SEARCH WITHIN CONTENT
Citation Information : Connections. Volume 39, Issue 1, Pages 1-20, DOI: https://doi.org/10.21307/connections-2019-008
License : (CC-BY-4.0)
Published Online: 24-December-2019
This paper investigates whether the Maidan Revolution in Kyiv (late 2013–early 2014) and the ongoing armed conflict in Eastern Ukraine (early 2014) have been reflected in the collaboration networks of Ukrainian and Russian organizations in Sweden between 2013 and 2016.
I use ERG models to account for the probabilities of ties between the organizations, depending on the network structure and individual attributes such as ethnic identification and the choice of a side to support in the conflict.
Results suggest that it is support for a certain side in the conflict, and not ethnic self-identification, which drives the clustering of the networks during the most violent period.
In order to increase their influence and achieve certain other goals, most organizations tend to collaborate with other organizations that they believe share their perspectives and attitudes (Portes et al., 2008). Ethnic organizations mostly base their activities on perceptions of common “routes” and “roots” and tend to collaborate with similar others, and thus have a quite homophilous collaboration network. However, when a violent armed conflict in the homeland arises, it can be brought closer to the everyday space of diaspora through modern media and globalization processes (Baser, 2015; Brubaker, 2005; Féron, 2017; Féron and Lefort, 2019; Jabri, 2007; Oberschall, 2000). For example, the development of modern media has allowed a lot of war and conflict-related events to be available and witness-able across the geographical spaces simultaneously (Ukrainian Revolution at the end of 2013 – beginning of 2014 has been streamed by multiple channels online; Mosul battle streamed online via Facebook, and others).
A lot of ethnic/diasporic organizations in this context may mobilize their activism in order to show their support for or discontent with the events, especially if they perceive themselves as a group under attack (Oberschall, 2000). For example, Féron (2017; Féron and Lefort, 2019) discusses the case of conflict-generated diasporas that emerge as a direct response to the armed conflict in the home country. In addition, Baser (2005) looks more closely at the realities of Turks and Kurds living in Germany and Sweden and compares their experiences, which lead to different outcomes for the relationships between the two groups in these specific contexts. Other research on the interconnection of war in the homeland and diasporas include multiple case studies such as Palestinian, Irish, Armenian, Tamil, Rwandan Tutsi diasporas as well as studies of intergroup relations in the country of settlement, e.g., Sikh–Muslim relations in Britain (Féron, 2017; Koinova, 2016; Moliner, 2007; and others). However, this type of research is still quite scarce, while most of the diaspora studies focus mostly on the ways in which diasporas can affect the peace-building processes in their home countries, as well as the political unrest that they can be a part of in the country of settlement (Demmers, 2002; Féron, 2017).
Nevertheless, all this research, among others, show that an ongoing armed conflict in the home country can have the potential to affect the collaboration networks, which different ethnic organizations build with each other. Thus, shared ethnicity can lose its importance in how an organization decides to form a connection with another, and shared attitudes about the conflict can become a leading mechanism in forming collaborations. In other words, an ethnically homophilous collaboration network may reorient itself into clustering by attitudes toward the conflict, and thus actively choose to become homophilious based on that perceived value. An organization can, therefore, give a lesser degree of consideration to collaborate with another one solely on the basis of that organization’s claim of ethnic belonging, and choose the collaborations based on the perceived agreement on the politics in the home country. Alternatively, there might be a reconfiguration of the meaning of ethnic belonging, namely the attitudes toward the conflict may become a substantial part of identifying the potential collaborator as a “true” co-ethnic or not, and thus leading to the action of working together or rejection of any association.
In this paper, I focus on the collaboration networks of Ukrainian, Russian, and Russian-speakers’ organizations in Sweden to see the effects that their identified ethnicity and stand toward the conflict may have had on their structure. I account for the Swedish context and trace the evolution of the network, including its growth through the creation of the conflict-generated organizations, i.e., organizations that both through their name and activity description claim the war in Eastern Ukraine to be their main focus, agenda, and reason for existence. I do not claim to see the causal relationship or to distinguish the exact impact of the war on the collaboration networks since I specifically focus on the 2014 through 2016 period that saw the beginning of war in Eastern Ukraine and was the most violent period in terms of casualties. However, I suggest that there are some indications that the reflections of war have at least been present in the studied collaboration networks and thus had a significant enough impact, especially in the Ukrainian organizations’ case. Focusing on this conflict is particularly interesting since it allows us to follow its development from the early stages onwards, as well as follow the changes in the diasporic communities, in the context of the relatively high freedom of organization that Sweden provides.
I suggest that the concept of homophily is useful to understand ethnic organizations’ collaboration before the armed conflict in Eastern Ukraine started, but the model of foci of activity is more applicable for the analysis of collaborations based on their attitudes toward the armed conflict. I suggest that engaging in activities that (do not) support a certain side in the conflict might have become a focus of activity for the organizations and reorganized the organizational field along the conflict lines that later became incorporated into the identification of the other organization’s ethnic belonging. Thus, ethnicity in terms of similar ideas on one’s “roots and routes” may become less steering for the collaboration decisions than identification with a certain side in the conflict.
The current armed conflict in Eastern Ukraine can be traced back to the beginning of the Maidan Revolution in November 2013 when the protests against then President Yanukovych and the Parliament of Ukraine became large scale due to the President’s sudden decision not to sign a trade pact with the European Union. By late February 2014, there were more than a 100 unarmed protesters that were killed, after which then President Yanukovych fled the country and a new Parliament was established (UN Documents for Ukraine). In March 2014, the Russian Federation annexed the Crimean peninsula, an autonomous region within Ukraine, claiming its ethnic belonging as Russian and the “necessity of defending the Russians in Crimea” (address by President of the Russian Federation Vladimir Putin). Soon thereafter, Russian-backed rebels started an insurgency in two eastern regions of Ukraine: Luhansk and Donetsk (The World Bank: Conflict in Ukraine, 2017; OSCE statements). The most violent period of the conflict occurred in 2014 and 2015. July 2014 also saw the downing of the MH17 flight by Russian-backed insurgents in Eastern Ukraine (Crash MH17, 2014).
The World Bank Organization estimates that over 4 million people in Eastern Ukraine, specifically the Donbas region, have been directly affected by the continuing conflict (The World Bank: Conflict in Ukraine, 2017). According to a 2015 OCHA report, the number of casualties due to the conflict was 30,729, with 21,396 wounded and 9,333 killed (OCHA, 2015). By 2016, the number of those killed reached more than 10,000 (The World Bank: Conflict in Ukraine, 2017; UN Documents for Ukraine). In addition to casualties, 2.7 million people have been displaced. At the time of this writing, the official ceasefire has often been violated, and recent developments in the Ukrainian–Russian relations (Russia seizing three naval ships in the Azov sea in late November 2018) point to a new phase of the crisis (UN Emergency Security Council Meeting on Seized Ukrainian Vessels).
Sweden is a country with a high level of participation in civil society, with an astonishing number of 251,000 “civil society” organizations, out of which 156,845 are non-profit (Statistikmyndigheten SCB). It is relatively easy to register an organization if it is non-profit, including even applying for funding through the state. Sweden is quite welcoming to different types of non-violent, non-terrorist, ethnic activism, which creates good grounds for practicing one’s ethnicity and taking part in homeland politics through demonstrations and other similar activities (Baser, 2015).
In the context of the Ukrainian–Russian conflict, Sweden has been very supportive of Ukraine. One example is the multiple visits by the Swedish foreign affairs ex-ministers Carl Bildt (March, 2014) and Margot Wallström in 2017. In addition, Sweden has been offering humanitarian, financial, technical, and even police training aid to Ukraine since the war unraveled, through organizations such as Sida and Riksbanken (The National Bank of Sweden) (Sveriges Riksbank and Sida: Ukraine). Moreover, within its Regional Strategy for Cooperation with Eastern Europe, launched in November 2014, Ukraine is the biggest recipient (European External Action Service – European Commission).
In 2018, there were 21,930 people born in the Russian Federation and 9,924 born in Ukraine living in Sweden (Statistikmyndigheten SCB). According to the Swedish Migration Office, of the people seeking asylum from Ukraine during the years of the conflict, only 32 applicants were approved in 2014 and just 29 in 2015 (Migrationsverket, 2017). Thus, it can be assumed that the Ukrainian population living in Sweden has not been significantly affected by migration due to the armed conflict in Eastern Ukraine. However, the economic and political situation of Ukraine has no doubt influenced the decision-making process of emigration from the country.
The history of Ukrainian and Russian populations’ settlements in Sweden is, to a great extent, speculative and mixed. Throughout the Soviet Union period, most people coming from any republic within the Union would often be counted either as Soviet or Russian. Therefore, it is not easy to describe a clear and distinct history of every population. However, some knowledge has been passed on, both through the official governmental institutions, organizations, and individual people. The information below is based on interviews with the representatives of different Russian, Russian-speaking and Ukrainian organizations, Russian and Ukrainian embassies, and Swedish Statistical Bureau.
When it comes to Ukrainians in Sweden throughout the twentieth century, one of the first bigger waves came as prisoners of the 1939 to 1940 Winter War between Finland and the Soviet Union as well as more coming during the later years of the Second World War. By the mid-1950s, the community had managed to create some organizations (Embassy of Ukraine in Sweden). Another wave of immigration to Sweden came with the collapse of the Soviet Union, and included mostly women. Thus, by 2000, there were 441 men and 1,018 women born in Ukraine living in Sweden (Statistikmyndigheten SCB). This statistics, however, has to be viewed with caution since the registration by country of birth might have been mixed up with accounts of registering Soviet Union and not a particular Soviet republic before the late Soviet Union collapse.
The first big wave of emigration from Russia in the twentieth century came with the First World War and the 1917 Revolution, following thereafter with emigration caused by the Second World War and finally, following the break-up of the Soviet Union. In the early 1990s, the Russian population in Sweden consisted mostly of women (Embassy of the Russian Federation in the Kingdom of Sweden). Thus, by 2000, there were 2,192 men and almost twice as much (4,331) women born in Russia living in Sweden (Statistikmyndigheten SCB).
Ukrainian and Russian organizations in Sweden also have interconnected, yet, distinct histories of existence in Sweden (most of this information is obtained from the interviews with representatives from the Ukrainian and Russian ethnic organizations). Many of these organizations were created before or after the Second World War, and some changed from the so-called Soviet “friendship” to independent nation-state-specific organizations. Most of them claim a very long history of existence even before they were (or formally could be) registered. Practicing activism based on cultural issues is rarely problematic in Sweden, and organizations that mostly focus on maintaining traditions and celebrations from the home country can easily apply for funding from the state to organize such activities. On the other hand, to be a completely politically focused organization or one primarily occupied with humanitarian or other aid can be limiting, in terms of funding from the Swedish state and require stronger argumentation. This is even more complex for openly political organizations (Lagar för ideell förening – Bolagsverket). Therefore, many organizations may find it easier not to register with the state, although they do exist and organize meetings and events for their members. Most of these organizations use online social media platforms, where they can freely converse and diverge from mainstream ideas and thoughts.
All in all, the Swedish context is one with a relatively high freedom of organizational engagement and expression, which makes it quite easy for the diasporic communities to organize and push their agenda, often relating to raising awareness about their homeland or their situation in the country of settlement. Therefore, studying the collaboration networks in this context is not limited by the legal or oppressive regime structures in which similar practices cannot take place.
Here, it is also important to note that diaspora organizations, although often claiming to represent the totality of the group rarely do so (Ragazzi, 2012). Most often, diaspora groups are comprised of people who have a very special connection to the idea of “homeland” and stronger ethnic identification than their average co-ethnic. Thus, this research cannot be generalized to the total populations of either Ukrainians or Russians in Sweden, but only to very specific diasporic organizations with a strong and institutionalized sense of ethnic belonging. In addition, the research opens an important arena for studies on inter-ethnic and diasporic relations in the times of war in the home country.
The idea behind the concept of homophily is often summarized by a saying “birds of a feather flock together,” famously applied by McPherson et al. (2001) and relates to the phenomenon that people tend to become friends with people who share some similar characteristics with them. Homophily is an ambivalent process that makes the flow of information faster for similar others, while at the same time implies that this flow of information is localized and not different from whatever the similar others already share (McPherson et al., 2001).
McPherson and Smith-Lovin (1987) distinguished between choice homophily and induced homophily. In their discussions, induced homophily covers the effect of group composition that is homogenous on the individual pairings with similar others. Similarly, Blau (1977) proposed that patterns of relationships including homophily are guided by relative group size and ability to gain contacts for in- and out-group. In other words, the opportunity structure within the homogeneous group/organization that a pair is in dictates that the pair is also homogenous. Thus, in this view, baseline homophily reflects the composition at large and is affected by the relative size and pool of potential contacts. On the other hand, choice homophily is an individual bias or propensity to connect with similar others (Coleman, 1958; McPherson and Smith-Lovin, 1987; Marsden, 1988). In other words, the composition at large has no effect on the homophily patterns in the group.
Multiple studies have pointed to how homophily can be stronger or weaker for different types of ties as well as different socio-demographic or behavioral/attitudinal categories within the given context. When it comes to socio-demographic categories, such as sex, gender, age, or ethnicity, studies have been variable. In the case of gender, homophily is especially interesting since the group sizes at large are almost equal. The fact that gender homophily is strongly present in different societies and groups showcases that there is an individual or structural bias, since the organizational foci are gendered, as are workplaces and other activities (McPherson and Smith-Lovin, 1987; Eder and Hallinan, 1978). Marsden (1987) after controlling for kin showed that network composition of people with whom others discuss important matters is strongly gendered. Further, Ibarra (1992) found that men have stronger sex homophilous ties than women. Moreover, women with homophilous ties received support from other women and instrumental access through network ties to men, in her study of an advertising firm (Ibarra, 1992). When it comes to age, homophily patterns depend on the type of ties studied (McPherson and Smith-Lovin, 1987). In addition, since school classes are grouped by age, a strong baseline age homophily is induced (McPherson et al., 2001). Age homophily has also been shown to persist longer, most probably due to friendships formed at a younger age (McPherson et al., 2001; Marsden, 1987, 1988).
Homophily based on ethnicity is a special case and has been explained by both contact opportunities (group size) and biases. Studies have shown that smaller ethnicized and racialized groups share more networks with majority groups (Blau, 1977; Marsden, 1988; McPherson et al., 2001). On the other hand, other studies (Marsden, 1987, among others) have shown that this pattern may be different for certain groups, where a smaller group shows a tendency for homophily despite the baseline expectation that smaller ethnic groups’ networks should include more majority group members. One explanation that these studies give for this anti-intuitive pattern is that some organizational foci are segregated by ethnicity and thus limit opportunities and create bias. Often, these overlap with social class and status. In the case of ethnic organizations, the process of defining ethnic belonging and cultural heritage becomes central and practical. These organizations are voluntary, and historically people have joined them in order to gain access to information networks and for work opportunities, among other things (Portes et al., 2008). Portes et al. (2008) write that in order to play a role in the nation-state politics on minorities, people often organize in a formal, stronger, way to exercise more power. Similarly, Ooka and Wellman (2006) in their study on ethnic groups in Toronto showed that newly arrived migrants tend to have more homophilous networks that can be explained by both ethnic segregation (of neighborhoods, voluntary organizations, language, schools, etc.) and hidden value homophily, like tastes and information. In the context of ethnic organizations, ethnicity is constantly made and maintained through various organized events and similar activities.
Scott Feld’s (1981) foci of activity model is both complementary and explanatory for the analysis. Feld suggests a theory of focused social ties based on the idea that social networks are organized through shared focus and joint activity. He states that individuals often have little choice in their association with certain foci. While some activities can be chosen by individuals who then create social networks around them (e.g., playing tennis), social foci can be better understood as social structures that systematically constrain the choices of relationship formations (e.g., only certain people play or want to play tennis). Thus, people who are tied to each other through their relations to these focal activities also tend to be homogenous in other characteristics. Feld derives three propositions from his model. First, since we meet to associate, most relationships originate in focused activities. Second, these foci are usually homogeneous. Third, if foci are homogeneous, the ties that are created there also tend to be homogeneous (McPherson and Smith-Lovin, 1987; McPherson et al., 2001).
The main basis of identification in ethnic organizations is (obviously) ethnicity, and since almost every organization’s activities evolve around the cultural heritage of the group they represent, it makes little sense for them to be heterogeneous in terms of collaborations with other ethnic organizations. Thus, unless two organizations share some similarities in their views on their “roots,” culture, or heritage, theoretically they should not have many reasons to collaborate. These similarities mostly relate to views on ethnicity and, in the case of Russian, Ukrainian, and Russian-speaking organizations, showcase a complex and specific ethnic boundary-making process. One example is Russian-speaking organizations, which can include people from almost every ex-Soviet country. Inclusion of different cultural and religious holidays in such organizations forms a pan- (often Slavic)-ethnicity that connects all the specificities. In the case of Ukrainian and Russian organizations, the question of “similarity” of traditions and culture in general has been a loaded political topic, especially during the last several years when the war unraveled. The boundary between Ukrainian and Russian identifications has shifted continuously and is usually drawn on language spoken and/or country of origin. In the case of ethnic organizations, where each represents a group of people that are homogenous, at least in terms of how they self-identify through ethnicity (as someone representing a certain culture through membership in an organization), the focus of activity is usually traditions from the “homeland,” such as dancing or celebrating religious holidays, and relates to an already established sense of ethnic belonging (see discussion of homophily). However, when the war in the homeland starts abruptly, mobilization of the sense of ethnic belonging may lead to reinventing different activities and renegotiating collaborations on the basis of the attitude toward the ongoing war. Ethnic organizations – due to their already strong connection to the idea of homeland (Jabri, 2007; Demmers, 2002; Vertovec, 1999) – may regard taking a stand as a necessary point of activity in relation to their identification with certain ethnicity. However, if it is relatively easy to assign ethnicity to an organization through the already existing name, for example, the attitude toward the war could become a more complex process that is also related to maintaining the established ethnic identification. This may require more work, and organizations might feel the pressure to become active in support of or discontent with the ongoing situation at home. Thus, the focus of their activities may shift, from primarily ethnicity-maintaining cultural events like celebrating shared perspectives on the history and traditions to relating primarily to the developments of the political situation currently ongoing at home.
Studies of ethnicity and inter-ethnic relationships usually tend to understand ethnicity as a characteristic at the core identity (Chow and Bowman, 2010). I believe taking the model of focused social ties (Feld, 1981) discussed above grants the possibility of studying ethnicity without these essentialist assumptions. It creates a theoretical possibility for understanding ethnicity and diasporic communities through action as constructed through maintenance of ethnic boundaries (Wimmer, 2008).
In the current study, I suggest that collaboration of ethnic organizations is often based on the perceptions of shared “routes” and heritage. However, in times of war in the homeland, a reorientation might take place, usually through activities such as demonstrations and different campaigns connected to the developments “back home.” In this way, the war in the homeland can become symbolically transported into the everyday of the diasporic organizations and thus become a focus of activity too. Often, in order to raise an awareness about the developments in the homeland, the best way for these organizations would be to gather as many people as they can. Hence, organize the similar others around them through activities related to the conflict that is happening thousands of miles away. At the same time, those who previously were non-political or even active at all might become mobilized to action as well. Therefore, the restructuring of the organizational field by the conflict attitudes might take place.
Put in other words, if the war in the homeland has no implication for the collaboration networks, they would probably be characterized either by same ethnicity pairs, or not dominated by ethnic identification at all (only structural network characteristics would matter for the collaborations in this specific case, such as, e.g., large and famous organizations would be more likely to receive invitations to collaborate). On the other hand, if collaborations tend to be dominated by pairs that share a similar attitude toward the conflict, this could be a potential indicator that the war in the homeland has had a shaping role in the evolution of the collaboration networks.
In this section, I aim to clarify three main hypotheses that follow from the theoretical discussion above:
There is some collaboration between Ukrainian, Russian, and Russian-speakers’ organizations during the period studied. The collaborations are not dominated by organization pairs with the same ethnicity.
This hypothesis suggests that there is some collaboration between Ukrainian, Russian, and Russian-speakers’ organizations due to shared religion or traditions as well as, in some cases, a common spoken language.
During the 2014 to 2016 period, Ukrainian organizations tended to collaborate with other Ukrainian organizations, and Russian organizations with other Russian organizations.
Organizations might have collaborated with each other along the homogenous narrative of a shared past and/ or language, similar ideas on common “roots,” etc., only within clear ethnic boundaries. This hypothesis refers to the possibility that during the Revolution and subsequent war, this line of organizational collaboration remained the same. This scenario would showcase that while ethnicity is the main focus of the organizations, the developments in Ukraine were not reflected in the processes of network clustering.
Organizations that share attitudes relating to the conflict tend to collaborate more with each other than with organizations with other attitudes during the period studied.
In this scenario, organizational field of collaboration networks have reoriented from primarily subjectively identified ethnicity to standpoints on the armed conflict in Eastern Ukraine. Thus, the third hypothesis suggests that the armed conflict in Eastern Ukraine can be regarded as a focus of activity for Russian and Ukrainian organizations in Sweden (see Fig. 1).
The organizations researched in this study do not officially help with accommodation, work, or legal issues for the newly arrived migrants. The organizations that were created in the earliest period of critical developments in Ukraine (late 2013–2014) were concerned with the protests, assessing them either positively and showing support or negatively and treating the revolution as a coup – in the latter case often connecting it to Western political power struggles and conspiracies, or nationalist organizations active in Ukraine. Later, with the beginning of the armed conflict in Eastern Ukraine, a few organizations were created to send humanitarian help, among other activities, and some were created to spread information about the political developments in Ukraine.
Organizational network data collection started in early 2017. The network data were collected retrospectively through interviews and from official Facebook pages and websites of different Russian, Russian-speaking, and Ukrainian organizations. The main sampling method was to trace each organization from the connections of the previous one until the referrals led to the same organizations that were already in the database. The criteria for actors to appear in the network were: (i) the organization is Russian, Ukrainian, Russian/Ukrainian-speaking or active in connection with the conflict in Eastern Ukraine and (ii) the organization is based in Sweden. Actors could not be a political party or a governmental agency. Some organizations based outside of Sweden were included in the data collection but are not included in the data set for this analysis. The edges in the studied networks are all positive referrals and no negative (e.g. if organization A states that they will never collaborate with organization B, they have been excluded). Since the referral tracing data collection method was employed, some specific issues about the network boundaries should be mentioned. The first criteria for appearing in the network included organizations that are somehow active in connection with the conflict in Eastern Ukraine. Some of such organizations were generated by the conflict in Eastern Ukraine, while others have a broader set of activities and included in their agenda only a few events and collaborations that had to do with the conflict in Eastern Ukraine. In the first case, the conflict-generated organizations have been included as an edge sender and receiver node. In the latter case, the organization was coded only as an edge receiver. This was done to limit the network to only those organizations that are primarily focused on the conflict or identify themselves as Russian, Russian-speaking, or Ukrainian, while including potential collaborations with organizations that have a broader set of activities and agenda overall. The main motivation for including those organizations only as edge receivers was also to account for the theoretical possibility that the antagonist organizations could be connected through these broader organizations. Therefore, if not included at all, the network structure could be seriously implicated. The cumulative data for the period from 2013 to 2016 included 352 edges between 86 different organizations. However, for this analysis, the final data set consisted of 59 organizations located in Sweden (including international ones with a chapter in Sweden) during the period 2014 to 2016.
Table 1 shows that there are 14 Ukrainian organizations and 6 Russian organizations that clearly identify as such. The category “mixed” includes organizations that have members that identify themselves as Russian, Ukrainian, or other countries that used to be part of the Soviet Union or are Russian-speaking. Some of these organizations also identify themselves as Slavic. The fact that there are more specifically Ukrainian organizations reflects the phenomenon described elsewhere (author’s other unpublished article), qualitatively, namely that the people who were not happy with the claims of neutrality of the pan-Slavic organizations could demand a clear standpoint and even leave the organization if that demand was not satisfied. Some of these people could also start their own organization. On the other hand, if an organization became more political during the war, some members might not have felt completely happy with such course of events, and leave the organization as well. Furthermore, many organizations in the data set have been created as a direct response to the conflict, while claiming no identification with a specific ethnicity.
It is important to also note here that since it has been possible to identify with a certain conflict side only after the conflict started, some of the already existing organizations had to choose whether they wanted to be neutral or not. Six out of the eight pro-Russian organizations have been created directly to acknowledge their attitudes toward the conflict. This is visible through the names of these organizations as well as their open declarations. Out of all the pro-Ukrainian organizations, only two were created during 2014 and started by organizing demonstrations to raise awareness about political developments in Ukraine during the month of the Maidan Revolution. The rest were pre-existing organizations that declared their support for the Ukrainian Revolution. An interesting case is portrayed by the fact that out of four explicitly Russian organizations, only one claims to politically align with the pro-Russian side in the conflict. However, out of the 14 Ukrainian organizations – 10 are also pro-Ukrainian. Another interesting case is that of organizations that do not claim any ethnic identification (although in most cases these are predominantly Russian-speaking) but have been created to support the pro-Russian side of the conflict.
Taking these issues into account, and the fact that out of the 59 organizations analyzed in the current research, eight organizations were created with an aim directly connected to the current Ukrainian–Russian conflict, it can be suggested that the collaboration networks have also been changing according to the developments in Ukraine.
To analyze the networks, I use exponential random graph models (ERGMs) as they provide a method for modeling the probability of tie formation simultaneously for both the node attributes, such as ethnicity or type of organization, and the structural network statistics, such as triadic relationships or reciprocity between the organizations. One of the most important features of the ERGMs is that they assume network self-organization, in other words, tie dependence on each other. In comparison to other methods used in social sciences that assume independence of the individual subjects of one another, ERGMs give a more intuitive conceptualization of social networks as based on interconnectedness of actors in them. ERGMs also allow a lot of freedom for the researcher, by allowing multiple parameters to account for in the model, both concerning summary network statistics but also actor attributes (Lusher et al., 2013). In addition, ERGMs have often been used for the analysis of organizational collaboration networks (Fischer and Jasny, 2017).
As for limitations, ERGMs require a complete network to perform well (Lusher et al., 2013). To the best of the author’s knowledge, all Ukrainian, Russian, Russian-speaking, or conflict-generated organizations active in Sweden from 2013 to 2016 have been included in the data set. However, the issues of network boundaries and inclusion criteria, as discussed above, might potentially have some implications for the parameters’ estimation of the ERGMs. In addition, since ERGMs are relatively new, in terms of development of the software, and are computationally intensive, they often lack convergence. Therefore, it is not always possible to compute every model, especially in the case of large networks (Lusher et al., 2013).
The attribute data for the nodes include information such as: type of organization (independent, umbrella, multinational), side taken by the organization in the conflict (pro-Russian, pro-Ukrainian, neutral/no explicit statement), and “ethnicity” of the organization (Ukrainian, Russian, mixed Russians and Ukrainians, other ethnicity, no connection to any ethnicity/nationality). The two attributes of most interest for the current paper are organizational “ethnicity” and organizational choice of sides in the conflict. I treat pro-Russian, pro-Ukrainian, or neutral as the three least complex attitudes toward the armed conflict in Eastern Ukraine that can be distinguished. By pro-Russian attitude, I mean following the official position of the Russian Federation Government on both the Maidan Revolution and the ongoing war. This means that organizations sharing this attitude believe that the Maidan Revolution was a coup, and that there has been a threat to the Russian population in Eastern Ukraine, which justifies Russian troops in the region, while also supporting the annexation of the Crimean Peninsula (OSCE Russian delegation statement, 2018; Address by the President of the Russian Federation from March 2014, retrieved March 2019). By pro-Ukrainian attitude, I mean following the official position of the Ukrainian Government that condemns the annexation of the Crimean Peninsula, does not believe there was a threat to the Russian population in Eastern Ukraine after the Maidan Revolution, and sees the revolution as getting rid of the corrupt government serving under the V. Yanukovych presidency (Ukraine Ministry of Foreign Affairs, 2019). By neutral attitude, I mean that either an organization explicitly stated its neutrality in the matters of the conflict or they never had any event, post, or statement about any of these events on their official web pages.
ERGMs are not suitable for the 2013 data set because it only dates back to September, and thus has a very limited amount of both nodes and edges. However, that year was not yet marked by war or numerous deaths during the protests. The most active period, as discussed in the previous sections, was between the years 2014 and 2015 and which settled down by 2016, at least in comparison to the previous years (Uppsala Conflict Data Program). Therefore, the analysis here will only concern the period from 2014 to 2016.
All the models include the term “edges,” which serves as an intercept in the ERGMs and is a baseline probability of the tie formation (Lusher et al., 2013, p. 175). All the models start with the same baseline model, which includes network statistics such as measures of reciprocity and (in)transitivity as well as geometrically weighted in- and out-degrees, which are more stable terms for controlling for degrees. They work by imposing a specific rate of decay by degree to control for the nodes with a higher degree to contribute less than those with lower degrees (capturing popularity and activity spread) (Lusher et al., 2013, p. 9; Morris et al., 2008). Every model also includes the term “intransitive,” which controls for the effects of triplets of type 111D, 201, 111U, 021C, or 030C (as per Davis and Leinhardt’s, 1972 typology) and thus relates to clustering of the networks. This term is useful for the observed data since, as will become evident later in the descriptive analysis of the networks, most cases are characterized with low transitivity indices. In sum, together with reciprocity, the intransitivity term controls for effects of connectivity of the triplets. This specification helps for model convergence and fit since it “powers” the geometrically weighted terms.
As for the node attributes that are covariates in the baseline models, these include type of organization, “ethnicity,” and the conflict side. Since some of the organizations are umbrella organizations or relatively bigger in size and fame than others, this measure captures the size of the node and thus also its attractiveness and popularity to a certain degree.
The second step adds a homophily term to test whether organizations that support a certain side of the conflict tend to form ties with other organizations that have exactly the same view on the conflict. Unfortunately, only for the 2015 models was it also possible to check the tendency of pro-Ukrainian organizations to specifically connect to other pro-Ukrainian organizations, and pro-Russian ones to other pro-Russian ones (the model could converge with this particular specification of the term, while all the other models for the other networks lacked convergence).
Finally, the third step adds a homophily term on “ethnicity” of the organizations and tests whether organizations that identify with a certain ethnicity tend to form ties with organizations with the same identification.
As will be shown with descriptive results, especially for Ukrainian organizations, since there may be little heterogeneity on the standpoint toward the conflict and ethnicity, I performed multicollinearity checks for all the models that have both ethnic identification and conflict side parameters, by running them together and separately and comparing the results. If the results were different, then the full model included both “ethnicity” and “conflict side.” If they were not different, then the best fitted model is shown.
To assess the fit of the models, I use the AIC (the smaller the better) estimate and the goodness-of-fit plots of the models given by the “Statnet:ergm” package in R software. The goodness of fit of the model is judged by the fit of degree distributions. I use p-values to assess the significance of the model parameters. Many argue that the p-value has a meaning only in relation to what it could mean in the data and to the external context of the analysis, thus having a strict approach and regarding a parameter as significant only up to 0.05 level can be unnecessary and limiting. In addition, even though the ERGM Statnet package uses p-values, the ERGMs use Monte Carlo maximum likelihood estimation, for which taking confidence intervals as significance estimators makes more sense (Lusher et al., 2013). Therefore, I regard a parameter in the model as significant if the p-value has a value of up to 0.1 (at four levels).
The period from 2013 to 2016 can be characterized by the growth of both the number of organizations and the amount of interorganizational connections for reasons discussed in the previous sections. The four networks are completely different in structure, size, and also the frequency of interactions or the number of edges (Table 1 and Fig. 2). Moreover, new nodes appear in the network throughout the different years; hence, they should be analyzed separately (Table 2).
There is little transitivity between the nodes during 2013 and 2014. In 2015, the transitivity index suggests that some sort of clustering was taking place. Similarly, the network plots (Fig. 2) seem to further indicate some clustering along the ethnicity lines, especially for the years 2015 and 2016. If we take a closer look at the network plots for all the years (Fig. 2), in all of which the size of the node is based on the node degree, the above network properties also seem to have a strong relation with the node attributes (in Fig. 2, the nodal ethnicity attributes are shown in color). In addition to the descriptive properties discussed above, the plots hint on the clustering along ethnicity lines, which is stronger starting from year 2014 in the data set.
Figure 3 shows further that most of the Ukrainian organizations in the data set are on the pro-Ukrainian side, while Russian organizations take both pro-Russian and neutral positions, and organizations that are mixed or do not self-identify with any ethnicity have a lot of heterogeneity. Moreover, the absence of pro-Ukrainian–Russian organizations as well as pro-Russian Ukrainian organizations can also suggest that the conflict’s effect on the organizational network structure and that, at least for some organizations, the conflict in Eastern Ukraine, may have become a focus of activity, or possibly something that partially defines the organizational identity.
However, it is impossible to say whether this clustering is statistically meaningful and whether it is due to the data collection, the structure of the network, the node attributes, or the intersection of all three without any statistical inference. The next section aims to test exactly this question.
I start with the results from the 2014 model, which are presented below (Table 3; Fig. A1). The best fit from all of the three steps performed in the model is shown by step two, which takes into account homophily based on the attitude towards conflict. Interestingly, none of the covariates in the model are significant. The fact that neither the network structure nor nodal attributes of interest are significant, can pinpoint to the lag in creation of organizations that were specifically pro-Russian as in comparison to those which were pro-Ukrainian. In addition, since all the steps are showing very similar AIC values, the geometrically weighted in-degree parameter’s significance and positive value in the baseline may give some information about the network. As discussed earlier, late 2013 to early 2014 saw many new organizations being created as a direct response to what was happening in Ukraine. A lot of these organizations aimed to show support or opposition toward the Revolution or, later, the developments in Eastern Ukraine. To be able to have a bigger impact, many of these organizations may have started to connect to other big organizations that were already established before and thus had more “power” in this particular field. This would contribute to the larger popularity (measured by geometrically weighted in-degree) of some actors in the network, showing that the network is centralized on in-degree. On the other hand, some organizations may have been very active in reaching out to many other organizations to make themselves known in this field. Therefore, the analysis of the 2014 network suggests that there is some potential clustering in the network. However, since neither ethnicity nor attitude toward the conflict is significant as well as reflecting that there is no homophily in the collaboration network based on shared attitudes toward the conflict – H1 is supported.
For the 2015 network (Table 4; Fig. A2), Step 2 of the models shows the best fit. Both network structure and node attributes matter for the log-odds of a tie between the organizations in 2015. The large negative value on the intercept (term, edges) means that the network is sparse. The “intransitive” term is significant and negative, suggesting the tendency for decreasing number of intransitive triplets. Interestingly, the reciprocity term is now positive and significant. These two terms taken together give indication toward the difference in the network structure and show that reciprocal connections in 2015 were more likely. The in-degree term is not significant for the network in 2015, as opposed to the out-degree. More precisely, it shows a relatively large negative value. The geometrically weighted out-degree term measures the activity spread; in cases where it is negative, it indicates that the majority of the actors in a network have similar levels of activity and thus, the network is not centralized on out-degree (Lusher et al., 2013). This suggests that in 2015, the organizations that were created in 2013 to 2014 in response to the conflict in Eastern Ukraine were already established within the structure at that point.
What is most interesting about Step 2 in the 2015 network model is the significant (at 0.1 level) homophily value on the supported side of the conflict term for pro-Ukrainian organizations, which suggests that pro-Ukrainian organizations tend to have ties with similar (pro-Ukrainian) organizations. While performing the sensitivity analysis, the fit of the model without controlling for ethnic identification was better. However, since the descriptive analysis showed that there is little variation for the Ukrainian organizations in their attitudes toward the conflict, it can be assumed that this model does capture ethnicity. However, Step 3 of the model shows ethnic homophily as non-significant, and does not converge if the control for conflict side attitude is taken out. Thus, it could arguably be understood that the attitude toward the conflict is a more steering parameter than ethnicity. Therefore, H3 is supported by the results from the 2015 network. This was the only model that managed to converge with detailed ethnicity specifications.
The model for the 2016 network (Table 5; Fig. A3) shows the best fit in the baseline. Unfortunately, due to technical issues related to the computations in the ERGMs, Steps 2 and 3 in the models for the 2016 network did not converge and thus are not presented here. After performing a sensitivity check, and to avoid multicollinearity, the better fit model in the baseline was the one using conflict side as a control, and not ethnicity. Again, the large negative value on the intercept means that the network is sparse, while geometrically-weighted in-degree being significant and positive indicates that larger organizations might be receiving more connections, and the non-significance of the negative “intransitive” term that controls for the intransitive triplets which all suggest that clustering within the network remains. The geometrically weighted out-degree term is still negative and indicates that most actors have a similar level of activity and that the network does not tend to be centralized on the out-degrees (Lusher et al., 2013).
Another significant covariate of the model is the conflict side, where the pro-Russian side has a significant and positive value. This shows that organizations that are pro-Russian tend to have higher log-odds of a tie with any other organization than those that do not take a clear side in the conflict. This may suggest that the significance of sharing a point of view with an organization that other organizations connect to does not have the same meaning anymore. One reason could be that in the year before, the definitions of the sides of support were already established and these organizations that share the same views have already connected; perhaps, it makes more sense to expand the connections, or not, to other organizations no matter what their view. Thus, H3 is supported.
The results show that the conflict in Eastern Ukraine may have become a focus of activity for many Ukrainian and Russian organizations active in Sweden. The clustering of the organizations along the conflict attitude lines is shown to be significant, especially during the most violent period of the war (2015). The pro-Ukrainian organizations seem to be more active and show a stronger tendency for homophily, especially in the 2015 network. The results also show that the side of the conflict that an organization takes might be a stronger driver to collaborating with other organizations, thus suggesting that the re-identification of the organization from only based on identification with a certain ethnicity to primarily conflict-oriented took place during the period studied. In other words, it is the attitude toward the conflict that might now define collaboration decisions, and not only the perception of similar “roots and routes.”
In addition, early 2014 saw some pro-Ukrainian organizations being created as a response to the Maidan Revolution and most Ukrainians proclaiming their political orientation (mostly pro-revolution) toward it. On the other hand, in late 2014 and 2015, a lot of pro-Russian organizations (six) were created as a response to the openly acknowledged Russian involvement, while the pre-existing Russian organizations claimed mostly neutrality. By late 2014, most of the organizations had a fixed political stance, including neutrality. These pro-Russian organizations rarely claimed to be Russian, and were mostly concerned with spreading the “truth” about the conflict in Eastern Ukraine. This has further driven and changed the field according to the conflict lines and pushed the organizations to collaborate only with those who share the same view on the conflict. More specifically, by 2015, the pro-Ukrainian organizations were more likely to collaborate with other organizations that identified as being pro-Ukrainian. Interestingly though, the pro-Ukrainian organizations were less likely to have ties and thus engage in collaboration by 2016, probably because all the pro-Ukrainian organizations had already collaborated with each other by that point. In addition, after most of the pro-Ukrainian organizations connected with each other, there were none left in the organizational field – and with no new organizations being created, by 2016 the pro-Ukrainian organizations were less likely to collaborate with others in general. On the other hand, the pro-Russian organizations tended to be more likely to collaborate with other organizations in 2016, which is interesting since they were not significantly different from the neutral organizations in 2014 and 2015. One explanation could be that many of the pro-Russian organizations became most active and established in the organizational field in the late 2014 and 2015, and therefore only in 2016 could they accumulate enough collaborations to be somewhat different from the neutral ones.
To summarize, the fact that the attitude toward conflict explains tie formation within the collaboration networks for 2015 and 2016 better than ethnicity suggests that conflict in the homeland can become a focus of activity for organizations in a third country, and in this sense rearrange the organizational field. Ethnic homophily probably present before, and based on already established relations and boundaries, loses its clustering potential when a new focus appears as an important factor for activism. This further suggests that ethnic boundary-making processes are contextual and evolving, not only in relation to the context of the country of residence but also with regard to the political developments in the home country. Similar to Féron (2017), I suggest that conflict may become de-territorialized from its geographical location using similar symbols and ideas; however, thereafter, it can become reshaped within the context of diasporic experiences and finally become autotomized from the original conflict. This can be the case with the Russian and Ukrainian organizations as they seem to have renegotiated collaboration practices from basing them on ethnic identification to verification of the attitude to the conflict of the other organizations. Furthermore, these results may indicate that the meaning of ethnicity for the studied organizations, especially those identifying as Ukrainian, has become intertwined with the perception of the conflict in Eastern Ukraine, and thus incorporating being pro-Ukrainian in a political sense, with the meaning of being Ukrainian within the multiple understandings of defining ethnicity.
Finally, although not generalizable, the results found in this study further suggest that armed conflicts can be “imported” or re-territorialized into other contexts and should be accounted for when studying ethnic groups, transnational communities, or diasporas whose “homeland” has been in an ongoing armed conflict.