The Language of Risk and Vulnerability in Covering the COVID-19 Pandemic in Swedish Mass Media in 2020: Implications for the Sustainable Management of Elderly Care

The consequences of the COVID-19 pandemic—in terms of climate, economy and social aspects—cannot yet be fully assessed, but we can already see how the pandemic is intensifying already existing socio-economic inequalities. This applies to different population groups, particularly the elderly. In this article, our goal is to identify the linguistic constructions of elderly citizens in Swedish mass media coverage of the COVID-19 pandemic in 2020 from a sociological and corpus linguistics perspective. More specifically, our aim is to explore the discursive formations of the elderly in Swedish media during the pandemic and how these formations relate to risk as well as the discursive constructions of in- and out-groups. Drawing on corpus-assisted discourse studies (CADS), inspired by discourse–historical analysis (DHA), we examine the media coverage of COVID-19 by three Swedish newspapers published during 2020: Aftonbladet, a national tabloid; Svenska Dagbladet, a national morning newspaper; and Dalademokraten, a regional morning newspaper. In this article, the news articles and their messages are considered performative to the extent that—for example, at the same time as a story is expressed—the elderly are at risk of becoming seriously ill due to COVID-19; moreover, a position of vulnerability for the elderly is simultaneously created. The result reveals that the elderly were constructed as an at-risk group, while visitors, personnel and nursing homes were constructed as being risky or a threat to the elderly.


Introduction
The COVID-19 pandemic is a multi-hazardous risk with disastrous consequences, including its compounding effects on climate-related, economic and social domains. Over the past year, we have witnessed how the pandemic has escalated socioeconomic inequalities around the world with unprecedented consequences [1]. One of the social groups that has been very much present in the communication and debate regarding pandemic risks and risk-groups are the elderly; it has been argued that there is an ageist discourse underlying how we have come to understand the pandemic, e.g., [2]. In a previous study on the development of crisis narratives and struggles over legitimacy during the first phase of the COVID-19 pandemic in Swedish news, we identified conflicting narratives in the public debate related to the elderly [3]. One such debate can be best described as utilitarianism and striving for herd immunity, and the other as viewing human beings-including the elderly-as ends in and of themselves, a kind of 'herd humanism'. The elderly, defined as an at-risk group in the pandemic, were both associated with narratives regarding herd immunity being sacrificed for the greater good and with discourses encapsulating humanism that argued for a re-inscription of values associated with ageing.
In a commentary that includes over 20 researchers from the field of ageing studies, the following question on ageism and COVID-19 was posed and answered: What does our society's response say about us? According to these researchers the pandemic has accentuated the public discourse that questions the value of elderly: The COVID-19 pandemic has accentuated the exclusion of and prejudice against older adults. The current crisis highlights a disturbing public discourse about ageing that questions the value of older adults' lives and disregards their valuable contributions to society [4] (p. 693).
Although this study did not include Sweden, there are other studies on the COVID-19 pandemic that confirm a similar situation in Sweden [5,6]. A few months after COVID-19 had begun spreading in Sweden, it became evident that Sweden-in comparison with its Nordic neighbours-had failed to protect the elderly [7]. Research has revealed that a lack of competence, hygiene routines, clear guidelines, low staff continuity, neglected resources and insufficient protective equipment exacerbated the spread of infection in institutional care homes for the elderly. These shortcomings did not solely arise during the pandemic; rather, they illuminated that elderly care has largely been de-prioritized for a long time [8]. However, inequality has not been generated by the pandemic itself; it is the already-existing unequal opportunities and positions in society that have enhanced the consequences of the pandemic [9].
Prior to the pandemic, existing research indicated that older citizens were often associated with discourses of elderly care, risk and vulnerability-not as a marker of old age, but instead as a challenge for society [10]. Putting these results into perspective, Zinn and Macdonald [11] revealed that risk reporting in print news media has changed over time and moved towards a greater emphasis on social groups in daily life (e.g., women, elderly and children) in stories. Nevertheless, simultaneously, these groups were related to less agency in linguistic expressions [11]. Thus, to better understand vulnerability management and resilience under disasters, we must investigate how everyday social groups such as the elderly are narrated about and constructed in terms of risk in public debate. To do this, we utilize Ruth Wodak's discourse, historical analysis (DHA) [12], as it is particularly well suited for uncovering discourses on inequality and discrimination [12] in quantitative corpus linguistic analyses as well [13]. Discursive analyses of risk and normalization processes are a means of uncovering how the 'language of risk,' or risk discourses delimits how something is defined or even what is possible to think and say at a certain time-for example, during the pandemic [14] (p. 1633). Risk discourses can be explored as discursive imperatives, underwritten by their accompanying values and underlying morals-for example, who is framed as being at-risk (or is 'a risk' for that matter)-is entangled with positions of subordination and decisions regarding which lives are morally worth saving and those which are disposable [15]. By beginning from discourses regarding the elderly and COVID-19 in media reports, we can capture the processes that define and represent the elderly in relation to the pandemic as well as what is considered 'natural' and taken for granted in these reports [16].

Problem and Aim
In Sweden, as in other countries, the elderly are one of the population groups that have suffered the most from the pandemic, not only because of a vulnerability to the virus itself but also due to the position of the elderly in Swedish society [5]. The concept of ageism appears to have arisen during the COVID-19 pandemic in Sweden [5] as well as in many other countries such as the US, UK and Germany [17][18][19][20][21]. Overall, the findings reveal that the elderly are viewed as victims, as fragile and an at-risk group, which are images that connect to wider social practices and discourses regarding the elderly and ageing. Ageist discourses can contribute to the exclusion and subordination of the elderly; in order contribute with knowledge regarding how ageist discourses might operate in times of crisis, this study follows up on our previous qualitative analysis of the ideological conflicts present in the framing of the pandemic with a quantitative corpus linguistics analysis of articles published in three Swedish newspapers during 2020, with a particular focus on the elderly. Our aim is to explore the discursive formations of the elderly in Swedish media during the pandemic and how these formations relate to risk as well as the discursive constructions of in-and out-groups. With the point of departure in discourse-historical analysis [12] (p. 72)-the following five research questions guide our analysis: 1.
How are elderly persons named and referred to linguistically in Swedish media? 2.
Which nouns, adjectives, verbs and adverbs are attributed to the elderly? 3.
By means of which arguments are discrimination and suppression expressed? 4.
From which perspectives or viewpoints are these arguments expressed? 5.
Are the respective utterances intensified or are they mitigated?

Empirical Context: Elderly Care in the Swedish Welfare System
In order to give the study context, a short background on the Swedish welfare system and how it affects the elderly care and nursing homes follows. The pandemic has actualized questions about welfare, elderly care and nursing homes by exposing many weaknesses in the Swedish welfare system. The Swedish welfare system is supposed to care for the citizens from kindergartens, schools, health care to nursing homes or elderly care, and has in many ways replaced the function of the family to secure care and welfare. The working environment in Swedish nursing homes and in the elderly care has worsened since 1990 when Sweden adapted and began to practice New Public Management (NPM) within the Swedish welfare system [22]. NPM has also invited healthcare companies with profit interests as actors in the Swedish elderly care, and research shows that profit interests has led to a lower number of employees, lower educational level and a lower percentage of permanent employees in the care facilities [23].
The elderly care has been de-prioritized for a long time in Sweden [5,8], but the pandemic has brought them into focus: unsafe employments, lack of good and functioning hygiene routines, safety equipment and lack of competence all were reasons for the high spread of COVID-19 in nursing homes. Therefore, in 2020, the Swedish government founded a new commission called the "Corona Commission" and gave them the task to investigate the situation in Sweden during the pandemic with a focus on the elderly; the commission concludes in the report The elderly care in the pandemic (2020) that the Swedish strategy of protecting the elderly has failed. The report highlights the need for greater expertise, reasonable working conditions and more staff as crucial for the Swedish elderly care [24]. It is in the face of this situation that the results of the study should be interpreted.

Previous Research and Theoretical Framework
There are numerous studies on media reporting on the pandemic and discourses around the pandemic [25]. The reporting on COVID-19 in media uses metaphors to support the public to grasp the pandemic and its consequences. Elements of ideology and political initiatives are also part of the framing of the pandemic by the mass media. For example, war frames are common in the reporting in China [26], the United Kingdom (UK) [27] and Sweden [28]. This is confirmed and further analysed by a US study that found that newspaper coverage on COVID-19 is highly politicized, network news coverage somewhat less so, and both newspaper and network news coverage are highly polarized. The findings suggest that the high degree of politicization and polarization in media coverage may have contributed to polarization in the attitudes of people toward the pandemic in the US [18]. As in our own previous study on the public debate in Sweden [15], struggles over ideological dominance were found both in US and Chinese newspapers [29], with clear nationalist anchoring in ideology bias practiced through the selection of topics and the tone of reporting [29].
The COVID-19 pandemic has highlighted how ageist language is employed and ageist stereotypes are used to characterize older adults [19,21]. Ageist language is used even though the disproportionately negative outcomes for older adults in the pandemic partially reflect the social and economic inequalities that are manifest throughout the life course of marginalized groups. They also reflect major problems with institutional living [20]. An American study showed that media sources consistently described older adults as being vulnerable during the COVID-19 pandemic [19]. The study showed that national news sources engaged in both explicit and implicit ageism, for example by associating different words describing weakness and exposure with the elderly [19]; see also [17]. Another study reveals that classifying older adults, based on age, as a higher risk group of COVID-19 is potentially reinforcing ageism [21], which is an aspect that can accelerate the risk of social isolation and heighten the levels of psychosocial distress. Skoog [5] argues that formulations of the elderly as the primary 'at-risk group' subsequently resulted in restrictions specifically directed towards this demographic, which has created a form of ageism that stigmatizes the elderly. Thus, all people over a particular age were treated like a homogeneous group and faced with more extensive restrictions than other groups. This is also confirmed in studies of newspaper photos related to the COVID-19 pandemic [17]. A Finnish study identified different social positions for age groups: children as controlled pupils and also as happy and playful; youth as future-oriented graduates and reckless partygoers; adults as experts, professionals, caretakers and active recreationists; the elderly as isolated loners. The results correspond to the positions of villains, heroes and victims, respectively [17]. The restrictions also reinforced the stigmatized feeling of previously independent older individuals, when many of the elderly suddenly became dependent on others. Thus, the management of the pandemic caused a mental construction of the elderly as being fragile and almost the only ones who could become ill with the virus [5]. Further, the Swedish public debate in the spring of 2020 mirrored this focus on restrictions for the elderly and other risk mitigation actions that the government and the Swedish Public Health Agency had implemented at the time [15].
The Performativity of Risk, Discourse and Its Normative Implications Following Fairclough [30], we consider discourse as a representation of a certain domain of social practice from a particular perspective, and the relationship between discursive practices and the settings in which they are embedded are considered as dialectic. Thus, as Wodak [12] (p. 63) indicates, "discourses as linguistic social practices can be seen as constituting non-discursive and discursive social practices and, at the same time, as being constituted by them." Discursive analyses of risk in the public debate can uncover how certain social groups and institutions are defined in terms of risk, based on underlying values and morals. According to Hunt [31], the boundary between normative judgements of risk and objective hazards has become blurred, and a hybrid between moral discourses and discourses of risk has been created. The moral components of risk discourses imply that individuals are expected to self-regulate based on the norms of what it means to live a righteous life [31,32]. The restrictions during the pandemic are obvious examples of this, but also media reporting and everyday talk regarding the risks associated with COVID-19 behave in a performative manner. As a subject, one acts in accordance with performative risk discourses, as the work of normalization processes, where the conduct of the individual is governed through moral discourses of responsibility, a process which then masks itself by framing the conduct as the outcome of free and individual choice [33].
Within the discourse of responsible risk avoidance, behavioural differences are often considered a matter of choice [34]: if you do not adhere to recommendations, it is considered a choice you have made. The concept of the 'right choice' or the imperative to act in a certain manner in accordance with public recommendations also has the effect of dividing people according to those who are considered to be more or less at risk. It is inescapable that talk of being 'at risk' carries allusions to death-the embodiment of biological materiality, the mental and spiritual extremity, the ultimate risk. Simultaneously, ageing and the inevitable hazards of growing older form another arena in which the autonomous self is expected to behave and act wisely. Thus, growing old has itself become a social, reflexive and managerial risk in which the political domain is bound up with ideological and philosophical questions of self-governance [35,36]. A critical discourse analysis of how older people are portrayed during the pandemic can also reveal how ageist discourses interact with 'responsibilisation' and risk discourses and, thus, reinforce existing unequal social practices and structures [37].
Such a theoretical analysis, with its point of departure in critical discourse analysis, can be used to disentangle the manner in which the performativity of risk is intertwined with the processes by which age is constructed in a system, and through which risk discourses can be mobilized and used to uphold other formations such as social norms. This normalizing act often occurs along contemporary hierarchies of power-the act of naming is performative precisely because it initiates the individual into the subjected status of a subject [38] (p. 121). This implies that media news, when naming risky or at-risk subjects, brings them into being subjects for risk discourses and their underpinning values and priorities. Similarly, macro-topics such as COVID-19 interrelates such discourses [12]. For example, from our own previous study, the elderly were described in an editorial piece as having dementia and being close to death anyway [3]. Thus, we turn to a critical discourse in our study-historical analysis-to enable critical investigation and challenge what has been taken for granted thus far [37] (p. 2).

Materials and Methods
The methodological framework used in this study draws on corpus-assisted discourse studies (CADS) [39,40], which is a combination of corpus linguistics and discourse analysis. The corpus linguistics quantitative analysis is used to identify, quantify and compare the most prominent terms that occur in the coverage of the elderly during the pandemic, combined with a qualitative exploration of broader linguistic patterns related to these representations. In our previous study [3], we identified ageism discourses in a selection of new articles; here, we apply DHA in the analysis of all published articles in three newspapers during an entire year to avoid the hidden danger in qualitative analysis of unintentionally singling out manifestations of the analyst's interest [13].
The corpus consists of 26,841 articles from the year 2020 containing the word 'COVID-19' or 'Corona' from three Swedish newspapers: an agenda-setting unbound conservative morning paper called Svenska Dagbladet, with approximately 750,000 daily readers; an independent social democrat tabloid newspaper called Aftonbladet, with almost 3,000,000 daily readers; and a local independent social democrat newspaper called Dalademokraten, with 30,000 daily readers. This mix of newspapers aims to represent both national and local perspectives as well as different political directions. The number of articles was low at the beginning of the year; subsequently, it was followed by a peak in April 2020, with over 4500 articles on Corona or COVID-19. The reporting then stabilized with a second smaller peak in November 2020 (see Figure 1). The pattern is the same for all three newspapers; the statistical analysis confirms that the kind of analyses that are relevant here do not reveal any differences among the three newspapers. The first two months of reporting focused on the economic consequences of the outbreak of a new Corona virus in Wuhan, China. From March-end onward, the mass media reporting became more diverse and began focusing on domestic consequences along with comparisons of how other countries were handling the virus as opposed to the Swedish strategy. In total, the number of word tokens in the corpus is 16,487,463 and the number of word types is 272,568. The former represents the total count of words in their raw forms, while the latter illustrates the number of unique words. Using corpus linguistics and the program AntConc, we searched for how the elderly are framed in the news by analysing which words are more likely to occur in close relation to the mentioning of 'elderly'. This is an effective method when it comes to merging the qualitative and quantitative aspects of a text mass where the linguistics are of interest. The method places specific terms within a context by systematically listing each context in which a defined word occurs [39] (pp. 71-72). The search revealed five words to the left and right of the search term 'elderly' ('Ä ldre*' in Swedish). The most frequently occurring words were listed and subsequently analysed (see Table 1). The word classes we used are nouns, adjectives, verbs and adverbs; we consequently excluded non-discursive classes, such as pronouns and prepositions. This choice was made in order to follow Baker's [40] example, who suggested the exclusion of grammatical function words in this part of the process.  In total, the number of word tokens in the corpus is 16,487,463 and the number of word types is 272,568. The former represents the total count of words in their raw forms, while the latter illustrates the number of unique words. Using corpus linguistics and the program AntConc, we searched for how the elderly are framed in the news by analysing which words are more likely to occur in close relation to the mentioning of 'elderly'. This is an effective method when it comes to merging the qualitative and quantitative aspects of a text mass where the linguistics are of interest. The method places specific terms within a context by systematically listing each context in which a defined word occurs [39] (pp. 71-72). The search revealed five words to the left and right of the search term 'elderly' ('Äldre*' in Swedish). The most frequently occurring words were listed and subsequently analysed (see Table 1). The word classes we used are nouns, adjectives, verbs and adverbs; we consequently excluded non-discursive classes, such as pronouns and prepositions. This choice was made in order to follow Baker's [40] example, who suggested the exclusion of grammatical function words in this part of the process. Further, the quantitative aspect of this study is based on corpus linguistics. Corpus linguistics is a beneficial method when certain aspects of a large amount of text are of interest [40]. It is also a suitable method for CADS and DHA [13]. A number of routine processes and procedures are involved when searching a corpus in order to recover, organize and display linguistic information. The three steps that were used in this article were word frequencies, collocations and concordances [40], which also function as the framework for the analysis. The first step, analysis of frequencies, was used to identify how often the word-form that is under study-the so-called 'node'-occurs; in our case, the node is 'elderly'.
The next step of the analysis was to identify collocates, which Stubbs [41] (p. 21) defined as 'frequent co-occurrence'. Thus, in this stage, we investigate the node of interest's most frequent co-occurrence with other words. In our analysis, five words to the left and right of the node were analysed in the search of the most frequent co-occurrences. A collocates analysis using log-likelihood as the statistical measure has made it possible to analyse how the elderly are described by showing which other words appear in close proximity to the searched term. The higher the value of the likelihood, the stronger the correlation with the investigated node-that is, 'elderly'. Collocations are used in order to discover repeated or typical lexical choices that are used in media reporting on COVID-19 and the elderly. Such recurrent words are understood as reflecting practices by which communities express, interpret and evaluate the elderly, thereby indicating how they are framed in discourse. As Stubbs [41] (p. 188) suggested, we do not view collocations simply as lexical items, we also see them as having the possibility to act as 'nodes around which ideological battles are fought'.
The third step is the concordance analysis. Overall, the analysis of the frequent lexical words reveal a few of the most important concepts in the corpora, but a more detailed analysis of these lexical items in the present context is crucial. Context plays a significant role in signalling the relationships among particular words, which is impossible to achieve by merely considering word frequencies and co-occurrence alone. To understand the context of the words occurring in close relation to the elderly, a concordance analysis has been used in order to analyse the context in which the words occur [42]. Concordances are lists with lines that display all the occurrences of a search term. Following others who have used CADS, we believe that the concordance lines express social processes and phenomena. In order to investigate how they reflect social practices, we have used DHA as developed by Wodak [12], which has also been used by others in combination with corpus linguistic analysis [43].

Discourse Analytical Strategy
The specific discourse-analytical approach, discourse-historical analysis (DHA), applied in this study was first developed in order to trace the constitution of negative and discriminating images in public discourse, particularly discursive strategies to present 'us' positively and 'them' negatively [13]. Therefore, DHA is particularly well suited for analysing racist and discriminating discourses [35]. In our case, we are interested in how the elderly were framed in Swedish media during the COVID-19 pandemic and the normative effects it might have. Following DHA, the first step is to identify if stereotyped discourses are present in the investigated corpus [12]; as mentioned in the introduction, we already found tendencies of ageism discourses in the Swedish public debate on COVID-19 in a previous study [3]. The second step is to identify discursive strategies, and the last step is to examine the linguistic means and the context-dependent linguistic realizations of discrimination or, as in this case, ageism [12]. In this analysis, we place the main focus on the second step (see also [43]) to analyse the discursive strategies, as we have already identified the presence of ageism discourses in our previous study [3]. Wodak [12] (p. 73) defines a discursive strategy as "a more or less accurate and more or less intentional plan of practices (including discursive practices) adopted to achieve a particular social, political, psychological or linguistic aim". Five strategies, first practiced by Wodak [12] (p. 73), are often used in DHA, each with an objective and particular devise to perform the specific discursive strategy [37] (p. 29), [42]: (1) Referential or nomination strategy, which aims at focusing on the discursive construction of in-groups and out-groups and is realised through discursive devices such as membership categorizations, metaphors, and metonymies and synecdoches. (2) Predication strategy, which aims at labelling social actors more or less positively or negatively and is realised through stereotypical attributions of traits and implicit and explicit predicates. (3) Argumentation strategy, which aims at justifying positive or negative attributions and is realised through the use of topoi for justifying political inclusions or exclusions. (4) Perspectivation strategy, which aims at positioning the speaker's point of view and is realised through discursive devices, such as report, description, narration, or quotation of events and utterances. (5) Intensification strategy, which aims at modifying the epistemic status of a proposition and is realised through devices that intensify or mitigate the illocutionary force of utterances.
To perform the last step of the CADS, the concordances for each theme were read and analysed according to these discursive strategies, which we have translated into our five research questions: (1) how are elderly persons named and referred to linguistically in Swedish media? (2) Which nouns, adjectives, verbs and adverbs are attributed to them? (3) By means of which arguments are discrimination and suppression expressed? (4) From which perspective or viewpoint are these arguments expressed? (5) Are the respective utterances intensified or are they mitigated?
In the analysis, we read the words closest to the node in each theme to achieve a deeper and contextualized understanding of the material and attempted to answer the research questions. The following section presents the results from our analysis. In our analysis, we organized our collocates into themes based on the discursive strategies that correspond with our abovementioned research questions.

Results
The presentation of our results follows the three analytical stages of CADS: first, we provide a descriptive presentation of the node 'elderly' in the COVID-19 corpus, then the identified and analysed collocations are described, and, lastly, the analysis of the results from the concordance analysis are presented.

Frequency of 'Elderly' as a Node in the Investigated Corpus
In our corpora consisting of articles that include the word COVID-19 published in three daily newspapers during 2020, the node being investigated is 'elderly'. If we examine the occurrence of this node, the distribution is similar to the distribution of the total number of articles that mention COVID-19 (see Figure 1). It was not until March 2020 that the node occurred, but the number of occurrences increased rapidly and the reporting peaked in April 2020, with 2532 mentions of the node. During the summer months, the node occurred less frequently; however, during autumn, the occurrence of the node increased again and there was a second smaller peak in November 2020. This implies that the node 'elderly' was mentioned frequently in the articles that covered the pandemic. It is, of course, possible that one article may have several references to the elderly and another article may have none; nevertheless, elderly is an important node in the reporting on COVID-19 in Swedish mass media during the year 2020.
The relative frequency of the node in relation to the number of times COVID or COVID-19 occurs in the corpus was greatest during the peaks-April-May and November-December-and least during the summer months. The peaks in the occurrences corresponded to the two waves of infection in Sweden during 2020, which indicates that not only did a relatively large proportion of the corpus include the node but also that the mentioning of the elderly in mass media reporting followed the two waves of COVID-19 infections.

Collocations: Words That Co-Occurring with the Node 'Elderly'
Here, we turn to the second step in CADS: a collocate, or co-occurrence, analysis. We used a log-likelihood procedure as the statistical measure to analyse which other words appeared in close relation to the node elderly . We have used the 20 most frequent collocations to present the most prominent discourses regarding the elderly in the context of the pandemic. Table 1 presents the collocations of 'elderly'-first, the rank of the word, with the first word having the rank of 14. Words without discursive meaning in the context have been deleted from the table, thereby leaving only meaningful words in the analysis. The next column presents the frequency of the word and then how many times the word appears before the term 'elderly' (left of the search term) and how many times the word appears after the word (right of the search term). The threshold for an α-level of 0.05 is a loglikelihood stat of 3.8 [44] (p. 209), thereby implying that the risk of the collocation occurring by chance is very small with stats higher than 1000. Further, the last two columns present the collocation in the original language (Swedish) and the translated word in English.
Some words can easily be considered as merely terms that describe the elderly without adding any valuable information. 'Persons' and 'people' are used along with 'elderly'; moreover, 'elderly persons', 'elderly people' and the word 'COVID' do not add anything since the entire corpus addresses COVID-19 and Corona. After eliminating these words marked with an asterisk (*), 17 signifying words remain, in Table 1. The collocates are single words and to signify the meaning of the collocates; we need to dig deeper into the material by using the context around the single words-that is, the concordances.

Concordances: Contextualisation of the 'Elderly' Node
In the analysis of concordances, we turn to the discursive strategies of DHA [12], as previously described. The first step for our analysis was to organise the 17 remaining collocates from the perspective of how they appeared in the concordances in relation to the discursive strategies presented earlier; the result of that analysis is presented in Table 2 below. With regard to the referential strategy, the elderly were referred to using collocates, such as 'Sweden's, 'our' and 'the country's' elderly, which can be interpreted as discursive constructions of a national in-group. Table 3 presents typical collocates that represent 'Sweden', 'our' and the country's'. 'Our' elderly is often used to describe the elderly in the corpus and belonging to 'Sweden' or 'the country'. It is a clear in-group reference where referential strategies are used, such as the country's elderly, our elderly, the elderly in Sweden, etc. This group is also 'ours' in the sense that they need protection and our help within the framework of Swedish society. The material emphasizes how the elderly in Sweden must be protected or how Swedish society has failed to protect 'our' elderly, particularly in nursing homes. Note that the Swedish model for elderly care is strongly criticized as it has not been able to protect the elderly from infection. Table 3. Categorization of collocates according to referential strategies.

Collocates Illustrative Example of Concordance
Our Municipalities to cut further on their elderly care. Society's debt to our elderly is enormous.
The country's It is obvious that Sweden has let down the country's elderly. It has both the Prime Minister and the public health authority Sweden Should be able to wrap a protective ring around our sick and elderly in Sweden in corona times.
In the Swedish context, the invoking of 'our', 'Swedish' and the 'country's' elderly could also be interpreted based on what Barker [45] terms the duality of the Swedish welfare state, referring to its simultaneously inclusionary and exclusionary character. Sweden leans towards 'welfare nationalism', which is a form of protectionism in terms of the welfare system that involves desperately trying to sustain it by excluding others or making it difficult for them to belong [44] (p. 17). Barker [45] explains that, in Sweden, there is a pattern of excluding people who are considered 'undeserving'. As this practice has been questioned as part of the pandemic debate on whether the elderly are deserving or, as indicated in previous analysis, disposable in the struggle for herd immunity [3].
The referential strategy employed in our corpus could be understood as a strategy towards the discursive construction of an in-group, which we as individuals and Swedish society at large are responsible for caring about. Thus, the inscription of value in the elderly considers their past efforts, as 'our elderly' is a term constructed to include those who have previously contributed to society and, therefore, society has a debt to pay to them. The use of words such as 'our elderly' and 'the country's elderly' can evoke thoughts with a clearly nationalistic focus by portraying the Swedish elderly as people who built the country. This discursive device makes it possible to pit groups against each other and raise a question regarding the elderly who do not belong to the category of 'ours'.

Prediction Strategy: Which Nouns, Adjectives, Verbs and Adverbs Are Attributed to the Elderly?
With regard to the prediction strategy, the elderly were depicted as 'ill' or 'fragile', which is a rather stereotypical representation of the elderly as vulnerable, weak, ill and vulnerable to infection. Table 4 illustrates this with examples of concordances, including 'fragile' and 'ill'. 'Fragile' is used both to describe the state of elderly people as well as to describe that elderly and other fragile people must be protected during the pandemic. This word is often found in the context of describing the goal of the Swedish strategy-'to protect the fragile and elderly'. Thus, the elderly are described as a homogeneous group that is vulnerable and in need of protection. The context is the spread of COVID-19 to nursing homes and then the rapid spread among residents, which led to many elderly people becoming ill and a few dying. Thus, the image of the elderly is greatly simplified and unidimensional.

Collocates Illustrative Example of Concordance
Fragile Deals with the fact that an awful lot of elderly and fragile people have been swept away, probably Ill in general, those who become seriously ill and require intensive care are the elderly and preferential Moreover, the homogenous nature of the elderly category is problematic because it makes it easier to portray the elderly as a homogenous group. Collocating the term with words such as fragile and ill is the result of an ageist discourse [19]. The word 'fragile' was used to describe the state of the elderly as well as convey the idea that the elderly and other fragile people must be protected during the pandemic. It was often used when describing the goal of the Swedish strategy, which is to protect the fragile and elderly. As previous studies have shown, the word 'ill' has also been used in the context of the COVID-19 pandemic, since this disease primarily affects the elderly [5].

Argumentation Strategy: By Which Means Are Arguments of Discrimination and Suppression Expressed?
The argumentation strategy was used in relation to the 'elderly' node in our corpora and associated with words such as 'protect' and 'risk groups', which we interpreted in terms of inclusion and exclusion in the political landscape (see Table 5). These words appeared in concordances that were focused on the need to be cared for and protected. In other words, the Swedish strategy was based on the goal of protecting the elderly (who were often referred to as 'our elderly') and the debate centred around whether Sweden has succeeded in doing so. Elderly people are also portrayed as being without their own ability to act, although they must instead be helped and taken care of by society. The elderly have helped to build the country, but now that the corona epidemic is threatening their health and lives, we must hurry to their rescue. The elderly are worthy of our care but, simultaneously, are deprived of the opportunity to act independently; however, they are often described in the same way as children, as those who must be protected and taken care of but are not full-fledged members of society.
Positioning the elderly as a risk group to be protected can be considered as a form of political inclusion; however, it can also be seen as a means to turn them into a group that needs to be governed or is legitimate to govern. In accordance with Skoog [5], we understand that formulating the elderly as the primary 'at risk group' can have implications on the measures that are possible to implement. For example, in Sweden, elderly focused restrictions were implemented, such as visitor curfews at nursing homes. Many of the words that create the context for the 'elderly' node describe aspects of institutionalised care, such as special nursing homes, home care, personnel, visits and relatives (see Table 6). These words can be understood through a perspectivation strategy, which refers to a perspective from the speaker's perspective. Although the critique is of a more general political character, the framing of nursing homes and elderly care produces a negative image of institutional elderly care. Municipalities did not inform the authorities about the virus in specific elderly nursing homes, which created concern among relatives, but the National Board of Health and Welfare can When reading the concordances in context, it is evident that these sentences are often embedded in a narrative regarding the large number of deaths among the elderly in nursing homes along with either the failure of the Swedish pandemic strategy or the historical governance of elderly care. The perspectivation strategy made it possible to question and criticize elderly care from the perspective of the relatives and friends. This reveals how the framing of the pandemic in relation to institutional elderly care was used to illuminate the deficient structure and organisation of such care, including financial shortages and the large number of personnel without professional training and that are on temporary employment contracts and move among elderly homes. Personnel, nursing homes, home care and visitors were then framed as posing infection risks to the elderly. Thus, the elderly were constructed as an at-risk group, while visitors, personnel and nursing homes were constructed as risky or a threat to the elderly.

Discussion: Intensification Strategy
In the present study, we used a combination of corpus linguistics and discourse analysis, based on corpus assisted discourse studies (CADS) [40], to identify constructions of the elderly in Swedish mass media coverage of the COVID-19 pandemic during the year 2020. In this last discussion section, we summarize the results and discuss the intensification strategy. The intensification and mitigation strategies concern how the epistemic condition of the elderly and others are emphasized or softened in the corpus [37]. As mentioned in the introduction, in the spring of 2020 it became evident that Sweden had failed to protect its elderly, mainly because it had de-prioritised elderly care for a long time [7,8]. Our results confirm that the word 'elderly' was frequently mentioned in articles on COVID-19 in 2020, particularly during the spring. The pattern was the same in all three studied newspapers, thereby indicating homogeneity in news reporting during the period under investigation.
Our aim with this study has been to explore the discursive formations of the elderly in Swedish media during the pandemic and how these formations relate to risk and the discursive constructions of in-and out-groups. Through our analysis thus far, we have showed that that the elderly were labelled and referred to through the perspective of Swedish welfare exceptionalism, in which their in-group position is emphasized through focus on being 'ours'. When comparing this with previous studies, it is also possible to read this referential strategy as a means for journalists to re-inscribe value into the category of elderly who are perceived to be sacrificed in the struggle for herd immunity [3]. This tendency appears to be less evident in studies of the framing of elderly in media reports during the pandemic outside Sweden. Although it might not be intentional, this rhetoric has also opened up for a pitting of groups against each other, where, for example, our elderly must be safeguarded before or instead of other vulnerable groups, such as immigrants.
With regard to the nouns and adjectives attributed to the elderly, our study revealed that it was a rather stereotypical and homogenous framing of the elderly as vulnerable, weak, ill and exposed to infection. This result conforms to those of previous studies both within and outside of Sweden, and as these studies have already stated that such framing performs both explicit and implicit ageism [17,19,21]. By means of which arguments is this discrimination and suppression expressed? The positioning of the elderly as a risk group to be protected is considered a form of political inclusion; simultaneously, it is a means to turn them into a group that needs to be governed or is legitimate to govern. This benevolent form of ageism was manifested in the Swedish context through the specific restrictions that were developed for people aged above 70 years and their care homes. The perspectivation strategy present in our corpus reveals how the pandemic made it possible to question and criticize elderly care from the perspective of relatives and others and, as has also been indicated in other studies, to discuss major problems with institutional living [20]. Personnel, nursing homes, home care and visitors were then framed as posing infection risks to the elderly. Now, what is remaining is to answer the question regarding whether the respective utterances are intensified or mitigated. In studies similar to ours, it is revealed that the media communication on the COVID-19 pandemic employs ageist language and ageist stereotypes to characterize the elderly [19,21]. Thus, the management of the pandemic caused a mental construction of the elderly as being fragile and as almost the only ones who could become ill due to the virus [5]. The Swedish public debate in the spring of 2020 mirrored this focus on restrictions for the elderly and other risk mitigation actions that the government and the Swedish Public Health Agency implemented at the time [15]. Considering this stereotypical formation of the elderly in the corpus, we see this as an intensification strategy which modifies the understanding of the elderly in society. However, as such, the discursive formulation of the elderly as an in-group, 'our elderly', also works as a device to strengthen the definition of 'us' and those who belong in Swedish society. This language of risk was entangled with positions of subordination associated with age, and, as a subject, one tends to act according to performative risk discourses, as the work of normalisation processes [33]. The normalised understanding of old age as a manageable risk [36] was confirmed and enhanced during the pandemic. Thus, to better understand vulnerability management and resilience under disasters, we need to investigate how common social groups, such as the elderly, are constructed in terms of risk in the public debate. We attempted to disentangle the manner in which the performativity of risk is intertwined with the processes by which age is constructed within a system and through which risk discourses can be mobilised and used to uphold other formations. The elderly were normalized as ill and in need of protection-that is, an at-risk group-and this was considered natural and taken for granted [16,37]. We found similar patterns as those found by Zinn and MacDonald [11] in their study of how risk reporting in print news media has shifted towards a greater emphasis on what they call 'everyday social groups'-for example, women, the elderly and children-but with the display of less agency in the linguistic expressions related to the elderly. Thus, the elderly were also homogenised, as we did not find any examples in our collocates where the elderly category was non-homogenised or deconstructed, at least not in any of the concordances that we read.