ilable at ScienceDirect

Electoral Studies 36 (2014) 15e27

Contents lists ava

Electoral Studies

journal homepage: www.elsevier .com/locate/electstud

Ethnicity and electoral manipulation in Russia

Regina Goodnow a, Robert G. Moser b, *, Tony Smith b

a Department of Political Science and Public Administration, University of Wisconsin-La Crosse, La Crosse, WI 54601, USAb Department of Government, University of Texas at Ausin, Austin, TX 78712, USA

a r t i c l e i n f o

Article history:Received 30 January 2013Received in revised form 30 March 2014Accepted 21 May 2014Available online 16 June 2014

Keywords:Electoral manipulationEthnicityRussia

* Corresponding author.E-mail addresses: rgoodnow@uwlax.edu (R.

austin.utexas.edu (R.G. Moser), kievkowboy@austin

http://dx.doi.org/10.1016/j.electstud.2014.05.0050261-3794/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

To what extent does electoral manipulation follow ethnic lines in Russia? Using an originaldataset based on raion-level data, we find that the “ethnic component” of electoralmanipulation is more nuanced than previous studies have suggested. Electoral manipu-lation was most prevalent in majority-minority raions across ethnic and non-ethnic as wellas richer and poorer regions. We argue that concentrations of ethnic minorities provide:(1) greater incentives for electoral manipulation by the central state and regional elites inorder to signal political dominance and (2) greater capacity to carry out electoralmanipulation through networks of local co-ethnic elites. However, multilevel analysessuggest that the extent of electoral manipulation was also strongly contingent on regionalcontext. Electoral manipulation was significantly higher in the more politically volatileMuslim regions, while socioeconomic differences among regions, by contrast, had nodiscernible effect.

© 2014 Elsevier Ltd. All rights reserved.

1. Ethnicity and electoral manipulation: the case ofRussia

Scholars have long noted the dilemmas that an ethni-cally divided society faces when trying to democratize.Ethnic divisions can promote armed conflict, mass violence,discriminatory state policies, and regime illegitimacy.Moreover, recent scholarship has suggested that democ-ratization may exacerbate these problems by providingelites with incentives and opportunities to use nationalismfor electoral gain (Mansfield and Snyder, 2005; Snyder,2000; Wilkinson, 2004). Those hoping to manage theseproblems often look to institutions to either co-opt mi-norities by offering security through greater autonomy andrepresentation (Lijphart, 2004; Saideman et al., 2002) orbridge ethnic differences through institutions designed tofoster multi-ethnic electoral coalitions (Horowitz, 2003;Reilly, 2002). Throughout this literature, there is a

Goodnow), rmoser@.rr.com (T. Smith).

presumption that the main threat from ethnic diversity isconflict and competition between ethnic groups.

In this paper, we offer a different perspective on theinteraction between ethnicity and failed democratization.We argue that, under certain conditions, geographicallyconcentrated ethnic minorities may contribute to authori-tarianism through collusion with (rather than competitionagainst) central authorities, especially in the practice ofmanipulating elections. Ethnic minorities may become animportant “constituency” for a competitive authoritarianregime due to the political economy of elections dominatedby patronage. Under these types of elections, central au-thorities and regional elites enter into a patronage rela-tionship in which the former provides economic andpolitical benefits to targeted regions in exchange for votesdelivered by regional authorities during national elections.Regions with large concentrations of ethnic minorities areparticularly susceptible to these political arrangementsbecause central authorities may view such regions as aparticularly formidable potential threat to their authority.Central authorities may thus have additional incentives tomanipulate elections in these regions in order to signal

R. Goodnow et al. / Electoral Studies 36 (2014) 15e2716

political dominance and deter opposition (see Simpser,2013). Moreover, regions with high concentrations ofethnic minorities may provide more conducive environ-ments for electoral manipulation because they tend to beisolated regions with low levels of socioeconomic devel-opment, weak civil societies, and long histories of patron-eclient relations. Finally, local elites in minority regionshave resources, in the form of ethnic networks, that maymake it easier to engage in patronage and other forms ofelectoral manipulation (see Hale, 2003). Thus, our argu-ment is two-fold. In Russia, there tends to bemore electoralmanipulation in regions with concentrations of ethnic mi-norities because: (1) the state feels particularly vulnerablein minority regions and thus has more political reasons toengage in electoral manipulation and (2) it is easier toconduct electoral manipulation in minority regions due tothe existence of minority-based machine politics.

Russia offers an excellent example of this phenomenon.Despite centralizing policies that undermined regionalautonomy and a devastating war in Chechnya that hasspawned violence across the Caucasus, non-Russian re-gions have provided some of the strongest support forVladimir Putin and his “party of power” (Marsh andWarhola, 2001). Scholars have noted that evidence ofelectoral malfeasance, such as extremely high levels ofvoter turnout and support for the party of power, has beenmore endemic in Russia's ethnic republics than in itsRussian-dominated oblasts (see, in particular, Myagkovet al., 2009). However, analyses of electoral fraud inRussia have tended to emphasize the detection of fraud andits magnitude on a national scale. In this paper, we examinegeographic patterns to try to uncover the determinants ofelectoral manipulation. While various forms of electoralmanipulation have reportedly become widespread inRussia, they are not uniformly experienced across Russia'svast territory. By carefully examining where electoralmanipulation has and has not tended to occur withinRussia's 2700þ raions (which are roughly equivalent to U.S.counties), we hope to gain some leverage on how socio-economic and demographic factors affect electoral manip-ulation. In particular, we use raion-level electoral andethnicity data to explore the potential correlation betweenthe demographic concentration of ethnic minorities inRussia and irregular electoral patterns.

The linkage between ethnicity and electoral manipula-tion in Russia is important in practical terms for thedevelopment of the postcommunist Russian state, but italso sheds light on broader issues of democratic consoli-dation in ethnically divided societies. Many scholars focuson the interrelationship between ethnic conflict anddemocratization (Snyder, 2000; Wilkinson, 2004) and athriving body of literature has arisen around how to designdemocratic institutions to mitigate the prospect of inter-ethnic conflict. Two competing perspectives have domi-nated this debate. Scholars led by Arend Lijphart haverecommended proportional representation, decentraliza-tion, and the regional autonomy of ethnic minorities as amethod of cooptation within a broader system of con-sociationalism (see Lijphart, 1977, 2004). A rival school ofthought known as centripetalism has countered thatconsociational systems lock in ethnic divisions and that

electoral systems in divided societies need to provide in-centives for voters to pool their votes behind moderate,multi-ethnic coalitions (Horowitz, 1985, 2003; Reilly,2002). Regardless of whether the institutional remedycalls for greater inclusion or incentives for multi-ethniccoalition-building, both consociational and centripetal ap-proaches view the problem of ethnic division in terms ofethnic groupsmobilizing against one another or the state inexclusive organizations. By examining the relationshipbetween ethnic minorities and electoral manipulationbenefiting the dominant party in a competitive authori-tarian system, the Russian experience with ethnic electoralmobilization provides an alternative dilemma e the pros-pect of patronage systems and ethnic electoral machinesthat work in tandem with a dominant party to underminepolitical competition.

Russia is a particularly interesting and valuable case forthe study of ethnic politics because it contains manydifferent ethnic groups that vary considerably along de-mographic, socioeconomic, and cultural lines. Moreover,Russia's system of ethnic federalism politicized and insti-tutionalized ethnic identity by promoting geographic con-centration and reduced assimilation of the groups withethnic homelands, which served to enhance their politicalcapacity (Gorenburg, 2003). Finally, the country's system ofethnic federalism promoted a degree of regional autonomyand minority control over regional politics, particularlyduring the 1990s, when President Boris Yeltsin ruled over asystem of asymmetrical federalism that provided ratherwidespread autonomy to regional elites in non-Russianregions (Stoner-Weiss, 2001). Subsequent recentralizationof state authority under President Vladimir Putin hasgreatly curtailed the power of regional elites but alsoarguably increased incentives for electoral manipulationalong ethnic lines to signal political control in previouslyrestive regions. All of these factors provide an interestingcontext in which to examine the interplay betweenethnicity and elections in an atmosphere of competitiveauthoritarianism.

We will ascertain the differences (or lack thereof) in theincidence of electoral manipulation between Russian andnon-Russian regions using electoral and census data dis-aggregated to the raion level, which is the smallestgeographic unit at which data are available. Empirically, weexamine a relatively basic set of questions: Are raions withhigh concentrations of non-Russian minorities more likelyto show signs of electoral manipulation than raions whereRussians predominate? If so, are these patterns mitigatedby contextual factors such as rural/urban divisions orregional characteristics such as levels of economic devel-opment, resource-dependent economies, or ethnic differ-ences between Muslim and non-Muslim minority regions?

2. Electoral manipulation in Russia

It is important to demarcate what types of phenomenawe are trying to capture when studying electoral manipu-lation in a competitive authoritarian regime such as Russia.Electoral manipulation is a multifaceted enterprise thatcomes in a wide variety of forms ranging from illicit acts ofelectoral fraud such as ballot-box stuffing, misreporting

R. Goodnow et al. / Electoral Studies 36 (2014) 15e27 17

electoral returns, vote-buying, and the illegal denial or re-striction of access to polling stations to acts that are notnecessarily illegal but nonetheless use unacceptable meansto change electoral outcomes in favor of a particularcontestant such as various forms of pressure on voters andvarious forms of patronage (see Lehoucq, 2003; Simpser,2013).

In this paper, we consciously employ the broader termelectoral manipulation to capture the wide range of activ-ities that render an electoral context manipulated (asopposed to clean) (Simpser 2013: 32); rather than thenarrower term electoral fraud, which is often defined asactions that attempt to affect elections and break the law(Lehoucq, 2003: 235). Academics, journalists, and activistshave provided extensive indirect and direct evidence thatboth types of activities e electoral manipulation such ascoercive pressure and threats toward voters as well asillegal acts of electoral fraud such as ballot-stuffing, mul-tiple voting, and officials miscounting votes and changingelectoral results e have occurred in Russia (Myagkov et al.,2009; Fish, 2005;White, 2011).1 Our analyses using indirectmeasures of electoral manipulation cannot differentiatebetween illegal acts of electoral fraud (e.g., ballot-stuffingor tampering with results) and acts of electoral manipula-tion (e.g., inappropriate pressure, threats, or bribes to vote acertain way). Instead, our goal is to analyze geographicvariation across raion-level electoral contexts that haveresults indicative of some form of electoral manipulationand use various socioeconomic and demographic correlatesto ascertain where relatively manipulated electoral con-tests tend to occur in Russia.

2.1. High voter turnout as an indicator of electoralmanipulation in Russia

While the myriad methods of electoral manipulationcomplicate its conceptualization, its clandestine nature alsopresents scholars with severe obstacles to empiricalobservation and measurement. As Lehoucq (2003: 235)notes, instances of electoral manipulation are acts “thatonly its victims want publicized.” Therefore, sources thatallow scholars a window into these behaviors are severelylimited and often biased in one way or another. Ethno-graphic observation, opinion surveys, and legal challengesof electoral results and practices offer direct measures ofthe behavior in question but suffer from problems ofpartisan bias, misreporting by respondents and observers, alack of comprehensive coverage of the variety of ways to fixan election, as well as logistical problems of observers notbeing everywhere at all times (Lehoucq, 2003; Carothers,1997).2

Direct measures of electoral manipulation are notavailable in any systematic fashion for Russian elections.

1 Western media outlets have provided extensive coverage of instancesof Russian electoral fraud and manipulation. Within Russia, Golos, anNGO devoted to exposing electoral irregularities, has provided extensiveand systematic evidence of electoral manipulation, including video evi-dence of ballot-stuffing. Author interview in Moscow (June 2014).

2 For a defense of using legal challenges to electoral results as a proxyfor electoral fraud see Ziblatt (2009).

Therefore, we follow the example of scholars (e.g., Powell,1989; King, 2001; Myagkov et al., 2009; Mebane andKalinin, 2009; Lukinova et al., 2011) who have adopted anindirect approach to gauge electoral misconduct basedlargely on quantitative indicators, usually unrealisticallyhigh voter turnout levels and anomalous distributions ofvotes across parties. There are several advantages to thisapproach. First, the evidence of fraud is mined from officialelectoral results and is thus available across a wide range ofcountries as well as within countries over time, providingopportunities for comparative analyses. Second, thisapproach provides a systematic and uniform application ofcriteria presumed to indicate manipulation for all electoraladministrative units (in our study raions) rather than amethod dependent on either uneven coverage by monitorsor the opinions of respondents to a survey whose answersmay depend on interpretations of the question or willing-ness to respond candidly. Third, it provides a way toexamine variation across time and space in the occurrenceof electoral manipulation by differentiating between re-gions where fraud seemed more prevalent and regions thathad cleaner elections. This allows for the investigation ofcausal relationships linking environmental factors foundwithin electoral units and the occurrence of irregularelection results. Of course, such indirect proxy measuresare not perfect and suffer from some of the same limita-tions as other measures of electoral manipulation, such asapproximations that under- or over-report the magnitudeof the behavior. However, in this study we are not trying tomeasure how many votes were “stolen” in a given electionbut rather are striving to capture general patterns of whereelectoral manipulation seemed more or less prevalentacross Russia and then test whether those regions deemedto experience more manipulated elections shared crucialcommon characteristics. With this in mind, we argue thatas long as our proxy measure provides a sufficientlyconvincing approximation of manipulation, we can use it toexplore patterns of electoral manipulation, even if it doesnot fully capture the extent and complexity of thephenomenon.

We use very high voter turnout as our proxy for elec-toral manipulation. There are two reasons to suspect thatvery high turnout is a good indicator of electoral manipu-lation. First, as Myagkov et al. (2009) note, in the absence ofcompulsory voting, it is quite plausible that regions thatexperience extremely high percentages of voters going tothe polls are experiencing some form of electoral manip-ulation e voter pressure, ballot-stuffing, vote-buying,tampering with results, patronage-based mobilization,and/or some combination of these practices e especially ifmany other regions within the same country experiencemuch lower voter turnout. Moreover, the distribution ofvotes across parties also strongly suggests that electoralmanipulation played a role in high voter turnout in Russiabecause a single party, United Russia, benefitted dis-proportionally from large increases in turnout (seeMyagkov et al., 2009 on this point as well).

As the empirical data plots in Figs. 1a and b, and 2a andb demonstrate, higher voter turnout only benefitted theparty of power e United Russia received an increasingshare of the vote as reported turnout increased, in contrast

Fig. 1. (a and b) Turnout and United Russia's vote share for 2007 and 2011.

R. Goodnow et al. / Electoral Studies 36 (2014) 15e2718

to other parties, whose vote shares were relatively stableat lower turnout levels but actually decreased with higherturnout levels. As Myagkov et al. (2009) and Simpser(2013: 173) have pointed out, there is no logical reasonwhy vote share for individual parties or candidates shouldvary in a systematic way with voter turnout. More votersat the polls should mean more supporters for all parties incompetition at roughly the same vote share they garneredin other, lower-turnout contexts. But in Russia, as Fig. 1aand b starkly display, United Russia saw a dramatic andconsistent increase in vote share with increases in voterturnout. Given these voting patterns, we operationalizeelectoral manipulation in a dichotomous fashion, classi-fying raions as fraudulent when reported voter turnoutwithin the raion was at least as high as 75%. The 75%turnout threshold is a rather conservative estimate ofpotentially manipulated compared to non-manipulatedraions for a context without mandatory voting laws. Italso covers a greater number of potential instances ofmanipulation across a greater diversity of regional con-texts than would a higher turnout threshold. Moreover, at-test confirmed the statistical significance of the different(and larger) mean in United Russia's vote share for thegroup of raions with more than 75% turnout compared tothe group of raions with less than 75% turnout.3 Finally,the 75% threshold offers a significant distinction in thenature of voter turnout within Russian elections. For the2007 election, 26% of the raions in our dataset had turnoutlevels above the 75% threshold; the average turnout forraions above that threshold was 86% while the averageturnout for raions under the threshold was 61% e a 25%difference. For the 2011 election, 20% of the raions in ourdataset had turnout levels above the 75% threshold; theaverage turnout for raions above that threshold was 87%while the average turnout for raions under the thresholdwas 57% e a 30% difference.

Of course, there is an alternative explanation for higherthan normal voter turnout in Russia. Rather than being aproduct of electoral manipulation, it could be the case thatraions with a combination of very high voter turnout andvery high vote share for United Russia are simply areascharacterized by a high concentration of voters who areenergized supporters of the regime and President Vladimir

3 Using other turnout thresholds (e.g., 80%, 85%, and 90%) producedcomparable results in our quantitative analyses of the determinants ofelectoral manipulation. More specifically, the significant regional effectsfrom using the 75% threshold (presented in Tables 2 and 3 in theempirical section) were also significant at higher thresholds. Similarly,the variables with no significant effects at the 75% threshold, includingthe intercept effects for regional levels of economic development andnatural resources wealth, were also insignificant when using higherthresholds. (The one exception was a significant negative effect for thelevel of economic development when using a 90% threshold). We alsonoted one deviation from the otherwise consistently significant keyraion-level “percent minority” variable; it was no longer significant at the90% threshold when we accounted for whether or not raions were inethnic federal regions. This could be explained by the much smallernumber of raions meeting that very high threshold and their concen-tration within non-Russian regions. On the whole, even with these dif-ferences, the conclusions that we draw from using the 75% threshold areconsistent with results from even more conservative indicators of sus-pected manipulation.

Putin. Indeed, one of the persistent characteristics ofRussian politics under the leadership of Vladimir Putin hasbeen his consistently high approval ratings hovering at 70-80% throughout the period, according to opinion polls(Treisman, 2012). This has led even scholars who haveasserted that the Russian state has perpetuated large-scaleelectoral manipulation also to concede that Putin and hisparty would have won national elections during the 2000squite easily and decisively in completely clean electoralcontests (Fish, 2005; Myagkov et al., 2009; Mebane andKalinin, 2009; Treisman, 2012).

Why not attribute extremely high voter turnout incertain raions in Russia to positive signs of genuine ideo-logical or personal support for Putin and his party andoverall regime support rather than our pessimistic asso-ciation of higher turnout with electoral malfeasance?Several patterns militate against the more positive view ofextremely high turnout in Russia and point toward ourposition that high turnout is a reasonable and evencompelling proxy for electoral manipulation. First, averagevoter turnout nationwide was not overly high (64% in2007 and 60% in 2011), suggesting that high approvalratings for Putin and his party did not uniformly translateinto extremely high voter turnout across the country. Infact, the raions we identify as having high turnout are arelatively small minority (about 20%) of the total raions inthe country. Second, patterns of very high voter turnout inRussia did not correspond to socioeconomic trends thattypically drive greater voter participation e for example,excessively high turnout was not positively correlatedwith more prosperous raions. Finally, raions with

Fig. 2. (a and b) Turnout and vote share for select parties in 2007 and 2011.

R. Goodnow et al. / Electoral Studies 36 (2014) 15e27 19

extremely high turnout were located in precisely thoseregions that one would expect to experience some of thegreatest degree of discontent and disillusionment with theRussian state. It is difficult to imagine that results showinggreater than 95% turnout and 95% support for UnitedRussia in places like Chechnya or Dagestan, where violentattacks against the state are a regular occurrence, aredriven by anything other than massive electoral manipu-lation. As will be explained below, the more persuasiveexplanation for the geographic distribution of extremelyhigh turnout is that these are areas where elections arehighly controlled and manipulated. Simpser asserts thatthe very fact that Putin did not need to fix the elections inorder to win them is evidence that the flagrant manipu-lation of elections in Russia is used to signal dominancerather than win elections (Simpser, 2013: 176e183). Weagree. Our analysis showing the correlation betweenconcentrations of ethnic minorities and electoral manip-ulation at the raion level in Russia suggests that ethnicityplays a crucial role in the targeting of electoral manipu-lation for political purposes. Higher concentrations ofethnic minorities in raions increase the incentives for thestate to signal dominance due to perceived vulnerabilityand increase the capacity to perpetuate electoral manip-ulation due to tendencies toward ethnic-based politicalmachines (Hale, 2003).

3. Determinants of electoral manipulation

Lehoucq's (2003) review of the scholarly literature onelectoral fraud notes that there is very little systematicanalysis of the determinants of electoral manipulation andeven less general consensus regarding common causes thatcan explain variation in these practices. For example, somescholars have found that complaints of legal infractions ofelectoral laws are more common in urban areas than ruralareas (Lehoucq, 2003). However, Lehoucq (2003: 251)rightly notes that this may be a function of the propensityof urban populations to mobilize for their political rightsand elites to engage in more detectable forms of electoralmanipulation in urban settings than in rural ones whereunited networks of political machines and economic elitesmay be able to use more subtle modes of patronage andelectoral “influence” to produce the results they desire.

Most commonly, scholars assume that elites manipulateelections in order to win and thus expect manipulation tooccur in situations where it will have the greatest impacton election results (Lehoucq, 2003: 249e251; Ziblatt, 2009:2). Therefore, manipulation is expected to be most preva-lent in regions that have greater electoral competition,assuming that elites' efforts to skew results in their favor incompetitive electoral environments will have the greatestimpact on the distribution of seats and thus political powerto parties. Electoral manipulation in regions where theincumbent party already enjoys widespread support pre-sumably would be unnecessary and, conversely, efforts tomanipulate results in situations where victory is a long-shot would be risky (because it would take more egre-gious forms, making it more easily detected). Therefore,fixed elections are expected to occur more often in morecompetitive districts and under electoral rules such asmajoritarian or plurality systems where they can have thegreatest impact on who wins and loses elections (Ziblatt,2009: 3).

Simpser (2013) challenges the notion that electoralmanipulation is primarily an “election-winning” phenom-enon. Rather, he argues that often elites manipulate elec-tions in contexts and under circumstances that do notsubstantially contribute to simply winning the election athand. Electoral manipulation is often carried out in cir-cumstances in which the incumbent authorities wouldhave won the election without resorting to cheating.Moreover, many acts of manipulation are brazen and notconcealed from public view. Simpser argues that suchbehavior serves important functions by signaling politicaldominance of the incumbent forces, making apparent thepower disparity between those in power and those in thenascent opposition, and, thus, deterring future politicalmobilization against the state (Simpser, 2013: pp.276e283). As noted above, scholars observing Russianelections have also claimed that Vladimir Putin and hisparty would have won elections in the 2000s withoutresorting to electoral manipulation (Fish, 2005; Myagkovet al., 2009; Mebane and Kalinin, 2009; White, 2011;Treisman, 2012), raising the question of why and underwhat conditions such manipulation has taken place.

Ziblatt (2009) notes that economic inequality was a keyfeature of the regions that experienced the highest number

R. Goodnow et al. / Electoral Studies 36 (2014) 15e2720

of legal challenges against electoral malfeasance in 19th-century Germany. He argues, in part, that powerful eco-nomic interests had great resources at their disposal toperpetuate electoral fraud, thus introducing the role of thecapacity of powerful interests to translate political andeconomic advantages into electoral levers of control andmanipulation (Ziblatt, 2009: 3).

In a related body of literature, scholars on Americanpolitics have long noted the connections between ethnicminority groups, urban political machines, and electoralmanipulation. Patronage-based political machines inAmerican cities were often, although not exclusively, basedon clientelistic ties with ethnic groups in which the partyprovided patronage in the form of jobs, social services, andopportunities for social mobility in exchange for loyalty tothe party at the polls (Dahl, 1961; Cornwell, 1964;Wolfinger, 1972; Inglot and Pelissero, 1993). The key as-pects of this relationship included the party's access to stateresources that could be distributed to loyal groups, thesocioeconomic vulnerability of target groups making themlikely targets of machine politics, and the influence ofethnic voting that made it easier to mobilize as a votingbloc (Cornwell, 1964).

3.1. Why might regions with minority concentrations inRussia experience more manipulation?

As exemplified by Kanchan Chandra's (2004) study ofethnic parties in India and the scholarship on Americanurban political machines (see, especially, Cornwell, 1964),patronage and ethnicity often appear to go hand in hand.Chandra offers a compelling model for explaining this linkbetween ethnicity and patronage based on notions of apatronage-democracy, marked by a large public sector inwhich the state has control over jobs and services, and anelectoral environment of limited information. Under suchconditions, which arguably also characterizes the post-communist Russian state beginning under Yeltsin, votersand elites develop reinforcing expectations of ethnicfavoritism. In a self-reinforcing cycle, voters tend to supportco-ethnics because they expect politicians to provide ma-terial benefits to members of their own group and politi-cians feel compelled to fulfill this expectation of ethnicfavoritism in order to attain and retain office (Chandra,2004: Chs. 3e4).

Building from the work on ethnic-based patronage andthe determinants of electoral manipulation, one mightexpect that geographically concentrated ethnic minoritiesin Russia may provide a favorable environment forpatronage and electoral machinations undermininggenuine electoral competition for reasons related to bothincentives on the part of national and regional elites forcontrol over electoral processes as well as opportunitiesand resources that make such control more possible. First,as suggested by Simpser (2013), national elites in Russiamay have increased incentives to engage in widespreadfixing of election results in ethnic regions as opposed toregions dominated by the majority ethnic group due to aperceived need for greater control in ethnic regions. Afterall, ethnic mobilization within the non-Russian union re-publics was a principle cause of the collapse of the Soviet

state into 15 successor states (Beissinger, 2002) as well asdrives for autonomy by ethnic republics in the 1990s(Treisman, 1997; Gorenburg, 2003).

Second, as Hale (2003, 2007) argues, the institutionallegacy, political economy, and potential for ethnic mobili-zation found within ethnic republics in Russia have pro-vided these regions with greater resources for patronage-based machine politics than other areas of the countrywith much smaller minority populations. While regionalgovernors across Russia had significant access to resourcesthat could be exchanged for electoral support, minorityelites in ethnic republics had additional incentives andmechanisms to “buy” electoral support. Governors inethnic regions could provide specific benefits in education,employment, contracts, and status to the titular minoritygroup (e.g., Tatars in Tatarstan, Bashkirs in Bashkortostan)in exchange for electoral support from the target group.Moreover, this exchange could be monitored and enforcedmore easily since geographic concentrations of differentethnic groups allowed clear indicators of the “ethnic vote”(Hale, 2007: 231). More importantly, the bonds of ethnicityproduced norms and expectations on the part of minorityelites and minority voters within ethnic regions that madestrict enforcement less necessary. Since ethnicity waspoliticized and institutionalized in ethnic-based regions,there was an understanding on both sides of the exchangethat the minority elites in ethnic regions would provideadvantages for their co-ethnic voters and, in exchange,could count on strong electoral support from their clients(Hale, 2007: 231). Given this environment, national eliteshad greater incentives to co-opt the minority elites inethnic regions within their own clientelistic exchange e

economic and political benefits in exchange for electoralsupport in national elections e not only to send a signal ofcontrol alluded to above but also because the minorityelites could deliver the votes.

In short, beginning well before the emergence ofwidespread electoral manipulation under Vladimir Putin inthe 2000s, the political context within ethnic republicsprovided both the incentives (in the form of conveyingpolitical dominance in regions seen as potentially volatileand threatening to state power) and resources (in the formof elite networks, political machines, and co-ethnic codes ofloyalty and deference) necessary for the coordination ofhigher levels of electoral manipulation.

4. Ethnic federalism and ethnic electoral politics inRussia

Russia has a significant minority population e close to20% e similar in size to many countries in the region suchas Bulgaria, Romania, and Slovakia. However, unlike thesestates, Russia lacks a single major minority group. Instead,the non-Russian population is composed of literally dozensof very small groups. The largest minority group of theRussian Federation, Tatars, makes up less than 4% of thepopulation and only three groups (Tatars, Ukrainians, andChuvash) comprise over 1% of the country's population.

The composition of Russia's ethnic population cannot beproperly understood without acknowledgment of itsinstitutionalization within an ethnofederal structure.

4 Census data on the ethnic composition of Russian raions at the raion-level were collected by the author during fieldwork in Russia in 2003.Electoral results for parliamentary elections from 1995 through 2011 arenow available at the Russian CEC website.

R. Goodnow et al. / Electoral Studies 36 (2014) 15e27 21

Russia is composed of 83 subjects or regions that aredivided into 21 republics, 46 oblasts, nine krais, two cities(Moscow and St. Petersburg), four autonomous okrugs, andone autonomous oblast (the Jewish Autonomous Oblast).Republics, autonomous okrugs, and the Jewish Autono-mous Oblast make up the 26 regions named after a non-Russian ethnic group. Overlaying these federal subjectsare seven federal okrugs, which were part of Putin's pro-gram of power verticalization.

As Harris (1993: 571) has demonstrated, Russia's ethnicfederalism has had a strong influence on the ethnic identityof non-Russians. Federal ethnic groups provided with“ethnic homelands” managed to retain their native lan-guages and resist assimilation efforts to a much greaterextent than (often larger) non-federal minority groups.Moreover, regional governments in ethnic homelands havecultivated ethnic mobilization in centereperiphery re-lations. The very designation of ethnic federal units hasproduced incentives and resources for the promotion of anethnic cadre of elites and mass mobilization tied toethnicity (see Roeder et al., 1991; Treisman, 1997;Gorenburg, 2003; Hale, 2003).

In many cases, the titular nationality does not constitutea majority within its designated ethnic region and com-prises a majority in relatively few (former) single-memberelectoral districts. Despite these demographic obstacles,these groups managed to attain significant representationin the national legislature, which suggests that the sym-bolic power of an ethnic homeland is an important factor inthe representation of minorities in the State Duma (Moser,2008). Moreover, raion-level analysis of voting patterns inRussian SMD elections in 1995 shows a significant level of“ethnic voting” driven by the demographic concentrationof non-Russian minorities in single-member districts.Goodnow and Moser (2012) have shown that majority-minority districts had a significant impact on the voteshare of minority candidates who won election as well asminority voter turnout. Minority winners tended to gainlarger vote shares in electoral districts with a majority non-Russian population and majority-minority districts tendedto increase the size of the effect of raion-level minorityconcentration. In other words, minority voters tended tosupport minority candidates in larger numbers when sit-uated in majority-minority districts. Majority-minoritydistricts tended to have a similar amplifying effect on mi-nority voter turnout (Goodnow and Moser, 2012).

In short, ethnic federalism has worked in ways thatscholars recommending decentralization in ethnicallydiverse countries would expect. Ethnic-based decentral-ization aided the representation and empowerment ofnon-Russian elites, especially in the 1990s when regionsenjoyed greater autonomy, by facilitating demographicconcentration and providing material and symbolic re-sources that encourage ethnic voting and other forms ofethnic political mobilization.

However, as the works of Myagkov, Ordeshook, andShakhin (2008, 2009) and others have powerfully demon-strated, the ethnic-based regions that lie at the heart ofRussia's ethnic federalism have also produced pockets ofelectoral manipulation as well as other manifestations ofauthoritarian politics (McFaul and Petrov, 1997). Myagkov

and Ordeshook (2008) show quite clearly that evidence ofelectoral manipulation was initially located in large part inethnic republics. For example, in the 1996 presidentialelection they show that suspicious voting patterns(implausible shifts of votes in the second round of theelection that suggest vast majorities who supported theCommunist challenger, Zyuganov, in the first roundswitched to support President Yeltsin in the second round)were concentrated in ethnic republics. These patterns wererelatively rare overall compared to later elections in whichelectoral manipulation was believed to be more rampantbut were almost exclusively located in raions in ethnic re-publics, namely three e Bashkortostan, Dagestan, andTatarstan. Only 12% of raions suspected of fixed elections in1996 were located in Russian oblasts (Myagkov andOrdeshook, 2008: 4e5). This discrepancy in the incidenceof manipulation between republics and oblasts extendswell into the Putin years. Myagkov and Ordeshook's (2008)examination of the distribution of voter turnout from 1996through 2003 clearly shows that this particular proxy forelectoral manipulation (excessively high turnout) wasfound almost exclusively in ethnic republics while Russianoblasts had normal patterns of voter turnout (see alsoMebane and Kalinin, 2009). It was only beginning in thepresidential election of 2004 that widespread evidence ofmanipulation appeared in Russian regions and even thensuch behavior wasmuchmore prevalent in ethnic republics(Myagkov and Ordeshook, 2008; Mebane and Kalinin,2009).

In this paper, we utilize the ethnic and electoral varia-tion within regions to more systematically examine therelationship between ethnicity and electoral fraud in Rus-sia's two most recent parliamentary elections in 2007 and2011, when electoral manipulation was deemed to havespread beyond the ethnic republics. We show that even inthese elections the incidence of electoral manipulation inraions with large concentrations of ethnic minoritiesgreatly outstripped that of majority-Russian raions, evenwhen controlling for other factors such as degree of ur-banization, federal status (ethnic vs. Russian region), andsocioeconomic development.

5. Empirical analysis

5.1. Data: Russia as a test case of the impact of ethnicity onelectoral manipulation

The analyses that follow are based on electoral datafrom the Central Election Commission of Russia, ethnicitydata from the 2002 All-Russian National Census, and so-cioeconomic data from Rosstat, the Russian State StatisticsService.4 The election and census data are reported at thelevel of each raion in the country, which gives us approxi-mately 2700 observations for each election year on whichto base our findings. In addition to the ethnicity data, the

5 Since religion and ethnicity are closely associated in Russia, we useethnicity as a proxy for religious tradition. Certain ethnic groups arecoded as Muslim (e.g., Tatars, Bashkirs, Ingush) while the Russian ma-jority and other non-Russian minorities are coded as non-Muslim (e.g.,Buryat, Kalmyk, Mordva).

6 Data on the percentage of regional economies based on the export ofnatural resources come from Bradshaw (2006).

R. Goodnow et al. / Electoral Studies 36 (2014) 15e2722

Rosstat website reports the percentage of the rural popu-lation in each raion, which, in combination with informa-tion about each region's Gross Regional Product (GRP) andstatus as a resource-rich region, provides a measure of so-cioeconomic welfare.

Ethnic regions (republics and autonomous okrugs) havea preponderance of raions with a high concentration ofethnic minorities, while majority-Russian regions (oblastsand krais) are predominantly made up of majority-Russianraions. However, because ethnic regions often contain alarge proportion of Russians, there is a large enoughnumber of raions with substantial Russian populations toallow for meaningful comparisons between pockets ofRussian and minority voters within different electoral en-vironments. Indeed, nearly 34% of raions in non-Russianregions in our sample have a majority-Russian popula-tion. Pockets of minority concentration within Russian re-gions are less common, with approximately 3% of raions inRussian regions comprised of a majority-minority popula-tion. This variation in the concentration of minority pop-ulations within and across different regional contexts givesus considerable analytical leverage for isolating the po-tential causes of electoral manipulation. We can use thevariation within federal regions to examine not onlywhether there are differences in electoral patterns betweenRussian and non-Russian regions but also whether thereare significant differences between raions with minority-dominant populations and Russian-dominant populationswithin regions. This should allow a finer grained analysis ofwhether electoral manipulation in regions with large mi-nority populations was a general phenomenon found inRussian and non-Russian raions alike or an outcome thatwasmost common inminority-dominated raions. If it is thelatter, this would indicate more clearly than region-levelanalyses that there may be an ethnic component to elec-toral manipulation in Russia.

5.2. Independent variables: factors explaining the variation inelectoral manipulation

As noted above, previous scholarship, especially byMyagkov et al. (2009), has already firmly established thatelectoral manipulation was much more prevalent in ethnicminority regions than in Russian ones. Our primary intentin this paper is to study variationwithin ethnic regions andwe expect several factors to influence the degree of elec-toral manipulation found within ethnic republics. If, as weargue above, electoral manipulation is driven by incentivesto control politics and the capacity to coordinate such ac-tivities, we expect that raions with characteristics that in-crease one or both of these dynamics will experience ahigher incidence of electoral manipulation. We argue thatethnicity may play a vital role in this process and thuselectoral manipulation will be more prevalent in local en-vironments (raions) with high concentrations of ethnicminorities.

Thus, the variable of greatest interest is ethnicity, morespecifically the ethnic composition of raions (% non-Russian) located within different types of federal units.Raions with concentrations of ethnic minorities may beperceived as a potential threat to the state and thus require

“special efforts” to maintain control and also may possessresources embedded in ethnic mobilization that promotethe use of patronage and electoral machine politics (Hale,2003).

Similarly, we expect five other broader contextualcharacteristics of raions e urban versus rural raions andlocation in: Muslim regions, the Caucasus region, andresource-rich regions e to affect the likelihood that a givenraion will experience electoral manipulation. FollowingMyagkov and Ordeshook (2008, 2009), we expect urbanraions to be less susceptible to electoral manipulation thanrural raions (Rural) because the former provides feweropportunities for elite intervention into the electoral pro-cess than the latter. We also expect raions located inMuslim ethnic regions (Muslim) to be more manipulatedthan raions in non-Muslim or Russian regions since in-centives for national authorities to control Muslim regionsmay be higher given the experience with the Chechen Warand broader violence in the Caucasus as well as the prev-alence of drives for autonomy during the 1990s, whichwere strongest among Muslim ethnic republics (Treisman,1997).5 For similar reasons (greater incentives and re-sources to manipulate elections), we expect raions locatedin the Caucasus (Caucasus) and resource-rich regions(Resource) to be more likely to experience manipulatedelections.6 We also control for economic development atthe regional level, expecting that regions with higher levelsof prosperity will have higher turnout than regions withlower levels of prosperity.

Fig. 3a and b and Table 1 provide a descriptive snapshotof the occurrence of electoral manipulation in Russianraions possessing the characteristics we outlined above.These descriptive statistics showcase that both regional-and raion-level factors seem to be associated with thepresence of suspiciously high voter turnout (our proxy forelectoral manipulation). Raions with high voter turnoutwere found much more often in non-Russian raions thanRussian raions. More importantly, while electoral manip-ulation in general also appeared to be more prevalent inthose regions where we would expect (non-Russian re-publics, the Caucasus, Muslim regions, resource-rich areas),even in these contexts the gap between non-Russian andRussian raions remained intact.

5.3. Analytical approach

To explore these potential causes of electoral manipu-lation, we specify a series of logistic multilevel models thatconsider the effects of the varying proportions of ethnicand rural populations in raions on the probability thatelectoral manipulation will occur (i.e., that turnout will begreater than 75%), in addition to the contextual influence ofparticular regional characteristics on these variables'

Table 1Electoral manipulation and different electoral environments in 2007 and 2011.

2007

Turnout>75%

T<

All raions 700 (26%) 2Majority minority raions 326 (73%)Majority Russian raions 374 (17%) 1Rural raions 518 (36%)Urban raions 182 (14%) 1

Raions in:Ethnic federal regions 378 (65%)Maj. minority raions 309 (79%)Maj. Russian raions 69 (36%)

Russian federal regions 322 (15%) 1Maj. minority raions 17 (31%)Maj. Russian raions 305 (15%) 1

Muslim regions 234 (94%)Maj. minority raions 206 (96%)Maj. Russian raions 28 (85%)

Caucasus regions 165 (58%)Maj. minority raions 99 (89%)Maj. Russian raions 20 (12%)

Natural resource regions 193 (45%)Maj. minority raions 131 (75%)Maj. Russian raions 62 (24%)

Poorer regions 335 (27%)Maj. minority raions 174 (72%)Maj. Russian raions 161 (16%)

Fig. 3. (a and b) Ethnicity and electoral fraud in 2007 and 2011.

R. Goodnow et al. / Electoral Studies 36 (2014) 15e27 23

intercepts and slopes. The regions vary according towhether they are ethnic republics or autonomous okrugs(compared to oblasts or krais), majority-Muslim, located inthe Caucacus, and on the basis of their GRP's and naturalresource wealth. An advantage of the multilevel approach,compared to standard logistic regression, is that the dif-ferences in intercepts and slopes from the regional controland interaction terms can vary randomly across regions.

The equation for the first region type e ethnic versusnon-ethnic regions e is

log�

pij

1�pij

�¼b0þb1minorityijþb2ruralijþb3ethnicj

þb4ethnic*minorityijþb5ethnic*ruralijþuj

where, pij is the probability that the dependent variable isequal to one, i.e., an instance of potential electoralmanipulation, for raion i in region j. b0 is the constantterm and equal to the log-odds that manipulation willoccur when the variables are equal to zero. b1 and b2 arethe region-specific effects on the log-odds of manipula-tion occurring from a 1% increase in raions' minorityand rural populations, respectively. b3 is the direct effectfrom being in an ethnic region on the log-odds thatmanipulation will occur, and b4 and b5 are the moderatingeffects on the slopes of the two respective raion-levelvariables from being in an ethnic region. When expo-nentiated, the coefficient estimates are odds ratios,comparing the odds of a raion being manipulated toraions in the same region for each unit change in therespective independent variables. The random

2011

urnout75%

Turnout>75%

Turnout<75%

010 (74%) 553 (20%) 2160 (80%)118 (27%) 286 (65%) 157 (35%)892 (83%) 267 (12%) 2003 (88%)904 (64%) 401 (28%) 1019 (72%)106 (86%) 152 (12%) 1141 (88%)

203 (35%) 333 (58%) 238 (42%)80 (21%) 275 (72%) 108 (28%)

123 (64%) 58 (31%) 130 (69%)807 (85%) 220 (10%) 1922 (90%)38 (69%) 11 (18%) 49 (82%)

769 (85%) 209 (10%) 1873 (90%)14 (6%) 220 (88%) 29 (12%)9 (4%) 199 (92%) 17 (8%)5 (15%) 21 (64%) 12 (36%)

119 (42%) 138 (48%) 150 (52%)12 (11%) 108 (96%) 4 (4%)

153 (88%) 30 (17%) 146 (83%)237 (55%) 167 (39%) 264 (61%)44 (25%) 115 (66%) 60 (34%)

193 (76%) 52 (20%) 204 (80%)905 (73%) 260 (21%) 981 (79%)68 (28%) 167 (69%) 76 (31%)

837 (84%) 93 (9%) 905 (91%)

R. Goodnow et al. / Electoral Studies 36 (2014) 15e2724

component of the model, or level-two residual uj, reflectsthe effect on the log-odds that manipulation will occurfrom raions located in region j.

We specify the above model for both the 2007 and 2011parliamentary elections, in addition to four additionalmodels for each election year that account for the directand interactive effects of the remaining region types. Tables2 and 3 contain the results.

5.4. Discussion of the results

With just one exception, the variables accounting forthe percentage of ethnic minority and rural populations atthe raion-level had positive and statistically significant ef-fects on the chance of electoral manipulation occurring. For

Table 2Multilevel logistic analysis of electoral fraud in the 2007 Russian parliamentary

Model 1 Model 2

Intercept �3.87 (0.353)*** �3.76 (0.303)***

Raion-level:Percent minority 0.0148 (0.006)***

OR 1.01500.0256 (0.005)***OR 1.0260

Percent rural 0.0183 (0.002)***OR 1.0186

0.0191 (0.002)***OR 1.0194

Region-level intercept effects:Ethnic region 1.3612 (0.708)*

OR 3.9011Muslim region 4.566 (1.42)***

OR 96.2445Caucasus region

Economic development

Natural resource wealth

Regionelevel slope effects:Percent minority*Ethnic region 0.0208 (0.010)**

OR 1.0211*Muslim region 0.0026 (0.017)

OR 1.0027*Caucasus region

*Economic development

*Natural resource wealth

Percent rural*Ethnic region -0.0010 (0.005)

OR 0.9989*Muslim region -0.0107 (0.008)

OR 0.9893*Caucasus region

*Economic development

*Natural resource wealth

Between region variance:LR statistic 650.13*** 586.91***

Model fit:Wald statistic (df) 154.19 (5)*** 164.51 (5)***

Notes: The reported coefficients were estimated in STATA using adaptive Gaussia75% in the raion, and zero otherwise; level-1 n ¼ 2710, level-2 n ¼ 84; ***p < 0.

every percent increase in each of these variables, the oddsof a raion beingmanipulated increased by 1e3%, dependingon the model.

At the regional level, the variables accounting forparticular contexts are especially revealing. Surprisingly,the relative financial and resource wealth of regions had noconnection to the occurrence of manipulation either inde-pendently or by moderating the effects of ethnic raions(models 4 and 5). The only effect concerned rural raions in2011, which were at slightly higher risk of being manipu-lated in resource-rich regions than similar raions in non-resource regions. In contrast, the regions classified asethnic, Muslim, or Caucasus had increased odds ofmanipulation compared to other regions, with interestingvariations in the respective intensities of their effects.

elections.

Model 3 Model 4 Model 5

�3.78 (0.336)*** �3.38 (0.369)*** �3.54 (0.345)***

0.0271 (0.005)***OR 1.0275

0.0335 (0.006)***OR 1.0342

0.0310 (0.005)***OR 1.0315

0.0208 (0.002)***OR 1.0211

0.0179 (0.003)***OR 1.0182

0.0170 (0.002)***OR 1.0172

2.466 (1.063)**OR 11.7829

-0.0798 (0.152)OR 0.9233

0.4232 (0.939)OR 1.5270

0.0098 (0.015)OR 1.0099

-0.0033 (0.004)OR 0.9966

-0.0072 (0.012)OR 0.9928

-0.020 (0.006)***OR 0.9801

-0.0003 (0.002)OR 0.9997

0.0070 (0.006)OR 1.0071

725.67*** 673.99*** 710.87***

152.03 (5)*** 144.09 (5)*** 141.99 (5)***

n quadrature (AGQ); the dependent variable is 1 if turnout is greater than01, **p < 0.05, *p < 0.1.1.

Table 3Multilevel logistic analysis of electoral fraud in the 2011 Russian parliamentary elections.

Model 1 Model 2 Model 3 Model 4 Model 5

Intercept �4.66 (0.467)*** �4.17 (0.391)*** �4.49 (0.419)*** �4.00 (0.457)*** �3.97 (0.448)***Raion-level:Percent minority 0.0103 (0.008)

OR 1.01040.0172 (0.005)***OR 1.0174

0.0178 (0.005)***OR 1.0180

0.0274 (0.007)***OR 1.0278

0.0242 (0.006)***OR 1.0245

Percent rural 0.0144 (0.003)***OR 1.0145

0.0141 (0.002)***OR 1.0142

0.0197 (0.003)***OR 1.0199

0.0139 (0.004)***OR 1.0141

0.0133 (0.003)***OR 1.0135

Regionelevel intercept effects:Ethnic region 2.647 (0.853)***

OR 14.1145Muslim region 4.322 (1.532)***

OR 75.3893Caucasus region 3.737 (1.393)***

OR 41.9910Economic development -0.0007 (0.154)

OR 0.9992Natural resource wealth -0.0181 (1.100)

OR 0.9820

Regionelevel slope effects:Percent minority*Ethnic region 0.0087 (0.011)

OR 1.0088*Muslim region 0.0074 (0.016)

OR 1.0074*Caucasus region 0.0173 (0.016)

OR 1.0175*Economic Development -0.0038 (0.005)

OR 0.9961*Natural resource wealth -0.0090 (0.012)

OR 0.9910Percent rural*Ethnic region 0.0074 (0.005)

OR 1.0075*Muslim region 0.01852 (0.008)**

OR 1.0187*Caucasus region -0.019 (0.006)***

OR 0.9811*Economic development 0.0015 (0.002)

OR 1.0015*Natural resource wealth 0.0120 (0.005)**

OR 1.0121

Between region variance:LR statistic 635.04*** 558.72*** 663.01*** 677.25*** 693.61***

Model fit:Wald statistic (df) 109.41 (5)*** 109.47 (5)*** 109.97 (5)*** 93.93 (5)*** 95.11 (5)***

Notes: the reported coefficients were estimated in STATA using adaptive Gaussian quadrature (AGQ); the dependent variable is 1 if turnout is greater than75% in the raion, and zero otherwise; level-1 n ¼ 2712, level-2 n ¼ 83; ***p < 0.01, **p < 0.05, *p < 0.1.

R. Goodnow et al. / Electoral Studies 36 (2014) 15e27 25

The most striking result was the very large increase inthe odds of electoral manipulation for raions located inMuslim regions (model 2). These odds were as high as 96 in2007 and 75 in 2011. The more rural raions in these regionsalso had a higher chance of manipulation in 2011 thansimilar raions in other regions. For raions in Caucasus re-gions (model 3), the odds of manipulation were consider-ably higher than the odds of manipulation in non-Caucasusregions, at 12 in 2007 and 42 in 2011, but these were muchlower than the odds of manipulation in Muslim comparedto non-Muslim regions. An interesting result for the Cau-casus regions is the diminished influence of rural raions onthe odds of manipulation occurring within this areacompared to the rest of the country. Lastly, raions in thebroader category for ethnic regions (model 1), which

comprised all raions in republics and autonomous okrugs,were still at greater risk of manipulation than raions inoblasts and krais, but to a much lesser extent than in thenarrower Muslim and Caucasus categories. The odds ofelectoral manipulation in ethnic regions were four times aslarge as non-ethnic regions in 2007 and 14 times as large in2011. Also of interest, the effect from the percentage ofminorities in raions on the odds that manipulation willoccur was intensified in the context of ethnic regions dur-ing the 2007 election.

6. Conclusion

In this paper we examined the relationship betweenethnicity and electoral manipulation by asking whether

R. Goodnow et al. / Electoral Studies 36 (2014) 15e2726

discernible differences exist in the incidence of suspiciouselectoral results between raions with different concen-trations of Russian and non-Russian populations. We alsoconsidered the impact of various social, political, andeconomic factors that could affect the occurrence ofmanipulation independently and in combination withethnicity. Our initial findings suggest that ethnicity, asexperienced at the local (raion-level), has a prevailing ef-fect on the occurrence of electoral manipulation. In a va-riety of contexts, raions with high concentrations of non-Russian minorities tended to experience voting patternsindicative of electoral manipulation more often thanRussian raions. One implication of this finding is thatelectoral manipulation in Russia seems to be a moredecentralized process than often depicted by region-basedanalyses. If regional leaders were delivering the vote onthe orders of national elites in a centralized fashion, onemight expect more uniform evidence of fixing electionresults within regions rather than the patchwork patternof high voter turnout in raions with greater concentrationsof ethnic minorities or more rural populations. We cannotdiscern from this study the precise mechanisms of howelectoral manipulation works in Russia but it hints thatlocal elites and populations are a central component of theprocess. Further research is clearly required to investigatethese processes.

The much higher levels of manipulated raions withinMuslim versus non-Muslim regions suggest that the type ofethnic cleavages in Russia may play a role in this dynamicas well. This is most likely due to historical experience withmobilization against the state, which increases theperceived threat posed by these regions, raising incentivesto convey political control. One possible extension of thisresearch would be to further investigate variance of elec-toral manipulation among raions with concentrations ofdifferent types of non-Russian ethnic groups.7

Another avenue for further research is investigation ofchange over time. Comparison of our cross-sectional datasuggests interesting similarities and differences betweenthe elections of 2007 and 2011. Ethnicity appears to havehad a significant impact in both elections but the gap be-tween Russian and non-Russian raions is greater in 2011(see Fig. 3a and b), which is surprising given that the 2011election actually witnessed a substantial decline in the voteshare for United Russia as it was under greater pressure toavoid the egregious manipulation of the previous election.Arguably, as electoral manipulation spreads more widely ittends to encroach on regions without concentrations ofethnic minorities and, conversely, when it recedes

7 For example, we did a preliminary examination of Bashkortostan,which is one of the more ethnically diverse regions in Russia with largeconcentrations of the titular group, Bashkirs, but also significant numbersof Russians and other non-Russian minorities, most notably, Tatars. It wasfound that, while the largest gap in the incidence of manipulation wasfound between Russian versus non-Russian raions, there was also asignificantly larger number of manipulated raions among the majority-Bashkir regions as opposed to raions with simply a majority-minoritypopulation comprised of other minority groups as well. This suggeststhat titular minorities may provide even greater capacity to manipulateelections than other minorities.

manipulation still persists in those areas with the greatestincentives to convey control and the resources to coordi-nate manipulation. Moreover, different types of elections,such as presidential elections, may also affect electoralmanipulation by raising the stakes and thus the incentivesto demonstrate political control. Further analyses of moreRussian elections (both parliamentary and presidential) areneeded to see if these relationships bear out.

References

Beissinger, M., 2002. Nationalist Mobilization and the Collapse of theSoviet State. Cambridge University Press, Cambridge.

Bradshaw, M., 2006. Observations on the geographical dimensions ofRussia’s resource abundance. Eurasian Geogr. Econ. 47 (6), 724e746.

Carothers, T., 1997. The observers observed. J. Democr. 8 (3), 17e31.Chandra, K., 2004. Why Ethnic Parties Succeed. Cambridge University

Press, Cambridge.Cornwell, E., 1964. Bosses, machines, and ethnic groups. Ann. Am. Acad.

Polit. Soc. Sci. 353, 27e39.Dahl, R., 1961. Who Governs? Yale University Press, New Haven.Fish, M.S., 2005. Democracy Derailed in Russia. Cambridge University

Press, Cambridge.Gorenburg, D., 2003. Minority Ethnic Mobilization in the Russian Feder-

ation. Cambridge University Press, Cambridge.Goodnow, R., Moser, R., 2012. Layers of ethnicity: the effects of ethnic

federalism, majorityeminority districts, and minority concentrationon the electoral success of ethnic minorities in Russia. Comp. Polit.Stud. 45 (2), 238e264.

Hale, H., 2003. Explaining machine politics in Russia’s regions: economy,ethnicity, and legacy. Post-Soviet Aff. 19 (3), 228e263.

Hale, H., 2007. Correlates of clientelism: political economy, politicizedethnicity, and postcommunist transition. In: Kitschelt, H.,Wilkinson, S. (Eds.), Patrons, Clients, and Policies: Patterns of Dem-ocratic Accountability and Political Competition. Cambridge Univer-sity Press, Cambridge, UK, pp. 227e250.

Harris, C., 1993. The new Russian minorities: a statistical overview. Post-Soviet Geogr. 34.

Horowitz, D., 1985. Ethnic Groups in Conflict. University of CaliforniaPress, Berkeley.

Horowitz, D., 2003. Electoral Systems: A Primer for Decision Makers. J.Democr. 14 (4), 115e127.

Inglot, T., Pelissero, J., 1993. Ethnic political power in a machine city:Chicago's poles at Rainbow's end. Urban Aff. Q. 28, 526e543.

King, R., 2001. Counting the votes: South Carolina's stolen election of1876. J. Interdiscip. Hist. 32 (1), 169e191.

Lehoucq, F., 2003. Electoral fraud: causes, types and consequences. Annu.Rev. Polit. Sci. 6, 233e256.

Lijphart, A., 1977. Democracy in Plural Societies: A Comparative Explo-ration. Yale University Press, New Haven.

Lijphart, A., 2004. Constitutional design for divided societies. J. Democr. 15(2), 96e109.

Lukinova, E., Myagkov, M., Ordeshook, P., 2011. Ukraine 2010: wereTymoshenko’s cries of fraud anything more than smoke? Post-SovietAff. 27 (1), 37e63.

Mansfield, E., Snyder, J., 2005. Electing to Fight: Why Emerging. De-mocracies Go to War. MIT Press, Cambridge.

Marsh, C., Warhola, J., 2001. Ethnicity, ethno-territoriality, and the polit-ical geography of Putin's electoral support. Post-Soviet Geogr. Econ.42 (No. 4), 1e14.

McFaul, M., Petrov, N., 1997. Russian electoral politics after transition:regional and national assessments. Post-Soviet Geogr. Econ. 38 (9),507e549.

Mebane, W., Kalinin, K., 2009. Comparative election fraud detection. In:Prepared for Presentation at the Annual Meeting of the AmericanPolitical Science Association, Toronto, Canada, Sept 3e6, 2009.

Moser, R., 2008. Electoral systems and the representation of ethnic mi-norities: evidence from Russia. Comp. Polit. 40 (3), 273e292.

Myagkov, M., Ordeshook, P.C., February 1, 2008. Russian Elections: AnOxymoron of Democracy. Available at: SSRN. http://ssrn.com/abstract¼1090975.

Myagkov, M., Ordeshook, P., Shakin, D., 2009. The Forensics of ElectoralFraud: Russia and Ukraine. Cambridge University Press, New York.

Powell, L., 1989. Correcting for fraud: a quantitative reassessment of theMississippi ratification election of 1868. J. South. Hist. 6, 633e658.

Reilly, B., 2002. Electoral systems for divided societies. J. Democr. 13 (1),156e170.

R. Goodnow et al. / Electoral Studies 36 (2014) 15e27 27

Roeder, P., 1991. Soviet federalism and ethnic mobilization. World Polit.43, 196e232.

Saideman, S., Lanoue, D., Campenni, M., Stanton, S., 2002. Democratiza-tion, political institutions, and ethnic conflict: a pooled, cross-sectional time series analysis from 1985e1998. Comp. Polit. Stud. 35(1), 103e129.

Simpser, A., 2013. More than Winning: Why Governments and PartiesManipulate Elections. Cambridge University Press, Cambridge.

Snyder, J., 2000. From Voting to Violence: Democratization and Nation-alist Conflict. Norton Books, New York.

Stoner-Weiss, K., 2001. The Russian central state in crisis: center andperiphery in the post-soviet era. In: Barany, Z., Moser, R. (Eds.),Russian Politics: Challenges of Democratization. Cambridge Univer-sity Press, New York.

Treisman, D., 2012. The Return: Russia’s Journey from Gorbachev toMedvedev. The Free Press, New York.

Treisman, D., January 1997. Russia's ethnic ‘revival’- the separatist activismof regional leaders in a postcommunist order.WorldPolit. 49, 212e249.

White, S., 2011. Elections Russian-style. Europe-Asia Stud. 63 (4),531e566.

Wilkinson, S., 2004. Votes and Violence: Electoral Competition and EthnicRiots in India. Cambridge University Press, Cambridge.

Wolfinger, R., 1972. Why political machines have not withered away andother revisionist thoughts. J. Polit. 34, 365e398.

Ziblatt, D., 2009. Shaping democratic practice and the causes of electoralfraud: the case of nineteenth-century Germany. Am. Polit. Sci. Rev.103 (1), 1e21.