Electoral authoritarian regimesFootnote 1 are now the most common non-democratic regime type (see, for example, Schedler Reference Schedler2013). Despite extensive research on electoral turnout in developed democracies, we know relatively little about the determinants of electoral participation in (electoral) autocracies. Electoral mobilization and the ‘socioeconomic roots’ of electoral authoritarianism are crucial for understanding electoral authoritarian regime dynamics and democratic regression. Yet, electoral behaviour has been less studied in the electoral authoritarianism literature. Most of the existing literature focuses on incumbents’ coercive and organizational capacity as well as electoral fraud in explaining authoritarian resilience (see, for example, Levitsky and Way Reference Levitsky and Way2010; Schedler Reference Schedler2013). While these are undoubtedly important, the socioeconomic roots of electoral authoritarianism and the ability of the regime to mobilize its citizens in elections also need attention. In contrast, the literature on dominant party regimes has emphasized that electoral clientelism and socioeconomic dependency can sustain non-democratic regimes even in the absence of overt coercion and large-scale fraud (see, for example, Greene Reference Greene2010; Magaloni Reference Magaloni2006).
This article studies the socioeconomic roots of electoral authoritarianism by using original local-level data from a prominent contemporary electoral authoritarian regime, Russia. The article shows how electoral mobilization of certain institutionally and socioeconomically state-dependent demographic sectors was a key part in Russia’s transition from a competitive to hegemonic authoritarian regime between 2000 and 2004. The theoretical part of the article focuses on accounting for the social and institutional context that the regime was able to draw on in electoral mobilization. First, building on previous work on post-Soviet patronage politics (Allina-Pisano Reference Allina-Pisano2010; Frye et al. Reference Frye, Reuter and Szakonyi2014; Hale Reference Hale2003, Reference Hale2007) the article outlines how particular structural features of post-Soviet social geography alleviate the commitment problems inherent in clientelistic exchanges. Previous work has suggested that voter mobilization and ‘machine politics’ in Russia have concentrated on specific sectors of the population, such as the rural areas and ethnic minority areas (Hale Reference Hale2003, Reference Hale2007; Matzusato Reference Matzusato2001; Reisinger and Moraski Reference Reisinger and Moraski2010). This article builds on that analysis and extends the theory on hierarchies in resource allocation at the local level and outlines how these created opportunities for turnout mobilization and monitoring. This article contributes to the electoral authoritarianism literature by showing how socioeconomic dependency on the state and the targeting of certain demographic constituencies can account for democratic regression even without large-scale coercion. The findings of this article also add to the influential literature on Russian electoral geography (for a review, see Clem Reference Clem2006), as well as the voluminous literature on electoral fraud in Russia (see, for example, Mebane and Kalinin Reference Mebane and Kalinin2009; Myagkov et al. Reference Myagkov, Ordeshook and Shakin2009). While electoral fraud and curtailing the opposition’s opportunities to contest the election fairly were clearly part of Russia’s authoritarian trajectory in the 2000s, this article stresses that an important part of the explanation lies in the complex legacies of Russia’s socioeconomic transition and the regime’s ability to harness these for electoral mobilization.
The article uses an original, author-collected local-level data set and multilevel modelling, which allow hypotheses of voter mobilization in Russia to be tested at the rayon (equivalent to US county) level, while also controlling for several alternative explanations of turnout patterns from the comparative electoral participation literature. Previous studies have yielded valuable insights using region (‘province’) level data,Footnote 2 but have not been able to account for the considerable within-region heterogeneity and local-level variation in Russia. As such, the findings of this article also complement the important work of Robert Moser and his co-authors who have used rayon-level data to investigate the determinants of ethnic voting and electoral fraud in Russia (Goodnow and Moser Reference Goodnow and Moser2011; Goodnow et al. Reference Goodnow, Moser and Smith2014).
Based on the theoretical framework, the article develops a set of hypotheses of turnout mobilization in rural and minority ethnicity localities which are tested on an original local-level data set which matches rayon (TIK)Footnote 3 -level electoral results with rayon-level socioeconomic data from the Russian 2002 census and region-level data from the Russian federal statistical service, Goskomstat. The article examines several multilevel models on turnout in the Russian presidential elections in 2000 and 2004, and finds strong support for the hypotheses, even when controlling for community size, other socioeconomic characteristics of the rayons, electoral competitiveness and many region-level characteristics. Turnout mobilization patterns become more pronounced as the regime becomes more hegemonic in 2004. In addition, these effects differ in significant ways between the ethnic titular republics and predominantly Russian regions. The results also hold when a proxy for electoral fraud is included in the models.
The article is structured as follows. The next section discusses the ‘signalling’ function of elections and pressures for electoral mobilization in authoritarian regimes. The article then develops the theoretical framework and shows how the peculiarities of post-Soviet transition created conditions amenable for patronage based turnout mobilization in Russia. This section also presents a set of testable hypotheses in predicting variation in rayon-level turnout patterns in 2000 and 2004. The subsequent section outlines the data and is followed by the statistical analysis and discussion of the results. The last section concludes.
Electoral Mobilization in Authoritarian Regimes
Electoral mobilization strategies differ between democracies and electoral autocracies. In democracies political parties and the media invest considerable efforts in electoral mobilization (see, for example, Rosenstone and Hansen Reference Rosenstone and Hansen2003) and try to mobilize sectors which do not participate so actively (such as the less wealthy, less educated, or the young). In contrast, authoritarian electoral mobilization tends to be non-programmatic and based on instrumental exchanges. To ‘deliver’ high turnout (or high overall electoral results), electoral autocracies often target the demographic groups which are most socioeconomically vulnerable and dependent on the state, such as poor or illiterate voters, or those in rural areas (Blaydes Reference Blaydes2006; Magaloni Reference Magaloni2006). Electoral clientelism – the distribution of targeted (material) benefits in return for electoral support – can also take place in democracies (see Kitschelt and Wilkinson Reference Hale2007). However, electoral clientelism can be an especially powerful tool in authoritarian regimes if the state has a large role in the economy and the regime is able to use its control of economic resources in voter mobilization (see, for example, Gervasoni Reference Gervasoni2010; Greene Reference Greene2010; McMann Reference McMann2006).
Russia is a good case to use to examine authoritarian electoral mobilization as the country’s transition to market was difficult, and some economic sectors, even if formally privatized, remained dependent on state support. Electoral participation was mandatory in the Soviet Union, and the regime invested considerable efforts in electoral mobilization (Karklins Reference Karklins1986). With the introduction of the democratic constitution in 1993 the citizens acquired the right to vote in multiparty elections, but also the right not to participate in elections if they opposed the regime or if there were no credible opposition candidates (voting is not compulsory in Russia). Overall turnout levels have declined in Russia since 1991. Yet, there are striking differences in turnout levels within the country: for example, in the 2004 presidential elections there was an almost 60 percentage point difference between the highest and the lowest rayon-level turnout in Russia.
This article focuses on turnout mobilization in Russian presidential elections in 2000 and 2004. Turnout levels are important in electoral authoritarian regimes, as low levels can ‘signal’ regime unpopularity (despite a hegemonic victory by the incumbent) and delegitimize the elections results (Magaloni Reference Magaloni2006; Simpser Reference Simpser2013). Both the 2000 and 2004 elections featured a genuinely popular incumbent and a rather weak opposition, and there was little uncertainty about the outcome.Footnote 4 Turnout mobilization was even more important in 2004, as the elections were in practice non-contested, and parts of the opposition advocated electoral abstention as a protest against the regime (Sakwa Reference Sakwa2005). In such conditions, the central dilemma for the regime was thus not simply to win the elections, but to ensure a ‘decent’ turnout to ‘signal’ regime popularity, to legitimize the outcome and to guarantee that the minimum turnout of 50 per cent, as stipulated by the electoral code, was met.
Patronage Politics in the Post-Soviet Space
Patronage Politics: Contingency and Commitment Problems
The literature on electoral clientelism has challenged the programmatic ‘responsible party government’ model of voting and stressed that voting can also be based on a different, patronage-based, linkage (Kitschelt and Wilkinson Reference Hale2007: 2) in which vote choices are based on instrumental calculations or may not even be autonomous. Electoral clientelism refers to the distribution of targeted (material) benefits given in return for electoral support and can involve either vote buying (that is, mobilizing support for a particular candidate/party) or turnout buying (mobilizing turnout) (see, for example, Nichter Reference Nichter2008). Patronage politics – a subcategory of clientelism – refers to the ability of politicians to use state resources, such as access to public sector employment or welfare, for electoral gain (Stokes Reference Stokes2007).
Voting is considered a low-cost (and relatively low expected benefit) decision in advanced democracies (see, for example, Aldrich 1993). In contrast, under patronage politics, not voting or voting for the opposition can incur high costs, as the contingency of the clientelistic exchange makes the costs and benefits of voting fundamentally different (Kitschelt and Wilkinson Reference Hale2007). Clientelism can involve either rewards, such as targeted benefits, or be based on a threat of the withdrawal of those benefits (a ‘punishment regime’ (Magaloni et al. Reference Magaloni, Diaz-Cayeros and Estevez2007)). Under the latter, access to benefits is dependent on (collective) voting behaviour, and there are implied sanctions for non-voting or voting for the opposition, such as losing one’s or the community’s livelihood. Clientelism does not always entail explicit coercion. Under repeated and sustained interactions, which create unwritten ‘rules of the game’ and expectations of behaviour, clientelism may not even need enforcement (Kitschelt and Wilkinson Reference Hale2007).
Electoral clientelism involves complex delegation and entails several commitment problems. How can the brokers check whether the clients fulfil their share of the bargain – that is, turn out to vote or vote in a specific manner – especially when the ballot is secret? In Susan Stokes’s (Reference Stokes2005) model the effectiveness of electoral clientelism depends on several things, such as the value of the benefits and sanctions to the clients, and the effectiveness of the monitoring regime. Several authors have noted how peculiarities of the post-Soviet political economy and social geography can be particularly amenable to solving these commitment problems (Allina-Pisano Reference Allina-Pisano2010; Frye et al. Reference Frye, Reuter and Szakonyi2014; Hale Reference Hale2007). These include the role of private (but state-connected) enterprises in welfare and social service provision, the continued (in)direct role of the state in some sectors of the economy and the legacies of Soviet social geography which mean that voting precincts can overlap with, for example, agricultural enterprises.
The post-Soviet patronage model is based on the politicization of state-connected goods and services, such as the continued access to jobs and welfare services or the provision of public infrastructure, which are contingent on collective voting behaviour. It involves a ‘tacit renegotiation of the social contract threatening loss of settled entitlements like public infrastructure, social services, and compensation for labour’ (Allina-Pisano Reference Allina-Pisano2010: 374). As such, the logic is fundamentally ‘subtractive’ and based on the threat of the withdrawal of these essential goods and services (Allina-Pisano Reference Allina-Pisano2010). As Frye et al. argue, the high sanctions involved in this model are likely to reduce the likelihood that the client will renege on his or her voting commitment (Frye et al. Reference Frye, Reuter and Szakonyi2014). The politicization of existing benefits also makes them cost-effective for the machine (see Stokes Reference Stokes2005).
The severity of the clientelistic commitment problem also depends on the effectiveness of voter monitoring (Stokes Reference Stokes2005: 321). Under the secret ballot, turnout monitoring is naturally easier than observing vote choice, but it still involves checking whether people actually participate in elections. Parts of post-Soviet social geography can be particularly amenable to voter monitoring. Many people vote in the communities where they reside: voting precincts can be composed entirely of former collective farms/villages, factory housing or student dormitories (Allina-Pisano Reference Allina-Pisano2010; Frye et al. Reference Frye, Reuter and Szakonyi2014; Hale Reference Hale2007). This enables community-level monitoring of turnout, but still does not solve the problem of individual-level monitoring. However, in the post-Soviet space, brokers – such as farm managers or local administration officials – reside in the communities themselves and have detailed information about people’s behaviour. Moreover, brokers could on some occasions have a direct role in administering the elections. Many Russian enterprises organize transport to the polling stations for the employees, or polling stations can even be located in agricultural enterprises (Allina-Pisano Reference Allina-Pisano2010; Frye et al. Reference Frye, Reuter and Szakonyi2014). These practices enable very effective turnout monitoring in small communities. The pioneering survey by Frye et al. (Reference Frye, Reuter and Szakonyi2014: 196) found that 25 per cent of employees felt that their employers tried to ‘influence their decision to turn out to vote’ in the 2011 parliamentary elections in Russia, and that the pressure applied was connected to the continuation of their employment or compensation for labour. The following sections detail further how peculiarities in two sectors, post-Soviet agriculture and institutionalized ethnicity, alleviate the clientelistic commitment problems.
Post-Soviet Agriculture and Patronage Linkages
Two structural features of rural areas are thought to make them susceptible to clientelistic pressures. First, rural communities are small and geographically concentrated, which makes monitoring of voting easier (Stokes Reference Stokes2005). Second, agrarian areas tend to be poorer and more dependent on government transfers, increasing their vulnerability to economic sanctions.
Most of the former Soviet collective farms did not become economically viable after privatization. Complex webs of barter transfers between the regional authorities and agricultural enterprises (so-called ‘commodity crediting schemes’) ensured continued production in the collectives despite their unprofitability. Importantly, these schemes enabled continued employment – ‘a safety net for collective farm employees’ – regardless of the economic performance of agricultural enterprises (Amelina Reference Amelina2000, Reference Amelina2002). The employees could receive a significant proportion of their wages in the form of in-kind payments, such as consumer goods or commodities that could be used for food production in private plots (Allina-Pisano Reference Allina-Pisano2008). This non-monetary compensation benefited the workers in an inflation-prone economic environment but also tied them ‘physically’ to their place of employment (where they also voted). The collective farms also continued to provide social services and ad hoc financial help (coordinated by the farm manager) to the members. Farm managers were thus in control of many levers over the employees, such as their employment, compensation for labour and social welfare. On some occasions, farm managers could be responsible for administering elections (Allina-Pisano Reference Allina-Pisano2010), enabling effective turnout monitoring. In turn, the farm managers were directly dependent on higher-level authorities for their own employment and farm subsidies. The ‘community crediting schemes’ were highly politicized, and demands for credit repayments were relaxed when it was politically expedient for the regional governments (Amelina Reference Amelina2000: 488). In sum, the Russian agricultural sector fulfils many of the structural conditions associated with effective patronage politics and voter mobilization: first, the sector is characterized by a complex socioeconomic dependence on the state, leading to strong potential sanctions. Second, the peculiar social geography of the former collective farms facilitate voter monitoring. Accordingly,
Hypothesis 1: Rayons with a higher proportion of agriculturally employed population should have higher turnout rates.
Hierarchical Local Administration and Institutionalized Ethnicity
Ethnicity and ethnic ties have long been associated with clientelism in comparative literature. According to Kanchan Chandra’s (Reference Chandra2007) model, ethnic favouritism can be one way of mitigating the clientelistic commitment problem. Several authors have noted how Soviet institutionalized ethnicity created structural conditions for voter mobilization (Hale Reference Hale2003, Reference Hale2007). The mechanism hypothesized by Henry Hale is based on the ethnocultural legacies of the Soviet Union. ‘Titular republics’, subnational units named after the ‘titular’ ethnicity (such as the Bashkirs in Bashkortostan), had a higher constitutional status in Russia under the Soviet regime than other regions. After the fall of the Soviet Union, Russian federalism continued to be de facto asymmetrical, and the privileging of the titular nationalities intensified in the republics. Hale hypothesizes that ethnic networks, which were already cultivated under the Soviet Union, could be amenable to clientelistic electoral exchanges whereby the preferential treatment of ethnic minorities ‘in education, state employment, territorially concentrated investment and status’ would be ‘exchanged’ for electoral support: ‘Where ethnicity and geography coincided, it was possible to monitor the “ethnic vote” and allocate ethnic rewards and punishments accordingly’ (Hale Reference Hale2007: 231). Accordingly:
Hypothesis 2: Participation rates should be higher in rayons with a higher proportion of non-Russian population.
Local administration structures also diverged between the titular republics and the rest of the regions (oblasts and krays) due to asymmetrical federalism. Tomila Lankina (Reference Lankina2004) has described how local administration structures, which were previously elected, were replaced by hierarchically appointed administrations in the titular republics. Revenue allocation also diverged between the region types: whereas in oblasts and krays the local level had more control over revenue, local bodies in the republics depended more on top-down transfers from the regional authorities (Freinkman and Yossifov Reference Freinkman and Yossifov1999). It could therefore be hypothesized that these vertical, often ethnically dominated, structures of local administration could be amenable to turnout mobilization in elections.
Given the considerable differences in institution building between the titular republics and the rest of the regions, this article tests whether the association between agricultural employment and ethnicity and turnout are moderated by the regional context by including two cross-level interactions in the models. In particular, if local administration networks in titular republics are instrumental in ethnic mobilization, the association between ethnicity and turnout should be more pronounced in the titular republics than in the rest of the regions. Accordingly:
Hypothesis 3: The effect of ‘ruralness’ on turnout is moderated by the regional context.
Hypothesis 4: The effect of ethnicity on turnout is more pronounced in titular republics.
Data and Alternative Explanations
The hypotheses outlined above are tested on an original data set which combines rayon-level Russian presidential electoral results from 2000 and 2004Footnote 5 with rayon-level socioeconomic data from the 2002 census and region-level data from the Goskomstat statistical service.Footnote 6 The rayons are the smallest subnational units for which both electoral and demographic data are available (Clem Reference Clem2006).
The degree of ‘ruralness’ of a rayon is operationalized by the percentage of agrarian employees per rayon (AgrEmp) taken from the 2002 census. The variable NonRussian measures the percentage of the rayon’s population that is not ethnically Russian,Footnote 7 and comes, too, from the 2002 census.
The subnational setting means that many of the institutional or election specific controls which are commonly used in cross-national studies, such as the timing of the elections, the electoral system or the ‘salience’ of the elections, do not vary here (see, for example, Blais Reference Blais2006; Franklin Reference Franklin2004). There are no complicated electoral registration procedures in Russia, and voting is non-compulsory. However, there is considerable economic and social variation between and within the Russian regions which the analysis controls for.
Much of the literature on developed democracies has found that socioeconomic resources, such as age, education, income, gender and occupation, are related to turnout (see, for example, Dalton Reference Dalton2008). Previous individual-level studies have also shown that age and level of education are related to electoral participation in Russia (Colton Reference Colton2000; McAllister and White Reference McAllister and White1998; White and McAllister Reference White and McAllister2007), and thus the models include the percentage of pensioners in the rayon (Pensioners) which proxies above pension age, and the percentage of people with higher education in the rayon (Higher Ed), with data from the 2002 census. The models also control for rayon size (Voters), which is operationalized by the number of registered voters per rayon (logged), as both social-psychological and clientelistic models of electoral participation predict that turnout should be higher in smaller communities, due to closer social ties and the ease of monitoring (Stokes Reference Stokes2005; Remmer Reference Remmer2010). The closeness of the elections – their competitiveness – is often thought to engender greater participation as the ‘probability of casting a decisive vote increases with the closeness of the elections’ (Blais Reference Blais2000: 43). Thus the models also test whether electoral competitiveness at the rayon level (smaller margins) is associated with higher turnout (Margin), in line with the findings from established democracies (Blais Reference Blais2000: 59).
In the early to mid-2000s there were great differences between the Russian regions in both institutional autonomy, institutionalized ethnicity and regional wealth. The 21 titular republics were constitutionally privileged due to Soviet ethno-federal policies and were able to write their own constitutions. Titular republic elites were able to control regional institutional choices in the 1990s to a much greater degree than the elites in the non-titular krays and oblasts (which were constitutionally subjugated to the federal centre). Thus the models include a region-level control for titular republic status (Republic). The models hence also control for gross regional product (GRP) per capita in the year prior to the elections (GRP pc), as well as the share of natural resource extraction in the region’s industrial production (Nat resources), to control for differences in regional wealth and economic structures. The data for both come from Goskomstat.Footnote 8 The models also account for the size of the region by the number of rayons per region (Reg size). Goskomstat does not provide regional data for all the autonomous okrugs/oblast (AOs), and thus the models do not include observations from them. Table 1 presents descriptive statistics of all the variables.
Summary Statistics
Ideally, we would draw on both individual and aggregate data to investigate electoral mobilization in Russia. Unfortunately, apart from the pioneering workplace survey of Frye et al. (Reference Frye, Reuter and Szakonyi2014), almost no individual-level surveys on Russia have included questions on clientelism or voter mobilization. Hence, local-level aggregate data is the best data that are currently available to study variation in electoral patterns across the whole of Russia and also the only data that currently allow us to compare trends across several elections.
Empirical Analysis and Discussion
The analysis is conducted by estimating hierarchical linear (random intercepts) models of turnout clustered in regions.Footnote 9 Multilevel models take into account the nested nature of the data, and they are especially suitable for testing theories that link variables measured at two levels. The results for the 2000 presidential elections are presented in Table 2, and for the 2004 in Table 3. The first model in each table contains the two main explanatory variables and controls, and the second model includes two cross-level interactions. The third model includes a proxy for electoral fraud.
Determinants of Turnout in the Russian 2000 Presidential Elections
Note: Standard errors in parentheses.
*p<0.05, **p<0.01, ***p<0.001
Determinants of Turnout in the Russian 2004 Presidential Elections
Note: Standard errors in parentheses.
*p<0.05, **p<0.01, ***p<0.001
Model 1 offers strong support for Hypotheses 1 and 2. Even when controlling for rayon size, the socioeconomic characteristics and electoral competitiveness, the share of agricultural workers in a rayon has a positive, substantive and statistically highly significant association with turnout, which supports the hypothesis that electoral mobilization and political participation are higher in the more agricultural rayons. A higher proportion of non-Russians in a rayon’s population is also associated with higher rates of political participation, supporting the second hypothesis that rayons with higher concentrations of non-Russian ethnicities would be more susceptible to turnout mobilization. The substantive effects are also large. By translating the effects into marginal effects (with all other variables set at their mean), it can be estimated that once the share of agrarian population increases from the median to the maximum value, electoral participation increases by over 15.5 percentage points. A similar increase in the non-Russian share increases turnout by 5.5 percentage points.
Most control variables are in the expected direction, and all the rayon-level controls are highly statistically significant. The negative association between rayon size and turnout – that is, that turnout is higher in smaller rayons – is in line with the expectations from both the clientelism literature and also the social-psychological models of turnout, while the positive coefficient for margin is more consistent with the mobilization hypothesis than the rational choice model of political participation. The finding that the presence of a larger share of the population with higher levels of education is associated with increased turnout is consistent with findings from industrial democracies and previous individual-level studies from Russia. The association of higher turnout with a larger proportion of pension-age population is also in line with previous individual-level work and the findings of Hale (Reference Hale2007). Turning to region-level controls, most of these are in the expected direction, although only two (wealth and size) are significant. Regional wealth has a negative association with turnout, which could suggest that pressures to mobilize are greater in regions without significant economic resources. Interestingly, turnout is higher in larger regions, in line with the findings of Hale (Reference Hale2003, 2007).Footnote 10
In Model 2 two cross-level interactions are added to test whether the demographic effects on turnout are moderated by the region type (titular republic or oblast/kray/federal city). The negative and significant first interaction suggests that the impact of ‘ruralness’ on rayon turnout is smaller in republics than in other regions. Turnout mobilization in agrarian areas thus seems to be greater in non-republics (that is, in mostly ethnically Russian oblasts and krays). This could suggest that the hierarchical administrative structures in titular republics were able to mobilize turnout in urban and rural areas, but in other regions mobilization is concentrated in the agricultural areas.Footnote 11 In contrast, the coefficient for the second interaction term is positive and significant, suggesting that the impact of the share of non-Russian population on rayon turnout is greater in titular republics than in the other regions. This finding supports the expectation that ethnic networks would be particularly amenable to turnout mobilization in titular republics.
As can be seen from Table 3, the results for 2004 are broadly similar to those for 2000, although there are some interesting changes. Looking first at Model 4, we can see that the association between the share of agrarian employees and turnout, and the association between the share of non-Russians and turnout are even stronger in 2004. Based on Model 4, once the share of agrarian population increases from the median to the maximum value, electoral participation increases by 19 percentage points. A similar increase in the non-Russian share increases turnout by 5.7 percentage points. This supports the expectation that turnout mobilization would intensify as the elections become less competitive. The coefficients for regional-level variables, such as republic status,Footnote 12 regional wealth and natural resource extraction grow markedly, and all are significant in this model. However, the coefficient for region size is no longer significant. The most interesting difference is seen in the cross-level interactions in Model 5. The impact of the share of agrarian employees on rayon turnout is again less pronounced in republics, suggesting that in oblasts and krays the mobilization concentrated mostly on agrarian areas. In contrast, the interaction term between the percentage of non-Russian population and republic status is not significant. While ethnicity on the whole is associated with higher rayon-level turnout, there is no longer a significant difference in the impact of ethnicity on turnout between the titular republics and the rest of the regions in 2004.Footnote 13
An extensive literature has now demonstrated the existence of electoral fraud in Russian elections (see, for example, Mebane and Kalinin Reference Mebane and Kalinin2009; Myagkov et al. Reference Myagkov, Ordeshook and Shakin2009). Although ‘isolating’ the effect of electoral fraud from other types of ‘electoral malfeasance’ is empirically challenging, some proxies can be used to control for the most ‘incredible’ electoral results. Models 3 and 6 include a proxy for electoral fraud. Rayons where both electoral turnout and votes for the winning candidate exceeded 90 per cent (a highly improbable result, given the tendency of turnout to decline as electoral competition declines)Footnote 14 were coded into a dummy variable proxying electoral fraud (Over90). While the coefficient is positive and statistically significant in both models, the size of the other coefficients is not greatly reduced, which suggests that the results hold even when the most ‘improbable’ outliers are controlled for. However, further precinct-level data analysis is needed to develop measures that would allow us to fully ‘isolate’ the effect of electoral fraud.
To check the robustness of the agricultural employment variable, Models 1 and 4 were rerun, replacing it with a dummy variable (Rural) denoting whether the rayon’s population was predominantly rural (‘1’) as opposed to urban/mixed (‘0’) (coded from an alternative data source, Goskomstat Regiony Rossii statistics).Footnote 15 As the Models 7 and 8 on Table 4 show, rural rayons have a positive and highly statistically significant association with turnout in both 2000 and 2004. The other coefficients remain substantively similar. Running the models with region fixed effects (not shown here) also results in very similar results.
Models 7 and 8
Note: Standard errors in parentheses
*p<0.05, **p<0.01, ***p<0.001
In summary, the results show clear evidence of the relationship between demographic characteristics of the rayons that are associated with clientelistic mobilization and higher rayon-level turnout. These associations grow substantially stronger in 2004, suggesting that the decline in electoral competitiveness (and the subsequent decline in interest in the elections) created pressure to mobilize certain ‘pliable’ sectors of population to guarantee a high overall turnout. It should also be noted that the ‘behaviour’ of other variables, such as the community size or the margin of victory, were more in line with expectations from the clientelism literature than patterns found in developed democracies. The regional context clearly moderated the relationship between the demographic variables such as agrarian employees and turnout. The association between agrarian employment and turnout was more pronounced in the predominantly ethnically Russian regions than in the titular republics, and this effect stayed constant in both electoral cycles. In contrast, the impact of ethnicity on turnout was greater in titular republics than in the rest of the regions in 2000, supporting the expectations that ethnic networks would be particularly amenable to electoral mobilization in those areas. However, despite a strong overall association between ethnic composition and rayon turnout in 2004, the impact of ethnicity was no longer moderated by the region type in these elections. This could suggest that, by 2004, the easily identifiable ethnic constituencies would have been targeted outside the titular republics, too.
Some alternative explanations cannot be fully ruled out without individual-level data (but are unlikely given the context). First, ethnic voting – that is, that ethnic minorities would (voluntarily) participate at higher rates than the rest of the population in order to ensure greater political representation – could be an alternative explanation for the observed association between higher shares of non-Russian population and rayon-level turnout. Yet, minority ethnic mobilization usually takes place in the context of ethnic party systems or when there are co-ethnic candidates on the ballot (see, for example, Barreto Reference Barreto2007; Horowitz Reference Horowitz1985). There were no ethnic party candidates in the 2000 or 2004 elections, nor did any of the candidates campaign on a specifically ethnic platform. Therefore it is unlikely that candidate positioning would account for the greater levels of minority ethnicity participation. Second, historically, turnout rates were higher in the rural areas in some European democracies, and we could observe a similar phenomenon also in Russia. However, the plentiful qualitative evidence of clientelistic, contingent electoral mobilization in Russian rural areas (see, for example, Allina-Pisano Reference Allina-Pisano2010) suggests that the high turnout rates in Russia were not simply a result of different ‘social norms’ or (non-coercive) ‘social pressure’. It is clear, though, that we need more individual-level data and micro-level studies to probe the findings of this article further.
Conclusions
This article has contributed to the literature on electoral mobilization in electoral autocracies and has shown how electoral mobilization of certain institutionally and socioeconomically state-dependent sectors was a key part in Russia’s transition from a competitive to hegemonic authoritarian regime between 2000 and 2004. Building on previous work on patronage politics in Russia, the article outlined how vertical administrative and resource dependency from the regional authorities created pressures for turnout mobilization in specific demographic localities: agricultural rayons and those inhabited by ethnic minorities in the titular republics. The article also detailed how socioeconomic dependency on top-down barter arrangements in the agricultural sector and fiscal dependency in the titular republics created opportunities to mobilize different sectors of the population in elections. Both of these mechanisms were influenced by complex legacies of Soviet economic and social control. Thus to fully account for electoral authoritarian regime trajectories we need to examine the ‘socioeconomic roots’ of authoritarianism as well as electoral fraud and political repression usually associated with these regimes. Also, as blatant fraud has become increasingly possible to detect, it could be expected that more emphasis will shift to electoral clientelism in electoral autocracies in the future.
The original rayon-level data set used in this article enabled the testing of the hypotheses related to electoral mobilization with multilevel models and local-level data while also considering several alternative explanations from the comparative turnout literature. The results support the expectations that electoral mobilization targeted certain demographic groups, such as the rural and minority ethnic communities. The article also uncovered interesting variation in the intensity of these effects between the titular republics and the mainly ethnically Russian regions, which may be related to fiscal regime differences between the region types. In addition, it should be noted that the ‘behaviour’ of other variables, such as community size, was in line with the expectations of the clientelism literature. The effects were more pronounced in 2004, supporting the expectation that the declining competitiveness of the elections would increase pressures for turnout mobilization.
Despite the fact that economic dependency was one of the mechanisms of political control in the Soviet Union there is still rather little literature that would explore how economic dependency on the state undermines democratization in the post-Soviet space.Footnote 16 This article has shown that these mechanisms need to be considered when studying the dynamics of democratic regressions. Given that patronage mobilization opportunities are related to the Soviet structural legacies, this framework should be generalizable to many other former Soviet countries.
Acknowledgements
Research for this article was funded by the Academy of Finland (project number 258190). Previous versions were presented at the Annual Meeting of American Political Science Association, Chicago, 2013, at the Åbo Akademi political science research seminar and the Finnish Political Studies Association annual meeting in Tampere, 2014. I wish to thank the anonymous reviewers, as well as Kenneth Greene, Gulnaz Sharafutdinova, Tomila Lankina, Mikko Mattila, Kimmo Grönlund, Henrik Serup Christensen, Catharina Groop, Alexey Bessudnov, Alisa Voznaya and participants at the presentations for valuable comments, and Gordey Yastrebov for great help during the data collection. All errors naturally remain my own.
Supplementary Material
To view the supplementary material for this article, please go to: http://dx.doi.org/10.1017/gov.2015.20
