reflect understandings of democracy on a general level within a society.¹¹⁰ At least, this assumption is supported by the relatively strong correlations between our data and the findings on popular conceptions of democracy from open-ended survey data.

Obviously, however, within media discourse, democracy has many more dimensions than those usually ascribed to the concept in survey studies: as politics, as regime form, as virtues, as belief system, and as nation-building. In relation to previous research on the meaning of democracy, we thus believe that the main contribution of our analysis is that our democracy tree broadens our understanding of the associated meanings, ideals, and standards of the concept democracy in everyday language, i.e. how it fits into a “semantic field” of related concepts.¹¹¹ One important finding, when tracking the least common denominators of the D-word, is that our study indicates that the different dimensions of democracy exist in the media discourse in all or most parts of the world, and the more abstract the conceptions or categories are formulated, the more common the denominators are. However, when we analyse the dimensions in more detail comparing both the emphasis and content of the different dimensions of democracy through the terms that connote democracy within a specific semantic field, common denominators across the globe are few or non-existent.

As a concluding remark, we believe our analysis may help explain why there is such little cross-cultural variation in citizens (strong) support for democracy across the globe,¹¹² if media is important for democratic learning as suggested by many scholars.¹¹³ In everyday language, as analysed through the media discourse, the concept democracy connotes governance, values and outcomes no matter the cultural region or regime type. Expressed in these vague terms within a “semantic field” – and in line with the main role of the media to inform their citizens about politics as well as to monitor and (hopefully) critically scrutinize the actions of the regime – democracy is in the media discourse closely related to day-to-day politics and political practices, just like citizens “individually expressed meanings of democracy”.¹¹⁴ Everyday politics is democracy, or democracy is everyday politics. In this perspective, support for democracy may thus only indicate support for the actions of the prevailing regime, indicating why “large portions of ordinary citizens mistakenly identify authoritarian political practices as being those of democratic rule”.¹¹⁵ Instead of being a beacon for liberal democracy, media as a main source of information about democracy seems to legitimize any regime as democratic.

Acknowledgements

Thanks to Dorothée Behr, Staffan I. Lindberg, Ludvig Beckman, Hanspeter Kriesi, Jan Teorell and Annabelle Lever for their comments on previous drafts of the manuscript. Special thanks to Sofia Axelsson for help with data management, and for organizing and administering annotators and translations.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Vetenskapsrådet [grant number 2017-02429].

Notes on contributors

Stefan Dahlberg

Stefan Dahlberg is a chair professor in political science at Mid Sweden University and Professor II at the Department of Government, University of Bergen. He is the PI for the Swedish European Value Study (EVS) and serves as a member of the EVS scientific committee. His research interests include comparative politics and political behaviour.

Ulf Mörkenstam

Ulf Mörkenstam is a professor in political science at Stockholm University and the Institute for Future Studies, Stockholm. His research interests are within democratic and political theory with a specific focus on the rights of indigenous peoples.

Notes

1 See Dalton, Shin, and Jou, “Understanding Democracy”; Norris, Democratic Deficit; Welzel and Alvarez, “Enlightening People”; Cho, “Global Citizenries.”

2 Brunkert, “Overselling Democracy.”

3 Sen, “Democracy as a Universal Value,” 12, our emphasis.

4 Shin, “Assessing Responses,” 1.

5 See, e.g., Haerpfer et al. Democratization; Shin, “Democratic Deconsolidation”; Lu and Shu, Understandings of Democracy.

6 Shin, “Assessing Responses”; Kirsch and Welzel, “Democracy Misunderstood”; Zagrebina, “Concepts of Democracy”; Davis, Goidel, and Zhao, “Meanings of Democracy”; Lu and Chu, Understandings of Democracy.

7 Welzel, Freedom Rising, 310.

8 Cf. Gallie, “Essentially Contested Concepts.”

9 Ulbricht, “Perceptions and Conceptions,” 1396.

10 See, e.g., Welzel, Freedom Rising; Ulbricht, “Perceptions and Conceptions”; Lu and Chu, Understandings of Democracy.

11 See, e.g., Mutsvairo and Rønning, “Janus Face”; McCombs and Valenzuela, Setting the Agenda.

12 Norris, Democratic Deficit, 7.

13 Zagrebina, “Concepts of Democracy,” 176.

14 Arendt, “Media and Democratization.”

15 Loveless, “Media Dependency,” 176.

16 Turney and Pantel, “From Frequency to Meaning.”

17 Wittgenstein, Philosophical Investigations.

18 Chapman et al. “Under the Veil,” 16.

19 The research project “Linguistic Explorations of Societies” (LES), is located at the University of Gothenburg, and is an interdisciplinary research project that cuts across the disciplines of political science, computational linguistics and computer science. The LES-lexicon can be accessed via: https://www.gu.se/en/linguistic-explorations-of-societies.

20 For detailed information about the lexicon, see Dahlberg, Axelsson, and Holmberg, “Democracy in Context”; Dahlberg et al. “Distributional Semantic Lexicon.”

21 Lenci, “Distributional Semantics”; Sahlgren, “The Distributional Hypothesis.”

22 Firth, A Synopsis of Linguistic Theory, 11.

23 Lenci, “Distributional Semantics,” 3.

24 Sahlgren, The Word-Space Model.

25 Turney and Pantel, “From Frequency to Meaning.”

26 Wittgenstein, Philosophical Investigations.

27 Editorial media refers to traditional forms of media, such as newspapers, magazines, and online news outlets, where the content is typically created and curated by professional journalists, editors, and writers. In contrast, social media consists of forums and blogs where the texts are user-generated and based on informal interactions.

28 Dahlberg et al. “Distributional Semantic Lexicon.”

29 See online Appendix A2 for an overview of top ten URLs by language and country.

30 In the analyses, we have weighted the countries according to term frequency, since all countries do not have the same number of terms representing them due to: (a) multiple languages; (b) error terms coded as missing; and (c) conflicting descriptions made by the translators. Overall, the inclusion of frequency weights does not alter the results in any substantial direction.

31 Cf. Schaffer “Thin Descriptions,” 319–326.

32 For an overview of all languages, countries, original terms used to query the lexicon as well as all translated terms, see table A1 in online appendix.

33 For a more detailed data description and overview of data size, words, web-documents and URLs by language and country, see online Appendix A3.

34 Frankenberger and Buhr, “‘For Me Democracy Is … ’.”; Hu, “Popular Understandings in Contemporary China.”

35 König, Siewert, and Ackermann, “Conceptualizing and Measuring Preferences.”

36 Cf. Shin, “Assessing Responses”; Shin and Kim, “How Global Citizenries Think”; Frankenberger and Buhr “‘For Me Democracy Is … ’.”; Hu, “Popular Understandings in Contemporary China.”

37 See, in turn, Chapman et al., “Under the Veil”; de Regt, “Arabs Want Democracy,” 40; Crow, “The Party’s Over,” 46–47; Hernández, “Europeans’ Views of Democracy”; Lu and Shi, “Battle of Ideas,” 26; Chang et al., “Popular Value Perceptions,” 349; Zhai, “Popular Conceptions,” 249–250; Tessler et al., “New Findings,” 91–92; Oser and Hooghe, “Democratic Ideals,” 714–715.

38 Dalton, Shin, and Jou, “Understanding Democracy,” 145.

39 Quaranta, “The Meaning of Democracy,” 861.

40 See, e.g., Mattes and Bratton, “Learning About Democracy”; Dalton, Shin, and Jou, “Understanding Democracy”; Canache, “Citizens’ Conceptualisations.”

41 Dalton, Shin, and Jou, “Understanding Democracy,” 143–147.

42 Canache, “Citizens’ Conceptualisations,” 1141.

43 Ferrín and Kriesi, “The European Verdict”; Lu and Chu, Understandings of Democracy.

44 Cf. Berinsky, “Measuring Public Opinion with Surveys.”

45 Cf. Schaffer, Democracy in Translation; “Thin Descriptions.”

46 See, e.g., Schaffer, “Thin Descriptions”; Gillman, “Ideals Without Institutions”; Frankenberger and Buhr, “‘For Me Democracy Is … ’.”

47 Schaffer, “Thin Descriptions”; Hu, “Popular Understandings”; Brooks, Ngwane, and Runciman, “Decolonising and Re-theorising.”

48 Gillman, “Ideals Without Institutions,” 438, 435.

49 See, e.g., Schaffer, “Thin Descriptions”; Brooks, Ngwane, and Runciman, “Decolonising and Re-theorising.”

50 See, e.g., Verba, Schlozman, and Brady, Voice and Equality; Lu and Chu, Understandings of Democracy.

51 See, e.g., Crow, “The Party’s Over”; Zhai, “Popular Conceptions”; Lu and Shi, “Battle of Ideas”; Landwehr and Steiner, “Where Democrats Disagree”; Shin and Kim, “How Global Citizenries Think.”

52 See, e.g., Inglehart, “The Renaissance of Political Culture”; Kirsch and Welzel, “Democracy Misunderstood.”

53 See, e.g., Chang, Wu, and Weatherall, “Popular Value Perceptions.”

54 See, e.g., Tessler, Jamal, and Robbins, “New Findings”; de Regt, “Arabs Want Democracy.”

55 See, e.g., Shin, Dalton, and Jou, “Popular Conceptions of Democracy”; Ferrín and Kriesi, “The European Verdict”; Quaranta, “The Meaning of Democracy.”

56 Canache, “Citizens’ Conceptualisations,” 1135.

57 See, e.g., Welzel and Moreno Alvarez, “Enlightening People”; Shin, “Assessing Responses”; Ulbricht, “Perceptions and Conceptions”; Kirsch and Welzel, “Democracy Misunderstood”; Zagrebina, “Concepts of Democracy”; Davis, Goidel, and Zhao, “The Meanings of Democracy”; Lu and Chu, Understandings of Democracy.

58 Shin, “Assessing Responses,” 27.

59 Kirsch and Welzel, “Democracy Misunderstood,” 60.

60 Welzel, Freedom Rising.

61 Kirsch and Welzel, “Democracy Misunderstood,” 87.

62 See, e.g., Lu and Shi, “Battle of Ideas”; Ulbricht, “Perceptions and Conceptions”; Zagrebina, “Concepts of Democracy.”

63 Ulbricht, “Perceptions and Conceptions,” 1419.

64 See, e.g. Park, “Learning Politics”; Voltmer and Rawnsley, “The Media”; Andersen and Strömbäck, “Media Platforms.”

65 Nelson, Clawson, and Oxley, “Media Framing”; Norris, Democratic Deficit; McCombs and Valenzuela, Setting the Agenda.

66 Reuter and Szakonyi, “Online Media in Authoritarian Regimes”; Ulbricht, “Perceptions and Conceptions.”

67 Lu and Chu, Understandings of Democracy, 8.

68 Welzel, Freedom Rising, 5.

69 König, Siewert, and Ackermann, “Conceptualizing and Measuring Preferences.”

70 Dalton, Shin, and Jou, “Understanding Democracy,” n. 12.

71 Gillman, “Ideals Without Institutions.”

72 For a more detailed description of the coding, see A4 in online appendix.

73 Shin and Kim, “How Global Citizenries Think,” 235.

74 See, e.g., Shin and Kim, “How Global Citizenries Think”; Zhai, “Popular Conceptions”; Lu and Chu, Understandings of Democracy; Wu and Wu, “Regime types”.

75 Schumpeter, Capitalism, Socialism and Democracy.

76 See, e.g., Dahl, Polyarchy, Democracy and Its Critics.

77 Dahl, Dilemmas of Pluralist Democracy, 11.

78 See, e.g. Chapman et al., “Under the Veil.”

79 Dalton, Shin, and Jou, “Popular Conceptions”; Chang, Wu, and Weatherall, “Popular Value Perceptions”; Lu and Chu, Understandings of Democracy.

80 See, e.g., Lipset, “Some Social Requisites of Democracy”; Easton, “A Re-Assessment of Political Support.”; Marshall, Citizenship and Social Class; Bågenholm et al., Oxford Handbook.

81 See, e.g., Dalton, Shin, and Jou, “Understanding Democracy”; Canache, “Citizens’ Conceptualisations.”

82 See, e.g., Welzel, Freedom Rising; Lu and Shi, “Battle of Ideas”; Shin, “Democratic Deconsolidation.”

83 In our third coding round, we coded the data anew to verify the consistency and transparency in the coding.

84 Dahlberg, Axelsson, and Holmberg, “Democracy in Context.”

85 Schmitter and Karl, “What Democracy Is … ,” 76.

86 Rose, “Democratic and Undemocratic States.”

87 See, e.g., Frankenberger and Buhr, “‘For Me Democracy Is … ’,” 393.

88 See, e.g., Newton, “Political Support”; Lu and Chu, Understandings of Democracy.

89 See, e.g., Kriesi and Ferrín, How Europeans View Democracy; Zhai, “Popular Conceptions.”

90 See, e.g., Dalton, Shin, and Jou, “Understanding Democracy,” n. 15–16.

91 See, e.g., Gillman, “Ideals Without Institutions.”

92 See, e.g., Dalton, Shin, and Jou, “Understanding Democracy,” 144; Welzel, Freedom Rising.

93 See, e.g., Gillman, “Ideals Without Institutions.”

94 Norris, Democratic Deficit; Márquez-Ramirez et al., “Detached or Interventionist?”; Žuffová, “Fit for Purpose?.”

95 Chapman, “Under the Veil,” 16.

96 Dalton, Shin, and Jou, “Understanding Democracy”, “Popular Conceptions”; Shin, Dalton, and Jou, “Popular Conceptions.”

97 Dalton, Shin, and Jou, “Understanding Democracy,” 147.

98 Ibid., 154–155, n. 15–16.

99 When aggregating our democracy classifications from the lexicon, we weighted the terms by: (a) vector similarity (cosine similarity in lexicon); (b) neighbouring term frequency in lexicon; and (c) term frequency by country (since some countries are represented with up to three languages).

100 In the correlation between outcomes and social benefits, Indonesia ends up as a severe outlier with a Cook’s distance of 1.4. After a closer look into the data sources, we found that 63 percent of the respondents in the data from Dalton et al. “Popular Conceptions,” answered “Don’t know”. For this reason, we have excluded Indonesia in all three correlations. Hungary is another outlier with severe influence.

101 Welzel, Freedom Rising. For a detailed overview of the country classification into cultural regions, see Appendix 1.

102 See, e.g., Welzel, Freedom Rising, “Democratic Horizons”; Kirsch and Welzel, “Democracy Misunderstood”; Kruse, Ravlik, and Welzel, “Democracy Confused.”

103 Akaliyiski, Welzel, and Hien, “A Community of Shared Values?”

104 See, e.g., Lu and Shi, “Battle of Ideas”; Chang, Wu, and Weatherall, “Popular Value Perceptions”; Shin, “Democratic Deconsolidation”; Zhai, “Values Change and Support.”

105 Lührmann, Tannenberg, and Lindberg, “Regimes of the World.”

106 See, e.g., Ulbricht, “Perceptions and Conceptions”; Zagrebina, “Concepts of Democracy”; Lu and Shu, Understandings of Democracy.

107 At this abstract level, the results remain unaffected even after controlling for press freedom using the Freedom of the Press score (“Freedom House 2018”), which is constructed by combining ratings from four components: laws and regulations, political pressures and controls, economic influences, and repressive actions.

108 Ulbricht, “Perceptions and Conceptions.”

109 Chapman et al., “Under the Veil.”

110 When using the World Bank (World Development Indicators) data to adjust for the percentage of individuals in the population using the Internet, this variable did not account for the variations in abstract conceptions of democracy between countries.

111 Cf. Schaffer, “Thin Descriptions.”

112 See, e.g., Dalton, Shin, and Jou, “Understanding Democracy”; Cho, “Global Citizenries”; Kirsch and Welzel, “Democracy Misunderstood”; Lu and Chu, Understandings of Democracy.

113 See, e.g. Norris, Democratic Deficit.

114 Frankenberger and Buhr, “‘For Me Democracy Is … ’,” 393.

115 Shin, “Assessing Responses,” 25.

Appendices

Appendix 1. Country divisions based on cultural regions (Welzel 2013)

Table

Catholic Europe		Latin America
Country	Frequency	Country	Frequency
Andorra	30	Argentina	32
Austria	30	Bolivia	28
Belgium	65	Brazil	60
Cyprus	60	Chile	32
France	30	Colombia	30
Greece	30	Costa Rica	29
Italy	30	Cuba	31
Luxembourg	57	Dominican Republic	31
Portugal	30	Ecuador	30
Spain	62	El Salvador	30
Total	424	Guatemala	30
		Honduras	31
		Mexico	32
China	60	Nicaragua	30
Hong Kong	52	Panama	30
Japan	2	Paraguay	31
Taiwan	43	Peru	30
Thailand	27	Puerto Rico	30
Vietnam	30	Uruguay	31
Total	214	Venezuela	32
		Total	640
English West
Australia	30	North Africa and Middle East
Canada	62	Algeria	30
Ireland	30	Egypt	59
New Zealand	30	Iran	14
United Kingdom	30	Israel	30
United States	60	Morocco	39
Total	242	Tunisia	30
		Turkey	35
Ex-Communist East		Total	237
Armenia	30
Azerbaijan	29	Protestant Europe
Belarus	32	Denmark	30
Bulgaria	30	Finland	59
Kazakhstan	31	Germany	63
Kyrgyzstan	29	Netherlands	32
Moldova	57	Norway	30
Romania	30	Sweden	39
Russia	62	Switzerland	90
Ukraine	62	Total	343
Total	392
		South Asia
Ex-Communist West		Brunei	28
Czech Republic	30	India	30
Estonia	30	Indonesia	60
Hungary	30	Malaysia	59
Latvia	29	Pakistan	29
Lithuania	32	Philippines	30
Poland	32	Singapore	30
Slovakia	30	Sri Lanka	30
Slovenia	29	Total	296
Total	242
		Sub-Saharan Africa
		Cameroon	30
		Gabon	29
		Ghana	29
		Kenya	59
		Mali	30
		Namibia	20
		Nigeria	30
		Senegal	30
		Somalia	21
		South Africa	59
		Zambia	28
		Zimbabwe	30
		Total	395

Comment: The division is based on a categorisation made by Welzel (2013). The frequencies denote the number of terms per country. 15 terms from editorial media and 15 from social media. For some countries a few terms have been coded as noise or error terms, which means that they do not quantify up to 30, and for some countries the translators have disagreed on the translation, which means that they can exceed 30 terms. For approximately 19 countries, we have two languages and for Switzerland even three, which means that we have 60 (90) terms for these countries.

Appendix 2. Country divisions based on regime types (Lührmann et al. 2018)

Table

Liberal Democracy		Electoral Democracy		Autocracy
Country	Terms	Country	Terms	Country	Terms
Andorra	30	Argentina	30	Algeria	30
Australia	30	Bolivia	30	Armenia	30
Austria	30	Brazil	60	Azerbaijan	28
Belgium	65	Bulgaria	30	Belarus	32
Canada	62	Chile	30	Cameroon	30
Costa Rica	29	Colombia	30	Egypt	59
Cyprus	60	Dominican Republic	31	Gabon	29
Czech Republic	30	Ecuador	30	Honduras	31
Denmark	30	El Salvador	30	Iran	15
Estonia	30	Ghana	29	Kazakhstan	31
Finland	59	Greece	30	Malaysia	59
France	30	Guatemala	30	Nicaragua	30
Germany	63	Hungary	30	Pakistan	29
Ireland	30	India	30	Russia	62
Italy	30	Indonesia	60	Singapore	30
Latvia	29	Israel	29	Turkey	35
Luxembourg	57	Kenya	59	Ukraine	62
Namibia	20	Kyrgyzstan	29	Venezuela	31
Netherlands	32	Lithuania	32	Zambia	28
New Zealand	30	Mali	30	Zimbabwe	30
Norway	30	Mexico	32	Brunei	28
Portugal	30	Moldova	57	China	60
Slovenia	29	Nigeria	30	Cuba	31
Spain	61	Panama	30	Hong Kong	52
Sweden	39	Paraguay	31	Morocco	39
Switzerland	90	Peru	30	Somalia	20
Taiwan	42	Philippines	30	Thailand	27
United Kingdom	30	Poland	32	Vietnam	30
United States	60	Puerto Rico	30	Sum:	998
Uruguay	31	Romania	28
Sum:	1218	Senegal	30
		Slovakia	30
		South Africa	59
		Sri Lanka	30
		Tunisia	30
		Sum:	1198

Appendix 3.

A. Dimensions of democracy across cultural regions

Figure a1. Dimensions of democracy across three different cultural regions (relative frequency within regions)

Figure a2. Dimensions of democracy as governance across three different cultural regions (relative frequency within regions)

Figure a3. Dimensions of democracy as values across three different cultural regions (relative frequency within regions)

B. Dimensions of democracy across regime types

Figure a4. Dimensions of democracy in three different regime types (relative frequency within regime types)

Figure a5. Dimensions of democracy as governance in three different regime types (relative frequency within regime types)

Figure a6. Dimensions of democracy as values in three different regime types (relative