Public policy and elections in authoritarian regimes: evidence from the policy on native languages in Russia

Abstract How might public policy changes affect electoral support for authoritarian regimes? Missing from the existing scholarship, which focuses mostly on regimes generating political budget cycles and manipulating electoral rules, is an exploration of how non-fiscal and non-electoral policies may impact incumbent support. We examine this issue with electoral and census data from one of the world’s most prominent authoritarian regimes – Russia – to evaluate the regime’s 2017 change to the policy governing native language instruction, which curtailed minority students’ ability to learn their native languages and faced opposition in some ethnic regions. Examining panel data on presidential elections using fixed effects models, our results reveal that the regime’s support decreased in titular minority areas in 2018. The results also indicate that some of these patterns emerged in previous national legislative elections and thus cannot be solely attributed to the policy change.


Introduction
How might public policy changes affect electoral support for authoritarian regimes at the subnational level? In democracies and autocracies alike, governments utilise public policy strategically to draw non-voters to the polls, capture new voters or reinforce support from existing loyalists (see, e.g. Kramer 1971;Fair 1975;Nordhaus 1975;Tufte 1978;Ames 1987;Pepinsky 2008;Guo 2009;Bechtel and Hainmuller 2011;Layton and Smith 2015;Higashijima 2010Higashijima , 2016. In authoritarian regimes like Russia, Mexico, Malaysia, Egypt and China, the primary scholarly focus has been on political budget cycles, whereby regimes manipulate fiscal policy and engage in pre-electoral spending to mobilise voters (see, e.g. Alesina et al. 1997;Gonzalez 2002;Magaloni 2006;Pepinsky 2008;Blaydes 2011). Missing from this research on public policies in autocracies, however, is an exploration of the potential electoral effects of non-fiscal policies that might impact incumbent support (Cf. Shkel 2019).
In Russia, one of the world's most prominent electoral authoritarian regimes, incumbents have been able to mobilise electoral support in areas with dense ethnic minority populations, that is, in the ethnic republics and autonomous okrugs 1 of the country's federal system (Myagkov et al. 2009;Reisinger and Moraski 2010;Goodnow et al. 2014;White 2015White , 2016White , 2021Moser and White 2017;Saikkonen 2017;White and Saikkonen 2017;Saikkonen and White 2021). However, the Kremlin recently introduced a policy with the potential to impact its relationship with ethnic minority areas and therefore perhaps also its electoral fortunes in those areasin 2017, President Vladimir Putin proposed legislation that changed the rights of ethnic minorities living in ethnically designated areas to receive formal instruction in their mother tongue. What are the potential electoral ramifications of this policy change for the regime?
In this paper, we examine these questions vis-à-vis the Kremlin's change to the native language policy. This legislation rendered instruction in minority languages optional and severely curtailed classroom time dedicated to those languages. Qualitative evidence and survey data indicate that some titular ethnic minorities viewed this policy as an attack on their political and cultural sovereignty and identity as minority peoples in Russia's multi-ethnic federation (Yusupova 2018;Lyubimov et al. 2018;Shkel 2019;Arutyunova and Zamyatin 2021), although this view was not universally held among minority groups (Idel Real 2021). Some ethnic regions also experienced popular protests, submitted formal complaints to the Kremlin and engaged in online activism ranging from petitions to the development of non-governmental organisations in opposition (Lyubimov et al. 2018;Yusupova 2018;Arutyunova and Zamyatin 2021). With seemingly a sole exception (Shkel 2019), scholars have not explored the potential electoral effects of this policy. We investigate this policy change with respect to the regime's relationship to these electoral constituencies by evaluating its performance in these areas in national elections that occurred prior to and following the legislation's passage. To that end, we examine panel data on Russian presidential elections between 2008 and 2018.
We anticipate that opposition among ethnic minority constituents may have impacted regime performance in these areas (in terms of both Putin's vote share and voter turnout) and augmented support for the opposition, namely the Communist Party of the Russian Federation (KPRF). As noted by Shkel (2021) and Arutyunova and Zamyatin (2021), the policy was unpopular among ordinary titular minorities in ethnic republics (Cf. Idel Real 2021). In other words, the opposition seemed concentrated in the electoral environments that had hitherto reported robust incumbent support and electoral turnout (Myagkov et al. 2009;Kalinin 2009, 2010;Goodnow et al. 2014;White 2015White , 2016White and Saikkonen 1 Russia's federal system has 83 regions, discounting the illegal annexations that occurred in 2014. There are six different administrative designations within the system: oblast, krai, autonomous oblast, city of federal significance, republic, and autonomous okrug. Republics and autonomous okrugs are designated areas for a specific ethnic minority that lives there, the so-called titular minority. 2017; Moser and White 2017;Saikkonen 2017). We hypothesise that the possibility of protest voting among titular groups may cause these localities to report flagging regime support and increasing support for regime challengers.
We examine electoral support with two-way fixed effects models that use unique rayon-level electoral and sociodemographic data. We find evidence that is consistent with the expectation that the policy negatively impacted regime support in 2018 in areas with larger proportions of titular minorities. Because we also find that some of these patterns predated the policy's introduction in 2017, the policy change was likely not the exclusive cause of the associations that we uncover.
Naturally, predictors of candidates' vote shares and voter turnout in any country are multicausal. Our analysis draws attention to the potential impact of the policy change while also acknowledging the inherent complexity of evaluating the extent to which any one factor (such as a policy change) could influence the various electoral outcomes that we investigate. We do not claim that the language policy change was of overwhelming importance in titular areas but rather posit that the electoral patterns that we observe are consistent with the policy exercising at least some impact alongside other considerations. These considerations make ruling out all other macrolevel influences and formally validating the electoral and causal effects of the language policy challenging. However, we provide evidence that is consistent with the policy's impact and, as such, offer tentative and modest conclusions, rather than dispositive proof. This paper is organised as follows. Section 2 presents the existing scholarship on the connection between public policy and authoritarian electoral performance. Section 3 considers native language policies in the USSR before analysing this policy's formulation, implementation and monitoring. We examine the extent to which the policy represented a change in centre-periphery relations and detail the regional response. The subsequent section presents our specific hypotheses related to the anticipated electoral effects of the policy. Section 6 presents our empirical approach and data. Section 7 discusses our empirical results, followed by a concluding section.
Public policy and electoral performance in authoritarian regimes Existing research has shed light on the relationship between public policy and electoral performance in authoritarian regimeslike their democratic counterparts, autocratic incumbents leverage economic policy opportunistically to augment their electoral support but also routinely manipulate electoral rules more egregiously and conspicuously than their freely elected peers. A flora of research across regime types has confirmed the existence of political budget cycles, whereby regimes conspicuously boost pre-election spending to signal competence to voters (see, e.g . Nordhaus 1975;Ames 1987;Rogoff and Sibert 1988;Rogoff 1990;Alesina et al. 1997;Magaloni 2006;Blaydes 2011;Higashijima 2010Higashijima , 2016. Indeed, Alesina et al.'s (1993) finding that "pre-electoral manipulations of policy instruments for 'signaling' purposes are likely to be the norm, rather than the exception" illustrates the worldwide pervasiveness of the practice (p. 3). While most comparative policy research finds that incumbents benefit from fiscal manipulation, other scholarship indicates that sophisticated and well-informed voters are less likely to support those who rely on these policies (Peltzman 1992;Brender 2003;Brender and Drazen 2005;Drazen and Eslava 2005). For example, Drazen and Eslava's (2005) investigation of Colombia found that voters rewarded incumbents who pursued pre-election expansionary spending, but not if that spending created large election-year deficits (for other comparative studies of political budget cycles, see, e.g. Ames 1987 on Latin America; Peltzman 1992 on the USA; Alesina et al. 1997 on OECD countries;Kneebone and McKenzie 2001 on Canada;Gonzalez 2002 andMagaloni 2006 on Mexico;Brender 2003 on Israel).
Much of the research on election cycling in dictatorships has focused on singleparty regimes like Mexico, Malaysia, the Philippines, China and Egypt. Gonzalez's (2002) study of the PRI in Mexico reveals its "strong systematic use of public spending in infrastructure as a means to earn votes prior to federal elections" and indicates that fiscal policy manipulation was most egregious during the election quarter itself (p. 205; see also Magaloni 2006). Similarly, Pepinsky's (2008) study of Malaysia under the United Malays National Organization (UMNO) argues that the regime utilised fiscal policy strategically and systematically to conserve domestic legitimacy, demonstrate political dominance to the opposition and ultimately maximise votes. This body of scholarship also uncovers local political budget cycles in autocracies, namely China and the Philippines (see, e.g. Guo 2009;Labonne 2016). Generally, closed regimes are more likely to engage in fiscal manipulations that have conspicuous and immediate effects on their electoral constituencies as opposed to those with longer-term effects (Nordhaus 1975). This phenomenon is exemplified in Russia, where the Putin regime announced one-time payments to all pensioners (who account for nearly 40% of the electorate) in advance of elections, but strategically postponed increasing the pension age, which was widely unpopular (Aasland and Cook 2016).
Another major strand of scholarship on the electoral effects of policy in dictatorships exposes how incumbents fashion electoral policies to preserve their power. In Russia, for example, the regime has manipulated electoral policy multifariouslyit has banned regional parties from competing, introduced strict and virtually insurmountable obstacles to the party registration process, and increased the threshold for legislative representation to circumscribe pluralism, among other initiatives (see, e.g. Wilson 2006;Smyth et al. 2007;McAllister and White 2008;White 2016). Similarly, the PRI regime in Mexico extensively engineered public policy to diminish opposition success and increase its own electoral fortunes. Decades prior to Russia's changes to electoral policy in the early 2000s, the Mexican PRI generated and subsequently augmented rules requiring political parties to become national organisations with membership thresholds in 1946 and 1954, respectively (Magaloni 2006). In Egypt, Blaydes (2011) notes that the Mubarak regime exploited electoral rules to systematically weaken the Muslim Brotherhood.
Of course, distributive policy or policies governing electoral competition may condition who turns out to vote and where, which parties and candidates are on the ballot, and therefore whom voters support. Conversely, our understanding of the electoral effects of non-fiscal and non-electoral policy promulgated in dictatorships is exceedingly sparse (Cf. Shkel 2019), perhaps because of an implicit presumption that policy outside of those areas would not influence electoral outcomes. Shkel's study (2019) investigates the extent to which the regions' changing relationship to the federal centre was impacted by the language policy with a particular focus Bashkortostan, Tatarstan, Komi, Chuvashia and Sakha (Yakutia). In two regions (Bashkortostan and Sakha (Yakutia)), support for Putin plummeted from the 2012 to the 2018 elections, while support in two other regions (Tatarstan and Chuvashia) was nearly unchanged (Shkel 2019, 103). Therefore, the electoral effects of this policy appear uneven and asymmetrical across regions.
We extend this work by presenting the first cross-regional evaluation of the extent to which the policy may have affected the regime's electoral support in the 2018 presidential elections. We use electoral data from all of Russia's counties (rayons) and pair that data with information from the All-Russian National Census, the Russian State Statistics Service (Rosstat). Since we use panel data on Russian presidential elections, our data allow us to uncover electoral effects that may be consistent with the policy change across all regions as well as specific subsets of regions that we expect were the most affected by it, while controlling for unobserved heterogeneity at the local level.
The 2017 native language policy: formulation, implementation and monitoring Since the Soviet era when the Communist regime granted titular languages legal status as "state" languages in its waning days, preserving native languages has been important to titular minorities living in the ethnic republics and autonomous okrugs (Chevalier 2006). Even before that time, the right to receive instruction in these languages was codified in the 1936 Soviet Constitution. These constitutional provisions existed alongside the universal requirement to learn Russian language, a policy promulgated via decree in 1938 (Ibid). The 1991 law On the Languages of the Peoples of the Russian Federated Socialist Republic and the 1993 Constitution of the Russian Federation continued to allow the elevation of titular languages as "state" languages in addition to the designation of Russian as the "national" language in geographic areas designated for ethnic minorities (Ibid). As a result, regional legislatures in these areas provided legal status for titular language(s)ranging from one "state" language to the 13 that were legally recognised in Dagestan (Bowring and Borgoyakova 2016). Approximately one-third of these regions also advanced policies requiring the compulsory study of titular languages in schools, including for ethnic Russians, in their regional constitutions (Arutyunova and Zamyatin 2021). The requirement for ethnic Russians living in those areas to learn the minority language did not exist during the Soviet era and was viewed as a problem (Ibid). In 2004, the Constitutional Court upheld Tatarstan's right to make compulsory the study of native languages regardless of ethnicity and the 2012 Federal Law "On Education in the Russian Federation" (FZ-273) preserved this possibility in the ethnic regions (Ibid).
In July 2017, President Putin convened the Council for Interethnic Relations in Yoshkar-Ola, the capital city of Mari El Republic alongside the northern banks of the Volga River. His seemingly ad hoc comments effectively abolished the compulsory study of non-Russian languages in ethnic republics and autonomous okrugs (Putin 2017). Putin commented that although the teaching and study of minority languages was a constitutionally guaranteed right, such study should be voluntary and that forcing a person to learn a non-native language was unacceptable (Ibid). Thus, learning any language other than Russian would henceforth be optional, forcing students to choose between receiving instruction in their native language or other subjects that may better qualify them for university admissions or workforce entry.
Putin's proposed change to the language policy attempted to resolve to a problem articulated by ethnic Russians living in the titular republicsin Tatarstan, ethnic Russian parents complained that their schoolchildren were compelled to learn Tatar, even though they were not a titular minority (Lyubimov et al. 2018). Indeed, after regional authorities in Tatarstan dismissed their complaints, thousands of parents complained directly to Moscow (Seddon 2018). Although the lawsuit was ultimately rejected, one mother sought damages for what she described as "enormous material harm" to her son on account of his compulsory instruction in Tatar (Chapman 2017). Parental complaints of this nature were reported in other titular republics as well (Ibid).
Per Putin's comments, the policy was implemented by local school administrators and related regional authorities (Putin 2017). These actors were required to align language instruction with the policy by December 1, only 6 months after Putin made his remarks, by increasing the volume of Russian language instruction and concomitantly decreasing that of titular languages (Business Gazeta 2018). Teachers in Tatarstan expressed incredulity at the requirement that they dramatically alter instruction during the school year itself (Baranov and Afonina 2017). Additionally, in cases where regional laws addressed language instruction, legislation would require amendment or repeal so as not to contravene the new federal policy (Jankiewicz and Knyaginina 2019). Prior to the directive, regional legislation in 8 of the 22 ethnic regions required mandatory native language instruction (Jankiewicz and Knyaginina 2019). For example, the Law of the Republic of Tatarstan and the Law of the Chechen Republic stipulated that Russian and Tatar or Chechen, respectively, were studied in equal amounts and constituted compulsory subjects (Ibid).
Putin also declared that widespread policy monitoring would be required (Putin 2017). In September 2017, Presidential Special Order 1710 detailed how monitoring would proceed and compliance would be enforcedthe General Prosecutor's Office of the Russian Federation in conjunction with the Federal Service for Supervision in Education and Science (Rosobrnadzor) would "audit schools in these areas to verify whether minority language study was voluntary or compulsory" and employees found to be in violation would be fired (Yusupova 2018). These two agencies were tasked with monitoring the policy's implementation because educational content is regulated by the federal government in conjunction with local schools, but not at the level of the country's administrative subjects, that is, regions (Jankiewicz and Knyaginina 2019). Putin instructed the Prosecutor General to complete inspections of all schools in the ethnic regions by November 30 (Business Gazeta 2018). Starting with Tatarstan and Bashkortostan, inspectors began their work and found that titular languages were taught compulsorily and widespread termination of teachers ensued (Ibid). Violations of varying severity were found in all ethnic republics (Radio Svoboda 2017). The Prosecutor's Office required local school authorities to hold meetings with parents and require them to provide a written statement testifying to their children's native language (Baranov and Afonina 2017).
Despite the centre's jaundice-faced relationship to the regions under Putin, it was somewhat surprising that the policy was enforced and monitored by federal agencies as coercively as it was. In addition to the fact that the policy fell within the purview of those two agencies, monitoring was likely conducted by them to prevent subnational agency loss. Elected regional officials may not have enforced it as strictly or intrusively as dictated by the Presidential Order because many regional elites in the ethnic republics are themselves members of the titular group, against whom the policy was directed (Aliyev 2018). Putin himself seemed to recognise the prospect of agency loss, as he "stress[ed] [ : : : ] that school administrations and regional authorities should guarantee [the] right[s]" associated with language instruction (Putin 2017). Indeed, some governors of the ethnic republics directly challenged the policy and presumably could have erected nearly insurmountable obstacles to implementation in their respective areas. Although the Kremlin could have implemented an incentives scheme to reduce this potentially significant agency loss, it instead favoured the Prosecutor General's office, whose head is nominated by the President and rubber-stamped by the legislature. Therefore, because the policy may have run counter to regional elites' interests, the centre may have worried that the implementation and monitoring of the policy could suffer if not controlled vigilantly.
In response to Putin's comments, the Russian government adopted new amendments to 2012's "On Education in the Russian Federation." Even though the language policy had already been promulgated and implemented in the ethnic regions based on Putin's directives in 2017, these amendments codified his earlier comments making the learning of minority languages voluntary (based on parents' request) and limiting their instruction to a maximum of 2 hours per week. The bill was co-authored by two Kremlin loyalists representing Chuvashia (Lyubimov et al. 2018). In the State Duma, the bill passed by a vote of 373 to three and nearly all the deputies representing the ethnic republics voted in favour (Ibid).

Regional responses to the Kremlin's policy change
Responses to the new policy varied. Along a continuum, the Kremlin's actions were viewed anywhere between generally unremarkable on one side to provocative and politically divisive on the other. Rustam Minnikhanov, the president of Tatarstan, alongside his minister of education declared that Putin's comments did not apply to the region and that Tatar language classes would remain mandatory. The minister of education was subsequently dismissed because of the controversy (Chapman 2017). Although authorities blocked protestors from holding rallies in the capital, activists managed to distribute Tatar alphabet textbooks to local members of parliament there and organised flashmobs to evade detection by authorities in advance (Yusupova 2018). At least in Tatarstan, the policy's introduction coincided with the Kremlin's refusal to renew its bilateral federation treaty with the region, the last ethnic republic that had such an agreement, which had allowed it a special relationship with the centre (Halbach 2018). Shkel (2019) notes that regional elites in Tatarstan embarked on a "consolidated protest against the actions of the federal centre and unambiguously positioned themselves as protectors of the preferences granted to the titular ethnic group" (p. 96).
Other ethnic republics joined Tatarstan in opposition to the Kremlin's offensive. In its capital city, Bashkortostan witnessed an unsanctioned protest that reached upwards of 2,500 people, the largest mobilisation that had occurred in the region for any reason (Yusupova 2018;Arutyunova and Zamyatin 2021). At that protest, the change from mandatory to voluntary study of Bashkir was called an "ethnocide against Bashkirs" (Arutyunova and Zamyatin 2021, 12). Several ethnic republics, including Altai, Bashkortostan, Kabardino-Balkaria, Tatarstan and Tuva, passed resolutions or sent letters to the Kremlin formally opposing the bill (Lyubimov et al. 2018). One Russian specialist even speculated that the new policy may have been responsible for an uptick in conflict incidents and violence in certain ethnic republics, such as Chechnya and Ingushetia, following Putin's comments (Hauer 2018). In some of these areas, opponents argued that it aimed to progressively Russify the country's hundred-plus ethnic minority groups by diminishing minority culture and the country's status as a federation (Hauer 2018;Lyubimov et al. 2018). Some activists argued that the Kremlin's actions were also an attempt to eradicate the existence of ethnic minorities themselves; "no language-no people" was a frequent slogan of the protest movement (Arutyunova and Zamyatin 2021). Therefore, some titular minorities interpreted the policy as a targeted assault by the centre on the cultural and political autonomy of the ethnic republics.
Approximately 1 year after Putin's remarks, some ethnic minorities created an official nongovernmental organisationthe Democratic Congress of the Peoples of Russiadesigned to oppose the policy, promote cultural diversity, resist initiatives to alter Russia's administrative structure and resist efforts to undermine the constitutional protections afforded to minority languages (Yusupova 2018;Lyubimov et al. 2018). Additionally, grassroots activists produced an online campaign imploring the Kremlin to abandon its policy using the viral hashtag #StopLanguageGenocide (Ibid). Petitions also circulated on Change.orgone petition described as "No law against native languages" received upwards of 40 thousand signatures (Arutyunova and Zamyatin 2021, 12).
Beyond changing laws surrounding the teaching and learning of native languages in the ethnic regions, this policy removed the "last element of the republics' status" and thereby contributed to the progressive weakening of federalism that had been undertaken as a matter of priority during Putin's tenure (Arutyunova and Zamyatin 2021, 5). Indeed, as mentioned, one of the primary distinctions between the ethnic republics and regions with other administrative designations was the right to establish titular state languages that could be elevated to equal status with Russian (Jankiewicz and Knyaginina 2019). When Putin took office at the turn of the millennium, the Kremlin embarked on a crusade to fundamentally transform the federal structure by introducing a new layer of administration between the centre and the regions, altering the share of government revenue available to the regions, and abolishing the direct election of regional governors in favour of executive appointment, among other notable reforms (Petrov and Slider 2007;Treisman 2011). Crippling the federal structure and reinforcing the "power vertical" has been politically sensitive in some ethnic regions because the administrative structure designates areas for ethnic minorities and allows them a degree of political and economic autonomy commensurate with that administrative status (Halbach 2018). Situating the language policy within this broader institutional context, titular minorities residing in their own ethnic areas may view preserving their native languages as a critical aspect of their autonomy from the federal centre and the encroachment on their linguistic rights as an unwarranted and unconstitutional intrusion (Cf. Idel Real 2021). Furthermore, as mentioned, more than one official language was recognised in these areas, signifying that the status of minority languages was equal to that of Russian. Therefore, although ethnic regions had absorbed the Kremlin's institutional reforms before, this policy presented a uniquely severe assault on the identity and culture of some ethnic minorities.

Hypotheses
As detailed above, this reform generated opposition among some titular constituents that may translate into diminished support for the regime. Before the policy change, ethnic regions had a robust system of ethnic institutions that dated to the Soviet era but were preserved post-collapse, which included territorially designated areas for non-Russian minorities, independent education systems that allowed for schooling across subjects in the titular language, laws protecting minority languages and various forms of ethnic privileging that elevated the position of titular minorities within regional political institutions (Gorenburg 2003). The multifaceted institutionalisation of ethnicity in Russia's system of ethnic federalism has elevated ethnic identity as the "dominant form of self-identification among non-Russians" (Gorenburg 2003, 3). Additionally, language has been considered the "most important characteristic that distinguishe[s] ethnic groups from each other" (Gorenburg 2003, 37); therefore, from the perspective of some titular minorities, native languages represent a critical asset worth defending against attempts at encroachment by the ethnic majority. Indeed, these groups view their respective cultures and languages as public goods that hold intrinsic value (Gorenburg 2013).
Long-standing and deeply embedded ethnic institutions coupled with the dominance of ethnic identification and the centrality of language to that identification "creates a number of preferences for the titular ethnic groups, which increases the relevance for them of national issues related to the preservation of the privileged status of their language and culture" (Shkel 2021, 21). The Kremlin's actions in 2017 compromised a central pillar of ethnic institutions in these areas, and some titular minorities likely perceived the reform as a direct threat to the most salient dimension of their self-identification. Policies viewed by minorities as discriminatory and threatening may play a key role in fomenting a grievance among those with a shared ethnicity (Giuliano 2011). As a result, it is reasonable to expect that some non-Russians likely suffered acutely from a psychological loss of these powerful symbolic resources.
Since some regional presidents and local co-ethnic elites were also opposed to the policy, the depth and breadth of opposition among their constituents was likely directed squarely at federal authorities. Indeed, focus groups conducted with titular respondents about the language policy indicated that they "blamed the federal authorities" instead of co-ethnic regional elites (Shkel 2021, 14). It is important to note that these groups have challenged the federal government in the past, both during the Soviet and post-Soviet era, and have utilised elections as a forum for protest (Gorenburg 2003). For example, Tatars boycotted federal elections between 1991 and 1993while nationwide turnout in the 1991 presidential election was 75%, participation in Tatarstan plummeted to 35% (Gorenburg 2003, 136), demonstrating the efficacy of their resistance to the centre. Moreover, titular areas learned that opposing incumbents in national elections could generate dividendsregions that voted against Yeltsin in 1991 and against pro-Kremlin Russia's Choice in 1993 were more likely to extract largesse from the federal government in the form of intergovernmental transfers (Treisman 1996). In short, the oppositional electoral strategies pursued by ethnic minorities "paid off far better than complaisance" (Treisman 1996, 322). Therefore, the retribution that may be meted out because of the 2017 reform would likely influence national elections rather than regional contests because of regional elites' shared opposition to the policy as well as ethnic minorities' history of protest voting against national incumbents that generated dividends for the Kremlin's opposition at the time. The language reform may have provoked titular minorities into fighting for linguistic rights via electoral protest, an avenue of opposition that had proved effective in the past.
The Kremlin's actions seemingly disregarded the preferences of some of those in the ethnic regions and titular minorities signalled their opposition multifariously, as detailed above. This increased the likelihood that the incumbent vote would be affected in ethnic areas, although electoral outcomes in those areas are undoubtedly impacted by other factors too, such as economic concerns (Idel Real 2021). However, we predict that the policy change should depress regime support in rayons with higher shares of titular minorities if regime supporters stayed home rather than voting or voted for the opposition as hypothesised in H2.
Thus, we expect: H1: The association between the proportion of titular minority residents in a rayon and the regime's vote share should be negative after the policy change.
Concomitantly, we anticipate that titular areas might begin to direct electoral support to the Kremlin's opposition after the policy change. There were candidates from only two political parties that competed in the presidential elections in both 2012 and 2018the Communist Party of the Russian Federation (KPRF) and the Liberal Democratic Party of Russia (LDPR). Gennady Zyuganov, who has led the KPRF since the collapse of the Soviet Union, ran in 2012 and the Communists allowed Pavel Grudinin to compete for the presidency as their candidate in 2018. Vladimir Zhirinovsky, longtime leader of the ultranationalist and xenophobic LDPR, ran in both contests. Specifically, we expect votes in titular areas to accrue to candidates with party platforms that were oriented around the defence and resurrection of cultural rights. The KRPF's party platform stated that it "deems it necessary" to "ensure that cultural benefits are generally accessible" and to "protect the national cultures of all the country's peoples" (Communist Party of the Russian Federation (KPRF) 2022).
Consequently, we predict: H2: The association between the proportion of titular minority residents in a rayon and the KPRF's vote share should be positive after the policy change.
Finally, we explore another mechanism through which the regime's electoral outcomes may have shiftedvoter turnout. While some ethnic minorities may have expressed their dissatisfaction by voting for the regime's opposition in lieu of Putin as anticipated in H2, others may have decided not to vote altogether. Existing research substantiates the use of abstention as a method of electoral protest and argues that it "constitute[s] a significant political act rather than passivity" (Karklins 1986, 449; see also Alvarez et al. 2018;Uggla 2008;Moral 2016;Runciman 2016). Moreover, the policy changes may have also affected the ability of the local elites to mobilise voters to turn out. Therefore, variation in voter turnout is another important electoral outcome to explore with respect to the policy change.
Accordingly, we anticipate: H3: The association between the proportion of titular minority residents in a rayon and voter turnout should be negative after the policy change.

Data, indicators and estimation strategy
We inspect pro-regime vote shares and trends in electoral turnout in Russian presidential elections between 2008 and 2018, as the latter election was the first national contest held after the policy change had taken effect. We investigate the policy's electoral effects with rayon-level panel data from the 2008, 2012 and 2018 presidential elections. We also examine trends in incumbent vote shares in the 2011 and 2016 national parliamentary elections to assess the extent to which the trends that we observe in 2018 potentially predated the policy reform. Rayons (which are roughly equivalent to US counties) are the smallest units for which consistent electoral and sociodemographic data are available. The outcome variables (Putin's vote share, 2 KPRF candidate's vote share and voter turnout) are measured at the rayonlevel using data from the Central Electoral Commission of the Russian Federation. 3 To examine how the proportion of titular minorities in a rayon influenced electoral changes observed in 2018, we use an interaction term that specifies that the election took place in 2018 (binary variable denoting post-policy treatment) and the proportion of the titular minority population residing in given rayon (continuous variable) using data from the 2010 All-Russian National Census. 4 We employ linear regression models with both rayon and year fixed effects. The rayon fixed effects control for omitted variable bias and unobserved heterogeneity between rayons.

Empirical analysis
We begin by examining descriptive patterns in pro-regime electoral support and electoral turnout in the 2008, 2012 and 2018 presidential elections. The first graph 2 Although Putin did not run for president in 2008 due to constitutional term limits, we use Dmitry Medvedev's vote share to assess pro-regime support. Before the 2008 election, Putin was named Medvedev's Prime Minister.
in Figure 1 depicts trends in voter support for pro-regime presidential candidates (Dmitry Medvedev in 2008 and Vladimir Putin in 2012 and 2018) by the predominant ethnic composition of rayons across the country to preliminarily assess the patterns we are interested in. While support for these candidates was remarkably stable between 2008 and 2012 in majority titular rayons, there was a small decrease in 2018, which is consistent with the notion that the policy change may have impacted electoral outcomes in titular areas. However, it should be noted that, even if regime support declined from the previous elections in these areas, it was still very high in 2018. This contrasts with the conspicuous increase in regime support in rayons inhabited mostly by non-titular residents in 2018. The second graph in Figure 1 reveals trends in voter turnout across the same period and illustrates that it diminished somewhat in 2012 and 2018 from a record-breaking level of upwards of 90% in rayons inhabited mostly by titular minority residents in 2008. However, levels of electoral turnout do not show as significant of a decrease in the 2018 elections in rayons inhabited by majority titular residents as does electoral support for the regime.
To probe whether the electoral trends in the ethnic areas may have preceded the policy change (which took place in 2017), we examine similar patterns in the 2011 and 2016 parliamentary elections. As can be seen from Figure 2, a noticeable drop in United Russia's support in 2016 in majority titular rayons is apparent. Therefore, it appears that some of these electoral trends may have predated the policy's introduction in 2017. It is important to note that the graphs in Figures 1 and 2 depict electoral trends for rayons that are majority titular or majority non-titular as a binary indicator. However, in our subsequent statistical models, we leverage a continuous measure of the proportion of titular minorities at the rayon level to avoid creating arbitrary cut-off points in our data and increase the precision of our analysis.
The results of the main regression models are presented in Table 1. Based on our theoretical expectations, we anticipated that the policy change would negatively affect the regime's electoral support in 2018 in titular minority areas (H1). As evident in Table 1, the Model 1 results lend substantiation to this hypothesisindeed, when we examine panel data, we find that the "post-treatment" interaction term negatively impacts the regime's electoral support. The interaction term is highly statistically significant and of substantive magnitude.
We also hypothesised that there should be a positive association between the KPRF candidate's electoral support and the "post-treatment" interaction term (H2). The empirical results substantiate this hypothesis, seen in Model 2. Indeed, we find that the interaction term is positive and achieves the highest threshold of statistical significance. Furthermore, we expected that the association between voter turnout and the proportion of titular minorities would be negative (H3). Paralleling the results with respect to incumbent support in Model 1, our expectation about changes in the level of political participation is born out in the results of Model 3, as the interaction term is negative and highly statistically significant.
Next, we specify the same models with a subset of the data that only includes rayons located in federal republics or autonomous okrugs to detect at a higher degree of empirical resolution which types of rayons are driving the patterns we observe (see Models 4-6 in Table 2). The results for Models 4 and 5 are broadly similar to Models 1 and 2, although the magnitude of the coefficients for the interaction terms is smaller. In Model 6, the coefficient for voter turnout among the subsample is positive but does not achieve conventional levels of statistical significance, indicating that it did not exercise a systematic effect. The results from Models 4-6 are consistent with the theoretical logic that protest voting may have occurred. In short, votes for Putin declined, votes for the opposition candidate increased and turnout was not systematically influenced. Voters in these areas still went to the polls (as also demonstrated in Figure 1) but voted against the regime and in favour of the opposition candidate who campaigned on the defence and resurrection of cultural rights, as noted above. This finding also suggests that turnout-oriented mobilisation networks in ethnic republics may not have been harmed because such harm may have been reflected as a negative coefficient in Model 6. Since much of the previous qualitative evidence and survey results focus on the response to the policy in Tatarstan, Model 7 only includes rayons located in that republic. The interaction term is in the expected direction, which aligns with the results of the previous Notes: Robust standard errors are in parentheses (clustered at the rayon level). *p < 0.05, **p < 0.01, ***p < 0.001. models, although the coefficient does not achieve conventional levels of statistical significance. Collectively, our results are robust to various alternative specifications and approaches to examining the data and are consistent with our theoretical expectations. Using data on United Russia's rayon-level vote share in the 2011 and 2016 parliamentary elections, we subsequently probe the possibility that these electoral patterns preceded the 2017 policy change. Model 8 includes an interaction term between the year 2016 and the share of titular minority population in a given rayon to parallel the modelling strategy with respect to the post-policy change election that was the 2018 presidential contest. We find that the coefficient for this interaction term is again negative. These results suggest that the negative trends between the proportion of titular minorities and incumbent support in national elections preceded the 2017 policy reform. Therefore, while it can be argued that the policy's effects may have exacerbated these pre-existing trends, they are likely not the exclusive cause of the empirical associations that we observe when examining the presidential elections (Tables 2 and 3).
We also examine qualitative and survey evidence on the effects of the policy change. Newspaper reports on regional leaders' assessments on the policy change suggest that these were rather negative. For example, the President of Tatarstan Minnikhanov publicly mused that the Kremlin's investigations of local schools were "terrorizing" school directors and teachers and that "tomorrow we will organize the elections with them," intimating that the Kremlin should tread carefully in Tatarstan because it would be asked to deliver high levels of support for Putin in the 2018 presidential elections (Antonov 2017). Moreover, the Tatar President wondered whether "opposition structures" could use the conflict to their advantage and commented that the situation could reflect badly on President Putin so close to the elections (Antonov 2017;Shkel 2019). A few months after Putin's comments and around the time that the inspections began, approximately 60 members of the Tatar intelligentsia submitted a letter decrying the policy and reminding the regional government of their elected positions: "Dear deputies, elected by the people for a while [ : : : ] Your decision is our final frontier. Depending on how it will be, you will receive our support or alienation.
[ : : : ] There will be no one to support you. And this will certainly happen if the civil society of Tatarstan and its leadership do not show the determination to stand within the legal framework for their positions to the end. We urge you to think about this before finally deciding the fate of the Tatar language" (Real Time 2017, emphasis added, see also Chapman 2017).
This suggests that regional elites and titular groups alike expected the regime's electoral performance to suffer in the ethnic republics and that the opposition might be able to capitalise on the situation.
Other qualitative and survey evidence similarly reveals that the language policy negatively impacted regime support. For example, focus groups comprised of titular minority participants confirms that the issue had a "wide public resonance with many of them" (Shkel 2021, 3), that they "blame[d] the federal authorities" and that "the language problem [was] decisive in their political choice" (Shkel 2021, 14). Indeed, non-Russian participants indicated that the reform "undermined the confidence in the Russian president" (Shkel 2021, 10). Moreover, one Tatar participant stated that "20 million Muslims in Russia turned their backs on Putin after he said this in Yoshkar-Ola" and another said that "this decision greatly influenced us. And now we will not vote for him. [ : : : ] After such a statement, Putin lost our support" (Ibid). Furthermore, survey conducted before the policy reform indicated that 95% of Tatars wanted their children to speak the native language (Inkazan 2017)collectively, our results are consistent with the interpretation that titular groups translated their commitment to native language education and the salience of the policy change into electoral retribution.
However, other qualitative and survey evidence suggests that there was heterogeneity in the response among different ethnic regions and also among titular constituents who share the same ethnic identity. For example, the lack of media coverage pertaining to Buryatia or Karachay-Cherkessia suggests that the policy change seemed to galvanise neither public protests nor submission of formal letters of opposition to the Kremlin in some ethnic areas (Shkel 2021). Moreover, a survey conducted in summer 2018 reveals that some respondents in Tatarstan were also concerned with inflation, job access and corruption, which may have overshadowed the importance of the language policy among some voters (Idel Real 2021). Therefore, qualitative and survey evidence suggests that local-level responses to the language policy in ethnic minority areas varied.

Conclusion
To date, the literature on public policy in dictatorships has almost exclusively examined the influences of fiscal and electoral policies on regime support, perhaps because of an implicit presumption that policy outside of those two areas would likely not influence electoral outcomes. Our paper presents one of the first studies that focuses on the influence of non-fiscal and non-electoral policy on electoral support for incumbents in non-democratic regimes. In particular, we examine the effects of the change in the native language policy in Russia on support for the Putin regime. The change to the native language policy represented a real and symbolic final "offensive" against the self-determination rights of titular minorities and generated opposition in those areas of the country. However, despite the language policy's direct impact on ethnic minority areas that have long served as the regime's electoral strongholds, its electoral consequences have been hitherto left mostly unexplored (Cf. Shkel 2019).
We present the first systematic, subnational and subregional study of the electoral consequences of this policy reform with two-way fixed effects analyses of unique panel data of rayon-level electoral and sociodemographic information. Our statistical findings with respect to the presidential election data align with our theoretical predictionsthe policy change diminished pro-regime votes in areas with greater proportions of titular minorities in the 2018 presidential elections. However, our interrogation of these patterns with parliamentary election data also revealed that some of these trends predated the introduction of the policy change in 2017. This suggests that the electoral changes we uncover cannot be exclusively attributed to the policy reform. In addition, while on the whole suggesting considerable anger at the policy change that was directed at the federal regime, qualitative and survey evidence also reveal heterogeneity in local-level responses to the reform.
Naturally, there are limitations to our study. While our fine-grained data are particularly well suited for examining changes across all the rayons over time, it is not amenable for analysis of individual-level electoral behaviour. Therefore, our results should be complemented with individual-level studies that combine data on vote choice with information about the extent to which different types of voters opposed the policy.
In this article, we have undertaken some of the theoretical and empirical work needed to determine the extent to which non-fiscal and non-electoral policies matter for the electoral support of non-democratic regimes. A worthwhile task for future research would be to determine whether other public policies impacted the incumbent regime in Russia, as well as in authoritarian regimes around the world. Public policies may have discernible and perhaps unforeseen electoral consequences, even in regimes that rely on coercion rather than the consent of the governed.