Introduction
When Russia invaded Ukraine on 24 February 2022, the European Union (EU) started a race of solidarity in favour of the population fleeing the war. This represents a remarkable change in attitude regarding the EU reception policies of the last decade, demonstrating its empathy towards Ukrainians, since, at the time of the writing, the EU provides hospitality to almost six million Ukrainian refugees (UN Operational Data Portal, 2024).
Such a renewed openness catches the attention of commentators and scholars in Migration Studies, who ask about the occurrence of EU double-standards for the reception of refugees (Düvell and Lapshyna, Reference Düvell and Lapshyna2022; Traub, Reference Traub2022). Most of them notice that although a refugee is a refugee by law – meaning that the same rights should be guaranteed for all those entitled to obtain the refugee status – the attitude of the EU is different if it is a Middle Eastern and African refugee or Ukrainian one. Broadly speaking, except for the Ukrainians, in the EU there is an increased attitude of substantial refusal to welcome refugees and unwillingness ‘to express solidarity’ (Votoupalová, Reference Votoupalová2022), also encouraged by the rise of populist movements.Footnote 1 Furthermore, the reception policies put in place to welcome the Ukrainian population remain mostly applicable to this specific case, without promoting a structural change of the EU migration law and reception policy. EU countries are still largely hostile towards the arrival of non-Ukrainian refugees (Nashed, Reference Nashed2022; Spaggiari et al., Reference Spaggiari, Thompson and Papangeli2022).
In the light of this situation, we compare the Syrian and Ukrainian cases, the two largest and most recent refugee crises (Eurostat, 2023), to delve into the inconsistent stance taken by the EU towards asylum seekers from these countries. Specifically, we analyse the interventions in the European Parliament's (EP) plenary sittings and investigate two hypotheses. First, we assess whether a securitisation process (Buzan et al., Reference Buzan, Wæver and de Wilde1998: 21) is affecting the representation of the refugee as a threat in the EP context, with references to the case of Syrians. Second, we observe whether the representation of refugees as a threat can be discursively deconstructed when their reception is to be incentivised (Wæver, Reference Wæver and Lipschutz1995; Aradau, Reference Aradau2004), as in the case of Ukrainian refugees. Building on these assumptions, we explore which discursive constructions of hierarchies have driven the EU's behaviour in response to pressing migration crises.
The article is structured as follows. Section one illustrates the theoretical framework, drawing on a critical approach to Securitisation Theory (here after, Critical Securitisation Theory). Whether Copenhagen School and Securitisation Theory explain how threats can be socially constructed through language (Wæver, Reference Wæver and Lipschutz1995: 55; Buzan et al., Reference Buzan, Wæver and de Wilde1998: 26–33) and other ‘argumentative practices’ (Balzacq, Reference Balzacq, Cavelty and Mauer2009: 60; Reference Balzacq2011), a critical view to International Relations (IR) reveals the persistence of hierarchical structures in the current international order. This perspective allows us to conceptualise the representation of refugees and their reception in the EU as a potentially racialised and gendered practice.
Section two concerns the Computational Text Analysis (CTA) of interventions released in the EP and scraped from its website concerning the Ukrainian and the Syrian cases.Footnote 2 We consider the EP as the most representative institution of the widespread attitude of Member States and their populations on the issue of migration, which is key in the EU public debate (e.g., Farrell and Scully, Reference Farrell, Scully, Mair and Thomassen2011; Shackleton, Reference Shackleton, Tömmel and Verdun2018).Footnote 3 The time frame considered goes from July 2014, when the EU recorded a sharp increase in asylum-seeking application, to December 2022 and covers a corpus of 834 interventions. By means of the R programming language, a triangulation of different CTA tools is performed to substantiate the credibility and validity of our results. These are Word- and Bigram-Frequency Analysis, Term Frequency-Inverse Document Frequency test and Structural Topic Modelling. The method of CTA suits the critical theoretical angle of our investigation, demonstrating through a linguistic approach how EP plenary debates perpetuate hierarchies of power based on race and gender in the EU context.
In section three, we discuss the empirical findings and how, through a certain use and choice of words, refugees are either represented as a threat to EU security or as threatened subjects to rescue. In the case of Syrians, the representation of the threat is conveyed through stigmatising references to a violent and culturally distant population. Islam, declined in terms of the Islamic State and Islamic religion, is a recurring theme. In the case of Ukrainians, the emphasis on cultural proximity, but above all the presence of women and children as refugees reinforces European states' willingness to welcome them.
Moreover, and more importantly, our research succeeds in bringing to light two further social practices, both of which reveal stigmatising hierarchies that influence the EU's response not only in the case of Syria, but also in that of Ukraine. What emerges is both a de-personalization that helps justify some member states' anti-migration stances towards Syrian refugees, and a gender-based victimisation of Ukrainians, which instead contributes to the deconstruction of the representation of the refugee as a threat and facilitates the adoption of protection measures. Eventually, in the conclusions, we outline how our research helps expand the literature on Migration Studies while offering a significant empirical dataset on migration-related discourse at the EU level.
Towards a critical securitisation theory
In Securitisation Theory the concepts of ‘security’ and ‘threat’, are no longer objective conditions, but the results of social constructions (Wæver, Reference Wæver and Lipschutz1995, Reference Wæver2011; Buzan et al., Reference Buzan, Wæver and de Wilde1998; Balzacq, Reference Balzacq, Cavelty and Mauer2009, Reference Balzacq2011; Buzan and Hansen, Reference Buzan and Hansen2009, Reference Buzan, Hansen, Buzan and Hansen2012).Footnote 4 As Wæver (Reference Wæver2011: 471) argues: ‘the security quality does not belong to the threat but to its management’, in other words, to what is addressed as a threat to security, whether it is a real threat or not. In this process, language plays a performative role and is conceived as a securitising move (Buzan et al., Reference Buzan, Wæver and de Wilde1998: 26–33) that materialises the speaker's conceptualisation of and attitude towards an issue by means of specific lexical choices and discursive frames (Chong and Druckman, Reference Chong and Druckman2007). Nonetheless, consistent with the second generation of scholarship researching on securitisation (Stritzel, Reference Stritzel2007, Reference Stritzel2014; Balzacq, Reference Balzacq, Cavelty and Mauer2009, Reference Balzacq2011, Reference Balzacq2015), this also draws upon ‘sustained argumentative practice (that are) aimed at convincing a target audience to accept, based on what it knows about the world, that claim that a specific development is threatening enough to deserve an immediate policy to curb it’ (Balzacq, Reference Balzacq, Cavelty and Mauer2009: 60; emphasis in original). Securitisation, indeed, can be described as the process by which a legitimate referent object of security is discursively declared as existentially threatened by those in authority (Balzacq, Reference Balzacq2015). Political leaders are those who, acting on behalf of the state, often elevate threats to an urgency status in a way that is coherent with their interests and beliefs, and to legitimate their actions. This may include the adoption of exceptional measures to counter socially constructed threats, which entail and justify the suspension of founding norms or values (McDonald, Reference McDonald2008: 567).
Recently, Securitisation Theory has evolved and intertwined with Critical IR (Aradau et al., Reference Aradau, Huysmans, Neal and Voelkner2014; Vuori, Reference Vuori2014; Aradau, Reference Aradau2018), including post-colonialist (Seth, Reference Seth2011; Moffette and Vadasaria, Reference Moffette and Vadasaria2016; Rosenberg, Reference Rosenberg2019) and gender, in particular, feminist (Hansen, Reference Hansen2000; Tickner, Reference Tickner2004; Enloe, Reference Enloe2014; Bertrand, Reference Bertrand2018) theories. In a critical perspective, the notion of hierarchy is a pivotal component of politics, which is essentially determined by the choices of dominant actors. The international context is hence affected by hierarchical representations, shaped in the relationship between the Self and the Other, that ‘stratify, rank, and organise the relations not only among states but also other kind of actors as well […]’ (Mattern and Zarakol, Reference Mattern and Zarakol2016: 625). Race and gender are some of the main discriminating factors on which hierarchies in the contemporary world are based. As investigated by postcolonial and feminist studies, race and gender can be powerful tools for categorising, excluding and perpetuating mechanisms of domination and exploitation (Chakrabarty, Reference Chakrabarty2000; Zarakol, Reference Zarakol2010; Sabaratnam, Reference Sabaratnam2017; Barder, Reference Barder2021; Acharya, Reference Acharya2022). Likewise, they can impact on both securitisation and its opposite process, de-securitisation, i.e., the transition from an existential threat to a policy issue (Wæver, Reference Wæver and Lipschutz1995; Aradau, Reference Aradau2004), mirroring hierarchies within the existing international order. Prejudices related to race and gender can be therefore concomitant with the representation of a referent object as a threat to security or with defining a subject as threatened (Gray and Franck, Reference Gray and Franck2019: 278), or not considered worthy of security.
Under these premises, we compare the different discursive representations of the Syrian and Ukrainian refugees in the EU context (Wæver, Reference Wæver, Kelstrup and Williams2000; Aradau, Reference Aradau2004: 407; Alkopher and Blanc, Reference Alkopher and Blanc2017). We observe how even an international setting such as the EU, which should have been thought to have gradually overcome institutionalised inequalities, is still affected by hierarchies that are largely dependent on racialised and gendered logics.
Racialised and gendered practices affecting reception policies
The concept of race generally refers to the physical characteristics of a certain population with cultural characteristics, such as sharing a common language, history, and customs (Henderson, Reference Henderson, Manchanda and Shilliam2014, Reference Henderson2017). This is particularly intertwined with religion, which is often used as a marker of race (Nye, Reference Nye2019). As Kamal Pasha (Reference Kamal Pasha2017: 314) explains, there is an enduring relationship between religion and race, as well as a ‘religiously-code racialisation’. In the last two decades, it has largely been exemplified by the racialisation of Muslim people, which has intensified as a result of the terrorist threat following the 9/11 attacks (Cainkar and Selod, Reference Cainkar and Selod2018). In this process, race and religion are mixed to discriminate against individuals who fit the racial and religious archetypes of Islam (Ibid).
Consistent with a race-based discriminatory approach, peripheral regions, including the Middle Eastern and African countries, have been depicted by the West as the ‘house of war’ (Bilgin, Reference Bilgin2019; Zarakol, Reference Zarakol2010: 12). ‘Non-white countries’, still seen as part of the ‘underdeveloped’ world, are described as ‘deficient in essential domestic or international capabilities’ (Freeman, Kim, and Lake, Reference Freeman, Kim and Lake2022: 176). It follows that the alleged normality of a condition of war and violence outside the West becomes one of the elements that enable discrimination in terms of racial otherness. In contrast, the West, and mostly Europe, has been self-portrayed as the ‘house of peace’ and as a part of the ‘civilised’ and ‘developed’ world. In this context, the absence of war in the West is associated with ordinariness and, if conflict occurs, it is understood as a shocking event. When Russia attacked Ukraine, this was thereby experienced as an exceptional circumstance triggering a unique response to the ensuing humanitarian crisis.
Perceptions and representations of conflict can also be plagued by a gendered characterization of international politics, which in turn can have an impact on reception policies. As a critical thinking emphasises, gender is understood as a set of socially and culturally constructed variables corresponding to behavioural expectations. They imply the existence of hierarchies based on assumed or explicit preferences for men over women as well as for masculinities over femininities (Enloe, Reference Enloe2017).Footnote 5
Against this backdrop, we observe how, according to a biased representation of social reality, what is seen as the ‘worst gender traits’ is frequently exploited to blame the non-West while the ‘best gender traits’ are instead used to praise the West. In detail, non-Western and non-white countries are discursively depicted as both hypermasculine, and therefore aggressive and belligerent, but at the same time as effeminate, when the intention is to describe them as childish and primitive (Zarakol, Reference Zarakol2010: 11). Differently, the West has commonly self-assigned the role of those teaching how to behave, by adopting a patriarchal parental metaphor, as if it were the ‘father’ who is ‘[…] sometimes coercive, sometimes benevolent, but always knowing the best’ (Cohn, Reference Cohn1987: 697).
This suggests that race and gender interrelate each other and shape multi-level hierarchies. This is defined as ‘intersectionality’, in which different aspects of identity are assumed to discriminate against the same individual (Montoya and Killen, Reference Montoya and Killen2023). Nevertheless, intersectionality does not translate into an inseparability between identities, but each identity, be it race or gender, must be interpreted as a factor that can intersect with the others while remaining distinct. This allows to understand both the ambiguity and uniqueness of the dimensions of power and oppression (Hansen, Reference Hansen2000: 299). In the same vein, we consider race – including religion – and gender, to act on the perceptions and discursive representations of conflict and refugees, both as interrelated and single variables determining patterns of inclusion and exclusion by receiving countries (Stachowitsch and Sachseder, Reference Stachowitsch and Sachseder2019; Sachseder et al., Reference Sachseder, Stachowitsch and Binder2022). We investigate whether Syrian refugees are represented as a threat because of racial stigma, as non-white, non-Western and Muslim, thereby justifying policies of denial in the EP. And, if the de-construction of such representation is proceeding through the emphasis on a shared cultural background, as white, Western, and – Orthodox – Christians, which is also facilitating the reception of Ukrainian refugees as well as a rhetoric of emergency and solidarity in the EU. We then explore the implications of gender, asking whether there is an inclination to represent migrants as a threat when it comes to men, but a more welcoming approach when it concerns women. In the common imagination, women – and children – are rarely portrayed as a threat, but are instead seen as weak and helpless, particularly when they are away from a male figure. Eventually, we observe how these two expressions of discrimination, if so, intersect with each other.
Methodology: computational text-analysis
The empirical section of this article draws on CTA (Wilkerson and Casas, Reference Wilkerson and Casas2017; Benoit, Reference Benoit, Curini and Franzese2020). Advantages coming from the application of this method include ‘the ability to analyse […] complex and daunting electronic sources of text’ and systematically estimate ‘the policy positions of […] political actors’ (Laver and Benoit, Reference Laver and Benoit2002: 59). Furthermore, far from being a purely quantitative tool, CTA can help qualitative scholars ‘read critically into texts as discourses to uncover the patterns and connections of knowledge and power in the social structures that produced the texts’ (Benoit, Reference Benoit, Curini and Franzese2020: 462), consistent with the theoretical angle of this analysis.
The RStudio software is used to retrieve, pre-process and analyse the data. At first, we scrape from the EP website all plenary sittings' transcripts containing either the word Ukrain* or Syria*Footnote 6 in their title and aligning with the following temporal parameters. Both time frames start at a moment in which the matter of displaced populations seeking asylum or refuge became highly relevant to the EU. While in the case of Ukraine the EU's involvement in the refugee governance was evident since the outset of the conflict (24 February 2022), in the case of Syria, European states began to experience a sharp increase in asylum seekers only in 2014, almost three years after the outbreak of the war. That year the EU recorded an increase of 44% of the overall asylum-seeking applications, almost double compared to the year before (UNHCR, 2015). The end date instead, 31 December 2022, refers to the point in time when we decided to close our empirical investigation, considering the collected data sufficient for our analysis.
The scraping resulted in two data frames including the following information: (1) the transcripts of the interventions, (2) the languages in which the interventions are made, (3) the parliamentary groups to which the MEPs belongFootnote 7 and (4) the speakers' names. Although these variables are not included in the scope of our analysis, which considers the whole EP as a single unit of analysis, their insertion in our structured datasets' metadata could be beneficial to further research aiming at investigating inter-party/-speaker differences in migration-related discourses within the EP.
After translating all non-English interventions into English via the Google Translate software,Footnote 8 being the scope of the study limited to discourses performed around the topic of immigration, only those interventions containing the words refugee(s), asylum, (im)migration or (im)migrant(s) are kept for the analysis.Footnote 9 This list of keywords, adapted from Gabrielatos and Baker (Reference Gabrielatos and Baker2008), is constructed to consider different linguistic forms referring to in-coming human mobility.
The final datasets consist of a 355-MEP-intervention corpus on Ukraine and a 479-MEP-intervention one on Syria. The interventions count 301 and 236 words on average, respectively. Text pre-processing includes lowercase conversion and the deletion of punctuation, numbers and English stop words, as well as lemmatisation, i.e., the process of reducing words to their base or root form (lemma). Further manual custom-word removal, aiming to exclude from the analysis frequently recurring words involves: words with fewer than three characters, the possessive case (‘s), modal verbs (may, must, shall, have to, need, will), party names (PPE, S&D, Renew, etc.), speaker's qualifications and roles (mr, mrs, president, commissioner, etc.), forms of address and courtesy (ladies and gentlemen, dear, honourable, etc.).
Once these preliminary steps have been taken, absolute word and bigram frequencies are calculated for each of the two cases, to have an overall view of the most significant words and phrases employed in debates on Ukraine and Syria. Furthermore, a Term Frequency-Inverse Document Frequency (TF-IDF) test is conducted to highlight the most relevant and exclusive words employed in each of the two cases (Scott, Reference Scott2021). The test is performed only to those words in the Ukraine and Syria datasets that match the English dictionary provided by the hunspell R package for spell checking (Ooms, Reference Ooms2023). This is done to mitigate the impact of particular lexical items related to self-evident contextual differences between the two scenarios (e.g., names of politicians, political organisation and geographical references) and emphasise the exclusivity role played by words belonging to a common lexicon. The hunspell dictionary is only applied to the TF-IDF test, as the interpretation of its results solely rely upon word exclusivity, as opposed to the following topic modelling analysis, whose interpretation requires a deeper data examination that goes beyond the exclusive occurrence of terms.
The core of the study is represented by a Structural Topic Model (STM) analysis, a text-mining tool designed to provide a most-likely totality of topics covered in the EP, to have a more comprehensive and precise look at the ‘semantic aspects underlying correlations between words’ (Ponti et al., Reference Ponti, Tagarelli, Karypis, Elomaa, Hollmén and Mannila2011: 248) in our datasets. MEPs' interventions constitute the units of analysis of our STM. By clustering similar words – i.e., words that share a high probability of co-occurring in specific texts – STM explores latent semantic structures within a corpus and, in addition to other topic modelling techniques, includes specific document-level metadata as covariates – the cases of Syria and Ukraine in our instance – in the statistical analysis (Roberts et al., Reference Roberts, Stewart, Tingley, Lucas, Leder-Luis, Gadarian, Albertson and Rand2014; Roberts et al., Reference Roberts, Stewart and Airoldi2016; Roberts et al., Reference Roberts, Stewart and Tingley2019). To evaluate our model, prior to the analysis, a number K of expected topics is also defined by means of a trade-off method (Mimno et al., Reference Mimno, Wallach, Talley, Leenders and McCallum2011; Silge, Reference Silge2018) between word exclusivity – the unlikelihood of the most frequent words in one topic to overlap with those in another – and semantic coherence – co-occurrence of high-probability words in the same topic. By plotting all the results from the model testing a range of Ks from 10 to 50 (Figure 1), 21k, 22k, 28k, 29k and 36k seem to be the numbers of topics showing the best balance between high coherence and high exclusivity, with 28k performing slightly better than the others.
Figure 1. Coherence-exclusivity trade-off for K = 10 to 50.
A STM model is trained for each of these K values, and the corresponding top high-probability and frequent/exclusive words are stored (Figures 7–16 in the Annexes). The similarities among top-word figures for different Ks confirm the robustness of the model. We can thus conclude that the results of the STM analysis at K = 28 are reliable and can be interpreted with confidence.
Finally, the interpretation and labelling of the 28 topics triangulates an analysis of the highest-probability and most frequent and exclusive words with the use of the findThoughts function from the STM package, which displays the most representative interventions for each topic (Table 1 in the Annexes).
Results and discussion
The results of the analysis are presented and discussed so as to provide a convincing semantic interpretation of the MEP's discourses, first at a word level by taking into account relative – to the two cases of Ukraine and Syria – frequency and relevance of single words, then at a phrase/relational level by contextualising words with respect to their linguistic environment. The former task is performed by means of word-frequency (word clouds in Figure 2) and TF-IDF (Figure 3) analysis, the latter by means of bigram-frequency (Figures 4 and 5) and STM (Figure 6) analysis.
Figure 2. Word-frequency word clouds.
Figure 3. TF-IDF analysis.
Figure 4. Bigram frequency – Syria.
Figure 5. Bigram frequency – Ukraine.
Figure 6. 28k-STM topics and case proportion per topic.
First and foremost, given their exclusiveness, words referring to terrorism (jihadist, extremism) and religion (religious) hold a prominent place in the Syria corpus (Figure 3), whereas gender- and family-related terms (child, woman, abortion) are both frequent and exclusive in the case of Ukraine (Figures 2 and 3). These findings suggest how relevant such topics are in relation to the EP refugee-related agenda. Also, both databases include several terms that indicate a widespread recognition of the refugee-status by the EP to people fleeing from both conflicts. These embrace support, humanitarian and help, figuring among the most frequent in both cases; solidarity and protection in case of Ukraine; aid, security in the Syrian one (Figure 2).
Lastly, the term war, used in both contexts, comes nonetheless with sanction and vulnerability in the Ukraine dataset, and with peace, solution and reconciliation in the Syria one (Figures 2 and 3). Although being all of them semantically related to a situation of conflict, the former two terms focus on the image of Ukraine as a victim and on the implementation of measures to counterattack, whereas the latter directly refer to a termination of the conflict and, perhaps, of the migration crisis without explicitly taking sides. In this regard, the nature of each conflict might also have a key impact on defining the EU reception attitude towards refugees. Having the Ukrainian war sprung from a clear aggression by Russia, the distinction between offender and offended, and therefore the recognition of the targets worthy of solidarity, might have been easier to draw. By contrast, since the Syrian war takes the form of a civil war, in which the roles played by different actors cannot be identified on the basis of their nationality, the reception of the injured party might have been precluded by its unclear definition.
Several considerations made at a word level, are confirmed by the frequency distribution of bigrams (Figures 4 and 5) including the terms refugee(s), asylum, (im)migration and (im)migrant(s) and by the STM analysis (in Figure 6 the proportions in which interventions from the Syria and Ukraine datasets contribute to shaping each topic). Indeed, the relevant occurrence of bigrams like refugee woman and refugee child (Figure 5), as well the key role played by topic 13 about the crimes perpetrated against children and women (Figure 6) in the Ukraine dataset show the EP's special attention to vulnerable social groups in need for protection.
In this regard, a focus also emerges on the Russian soldiers' sex crimes to which Ukrainian women are subjected (interventions for topic 13, Table 1 in the Annexes). Such consistent victimisation of children and women, passive actors in a violence that is inflicted on them from the foreign invader, provides an argument to discursively justify the de-securitisation of Ukrainian refugees. Likewise, topic 19 about the persecution of Christian religious minorities in Syria and Iraq by ISIS (Figure 6) mirrors a certain sensitivity by MEPs towards Syrians and Iraqis of Christian faith, which contrasts with the more widespread hostility towards Islam, that is instead often associated with terrorism (topics 3, 21 and 24). Similarly, the sense of solidarity expressed towards Christians can barely be found in relation to Syrian refugees at large, often represented as part of a big anonymous crowd, only defined (Figure 4) by its custody location (camp), quantity (million, thousand, much, many) or its impetuous streaming (wave, flow). Further evidence of the de-personalisation of Syrian migrants can be found in the lack of linguistic elements suggesting any social subdivision, as it happens with vulnerable groups in the case of Ukraine (e.g., refugee child, refugee woman, asylum migration).Footnote 10
The lexical recognition of migrants' refugee status and need for help found in both cases at a word level, appears more explicitly phrased with respect to Ukrainians, considering the linguistic context and the semantic relations among words. This is exhibited by bigrams as help refugee, war refugee, welcome refugee, asylum migration, support refugee, refugee flee, accept refugee (Figure 5), which can all be found in Ukraine-related documents, as opposed to the presence of illegal immigrant among the most frequent expressions in Syria-related interventions (Figure 4). Moreover, topic 10 on refugee reception (Figure 6) also touches upon the activation of the Temporary Protection Directive for Ukrainian citizens. Although this measure had existed since 2001 and could have potentially applied to other situations, it was activated for the first time only on 4 March 2022, by unanimity, in reaction to the Ukrainian refugee issue. This remarks, once again, the strong sentiment of solidarity by all EU member states and institutions, given the exceptionality of this situation in comparison to the past. Such exceptionality is lexically expressed also in the Syria database, by mentioning a refugee/migration crisis (Figure 4) much more consistently than in the case of Ukraine (Figure 5) and, thus, framing the Syrian diaspora more as a European problem to solve rather than a humanitarian action to undertake.
The EU's quick and supportive action in favour of the civilian population in danger is discussed in several topics that, in most cases, refer to Ukrainians (topics 6, 11, 12, 14 and 28 in Figure 6). Some of these highlight the role of Eastern European countries and Moldova in devoting their efforts and funds to welcome refugees. Such finding is outstanding if we think that among these are Poland, Hungary and Slovakia, members of the Visegrad Group, which had firmly stood against an ‘open-door policy’ and the proposal of EU migrant quotas to redistribute refugees during the 2015/16 migration crisis (Szalai et al., Reference Szalai, Csornai and Garai2017).
Finally, although war is equally condemned in both cases (topic 1) in the name of territorial integrity, the STM analysis also shows, as anticipated by the word-frequency analysis, how differences in the framing of refugees run parallel to dissimilar conceptualisations of war. In the case of Ukraine, the EU's political stance against Russia and its president Vladimir Putin is clearly expressed (topic 27), whereas the only topic directly touching upon war in the Syria database refers to a ceasefire (topic 9), which implies a lesser involvement and a more neutral perspective on the part of Europe. A representative example of this point is provided by topic 28, where the dramatic nature of the situation is expressed on the one hand by displaying horrific images of war in the interventions about Syria, and on the other by proposing actions – sanctions to Russia, reconstruction of Ukraine, and so on – to undertake (interventions for topic 28, Table 1 in the Annexes). This contrast between a passive attitude and an active one, along with the greater focus on the refugee status of Ukrainians, suggest hierarchical differences in the conceptualisation of war. In other words, the EP seems to consider the Russian full scale invasion of Ukraine more warlike than the Syrian one and, therefore, considers the displacement of the Ukrainian population as a more direct outcome of war.
Besides confirming some points from the word-level analysis, the relations among words in our datasets leave room for further observations. First, and in connection with our last remarks on EU's solidarity towards Ukraine, the STM shows how the EU support is not only expressed in terms of refugee reception, but also in protecting and promoting Ukrainian culture (topic 15). This implies a sense of cultural proximity between Europe and Ukraine, which strengthens the bonds between the two and, therefore, EU's moral obligation to get involved in the preservation and reconstruction of Ukraine.
On the other hand, one of the most relevant topics concerns the EU's errors and powerlessness in relation to the Syrian war (topic 20). As opposed to the European unity and partisanship widely expressed in the case of Ukraine, this topic includes several interventions criticising either some EU member states for exacerbating the situation or European institutions themselves for playing a subordinate role in Syria compared to other international actors, such as the USA, Russia and Turkey (interventions for topic 20, Table 1 in the Annexes).
Nonetheless, despite the relevance of these self-blaming positions in relation to the outbreak and evolution of the Syrian conflict, no reference to the lack of effective measures by the EU to support Syrian refugees is particularly evident in MEPs' interventions. To this regard, although violence and human rights' violations are consistently mentioned also in the case of Syria (topics 7, 8, 19, 23, 24, 26 and 28), the only topic about helping refugees, topic 18, revolves around the EU Regional Trust Fund in Response to the Syrian crisis and the Emergency Trust Fund for Africa. The former aims at addressing ‘the critical needs of 5.6 million Syrian refugees’ (European Commission n.d.), the latter deals with ‘the root causes of instability, forced displacement and irregular migration’ (EU Emergency Trust Fund for Africa n.d.), thus including refugee and non-refugee migrants into a single discussion. A blend of issues of this kind fails in recognising and giving right to differences between forced and voluntary migration and among different pull and push factors, thus depriving Syrian refugees of their legal recognition in the international law. Moreover, by treating different policies and fundings targeting different types of migrants altogether, the EP shows greater concern about controlling human displacement rather than saving human lives. This attitude is confirmed by the prominence held in the Syrian dataset by topics related to the management of migration flows (topic 16) and migrants' weaponisation against the EU (topic 22).
Conclusions
The article investigates the discursive devices triggering different representations of refugees by comparing MEPs' interventions about Ukrainian and Syrian diasporas to the EU. This allows to assess the existence of implicit hierarchies in the international order, which are reiterated within the EU political institutions. In agreement with previous studies on refugee-related media narratives (McCann et al., Reference McCann, Sienkiewicz and Zard2023) our analysis shows how Syrian and Ukrainian refugees are differently portrayed within the EP, by means of specific lexical choices linking the semantic sphere of migration to race and gender. The importance of our contribution lies in the fact that this would demonstrate a re-actualised racist and gender discourse (Ibrahim, Reference Ibrahim2005: 164 in Moffette and Vadasaria, Reference Moffette and Vadasaria2016: 4), which could even lead to ‘drawing and (re)establishing boundaries’ (Jackson, Reference Jackson2006: 16) of what has hitherto been deemed acceptable and unacceptable in a human rights-conscious context such as the EU.
In doing so, we interconnect a theoretical framework based on Critical Securitisation Theory with a methodology relying on Computational Text-Analysis in a novel and, hopefully, promising way. Specifically, our main findings demonstrate a stronger focus on race-related topics in the case of Syria, mainly through discursive references to Islamic religion and terrorism. In this respect, we emphasise a de-personalisation of Syrians, represented as a mass of migrants, which serves to de-humanise the refugee (Bleiker et al., Reference Bleiker, Campbell, Hutchison and Nicholson2013) and de-legitimise their asylum-seeking status expressed by a semantic blend of refugees and non-refugee migrants. This is opposed to the emphasis on Ukraine's cultural proximity to the EU and on the need to protect harmless women and children. A spirit of unanimous solidarity prevailed within the EU countries, encouraging the granting of unprecedented temporary protection measures to Ukrainians. Nonetheless, our analysis reveals a gender-biased victimisation of the Ukrainian refugees. This awareness becomes even more relevant when linked to what Hansen (Reference Hansen2000) called the ‘silent security dilemma’. Considering security a speech act, those who are deprived of having their voice cannot fully emerge from the condition of insecurity. The individual, in this case the woman, who is represented as a victim to protect, is indeed prevented from becoming an active subject of security policies (Aradau, Reference Aradau2008: 144).
Eventually, although it is believed to bring a valuable contribution to the available literature and empirical datasets, this research does not yet provide a comprehensive exploration of the reasons behind the EU's different approach to the refugee crises. In this regard, methodological limitations include the lack of manual validation of automated translation in the pre-processing stage and the failure to assess for possible impacts on the European attitude to refugees due to the mismatch in our datasets' timeframes, as well as the need to extend the data collection beyond December 2022. Furthermore, automated content analysis should be complemented with more consistent close readings to further validate the researchers' modelling decisions and interpretation of the outputs. Being the TF-IDF and STM bag-of-words model, they focus on word semantics, by completely disregarding other communicative – verbal and non-verbal – dimensions, such as word order, grammar, audio-visuals, intonation, proxemics, gestures, etc., which could help provide a more complete picture of how the refugee crises have been framed in EU institutions. Also, the application of STM to a smaller unit of analysis than the whole intervention (such as the sentence) could provide a more fine-grained analysis of intra-textual semantic nuances. However, our structured database of MEPs' migration-related interventions can provide a starting point to explore discursive polarisation within the EP by means of tools for extracting opinions and sentiments, estimating party/speaker effects, and observing how the discourses have evolved over time, which we believe will further enhance the value of this contribution.
Funding
This research received no specific grant from any public or private funding agency.
Data
The replication dataset is available at http://thedata.harvard.edu/dvn/dv/ipsr-risp.
Supplementary Material
The supplementary material for this article can be found at https://doi.org/10.1017/ipo.2024.10
Acknowledgments
We would like to thank the reviewers for their valuable work and suggestions. We would also like to thank Matteo Dian, who contributed to the draft of the paper from which this manuscript was drawn. We are also grateful to Michela Ceccorulli, Laura Holderied, Vera Axyonova, Marco Albertini, Luca Pinto, Aidar Zinnatullin, Andrea Knapp and Andrea Capati for their helpful suggestions and comments.
Competing Interests
The authors declare none.
Author Contribution
For the purposes of the Italian evaluation, G. Giancaspro is author of sections ‘Methodology: Computational Text-Analysis’ and ‘Results and Discussion’ and F. Lucenti is author of sections ‘Towards a Critical Securitisation Theory’ and ‘Racialised and Gendered Practices Affecting Reception Policies’. Both authors equally contributed to the introduction and conclusions.
Open access
