Elite stigmatization of the unemployed: The association between framing and public attitudes

This paper uses a multi-methods approach to explore the social psychological construction of stigma towards the unemployed. Study 1a uses thematic analysis to explore frames used by political elites in speeches at U.K. national party conferences between 1996 and 2016 ( n = 43); in study 1b, we track the usage of these frames in six national newspapers ( n = 167,723 articles) over the same period showing an increase in the use of negative frames. Study 1c shows that these are associated with national attitudes towards welfare recipients using the British Social Attitude Survey. We ﬁnd the ‘Othering’ frame is correlated with negative attitudes towards the unemployed, even when controlling for the unemployment rate. This ﬁnding supports the claim that social attitudes are related to frames produced in the political and media spheres. We provide theoretical integration between social representations theory and framing which affords development in both domains.

Following the financial crisis in 2008, successive UK governments have implemented austerity measures to reduce public spending which has particularly impacted the welfare state (Reeves, Basu, McKee, Marmot, & Stuckler, 2013). These changes coincided with a hardening of media reporting and political rhetoric associated with unemployed people receiving welfare payments (Fletcher, Flint, Batty, & McNeil, 2016). Notions such as 'scroungers and shirkers' have become a prevalent part of public discourse (Jensen & Tyler, 2015;Patrick, 2016). Though it is often argued that this negative rhetoric is associated with attitude changes in the population, negatively impacting welfare recipients by stigmatizing them, this relationship has not been explored empirically.
Thus, this paper aims to investigate whether there is a relationship between (1) political discourse, (2) newspaper reporting, and (3) public attitudes towards the unemployed. Hence, we look at the association between the framing of a specific issue (unemployment) by politicians and its reproduction in national newspapers. We then test whether there is a relationship between the reproduction of political frames and negative attitudes towards the unemployed at a national level.
Specifically, we map the prevalence of discursive frames with a dictionary of words, derived from thematic analysis of political party leaders' speeches. We use the dictionary to indicate the presence of each frame in six national newspapers over 22 years to demonstrate how the prevalence of different frames has changed over time. The time This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. series is then examined alongside British Social Attitude (BSA) Survey data concerning the unemployed and unemployment. We find that negative media frames used when reporting about unemployment are correlated with negative attitudes towards the unemployed in the population, even when controlling for the actual unemployment rate.

Framing
Framing is a widely used concept in social psychology, political science, and communication and is defined as 'the process by which people develop a particular conceptualization of an issue or reorient their thinking about an issue' (Chong & Druckman, 2007, p. 104). Framing operates through communication; for example, economic discourse in the public sphere may be framed in ways that highlight certain elements (e.g., growth) and not others (e.g., average wage). Research on framing supposes that the prevalence, or exposure, to certain frames influences attitudes of those exposed to the frame. This is known as the 'framing effect'. Much research has explored how the 'frames in the communications of elites (e.g., politicians, media outlets, interest groups) influence citizens' frames and attitudes' (Chong & Druckman, 2007, p. 109). Framing in the context of elite communication is said to operate by; making new information available; making information which is already known accessible (priming) and/or making certain information more important for the evaluation of a target (Brewer, Graf, & Willnat, 2003;Chong & Druckman, 2007).
Therefore, framing can be considered a political process, often originating from political leaders (Jacoby, 2000). This conceptualization is known in the literature as emphasis framing. Accordingly, public opinion or social attitudes are developed through the interaction of political elites (high profile, senior) and media, whereby politicians frame issues in ways which are beneficial to their party-political goals (Druckman, 2001). This approach may entail emphasizing specific elements of an issue, such as individualized explanations for unemployment (Feather, 1985;Lewis, Snell, & Furnham, 1987), which, when reproduced in mass media, focus the public's evaluation of unemployment only in those terms (Nelson, 2004).
Moreover, research has shown that influential mass media (i.e., newspapers of record such as The Daily Telegraph) are narratively reproduced by other media forms such as tabloid press and digital news outlets (Wang & Shoemaker, 2011). Thus, frames used by politicians are likely to be widely shared in newspaper outlets and therefore highly accessible within the social milieu of their origin.
Overall, the literature suggests that framing operates through the reproduction of narratives used by political elites in elite media, which are then co-opted by other media sources. It is important to note that this process is likely to influence, and be influenced by, the attitudes and frames-in-thought (an individual's pre-existing considerations in evaluating a target) of the public in an interactive and iterative process. As such framing is not a unidirectional process, rather it informs and is informed by existing public opinion.

Social psychology and framing
Framing is closely related to the theory of social representations (Bauer & Gaskell, 1999;Elcheroth, Doise, & Reicher, 2011;Jovchelovitch, 2007;Moscovici, 2000) that has been directly deployed in framing research (Uzelgun & Castro, 2015). Social representations are the socially constructed, everyday knowledge that enables humans to interact with the world around them, including other humans, physical and metaphysical objects. The two theories are linked through their attention to knowledge production and common-sense making. Connecting social representations theory (SRT) and framing is empirically useful as SRT provides several concepts that offer analytical power to framing theory. For instance, SRT distinguishes between knowledge that is hegemonic (widely shared, almost universally accepted), emancipated (shared among sub-groups), and polemic (controversial notions and conflicts; Moscovici, 1988;Mouro & Castro, 2012).
Research on framing in mass media may track the conversion of specific representations from polemic to hegemonic or vice versa. Mapping out these transformations and transitions offers an inroad to understanding how framesand the ideas, values, and meanings they conveytravel and change in public spheres.
Social representations are developed in dialogue with others (Bauer & Gaskell, 1999). As such, representation entails the consideration of alternative ideas and other groups in their formation (Gillespie, 2008;Jovchelovitch, 1995). In any given public sphere, hegemonic, polemic, and emancipated representations originating within different interest groups coexist and come into tension.
Thus, from this perspective, framing entails a negotiation between politicians, the mass media, and the polity about the meaning of a specific issue. Politicians, in framing an issue, consider the expectations, beliefs, and possible reactions of the electorate and media in a self-other dynamic. This pattern fits well with what we have defined earlier as the framing process, aptly describing an interaction between different interests to define an issue. Thus, social representational dynamics are likely to underpin both the efficacy of frames used by politicians and media, but also the content and form they take.

The present context and study
Within the present study, it is important to note that UK welfare recipients have come to the forefront of political and media discourse in the context of austerity, following the financial crisis of 2008. It is argued that the crisis and resulting austerity precipitated changes to social security provisions including increased conditionality for out-of-work benefits such as Jobseekers Allowance (JSA)/Universal Credit (UC), which is the main form of assistance available to the unemployed (Dwyer & Wright, 2014). These changes are argued to have influenced the ways unemployed people and unemployment are discussed in the media, leading to a rise in negative representations (Jensen, 2014) and a general assumption that stigmatization of those receiving welfare benefits is hegemonic (Fletcher et al., 2016;Shildrick, MacDonald, & Furlong, 2014).
Specifically, academic and lay explanations of the rise of stigmatization of the unemployed locate its cause with media and political elites (Shildrick et al., 2014). These elite actors have marginalized welfare recipients to provide a pretext that justifies reduced and more conditional welfare spending through the creation of an anti-welfare common sense (Jensen & Tyler, 2015).
Empirically, this relationship would entail a positive association between negative media framing of the unemployed and negative attitudes towards the unemployed in the population. However, studies have not specifically investigated the relationship between political framing, media framing, and attitudes on a national level concerning unemployment. To investigate this, we conduct three related studies to understand possible framing effects on attitudes towards the unemployed in the United Kingdom.
In an exploratory analysis, we look at how politicians frame unemployment and the unemployed. Building on this analysis and based on previous literature, we hypothesize: H 1. Negative framing of the unemployed in news media will increase at a faster rate than other kinds of framing within the analysis period.
H 2. Negative framing of the unemployed will be positively associated with negative attitudes towards the unemployed at a national level.

Methods
To explore frames used by political elites, we investigated British Political Party leader's speeches at annual conferences 1 from the two main parties (Labour and Conservative). Speeches at annual conferences address members of the political parties, but also the nation at large and establish key policy initiatives and their rationale. These speeches are a key site where frames related to groups or issues within society are discussed explicitly.
We sampled speeches from 1996, when the current main welfare payment for unemployed citizens seeking work (JSA) was introduced, until 2016 when the new regime of UC began to be rolled out widely (n = 43 speeches). During this period, there were 10 party leaders (six Conservative, four Labour) of which five became (or were) Prime Minister. 2 Previous research has used such data to explore the construction of social representations and their parameters (Gleibs, Hendricks, & Kurz, 2018;Obradovi c & Howarth, 2018;Reicher & Hopkins, 1996).
We used thematic analysis conducted with Nvivo software. The analysis focused on politicians' talk about unemployment broadly, including welfare benefits and unemployed people specifically. We employ thematic analysis (Braun & Clarke, 2006) for its flexibility and focus on how unemployment and unemployed people are constructed. We utilize an inductive approach to the data analysis, concentrating on the semantic content of leader's speeches rather than latent meaning. We move from direct coding, which is descriptive, to summarization involving the interpretation of the overall meaning of similar codes (themes). The analysis followed an iterative process of close reading of the transcripts, followed by coding where political leaders discuss unemployment, then grouping the codes into sub-themes and finally overarching themes. These overarching themes are then taken as our frames throughout the rest of the paper (see Table 1).

Results
Overall, three prominent frames in the rhetoric of political leaders are evident. These are 'othering the unemployed', 'politics of unemployment', and 'welfare policy'. One of these frames, 'othering the unemployed', is decidedly negative. The other two frames are more neutral overall, as they can be presented as positive or negative depending on the project of the speaker. We discuss each of the three frames below, drawing on subthemes only to illustrate the different ways in which the frames manifest in political rhetoric.
Othering the unemployed 'Othering the unemployed' represents a frame deployed by politicians to discuss the individual attributes of unemployed people, and more broadly to distinguish the unemployed from other citizens based on normative cultural differences. The use of the term 'othering' denotes the sense of defining the unemployed as intrinsically different, and subordinate to, the 'average' British citizen.
Almost half (n = 21) of all speeches in the data corpus refer in some way to specific cultural norms of the unemployed that are responsible for their situation. For example: We're going to liberate people from the culture of welfare dependency with a Common Sense Revolution. It's time to insist that those who can work, must work. (emphasis added;William Hague, Conservative, 1999) It is made apparent here that the unemployed are 'choosing not to work', and this is proposed as a cultural norm of 'welfare dependency' in opposition to the rest of the society. This differentiation builds separation between 'us' and the unemployed, partitioning them as a cultural other. This notion is similar to the individualistic mode of explanation for unemployment argued by Lewis, Snell, and Furnham (1987). However, it goes further, considering that unemployed people have a shared culture (Likki & Staerkle, 2015;Shildrick et al., 2014) and by the same token are apart from the culture of the rest of society. The speaker (Hague) references a future project based on 'common sense'. This future project entails a society in which the culture of welfare dependency is abolished and those who practise it are realigned with the rest of society.
However, politicians do not only focus on the future project of the nation when othering the unemployed by ascribing cultural differences to them. They also appeal to the past, as a place where positive shared norms around work can be found: Decades ago, when we had a universal collective culture of respect for work, a system of unconditional benefits was good and right and effective. . . That culture doesn't exist anymore. In fact, worse than that, the benefit system itself encourages a benefit culture . . . So we will end the something for nothing culture (emphasis added; David Cameron, Conservative, 2008) Here, David Cameron appeals to a historical period when all citizens shared a culture of work. He argues that this culture no longer exists, having been replaced by a 'benefit culture', characterizing it by its 'lack of respect for work'.
Politicians draw on and attempt to create, a shared understanding of a distinct subculture of unemployment. This attempt is often signalled by an emphasis on state dependency and more recently a 'something for nothing culture'. This notional 'myth of voluntary unemployment' (MacLeavy, 2011, p. 5) is deployed as an affront to the historical national and cultural norms of British society as referenced in David Cameron's statement that in the past -'we had a universal collective culture of respect for work'.
Another strategy used by party leaders juxtaposes 'hard-working, law-abiding people' and 'ordinary, working-class' against the 'culture of benefits'. Here, 'hard-working' is used as a term which encapsulates British culture, clearly implying that those who do not work are excluded from the constituency of political elites. For instance: The Conservative Party has always stood for hard-working, law-abiding people. And we stand for them again today. (Ian Duncan Smith, Conservative, 2003) In 20 speeches, party leaders made direct reference to the 'hard-working', mainly defining them as the population to which their party was focusing their attention and policies. This helps to create a representation of who deserves support and who should be excluded. In some cases, the employed are directly contrasted with those who do not work: . . .hard working families who play by the rules are not going to see their opportunities blighted by those that don't. (Tony Blair, Labour, 2004) Other research in this area has noted similar findings in the way that the unemployed are not just defined, but also compared, with employed people (Gibson, 2009). We can interpret this kind of rhetoric as identity entrepreneurship (Gleibs et al., 2018;Reicher et al., 2005), where British identity is constructed around notions of hard work, effectively excluding unemployed from belonging within the national identity.
This frame of 'Othering the Unemployed' may set ingroup boundaries that are defined by engagement with the labour market. As such, those who are engaged with the labour market become part of the ingroup to whom politicians' direct rhetoric and policy, whereas those claiming welfare benefits are excluded (i.e., made as an 'other' in opposition to the ingroup norms of hard work). The 'other' here is demonized as a threat to cultural norms and values. The unemployed are represented as responsible for their own circumstances which necessitates radical action to eliminate the threat to the national project. This kind of rhetoric when shared widely may encourage an anti-welfare common sense (Slater, 2014) that is likely to be associated with negative attitudes towards the unemployed nationally. However, such an association is yet untested.

Politics of unemployment
The second frame is 'politics of unemployment'. It is often deployed to either aggrandize the achievements of one's political party or debase the record of another by referring to the rate of unemployment, job creation, or other statistical measures. This frame was present in 14 speeches.
We set out to create jobs. And we are succeeding. Unemployment is lower here than in any comparable country in Europe. In Britain it is falling. Across Europe it is not. (John Major, Conservative, 1996) Here, John Major attests to the conservative party's success in reducing unemployment at a faster rate than other comparable nations. This claim implies that the economy is doing well, and, by association, the Conservatives' economic policies are succeeding.
So what have we seen? We've seen recession, higher unemployment, higher borrowing. I don't think that's what people were promised. (Ed Miliband, Labour, 2012) In this quote, Miliband, rather than praising his own party, discusses the failure of the opposition (higher unemployment, higher borrowing) and questions their campaign promises. Economics are instrumental in electoral politics and the perceived economic aptitude of a party can be influential in elections. This frame generally represents how party leaders frame the economic circumstances related to the rates of unemployment. This comparison is done either by relation to previous British governments or by contrast to similar foreign nations.
The importance of this frame is to construct an account of economic and therefore political success or failure through unemployment. Notably, though, this frame is not indicative of the kinds of people who are unemployed and therefore can be influential in creating more sympathetic attitudes to unemployment. For instance, where unemployment is high, the electorate may be more compassionate towards the unemployed, because economic circumstances are challenging. This consideration could give rise to notions that unemployment is a matter of societal conditions and not reserved for a specific sub-culture (Lewis et al., 1987). Literature that seeks to understand attitudes towards the unemployed often distinguishes between individual and structural causes for unemployment (Bullock, 1999;Feather, 1985;Piff et al., 2020). The political frame can represent a structural cause for unemployment where high rates of unemployment or related issues are foregrounded.

Welfare policy
Finally, in the 'welfare policy' frame (n = 33 speeches), politicians use unemployment, and the dangers it poses, as a platform for supporting new initiatives. Through this analysis, we can trace the introduction of new policies and their perceived impact. For example, John Major (Conservative) in 1996 states: This week we Tories took a big step forward with the start of our new Job Seeker's Allowance. We do not want to pay people to stay on the dole. We do want to help them get back into work.
Political elites deploy the frame as a solution to the problems of either the welfare state broadly or unemployment specifically. Also, in some cases, politicians are explicit about the kinds of unemployed people who will benefit from new policies: We are adding today the option of self employment as part of the new deal. But they have to take one of the options on offer. We want single mothers with school age children at least to visit a job centre, not just stay at home waiting for the benefit cheque every week (Tony Blair, Labour, 1997) Here Tony Blair portrays an image of a single mother, conjuring the trope of the 'welfare queen' (Bullock, Fraser Wyche, & Williams, 2001;Chauhan & Foster, 2014;Fletcher et al., 2016). The policy solution, in this case, provides state assistance conditional on attending a jobcentre. Conditionality of welfare payments introduced during this period changed the welfare system drastically (Dwyer, 2004;Dwyer & Wright, 2014). More recent changes are an advancement of this idea: With us, if you're out of work, you will get unemployment benefit. . .but only if you go to the Job Centre, update your CV, attend interviews and accept the work you're offered. (David Cameron, Conservative, 2014) This more recent form of conditionality includes receiving assistance only if unemployed people 'accept the work they're offered'. The welfare policy frame, then, narrates the conditions upon which unemployed persons and others can receive assistance. The benefits of each initiative are outlined in terms of their impact either directly on the unemployed, or on fiscal savings (Fletcher et al., 2016).

Discussion
Each frame can, and often is, deployed alongside the others. Political elites may describe unemployed people in a way that frames them as an outgroup, while in the same narrative discussing the economic context and offering policy solutions. However, it is useful for answering our research questions about the development of frames over time and their association with attitudes, to separate these into distinct categories. Moreover, although used in conjunction, the frames that we have identified are both internally homogeneous and externally heterogeneous and refer to distinct rhetorical elements.
It is also important to note that the language used to invoke each frame has changed over time and certain phrases that were present in the early speeches are not present in later speeches, such as the notion of 'yob culture' to denote mainly working-class unemployed young men (McDowell, 2007). This development provides support for the analysis method. By directly examining language longitudinally, we can be confident that we have captured a variety of ways in which each frame is deployed and not only the current acceptable terminology.
Through this analysis, we have shown that politicians do frame the unemployed in negative ways in the context of party leaders' speeches. This understanding provides a useful first step in ascertaining whether negative frames have become more prevalent between the introduction of JSA and UC using an ecologically valid analysis of the ways the frames are deployed in naturalistic (for political elites) settings. However, this analysis does not provide us with information about how widely the frames are shared or whether the use of these frames has increased. We address this question in study 1b.

Methods
As we are interested in the prevalence of frames relating to unemployment, those newspapers that are most widely circulated are assumed to be the most precipitous of framing effects for the population at large. Furthermore, national newspapers have often been considered an important medium through which ideas about unemployed people are developed and transmitted (Bauer & Gaskell, 1999;Bullock, 1999;Bullock et al., 2001;Chauhan & Foster, 2014;Dorey, 2010;Fraser, 1994).
In the United Kingdom, newspapers have a political orientation and lend support to political parties; therefore, we have included a variety of newspapers with differing political orientations and reporting style. Specifically, we include the following: The Daily Mail (right-wing, tabloid, n = 16,708 3 ), The Daily Telegraph (right-wing, broadsheet, n = 26,227 3 ), The Mirror (left-wing, tabloid, n = 17,409 3 ), The Sun (right-wing, tabloid, n = 18,949 3 ), and The Daily Express 4 (right-wing, tabloid, n = 18,702 3 ). These newspapers represent the five most widely circulated newspapers over the 22-year period of the analysis. The Guardian (left-wing, broadsheet, n = 40,906 3 ) was added to the analysis to provide a full spectrum of political orientation and reporting style.
We collated keywords/phrases related to each frame from the political leaders' speeches into a dictionary that indicated the presence of the frames. We only use keywords which are direct quotations from leaders' speeches (see Table 1). By only using phrases used in the elite discourse, we solve issues of objectivity in researcher defined dictionaries, where word selection can be compromised by the method of selection or the researcher's hypotheses. The keywords/phrases obtained from the political speeches were used in a keyword search of the six selected UK newspapers over the same period (1996-2017 inclusive) through the Factiva digital archive. Where applicable, all search terms are truncated by use of an asterisk enabling returned results for all forms of the word. The search result is the number of articles containing each search word in each year in all six newspapers (n = 167,723 across all years including duplicates). A proxy for the total number of articles in each newspaper per year was obtained by using the search word 'the' and following the same process (n = 13,368,184 including duplicates). We therefore ascertain what proportion of the total number of articles contain the search word in question by dividing the number of search word hits in each year by the total number of articles in each year. Thus, in the analysis, increases in the use of a search term are increases in the number of articles using that term as a proportion of the total in that year. This is summed to give a total proportion for each frame. Following Phelps et al. (2012), keywords/phrases returning less than 20 articles in the peak year were removed leaving a total of 44 keywords/phrases to be included in the analysis (see Table 2). Readers will notice that the number of keywords/phrases used in each frame is unequal; this reflects the language used by politicians which is specific to each frame. However, the number of search words in each frame is not directly related to the number of articles retrieved (Table 3).
To understand how frame usage has changed over time in the media, we employed two statistical measures. The correlation between the proportion of articles containing the search word and linear time (Pearson's r), and the estimated mean annual change (EMAC).
The EMAC measure (Nafstad et al., 2013;Nafstad, Blakar, Carlquist, Phelps, & Rand-Hendriksen, 2007, 2009Phelps et al., 2012) is calculated using a relative linear regression slope. This is done by dividing the regression slope (number of articles predicted by the year) by the mean number of articles per year for each keyword and multiplying this figure by 100. For example, if the search word 'journal' has a slope coefficient of 0.1 and an 3 Excluding duplicates 4 Analysis for this newspaper begins in 1997 because the Factiva archives' records start for this paper in that year average of 10 articles per year for 20 years; then, the calculation would be 0.1/10 9 100. We would then report an EMAC of 1%, indicating an increase of usage by 1% each year over 20 years.
The EMAC accounts for factors not addressed by simple percentage calculations, including consideration for keywords that begin at different points in the time series (for a larger discussion of the EMAC, see Nafstad et al., 2009). We have varied the EMAC calculation from previous research that looks directly at the number of times a word is used by basing the calculation on the mean number of articles that include each search term. This is necessary because we do not have a valid comparison with the development of a large sample of popular words over time (e.g., Nafstad et al., 2013).

Estimated mean annual change (EMAC)
The developmental changes in the usage of the three frames we have identified are presented in Table 4. We see that the Othering frame has an EMAC increase of 2.7% (M = .0036, SD = 0.0013 5 ), r = .49, p = .021 n = 22. EMAC scores of 3% or more are considered high (Nafstad et al., 2013). This increase tells us that the Othering frame is becoming a more popular narrative over time, which adds further credence to the assertions of other researchers about the growing use of stigmatizing language to describe unemployed people (Fletcher et al., 2016;Friedli & Stearn, 2015;Gibson, 2009). This frame's usage peaked in 2013 during the height of the conservative parties' changes to social security provision. 2013 was also the year that UC was introduced to replace a range of means-tested social security benefits (Fletcher et al., 2016). These facts may suggest that the Othering frame is deployed during times of political change relating to the provision of social security.
The politics of unemployment frame has an EMAC of À1.24% (M = .0038 SD = 0.0013), r = À.24, p = .290 n = 22) indicating that the frames' use is slightly declining; however, given that the Pearson correlation is not significant, we conclude this is a more volatile frame that relies heavily on the context of use. It is notable that the lowest year in which this frame was present, as a proportion of all articles, was a year before the financial crisis (2007) and its peak year was after the financial crisis began (2009). This result indicates the ecological validity of the frames we have identified, given that they mirror the socio-political context at the time.
Finally, the welfare policy frame has seen a trivial increase over the analysis period of 0.61% (M = .0052, SD = 0.0019), r = .11, p = .621, n = 22. However, it is to be noted again that the correlation with linear time is not significant and therefore changing usage of the frame is not related to the passage of time but rather other contextual variables. To summarize, the Othering frame has seen the greatest increases using the EMAC measure and is significantly correlated with linear time. As such, H 1 is confirmed; negative framing of the unemployed in national news media is increasing at a faster rate than other frames we have identified. Still, plotting these results against major political events shows that these changes should be contextualized within the broader political landscape (Figure 1). We see falls in the use of all three frames especially following the start of the 'war on terror' and large increases following the financial crisis and the onset of austerity. Looking forward, we see a similar decline from the start of the Brexit referendum which is likely to continue until the United Kingdom leaves the EU. Nevertheless, the upward trend of 'othering the unemployed' is relatively stable from 2002 until the start of UC. We also note that there was not a sustained increase in the use of the 'politics of unemployment' frame even during what was a sustained financial crisis.

Discussion
The results of study 1b show that negative framing of the unemployed has become more prevalent in the analysed newspapers. The change in prevalence of negative framing of the unemployed is significantly associated with linear time.  Thus, we have provided evidence to reject the null hypothesis, that negative framing of the unemployed does not increase at a faster rate than other kinds of framing. However, this study does not show whether the increased use of framing has effects on the populations' attitudes towards unemployed people. That is, we have not shown what frames do. In study 1c, we look at the relationship between the use of frames and attitudes at the national level to test for possible framing effects.

Methods
To ascertain whether changes in the use of different frames have any relationship with overall national attitudes towards welfare recipients, we obtained BSA Survey data for the period 1996-2017 for five variables related to unemployment. BSA survey is a representative cross-sectional survey consisting of approximately 3,000 participants per year. The variables chosen concern attitudes towards welfare and welfare recipients and have been collected for a large majority of the analysis period. A time series of these variables is presented in Figure 2.
The first variable we analysed was the proportion of respondents who disagree/ strongly disagree with the item 'the government should spend more money on welfare benefits' (MOREWELF) on a five-point scale (N = 21 6 , M = .32, SD = 0.05). The second variable asks respondents to choose between two statements 'benefits for unemployed people are too low and cause hardship, OR benefits for unemployed people are too high and discourage them from finding jobs' (DOLE, N = 22, M = .50, SD = 0.10). Here, we take the percentage of people who agree with the latter. The third variable examines the percentage of people who agree/strongly agree with the statement 'if welfare benefits weren't so generous, people would learn to stand on their own two feet' (WELFEET, N = 21, M = .48, SD = 0.07). Fourth, we investigate the proportion of people who agree/ agree strongly 'Most people on the dole are fiddling in one way or another' (DOLEFIDL, N = 21, M = .36, SD = 0.05). Finally, we use the variable UNEMPJOB which asks, 'How much do you agree or disagree that . . . around here, most unemployed people could find a job if they really wanted one'. We take the percentage that agree or agree strongly (UNEMPJOB, N = 21, M = .61, SD = 0.07).
We also include a measure of the unemployment rate from the Eurostat database. The rate is the percentage of the working-age population in the United Kingdom who were unemployed in the reference week, available for work, and actively seeking work. This rate is distinct from the number of people claiming social security support because they are unemployed, which is known as the 'claimant count'. We also include this measure in the correlation analysis. The claimant count data were drawn from the Office of National Statistics (ONS) UK.

Results
To see whether there was an association between the use of the identified frames in national newspapers (Othering the Unemployed, Politics of Unemployment and Welfare Policy) and BSA measures related to unemployment, we conducted a correlational analysis. Table 5 summarizes the results and shows that the Othering frame is consistently associated with negative attitude measures in the population (except DOLEFIDL, which was uncorrelated with any other variables and UNEMPJOB which was negatively correlated with both Politics of Unemployment and Welfare Policy). Othering is also significantly associated with the unemployment rate (r = .64, p = .001), However, it is not significantly associated with the more direct measure, claimant count. This difference is of note because it suggests that the actual number of people claiming social security benefits is not an important prerequisite for heightened stigmatization of this group. However, the unemployment rate is highly correlated with the claimant count. We may speculate that where the unemployment rate rises, UK citizens, media, and politicians may be sensitized to possible future rises in the claimant count, contributing to further stigmatization of unemployed people who claim social security benefits.
Notably, our other frames are associated with UNEMPJOB and none of the other attitude measures. This indicates that increased use of these frames reduces negative attitudes towards the unemployed concerning their ability to find work. This adds credence to our earlier assertion that highlighting structural rather than individual causes of unemployment may ameliorate negative attitudes to the unemployed in the population. Overall though, we have shown that negative framing of unemployed welfare recipients is positively associated with negative attitudes in the population, supporting H 2 .
Given that we intuitively may suspect the rate of unemployment is a confounding variable in the association between negative framing and negative attitudes, we conducted multiple linear regression to test the effects of the Othering frame on attitudes when controlling for the unemployment rate. This process was done with each of the attitude variables, though UNEMPJOB and DOLEFIDL are not reported here due to nonsignificant correlations. The regression models for each of the other variables were Note. M and SD are used to represent mean and standard deviation, respectively. Values in square brackets indicate the 95% confidence interval for each correlation.
The confidence interval is a plausible range of population correlations that could have caused the sample correlation. *Indicates p < .05; ** indicates p < .01.
significant, and the Othering frame was a significant predictor of these negative attitudes even when controlling for the rate of unemployment (see Table 6).

General discussion
In study 1a, we provided evidence that politicians use at least three different frames to discuss the unemployed and unemployment. One of these frames, 'Othering the Unemployed' was decidedly negative. Our first hypothesis, that negative framing of the unemployed increases at a faster rate than other frames found support using the EMAC measure. We also found support for our second hypothesis that negative framing of the unemployed would be positively correlated with negative attitudes towards the unemployed. Additional evidence was found using multiple linear regression models where we controlled for the unemployment rate, showing that over and above the effect of the unemployment rate, there is a significant relationship between the Othering frame and negative attitudes in the population. Much of the framing literature does not unambiguously test relationships between political framing, media framing, and attitudes nationally. In this paper, we provide a specific test of this association between framing and attitudes towards the unemployed/ unemployment in the United Kingdom, tracing their usage, through both political communication and widely shared newspaper reporting, longitudinally.
Through this methodology of tracking frames through different mediums of communication, we support assertions from framing literature that suppose the prevalence of, and exposure to, frames influence the attitudes of citizens towards the object of the frame. In this case, leading to more negative attitudes towards the unemployed in the general population. This method is consistent with conceptualizations of framing that posit it as a political phenomenon originating with political elites (Jacoby, 2000) and not necessarily based on factual information (Hopkins, Sides, & Citrin, 2019) such as the actual unemployment rate.
However, framing theory provides only a partial account of where frames originate and how they are developed in the public sphere. Here, an integration of the SRT literature is useful to account for the development of frames through the interactivity of different actors in the public sphere to define the issues associated with unemployment. Social representations as we have described them, embed self-other relations in their constitution. That is, social representations are intersubjectively agreed social realities. When politicians seek to frame an issue, they must be aware of and consider, the possible reactions, motives, and beliefs of the polity. As such, framing is not a unidirectional relationship from political elites to citizens. Rather the assumed beliefs of citizens define acceptable and popular frames on issues of political import.
Different social representations of the same issue may exclude or diminish the veracity of other representations (Howarth, 2006); this can explain how the 'Othering' frame increases in use over time and in particular after the financial crises while other, competing frames remain stagnant. This process alludes to the development of increasing hegemony of the Othering frame to account for unemployment. The results support the theoretical hypothesis that 'otherising', which relies on social psychological processes of creating outgroups (Tajfel & Turner, 1979), is an effective frame in shaping social attitudes. Such a frame it appears is much more compelling than the policy or political domains which do not draw upon these social psychological processes. The results of this study have several implications. Foremost, they support both academic and lay assertions that stigmatization of the unemployed is related to political rhetoric and media elites. At a societal level, the ramifications of this may be the open acceptance of stigmatization of the unemployed, making negative attitudes towards the unemployed a common-sense, natural assertion. At the interpersonal level, there are likely to be impacts on the social interactions of unemployed people. Given that anti-welfare common sense is widespread it would be sensible for unemployed individuals to assume that that identity would be stigmatized by relevant others, thereby influencing their beliefs about others beliefs (Elcheroth et al., 2011). Finally, at the personal level, it would also be possible for unemployed people to internalize negative attitudes about unemployed people in general and apply those attitudes to themselves. These personally applied negative attitudes may partially explain the negative psychological effects associated with unemployment (McKee-Ryan, Song, Wanberg, & Kinicki, 2005;Wanberg, 2012).
Other research in this area has often suggested that negative media and political rhetoric towards marginalized groups would be associated with negative attitudes towards those groups. However, this study is the first instance (that we know of) where longitudinal data have been used to track this association over time within the context of unemployment in the United Kingdom, using ecologically valid data to ascertain how different narratives are deployed and change over time. As such, this study presents strong evidence that the unemployed have become a more stigmatized group over time worthy of the attention of researchers in social psychology and related disciplines.

Limitations
There are some limitations of this study. First, there are issues with how our dictionary was produced. Because we have used party leaders' speeches at national conferences, we may not have captured all the phrases which indicate the presence of the frames we have identified. Political speeches are a more contrived, formal mode of communication than everyday language (Bauer & Gaskell, 1999). As such, we may not capture the more derogatory phrases used in informal communication and therefore not fully capture all articles which invoke the frames we identified.
Secondly, this research is limited by the relatively low number of time-series observations used in the analysis (BSA, n = 21/22, Newspaper, n = 22). However, posthoc power analysis using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007) suggests that based on the mean correlation between negative framing (Othering) and attitudes (.59) then n = 22 obtains statistical power of .85. Nevertheless, other frames may have smaller effects that we are not able to capture.
We also note the limitations of correlational studies and the inability to understand causal effects. Still, we feel that experimental data are not appropriate for this paper, as the conditions of repeated exposure to frames in the societal context cannot be reproduced satisfactorily in laboratory settings. We have additional confidence in these results because we have controlled for (in study 1c) the effects of the actual unemployment rate on attitudes and have still obtained a significant effect of the Othering frame. However, we also attempted to explore whether a lagged time-series analysis would be appropriate (following Kellstedt, 2000;Russell Neuman, Guggenheim, Mo Jang, & Bae, 2014), but the number of data points was insufficient for a reliable analysis.

Further research
Further research is needed to ascertain the generalizability of the findings of this study with other target groups. It should be possible, using the methodology described here, to examine attitudes towards a wide array of stigmatized groups, if that stigmatization is prevalent in public discourse.
In addition, researchers interested in this topic may look to pin down the direction of causality in the relationship between political rhetoric, media reporting, and individual attitudes. Though it should be noted that our theoretical perspective (SRT) dictates that at any point in the causal chain, there is at least an implicit negotiation between self and other, such that each actor, be it, politician, newspaper, or individual is considering the representations, values, beliefs, and identities of others when making decisions about frames to employ (Gillespie, 2008).
As we have noted, political language and mass media can change the nature of social reality by presenting new knowledge about groups and objects in the social world. This influence is, at least in part, because these actors (politicians and media) are seen as being prototypical members of British society (Haslam & Reicher, 2007;Reicher et al., 2005). As such, knowledge production and common-sense making are about what we think others think of an issue. This notion has been variously described as meta-knowledge or metarepresentation (Elcheroth et al., 2011). To understand the effects of stigmatization on the unemployed, empirical research should investigate the relationship between representations and meta-representations in the context of stigmatized identities. That is, to what extent does stigmatization effect how unemployed people think about their own identities and how they think, others think, about their identities, i.e., meta-identification?

Conclusion
Nothing can be inferred from an individual's employment status about what kind of person they are. Yet, in this paper, we have shown that indeed, employment status is used to infer a variety of negative individual attributes, which designate unemployed people as a cultural other. When these modes of communication are deployed by political and media elites, they influence the attitudes of citizens towards unemployed people, often in stigmatizing ways. This relationship between framing and attitudes towards the unemployed does not go away when we include the actual unemployment rate.
These results provide cause for concern around the lived experience of unemployment considering the stigmatization that people who find themselves out of work face. Coping with a stigmatized social identity is beset with challenges and may reduce the ability of unemployed individuals to find work or seek support to do so. In this paper, we have provided evidence that this stigmatization exists, but more work is needed to understand what its effects are on work-related outcomes. Social scientists and policymakers would do well to turn their attention to understanding and creating systems and policies which would enable positive social identities to be sustained, even in unemployment. Such an approach would surely provide common benefit to society, as well as the individual.