Thesis Draft 20141004 1 Migration as an adaptive response to ethnic nationalism in Russia Michelle O’Brien A thesis in partial fulfillment of the requirements for the degree of Master of Arts University of Washington 2014 Committee: Steven Pfaff Stewart Tolnay Program Authorized to Offer Degree: Sociology 2 (c) Copyright 2014 Michelle O’Brien 3 University of Washington Abstract Migration as an adaptive response to ethnic nationalism in Russia Michelle O’Brien Chair of the Supervisory Committee: Steven Pfaff Sociology In this paper, I argue that migration responses to push factors differ along ethnic lines. I examine this hypothesis using panel survey data from the Russia Longitudinal Monitoring Survey, census data, and regional-level political data from Russia. I hypothesize that nationalist political parties send signals of anti-minority sentiment, which ethnic minorities interpret as threatening to their prospects in the region. This leads to a demand for an adaptive response, generating out- migration. I estimate an event history model and find that, although ethnic minorities do not demonstrate a higher propensity to migrate than the majority group, they do respond differently to political and economic push factors, in particular, to signals sent by nationalist parties. 4 In December of 2010, an ethnic riot erupted in Moscow (Barry 2010). The event appeared to be the result of a build up of ethnic tensions, fueled by xenophobic politics and sparked by episodic incidents of violence. While President Vladimir Putin has remained relatively conservative in his use of ethnicity in speeches, anti-minority sentiment in mainstream politics has been steadily growing, even as the country experiences more economic stability than at any other time after the fall of the USSR. The fourth largest political party in Russia is Vladimir Zhirinovsky’s so-called Liberal Democratic Party of Russia (LDPR). A stable ideological party, with varying regional success at the polls, it is rooted in deep-seated xenophobia and ethnonationalist rhetoric. Its success poses a threat to the status and life chances of ethnic minorities. Drawing on Hale’s (2008) relational theory of ethnic politics, I argue that this threat generates a demand for adaptive response. Two major venues through which actors may adapt are exit (i.e. migration) and voice (i.e. political advocacy). With opportunities for minority group advocacy severely limited in Putin’s Russia, this research focuses on migration as an adaptive response to ethnic politics. I hypothesize that nationalism, as a component of ethnic politics, contributes to out-migration of ethnic minorities in contemporary Russia. To arrive at the outcome of migration as an adaptive response in which minorities engage, two processes are necessary. First, an individual making the decision to migrate must interpret ethnic tensions as a threat to her life chances, and she must evaluate her future prospects in this ethnically charged framework. Second, the option of migration must be a viable one. That is, an individual must consider herself the plausible target of the threat of diminishing life chances, conclude that an adaptive response is required, and determine that the benefits of migrating outweigh the costs. In order to explain these processes, I employ the relational theory of ethnic 5 politics (Hale) and demographic theories. I estimate an event history model using regional, household, and individual-level data from Russian censuses and the Russia Longitudinal Monitoring Survey. I find evidence that political push factors affect minority groups differently than the ethnic majority, suggesting that the success of ethnonationalist politics in a region signals vulnerability to ethnic minorities, influencing migration decisions. Ethnic Tensions and Life Chances in Russia When does an individual interpret ethnic tensions as a threat to her own life chances? Why might minorities flee Russia in the face of ethnic tension, even if the sentiment is not directly aimed at or destructive towards them? Hale’s (2008) relational theory of ethnic politics distinguishes between ethnicity as a mechanism to reduce uncertainty and ethnic politics as the strategies a group engages in to increase member life chances. Identity, generally speaking, is a relational way in which actors make sense of their world. Ethnic identity, then, becomes salient when this particular point of reference “come[s] to have greater importance in people’s lives, when people’s lives are seen to be affected in more significant ways by the referent,” (2008:36). This is especially true when one group has the power to affect the other group’s average life chances, as through employment discrimination. Ethnic identification must not only be salient, but also be accessible (i.e. through contextual clues or repeated exposure) and must fit the situation (even if based on incomplete or wholly inaccurate stereotypes). In addition, ethnicity is a particularly “sticky” identifier, as it encompasses four categories that increase its accessibility and fit to social life: 1) ethnicity implies a common fate; 2) ethnicity tends to involve barriers to communication (both linguistically and culturally); 3) ethnicity carries visible physical differences that are hard to change; and 4) ethnic traits are usually correlated with less visible 6 traits (such as attitudes and values) (2008:42–43). Thus, while ethnicity may not be the only information in a given scenario, it is an important one for simplifying and making sense of social life. If ethnicity is pre-rational and exists as a mechanism to reduce uncertainty, then what people do in their newly simplified world is ethnic politics. Ethnic politics is a categorical way of thinking applied to widespread human desires. Hale defines these human desires as material gain, power, status, and/or security. Material well-being is a long-term desire, in which group members take a prospective approach to economic gains, potentially seeing a temporary tradeoff as a lifelong impediment. Power is a fungible good that may lead to other gains, such as economic prosperity or physical security. Status and self-esteem may be desires that, if left unfulfilled, lead to resentment and ethnic tension. Security can be considered a survival motive. It is likely to be important when “physical well-being is potentially at stake along group lines” (2008:54). When threatened with a reduction in life chances, then, ethnic identity becomes salient and actionable. Ethnicity certainly appears to be salient in Russia. The Russian language differentiates between ethnic Russians, russkii, and citizens of Russia, rossiiskii, and nationalist parties tend to follow suit, using the ethnic term for Russian when calling for a return of the country to its people (“Rossiya dlya russkikh” or “Russia for [ethnic] Russians”). Historically, the Russian Federation has taken a “fill-in-the-blank” approach to self-identified ethnicity. The Russian Census has respondents fill in their identified ethnicity; with no categorical options that one sees on the U.S. Census. In 2010, the Russian Census recorded over 160 ethnic groups residing within its borders. Table 1 shows the most numerous ethnic groups in the Russian Federation, as measured by the 2010 Russian Census. 7 [Table 1 about here] Although official statistics are unavailable, reports of discrimination and violence against ethnic “others” is common, and an oft-cited area of concern for the United Nations (Kulaeva et al. 2013; Verkhovsky, Kozhevnikova, and Sibireva 2010). Tension with minorities and migrants is evident in opinion polling as well as the media. Since 2000, the Levada Center has consistently reported that a majority of a representative sample supported the phrase “Russia for ethnic Russians” (Anon 2011a). In 2012, nearly 30% of respondents said that they can feel ethnic tensions in the city or town in which they live. When asked “Are violent ethnic clashes possible in Russia now?” 43% answered yes, and when asked the same question about their specific communities, 23% of respondents answered yes. Over 40% of respondents reported feeling annoyance or dislike toward residents of the “southern” republics (the Caucasus), and 42% reported that population restrictions should be placed on people from the north Caucasus ethnic groups living in Russia1. In a 2005 Levada Center opinion poll, a majority of respondents answered that they felt negatively about the employment of migrants in law enforcement, public service, and private enterprise. Similarly, the poll found a majority disapproval of migrants acquiring any type of property. The proportion of respondents who answered that they did not want a “foreigner” as a neighbor varied by origin of the “foreigner”, ranging from under 20% for Ukraine to over 70% for the Caucasus, Central Asia, or China2. 1 It is worth noting here that residents of the north Caucasus are citizens of the Russian Federation. The north Caucasus consists of the Russian oblasts of Dagestan, Chechnya, Ingushetia, North Ossetia, Kabardino-Balkaria, Karachay-Cherkessia, parts of Krasnodar Krai, Adygea, and parts of Stavropolskii Krai. 2 Any one of these ethnic group members may be, in fact, a citizen of Russia. 8 Ethnicity is a matter of importance to contemporary politics and politicians as well. The current president, Vladimir Putin (2012), has written publicly about ‘the ethnicity issue,’ both calling for harsh legal action against disrespectful internal migrants, and suggesting that ethnic groups must live in harmony in order to secure the Russian state. More recently, Putin explicitly used the Russian ethnic background and shared linguistic and cultural history as pretext for the invasion of Ukraine’s Crimean region, fueling fears about the implications of Putin’s bold geopolitical action (Conant 2014). The main ethnonationalist political party in Russia, the LDPR, targets ethnic minorities through its rhetoric exalting ethnic Russians above all others, and pledging to return Russia to its ethnic Russian inhabitants. Thus, ethnicity is salient for both ethnic Russians and ethnic minorities in Russia, primarily through a hierarchy constructed in which ethnic Russians are above all others. For this reason, I do not focus on any particular ethnic group, and instead focus on the division between ethnic Russians and ethnic non-Russians. I posit that the signal sent by ethnonationalist groups will resonate differently between ethnic Russians and ethnic non-Russians, while specific ethnic groups may not have differential interpretations of the same signal. For instance, minorities may interpret the protest banner that reads “Russia for ethnic Russians” as a signal of potentially worsening prospects. A policy of returning the Russian Federation to its “rightful” ethnic Russian owners will likely be threatening to minority individuals, regardless of specific ethnic group identification. Nationalism as a Signal of Worsening Prospects Nationalism in Russia takes many forms, and is far from new, demonstrated by a long history of tension as nationalist interests attempt to operate within a multi-ethnic state (Kappeler 2001; Smith 1995). In this research, I focus on the right-wing platform of ethnonationalism, in 9 which the primary building block of nationalist rhetoric is anti-minority sentiment. As an indicator of the success of ethnonationalism, I use the vote share of a single party explicitly representing this platform, the LDPR. LDPR is the major ethnonationalist force in the Duma and has been one of the most stable political parties in Russia since the mid-1990s (Hanson 2010). It is an ideological party, relying on rhetoric that defines the future of Russia under an ideal LDPR leadership. “The central features of [Zhirinovsky’s] political creed… have been remarkably clear and consistent since nearly the beginning of his career as a public politician in the late Soviet period,” (Hanson 2010:195). Those features have changed little from Zhirinovsky’s first publications in the early 1990s to the LDPR’s published programs in 2014. The political creed includes an aim to “return to the [ethnic] Russian people the status of nation-state… What’s good for Russians is good for all. For [ethnic] Russians, along with all those indigenous people of Russia, we will build our common Russian home,” and to spread a political ideology that reunites the ethnic Russian people toward the common goal of restoring Russian boundaries to their “historical territory”. The LDPR program explicitly demands a universal system in which the Russian language is studied by every one, without alteration. LDPR claims to aim to “defend the country from migrants” and maintain a “continuous regime of counterterrorism operation” in the north Caucasus (Zhirinovsky n.d.). This familiar nationalist rhetoric places the ethnic Russian on top of an ethnic hierarchy and demands for the restoration of land and status to the population, while defending it from the threats of immigration and foreign politics. The LDPR is stable, with “remarkable organizational capacity in many of Russia’s regions” (Hanson 2010:208) . It is widely recognized party that has gained not only traction in legislative elections, but also media attention. This is important, because it indicates that Russians know what this party stands for, and are familiar with the platform on which it runs. Its 10 success or failure in a region signals information to individuals living in that region, upon which these individuals may evaluate their future prospects in the region and determine the need for adaptation (Lohmann 1993). What, then, is the signal that the success of the LDPR sends? LDPR’s registered leader, Vladimir Zhirinovsky, is blatantly anti-immigrant, anti- minority, and misogynistic. He has been quoted as claiming to intend to “close [all of Russia’s] borders on the day after the election,” (Bidder 2008), and threatening to “seize Alaska from the United States, to launch a nuclear strike on Japan, to flood Germany with radioactive waste, and to occupy the Baltic states,” (Anon 2000). While some consider Zhirinovsky to be merely a showman on the Russian political scene, he has had major successes, including LDPR’s first- place win in the legislative election of 1993 and his own presidential run, which landed him third place in both 1991 and 2008 (Hanson 2010). Zhirinovsky is not the only outspoken nationalist representing the party. Andrei Lugovi famously sits in parliament for LDPR, despite being wanted in Britain for his alleged connection to the poisoning death of Alexander Litvinenko (Anon 2011b). Despite the tendency of Westerners to write off Zhirinovsky because of his theatrics, experts see the LDPR as an enduring, successful, ideologically-coherent political party (Hanson 2010). Because LDPR is explicitly nationalist and pro-ethnic Russian, it serves as a proximate signal of a broader anti-minority sentiment. The interpretation of this signal as future risk of harm to life chances should delineate along ethnic lines. Thus, ethnic minorities receiving this signal update their information about their status in the regional environment and determine that adaptation is necessary, while ethnic Russians would not. The signal that LDPR sends may be global, but the implications are often local. Although electoral laws change often in Russia, during the time period studied, seats in the Duma were allocated based on a proportional representation system (State Duma 2005). In all regional 11 parliaments, either a proportional representation or mixed system is in place, meaning that at least some of the parliament seats are awarded due to the party’s proportion of vote share (Ross 2014). For example, in the Amur Oblast, characterized by over 20% nationalist vote share, the LDPR has taken a corresponding 8 of the 36 regional parliament seats. Popular regional support for LDPR also has indirect but formal political implications, such as the 2012 appointment of party deputy Alexei Ostrovsky as governor of Smolensk, a region in which LDPR captured 14.75% of the vote share in 2011. What difference does it make to have party leadership in Russia’s regions? A case study of Pskov under bold LDPR leadership in the mid- to late-1990s revealed a criminal, mafia-like takeover of the economic and civic organizations in the region (Slider 1999). In this context, elected governor3 Evgeny Mikhailov employed young LDPR members instead of experienced bureaucrats, turned over nearly all of his predecessor’s reforms, and quickly established a strict, vertical hierarchy with himself at the top. At the time of publication (1999), Slider reported that federal authorities paid little attention and made no effort to curb any local abuses made by LDPR leadership. In 2002, reports of discrimination against Polish Catholics in Pskov emerged, with native Russian Orthodox leaders fearing the ‘Catholic expansion’ and plans to “interfere with children of our Orthodox families,” (Myers 2002). In this case, violence was not necessary, administrative hurdles aimed at the foreign influences effectively halted the building of a cathedral in Pskov, and led to expulsions of Polish Catholic religious leaders (ibid). Regional vote share capitalizes on the global signal sent by LDPR, but the presence and power of LDPR has had direct regional impact. 3 In the Putin era, laws regarding the selection of governors have frequently changed. In 2004, during his first term as President, Putin abolished Yeltsin’s initiative of direct elections of governors. In 2012, Medvedev pushed legislation to allow for the option of election of regional governors. In 2013, again-President Putin overturned the reform and began appointing regional governors (Ross 2014). 12 How strong must a signal be in order to contribute to the costly decision to migrate? Perhaps vote share, in a struggling (and regionally variable) democracy (McMann and Petrov 2000), may not be the proxy of choice in determining political signals. To further extend this discussion, I also include a regional measure of reported hate crimes. These data are from the SOVA Center of Information and Analytics, based in Moscow. The center reports hate crimes annually, and breaks the crimes down in a number of ways, including violence (murder, credible threats of murder, stabbing, beating) and vandalism (arson, graffiti, breaking of windows). Like lynching in the U.S. South, hate crimes in today’s Russia are characterized by the victim’s membership in an ethnic minority group. They are rare, and centered in places of political power. However, just as with Tolnay and Beck’s (1992) study on lynching, I expect that hate crimes will have an effect on out-migration of ethnic minorities. I argue that although locally varying and rare, these events will trigger an evaluation of worsening prospects for minority individuals. Migration as an Adaptive Response Migration is a costly event, taken on as an adaptive response to both push factors at the origin and pull factors at the destination, with consideration of both personal factors and intervening obstacles (Lee 1966). Different theoretical camps have suggested mechanisms at every level of motivation. Macro neoclassical economic theory attempts to explain labor migration with the wage differential caused by variations in supply and demand between origin and destination countries, while micro neoclassical economic theory focuses on individual decision-making in which a rational actor weighs the costs and benefits of migration (Massey et al. 1993). For economists and demographers using a rational decision-making approach to migration, these costs and benefits have been almost exclusively economic ones. The new 13 economics of migration pivots from the individual to the household unit as the primary location of decision-making in order to diversify risk and increase exposure to favorable markets as a whole unit, thus characterizing migration as a family-unit strategy to increase household resources (Stark and Bloom 1985). In short, these theories are all making claims about a decision to maximize life chances through migration. Although many demographic studies of migration have focused on migration as a strategy to maximize economic well-being, some case studies have emerged to consider other aspects of life chance maximization, particularly when conflict occurs in the sending or origin location. Tolnay and Beck (1992) find that, in the early 20th century, lynching significantly contributed to the out-migration of the black population from southern U.S. states. The authors use county-level data for a sample of southern states, and model the net out-migration of blacks as a function of the frequency of lynchings. They find that blacks were more likely to leave counties where they were denied education, and where they were economically disadvantaged (Tolnay and Beck 1992). Importantly, the authors conclude that while white lynchings did occur, and while white migration did occur, the effect was not significant. Thus, “[t]he threat of violence was salient only for blacks in the South, and represented a motivation for migration only for them,” (1992:112). This evidence encourages the argument of this research, that ethnicity will be salient for minority ethnic groups in Russia, but not necessarily motivate the adaptive actions of the ethnic Russian majority group. Conflict, more generally speaking, has been shown to contribute to out-migration. Some of the highest rates of emigration in the world have been attributed to conflict. Between 1990 and 2003, amidst armed conflict with Russia, Georgia lost an estimated 20% of its residents to out- migration (Gerber and Torosyan 2013). This stream of migrants is similar to what the U.S. South 14 experienced in black out-migration, mentioned above. Likewise, in the months after gun battles associated with a larger conflict in Nepal, researchers have found a 67% increase in the likelihood of out-migration (Williams et al. 2012). These authors address cases of armed, state- level conflict. However, studies of the effect of lynching in the American South or hate crimes in contemporary Russia, present unique challenges. The events are rare, and it is difficult to interpret the threat of an individual event net of other contextual factors, although that is precisely what our statistical models ask of the data. Further, the data I use include hate crimes and nationalist vote share, both of which operate at a lower intensity than a state-level armed conflict. Research has offered support to the notion that non-violent, non-national level events may affect migration. Work on nationalist policies geared to provide economic advantages to Francophones over the previously dominant Anglophones has shown the effect of such policies on out-migration of Anglophones from Quebec to other provinces of Canada (Pettinicchio 2012). The characteristic of speaking English was a significant predictor of migration, as was the interaction of being an English speaker and a professional, providing evidence that ethnonationalist policy that provided economic advantages to French speakers also contributed to emigration of Anglophones. Researchers used a 2005 survey to investigate differences in propensity to out-migrate among Asians and Europeans in Kyrgyzstan. In justifying their focus on ethnicity, the authors note “because the collapse of the Soviet Union brought to the fore and rearranged ethnic identities and because ethnicity has been a major factor in post-Soviet migration, [they] focus in particular on ethnic differences in migration intentions,” (Agadjanian, Nedoluzhko, and Kumskov 2008:621). Noting the rarity of ethnic violence, the authors find few differences among ethnic groups in their propensities to migrate. Finally, a qualitative study in 15 post-Soviet Uzbekistan investigated the political and economic motivations of out-migration (Radnitz 2006). The author finds that the forces leading to migration exist not in the political factors alone (such as discrimination, insularity, or ethnic violence) but in their interaction with economic factors. Thus, when faced with both a fledgling economy and ethnic tensions, Uzbeks were more likely to emigrate. Data & Methodology Does nationalism contribute to ethnic minorities’ propensity to migrate in contemporary Russia? In order to investigate this question, I use panel data from the Russia Longitudinal Monitoring Survey 2009-2012, combined with regional data from the Russian Census and the SOVA Center for Information and Analytics. The Russia Longitudinal Monitoring Survey– HSE, is conducted by the National Research University Higher School of Economics and ZAO “Demoscope” together with Carolina Population Center, University of North Carolina at Chapel Hill and the Institute of Sociology RAS, and is henceforth referred to as the RLMS4. The RLMS is considered the only nationally representative survey in Russia. This survey has been repeated annually since 1994. Figure 1 maps the 25 sampled areas of Russia’s 87 regions5, stretching over multiple economic regions. [Figure 1 about here] 4 The RLMS can be accessed at http://www.cpc.unc.edu/projects/rlms-hse 5 Russia’s regions, commonly referred to as ‘oblasts’, are federal subjects officially called ‘oblasts’, ‘krai’, ‘republics’, and ‘federal cities’. With the exception of the federal cities Moscow and St. Petersburg, regions are roughly equivalent to U.S. states. They vary in size, population, and autonomy. 16 Data are collected at both the household and individual level, with a unique household and ‘site’ identified for each observation. Household-level variables include a detailed roster, reasons for members to have left the household, income and expenditures, and home ownership (among many others). Variables at the individual level include birthplace, self-identified ethnicity, previous migration experience, language(s) spoken, religious practice, and labor force participation (among many others). For each year, I merge the household and individual files, matching each individual to her corresponding record in the household roster. Generally speaking, not all members of the household will be in the individual file. This requires that some members of the household receive an ethnicity indicator based on the other members of the household. This decision relies on the important assumption that while only one individual in the household may be an ethnic minority, even an inter-ethnic married couple will interpret anti- minority sentiment as a threat to their future prospects. Thus, theoretical justification may be made for similar migration decision-making and heightened ethnic awareness for households in which even one ethnic minority resides. Using this method, approximately 19,000 observations with missing ethnicity information are deemed ethnic Russians, roughly 2,600 are deemed ethnic minorities, and a little over 2,000 cannot be determined, and are dropped from the analysis. I have structured the data in a person-year file in which individuals are nested within households within regions. The RLMS provides repeated observations of households and individuals over time, in which respondents may enter or exit the dataset at any time due to new sampling that year or to non-response. The repeated household roster is a significant asset to the RLMS. Interviewers surveying a household in the first year in which the household enters the survey take detailed notes about every person living in the household, including birthdate, gender, and relationship to others on the roster. In subsequent years, the interviewer takes the roster table 17 back to the household and updates it. Thus, if a member of a household leaves, the survey captures a general timeline of migration (i.e. if the member is present in 2009 but absent in 2010, but is present in 2011, we can make some migration duration assumptions) as well as the reason for absence (i.e. why is the person no longer in the household) and the distance the individual moved (i.e. responses to the ‘why’ question include: lives in this building, but in a different household, lives in this settlement but at a different address, lives outside of this settlement, moved for university, died, and other). This nuanced coding of individual moves was new to the RLMS in 2009. Dependent Variable I define out-migration as an individual migrating by T2 given that the independent variables are measured at T1. Moves count as out-migration if the individual moves out of their household and current population center for any reason other than to attend university. The household must exist to report the individual’s migration, and therefore, whole household migrations are not considered. The RLMS has produced few studies on migration and mobility. One exception is Guido Friebel and Sergei Guriev’s work on the effect of in-kind payments received by workers on their mobility. The authors use Rounds 6 and 7 of the RLMS and measure intent to move in Round 6 as well as actual moving in Round 7. Friebel defines a “move” as a move to a different population center or “community”. The dependent variable move takes a value of 0 if an interviewed individual in round VI lived in the same community round VII and a value of 1 if interviewers were unable to find that individual in the same community in round VII. The value of the category move=1 thus also includes nonrespondents and people who died between the two rounds, meaning that is an imperfect measure of regional mobility. (Friebel and Guriev 2005:191) 18 In another study by Yuri Andrienko and Sergei Guriev, the authors use this same measure as an estimate of informal migration to supplement aggregated official statistics in a paper on the determinants of interregional mobility in Russia. This method of operationalizing migration and mobility is demonstrative of the limitations of the early rounds of RLMS, which the authors use for its inclusion of the variable indicating intention to move. Unfortunately, the authors using these measures of migration are also counting death and non-response. Using later rounds of the data, however, allow this research to differentiate between individuals no longer in the household due to death, university, or out-migration from the region. Many studies on the years immediately following the collapse of the Soviet Union take into account official statistics, and take advantage of the propiska, or internal passport, that has historically been a part of the story of Russian mobility since the 1930s (as well as before the revolution, in the Russian empire). The Soviet propiska system was officially abolished after the collapse, but registration continues in a few FSU countries, including Russia. In Russia, registration is necessary if a person lives in one place for 90 days or more. It is generally accepted that penalties are rarely enforced, that firms hire unregistered employees regularly, and that informal migration is significant, and according to some estimates, as high as formal migration. Internal migration is penalized at much lower rates than international migration if unregistered. This presents some challenges to studies using official migration statistics to measure interregional mobility. Explanatory Variables 1 - Vote Share 19 To each individual observation, I attach the official vote share data for the Liberal Democratic Party of Russia, which was collected from the Central Election Committee of Russia and compiled by Electoral Geography. This vote share is disaggregated to the region level, and is time-lagged so that each person-year receives the score of the most recent legislative election that occurred at least one year ago. Thus, an individual record in survey year 2011 receives the vote share for 2007, and the same record in survey year 2012 receives the vote share for the election 2011. In this way, the vote share is both time-lagged, and time-varying, although it does not change annually. I restrict my calculation of vote share to legislative elections instead of including presidential elections. Legislative votes are based more on party platforms than presidential elections, which may be more focused on individual candidates. Presidential elections in Russia in this time period are problematic to measure, as official corruption tends to be concentrated and highly public in presidential elections. In addition, with Vladimir Putin carrying the bulk of the vote share in these elections, the other candidates divvy up much less of the remaining vote share. Legislative elections allow the entrance of many more parties and have more meaningful results for smaller parties. While not ignoring the possibility of corruption during the height of Putin’s popularity, I note that much of the protest around vote rigging has occurred after 2012, which is outside the scope of this study. For the period sampled, none of the regions produced a vote share of 0% for LDPR or for turnout. Vote shares for the sampled regions are summarized in Table 2. [Table 2 about here] 20 Each sampled region has a corresponding, mutually exclusive site ID in the RLMS. Both the level of vote share and the percentage change in the share of votes captured by the LDPR contain regional and temporal variability. 2 – Hate Crimes In order to control for ethnically motivated factors outside of anti-minority sentiment, I include an indicator for hate crimes. I use annual hate crime count data from the SOVA Center for Information and Analysis, time-lagged by one year. The SOVA Center is a Moscow-based Russian non-profit organization that was founded in 2002. The center collects information about human rights violations in Russia and is funded by the Open Society Foundation, the National Endowment for Democracy, the Henry M. Jackson Foundation, the Russian Federation through the State Club Foundation, among many others. The center regularly publishes articles and books in addition to monthly counts of hate crime data, classified by region and victim identity. Systematic counts of violence were collected every month between 2009 and 2012. Table 3 shows a snapshot of the targets of hate crimes nationwide, over the last five years. [Table 3 about here] The SOVA Center classifies an incident as a hate crime if it is motivated by the target’s identity. Incidents of violence include murder, beating, stabbing, wounding, or threats of murder. Incidents of vandalism include breaking windows, arson, and graffiti. A major limitation to the SOVA Center’s statistics is gross underreporting and the inability for the SOVA Center to 21 identify minority status of many victims of apparent hate crimes. Therefore, the reported hate crime figures are very conservative estimates. 3 – Minority Status I operationalize minority status as non-Russian ethnicity. Often the terms nationality and ethnicity are used interchangeably in the Russian context. This is largely a result of imperial and Stalinist policies. Nearly always ethnicity is self-identified. This is true on the Russian census, which typically returns hundreds of ethnic group responses. In the RLMS, variables that capture ethnicity are as follows: 1 - “What nationality do you consider yourself?” Interviewers write down the answer in a blank space. 2 - “Were you born in the place of your current residence or elsewhere?” 1 – In another place 2 – In the place where I live now 7 – Doesn’t know 8 – Refuses to answer 3 - (for those who answered “in another place” for borndp) “In what republic of the former USSR were you born? I have constructed a variable called “minority” in which I give a “0” outcome to individuals who responded either that they were born in the place they were surveyed (all survey sites are in Russia), or that they were born elsewhere in Russia, and a “1” if they were born elsewhere, not in Russia or self-identify as a non-Russian ethnicity. Ethnic Russians make up the majority of the dataset, but with some significant categories of other ethnic groups. I create a household level variable representing minority households. Figure 2 shows the geographic dispersion of person- years contributed by Ethnic Russian and non-Russians in the RLMS sample from 2009 to 2012. 22 [Figure 2 about here] I have aggregated the 87 regions into their economic region counterparts. The economic regions are mutually exclusive. Ethnic non-Russians reside in every economic region and in the federal cities of Moscow and St. Petersburg, although some regions consist of a much larger proportion of minorities in the sample than others. Major cities are included, but not separately counted in the Full Sample calculation. Instead, they are counted as part of the Northwestern (St. Petersburg) or Central (Moscow) economic regions. Control Variables I control for demographic and economic conditions at the individual, household, and regional level. I calculate the respondent’s age using birth year and allow it to vary each year. The data are not restricted by a maximum age, but all children under the age of 18 are removed. I include gender in the model, with female respondents receiving a 1 and male respondents receiving a 0 on the indicator. Gender previously reported for the same individual is assumed to continue through to subsequent years. Individual respondents report their monthly income and home ownership status. However, for the vast majority of households, only one individual reports these figures. I carry these indicators forward to all members of the household. Although this means that the wages reported are not ‘real’ household wages, they are useful in generating relative income versus other respondents. For use in the model, I standardize the income figures, so that one unit increase is 23 equivalent to one standard deviation increase. This allows for the coefficient to be large enough to be legible in the model. Household ownership in the model indicates that the household in which the members reside is owned by someone living in the household, but does not differentiate between which members of the household own the home. Both income and home ownership status may vary by time. Once the household is owned, however, subsequent missing data may be replaced based on previous status in the same household. At the regional level, I include a standardized measure of the regional Gross Domestic Product (GDP) per capita in 2009. I also include a standardized measure of the regional population in 2010 according to the Russian Census. For both of these standardized regional levels, a unit increase is equivalent to one (sample) standard deviation increase. Neither variable is time varying, but both provide regional variability. I also include the year of the survey in the model to control for any period effects that may be unobserved. Table 4 summarizes the averages of these control variables for both ethnic Russians and minorities. [Table 4 about here] After merging the rounds of data, the result is a panel data set, with each row containing an individual who carries the properties of her self, her household, and her region. The total number of person-years in the data is nearly 90,000. Individuals who are gone from their households at the beginning of the case study time period are censored (dropped) entirely. Table 5 shows the migration status for each of the person-years by survey year. 24 [Table 5 about here] Reflecting the officially reported population make-up, ethnic Russians dominate the dataset. However, as Table 6 shows, approximately the same proportion of ethnic Russians and ethnic non-Russians migrate out each year. [Table 6 about here] Leveraging the longitudinal design of the RLMS, and the unbalanced panel construction described above, I estimate a discrete-time event history model (Allison 1984). I focus on a respondent’s first instance of migration, which simplifies the model and computational requirements, while also recognizing that it is impossible to distinguish a respondent who leaves every fall for a few months, returns in the spring, and leaves in the fall again and a respondent who is absent for multiple years. Thus, subsequent moves are not considered in this model. I restrict my analysis to those at risk of migrating out. Thus, all those in 2009 who are marked as absent (i.e. we don’t know when these individuals migrated out) are removed from the data. After any individual’s first migration out, they are also removed from the analysis. Right censoring occurs here if an individual does not migrate within this observation period, and no new information is available on that individual. The event history model allows for time varying explanatory and control variables. Because each observation is a person-year, indicators may change over time, as subsequent years for the same individual become additional observations. In this model, any individual may have between one and four observations, depending on when 25 they enter and/or exit the sample. Table 8 shows an example of this kind of repeated observation for an individual in the RLMS. [Table 7 about here] In the example shown in Table 7, a single individual has four observations in the data. Her age, income, home ownership status, exposure to nationalist vote share, and migration status are all allowed to vary by time. Repeated observations for each individual in the RLMS make it appealing to follow an individual and investigate their migration experience. However, repeated observations also generate statistical dependence between observations and are likely to underestimate p-values. To account for this dependence, I estimate clustered standard errors around the individual, so that standard errors are measured on the person and not the person-year. I expect to find that political factors affect ethnic minorities differently from ethnic Russians. I estimate a total of four event history models, outlined below. Each model uses a dichotomous indicator of whether or not a respondent migrates out this year as the dependent variable. I use two separate sampling frames for the models – 1) ethnic Russians only, and 2) ethnic non-Russians only. Within each sampling frame I model first the individual and household control variables (age, sex, home ownership, income, and survey year) and second the full model with regional-level variables included (GDP, Population, LDPR vote share, and count of hate crimes). 26 Findings & Discussion At top of Table 8, the correlations for ethnic Russians are coded in grey and at the bottom, the correlations for ethnic minorities are coded in white. Looking only at the bivariate relationship, LDPR vote share is negatively correlated for ethnic Russians but positively correlated for ethnic minorities. A plausible explanation for the opposing correlation signs might be that regions with an overrepresentation of ethnic Russians may be more likely to underreport hate crimes, whereas in more diverse regions, this may not be the case. We expect some variables to be correlated, such as population and GDP, for instance. [Table 8 about here] Table 9, below, shows the results from the event history models, controlling for individual, household, and economic variables. [Table 9 about here] 27 These findings support the hypothesis that the migration process for ethnic minorities in contemporary Russia is fundamentally different from that of ethnic Russians. Although both ethnic Russians and ethnic minorities are out-migrating, this outcome has a different relationship to nationalism among ethnic Russians and ethnic minorities. The regional LDPR vote share has a statistically significant positive relationship to out-migration among ethnic non-Russians, and a statistically non-significant relationship to out-migration among ethnic Russians. This implies a very different migration story for ethnic minorities from their ethnic Russian counterparts. Above and beyond regional economic conditions and individual factors, ethnic minorities are more likely to migrate when residing in a region with higher nationalist vote share. This is consistent with the argument that the nationalist vote share acts as a signal to minorities about their future prospects, which demands an adaptive response. In this context, some ethnic minorities decide to adapt by way of migration. In contrast to the migration story of individuals maximizing or diversifying their resources, the model for ethnic minorities suggests that, even controlling for economic factors, part of the migration story is the desire to protect life chances that are being threatened by the success of ethnonationalism. There is a large difference in the size of the coefficients of the nationalist vote share between ethnic Russians (.1463) and ethnic non-Russians (3.4997). I conducted additional tests that revealed that these two coefficients are not statistically significantly different. However, the small sample of ethnic non-Russians results in a large standard error for the vote share coefficient in that model. This makes it mathematically difficult to reject the null hypothesis that the coefficient for vote share is the same for ethnic Russians and ethnic non-Russians. The models share other similarities. The significance of the survey year indicator speaks to a period effect, net of all other factors. That is, with each subsequent year, the propensity to 28 migrate increases for all respondents. In both subsamples, a statistically significant negative relationship exists between age and migration, favoring younger respondents. This is consistent with existing demographic research on migration decision-making. I found no effect when adding an age-squared term to the model. In none of the models is home ownership significant, likely because of the high rates of home ownership among both subsamples. For the ethnic Russians, we find evidence of selectivity based on income, regional GDP, regional population, and residence in Moscow or St. Petersburg. This suggests that a respondent is more likely to migrate when experiencing lower income, but higher GDP and population. Respondents are much less likely to migrate if residing in Moscow or St. Petersburg. This tells a migration story of young ethnic Russian men striving to maximize their life chances through migration. The lower income but higher GDP and population suggests that men with lower income but access to resources in the community may be more likely to migrate. Respondents from extremely impoverished regions may not be able to shoulder the high cost of migration, despite the opportunity to maximize life chances. However, a different migration story emerges for ethnic minorities. While ethnic minorities are more likely to migrate while young, the gender effect disappears for this subsample. The statistical significance of income, regional GDP, and regional population also disappears. The regional factor that remains is the nationalist (LDPR) vote share. Conclusions and Implications This research has provided a difficult test for Hale’s theory of ethnic politics. Few other studies analyze the relationship between electoral processes and migration. Electoral processes and nationalist signaling are certainly less physically threatening than attacks to personal security. 29 However, these findings support the notion that electoral processes have consequences for social processes. Additionally, by using a nuanced and restrictive definition of migration, I restrict the analysis to migrants leaving their region for purposes other than university study. These are migrants poised to respond to contextual factors, without the inflationary effect of including young members of the household who leave to pursue higher education in major population centers like Moscow and St. Petersburg. The findings from this line of inquiry have implications for the way that the social scientists analyze the factors leading to migration in heterogeneous societies. In regional contexts of ethnic tension, migration decision-making may differ along ethnic lines. Including race and ethnicity indicators in the statistical models might improve research examining the causes of migration. Contextual factors may play a different role in decision-making of dominant and minority groups. Although I find no statistically significant effect of hate crimes, this indicator had tremendous statistical burden. The indicator was required to show effect after controlling for population, residence in a major city, and GDP, among other factors. It is a regionally-biased factor, and represents an act of contentious claims-making that occurs in places of political power. These places also tend to be diverse, densely populated and sources of in-migration and not necessarily out-migration6. I have argued that signals from the nationalist party conveys the idea of ethnic Russian superiority, and that ethnic minorities interpret this signal as threatening to their future prospects in a certain region. Targeted violence that appears to focus on phenotypical differences, specific ethnic groups, or particular symbols of minority status (such as attacking Muslims wearing the 6 Supplementary analysis revealed no effect of hate crimes when included as a rate, or when focusing on violence alone, without counting vandalism. Additional analyses found no effect when removing the indicator for population or residence in a major city. 30 hijab) may create acute fear within that minority group, but may not have the same aggregated effect of the signal that nationalist success sends. That is, when a specific ethnic group is targeted by violence, that ethnic group may respond7. However, when the superiority of ethnic Russians over all other groups is touted, then the signal is sent to all minorities, regardless of group identification. This underlines the importance of creating a dichotomous variable to represent minority status. Although, rightfully so, attacks against Chechens and migrants from Central Asia receive media and political attention, anti-minority sentiment against Jews, Catholics, Muslims, Westerners, Roma, and even Slavic minorities such as Ukranians and Belarussians is prolific. Thus, dividing ethnic groups not only reduces statistical power, but also forces the researcher to make too many arbitrary decisions about which minority group should rightfully fear the success of Zhirinovsky and the LDPR in their region. Equally as troublesome is the problem of underreporting of hate crimes, both for research and policy-making purposes. In 2010, the SOVA Center reported a decline in violence against ethnic minorities, noting that the decline could be a result of anti-violence measures. However, “it [had become] increasingly apparent that information about [racist violence] incidents fails to make it into the public sphere – and we learn about them not right after the crime is committed, but once the attacker receives his court sentence” (Kozhevnikova 2010). This underreporting may be occurring universally, or may be geographically biased, but with the current data, it is impossible to know. However, the correlation between LDPR vote share and hate crimes indicates that it is not an arbitrary factor, and that an explanation of ethnic minority migration 7 Subsequent analysis on specific targeted ethnic groups, such as those from Central Asia or the North Caucasus, reduced the power of the model to an unsatisfactory degree. Nearly 80 ethnic groups are represented in the survey data, and dividing the respondents in such a way dramatically reduces the statistical viability of the model. 31 decision-making might benefit from exploring more liberal operationalization of this variable, perhaps avoiding the use of official statistics. Limitations One limitation of this study is based on the structure of the data. The RLMS only has nuanced measures of migration from 2009 on, which limits the scope of the study. Additionally, individual economic factors are not ideal, and should be interpreted with care, due to the necessity of carrying forward information over an entire household. (The household survey contains the dependent variable of out-migration, as well as birthdate and gender of the household members. The ethnicity and monthly income indicators are measured in the individual interviews, and must be assumed for the rest of the household.) Future Research Nationalist vote share appears to have some effect on migration decision-making by ethnic minorities in Russia. These findings hold even when controlling for regional economic variables. I have argued that nationalist vote share is a proxy for anti-minority sentiment, and that this sentiment is interpreted by ethnic minorities as contributing to a context in which life chances are at risk. Thus, ethnic minorities interpret this context as demanding an adaption. In climates of limited political advocacy, this adaptation may be out-migration. Further research is needed to unpack and better nuance these political factors. As the measurement of hate crimes is a significant limitation to the indicator, a better representation of 32 violence against minorities should be explored. The nationalist vote share as a proxy for anti- minority sentiment certainly does not capture the universe of racist undertones in the Russian Federation. Indeed, scholars have remarked at the pace at which mainstream politics under Putin have adopted radical ethnonationalist stances (Hanson 2010). Content analysis focusing on media may allow for a more accurate measurement of anti-minority sentiment. An unexpected finding was the difference in the gender effect between ethnic Russians and minorities. Young Russian men were more likely to migrate, but the gender coefficient was not statistically significant for ethnic non-Russians. This is not likely a story of family reunification, as the only migration the data measures is that of an individual who leaves her household behind. Thus, these are not whole households moving, but rather individual women with the same propensity to migrate as individual men who are also ethnic minorities. Subsequent qualitative work may shed light on the decision-making process for minority women versus that of Russian women. 33 References Agadjanian, Victor, Lesia Nedoluzhko, and Gennady Kumskov. 2008. “Eager to Leave? Intentions to Migrate Abroad among Young People in Kyrgyzstan 1.” International Migration Review 42(3):620–51. Allison, Paul D. 1984. Event History Analysis: Regression for Longitudinal Event Data. Sage. Anon. 2000. “Zhirinovsky: Russia’s Political Eccentric.” BBC News, March 10. Retrieved July 3, 2014 (http://news.bbc.co.uk/2/hi/europe/667745.stm). Anon. 2011a. “Natsionalizm v sovremennoi Rossii (‘Nationalism in contemporary Russia’).” Levada Center, April 2. Retrieved (http://www.levada.ru/press/2011020407.html). Anon. 2011b. “Vladimir Zhirinovsky.” The Moscow Times, April 1. Retrieved July 3, 2014 (http://www.themoscowtimes.com/people/article/vladimir-zhirinovsky/434232.html). Barry, Ellen. 2010. “Russian Protests Erupt Over Soccer Fan’s Killing.” New York Times, December 12. Retrieved May 19, 2014 (http://www.nytimes.com/2010/12/13/world/europe/13russia.html?mabReward=relbias:s &contentCollection=Homepage&t=qry720%23/moscow+protest/from20101201to201012 31/&action=click®ion=Masthead&module=SearchSubmit&url=http://query.nytimes.c om/search/sitesearch/?action=click&pgtype=Homepage&pagewanted=print). Bidder, Benjamin. 2008. “Zhirinovsky’s Follies: Nuclear Threats and Busty Ladies in the Race for Second-Place in Russia.” Spiegel, February 28. Retrieved (http://www.spiegel.de/international/world/zhirinovsky-s-follies-nuclear-threats-and- busty-ladies-in-the-race-for-second-place-in-russia-a-538403.html). Conant, Eve. 2014. “Ethnic Russians: Pretext for Putin’s Ukraine Invasion?” National Geographic, May 2. Friebel, Guido, and Sergei Guriev. 2005. “Attaching Workers through In-Kind Payments: Theory and Evidence from Russia.” World Bank Economic Review 19(2):175–202. Gerber, Theodore P., and Karine Torosyan. 2013. “Remittances in the Republic of Georgia: Correlates, Economic Impact, and Social Capital Formation.” Demography. Retrieved May 22, 2013 (http://link.springer.com/10.1007/s13524-013-0195-3). Hale, Henry E. 2008. The Foundations of Ethnic Politics: Separatism of States and Nations in Eurasia and the World. Cambridge; New York: Cambridge University Press. Hanson, Stephen E. 2010. Post-Imperial Democracies: Ideology and Party Formation in Third Republic France, Weimar Germany, and Post-Soviet Russia. New York: Cambridge University Press. 34 Kappeler, Andreas. 2001. The Russian Empire: A Multi-Ethnic History. Routledge. Kozhevnikova, Galina. 2010. Manifestations of Radical Nationalism and Efforts to Counteract It in Russia during the First Half of 2010. Moscow: Sova Centre for Information and Analysis. Retrieved (http://www.sova-center.ru/en/xenophobia/reports- analyses/2010/07/d19436/). Kulaeva, Stephania, Olga Abramenko, Alexander Verkhovsky, and Simon Papuashvili. 2013. Implementation of the International Convention on the Elimination of All Forms of Racial Discrimination by the Russian Federation. Sova Centre for Information and Analysis. Retrieved (http://www2.ohchr.org/English/bodies/cerd/docs/ngos/SOVA_ADCM- IPHR_FIDH_RussianFederation_CERD82.pdf). Lee, Everett S. 1966. “A Theory of Migration.” Demography 3(1):47–57. Lohmann, Susanne. 1993. “A Signaling Model of Informative and Manipulative Political Action.” The American Political Science Review 87(2):319. Massey, Douglas S. et al. 1993. “Theories of International Migration: A Review and Appraisal.” Population and Development Review 19(3):431–66. McMann, Kelly M., and Nikolai V. Petrov. 2000. “A Survey of Democracy in Russia’s Regions.” Post-Soviet Geography and Economics 41(3):155–82. Myers, Steven Lee. 2002. “New Russian Cathedral Stymied by Interfaith Rift.” New York Times, September 10. Pettinicchio, David. 2012. “Migration and Ethnic Nationalism: Anglophone Exit and the ‘decolonisation’ of Québec: Migration and Ethnic Nationalism.” Nations and Nationalism 18(4):719–43. Radnitz, Scott. 2006. “Weighing the Political and Economic Motivations for Migration in Post- Soviet Space: The Case of Uzbekistan.” Europe-Asia Studies 58(5):653–77. Ross, Cameron. 2014. Russian Regional Politics Under Putin and Medvedev. Routledge. Slider, Darrell. 1999. “Pskov under the LDPR: Elections and Dysfunctional Federalism in One Region.” Europe-Asia Studies 51(5):755–67. Smith, Anthony. 1995. Nations and Nationalism in a Global Era. Cambridge, UK  ; Boston: Polity. Stark, Oded, and David E. Bloom. 1985. “The New Economics of Labor Migration.” The American Economic Review 75(2):173–78. 35 State Duma. 2005. On the Election of Deputies of the State Duma of the Federal Assembly of the Russian Federation. Tolnay, Stewart, and E. M. Beck. 1992. “Racial Violence and Black Migration in the American South, 1910 to 1930.” American Sociological Review 57(1):103–16. Verkhovsky, Alexander, Galina Kozhevnikova, and Olga Sibireva. 2010. Xenophobia, Freedom of Conscience and Anti-Extremism in Russia in 2009. Moscow: Sova Centre for Information and Analysis. Williams, Nathalie E., Dirgha J. Ghimire, William G. Axinn, Elyse A. Jennings, and Meeta S. Pradhan. 2012. “A Micro-Level Event-Centered Approach to Investigating Armed Conflict and Population Responses.” Demography 49(4):1521–46. Zhirinovsky, Vladimir. n.d. “A practical program for the Liberal Democratic Party.” Retrieved July 30, 2014 (ldpr.ru/party/Program_LDPR/A_practical_program_for_the_Liberal_Democratic_Party/ ). 36 Appendix - Figures and Tables Figures Figure 1 Map of the RLMS Sampling Sites in the Russian Federation 37 38 Tables Table 1. Ethnic Groups in the Russian Federation Ethnic Group Number of Residents % of Population Language Group Russians 115,889,107 79.83 Indo-European Tatars 5,554,601 3.83 Turkic Ukrainians 2,942,961 2.03 Indo-European Bashkirs 1,674,389 1.15 Turkic Chuvashs 1,637,094 1.13 Turkic Chechens 1,360,253 0.94 Caucasian Armenians 1,130,491 0.78 Indo-European Belarusians 807,970 0.56 Indo-European Kazakhs 653,962 0.45 Turkic Germans 597,212 0.41 Indo-European 39 Table 2 – LDPR Vote Share in RLMS-sampled regions Regions 2007 2011 Altai 6.45% 10.65% Amurskaya 10.13% 20.99% Chelyabinskaya 9.45% 11.77% Chuvashia 8.49% 10.67% Kabardino-Balkaria 0.41% 0.08% Kaliningradskaya 10.17% 14.10% Kaluzhskaya 8.23% 14.36% Khanty-Mansiisky AD 13.19% 22.53% Komi 11.42% 11.91% Krasnodarsky krai 8.07% 10.45% Krasnoyarsky krai 10.56% 16.99% Leningradskaya Oblast 8.64% 14.78% Lipeckaya 9.65% 14.40% Moscow city 7.14% 9.45% Novgorodskaya 9.55% 11.48% Penzenskaya 5.86% 10.12% Rostovskaya 5.36% 10.15% Saint-Petersburg city 7.48% 10.30% Saratovskaya 6.22% 7.24% Smolenskaya 11.99% 14.75% Tambovskaya 7.68% 7.09% Tatarstan 3.88% 3.48% Tomskaya 13.20% 17.85% Tulskaya 7.13% 9.21% Volgogradskaya 9.03% 13.28% ALL RUSSIA 8.15% 11.68% 40 Table 3 - Targets of Hate Crimes (Violence or Vandalism), 2009-2014 2009 2010 2011 2012 2013 Total Central Asia 132 103 49 42 62 397 Caucasus 96 48 22 18 29 213 Blacks 61 28 20 26 5 140 Arab world 2 2 5 1 10 Other Asian countries 48 21 13 5 6 93 Other "non-Slavic" 59 109 26 16 33 243 Jews 4 3 3 2 12 41 Table 4 – Descriptive Statistics8 of Explanatory and Control Variables, by Ethnicity Ethnic Russians Std. Dev. Ethnic non- Russians Std. Dev. Number of Individuals 30623 -- 4646 -- Number of Person-Years 76085 -- 11385 -- Median Age 43 17.82 45 17.69 Median Monthly Income (rubles) 10,000 19160.9 9,723 33176.19 Median 2009 GDP in Region of Residence 15,098 9649.469 15,290 8472.784 Median 2010 Population in Region of Residence 2,521,892 2,968,800 1,521,420 2,549,721 Median LDPR vote share in region of residence .0945 .0335932 .0849 .0492 Median Count of Hate Crimes in Region of Residence 4 42.75704 3 33.82214 Ethnic Russians Ethnic non-Russians Proportion in Owned Homes 91.23% 90.83% Proportion Female 57.31% 54.55% 8 Averages and proportions counted on the basis of person-years and not unique individuals. These figures can be interpreted as the average or proportion over the number of observations, and thus the average contributed to the model, instead of the raw average for individuals. 42 Table 6 – Proportion of Individuals in the Sample Migrating Out Each Year Ethnic Russian Majority Ethnic Non-Russian Minority Year Total # of Individuals Individuals Migrating Out This Year Percent Migrating Out Total # of Individuals Individuals Migrating Out This Year Percent Migrating Out 2009 3,961 0 (censored) n/a 560 0 (censored) n/a 2010 7,919 750 9.47% 1150 124 10.78% 2011 8,432 1,313 15.57% 1177 166 14.10% 2012 9,804 1,068 10.89% 1377 139 10.09% Table 7 – Example of Repeated Observation for Individual in RLMS Data Using an Event-History Model Year Individual ID Gender Age Ethnicity Income Home Ownership Vote Share Migrates Out This Year 2009 PERSON A FEMALE 32 Russian 12,000 No 11.2% No 2010 PERSON A FEMALE 33 Russian 15,000 No 11.2% No 2011 PERSON A FEMALE 34 Russian 15,000 Yes 11.2% No 2012 PERSON A FEMALE 35 Russian 16,000 Yes 14.5% Yes Table 5 – Count of Person-Years Represented in the RLMS Sample, 2009-2010 by Migration Status Year Does not migrate out this year Migrates out this year Totals 2009 11,549 0 (censored) 11,549 2010 24,274 1,152 25,426 2011 24,166 1,568 25,734 2012 25,446 1,333 26,779 Total 85,435 4,053 89,488 43 Table 8 – Correlation Matrix of variables included in the analysis Correlations with statistical significance indicators for the subsample of ethnic Russians are above the line (grey), Below the line (white) for the ethnic minority subsample. Migrates Out This Year Sex Age Std. Mo. Income Home Owned Survey Year Std. 2009 GDP Std. 2010 Pop. LDPR vote share Hate Crimes Count Major Cities9 Migrates Out This Year 1.00 -.0325* -.0730* .0017 -.0695* .0682* .0188* .0289* .0159* .0252* .0165* Sex .0047 1.00 .1230* -.0996* .0088* -.0059 .0080 .0052 -.0029 .0051 .0072 Age -.0549* .1207* 1.00 -.0136* .1758* .0043 .0156* .0262* -.0285* .0298* .0281* Std. Mo. Income .0189* -.0874* .012* 1.00 .0120* .0761* .1588* .1600* .0342* .1497* .1521* Home Owned -.0409* .0334* -.097* -.0689* 1.00 - .0129* .0245* .0615* -.0618* .0436* .0398* Survey Year .061* -.0030 -.006 .0613* -.0121 1.00 .0081* .0134* .4437* -.1034* .0020 Std. 2009 GDP .0529* .0367* .012* .1258* -.1117* .0296* 1.00 .5885* .0705* .6369* .6808* Std. 2010 Population .0396* .0013 .007 .0639* -.1600* .0326* .6755* 1.00 -.2837* .8614* .7750* LDPR vote share .0449* -.0027 .001 .0887* -.1020* .1746* .3364* .0609* 1.00 -.2640* -.2130* Hate Crimes Count .0283* .0115 .007 .0639* -.1055* - .0773* .6213* .8380* .0471* 1.00 .8356* Major Cities .0205* .0233* .010* .0606* -.0580* .0140 .6413* .7762* .0273* .8638* 1.00 9 Residence in Moscow or St. Petersburg city limits. 44 Table 9 – Results from Event History Models on Two Subsamples Model I Ethnic Russian Subsample Model II Ethnic non-Russian Subsample Coeff. CSE Coeff. CSE Ethnonationalist Factors LDPR .1463 1.196 3.4997*^^ 2.006 Count of Hate Crimes .0004 .0023 .0056 .0071 Individual and Household Factors Age -.0446*** .0030 -.0358*** .0075 Sex -.3417*** .0784 .17998 .2088 Home Owned .1622 .1360 .5227 .3762 Std. Mo. Income -.3572*** .0658 -.5585 .3655 Contextual Factors Std. 2009 GDP .0939* .0524 .0845 .1265 Std. 2010 Pop. .1611** .0766 .0793 .1887 Major Cities -.7045** .3060 -.7942 .9148 Survey Year .6762*** .0438 .2951*** .0744 Intercept -1362.39*** 87.98 -597.1734*** 149.77 *** p < 0.001 ** p < 0.05 * p < 0.10 in a two-tailed test ^^^ p < 0.001 ^^ p < 0.05 ^ p < 0.10 in a one-tailed test (for hate crimes and LDPR variables only)