Simulating pluralism: the language of democracy in hegemonic authoritarianism

ABSTRACT This article analyses the language authoritarian leaders use to legitimate their rule. It examines the official speeches of autocrats in hegemonic regimes and compares them to the rhetorical styles of leaders in closed or competitive regimes and democracies. While recent autocracy research has drawn most attention to the phenomenon of competitive authoritarianism, the survival strategies of hegemonic regimes are less explored. Thus, the study focuses on the simulation of pluralism as a key feature of hegemonic regimes. By installing non-competitive multiparty systems which merely pretend pluralism, these regimes maintain a strong grip on power. The study finds that the leaders of hegemonic regimes use a surprisingly democratic style of language to sustain this façade of pluralism. The dictionary-based quantitative text analysis of 2074 speeches of current leaders in 22 countries illustrates that compared to other autocracies, hegemonic regimes overemphasize the (non-existing) democratic procedures in their country to fake a participatory form of government and gain national and international legitimacy. The subsequent case studies of Uzbekistan, Saudi Arabia, and Russia further reveal the differences in context and motives for autocrats in hegemonic, closed, and competitive regimes to use autocratic or democratic styles of language.


Introduction
Does it matter what autocrats say? Recent contributions point to the linkages between the language style of authoritarian leaders and the survival of their regimes Windsor, Dowell, and Graesser 2015;Windsor et al. 2018). This article adds to this nascent strand of literature and studies the language authoritarian leaders use to legitimate their rule on national and international level. More concretely, I look at the rhetorical differences in the official communication of closed, competitive, and hegemonic authoritarian regimes.
Hegemonic authoritarian regimes are frequently seen as a residual category, comprising all those regimes which do not classify as competitive or closed autocracies (Donno 2013). Furthermore, the mechanism of how these regimes manage to survive remains nebulous since recent studies have drawn most attention to the phenomenon of competitive authoritarianism (Levitsky and Way 2010;Schedler 2013;Handlin 2016Handlin , 2017. While closed regimes have no elected legislative or a single-party rule, competitive regimes allow for several parties to compete during elections. In contrast to this, hegemonic regimes typically install non-competitive multiparty systems which merely pretend pluralism 1 while the regime maintains a strong grip on power (Brownlee 2009).
This article argues that the leaders of hegemonic regimes use a purposely democratic style of language to justify and perpetuate their non-competitive multiparty regimes. The rhetorical strategy of talking like democrats and faking a participatory style of government helps them to promote elite cohesion, gain national and international legitimacy and thereby prolong their rule. Interestingly, I find that this linguistic strategy of hegemonic regimes is systematically different to the official communication styles in closed or competitive regimes: Although politically more open, competitive regimes talk less about democracy. The findings further reveal that the leaders of closed regimes hardly refer to democracy in their speeches.
The empirical examinations in this article compare the language of political leaders by applying quantitative and qualitative text analysis. Due to the general problems of data availability in autocratic regimes, the quantitative text analysis is limited to those autocracies which provide on their official websites a collection of translated English speeches delivered by their leaders. This trend has spread out significantly during the last decade and different kind of regimesbe it Azerbaijan with its strong economic linkages to the West or isolated North Koreamaintain such websites to appear modern, attract foreign investment and gain international legitimacy. Typically, the translated and uploaded material is a good mix of speeches in front of very diverse audiences (national and international occasions). Overall, the text corpus for the dictionary-based logit scaling (Laver and Garry 2000;Lowe et al. 2011) comprises 2074 speeches of the leaders in 22 countries. The analysis focuses on autocracies but includes also a selection of speeches delivered by current leaders of democracies to better illuminate the differences between autocratic and democratic styles of language. The subsequent case studies are a qualitative and contextualized assessment of the language used by the leaders of hegemonic Uzbekistan, closed Saudi Arabia and competitive Russia to further highlight the rhetorical differences among autocrats.
The article proceeds as follows: The next section provides a conceptualization of autocratic and democratic styles of language. Subsequently, I elaborate on the simulation of pluralism as the defining feature of hegemonic authoritarianism and, based on this, formulate hypotheses for the analysis of the language of political leaders. After summarizing the major findings of the twofold empirical examinations, I consider alternative explanations for the results. The conclusion discusses broader implications of the findings and provides some recommendation for future research.

Autocratic and democratic styles of language
Recent autocracy research highlights the importance of legitimation for the survival of authoritarian regimes (Gerschewski 2013;Dukalskis 2017;Dukalskis and Gerschewski 2017;Schneider and Maerz 2017). Following a non-normative understanding of legitimation as the rulers' efforts to make the population believe in their legitimate authority (Weber 2002), most of these novel contributions focus on this process of institutionalizing persuasion. 2 Language is a powerful tool in this regard and autocracies organize spectacular events for spreading the word of their rulers (e.g. Adams 2010). In the age of the Internet, an increasing number of authoritarian regimes translate these speeches into English and publish them on shiny official websites to further strengthen their legitimacy (Maerz 2016). While the uploaded speeches are a highly diverse selection of transcripts from all kind of national and international occasions (economic summits, jubilees, national holidays, diplomatic receptions, etc.), such websites target primarily an international audience and serve the purpose of gaining mostly external legitimacy.
Several contributions assess public discourses and the language of autocrats in one or a small number of cases (March 2003;Megoran 2008;Omelicheva 2016;Maerz 2018). However, there is only little research about the rhetoric of authoritarian regimes which is based on large quantities of texts. The recently expanded toolbox of quantitative text analysis provides innovative approaches for analysing political texts (Grimmer and Stewart 2013) yet, this burgeoning literature has been mainly focused on democratic settings. Exceptions are Dowell, Windsor, and Graesser's (2015) and Windsor, Dowell, and Graesser's (2015) study of the language of Mao Tse-Tung, Fidel Castro, and Hosni Mubarak during natural disaster crises or Windsor et al.'s (2018) analysis of leaders' language and political survival strategies in the MENA region. By applying computational linguistic analysis, these contributions investigate about the leaders' mental constructs from a psychological perspective. Dowell, Windsor, and Graesser (2015) and Windsor, Dowell, and Graesser (2015), for example, illustrate how autocrats use more complex and formal language during crises to demonstrate authority and leadership strength. In addition, Windsor et al.'s (2018) analysis proposes that those leaders who use a more positive language and reduce blame and anxiety in their speeches have higher chances to survive times of political instability.
This article continues these first and fruitful explorations of how language can contribute to authoritarian persistence and finds that there are crucial variations among autocratic regimes concerning their use of formal languageparticularly if it concerns the leaders' claims to legitimacy. I distinguish between competitive, hegemonic and closed authoritarian regimes, as will be discussed further in the next section. The main argument of the article is that the leaders of hegemonic regimes maintain a strong rhetoric linkage to democratic procedures as their legitimacy foundation whereas other leaders of authoritarian regimes refer to more traditional sources of legitimacy in their speeches and apply a rigorously authoritarian style of language. While it is rather obvious that the leaders of hegemonic regimes also rule in an authoritarian manner, they refer to their country as democratic and speak strikingly often about reforms and democratic institutions. It is the goal of the article to further illuminate these different rhetorical patterns and thereby contribute to a better understanding of the varying mechanisms of legitimating authoritarian rule.
To analyse the rhetorical strategies of political leaders, I broadly distinguish between autocratic and democratic styles of language. The general ambiguity of language poses challenges when defining such categories. Autocrats make use of both concepts of language in varying degrees. The leaders of democracies also occasionally draw on autocratic categories of language, in particular populist and overly conservative ones. Therefore, measures of autocratic and democratic styles of language are fuzzy in this regard because they do not necessarily reflect the de-facto degree of authoritarianism or democracy. However, since this article does not aim at measuring degrees of authoritarianism but rather focuses on how different autocrats make strategic use of language, this twofold division is meaningful for the analysis. Figure 1 illustrates the multilevel concept of language which I construct as a framework to study the language of political leaders. By following the basic guidelines of Goertz (2006) for concept formation, the two categories of the first level represent the mutually exclusive antipodes of autocratic versus democratic language. The second level contains more fine-graded categories to distinguish whether the political leaders emphasize autocratic or democratic procedures and illiberal or liberal ideological orientations.
The theoretical differentiation between procedures and ideological orientations is meaningful because it accounts for the historical tension between democracy and liberalism and the concluding varieties of democracy. The question of how to define democracy is as old as the concept itself. Minimalist accounts of democracy merely demand for meaningful elections as a necessary requirement (cf. Schumpeter 1947). More comprehensive understandings of the term argue that without a government's adherence to liberal rights and principles such as freedom of speech, association and information or the rule of law, the electoral component of democracy becomes insignificant (Dahl 1971). Recent trends of autocratization illustrate that it is first and foremost these liberal aspects of democracy which are currently under assault . While neither the leaders of backsliding regimes such as Hungary or Poland nor long-standing autocrats openly refute democracy as such, they put renewed effort in defining and communicating their illiberal notions of the term (Buzogány 2017;Kirsch and Welzel 2018;Maerz and Schneider, forthcoming).
In line with this, current classifications of democracy show that while the compliance with liberal principles is a necessary condition for liberal democracy, the so-called electoral democracy type does not satisfy such liberal principles (Coppedge et al. 2018;Lührmann, Tannenberg, and Lindberg 2018, 63). This distinction between a regime's adherence to either liberal or illiberal principles is also reflected in my measurement of autocratic and democratic styles of language: the theoretical subdivision into democratic and autocratic procedures and liberal and illiberal ideological orientations is able to illustrate if political leaders who talk about democratic procedures have also internalized liberal values or are merely using their references to seemingly democratic procedures to maintain an illiberal regime.
The key words in each of the subcategories (autocratic and democratic procedures, liberal and illiberal ideological orientations) were selected to provide a brief and intuitive version of this article's understanding of the four concepts and serve as semantic orientation for building the dictionary for the analysis. Thus, the subcategory of autocratic procedures is emphasized by a comparatively frequent usage of terms which represent a strict and uncontested maintenance of power (e.g. defense, military, monarchy). Authoritarian law and order mean, for example, that autocrats construct a discourse of danger which condemns opponents as 'evil' and constantly threatens them with penalty, prosecution, and prison. Illiberal ideological orientations are expressed by overemphasizing traditional values, paternalism, and nationalismtypically done by a range of autocrats but currently also a distressing trend among Western right-wing populists. As opposed to this, the first subcategory of a democratic style of language implies that political leaders stress liberal values, the rights of women and minorities or human rights in general. This subcategory, however, is only concerned with political liberalism and does not refer to economic liberalism because liberal democracies typically display different economic models with varying economic liberties. The second subcategory of a democratic style of language captures if political leaders speak about the (seemingly) democratic procedures and implemented reforms in their (authoritarian) country. The section on the construction of the dictionary further highlights the differences between these two styles of language and provides word examples.
Based on this concept of language, the dictionary-based analysis of autocratic and democratic styles of language in this article is a straight-forward assessment of how often and when political leaders emphasize certain political procedures and ideological orientation. In this sense, my text-analytic approach is similar to those quantitative text analyses which examine party manifestos and speeches in democracies to infer policy positions on a left-right scale (e.g. Laver and Garry 2000;Laver, Benoit, and Garry 2003;Lowe et al. 2011;Proksch and Slapin 2012;Martocchia Diodati, Marino, and Carlotti 2018). Several of these left and right classifiers, particularly concerning liberal and traditional values, are also reflected in the concept of democratic and autocratic styles of language. Yet, due to the inclusion of autocratic settings, my analysis comes along with different assumptions about the intention of the speeches in authoritarian or democratic contexts.
One crucial difference between the speeches in democratic and autocratic contexts is that the latter follow a more distinct purpose of legitimation. Contrary to authoritarian regimes, the legitimacy of democracies relies on a range of concrete procedures and transparent processes and is not merely a façade controlled by the regime. The speeches of political leaders in democracies might also aim at strengthening the people's belief in the legitimacy of the government. Yet, their main intentions are to communicate diplomatic goals to an international audience or policy positions to potential voters and electioneer in a competitive struggle over political influence. Based on this, I argue that with decreasing degrees of electoral competition and growing authoritarianism, the leaders' speeches are less about policy positions and more about legitimating authoritarianism.

Simulating pluralism in hegemonic regimes
This article classifies autocracies in competitive, hegemonic, and closed regimes. Compared to other regime classifications which are based on discrete data (Cheibub, Gandhi, and Vreeland 2010;Wahman, Teorell, and Hadenius 2013;Geddes, Wright, and Frantz 2014), this categorization refers to continuous data and helps to distinguish the fully closed regimes from their more open counterparts. As illustrated in Figure 2, hegemonic and competitive regimes are generally seen as sub types of electoral authoritarianism which is situated between democracy and closed autocracy. Closed autocracies have no elected legislature or rule with a single-party regime. Competitive regimes allow for at least a minimum amount of real competition during multiparty elections. Hegemonic regimes are frequently seen as a residual category which includes all those regimes which do not classify as competitive or closed autocracies (Brownlee 2009;Donno 2013). This article zooms in on hegemonic authoritarianism and provides deeper theoretical reflections and new empirical findings about how this hitherto under-studied regime type manages to survive.
If compared to competitive regimes, the most crucial difference and also defining feature of hegemonic regimes is that even though they hold regular elections, they do not allow for real electoral competition, hence there is no meaningful level of contestation (Brownlee 2009, 524). Instead, they exhibit a hegemonic party system (Sartori 2005, 205-206) which is either clearly dominated by the party of the ruling regime or an artificial arrangement of several satellite parties to pretend contestation and pluralism.
Cameroon's ruling party, called the People's Democratic Movement, is one example of a hegemonic party next to which officially allowed opposition parties have no real chance to compete (Pelizzo and Nwokora 2016). Uzbekistan set up several satellite parties which have impressively democratic names, e.g. People's Democratic Party of Uzbekistan or the Liberal Democratic Party. At first sight, this multiparty system makes indeed the impression that the regime shares some of its power. Yet, at second sight, the agendas of these government-friendly parties hardly differ and it is rather evident that they have been installed by the regime as a farce while keeping full control over the party system and marginalizing any real opposition (Bader 2009).
How do hegemonic regimes manage to maintain these toy parliaments and rubber stamp politics over longer periods of time? As recent contributions highlight, authoritarian rulers rely not only on repression but institutionalize sophisticated strategies of cooptation and legitimation to ensure their survival (Gandhi 2008;Gerschewski 2013). Besides strictly limiting civil and political rights, hegemonic authoritarian regimes install non-competitive multiparty systems to co-opt strategic elites as loyal members of their own strictly controlled parties or grant them the privileges of official posts in other broadly impotent parties. In both cases, such institutional inclusion is not about a transfer of power but merely about simulating selected individuals the prospects of political participation while tying their fate to the regime's survival (Sartori 2005, 205-206). To credibly promote these non-competitive multiparty systems as seemingly pluralistic institutions and make their authority look legitimate, I argue that hegemonic regimes strategically use a democratic style of language. Speaking like democrats is one tool of legitimation which they use to disguise the fully fledged authoritarian nature of the regime and sustain the illusory façade of pluralism. Thus, I expect to find an inflated use of democratic terms in the speeches of hegemonic regimes. Based on this, I formulate a first hypothesis for the analysis of the official communication of hegemonic, closed or competitive regimes. H1: The leaders of hegemonic regimes use a democratic style of language in their official speeches.
Using democratic terms in their official speeches is not an exclusive characteristic of the leaders of hegemonic regimes. Yet, I assume that there are crucial variations in frequency and context of using such terms in hegemonic, closed and competitive regimes. I expect that the leaders of closed regimes hardly deploy them in their speechessimply because they do not set up non-competitive multiparty systems which need to be communicated as being democratic institutions. In their study of authoritarian claims to legitimacy, von Soest and Grauvogel (2017, 298) illustrate that closed regimes rely mostly on foundational myths and also ideologies as legitimacy foundations. As their name says, these regimes are politically closed and based on this, I expect their rhetorical strategies to focus on legitimating their heredity lines and other autocratic procedures of power maintenance: H2: Closed regimes hardly refer to a democratic style of language in their official communication.
Compared to hegemonic and closed regimes, competitive regimes are politically more open and closer to democracies. Hence, one could assume that the leaders of competitive regimes also use the most democratic style of language among all autocrats. Yet, I expect that the speeches of competitive regimes reveal less democratic terms than those of hegemonic regimes because competitive regimes invest not as much resources and rhetorical means in faking pluralism. As von Soest and Grauvogel (2017, 298) show, competitive regimes mostly rely on procedures and performance when formulating their claims to legitimacy. However, instead of broadly overemphasizing democratic procedures, I argue that their rhetorical strategy has to be more specific. Rather than simulating pluralism, they are forced to deal with the repercussions of allowing minimal amounts of real electoral competition and emphasize in their official communication those topics ('performance') and policies ('procedures') which help them to maintain the status quo.
H3: The leaders of competitive regimes use less democratic terms than those of hegemonic regimes.
To empirically test these hypotheses, the different regime categories need to be operationalized. Recent literature on regime classifications engages with the challenges of distinguishing between competitive authoritarianism and early forms of democracy (Bogaards 2012;Handlin 2017). To avoid ambiguity, this article relies on the rather minimal definition of autocracy by Geddes, Wright, and Frantz (2014, 317) which is based on the electoral process. 3 Concerning the borderlines between the three autocratic regime types, non-competitive multiparty systems are generally seen as a defining feature of hegemonic regimes (Sartori 2005;Brownlee 2009). Yet, operationalizations of this regime type have been rather coarse and classified several regimes which simulate pluralism as closed or competitive regimes, as I explain in more details in Table A1 in the Supplemental Material. Table A1 further illustrates how I follow Brownlee (2009, 524) and make use of the Database of Political Institutions (DPI) indicators on competitiveness in legislative and executive elections (Beck et al. 2001) but suggest two crucial refinements of the operationalization. The section about data and case selection provides further information about how many cases in each of these regime types the analysis refers to.

Comparing the language of political leaders
Methods and dictionary I use quantitative and qualitative techniques of text analysis to examine the language of autocrats. Given the novelty of analysing large quantities of speeches from authoritarian contexts, this mixed-method approach is most suitable and enables me to cross-validate the results. It is also in line with one of the basic principles of text analysis suggested by Grimmer and Stewart (2013, 4): automated text analysis can augment, but not replace thorough reading of texts. The quantitative dictionary-based analysis of 2074 speeches is preceded by the careful construction and validation of the dictionary and succeeded by qualitative case studies.
The dictionary comprises 241 terms and is based on the additive concepts of language in Figure 1: a democratic style of language is measured by word frequencies in the subcategories of democratic procedures and liberalism. The measurement of an autocratic style of language refers to the frequencies in the subcategories of autocratic procedures and illiberalism.
Generally, the dictionary contains key words with as little vagueness as possible. Some of the terms 4 were adopted from Laver and Garry's (2000) dictionary for estimating policy positions from political texts. Yet, because the validity of a dictionary for quantitative text analysis is highly context-specific (Maerz and Puschmann, forthcoming), I collected the large majority of the terms during the qualitative assessment of a stratified sample of speeches 5used as a pool of key wordsand with the help of recent literature on legitimation patterns and official rhetoric in authoritarian regimes (Megoran 2008;Matveeva 2009;Pruzan-Jørgensen 2010;Omelicheva 2016;von Soest and Grauvogel 2017;Maerz 2018). To validate the dictionary, I qualitatively checked the context of all codings in a sample of 110 speeches. 6 This robustness test allowed me to identify and exclude frequent negations and ambiguous terms and thereby ensures that the assigned terms in the dictionary do indeed measure the proposed concepts. 7 Examples of those terms which measure the emphasis of democratic procedures include the stemmed words democra*, election*, fair*, parliament*, reform*, transparen*, or vote*. The category on liberalism comprises evident terms such as free*, liberal*, tolera*, but includes also words such as discriminat*, authoritarian*, or repressi* since these terms are frequently used to demand for more political and civil rights. Following this logic vice versa, anarch*, chaos, destabili*, and obscen* or pervert* are in the categories of autocratic procedures and illiberalism. Other terms for autocratic procedures are defen*, protect*, securit* or stabil*, for illiberalism discipline, moral*, tradition*, patriot* or pride. 8

Data and case selection
This article studies the language of autocrats by looking at their publicly delivered speeches. A growing number of autocratic regimes sets up official e-government websites to propagate modernity and thereby gain national and international legitimacy (Maerz 2016). The case selection for the analysis is guided, first, by the installation of such websites and second, by the number of the uploaded and officially translated English speeches of the regimes' current leaders. To qualify as selected case, there had to be at least 10 translated speeches on the website of the respective regime. 9 This 2-step procedure of case selection resulted in a collection of 2074 speeches from the leaders of 22 countries, as shown in Table 1. The democratic cases were randomly selected by data availability. I included them primarily for illustration purposes and to contrast the language styles of democrats with those of autocrats. I also included Turkey and Hungary as so-called backsliding cases (Greskovits 2015;Blockmans and Yilmaz 2017) because I was curious whether and how their recent reverting to authoritarian practices is reflected in their official communication.
Due to the general problems of data availability for authoritarian contexts, a rather obvious constraint concerning data selection is that the collection of speeches is limited to those which are made publicly available in English on the official website of authoritarian regimes. This could bear the potential of selection bias. One problem, for example, could be that it is only certain types of regimes, countries or leaders that install such websites and upload their speeches in English. However, as Maerz (2016) illustrates, the habit of maintaining professional websites to gain internal and external legitimacy is not a  Table A2 in the Supplemental Material provides the sources of the speeches. c The cases of Saudi Arabia and Uzbekistan include also several speeches of the new leaders since 2015 and 2016 because leadership succession was not followed by regime change. d The speeches cover Rasmussen's I., II. and III. cabinet (2009-2011 and 2015-2017). unique characteristic of one particular regime type but rather a general trend in modern authoritarianism. Moreover, also if alternative classifications of authoritarian regimes in military, monarchic or party-based types (e.g. Geddes, Wright, and Frantz 2014) are applied, there is no regular pattern observable. The same is true concerning types of leaders: the selected regimes are ruled by various kinds of leadership groups. While several authoritarian regimes which offer numerous speeches on their website are resource-rich countries, other regimes such as Uzbekistan and Tajikistan illustrate that such country-specific characteristics vary among the selected cases.
Beside this, another potential selection bias could be that it is only a certain type of speeches which are made available on the website of authoritarian regimes. Yet, as the mentioned robustness tests with a random sample of the selected speeches in each case have shown, the collection of texts is a diverse mixture of speeches delivered at various national and international occasions. Hence, they address either a national or international audience (or both) and deal with a range of economic and political issues and are therefore representative for assessing the general language styles of political leaders.
The speeches are of varying number, length, and quality and cover diverging periods of time. Yet, while the summary of the findings provides confidence intervals for the estimated positions in each case, I broadly neglect these differences of range and time coverage for building the corpus because I assume that a regime consciously creates its 'electronic face' and thereby controls its perceived image in the Internet. In this sense, the analysis captures a double process of legitimation which addresses a national or international public (or both): the actual text of a speech delivered at some real-life national or international occasion and the subsequent selection of speeches published online by the respective regime. The corpus of 2074 speeches 10 was cleaned of punctuation, numbers and separators. Words were set to lowercase and stemmed before the frequencies of the (stemmed) dictionary terms were counted.

Statistical model
To get from the 'raw' counts of the dictionary application to a sound measurement of relative frequencies of autocratic and democratic styles of language, I adopt Lowe et al.'s (2011) spatial model of logit scaling. 11 I chose this model because their approach has no predefined endpoints, accounts for the relative difference between opposed categories, works also well for aggregated categories and is generally superior to earlier methods of scaling, as the comparisons of Lowe et al. (2011) illustrate. They apply the model to measure the relative balance of policy positions on a left-right scale with a logistic function of word counts attached to the positive and negative (left and right) side of the dimension. In this article, I assign the positive ends to democratic and the negative ends to autocratic styles of language. I denote the number of scores for democratic emphasis D and for autocratic emphasis A. The output u is an estimate of the position on the autocratic-democratic scale. Drawing on Lowe et al.'s (2011) empirical logit scale, this is defined as To improve the model structure, Lowe et al. (2011) add 0.5 to all counts. This is a standard statistical practice which can also be used as a measure to reduce bias in the case of estimating the proportions of categories. As explained by Lowe et al. (2011, 132), this makes particularly those estimates more stable which are derived from rather small counts. The logit is therefore expressed as The application of the model in this article comes with three constraints. First, the resulting scale does not account for differences in the language of autocrats over time. Yet, this is rather a problem of data availability since there are simply not enough speeches available in each case for measurements per year. Second, the model does not account for the varying size of data in each case, influencing the robustness of the respective results. Thirdly, I expect the word frequency counts to be affected by random noise and other bias. Due to the latter two constraints, I adopt Lowe et al.'s (2011, 134) simple Bayesian approach of computing standard errors and confidence intervals for the estimates of u. Given that a standard Beta prior over the proportions of A and D words with parameters of a A = a D = a implies a posterior distribution over position, this is approximated for A + D ≥ 10 as As Lowe et al. (2011, 134) explain, setting a = 0.5 corresponds to a symmetrical invariant Jeffreys prior, here over autocratic and democratic styles of language. This suggests the following 95% confidence interval: 12 [u D − 1.96s, u D + 1.96s] Figure 3 illustrates the findings of the dictionary-based logit scaling of 2074 speeches by political leaders from 22 countries. It shows the estimated position for each case on the scale of autocratic to democratic styles of language as negative to positive ends of both axes. Following the concepts of languages in Figure 1, the x-axis measures the leaders' emphasis of liberalism or illiberalism, the y-axis the leaders' emphasis of democratic or autocratic procedures. Similar to Medzihorsky, Popovic, and Jenne (2017, 6), I integrated the 95% region of the estimated confidence intervals as lightly colored ellipses around each case.

Quantitative analysis
It is striking that almost all hegemonic regimes in this analysis emphasize democratic procedures in a comparatively high degree. With the slight deviations of Singapore and Cameroon, they outperform all the other inspected cases of authoritarianism. Uzbekistan is at the very top of the scale, closely followed by Kazakhstan, Jordan, Tajikistan and Azerbaijan. Their scores of speaking about democratic procedures compare to those of democracies or even outrun them. Yet, a crucial difference is that the language of hegemonic regimes clearly accentuates illiberalism whereas the democrats in this analysis stress liberalism. While these findings confirm my first hypothesis and show that the leaders of hegemonic regimes overstate the talk about democracy, they also highlight that this language of democracy in non-democratic surroundings typically lacks elements of liberalism.
Regarding the closed regimes in this analysis, it is not surprising that in contrast to the other cases, they are all below the zero level of emphasizing democratic procedures or liberalism. They legitimate their authority by constantly emphasizing that they provide for stability, order, and unity. They stress traditional values and thereby justify their paternalist style of ruling, promote nationalism or refer to religion and remind the people that they see themselves called by God. This supports my second hypothesis: intensively talking about democracy is neither a plausible nor a particularly effective strategic tool in the longrunning monarchies of Saudi Arabia, Kuwait and United Arab Emirates. The isolated regime of North Korea also seems to draw mostly on other rhetorical strategies. While Saudi Arabia has the highest scores in stressing autocratic procedures in its official communication, it is North Korea which makes use of the most illiberal language in this analysis.
The findings concerning the competitive regimes in this analysis suggest that their leaders are talking about democracyyet, as assumed in the third hypothesis, in more moderate dimensions than hegemonic regimes. Uganda, Morocco, and Malaysia show average performances concerning the democratic procedures dimension. Due to a comparatively small amount of speeches from the current Malaysian prime minister Dato' Sri Mohd Najib, this distinct case of competitive authoritarianism 13 is based on a relatively coarse estimate. Yet, the more robust findings for Russia, the other typical case of competitive authoritarianism in this analysis, suggest that Putin is hardly talking about democratic procedures or liberal values if compared to hegemonic regimes. This is a counter-intuitive finding because hegemonic regimes are generally more closed, isolated and often also more repressive than competitive regimes.
The democratic cases in this analysis score as one would expect for the speeches delivered by the leaders of consolidated democracies: situated in the positive area of both axes, they speak explicitly of democratic procedures and stress liberal values. Interestingly, the two backsliding cases of Hungary and Turkey deviate from this pattern. Their reverting to less open forms of government manifests in the speeches of Viktor Orban and Recep Tayyip Erdoğan with illiberal styles of language which are similar to those of several authoritarian regimes.
It is generally noticeable that illiberalism outweighs in all of the observed authoritarian regimes. While Uganda is the only case which is closest to positive scores on liberalism, Juche-driven North Korea and religious Saudi Arabia have strikingly negative values on this dimension. Furthermore, all regimesalso the closed onesperform better in emphasizing democratic procedures than speaking about liberal values. This indicates that autocrats feign less about freedom, openness, equality and transparency but rather speak about seemingly democratic elections and reforms in their country. One reason for this could be that it is much more challenging to fake liberalism in illiberal surroundings. Hence, the illiberalism-liberalism scale can be seen as a rough predictor of authoritarianism and democracy, while the scale of democratic versus autocratic procedures points to the hypocritical cases of hegemonic authoritarianism which vapour about democracy while perpetuating a full-blown authoritarian rule.

Qualitative analysis
The qualitative analysis has the purpose to further illustrating the difference in language styles among autocrats by looking at the context of frequently used terms in the official communication of Uzbekistan, Saudi Arabia, and Russia. I chose these cases because they represent each of the authoritarian regime types: Uzbekistan is a hegemonic regime, the monarchic regime of Saudi Arabia is politically closed, Russia holds regular elections which are generally considered as (minimally) competitive (Beck et al. 2001;Hyde and Marinov 2012). The analysis assesses excerpts of the regimes' speeches with the help of Yoshikoder's key-word-in-context tool (Lowe 2015). This tool extracts for each term of the dictionary the five precedent and subsequent words in the respective texts, allowing to access the background of recurring terms and, based on this, select relevant text passages for further inquiries.

The Language of democracy in hegemonic Uzbekistan
Authoritarian Uzbekistan refers remarkably frequent to democratic procedures in its official communication. The qualitative analysis provides details about this intense rhetorical boasting of democracy in national and international speeches. Islam Karimov, the longrunning but in 2016 deceased president of independent Uzbekistan, constantly stresses in his speeches the regime's apparent aim of building a democratic state, intensifying the democratic transformation and 'consolidating democratic values in the minds of the people'. 14 However, similar to the justification patterns in other post-Soviet regimes (Roberts 2015, 154), he keeps pointing out that the country follows its 'own model of democratization', dispelling any critique about the failure to achieve this goal. After the death of Karimov in 2016, the regime managed a smooth leadership succession and installed Shavkat Mirziyoyev as president. Loyal to his predecessor, he praises the democratic reforms which have been apparently accomplished under Karimov's rule and likewise claims to be committed to 'building a free, democratic, humane state'.
Another iterative element in the official communication of the regime is the alleged democratic electoral process. Both leaders talk about their 'meetings with voters' and thereby pretend a citizen-orientated style of governance. Political parties in Uzbekistan exist at the pleasure of the president, share similar agendas and have no real influence in the personalist regime (Bader 2009;Beacháin and Kevlihan 2015). Yet, both leaders claim the elections to be guided by democratic principles. Despite the ridiculously high election outcomes of the last decades (Beacháin and Kevlihan 2015, 501), the regime speaks of competitive campaigns. Mirziyoyev comments for the most recent electoral results that 'all foreign observers openly and unanimously acknowledged that the elections were conducted in the atmosphere of sound competition and struggle among political parties'. This statement is in sharp contrast to the OSCE's final report about these elections which repeatedly finds a lack of competitiveness and lists a plethora of serious concerns about undue limitations on fundamental freedoms of expression, association and assembly during the electoral process (OSCE 2016, 1-2). However, as also boldly reported in local newspapers (Tashkent Times 2016), other international organizations such as the Shanghai Cooperation Organization (SCO), known for promoting authoritarian norms in Central Asia (Ambrosio 2008), support this pretense of democracy.
Repeatedly using a few selected terms such as 'democratic', 'free', and 'liberal' is a striking characteristic of Uzbekistan's linguistic strategy. Both presidents apply these buzzwords with great formulaicityas if the mantra-like repetitions would belie the national and international audience about the fact that Uzbekistan is neither a democratic nor a liberal country and, as the past decades have shown, has also not been en route of becoming one.
While the Uzbek regime strategically uses these buzzwords to simulate democratization and a participatory style of government, its leaders also apply a rather illiberal language. National values and traditions and the spiritual heritage of the Uzbek people are recurring motives. Furthermore, there are frequent references to the country's 'centuries-long history' and famous figures claimed to be the glorious ancestors of the Uzbek people (e.g. Amir Timur). As already assessed in great detail in other contributions (Kurzman 1999;March 2002), such historic reinterpretations and inventions of a long-standing Uzbek nationhood 15 serve the regime to foster national consciousness, patriotism, and unity among the people living in post-Soviet Uzbekistan. The regime claims legitimacy for its 'great leadership', often equated with Amir Timur's heroic qualities, by highlighting Uzbekistan's stability and contrasting it to the unstable conditions in neighbouring countries. Strict border controls and other means of repression are justified by security needs and the assurance of 'not allowing any evil to the country's doorsteps'. Combining empty phrases about democratization with such patterns of power maintenance further unmasks the former and current Uzbek presidents' double tracked strategy of speaking like democrats while upholding a strictly authoritarian rule.

The closed regime of Saudi Arabia
The speeches of Saudi Arabia's leaders are among the most authoritarian if compared to the rhetorical styles of the other cases in this analysisboth in terms of stressing illiberalism and speaking about autocratic procedures. In contrast to the official communication of hegemonic Uzbekistan, the closed regime uses the terms 'democratic' or 'democracy' not even once in any of the 63 speeches collected for the analysis. Although Saudi Arabia allows for regular municipal elections, the leaders mention these elections only twice in the assessed material, indicating their insignificance for shaping the politics of the country. Other terms emblematic for a democratic style of language are barely used: 'free' appears neither often nor in terms of freedom in the country but rather in the context of Saudi Arabia's international demands such as 'establishing a free and independent state of Palestine' or making 'the middle east free from atomic weapons'. 'Liberalization' is mentioned only once in economic contexts, the term 'liberal' does not occur at all. It is obvious that King Abdullah and King Salman avoid terms with a strong connotation of democracy and use more innocuous words and platitudes such as 'peace', 'harmony' and 'friendship'. Yet, also such sporadically applied terminology is outweighed by the frequent use of those words in this article's dictionary which represent an autocratic style of language.
Religion and the authority of the royal family are recurring topics in the official addresses of the long-standing Islamic monarchy. 16 While Uzbekistanalso a country in which the majority of the people is Muslimrarely refers to 'Allah' or 'Islam' in its official communication, these and other religious words are among the most frequently used terms in the speeches of Saudi Arabia. As perpetual formulas, both monarchs remind the people in their speeches that they see themselves called by God and led by his will. Other prominent subjects are the country's stability, security in the region and national unity. Sedition calls are considered evil and the people are constantly urged to preserve this unity.
In most of the official speeches of the Saudi kingdom, women are neither addressed nor mentioned. Although for the first time in Saudi Arabia's history women were allowed to run for office during the municipal elections in 2015, they are still subject to a deeprooted exclusion in Saudi Arabia (Al-Rasheed 2013). The absence of women in various spheres of life might also explain why in the majority of the speeches, the kings address their audience merely with 'dear brothers' and have only recently started to speak occasionally of 'brothers and sisters'. Despite these and other minimal signs of opening (Le Renard 2014), Saudi Arabia remains one of the politically most closed regimes in the world. Confirming my second hypothesis, the analysis of its official communication illustrates that the country is also linguistically ruled in a very authoritarian and unprogressive style.

The competitive regime of Russia
The analysis of speeches delivered by Vladimir Putin during his third term of office as president of Russia (2012Russia ( -2018 demonstrates that there is a slight prevalence of talking more about democratic than autocratic procedures. Yet, at the same time, Putin makes use of a comparatively illiberal style of language. The contextualized examination of the texts reveals that while Putin speaks about cultural and ethnic diversity in Russia and describes the country as open and self-respecting, he is also emphasizing traditional values, nationalism and paternalism. Compared to the other cases of authoritarianism, the president's references to the military are remarkably frequent. 'Russia's glorious military traditions', its academies, skills, achievements, and latest technologies are overly praised. The 'Great Patriotic War' (World War II) is a recurring topic, the memory about its heroes and sentences like 'Russia will always remain invincible' are to endow patriotism, national pride and unity among the people. Other frequent terms are 'defend', 'Russia's strength' and 'stability'. These terms are often mentioned in the context of fighting 'illegal drugs and migration', 'trafficking' and other crimes. Similar to Uzbekistan, the regime is depicted as protector from such threats, legitimizing the strict and repressive policies and measures in this regard.
Concerning democratic procedures, the most salient difference to the rhetorical strategy of hegemonic Uzbekistan is that while Putin talks about recent elections or requires fighting corruption and improving justice and transparency, his speeches broadly lack the cant about the great aim of building a democracy. Indeed, if compared to the Uzbek speeches, the terms 'democratic' or 'democracy' appear only occasionally, 17 indicating that speaking about democracy is not deemed to be an efficient legitimation strategy for the Russian regime.
One reason for why Putin is rarely applying the terms 'democratic' or 'democracy' could be Russia's short but resonating experiences with some forms of democracy under the presidency of Boris Yeltsin during the early 1990s. While the Russian constitution adopted in 1993 did not stipulate for today's superpresidentialism, it provided the environment for creating such a large presidential apparatus. As a consequence, the early forms of democracy in Russia were more and more identified with one single person (Fish 2005, 217). Yet, this also meant that the failures and declining popularity of Yeltsin were accompanied by a growing public scepticism about democracy. As Fish (2005, 224-225) writes, for many Russians the term 'democrat' even became a curse word at that time. Another, more general explanation for why the Russian regime is not excessively using the vague telos of democracy could be that the at least minimally competitive surroundings and de facto existing oppositional forces in the country push the president to talk about concrete policies and agendas. To win the highly asymmetrical but nevertheless existing struggle over political popularity and legitimacy, he rather stresses the successful performances of the regime and the strength of existing institutions. Naturally, this includes also the propagation of his own institutionthe well-established and non-democratic system of Putinism (Fish 2005, chap. 8.3;Laqueur 2015).

Alternative explanations
There are several aspects which could serve as alternative explanations for the findings of this article. For example, beside the applied regime categories of closed, hegemonic, and competitive regimes, other regime classifications, e.g. based on discrete data such as Geddes, Wright, and Frantz (2014), have to be considered. One question in this regard would be whether it is rather the party-based, personalist, military or monarchic nature which explains why some authoritarian leaders intensively speak about democracy while others make use of an authoritarian style of language. Yet, for the cases observed in this analysis, such categorizations do not bring meaningful insights: While some monarchic regimes stress illiberalism and autocratic procedures, others stick out with legends about democracy. Similar contradictions are to be found for the other categories.
Authoritarian diffusion as an international factor (Lankina, Libman, and Obydenkova 2016) could also explain why several authoritarian regimes strategically use the telos of democracy while others apply different linguistic tactics. Indeed, some of the high-performing cases of hegemonic regimes share similar historical experiences and geopolitical proximity. Yet, tracing the subtle processes of diffusion is methodologically challenging. In addition, such diffusion effects alone cannot explain why North Korea and Saudi Arabia make use of a particular illiberal rhetoric and hegemonic regimes of different regions apply similar styles of democratic language.
As another international aspect, the variation in language styles among autocrats could be rather a question of whether or not the respective regimes have strong linkages to Western democracies. Such linkages generally include the various efforts of the West to promote democracy in these countries. Hence, this would suggest that particularly those autocratic regimes which are targeted by these programmes make strategic use of a democratic style of language to pretend democratization and keep up the financial support and other benefits of these programmes (Bastiaens 2016, 141). However, while several of the hegemonic regimes in this article have been indeed the target of Western democratization efforts, the strength and mutual support for these linkages vary. Furthermore, current trends among some of these regimes (e.g. Uzbekistan or Tajikistan) to strictly restrict or even expel Western NGOs from their countries have been further weakened these linkages (Dupuy, Ron, and Prakash 2016;Ziegler 2016).
Beside this, factors such as a recent leadership change or even a regime downfall could also influence the choice of language style in the officially delivered speeches by these countries. Yet, the authoritarian regimes in this analysis are mostly persistent and, as the case studies have illustrated, leadership changes in Uzbekistan or Saudi Arabia have not significantly changed their democratic or autocratic language style. Furthermore, it is striking that those hegemonic regimes which intensively talk about democratic procedures and institutional reforms endure already for several decades (e.g. Tajikistan, Kazakhstan, Uzbekistan, Jordan). This indicates that the leaders of these autocracies deem the rhetorical strategy of legitimating their non-competitive multiparty system with a possibly democratic style of language as efficient.

Conclusion
This article examined autocratic and democratic styles of language by analysing 2074 speeches of political leaders from 22 countries. In addition, I conducted a contextualized and more fine-graded analysis of speeches delivered by the leaders of hegemonic Uzbekistan, closed Saudi Arabia and competitive Russia. While many of the recent contributions to modern autocracy research study the phenomenon of competitive authoritarianism, the main focus of this article was on hegemonic regimes and the simulation of pluralism as one of their key survival strategies. The analysis inquired whether this strategy of facilitating elite cohesion and gaining legitimacy is also reflected in the official rhetoric of hegemonic regimes. In this regard, I found evidence for my hypothesis that the leaders of hegemonic regimes use a democratic style of language. Compared to the other cases in this analysis, most hegemonic regimes even seem to overemphasize democracy and democratic procedures in their speeches to feign a clearly non-existing system of pluralism. Yet, as a crucial difference to leaders of democracies, the findings illustrated that this democratic language style of hegemonic regimes broadly lacks elements of liberalism. The reasons for this could be that speaking about liberal values in illiberal surroundings is not of strategic value for the hegemonic regimes when formulating claims to legitimacy in front of a national or international public. While the constant talk about seemingly democratic procedures and institutional reforms is deemed effective, a mock about diversity, openness, and transparency seems to be a much less convincing tool for the leaders of hegemonic regimes. In contrast to this linguistic tactic of hegemonic regimes, the findings showed that the leaders of closed regimes hardly talk about democracy. The results further suggest that also competitive regimes have less aspirations of using a democratic style of language in their official communication.
One constraint of the analysis in this article is the limited amount of cases in each of the three authoritarian regime types. Quantitatively assessing large numbers of speeches delivered by autocrats is a new sub field of autocracy research and this study is rather exploratory. It points to crucial differences concerning the linguistic strategies of autocraciesyet, particularly the findings for competitive regimes need to be tested in a larger amount of cases to further generalize them. Thus, future research needs to find better and more far-reaching methods of data collection for the quantitative analysis of the language of autocrats. Instead of referring to merely those cases of authoritarian regimes which upload English translations on their official websites, improved natural language processing (NLP), automated translation (de Vries, Schoonvelde, and Schumacher 2018; Maerz and Schneider, forthcoming;Proksch et al. 2019) and new text mining tools could help to include more cases and sources in multiple (non-European) languages.
The article discussed alternative explanations and illustrated that while some of the international factors might partly contribute to different language styles among authoritarian regimes, the observed variations suggest that regime-level features override other aspects. However, another task for future research would be to further scrutinize the (reciprocal) effects these language styles have on the international levelparticularly the language of democracy used by the leaders of hegemonic regimes. Thus, do they indeed gain legitimacy among the international audiences addressed in their speeches? In this regard, it would be fruitful to further disentangle the international from the national speeches. While this analysis traced democratic terms in both kinds of these speeches, a more fine-graded differentiation of the rhetorical styles and applied formulaicity is needed. Further accounting for such languagespecific aspects when dealing with hegemonic leaders would help Western democracies and the policy community at large to better identify whether their financially strong efforts of promoting democracy nurture fake democratic institutions to consolidate authoritarianism or real progress to more open forms of governments. Notes 1. Pluralism is understood here in its political sense, meaning a diverse distribution of political power including many separate elites rather than a single power elite, cf. Lijphart (1975, 2). 2. Due to general problems of accessing reliable data in authoritarian settings and especially because of the issues of conducting surveys in such surroundings, measuring the actual amount of legitimacy belief among the population of authoritarian countries poses great difficulties. First attempts are Frye et al. (2017) and Mazepus (2017). 3. A regime is autocratic if (1) the executive achieved power by other means than fair, competitive and free elections, (2) the executive achieved power by fair, competitive and free elections but changed these rules afterwards or (3) the military prevents the compliance with these rules or changes them. I use the term autocratic/authoritarian regime and autocracy interchangeably and as umbrella terms for the various autocratic subtypes examined in this article. 4. Laver and Garry's (2000) categories on institutions, law and order, values and groups were partly adopted and renamed. However, my dictionary does not include any economic terms because these terms are not explicitly autocratic or democratic (e.g. economic liberalism does not necessarily conform with political liberalism). 5. I chose the most recent international and national speech in each case to grasp current terminology. 6. I randomly picked 5% of the speeches in each case, excluding those taken as pool of key words. The additional use of unsupervised text mining methods, e.g. topic modeling, could be another method of (cross-) validation, similar as done in Maerz and Schneider (forthcoming). Yet, due to the comprehensive qualitative checks and combination with detailed case studies, the dictionary in this article seemed adequately validated for the analysis. 7. I used Lowe's (2015) Yoshikoder for constructing and validating the dictionary. 8. The full lists of words in each category of the dictionary are available at: https://dataverse. harvard.edu/dataverse/sfm. 9. The material of North Korea consists rather of letters than speechesyet, remarkably, all available texts were translated in nine different languages and published on the official website of the regime. The uploaded texts of Russia, Hungary and Great Britain include occasional interviews and dialogues or joint statements with other persons. These texts were filtered to get the leaders' statements only. Speeches uploaded in languages other than English were excluded. 10. I collected all speeches with the help of web-scraping techniques, comprising several packages in R as explained in Munzert et al. (2015). 11. For a more recent application, see also Proksch et al. (2019). 12. All operations were done in R (2018, v. 3.4.0.) with the quanteda (Benoit et al. 2017, v. 0.9.9.) and tm package (Feinerer and Hornik 2017. v. 0.7-1.). The replication files are available at: https://dataverse.harvard.edu/dataverse/sfm. 13. Malaysia and Russia score 7 in both DPI indicators and are therefore considered as typical competitive regimes (Beck et al. 2001, cf. Table A1 in the Supplemental Material). Uganda and Morocco are rather borderline cases: Uganda scores 6 in the legislative and 7 in the executive elections indicator but is generally considered as competitive (Levitsky and Way 2010, 32). Morocco has low scores in the executive elections indicator due to its monarchic nature, yet, allows for competitive legislative elections which are not based on a satellite party system (score of 7 and Stepan, Linz, and Minoves 2014). 14. All direct and indirect citations from the speeches refer to the corpus of texts for each case.
The Supplemental Material provides the sources for the text corpora. 15. The country of Uzbekistan as such has actually a rather short history, cf. Baldauf (1991). 16. Robustness tests have shown that also without including such religious and monarchic terms in the dictionary, Saudi Arabia still outperforms most of the other cases in terms of an autocratic style of language. 17. In the 27 speeches of Uzbekistan, the root 'democra-' is mentioned at least 44 times. In the significant larger amount of Russia's speeches (325), it appears only 37 times.