When we think of public policy speeches by authoritarian actors, we often picture grey apparatchiks ‘entertaining’ their audiences by reading aloud long and complex texts in mind-numbing monotones (Anderson Reference Anderson1996, Reference Anderson2014). This image is well-founded, because most of the time such speeches are indeed rather difficult to listen to. As we argue and demonstrate in this paper, the majority of officials tasked with public policy communication rarely make an effort to engage their audiences, typically delivering dreary, needlessly complex speeches filled with overly long sentences, as exemplified by the following excerpt: ‘Millions of roubles worth of industrial products were produced in excess of the plan; sold to the state: 199,000 tons of milk, or 6 percent more than in the same period last year; livestock and poultry – 52.8 thousand tons, or 14 percent more than last year’ (CPSU 1971, 296). Yet other officials, similarly to effective public speakers in democracies, attempt to capture the attention of their audience by crafting simple, clear messages.Footnote 1 What then explains whether, and when, authoritarian elites will go the extra mile, revising their policy reports to make them easier to comprehend? And what does this behaviour reveal about their own incentives and authoritarian politics in general?
We argue that the linguistic complexity of authoritarian speech is driven by strategic logic, determined by elite incentives to placate and co-opt the citizens in the service of regime stability and their own careers; effective policy communication is one of the tools at their disposal (Rauh Reference Rauh2023; Dewan and Myatt Reference Dewan and Myatt2012). However, not all officials are equally motivated to reduce the complexity of their policy reports and use vernacular. When addressing constituencies characterized by lower ruling party support and/or higher protest activity, where citizens are more likely to hold critical views and be receptive to opposition messages (Rosenfeld Reference Rosenfeld2021), elites have stronger incentives to appear responsive to the public (Guriev and Treisman Reference Guriev and Treisman2019; Aytaç Reference Aytaç2021). In such contexts, state officials are more likely to convey policy reports through colloquial and persuasive language. In contrast, when elites preside over regions where the ruling party enjoys dominance and/or protest activity is minimal, they lack incentives to act as if they care. Then, they will be more likely to merely read reports from the rostrum in a Brezhnevite fashion. In both scenarios, elites remain primarily accountable to the autocrat rather than the public, but those governing less loyal regions have stronger incentives to invest effort into their policy communication and to mimic democratic politicians.
We propose that the linguistic complexity reveals the incentives of elites: they speak simply when they want to be closer to constituents, to persuade and be understood; they use complex, ‘wooden’ language when they want to maintain disengagement, apathy, and depoliticization, dissociating the language of politics and policy from the vernacular (Anderson Reference Anderson1996, 148). Most of the time, because authoritarian actors generally require their subjects to be passive, a higher complexity in speech is their default (Lasswell Reference Lasswell, Lasswell and Leites1949, 29–31), contributing to the impression of themselves as experts in their policy domains that are too difficult and also unnecessary for ordinary citizens to comprehend. That is, because depoliticization is a default position to maintain, therefore, the rhetorical distance, reflected through complexity in speech, is equally a default, baseline position in speech for elites. Linguistic complexity is dynamic, however, and can be adjusted when elites find themselves having to additionally co-opt and persuade.Footnote 2
At the same time, while the autocrat values the ability of elites to manage and co-opt citizens, including through propaganda, simple, persuasive communication can also reveal political ambition. Effective communicators, by reaching a wider audience, may be perceived as potential threats to the autocrat (Gueorguiev and Schuler Reference Gueorguiev and Schuler2016). As a result, the logic of tailoring policy messages to autocratic audiences will be tempered by the opposite concerns over not being the proverbial tallest poppy to be cut down. Therefore, we also examine the consequences of clear speech for elite careers (Svolik Reference Svolik2012; Gorlizki and Khlevniuk Reference Gorlizki and Khlevniuk2020).
This paper offers the first comprehensive study of the general logic of political speech complexity in a non-democratic context. While it is widely studied in democracies (Benoit et al. Reference Benoit, Munger and Spirling2019; Decadri and Boussalis Reference Decadri and Boussalis2020; Lin and Osnabrügge Reference Lin and Osnabrügge2018; Spirling Reference Spirling2016; Rauh Reference Rauh2023), including through theoretical work (Dewan and Myatt Reference Dewan and Myatt2008, Reference Dewan and Myatt2012), the linguistic complexity of authoritarian policy communication remains largely undertheorized.Footnote 3 We argue and demonstrate that speech complexity is deployed strategically, driven by elite incentives to project a more positive regime image to less loyal constituents. We further find that simpler, more effective communicators advance their careers. Although we primarily test our argument on one subset of authoritarian elites delivering important policy reports, we validate it further by extending the analysis to very different types of speech and different types of elites. Because linguistic complexity affects all aspects of communication, it is likely to be one of the central yet underexplored tools of authoritarian propaganda and legitimation (Dukalskis and Gerschewski Reference Dukalskis and Gerschewski2017; Baturo and Tolstrup Reference Baturo and Tolstrup2024; Liu Reference Liu2022). An investigation of speech complexity, proposed herein, can shed new light on how authoritarian communication and propaganda really work.
Our argument has important implications for the literature on authoritarian accountability and co-optation (Kim and Gandhi Reference Kim and Gandhi2010; Cook and Dimitrov Reference Cook and Dimitrov2017; Jowitt Reference Jowitt1975; Magaloni Reference Magaloni2006). While it is well-established that dictatorships calibrate repression and public goods provision in response to regional or societal characteristics (Frantz and Kendall-Taylor Reference Frantz and Kendall-Taylor2014; Xu Reference Xu2021), we demonstrate that authoritarian responsiveness in service of co-optation extends beyond material strategies, such as social spending, and includes non-material, informational tools, such as attempts to reduce the ‘conceptual distance’ between elites and citizens (Anderson Reference Anderson1996, 145), by tailoring and adjusting policy communication according to the characteristics and protest potential of their audiences. In turn, because constituents tend to prefer politicians who communicate simply (Bischof and Senninger Reference Bischof and Senninger2025)Footnote 4 and engage with information presented in an accessible manner (Anderson Reference Anderson1983), simplicity in policy communication may shape public acquiescence and perceptions of the legitimacy of authoritarian rule (Aytaç Reference Aytaç2021; Carter and Carter Reference Carter and Carter2023).
To test our argument, we rely on a dataset of annual legislative addresses delivered by Russian governors from 2007 to 2021. Choosing it as a case study has several advantages. First, Russia’s large number of politically diverse federal subjects provides a unique context for a within-country large-N analysis (Baturo et al. Reference Baturo, Khokhlov and Tolstrup2025; Libman Reference Libman2022; Reuter and Robertson Reference Reuter and Robertson2012). Second, regions differ in terms of gubernatorial turnovers over time, and regional governors themselves vary in their backgrounds, allowing us to distinguish between regional and individual-level effects. Third, while the addresses follow a similar format and cover socio-economic policies, governors, however, have leverage over how they cover them. To our knowledge, this is one of the first studies to adapt established complexity metrics (Flesch Reference Flesch1948; Smith and Senter Reference Smith and Senter1967) to political texts in Russian. We validate our automated measures through human coding of a sample of speeches and supplement our analysis with alternative measures of linguistic complexity and emotiveness.
Because we rely on the observational data, to improve the identification further we also employ regression discontinuity and interrupted time-series analyses. We demonstrate that actors adjust speech complexity when their strategic context changes, and that complexity varies when the same individuals move between different governorships. For robustness, we extend our analysis to a separate corpus: social media posts made by governors. We also test our argument in a different empirical domain: legislative speeches by members of Russia’s lower house, the State Duma. Finally, we also supplement by analysing all available speeches made by several governors.
The Logic of Political Speech Complexity in Autocracies
To illuminate the logic behind speech complexity in autocracies, we propose an argument centred on the effects of audience characteristics on elite incentives for more effective policy communication.Footnote 5
Autocratic Elites and Public Policy Communication
Politicians generally adjust the complexity and tone of their public rhetoric to either reach a wider audience or target their core constituencies (Dewan and Myatt Reference Dewan and Myatt2012; Spirling Reference Spirling2016). Thus, they use more emotive rhetoric in debates that have a large general audience and less emotive and more complex rhetoric when their speeches are not broadcast (Osnabrügge et al. Reference Osnabrügge, Hobolt and Rodon2021; Gennaro and Ash Reference Gennaro and Ash2022). In democracies, due to accountability concerns, politicians put substantial efforts into tailoring speeches to their audience’s characteristics, and, for example, rely on simpler language when addressing poorer, less educated constituents (Bischof and Senninger Reference Bischof and Senninger2018; Lin and Osnabrügge Reference Lin and Osnabrügge2018; Decadri and Boussalis Reference Decadri and Boussalis2020).
At first glance, because authoritarian elites are primarily accountable to the autocrat, they may not need to be overly concerned with how their messages are received by the public. In this view, any observed differences in speech complexity across elites would be attributed to the nature of their respective policy portfolios or even their personal traits (Charteris-Black Reference Charteris-Black2011; Schoonvelde et al. Reference Schoonvelde, Brosius, Schumacher and Bakker2019).
At the same time, the accountability of authoritarian elites to the autocrat includes their ability to manage and co-opt the citizens in their constituencies (Kim and Gandhi Reference Kim and Gandhi2010; Cook and Dimitrov Reference Cook and Dimitrov2017; Magaloni Reference Magaloni2006). Even in closed autocracies, ‘most territorial leaders’ cannot be seen to be ‘completely unresponsive to demands from below’ (Gorlizki and Khlevniuk Reference Gorlizki and Khlevniuk2020, 2) – when and if such demands arise. One tool for managing constituents is policy propaganda, including through speech. To lower the costs of staying in power, dictatorships also promote development and spending on public goods to co-opt the public (Cook and Dimitrov Reference Cook and Dimitrov2017; Jowitt Reference Jowitt1975; Kim and Gandhi Reference Kim and Gandhi2010; Magaloni Reference Magaloni2006). In practice, co-optation measures and policies have to be publicized and communicated through speech (Rosenfeld and Wallace Reference Rosenfeld and Wallace2024).
In their communication with the public, contemporary autocracies emphasize policy responsiveness to the needs of the population by explaining policy plans and claiming credit for past successes (Aytaç Reference Aytaç2021; Baturo and Tolstrup Reference Baturo and Tolstrup2024; Guriev and Treisman Reference Guriev and Treisman2019). Such propaganda is often effective in deterring the public from challenging the government (Huang Reference Huang2015; Carter and Carter Reference Carter and Carter2023; Lutscher and Donnay Reference Lutscher and Donnay2024). It also influences public attitudes towards policy issues (Mattingly and Yao Reference Mattingly and Yao2022) and encourages specific behaviours, including regime support (Adena et al. Reference Adena, Enikolopov, Petrova, Santarosa and Zhuravskaya2015; Rosenfeld and Wallace Reference Rosenfeld and Wallace2024). While these efforts are bolstered by the ability to control the media, autocratic officials still need to craft effective messages to persuade and be understood (Aytaç Reference Aytaç2021; Huang Reference Huang2015; Maerz Reference Maerz2019). When aiming to appear responsive and effectively communicate with ‘the mass of the populace’ (Gill Reference Gill2018, 177), they rely on a range of linguistic and verbal tools, including the complexity of their speech (Townson Reference Townson1992).
Speech Complexity and Autocratic Audiences
Due to co-optation and policy legitimation concerns, authoritarian elites are motivated to convince citizens that the government is responsive to their demands and worthy of their support or acquiescence (Guriev and Treisman Reference Guriev and Treisman2019). In the process of persuasion, simpler language is often found to be effective as it improves individuals’ understanding and engagement with information (Anderson Reference Anderson1983). It also provides a heuristic whereby the speaker is perceived as one of the ordinary citizens (Bischof and Senninger Reference Bischof and Senninger2018, Reference Bischof and Senninger2025), in turn making constituents view the speaker and their message in a more positive light (Dewan and Myatt Reference Dewan and Myatt2012). In the same way as public petitions, which help dictators to obtain information from citizens while serving as channels for building trust and co-opting the public (Dimitrov, Reference Dimitrov2015), simple communication contributes to citizens’ better understanding of policy while also portraying politicians as accountable (Bischof and Senninger Reference Bischof and Senninger2018, Reference Bischof and Senninger2025). This, in turn, contributes to regime legitimacy.
At the same time, authoritarian actors require their constituents to remain docile and eschew paying attention to or questioning how they conduct politics and implement policies. Then, relying on complex language generally serves the goal of maintaining popular disengagement and depoliticization (Anderson Reference Anderson1996; Lasswell Reference Lasswell, Lasswell and Leites1949, 29–31), signalling to the ordinary citizens not to engage with too complex policies, leaving their formulation and implementation to the experts – authoritarian elites. As pointed out by Anderson (Reference Anderson1996, Reference Anderson2014), ‘wooden’, complex speech, because it allows elites to distance themselves from the public and signal the lofty, inaccessible nature of their policy domains, has in fact typically been the default option in autocracies. Given that simpler speech may sometimes help in persuading the citizens that the government is responsive, while more complex speech serves the overarching interest of depoliticization and public apathy, what then explains how authoritarian actors will speak while addressing their audiences?
Specifically, we propose that the marked differences in linguistic complexity across authoritarian elites are primarily driven by the type of audience they address. When elites have stronger incentives to be persuasive and clear, they prioritize simple language in their policy speeches. Conversely, when their strategic context allows elites to ignore their audience, dry, monotonous speech in impenetrable, complex language is more likely to follow.
Assuming that regime support is not uniform across regional constituencies and societal groups – a plausible assumption for all but the most repressive autocracies – and that there exist significant within-country political, economic, and other societal differences, the ease with which elites are able to control the citizens and manage regime-controlled elections, inter alia, is also likely to differ (Magaloni Reference Magaloni2006; Reuter and Robertson Reference Reuter and Robertson2012). That is, in our framework, the audience is the public, not the autocrat, but elites ‘work towards’ their public audience because they are accountable to the autocrat for control of their constituencies. We can further simplify and distinguish between two types of public audiences: those with stronger loyalty to the regime, which are easier to govern, and those with weaker loyalty, which require more effort to manage.
Will authoritarian elites use simpler speech for their more or less loyal audiences, however? A priori, we can draw two contrasting perspectives. First, because elites have a stronger affinity with more loyal audiences, they may therefore make an effort to be understood by them, using simpler language. The second, opposite perspective is that elites have stronger incentives to perform well and speak more persuasively and simply when addressing not more but less loyal audiences.
One important caveat is that in the context of everyday bureaucratic politics under dictatorship, and regarding our main empirical domain, that of verbal policy reports, complex speech is typically the baseline, default position (Anderson Reference Anderson1996, Reference Anderson2014; Gill Reference Gill2018; Lasswell Reference Lasswell, Lasswell and Leites1949). Simply put, elites can always read reports, prepared by their subordinates, from the rostrum without adjusting them for the live audience, much like Soviet regional heads who ‘read aloud from prepared texts from the same blue notebooks’ (Gorlizki and Khlevniuk Reference Gorlizki and Khlevniuk2020, 271). Since policy reports require adjustments to be more persuasive and listenable, and elites can always task their staff to edit them accordingly, higher complexity in live addresses signals not only a strong ‘conceptual distance’ between the officials and the public (Anderson Reference Anderson1996), but also a notable lack of motivation to connect with the audience.
Simpler speech and more loyal audiences
According to the first perspective, elites simplify their reports to craft a favourable image of themselves and their regimes (Carter and Carter Reference Carter and Carter2023; Rosenfeld and Wallace Reference Rosenfeld and Wallace2024) when addressing more loyal constituents. For example, when elites need their less educated or poorer constituents to comprehend their policy propaganda, they will, among other things, employ speech ‘with tangy proverbs’ in order ‘to communicate with peasants better’ (Taubman Reference Taubman2003, 28, 271). Similarly to democratic incumbents, who often target their core constituencies – a strategy with low average marginal returns but more certainty of securing victory, in contrast to targeting swing voters (Gerber Reference Gerber2004) – authoritarian elites also make an effort to be understood by their base supporters, whom they placate with public goods and simpler communication. In contrast, when addressing audiences known for lower regime support, which may also be dominated by the regime opponents, elites are less likely to bother with effective persuasion and may instead rely on other tools of governance, such as repression (Svolik Reference Svolik2012; Xu Reference Xu2021).
Furthermore, while the proposed view of ‘wooden’ policy speech as the default position in autocracies is consistent with our case study and examples that follow, elites may also conceivably use or even increase the complexity to obfuscate (Bischof and Senninger Reference Bischof and Senninger2018, 474; Rauh Reference Rauh2023), to intimidate potential regime opponents (Guriev and Treisman Reference Guriev and Treisman2019; Baturo and Tolstrup Reference Baturo and Tolstrup2024), and, as discussed above, to maintain ‘conceptual distance’ from the public, signalling that politics is something of no concern to regular citizens (Anderson Reference Anderson1996, 148). Then, authoritarian actors will be more likely to intimidate, obfuscate, and/or distance from their less loyal constituents, not their supporters (Baturo and Tolstrup Reference Baturo and Tolstrup2024), by using unnecessarily complex, technical speech. Whether complexity arises from the default position in policy reports or as a deliberate tactic, we expect to observe higher complexity in reports addressed to less loyal, rather than more loyal, audiences, if the first theoretical perspective holds.
Simpler speech and less loyal audiences
The second perspective, which we subscribe to, is that elites will do the opposite and attempt to be more persuasive when they have to compete for attention in a political marketplace, such as when they address less loyal audiences. When their constituencies are secure for the regime and remain docile, elites lack the incentives to additionally placate, persuade, or entertain their audiences and, as a result, they will use duller, more complex language in order to maintain the existing public apathy and disengagement. The logic is somewhat similar to that often observed in democratic politics, where actors must exert more effort, whether through spending or campaigning more effectively, in more competitive constituencies, often marked by higher polarization (Hirsch Reference Hirsch2023; Meirowitz Reference Meirowitz2008).
Generally, authoritarian elites have stronger incentives to respond to the interests of more organized groups, which can impose higher costs on them through protest activity or voting against the ruling party – actions that displease the autocrat and can lead to their dismissal (Reuter and Robertson Reference Reuter and Robertson2012). In regions with stronger political opposition or a higher share of economically independent citizens (Beissinger Reference Beissinger2022; Greene Reference Greene2014), public officials face stronger competitive pressures (Magaloni Reference Magaloni2006). Also, when the audience has more choice in whom to listen to, authoritarian actors have to speak clearly to divert from such viewpoints (Dewan and Myatt Reference Dewan and Myatt2012); officials ‘cannot simply babble; if the noise in her speech is too large, she will be ignored’ (Dewan and Myatt Reference Dewan and Myatt2008, 361). Therefore, authoritarian elites will have stronger incentives to ensure support in the protest-prone regions and be more likely to reduce ‘the conceptual distance separating rulers from the people’ (Anderson Reference Anderson1996, 145) by employing simpler language. As a result, authoritarian officials will behave less like technocratic bureaucrats and more like street politicians, emulating the approachable communication style typical of democratic politics.
In contrast, elites who control their regions as substate dictators in their own right (Gorlizki and Khlevniuk Reference Gorlizki and Khlevniuk2020) have weaker incentives to make their messages comprehensible to citizens. For them, the more decisive factor is their ability to deliver votes for the ruling party (Buckley et al. Reference Buckley, Frye, Garifullina and Reuter2014; Reuter and Robertson Reference Reuter and Robertson2012). When elites are confident in their ability to secure pro-regime votes and control the population, they are less concerned with appearing engaging and responsive. Consequently, and invoking the clichéd imagery of grey party apparatchiks of old, mumbling for hours without any concern for their audience, such officials will use the bully pulpit to simply go through the motions, boring their captive listeners with the barrage of statistics or vocalizing technical reports, all without any real attention to effective communication.Footnote 6
Hypotheses
The proposed framework allows us to examine the incentives of authoritarian elites through speech. As previously outlined, elites are accountable to the dictator for policy failures, dissent, or delivering votes for the ruling party (Buckley et al. Reference Buckley, Frye, Garifullina and Reuter2014; Svolik Reference Svolik2012; Gorlizki and Khlevniuk Reference Gorlizki and Khlevniuk2020). Based on this, and according to the second – simpler speech for less loyal audiences – perspective, we draw that when there is potential for collective action and the electoral dominance of the ruling party is not guaranteed, elites will have stronger incentives to appear responsive to the public, manifesting in simpler, more persuasive speech. Therefore, our main hypotheses are as follows:
Hypothesis 1: Elites in regions with higher ruling party vote shares employ more complex language than in regions with lower shares.
Hypothesis 2: Elites in regions with lower protest activity will use more complex language than in protest-prone regions.
It is clear that the first perspective – simpler speech for more loyal audiences – logically implies the mirror opposite effects to those stated in the hypotheses. Also, as an additional hypothesis, we consider the possibility that elites have personal electoral incentives to simplify their speech, especially when they are directly elected rather than appointed by the autocrat (Chang and Higashijima Reference Chang and Higashijima2023; Buckley et al. Reference Buckley, Frye, Garifullina and Reuter2014). Generally, unelected professionals lack incentives to attract large audiences and tend to use more technical language, in contrast to elected politicians (Gray and Baturo Reference Gray and Baturo2021). In turn, experimental evidence suggests that voters prefer candidates with simpler campaign messages (Bischof and Senninger Reference Bischof and Senninger2025). Thus, actors who have to directly appeal to large audiences, even under autocracy, have stronger incentives to build identification with voters, thereby reducing the conceptual distance to the public by using vernacular and simpler language (Anderson Reference Anderson1996). Accordingly:
Hypothesis 3: Directly elected elites will use less complex language compared to indirectly elected or appointed elites.
The proposed hypotheses imply that elites use simpler speech not only when they already govern in less loyal constituencies but also that they adjust their speech complexity when their strategic context changes.
Alternative and Supplementary Explanations
Our theoretical framework is centred on the effects of constituency characteristics on the incentives and, consequently, the speech complexity of authoritarian elites. However, distinct communication styles may also be driven by individual personal, professional, or educational traits (Schoonvelde et al. Reference Schoonvelde, Brosius, Schumacher and Bakker2019; Decadri and Boussalis Reference Decadri and Boussalis2020). We therefore take relevant individual factors into account in the analyses. Furthermore, since governors with different traits may be strategically placed in specific regions by the autocrat, we also address the potential selection bias in the supplementary appendix and discuss it further below.
While simpler speech in more difficult-to-govern regions may help authoritarian elites engage and co-opt their audience, does it also advance their careers? Existing literature suggests that politicians who give clear expression to their views tend to be more successful than those who do not (Dewan and Myatt Reference Dewan and Myatt2012; Pennebaker Reference Pennebaker2011). If the logic of speech complexity holds, we can expect that the autocrat will reward more effective communicators. At the same time, clearer communicators often attract larger audiences (Dewan and Myatt Reference Dewan and Myatt2008, 361), and the clarity of their messages may be interpreted not only through the prism of effective communication but also as an indication of their ambition to reach a broader, nationwide audience (Charteris-Black Reference Charteris-Black2011, 225). Thus, conveying ideas in plain and convincing language can also be perceived as a threat to the autocrat (Gueorguiev and Schuler Reference Gueorguiev and Schuler2016).Footnote 7 Because it is not clear whether the autocrat’s security concerns will temper the demand for more effective communicators, in the appendix we therefore also explore the potential consequences of speech complexity on political careers.
Furthermore, if we extend our framework, it also follows that elites deploy linguistic simplicity, as determined by their incentives to appear more responsive in their constituencies, in order to influence societal attitudes. That is, another implication is that more effective policy propaganda through simpler speech may – or elites believe that it may – have an impact on public opinion. Because investigating this conjecture requires a very different research design,Footnote 8 we have to leave it out for future research; the results of the supplementary analyses are, however, in line with expectations.Footnote 9
Data and Methods
Text Corpus of Elite Speeches
We test our hypotheses using a text corpus of annual public speeches delivered by regional governors in Russia between 2007 and 2021. We leverage the fact that all governors are obliged to make annual public policy addresses in a relatively similar format, summarizing the past year’s policy achievements and setting goals for the near future, to some extent resembling presidential addresses (Baturo and Mikhaylov Reference Baturo and Mikhaylov2013), albeit at the regional level. Annual addresses are among the most important acts of public communication for governors (Baturo et al. Reference Baturo, Khokhlov and Tolstrup2025). The speeches are broadcast on TV, YouTube, and other social media platforms, published on regional government websites and in newspapers, and widely discussed in the media.Footnote 10 Because the annual addresses cover comparable substantive areas, while allowing governors considerable discretion in how they present content and engage their audiences, they provide a highly suitable empirical source to test our argument. Our main text corpus includes 924 speeches in total. On average, they are relatively long and cover a broad range of economic and social policy areas (see the appendix). As explained below, we also conduct supplementary analyses using different text corpora.
The Dependent Variable: Complexity of Communication
Scholars analysing democratic political speech often apply complexity metrics originally developed in other domains of social science, such as the Flesch–Kincaid Grade Level test, FK (Flesch Reference Flesch1948) or the Simple Measure of Gobbledygook, or SMOG (McLaughlin Reference McLaughlin1969). Following a similar approach, for each speech in our corpus, we calculate the Automated Readability Index, ARI (Smith and Senter Reference Smith and Senter1967), and, for robustness, also the FK (Flesch Reference Flesch1948).Footnote 11 The ARI formula is given by:
$${\rm{ARI}} = 6.26\; \times {{{\rm{number}}\;{\rm{of}}\;{\rm{letters}}} \over {{\rm{number}}\;{\rm{of}}\;{\rm{words}}}} + 0.2805 \times \;{{{\rm{number}}\;{\rm{of}}\;{\rm{words}}} \over {{\rm{number}}\;{\rm{of}}\;{\rm{sentences}}}} - 31.04$$
The calculated ARI scores range between 7.4 and 28.8 (μ = 17.5, σ = 3). The ARI values approximately correspond to the number of years of education required to understand the text (Flesch Reference Flesch1948). In our sample, 47 per cent of speeches (438 in total) exhibit very complex language (ARI > 17), appropriate for an audience with a PhD degree or multiple university diplomas. In turn, 39 per cent (364 texts) are suitable for university students and below (15 to 17).
Two examples illustrate the substantive differences between simpler and more complex texts. In 2012, the governor of the Altai Krai, Alexander Karlin, delivered a report packed with detailed statistical information on each aspect of the economy, including the production of sausages and whole-milk products:Footnote 12
Since 2005, our food and processing industry has shown only positive development rates. In 2011, the increase in production amounted to 4.1%. And in the manufacture of sausages and whole-milk products, the highest result for the entire post-Soviet period was achieved. Over the past year, industry enterprises have mastered the production of 210 new types of food products and 214 types of dietary supplements, phytoproducts, processed antler raw materials and beekeeping products; received 5 licenses for the production of medicines.
In contrast, the head of the Yamal-Nenets region, Dmitry Artyukhov, employed a much more accessible style in his 2021 speech when describing advances in demographics (which, admittedly, he could instead have explained in much more difficult to listen to language):
Dear countrymen! I have very good news – last year, the birth rate in Yamal increased. Unfortunately, this is not the case in all regions of our country. But Yamal is in the black. We give each baby a gift with everything you need, from diapers to winter overalls. In total, there are 47 useful things. We have already handed over 12,000 such sets. We will definitely continue this good tradition. What is one of the main questions for a family? Of course, your home. To do this, there are various support measures, including preferential mortgages and mother capital. But the Young Family program occupies a special place.
As a test of face validity, Figure 1 presents the results of the keyness analyses for the easiest and most difficult-to-read groups of texts, as sorted based on the ARI scores. This method identifies words that appear significantly more frequently in one set of texts than in others, capturing how ‘key’ a word is within a group, in contrast to other texts. In simpler texts, the distinctive words tend to be short and commonly used in everyday communication. By contrast, more complex or technical vocabulary distinguishes the most difficult-to-read texts. These results suggest that the ARI scores effectively capture variations in text complexity, distinguishing between more accessible and more difficult speeches.
Figure 1. Easy and difficult speeches: Keyness results.
Note: analyses for the top-10 ranked (easy) and bottom-10 (difficult) addresses. Tokens translated from Russian are on the right subplot.
To further validate the ARI scores, a random sample of 20 per cent of texts was drawn and read by two coders independently. Each coder categorized 185 documents as either ‘easy’ or ‘difficult’, achieving a high level of intercoder agreement.Footnote 13 For additional validation, we separately regressed the binary complexity measures assigned by each coder on the ARI scores and the predictors from the main models (see Table 15A in the appendix). The results show that the ARI score is strongly correlated with the coders’ assessments of complexity, providing further evidence of the ARI’s validity for measuring text complexity in Russian.
Because elites who use linguistic simplicity may also try to be persuasive by increasing the emotional appeal of their speech (Osnabrügge et al. Reference Osnabrügge, Hobolt and Rodon2021) and by emphasizing a direct personal connection with constituents (Liu Reference Liu2022; Pennebaker Reference Pennebaker2011; Slapin and Kirkland Reference Slapin and Kirkland2020), we additionally construct and employ two alternative indicators: speech emotiveness and the share of the first-person plural pronouns (see Table 5A and Figure 9A in the appendix). These measures capture whether the logic of adjusting to audiences operates not only through linguistic simplicity but also through distinct, observable manifestations of elite incentives to ‘act as if they care’.
Explanatory Variables
To account for the hypothesized audience effects on the text complexity (Hypotheses 1 and 2), we include the share of votes for United Russia (UR) in the last State Duma elections in a region before a legislative address (ICSID 2023), as well as the number of protests per region (Lankina Reference Lankina2015). Following Lin and Osnabrügge (Reference Lin and Osnabrügge2018), we also control and proxy for the linguistic skills and socio-economic status of the audience by including aggregate measures: the share of employed people with a university education, and the regional unemployment rate. We use the log of gross regional product (GRP) per capita to account for economic development. To control for the length of the annual addresses, we include the number of tokens per text (word count). To test Hypothesis 3, we use a binary indicator that equals 1 if the governor was appointed by the president and then indirectly elected by the regional legislature for the current term, and 0 if the governor was directly elected.
Our central assumption is that elites adjust the complexity of their speech to align with the characteristics of their constituencies. However, linguistic style is often influenced by a speaker’s own education and professional background (Schoonvelde et al. Reference Schoonvelde, Brosius, Schumacher and Bakker2019; Decadri and Boussalis Reference Decadri and Boussalis2020). Specifically, we expect that more politically experienced and educated governors, compared to newcomers and outsiders, may employ a simpler style, reflecting their experience or confidence as public speakers and less of a need for prompts such as reading aloud from written reports. Furthermore, certain elites may undergo training to improve their leadership skills, including public communication. Therefore, we include additional controls: the number of years in office, a PhD degree, governor’s age, their background in business or security service (ICSID 2023), and whether they graduated from the ‘School for Governors’ under the auspices of the presidential administration (Pertsev Reference Pertsev2018). The appendix provides a more detailed description of variables.
What Explains the Complexity of Speech?
Table 1 presents the results of the regression analysis. The unit of analysis is a region-year for the individual annual address delivered by the regional governors in 2007–21, with the ARI as a dependent variable in Models 1–8. In Models 9–10, we use two alternative dependent variables. Models 1–7 and 9–10 include region-level fixed effects, while Model 8 accounts for governor-level fixed effects.
Table 1. Linguistic complexity of governor annual addresses
Note: models 1–7, 9–10 are panel regression models with regional fixed effects; 8 with governor fixed effects; 9 is fixed effects ordered probit with emotiveness index as the dependent variable. ‘We’ share is the dependent variable in 10. *p < 0.1, **p < 0.05, ***p < 0.01.
Across all specifications in Table 1, the coefficients for the Duma UR vote share and regional Protests are statistically significant and exhibit the expected signs. Therefore, there is strong support for Hypotheses 1 and 2: governors in regions with lower UR vote shares and higher protest activity tend to use simpler language. That is, the results align with the second theoretical perspective on audience and speech complexity dynamics, suggesting that elites ‘work towards’ less loyal, rather than more loyal, audiences. Specifically, the top subplot of Figure 2 that visualizes the marginal effects shows that as the UR vote share goes up from its lowest to its highest value, the expected ARI increases from a more colloquial language score of 16.6 to a denser, more technical language at 18.9, or nearly one standard deviation of the ARI variable. In turn, comparing the average speech of a governor presiding over a docile, protest-free region (for example, Yuri Kokov of Kabardino-Balkaria) to that of a protest-prone region (for example, Valentina Matvienko of Saint Petersburg), the average ARI decreases by 1.5 points, from 17.6 to 16.1. While the effects displayed in Figure 2 can only be interpreted hypothetically, in the Further Analyses section below we demonstrate how governors in fact adjust their language when their strategic context changes.
Because constituencies with lower regime support and stronger protest activity are often those with a higher share of economically independent and educated citizens (Beissinger Reference Beissinger2022; Chenoweth and Ulfelder Reference Chenoweth and Ulfelder2017; Greene, Reference Greene2014), we include three indicators for relevant socio-economic characteristics as controls. As shown in Column 3 of Table 1, the coefficient for the GRP per capita is statistically significant, while the coefficients for the other two socio-economic indicators have the expected sign of effects but are not significant. Column 4 further reveals that the coefficients for all three lose statistical significance and their effects are absorbed by the political indicators.Footnote 14 That is, unlike democratic politicians, authoritarian elites tend to present complex issues through simpler language when addressing wealthier, more educated citizens. However, this primarily occurs because such constituents are typically found in more difficult-to-govern and less politically loyal regions.
Turning to Hypothesis 3, we include an indicator for whether governors are appointed, in Models 5–7. While the coefficient is positive and statistically significant in Column 5, meaning that appointed governors tend to use more complex language than those who are directly elected (by 0.5 in ARI), the effect is not significant in Columns 6 and 7. We therefore investigate this hypothesis further while examining a different subset of elites, as discussed below.
In terms of individual effects, which we control for in Column 7, better educated governors tend to use a plainer communication style, in contrast to longer-serving elites and those with prior careers in business and security services, while the main results are not affected. As an additional test, Column 8 reports the specification with governor-level fixed effects, showing that the results are in line with those estimated with regional fixed effects. Furthermore, in the appendix, we investigate whether governors with specific traits are systematically appointed to particular regions. We find no evidence to suggest that the variability in speech complexity reflects the autocrat’s strategic decisions to install better communicators in less loyal regions.Footnote 15
We also include specifications with two alternative dependent variables. As detailed in the appendix, we construct the five-point index of emotiveness, where a score of 1 stands for a dull, ‘wooden’ language and 5 for highly emotive language. We expect that governors who use simple, easy-to-comprehend language may also score high on emotiveness when they capture the public’s attention and make direct appeals to their emotions. Likewise, to build and emphasize a personal connection with their audiences, governors are likely to frequently use first-person plural pronouns, particularly ‘we’ (Liu Reference Liu2022). As seen from Column 10, the logic of using the ‘we’ share appears to follow that of simple speech. However, the results in Column 9, Table 1, are less clear-cut: while governors from more economically developed regions exhibit more emotion, the coefficients for the political variables are not significant.Footnote 16
Linguistic Complexity and Members of Parliament
Do different authoritarian actors, beyond regional governors, also adjust their speech to their constituencies?
While we find no evidence that the traits of individual governors systematically reflect the characteristics of their regions, in line with the literature (Schoonvelde et al. Reference Schoonvelde, Brosius, Schumacher and Bakker2019; Decadri and Boussalis Reference Decadri and Boussalis2020), we assume that such traits are observable. If governors of politically important regions know that the autocrat pays particular attention to their regions and their own performance, they are likely to use more effective communication as a result, and it may not necessarily be connected to their traits. While this still implies that regional audience characteristics influence speech, even if through the prism of the autocrat’s attention, we can additionally test whether elites adjust to their constituencies directly by extending the argument to a different empirical domain: the lower chamber of the Russian parliament, the State Duma, where the autocrat has less direct oversight of the speeches of a much larger group of elites. We have collected speeches of all members of parliament (MPs) on the floor and supplemented them with the MPs’ biographical information and regional socio-economic characteristics, relying on and extending data from Shirikov (Reference Shirikov2022). In the period of 2003–07 and then again since 2016 (convocations 4, and 7 and 8), MPs were elected under a mixed electoral system – half in single-member districts and half through party lists. In contrast, from 2007–16 (convocations 5 and 6), all MPs were elected via party lists.
Table 2 reports the results of the regression analysis for the State Duma corpus. Since we are primarily interested in hypothesized regional audience effects, the analysis is based on a sample of majoritarian MPs from convocations 7 and 8, as the required regional-level data for convocation 4 are unavailable.Footnote 17 The results in Columns 1–4 indicate that MPs from regions with lower UR vote shares tend to use simpler language. In contrast to governors, MPs do not appear to adjust the speech complexity in response to protests. However, compared to the mixed results regarding governors’ electoral incentives, we find strong evidence in favour of Hypothesis 3 for authoritarian MPs. Specifically, we test whether speech complexity changes when the same individuals transition from being elected in single-member districts (SMD), with stronger incentives to be understood by their constituents, to being elected via party lists, where they depend more on their party leaders instead (Chang and Higashijima Reference Chang and Higashijima2023). Therefore, Models 5 and 6 are estimated on a subsample of the same MPs who were first elected in SMD in 2003 and then elected from party lists in 2007. As indicated by the positive and significant coefficients for the Change from SMD to PR variable in Columns 5 and 6, MPs speak more simply when they hail from specific constituencies to which they are directly accountable, compared to a period when their accountability has been dispersed across the party lists.Footnote 18 In turn, in Models 7 and 8, we test the opposite, and find that the deputies, after initially being elected under SMD in 2016 or 2021, rely on simpler rhetoric, in contrast to their rhetoric when they were elected under party lists previously, in 2007 or 2011.
Table 2. Linguistic complexity of State Duma deputies’ speeches
Note: models 1–4 are panel models with regional fixed effects; models 5–8 are ordinary least squares models.
* p < 0.1,
** p < 0.05,
*** p < 0.01.
The results hold after controlling for various individual characteristics, ruling party membership, and incumbency (whether the MP already served in the Duma before the current convocation). Furthermore, the bottom subplot of Figure 2 reveals that changes in the method of election are associated with changes in ARI of approximately 0.3. In contrast to the effects observed for governors, this is a substantively smaller, yet statistically significant, effect, likely due to a much larger number and greater diversity of parliamentary texts per speaker.
Further Analyses
Our theoretical framework implies that governors not only reflect their audience’s characteristics, but, as indicated by the results in Table 2, also adjust their speech as their own context alters. We further improve the identification as follows. First, we utilize the fact that several officials have served as governors of very different regions concurrently, allowing us to assess whether the same individuals change their rhetoric when addressing very different constituencies. Second, we employ a quasi-experimental interrupted time-series design using a separate dataset, that of governors’ social media posts, to analyse how actors react to two unexpected events (Hypotheses 1 and 2). Third, through a regression discontinuity design, we examine the speech complexity of MPs who narrowly won on SMD tickets versus those who lost and were elected via party lists (Hypothesis 3).
First, we turn to the three available cases in which the same individuals served in more than one region concurrently and for which we have text data. We leverage such transfers to study whether these individuals retain their communicative styles across different regions. The first governor, Nikolay Merkushkin, served in two very distinct regions: first as the head of Mordovia, a titular nationality republic (1995–2012), and later as governor of the Samara region (2012–17). Born in 1951, Merkushkin pursued a career in the Communist Party after receiving his engineering degree and remained in politics after 1991 (ICSID, 2023). Oleg Kozemyako, a former businessman, served as governor of the Amur region (2008–15), then the Sakhalin region (2015–18), and, most recently, of Primorie since 2018.Footnote 19 Sergey Sobyanin was governor of the Tyumen region (2001–05), and after a period in federal politics, he became the mayor of Moscow in 2010 (ICSID, 2023).
Figure 3 illustrates how speech complexity changes when the same individual, while holding the same office and delivering a standard annual address, adjusts to the different constituencies in various regions.Footnote 20 When Merkushkin served as governor of economically poor Mordovia, a region with income per capita below the national median and notorious for its high concentration of penal colonies, he did not make an effort to simplify his annual addresses, instead relying on dense language. In a typical speech from that time, in 2009, he appeared to simply read from a document, peppered with numbers and factual details:
Small and medium-sized businesses should play a very important role in solving the problems associated with the aggravation of the situation in the economy. The republic has formed a fairly wide network and is recognized as one of the best networks in the country for its support. Based on the results of the winning federal tender, in December this year we will create another infrastructure facility, the Regional Fund for Mixed Investments in Small and Medium-Sized Businesses. 200 million rubles will be allocated for its creation (according to the 50:50 principle with the federal budget).Footnote 21
Figure 3. Speech complexity and governor transfer to different regions.
Note: the lines display predicted values of the ARI as the dependent variable and the respective regional indicator as the explanatory variable. MO: Republic of Mordovia; SAM: Samara; AMU: Amur; PRI: Primorsky Krai; SAK: Sakhalin; MOW: Moscow City; TYU: Tyumen.
In contrast, when Merkushkin was at the helm of the much more prosperous Samara region, which ranks in the top decile of economically developed regions and is home to many important industries such as aerospace and defence, with its capital city and major cultural and economic centre of Samara on the Volga River, he adopted a more typical ‘politico’ style, making direct and emotional appeals to his audience. In 2016, his address scored one of the lowest levels of complexity in the text corpus, a stark contrast to the speeches he delivered in the more economically depressed Mordovia:
We have many talented, successful entrepreneurs and large owners. We are ready to help you on any terms with information, with projects, with paperwork, and with various forms of public–private partnerships. Let’s make the Samara region one of the best in the country together. And you, yourselves, your children, and your grandchildren will be proud of you.
Compared to the stark differences in the case of Merkushkin, the socio-economic disparities between Amur, Primorie, and Sakhalin, where Governor Kozhemyako served – all located in the far east – are less pronounced. Correspondingly, Kozhemyako does not adjust his speech complexity as dramatically. Still, when he governed the Amur region, which is in the bottom quartile in terms of development, his 2009 speech exemplifies how one could easily lose their audience, with a high ARI score of 21.9. Upon becoming the governor of Sakhalin, a region known for its natural gas and oil industries, he noticeably switched to a more accessible and engaging language, with direct appeals to his audience. In turn, Sergey Sobyanin presided as the head of Moscow from 2010, one of the most politically important, socio-economically developed regions, with the highest protest record. Sobyanin also faced Alexei Navalny, an opposition leader and successful communicator, as his opponent in the 2013 gubernatorial election (Dollbaum et al. Reference Dollbaum, Lallouet and Noble2022). Figure 3 shows that Sobyanin of Moscow is distinguished by very low complexity scores in his policy speeches, in comparison to his earlier, denser speaking style as the governor of the Tyumen region (2001–05).
Russia’s governors are not typically reshuffled across regions, so we can only rely on a limited number of observations. As a second, different test, we conduct an interrupted time-series analysis to study whether all governors adjust speech complexity following two important and unexpected events. Due to the sparsity of data on annual addresses, we turn to a finer-grained dataset, that of governors’ posts on VKontakte, Russia’s most popular social media platform, which plays a crucial role in direct communication between elites and citizens (Dollbaum and Kim Reference Dollbaum and Kim2024; Khokhlov Reference Khokhlov2024). The data are available for 2020–23, and we focus on whether governors alter their public communication in response to two pivotal events during this period: the 2021 protests following the unexpected return of the opposition leader Navalny to Russia, and the announcement of partial military mobilization in 2022. Both events can be regarded as ‘as good as random’.Footnote 22
Specifically, we fit four multiple-group interrupted time-series specifications (Linden Reference Linden2015) to further test Hypotheses 1 and 2: whether governors from more protest-prone regions, such as Moscow and Saint-Petersburg (Lankina Reference Lankina2015), as well as those from regions with lower UR shares, reduce their speech complexity in response to two events, compared to governors from regions with higher UR vote shares and lower protest activity. Altogether, the sample includes 115,565 posts by 94 governors. Because many governors do not post daily, we aggregate the posts at the weekly level. Moscow and Saint-Petersburg are categorized as high-protest regions (with 189 and 129 recorded protests in 2021, respectively), while regions with UR vote shares below 50 per cent in recent federal legislative elections are classified as low ruling party share regions.
Figure 4 displays changes in the average complexity of social media posts following the 2021 protests (left) and the 2022 mobilization (right), for two groups of regions (arranged in rows). The downward short-dashed lines following the events indicate that the governors respond by making their posts simpler. Specifically, both events lead to a statistically significant weekly decrease in ARI among governors from regions with lower UR vote shares, in contrast to those with higher vote shares (as explained further in the appendix). The estimated weekly decrease in ARI is 0.03 and 0.06, respectively, over a three-month period, resulting in average posts becoming simpler by 0.36 and 0.72, which represents a substantially large effect. Following Navalny’s return, there is a significant decrease in ARI among governors from protest-prone regions (bottom left), but not following the military mobilization (bottom right). It is conceivable that governors did not reduce the complexity of messages in the latter case, as citizens of the most protest-prone regions, such as Moscow, were largely able to avoid conscription for the war against Ukraine (Dollbaum and Kim Reference Dollbaum and Kim2024).
Figure 4. Change in text complexity following protest onset and war mobilization announcement.
Note: the results are based on interrupted time-series analyses for two events (in columns) and two groups of regions (in rows). Treatment groups are regions with UR vote shares below 50 per cent (upper) and high-protest regions (bottom). The data are from Khokhlov (Reference Khokhlov2024).
Third, we further test Hypothesis 3 by leveraging the unique feature of the mixed electoral system for the State Duma elections, specifically the dual candidacy, where candidates can run simultaneously in both SMD and the party lists (Woller Reference Woller2024). Simply put, many MPs who come close to being elected under SMD but lose by a small margin, are then able to be elected under the party list. As a result, we can test whether MPs who narrowly lost in SMD elections, but entered the Duma from the party lists nevertheless, and thus do not face direct accountability to voters, exhibit less engaging speech compared to those elected directly through SMD.
Following Woller (Reference Woller2024), we conduct a regression discontinuity analysis for dual candidates in convocations 3 and 4.Footnote 23 As the treatment variable, we use the margin of vote share for the dual SMD candidates. A positive margin indicates that the candidate won and entered the Duma as SMD deputy. A negative margin implies that the candidate lost the SMD election but still entered the legislature via the party list, due to their high ranking on the party list. Specifically, we compare the speech complexity of the dual candidates (SMD winners and SMD losers/list-elected) in the Duma during the one-year period following the election.
Figure 5 indicates that MPs who could have been elected under SMD but failed and entered the Duma indirectly tend to speak with a higher degree of complexity, likely due to their lower affinity with constituents. In contrast, dual candidates who succeeded in their SMD races and maintained more direct links to their constituents exhibit significantly simpler rhetoric, providing an additional support for Hypothesis 3. Table 20A in the appendix reports the results in full, including for the alternative bandwidths.
Figure 5. Regression discontinuity in time results.
Note: Estimated for the −10 to 10 per cent bandwidth of the vote share margin on the sample of dual Duma candidates, the third and fourth convocations; the graph displays average ARI scores during the first year following the election, with the size weighted by the number of MPs elected under particular margins.
Further analyses regarding our argument are possible. In the appendix, we demonstrate that better communicators survive in office longer and/or get rewarded by being promoted to federal posts, compared to elites who rely on more complex language (Table 25A). That is, we do not find evidence in favour of the opposite perspective – that ambitious governors try to avoid being perceived as popular alternative leaders and opt to intentionally speak in a more complex manner so as not to attract the wrath of a jealous autocrat. Instead, the results imply that the autocrat regards effective communication as important for regime stability.
The appendix provides further details on the data and variables, specifications addressing selection concerns (Table 2A), additional analyses with alternative complexity measures based on listenability metrics and word-embeddings (Tables 3A–6A), and specifications with alternative samples (Tables 7A–9A). The appendix also includes additional mediation and instrumental-variables specifications to probe the relationship between political and socio-economic characteristics (Tables 10A–13A), and validation of the ARI measure through human coding, along with random forest estimates as discussed in the data section. Furthermore, we provide supplementary analyses of all public speeches, not only annual addresses or VK posts, made by governors in two regions (Table 18A); as well as tests demonstrating that simpler speech attracts more public engagement (Table 24A).
Conclusions
Even though making official speeches is a ‘major public activity’ of authoritarian elites (Gill Reference Gill2018, 176), we know relatively little about the logic of effective communication under dictatorship. In this paper, we systematically examined the logic of linguistic simplicity by authoritarian elites, proposing that it reveals important information about their incentives to either remain aloof and disincentivize citizens from politics and policy or reduce their distance from the ordinary citizens in order to appear more responsive. Furthermore, while complex, ‘wooden’ speech is usually a default under authoritarianism (Anderson Reference Anderson1996, Reference Anderson2014; Gill Reference Gill2018; Lasswell Reference Lasswell, Lasswell and Leites1949), we have demonstrated that elites, however, reduce complexity when their strategic context changes. As with Mikhail Gorbachev, who, as a regional party secretary, ‘spoke like a character out of Orwell, turning language to dust’ (Remnick Reference Remnick1990), only to become an effective communicator as a party secretary when addressing the national audience from 1985 onward, we find that contemporary authoritarian elites also deploy speech complexity depending on their audience. Specifically, they tend to speak more simply in settings where citizens’ opinions cannot be completely discarded.
We have also found that more effective communicators are more likely to be promoted. This raises the question of why not all governors simplify their speech. Our analysis does not suggest that elites with specific traits are ‘selected’ for regions that are more challenging and require more effective communicators. It is possible that more ambitious elites use effective communication as a strategy to advance their careers, but without psychometric evaluations, we are unable to measure such unobserved traits. Governors who rely on more complex language may also underestimate the importance of clear communication. We leave this intriguing question for future research, concluding here that, with respect to policy speech, more effective communication serves as a valuable tool for career advancement in autocracies.
Text-as-data methods have provided valuable insights into agenda-setting, preferences, and the strategic use of language in democratic politics (Benoit et al. Reference Benoit, Munger and Spirling2019; Osnabrügge et al. Reference Osnabrügge, Hobolt and Rodon2021; Slapin and Kirkland Reference Slapin and Kirkland2020). By explaining the logic of how elites adjust their public messages in the service of autocratic stability, this paper contributes to the study of public speech in autocracies, which has received comparatively less attention. Future research can expand this approach to other languages and regimes, also investigating whether simpler speech influences public attitudes (Bischof and Senninger Reference Bischof and Senninger2025) under authoritarian rule and how simplicity and emotiveness are related, as well as whether actors can maintain or reduce distance from their audience in speech using other tools of language.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0007123425100914
Data availability statement
Replication data for this article can be found in Harvard Dataverse at https://doi.org/10.7910/DVN/SQWI1A
Acknowledgements
We would like to thank the editor and anonymous reviewers, as well as Raymond Hicks, Julia Gray, Daniil Romanov, Lanabi La Lova, Anastasia Ershova, Jakob Tolstrup, and Paul Chaisty. This paper was earlier presented at EPSA 2023 and COMPTEXT 2023 conferences. We are grateful for the excellent feedback received.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors. One of the authors has carried out this research while on a visiting professorship at the University of Aarhus and would like to acknowledge the host institution.
Competing interests
The authors have no competing interests.
Open access
