How divided is the attitudinal context for policymaking? Changes in public attitudes to the welfare state, inequality and immigration over two decades in Britain

If public attitudes towards the welfare state, inequality and immigration are becoming increasingly polarized, as recent political events might suggest, the space for progressive social policies is more constrained. Using data from the British Social Attitudes Survey (BSA) spanning 23years, we analyse trends in these attitudes, examining whether there has been divergence between those who have been more and less exposed to disadvantage through changes in the economy and the welfare state across more than two decades. Taken in this longer term context, and examining characteristics not previously considered in relation to public attitudes such as lone parenthood and disability, we find little evidence of polarization in attitudes to welfare, inequality and immigration and even some evidence of attitudinal gaps narrowing. We conclude that given this lack of division, there may be greater room for more pro-welfare and progressive policies than the prevalent narrative of polarization suggests. interaction in is the for levels of country/region and household composition.


| INTRODUCTION
We live in a society with increasingly polarized public attitudesor so recent political events would suggest. Globally, we are invited to consider the electoral success of parties and candidates promoting intolerance and other illiberal values including Trump in the US in 2016, Morawiecki in Poland in 2017 and Bolsonaro in Brazil in 2018. In the UK, a divisive campaign leading up to the Brexit referendum (a split of 51.9% to Leave and 48.1% Remain 1 ) was followed by negotiations about Britain's withdrawal from the EU highlighting further differences between the four nations and between Leavers and Remainers. Most recently, the Black Lives Matter protests, and the counterprotests in response, have brought attention to the deep racial divides that persist across the world.
If public attitudes have become more polarized, as these events suggest, this has important consequences for social policy. Attitudes not only inform people's voting behaviour but also provide the ongoing context in which policy-making takes place, with the potential to be a constraining or enabling force, potentially shaping the future policy agenda (Chung, Taylor-Gooby, & Leruth, 2018). If attitudes become increasingly divided, then policy solutions are difficult to achieve (Ford & Lymperopoulou, 2016).
Some political commentators link apparent attitudinal polarization to the different experiences of the post-war baby boomers and subsequent generations, positing that the belief that your generation will do less well than your parents' represents a fundamental breach of an implicit generational contract (Pew Research Centre, 2018;Social Mobility Commission, 2018;Willets, 2011), although this is challenged by other evidence suggesting that age and birth cohort are less significant than education in explaining attitudes and expectations of the future (Shrimpton et al., 2017). Other commentators look to explanations based on the experience of being economically "left behind" (Goodwin & Heath, 2016). For example, Walker and Corbett (2019) cite the growth in inequality in incomes and wealth, especially housing wealth, and the residualization of social security, as among the common threads underlying the Brexit vote, Trump's electoral success and the rise of right-wing nationalism. Other authors have emphasized the role that immigration, and especially changes in the rate of immigration, can have on public attitudes, particularly attitudes towards welfare provision (Alesina, Glaeser, & Glaeser, 2004;Eger, 2010).
But the connection between experience of inequality, or of immigration, and polarization of attitudes is contested. Iversen and Soskice (2015) argue that in fact countries with higher economic inequality tend to have less polarized political attitudes, and Kymlicka and Banting (2006) find no evidence that higher levels of ethnic diversity arising from immigration undermine support for the welfare state. Moreover, existing evidence of polarization in attitudes is itself incomplete: differences in attitudes between advantaged and disadvantaged population sub-groups, and change over time in these differences, is complex and no unifying pattern emerges (Duffy, Hewlett, McCrae, & Hall, 2019).
In this paper, we aim to fill some of these gaps, examining attitudinal differences between groups in Britain defined by a range of socio-economic characteristics and the extent of polarization or convergence over a 23-year period. Britain provides a useful test case to consider changes in these attitudes because it is characterized by high levels of income inequality compared to most other OECD countries (OECD, 2020). Also, like many countries, Britain has seen significant cuts to social security (Cooper & Hills, 2021) and a shift towards increased conditionality (Wright, Fletcher, & Stewart, 2020). Finally, in common with many other countries, Britain has experienced relatively high levels of immigration (Cangiano, 2019;Office for National Statistics, 2020). Thus if inequality, austerity and immigration are drivers of polarization in attitudes, we would expect to see such polarization in the British case.
We examine attitudes to social protection (or "welfare"), economic inequality and immigration, among groups defined by age, educational qualifications, parenthood, disability and country/region within Britain. Reflecting on these results, we consider whether the observed trends and differences can be characterized as "polarisation" and whether the changes in differences in attitudes between groups appears to be linked either to rising inequality or to changes in immigration. We show that neither hypothesis is straightforwardly compatible with the evidence we present. Our results provide an account of the public "attitudinal capital" Britain had at the start of the pandemic and reveal that this was not as depleted as political events might lead us to believe. Britain was not as divided or as hostile as expected, and this has undoubtedly been a strength. This was reflected during the early stage of the COVID-19 crisis in expressions of mass public support for the National Health Service and welfare state more broadly and widespread concerns about inequality. This reaction is in contrast to the public resistance to  policies in some other countries, including the US which saw anti-lockdown protests in some states (Dyer, 2020;Fairchild, Gostin, & Bayer, 2020;Galea & Abdalla, 2020). The paper concludes with the observation that over-stating the divisions in public attitudes towards key institutions of social protection and their underlying drivers is a disservice to policy analysis and development. Instead, taking the long view of trends in public attitudes and examining a range of potential socio-demographic stratifications, we argue that even in countries like Britain which have experienced high levels of inequality, harsh austerity and significant immigration, there is a window of opportunity to capitalize on the space for more progressive policymaking. Many commentators have highlighted harshening attitudes towards social security claimants following negative discourses prevalent in political rhetoric and mainstream media (Deeming, 2015;Geiger, 2018;O'Grady, 2017). Whilst some have argued there has been a trend towards individualism (Santos, Varnum, & Grossmann, 2017), taking a longer-term view suggests that public support for welfare spending has fluctuated over time rather than continuously declined (Hudson et al., 2016). Attitudes towards collective provision are cyclical in response to levels of government spending, with the public supporting greater taxation and more government spending following times of austerity and withdrawing this support following periods of higher government spending (Pearce & Taylor, 2013, p. 36). Though this cyclical pattern was not strongly evident in Britain immediately after the Coalition Government austerity 2010-2015 (Curtice & Ormston, 2015), by 2016, there appeared to be a shift back in favour of increased taxation and increased spending on health, education and social benefits and more support for redistribution (Curtice, 2016). The trend continued in 2017, with a marked drop in the proportion of people that think the generosity of welfare benefits creates dependence and a steep rise in respondents who believe cutting benefits would damage too many people's lives, according to the British Social Attitudes (BSA) survey (Kelley, Warhurst, & Wishart, 2018).
In relation to inequality, a high proportion of the British public think that the gap between high and low incomes is too large: in the latest BSA data (2018), 78% of people agreed, and this has changed little since it was first measured in 1983 (Clery & Dangerfield, 2019). However, a relatively high degree of concern about inequality does not necessarily translate into a belief that government should redistribute from better off to worse off or that government should increase spending on benefits for the poor (Hills, 2002).
In terms of attitudes about immigration, an analysis of European Social Survey (ESS) data found that on aggregate the UK became more positive about the economic benefits, and slightly more negative about the cultural impact, of immigration over the period 2002 to 2014 (Ford & Lymperopoulou, 2016). The large literature on immigration attitudes highlights many factors relevant to understanding change in these attitudes, including levels and changes in immigration (Coenders & Scheepers, 1998). Group threat theory proposes that a dominant group feels threatened by outside groups (Quillian, 1995). This could be exacerbated by economic pressures, with an increase in racial prejudice during periods of economic downturn when there is increased competition for scarce jobs (Johnston & Lordan, 2016). Ramos, Thijssen, and Coenders (2021) draw on group threat theory to explain their findings that cultural differences (and thereby perceived cultural threat) between minority and majority groups drive discrimination; this suggests the characteristics of migrant populations may be relevant in understanding why attitudes about the cultural impact of immigration in Britain have not softened as much as attitudes about the economic impact of immigration.
2.2 | How do attitudes to inequality, the welfare state and immigration vary across groups?
There has been a strong focus in the literature on how attitudes to inequality, the welfare state and immigration differ by socio-economic status, as variously measured by social class, income and education because self-interest has conventionally been identified as a potential influence on attitudes, alongside individuals' ideological positions and current policy provision (Chung et al., 2018). Whilst it has been argued that people's values and beliefs are more important in explaining attitudes towards redistribution than their material circumstances (Sefton, 2005), in some cases, the two are not so easily separated. Park, Phillips, and Robinson (2007) find that personally experiencing poverty at some point in life is one of the main predictors of being "liberal". 2 Edmiston (2018) confirms that first-hand experience of relative deprivation is important in attitude formation about inequality and the role of the welfare state, partly because auto-biographical experience results in better understanding about the relationship between structure and agency. In-line with these findings, Dunatchik et al. (2016) find that people on low incomes tend to have less harsh views towards unemployed people and are more supportive of spending on benefits.
Age and education are key explanatory characteristics of attitudes towards immigration. Young degree holders are more likely to be positive about both the EU and immigration (Curtice & Montagu, 2019). Education may interact with geography: people of all educational levels were more likely to vote Leave in the Brexit referendum if they lived in a low skill area, with effects being strongest for the more highly qualified (Goodwin & Heath, 2016). Macroeconomic changes also matter. In Ireland, as unemployment rose following the last recession, attitudes to immigrants became more negative, and the attitudes of people with lower levels of education appeared to change the most (McGinnity & Kingston, 2017). This is consistent with Polavieja's (2016) evidence that degree of exposure to "occupational" and "environmental" competition from immigrant workers is associated with strength of anti-immigrant sentiment across Europe. However, a recent list experiment provides an important challenge, finding that those with higher levels of education are more likely than others to "mask" their true attitudes about immigration, hiding their negative views with more socially desirable responses (McGinnity, Creighton, & Fahey, 2020).

| Research questions
The dominant narratives around public attitudes lead us to expect wide and increasing differences between social groupsespecially by age, educational or skill level and geography. To investigate further the extent of polarization in public attitudes and the degree to which this is associated with the experience of being "left behind", we focus on characteristics that define groups who lost out in Britain during the economic recession and austerity that followed the financial crisis in 2008 and those who might be most exposed to any adverse effect of immigration on wages.
The former includes young adults, disabled people and families with (young) children, especially lone parents (Equality and Human Rights Commission, 2018; Hills, Cunliffe, Obolenskaya, & Karagiannaki, 2015;Lupton, Burchardt, Hills, Stewart, & Vizard, 2016). Young adults experienced a fall in their wages in real terms between 2005/6 and 2015/16 and also lost out relative to older adults in terms of household incomes and wealth (Obolenskaya & Hills, 2019). Disabled people and lone parents experienced not only reductions to benefit rates and entitlements (Tucker, 2017) but also significant cuts to services on which they depend, such as social care (Burchardt, Obolenskaya, & Hughes, 2020) and children's centres (Stewart & Reader, 2021). There are also important geographical differences (Obolenskaya, Lupton, & Provan, 2016), with London having markedly different outcomes to the rest of England, and differences between the constituent countries of the UK; we therefore include broad geographic location as part of our analysis.
The Migration Advisory Committee (2018) finds that any effect of immigration on depressing wages is concentrated among the low-waged and low-skilledwhich we proxy through identifying those with no or low educational qualifications. Young adults, disabled people and lone parents are over-represented in low-waged work, so a focus on these groups is also relevant in terms of a possible polarising effect of immigration on attitudes.
While the characteristics of age, educational level and geographic location, included in our analysis, have been central to the debate hitherto, there is little existing evidence on the attitudes of disabled people and parents; this is therefore a significant contribution of our paper. 3 In relation to all of the groups, we ask not only whether the attitudes of the disadvantaged are significantly different from those of more advantaged groups but also whether any gap between them has widened or narrowed over time.
We focus on attitudes related to the welfare state, inequality and immigration as these attitudes are most relevant to the economic and societal changes that have taken place over the period under consideration and have been found to be related to each other (Vandoninck, Meeusen, & Dejaeghere, 2018). Furthermore, these attitudes are particularly important aspects of the context in which social policy-making takes place. In this analysis, our overall aim is to investigate how the attitudinal context for social policymaking has changed over the last two decades. Specifically, we seek to answer: • To what extent, if at all, is there reduced support overall for the welfare state, less concern about inequality and more concern about immigration?
• Have views become more polarized between more and less advantaged groups?

| DATA AND METHODS
We analyse BSA data over a 23-year period from 1995 to 2018. The BSA is an annual attitudinal survey with a representative sample of 3,000-4,000 respondents. Some questions are asked in all years, some on a rotating basis and some ad hoc. The years selected for analysis were constrained by the availability of consistent measures. BSA is the pre-eminent source of data on long-run trends in public attitudes in Britain (Sundberg & Taylor-Gooby, 2013).
Appendix 1 in Data S1 gives definitions of our key characteristics of interestdisability, parenthood, age, education and country/regionand the variables used to identify them. Appendix 2 in Data S1 gives the frequencies and sample sizes for each characteristic in the analysis years.
The outcomes of interest are respondents' attitudes to the welfare state, inequality and immigration. As is common in attitudinal research, we measured these attitudes using indices, which tend to more reliably capture respondents' underlying attitudes than single measures (DeVellis, 2003), and avoid some of the difficulties associated with sensitivity to particular question wording (Sumino, 2018). The variables included in the indices are shown in Box 1.
The index of attitudes to the welfare state is based on a derived variable frequently used in BSA analysis. 4 Interviewees are invited to respond to eight items, categorizing each on a 5-point scale from "strongly agree" to "strongly disagree". Responses are scored (in reverse where necessary), summed and divided by the number of items to produce a final 5-point scale. A higher score indicates greater hostility to welfare.
For concern about inequality, we created an index-based loosely on analysis of liberals and sceptics by Park et al. (2007), using three items as shown in Box 1. The items are the best available given the constraints of the data.
A higher score indicates less concern about inequality.
For attitudes to immigration, our options were constrained by the years in which questions were asked. Nevertheless, we decided this was an important attitude to include given the significance of immigration for recent political events and that our focus is on the attitudinal context for policy-making. For 1995For , 2003 and 2013, we created an index based on six items as shown in Box 1. These years are unfortunately not years in which disability and

Welfarism index
Index based on eight items measured on 5-point "strongly agree" to "strongly disagree" scales: • The welfare state encourages people to stop helping each other.
• The government should spend more money on welfare benefits for the poor, even if it leads to higher taxes.
• Around here, most unemployed people could find a job if they really wanted one.
• Many people who get social security do not really deserve any help.
• Most people on the dole are fiddling in one way or another.
• If welfare benefits were not so generous, people would learn to stand on their own 2 feet.
• Cutting welfare benefits would damage too many people's lives.* • The creation of the welfare state is one of Britain's proudest achievements.* *In 1995, these two items were replaced with: "People receiving social security are made to feel like second class-citizens", and "The welfare state makes people nowadays less willing to look after themselves".
In the final index, higher scores indicate stronger anti-welfare attitudes. Range 1 to 5, interquartile range 2.5 to 3.4.

Inequality index
Index based on three items measured on, or converted to, 3-point scales: • Government should redistribute income from the better off to the less well off (agree/ neither agree nor disagree/disagree) • Level of benefits for unemployed people (too low and cause hardship/neither/too high and discourage from finding jobs)* • Ordinary working people do not get their fair share of the nation's wealth (agree/neither agree nor disagree/disagree) *Three additional response categories accounting for very small proportions of respondents are not included in the construction of the index.
In the final index, higher scores indicate less concern about inequality. Scale 0 to 2.3. Interquartile range 0.7 to 1.5.

Immigration index
For 1995 to 2013: Index based on six items measured on scales with varying number of points, standardized to a range 0-1: • Immigrants increase crime rates parenthood are recorded consistently, so our analysis of this index is restricted to age, education and geography. The same immigration questions were not asked in later years. However, two related questions were asked from 2011 onwards, and we use these in Section 4.2 below to give a sense of the direction of overall trends in attitudes to immigration. We do not use them in our analysis of divergence over time because it is not possible to construct a consistent index over the full period. For both the index and the two separate questions, a higher score indicates greater hostility to immigration.
All analysis presented includes appropriate sample weights and adjusts for the complex survey design (Phillips, Curtice, Phillips, & Perry, 2018a).
We begin by analysing the relationship between our characteristics of interest and each of the three attitudinal indices in 2018 by estimating linear regression (OLS) models. The indices are therefore treated as continuous variables; this approach has been shown to be a reasonable approximation (Phillips, Curtice, Phillips, & Perry, 2018b).
Because the characteristics of interest are associated with each other (see cross-tabulations in Appendix 2 in Data S1), we include all characteristics simultaneously in the models in order to estimate the association of each characteristic with each attitudinal index, independently of the other characteristics in the regression. 5 However, we do not presume that these associations are causal. Based on these regression models, we calculate average predicted scores 6 for each attitudinal index for each characteristic category. Plots of the predicted scores and associated confidence intervals are presented in the main text, with y-axis scales reflecting the interquartile range of the index, and the full regression results are provided in Appendix 3 in Data S1.
We then look at trends over time in each of these attitudes based on the average attitudinal scores for the whole sample. Where possible, we present results for 5-year intervals from 1995 to 2015 and add 2018 as the most recent year available. However, for the immigration index, we have to use wider intervals (1995,2003,2013) and a separate series for 2011 to 2018.
Finally, we estimate models interacting each characteristic in turn with year, in order to analyse whether differences in attitudes between groups diverged, converged or remained the same over the period. These are plotted for visual inspection, and the underlying interaction coefficients (shown in Appendix 3 in Data S1 and reported in the text) are assessed for statistical significance.  Table 1 in Data S1.
Having a disability is associated with a significantly lower Welfarism and Inequality index score, indicating more support for "welfare" and more concern about inequality compared with respondents who do not have a disability. This holds both before and after controlling for other characteristics such as age, with which disability is strongly correlated.
On the whole, having lower levels of educational qualifications is associated with stronger anti-welfare attitudes, less concern about inequality and stronger anti-immigration attitudes, compared with people who have a degree.
Overall age is not significantly associated with any of the attitudes, after controlling for other characteristics, with a few exceptions: those aged 55-74 have significantly lower scores on the Welfarism index indicating stronger support for "welfare" compared with the youngest age group. Those aged 75+ have less concern about inequality than those aged 16-34. Anti-immigration attitudes tend to be stronger for older respondents, though the relationship is not statistically significant. It appears that the "generational divide" in attitudes is considerably less prominent once other characteristics, especially education, are taken into account.
Lone parent households have significantly lower Welfarism scores indicating more pro-welfare attitudes compared with couple parent households. A similar pattern is found for the inequality index, with lone parents having more concern about inequality compared with couple parent households.
On the whole country/region is not a significant predictor of these attitudes once other characteristics are controlled for. Those living in London display more concern about inequality and weaker anti-immigration attitudes compared with the rest of England, though these relationships are only significant at the 90% level. In relation to Immigration, respondents in Wales expressed more anti-immigrant sentiment than respondents in England (significant at 95%).
F I G U R E 1 Predicted attitudinal index scores, by characteristic. Higher scores indicate stronger anti-welfare attitudes, less concern about inequality and stronger anti-immigration attitudes. Note: Predicted scores on the index are indicated by the diamond and 95% confidence intervals by the whiskers. is even lower than that observed in 1995, suggesting less hostile attitudes on average about "welfare".
The average Inequality index score for the sample as a whole increased significantly between 1995 and 2005, indicating a decrease in concern about inequality, and this lower level of concern was sustained in 2010 ( Figure 2b).
Between 2010 and 2018, the average Inequality index score decreased (i.e., more concern about inequality). Concern about inequality has therefore followed a similar pattern to attitudes to the welfare state, although concern about inequality has not yet returned to the level observed in 1995.
In Figure 2c Higher scores indicate stronger anti-welfare attitudes, less concern about inequality and stronger anti-immigration attitudes. Note: Average scores on the index are indicated by the spot and 95% confidence intervals by the whiskers. Source: authors' calculations using BSA 1995BSA -2018 Immigration index, and so we should be cautious about drawing comparisons, it does suggest that since 2013 attitudes about immigration, at least in relation to the economic and cultural impact, have become more positive, decreasing by two points on an 11-point scale.
Taking these three trends together suggests a general picture of reduced support for the welfare state, reduced concern about inequality and increasing concern about immigration between 1995 and the early/mid 2000s, levelling off in the period 2005-2010/13, followed by a reversalmore dramatic in the case of Welfarism and less pronounced but still significant in relation to Inequality and Immigration.
In the next section, we investigate the extent to which these trends in attitudes were shared in common by groups with different characteristics and economic fortunes over the period.  Figure 3a shows that non-disabled people were more anti-welfare than disabled people at the beginning, middle and end of the period. There was a period of divergence in-between. Between 2000 and 2010 respondents with a disability diverged from non-disabled respondents, becoming even more pro-welfare in their attitudes. The gap between the two groups then narrowed again by 2015 as non-disabled respondents "caught up" with their disabled counterparts, becoming more pro-welfare after 2010. The gap between the two groups is slightly wider in 2018 compared to 1995, though this difference is only marginally significant. Figure 3b shows there has been a convergence of attitudes towards the welfare state across age groups over the period. Up to 2010, younger age groups broadly became more anti-welfare, whilst older age groups broadly became more pro-welfare. Then from 2010 onwards, all age groups became more pro-welfare and more similar to each other in their attitudes compared with the start of the period. In 1995, those aged 55-74 and 75+ were more anti-welfare compared with younger age groups, but by 2018, none of the age groups are statistically different from each other. The interaction effects show that there was a significant narrowing of the attitudinal gap between the two older age groups and the youngest age group over the period. This presents a contrast to the more familiar narrative of a "generational divide" in attitudes, which is created when raw differences between age groups are presented, without controlling for education and other characteristics.
A similar story of convergence is found between lone parents and couple parents. In 1995, lone parents had distinctively more pro-welfare attitudes compared with parents living with a partner and respondents without children ( Figure 3c). However, all three groups became more anti-welfare in the 1990s and attitudes converged in 2015 as lone parents became more anti-welfare and more similar in their attitudes to couple parents. Over this period, the proportion of lone parents with some paid work rose from 44% (in Spring 1996) to 67% (in Spring 2018), converging with employment rates of couple parents (Office for National Statistics, 2021), and this may underlie their converging attitudes, despite the fact that a high proportion of lone parents in work continue to receive tax credits/Universal Credit. By 2018 both lone parents and couple parents became more positive in their attitudes about the welfare state and had more similar attitudes to each other than in 1995, though the difference between the years is only marginally significant. Figure 3 follow a pattern of parallel trends. There has been no significant change over this period in the Welfarism attitudinal gap by country/region, controlling for other characteristics. By education, degree holders started the period with more pro-welfare attitudes than less well-educated respondents but became more anti-welfare from 1995 onwards, converging with other groups, such that by 2010, they were no longer statistically significantly different. They diverged again after that, as their attitudes became more prowelfare faster and to a greater extent than non-degree holding respondents. Nevertheless, the gap in anti-welfare attitudes by education is not significantly different in 2018 than in 1995 (see Appendix 3 in Data S1).

| Concern about inequality
Attitudes about inequality can also be characterized as parallel trends between groups or in some cases convergence after a widening of the attitudinal gap in the middle of the period. The attitudes of disabled and non-disabled respondents towards Inequality follow a similar trend to their attitudes towards Welfarism (see Appendix 3 in Data S1 for all results). Although there is a slight divergence in attitudes towards Inequality between disabled and non-disabled people in the middle of the period, by 2018, the attitudinal gap is not statistically significantly larger than it was in 1995. Disability is not a more important predictor of attitudes about inequality in 2018, controlling for other characteristics, than it was in 1995. For all other characteristics, the interaction effects show attitudes across groups changed in parallel, with no widening of the differences in attitudes. This is the case for different age groups, levels of education, country/region and household composition.

| Attitudes towards immigration
Finally, there is no sign of increased division in attitudes to immigration according to our results. Degree holders stand out as having the least anti-immigration attitudes at all three time-points (Figure 4), but the trend is similar for all groups and the attitudinal gap remains the same throughout the period (see Appendix 3, Table 12 in Data S1).
The trend towards more anti-immigrant attitudes in the mid-2000s that we noted in the previous sectionand the moderating of attitudes thereafterwere driven by the youngest (16-34) and oldest (75+) age groups (after controlling for education) (not shown in the Figure). The middle-aged groups demonstrate more stable attitudes over time in this respect. An initial divergence between young and middle-aged in 2003 was reversed by 2013, as confirmed by the interaction coefficients.
Finally, England (excluding London) and Wales appear to diverge in their attitudes towards immigration from London and Scotland in the first part of the period, and this is statistically significant as far as the gap between Scotland and the rest of England is concerned (Appendix 3, Table 14 in Data S1). But once again, this is corrected in the second period, with Scotland moving on an exactly parallel trend with Wales and converging with the rest of England by 2013.

| DISCUSSION AND CONCLUSION
The search for evidence of divergence in attitudes in Britain between groups defined by characteristics prominent in narratives of polarization, such as age, skill-level and broad geography, has yielded very little support. Neither have we found characteristics less frequently discussed but associated with experience of economic hardship and exposed to public spending cuts such as disability and parenthood status to be consistently associated with a divergence in attitudes towards welfare or inequality. Nevertheless, there are more subtle differences between groups, attitudes and trends which belie any simplistic summary.
There are of course some limitations to the analysis we have reported. The three attitudinal indices can give only partial insight into the full complexity of respondents' views and do not capture the nuances that qualitative or F I G U R E 4 Trends in predicted immigration index score by education. Higher scores indicate stronger anti-immigration attitudes.
Note: Predicted attitudinal index scores are plotted for interactions between the selected characteristic and year. Controls are as shown in Figure 1. Predicted scores are indicated by the spot and 95% confidence intervals by the whiskers. Source: authors' calculations using BSA 1995BSA -2018 deliberative approaches can (Taylor-Gooby, Leruth, & Chung, 2019), nor do they allow us to gauge the salience of the different issues in people's overall political judgements. We are constrained by the years and indicators available in the dataset and cannot be sure how observations within our 5 year (or in the case of Immigration, 8 to 10 year) windows would affect the trends we measure, although sensitivity testing for different years and measures gives some reassurance on that point. The inclusion of additional socio-demographic characteristicssuch as ethnicity or more fine-grained geographical markersnot available given the sample size could reveal axes of difference or divergence that we have not been able to consider. These point to avenues for further research, if new data become available. Nevertheless, the analysis of three attitudinal indices comprising 17 individual items measured on a reasonably consistent basis in a nationally representative sample over a period of up to 23 years provides a solid evidence base from which to draw some conclusions.
Overall, we find that some groups who are disadvantageddisabled people, lone parentshave more prowelfare attitudes and are more concerned about inequality but othersthe young, the less well-educatedare not.
We do not find that disadvantaged groups have diverged from others in their attitudes regarding the welfare state, inequality or immigration. Indeed rather than the polarization and deepening divides in attitudes between people with different characteristics and experiences that political developments in Britain might lead us to expect, the trends we detect pre-pandemic are either of stability or convergence. These findings have implications for other countries that have undergone similar social and economic changesrising inequality, rolling back of social security and increases in immigrationover the last quarter of a century. They bring into question prominent theories about attitudes based on experiences of such changes, including theories of being "left behind" economically as a driver of division in public attitudes towards welfare or immigration (Goodwin & Heath, 2016), or of increasing economic inequality resulting in attitudinal polarization (Walker & Corbett, 2019).
How can we make sense of these results? The first implication is that we may need to look for different explanations for trends in attitudes overall, for differences in attitudes between groups at a point in time and for whether those differences increase or decrease over time. It is tempting to search for a single underlying driver of these phenomena ("immigration" or "inequality" or "generational divides"), but the evidence presented here suggests they should not be conflated. The trend in the average Welfarism index, for example, is broadly consistent with the cyclical pattern or "thermostat effect" described by Pearce and Taylor (2013), with support being withdrawn during periods of expansion of welfare spending and restored during periods of austerity. The increase in hostility to immigrants, as measured by the average Immigration index, coincided with a period of increased net and gross migration to the UK (Office for National Statistics, 2018), and both the hostility and the level of net and gross migration levelled off at a similar time, consistent with the observation that rapid population changeor possibly media reporting of itis perceived as threatening by some (Quillian, 1995). However, neither public spending nor immigration provides plausible explanations for the differences we observe between disabled and non-disabled people or between low-skilled and high-skilled individuals, nor for how these differences have evolved over time. Some of the groups most exposed to gains and losses from welfare spending, or from immigration, exhibit stronger support for the welfare state, whilst others do not, and some have converged in their attitudes over time with more advantaged groups, whilst others have remained stable.
The second implication is that it is mistaken to lump together attitudes towards the welfare state and concern about inequality, let alone attitudes to immigration. These are three independent outcomes and neither the trends over time nor the differences between groups follow consistent patterns. This should hardly come as a surprise, yet this complexity is inconsistent with the attempt to characterize the country as a "divided Britain" or the period as one of "increasing polarization". Thus while the trend in overall net migration presents a rough correspondence to the overall trend in attitudes to immigration, it is of no help at all in explaining the trend in attitudes to the welfare state or the trend in concern about inequality. There is no support here for the wider thesis that immigration undermines pro-welfare attitudes (Eger, 2010;Mau & Burkhardt, 2009). Differences between groups cut across one another when multiple outcomes are considered, rather than aligning neatly to define a "pro-welfare, pro-immigration, concerned about inequality" pole and its opposite. The evidence we have presented is for Britain, but the observation that different public attitudes relevant for the context of social policy-making do not follow the same trends holds across many European countries (Meuleman, van Oorschot, & Laenen, 2020;Svallfors, 2011), and we suspect that the same would be true for the non-alignment of differences between groups across attitudes.
This analysis has demonstrated that attitudes about the welfare state, inequality and immigration are distinct: they follow different trends over time and have different associations with socio-demographic characteristics, and these associations play out differently over time across the different attitudes. What is clear from this is that there is not a simple story of polarization of attitudes. Over-stating the divisions in public attitudes towards key institutions of the welfare state and the underpinnings of social solidarity, whether for dramatic effect or through over-simplification of the evidence, has not served policy analysis well. Taking the long view of trends in attitudes, and examining a range of potential socio-economic and demographic stratifiers, we can see that even in countries like Britain which have experienced high levels of inequality, harsh austerity and significant immigration, there is considerable shared "attitudinal capital" to be drawn on. This has undoubtedly been a strength in the context of policymaking in response to the many challenges created by the pandemic.
Given the overall trends in attitudes pre-COVID, there is greater room attitudinally for more pro-welfare and inequalityreducing policies now than was the case, for example, in 2010. Progressive policymakers have a window of opportunity to capitalize on this space and should not be deterred by exaggerated accounts of division and polarization.

ACKNOWLEDGEMENTS
We are grateful to Nuffield Foundation for funding this research as part of the Social Policies and Distributional Outcomes in a Changing Britain research programme. We would like to thank Fran McGinnity, Howard Glennester, Polina Obolenskaya, Polly Vizard, John Hills, Ruth Lupton and Mark Franks in particular for their helpful feedback on earlier versions of the paper. The data used are copyrighted to NatCen and deposited in the UK Data Service. We are grateful to NatCen for the use of the data and UKDS for making them available, as well as to the respondents of the survey. Responsibility for errors and interpretation remain with the authors.
2 With "liberal" values associated with being supportive of redistribution and emphasizing structural factors in explaining why some people are in poverty, in contrast to sceptics' who tend to emphasize personal behaviour in explaining people's poverty.
3 Though we are interested in disadvantaged groups we do not measure income in the analysis as the relationship between low income and a range of attitudes has been analysed in detail by others (Dunatchik et al., 2016;Sumino, 2018;Taylor, Saunders, & Toomse-Smith, 2017) and for most of the years in our analysis the income measures in the BSA are not ideal.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article.
How to cite this article: Cooper, K., & Burchardt, T. (2021). How divided is the attitudinal context for policymaking? Changes in public attitudes to the welfare state, inequality and immigration over two decades in Britain. Social Policy & Administration, 1-18. https://doi.org/10.1111/spol.12739