Introduction
The last few electoral cycles in Europe have been unkind to the continent's historic centre-left, with social democrats in Austria (2019), Belgium (2019), France (2017), Germany (2017), Ireland (2020), and the Netherlands (2019) all registering post-war electoral nadirs. Britain's Labour Party is no exception. In 2019, Labour gained its lowest seat share since 1935, losing several symbolically important ‘heartland’ constituencies held almost continuously since the adoption of universal suffrage (Goes Reference Goes2020). Autopsies of the campaign revealed disproportionate losses among manual workers and those with low incomes (Fieldhouse et al. Reference Fieldhouse2023; Goodwin and Heath Reference Goodwin and Heath2020; McDonnell and Curtis Reference McDonnell and Curtis2019); painful statistics considering that these were the exact groups for whom the party was founded to represent (Labour Party 2021a; Thorpe Reference Thorpe2015, 8–35). For many, Labour had alienated its working class base once too often in recent decades, with equivocation over ‘Brexit’ being the final straw (Bickerton Reference Bickerton2019; Goodwin and Heath Reference Goodwin and Heath2020). Yet, the narrative of a working class exodus from Labour perhaps oversimplifies things. Among UK parliamentary constituencies in the top quintile of deprivation – a composite measure encompassing average attainment in income, employment, housing, etc. – Labour did indeed lose traditional, underprivileged strongholds like Stoke-on-Trent North and Bishop Auckland. However, it held others, such as Bradford West and Bethnal Green, with increased majorities (Rae Reference Rae2021; Wainwright Reference Wainwright2019). There is one obvious difference between these four working class constituencies: while the first two are over 90 per cent white, the latter are over 50 per cent black and Asian (Barton Reference Barton2020). Data from the British Election Study Internet Panel (BESIP) following the 2019 election establishes an ethnic divide among the working class. Among lower technical, routine and semi-routine employees (encompassing service workers such as cleaners and carers, as well as industrial workers), Labour support was around 59 per cent among ethnic minorities, but only 29 per cent among whites.Footnote 1 Labour failed to connect with the white working class (WWC) specifically.
This paper investigates whether perceptions of the party's representation of different demographics can help explain this puzzle. Evidence suggests that Labour has maintained its reputation as the party for minorities more successfully than its reputation as the party for the (white) working class (Heath et al. Reference Heath2013, 13). Others argue that Labour's focus on ethnic minority interests (among other ‘new’ social movements) has reduced its standing among WWC voters (Ehsan and Stokes Reference Ehsan and Stokes2021; O'Neill Reference O'Neill2021). Similar claims have been made about the Democratic Party in America (Fukuyama Reference Fukuyama2018; Lilla Reference Lilla2016). These authors allege a shift in the priorities of social democrats from addressing class-based economic issues to ‘identity politics’ and the mobilization of a range of sociocultural grievances associated with historically marginalized, principally non-economic, social groups, such as ethnic minorities (Berman and Snegovaya Reference Berman and Snegovaya2019). This has, the argument goes, eroded attachment to the centre-left among ‘traditional’ supporters, who view efforts to represent such groups as, at best, competition for attention and, at worst, actually harmful to their interests (Lilla Reference Lilla2016).
This argument echoes Przeworski and Sprague's (Reference Przeworski and Sprague1986) identification of a ‘dilemma of electoral socialism’. Reaching out to middle class voters in response to the declining size of their working class base would reduce the latter's loyalty to the centre-left by eroding the parties' pro-working class brand (see also Evans and Tilley (Reference Evans and Tilley2017), Heath (Reference Heath2015), and Thau (Reference Thau2021)). Thus far, however, few, if any, have explored whether perceptions of social democratic attempts to represent non-whites have had a similarly demobilising effect on WWC support. Is this a new dilemma of electoral socialism? Do WWC voters view Labour as disproportionately focused on minorities to the exclusion of their interests? While feelings of displacement have been ascribed to WWC subjects in influential ethnographies (Gest Reference Gest2016, 39; Hochschild Reference Hochschild2016, 137–39), we are unaware of any study that has investigated these questions directly.
We proceed as follows. First, we note that Labour's electoral coalition has changed from one dominated by the WWC to a tripolar alliance highly reliant on non-whites and white graduates. We hypothesize that the WWC, more so than their graduate co-ethnics, may react negatively to the party's greater association with non-white representation. We distinguish between ‘ethnocentrism’ and ‘relative political deprivation’ based explanations for this. We then test these claims using BESIP data. Our findings partially support the thesis that associations with minority interests can undermine social democrat WWC support. First, the WWC are more likely than others to say that Labour represents minorities better than the working class. Second, subjective ratings of how well Labour represents both groups are less positively correlated among WWC respondents. Third, net of several attitudinal controls, the WWC are less likely to support Labour when they associate the party with minority interests. These results cannot be fully explained by the WWC's higher average levels of ethnocentrism. Instead, we argue that a substantial proportion of the WWC believe they have been relatively deprived of representation vis-à-vis other Labour coalition members in recent years and have reacted negatively. We conclude by discussing the implications of our findings and the extent to which they might be generalizable.
A New Dilemma of Electoral Socialism?
Analysts of British politics often claim that Labour, like other European centre-left parties, is struggling to hold together its twenty-first-century coalition of, on the one hand, industrial and lower-skilled service workers and, on the other, an increasingly large cohort of university graduates (Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2020; Ford and Goodwin Reference Ford and Goodwin2017; Gingrich Reference Gingrich2017; Harris Reference Harris2022; Oesch and Rennwald Reference Oesch and Rennwald2018; Piketty Reference Piketty2020). Following Sobolewska and Ford (Reference Sobolewska and Ford2020, 70–3), we conceptualize Labour's support as having changed over time from domination by the (formerly, overwhelmingly white) working class to a ‘tripolar’ coalition between a reduced white working class (WWC), and also white graduates and ethnic minorities.
We define the first group as white, lower technical, routine or semi-routine employees without a university degree; the second as white university graduates and current students; and the last as all non-white or mixed-race respondents regardless of occupation or education.Footnote 2 Using nationally representative British Election Study (BES) data from 1987 and 2019 (two defeats where the party gained roughly one-third of votes nationally), Fig. 1 details the percentage of the entire electorate and Labour voters who belonged to each group in either year. In 1987, over half of Labour's votes came from the WWC, despite this group making up less than two-fifths of the overall electorate. Graduates and non-whites – despite the latter being heavily pro-Labour – collectively contributed just 17 per cent of the party's votes due to their relative scarcity. In 2019, after three decades of post-industrialization, educational expansion, rising diversity, and, consequently, shifting Labour Party strategy (Evans and Tilley Reference Evans and Tilley2017; Gingrich Reference Gingrich2017; Sobolewska and Ford Reference Sobolewska and Ford2020), this had all changed. While these three groups collectively contributed over two-thirds of the party's support, just as in 1987, the relative contribution of each had shifted dramatically. Due to dwindling numbers, rising abstention, and increased support for rival parties, only 18 per cent of Labour supporters were from the WWC, a statistical underrepresentation. By contrast, the enlarged white graduate and minority voter blocs were now overrepresented, making up 26 per cent and 13 per cent of the entire British electorate but 38 per cent and 20 per cent of Labour supporters. Labour had moved from near dependency on a single constituency to a diverse, ‘tripolar’ coalition.
Labour's tripolar coalition is large but potentially fragile, given that its members differ in their values. In line with previous studies (Evans and Tilley Reference Evans and Tilley2017, 83; Harris Reference Harris2022; Sobolewska and Ford Reference Sobolewska and Ford2020), Appendix 1 shows that the three groups are ideologically similar on economic left-right issues, but the WWC are notably more authoritarian and ethnocentric. The relative social conservatism of ethnic majority working class populations has long been recognized (Lipset Reference Lipset1959). However, many analysts have highlighted that, as the political salience of immigration and international integration has risen, tensions between social democratic parties and their base have become full-on schisms, and some working class voters have drifted toward conservative and right-wing populist parties that better represent their positions on these issues (Ford and Jennings Reference Ford and Jennings2020; Hooghe and Marks Reference Hooghe and Marks2018; Kriesi et al. Reference Kriesi2008; Oesch and Rennwald Reference Oesch and Rennwald2018; Rydgren Reference Rydgren2013; although, see Abou-Chadi, Mitteregger, and Mudde Reference Abou-Chadi, Mitteregger and Mudde2021 for a critique of this narrative). Indeed, Labour's WWC problem can partially be explained by their recent positions on several controversial ‘sociocultural’ issues (Ford and Goodwin Reference Ford and Goodwin2017). Sobolewska and Ford (Reference Sobolewska and Ford2020) and Evans and Tilley (Reference Evans and Tilley2017, 83) demonstrate that the working class have long displayed Euroscepticism and opposition to immigration, putting them at odds with the rhetoric and policy of recent Labour leaders and most graduate voters. Given that Britain's ethnic minority population are more liberal on border regulations and were notably less likely than whites to support leaving the EU (Clarke, Goodwin, and Whiteley Reference Clarke, Goodwin and Whiteley2017, 154–5; Heath et al. Reference Heath2013, 66–7), one could surmise that the black and Asian working class were less disillusioned with Labour's increasing internationalism since the 1980s.
That said, particular divisive issues are only partially explanatory. First, Labour's leadership and positions have always been more socially liberal than their prototypical supporters. In the 1960s, Labour was far more popular among the working class than the more socially tolerant middle classes, despite abolishing the death penalty, legalizing homosexuality and abortion, and enacting several landmark race relations acts (Evans and Tilley Reference Evans and Tilley2017, 68–79; Sobolewska and Ford Reference Sobolewska and Ford2020, 190–8; Thorpe Reference Thorpe2015, 173–80). Second, Labour's social liberalism poorly serves its average non-white supporter in other ways. While ethnic minorities are relatively permissive on issues of immigration and policing, they are more conservative on questions of gender and homosexuality (Collins, Drinkwater, and Jennings Reference Collins, Drinkwater and Jennings2020, 23; Sobolewska and Ford Reference Sobolewska and Ford2020, 70–2), yet there has been no similar anti-Labour backlash from this community (Martin Reference Martin2019).
We do not deny that the role of high-profile ‘issues’ is a challenge that Labour and other social democratic parties face in trying to maintain their modern electoral coalition. However, considering the shifting balance of power within the party's electorate, we contend there is also the potential for an even deeper conflict over exactly who Labour fundamentally belongs to, whose interests it should prioritize, and how achievable positive sum representation of different social groups is in practice. In a nutshell, we argue that the rising importance of non-white voters to Labour's overall vote share and subsequent increased efforts by Labour to portray itself as the party for ethnic minorities may alienate WWC voters.
This perspective is not entirely novel. Popular understandings of parties as defenders of specific social groups rather than custodians of particular ideological traditions have long been acknowledged (Butler and Stokes Reference Butler and Stokes1974; Campbell et al. Reference Campbell1960), and such theories have resurged in recent years (Huddy Reference Huddy, Sears and Levy2013), especially within the USA where associations of ‘Democrat’ or ‘Republican’ with particular ethnoreligious identities are deeply ingrained (Achen and Bartels Reference Achen and Bartels2017). Put simply, awareness of the ‘sort of people’ that each party tends to help is often more widespread than accurate knowledge of politicians' stances, ideology, and policies (Achen and Bartels Reference Achen and Bartels2017; Butler and Stokes Reference Butler and Stokes1974, 336–7; Campbell et al. Reference Campbell1960; Smith Reference Smith2019), and assessments of the links between parties and social groups seem to help shape political identities (Ahler and Sood Reference Ahler and Sood2018). When a voter believes that a party is more closely associated with an ‘outgroup’ they dislike – whether because they deem it a rival for resources or because of base prejudice – the likelihood of them supporting the said party declines (Kane, Mason, and Wronski Reference Kane, Mason and Wronski2021; Mason, Wronski, and Kane Reference Mason, Wronski and Kane2021; Zhirkov and Valentino Reference Zhirkov and Valentino2022).
In consequence, the public's understanding of who ‘fits’ where politically is not just influenced by policy and high profile ‘issues’ but also by cheaper heuristics such as a party referring to one's group in speeches, nominating co-group candidates and, crucially, avoiding associating itself with the representatives, emblems, and stances of rival groups (Evans and Tilley Reference Evans and Tilley2017; Heath Reference Heath2015; Lupu Reference Lupu2016; Robison et al. Reference Robison2021). This makes group voting distinct from policy voting per se. For instance, in Europe, social democrats' appeals to the growing middle classes are argued to have undermined working class perceptions that these were parties ‘for them’ (Evans and Tilley Reference Evans and Tilley2017; Przeworski and Sprague Reference Przeworski and Sprague1986). It was not just more centrist policies on tax and redistribution that demobilized the working class, however, but also the changing demographic make-up of the party elite and their abandonment of references to workers' lifestyles and challenges in their rhetoric (Evans and Tilley Reference Evans and Tilley2017; Heath Reference Heath2015; Robison et al. Reference Robison2021; Thau Reference Thau2019, Thau Reference Thau2021). In Britain, the change was dramatic. While widespread in the mid-twentieth century (Butler and Stokes Reference Butler and Stokes1974), perceptions of Labour as a party for the working class declined substantially during the 1990s and 2000s (Evans and Tilley Reference Evans and Tilley2017, 43; 130–5).
Having previously alienated sections of its working class electorate by appearing too closely linked with middle class aspirations and interests, might Labour be in danger of losing the remainder over its appeals to minorities? Labour has certainly redoubled efforts to appeal to the non-white electorate. The party greatly expanded their stock of minority MPs, established and granted executive committee representation to an affiliated society for all non-white members (‘Labour BAME’), and made increasingly explicit appeals to the challenges facing non-white Britons, including vocal support for the Black Lives Matter campaign (Gregory Reference Gregory2020; Labour Party 2021b; BBC News 2019). Critics argue that Labour's alleged fixation with non-white voters undermines its ability to adequately represent its traditional constituencies, with the relative paucity of references to the working class in recent party publications being highlighted as indicative (Ehsan and Stokes Reference Ehsan and Stokes2021). Working class understandings of representation as a zero-sum game may also have been enhanced because minority representation in parliament accelerated when the number of WWC MPs declined precipitously (Campbell and Heath Reference Campbell and Heath2021, 1; 710–11). Whatever the exact cause, in 2010, only 63 per cent of white British people believed that Labour looked after working class interests ‘very’ or ‘fairly’ well, whereas 80 per cent said this of the party's representation of black and Asian people (Heath et al. Reference Heath2013, 93–4).
A non-trivial proportion of whites saw Labour as a party for ethnic minorities rather than the working classes per se, but did this erode attachments to Labour? Surprisingly, the relevance of a new (racialized) dilemma of electoral socialism has not been directly tested. True, several observational and experimental studies have examined the impact of socially liberal positions on race and immigration on centre-left support, with some reporting a positive effect (Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2020; Häusermann et al. Reference Häusermann2021) and others uncovering more negative outcomes (Angelucci and Vittori Reference Angelucci and Vittori2023). However, these studies cannot answer our research questions because they measure responses to observed or experimentally manipulated positions on policies rather than assessments of how well parties represent different groups. More relevantly, Dancygier (Reference Dancygier2010) argues that the potential political power of non-whites can exacerbate ethnic tension because natives may fear that local politicians will no longer be responsive to their needs. However, neither that nor her subsequent research on the ‘inclusion dilemma’ facing centre-left parties who field Muslim candidates (Dancygier Reference Dancygier2017) explicitly measure the extent to which native working class populations associate these parties with minorities. We fill this gap by directly measuring voters' associations of the Labour Party with different social groups. We hypothesize that the association of the centre-left with non-whites will erode their WWC support; however, we also advance previous discussions theoretically by distinguishing between two potential mechanisms.Footnote 3 First, underlying hostility to other racial groups (ethnocentrism) and, second, fear of losing representation of one's own group (relative political deprivation).
Why Might Minority Representation Reduce the Centre-Left's White Working Class Support?
Ethnocentrism
Ethnocentrism is the tendency ‘to reduce society to us and them’ (Kinder and Kam Reference Kinder and Kam2009, 37) and to combine ‘a positive attitude toward one's own ethnic/cultural group … with a negative attitude toward the other’ (Adorno 1950 in Motyl Reference Motyl2001, 153–4). Ethnocentric voters may simply react negatively to Labour's attempts at improving the standing of non-whites due to outgroup prejudice. Ethnocentrism is a reliable predictor of political behaviour in America, particularly increasing Republican support among lower-status whites (Kam and Kinder Reference Kam and Kinder2012; Kinder and Kam Reference Kinder and Kam2009). Furthermore, animus towards social groups commonly associated with a rival party – which, in America, often means particular races or religions – has been shown to drive the recent hardening of partisan division and polarization in US politics (Kane, Mason, and Wronski Reference Kane, Mason and Wronski2021; Mason, Wronski, and Kane Reference Mason, Wronski and Kane2021; Robison and Moskowitz Reference Robison and Moskowitz2019; Zhirkov and Valentino Reference Zhirkov and Valentino2022). Conversely, ethnocentrism is rarely measured in Britain, where the small size and heterogeneous nature of the minority population, in addition to its early enfranchisement, has historically limited the political capital available from mobilizing racial grievances (Laniyonu Reference Laniyonu2019; Sobolewska and Ford Reference Sobolewska and Ford2020, 15–16). That said, given evidence of racial prejudice in contemporary Britain (Sobolewska and Ford Reference Sobolewska and Ford2020, 61) and accusations of racism being commonly levelled at critics of social movements such as Black Lives Matter (Mason Reference Mason2021), it is prudent to control for the possibility that ethnocentrism drives WWC suspicion of minority representation.
Any gap between the WWC and white graduates in their response to associations of Labour with non-white representation should dissipate once controlling for ethnocentrism if that is the cause. Firstly, graduates are often less attached to their own ethnic ‘in-group’ than the working class, possibly because a higher social status reduces the psychological need to base one's identity and esteem on the achievements of the broader national community (Kaufmann Reference Kaufmann2018; Shayo Reference Shayo2009; Sobolewska and Ford Reference Sobolewska and Ford2020, 66–8). Secondly, copious research links university education with reduced hostility to ethnic ‘out-groups’, although the exact mechanism is disputed (Scott Reference Scott2022; Sobolewska and Ford Reference Sobolewska and Ford2020, 66–8).
Relative Political Deprivation
Relative deprivation is a psychological concept denoting feelings of resentment and frustration that emerge when one believes that one's in-group has been unfairly deprived of resources relative to others (Gurr Reference Gurr1970; Siroky et al. Reference Siroky2020; Stouffer et al. Reference Stouffer1949). While these resources are often material, such as jobs and housing (Dancygier Reference Dancygier2010; Gurr Reference Gurr1970; Pettigrew et al. Reference Pettigrew2008), it is possible that perceptions of another group enjoying greater responsiveness from (ostensibly) one's own representatives could catalyze specific feelings of relative political deprivation. Members of the WWC, in whose interests the party was founded and from whose ranks many party representatives were once sourced (Evans and Tilley Reference Evans and Tilley2017; Labour Party 2021a; Thorpe Reference Thorpe2015, 8–35), may feel aggrieved that a ‘new’ social group appears to be usurping their position as Labour's primary constituency. Moreover, particularly following the loss of descriptive and substantive representation to the middle classes since the 1980s (Evans and Tilley Reference Evans and Tilley2017; Heath Reference Heath2015), it is possible that WWC voters may come to interpret social group representation in zero-sum terms, where any further attention given to other groups comes directly at their expense (Norton and Sommers Reference Norton and Sommers2011). Such feelings would be compatible with the broader sense of displacement and of having fallen behind others ‘cutting in line’, as ascribed to the subjects of Gest (Reference Gest2016, 39) and Hochschild's (Reference Hochschild2016, 137–9) ethnographies of the WWC in East London and rural Louisiana, respectively.
Crucially, we do not expect the rest of Labour's electoral coalition – non-whites and graduates – to necessarily perceive a trade-off or react negatively when they associate Labour with minority interests. Black and Asian voters may simply view the ‘working class’ and ‘ethnic minorities’ as largely synonymous rather than as rivals. Even if they do not, they have a clear self-interest in Labour positioning itself foremost as a party for non-whites. White graduates have several reasons to be more comfortable with Labour's focus on ethnic minority interests (apart from their aforementioned social liberalism). First, they should not personally feel any sense of political deprivation vis-à-vis minorities: in recent decades, substantive and descriptive representation of graduates within the party has increased substantially (Evans and Tilley Reference Evans and Tilley2017; Quilter-Pinner et al. Reference Quilter-Pinner2022). Second, left-wing critics of ‘identity politics’ have argued that wealthier groups may even derive material benefits from social democrats associating more closely with minorities than the working class, given that they have less to fear from a politics centred on symbolic gestures and representation rather than competing economic interests (Lilla Reference Lilla2016; Walters Reference Walters2006). Third, experimental evidence from America suggests that whites who are cognisant of the historical injustice and barriers facing minorities can become less sympathetic to problems facing poorer members of the ‘advantaged’ ethnic majority (Cooley et al. Reference Cooley2019). Accordingly, higher-status whites might believe the prioritization of blacks and Asians over the WWC is ethically justifiable. Accordingly, WWC concerns about relative political deprivation are unlikely to be shared by other members of Labour's coalition. Indeed, research indicates that graduates are generally more supportive of increasing the number of Muslims and non-whites in parliament than are working-class identifiers (Campbell and Heath Reference Campbell and Heath2021, 1719).
Accordingly, our hypotheses are as follows. Relative to white graduates and ethnic minorities, we expect that:
H1: White working class (WWC) voters are more likely to perceive Labour as better representing ethnic minorities than the working class.
H2: WWC voters are more likely to perceive a trade-off between minority and working class interest representation.
H3a: WWC voters are less likely to support Labour if they perceive the party to be closely associated with minority representation.
H3b: WWC voters are particularly less likely to support Labour if they perceive the party to be more closely associated with the representation of minorities than with working class interests.
And, lastly, if our thesis about relative political deprivation is accurate:
H4: The distinctiveness of WWC voters in H1–H3 is not reducible to their higher levels of ethnocentrism.
Measurement and Modelling
To test our hypotheses, we must know whether individuals associate Labour with particular social groups and their level of ethnocentrism. Suitable measures of both are available in the BESIP study (Fieldhouse et al. Reference Fieldhouse2021), a longitudinal online survey that has, since 2014, been periodically fielded to nationally representative samples of around 30,000 British adults recruited by the leading British internet polling company, YouGov.Footnote 4
Perceptions of Labour's ability to represent particular groups are taken from successive BESIP items asking respondents ‘how closely [do they] think the Labour Party looks after the interests of’ “working class”, “black and Asian”, and, to help capture the “classic” dilemma of electoral socialism, “middle class” people, from 0 (“not at all closely”) to 3 (“very closely”). Ideally, we would measure the perceived representation of ‘white working class people’ specifically, but, unfortunately, no such survey item exists. That said, our general ‘working class’ item is a suitable proxy. As bemoaned by left-wing activists, class in Britain is regularly discussed as a particular (racialized) cultural tradition (flat caps, northern accents) rather than as an economic classification, with ethnic minorities often treated as a classless collective outside this group rather than (an increasingly large) part of it (Sandhu Reference Sandhu2018). Similarly, Labour's appeals to a distinct ethnic minority community (Labour Party 2021b) may also encourage perceptions that this group fundamentally differs from the party's conventional clientele (that is, the working class). Furthermore, asking a separate question about perceptions of ‘black and Asian interest’ representation may encourage respondents to focus on the white population when considering working class representation. When modelling the effects of Labour's perceived representation of minorities relative to the working class (rather than perceptions of absolute representation of either group), we subtract the ratings of working class representation from evaluations of how well Labour represents blacks and Asians. This generates a −3 to +3 scale where higher values indicate that representation is relatively biased towards minorities.Footnote 5
While these questions were fielded to some respondents in several waves, we mainly use the measures taken in Wave 10 of the BESIP (December 2016), as this is proximate to the wave containing our measure of ethnocentrism (see below) and, unlike other waves, it also fielded the questions on representation to the full sample. In the final part of our analysis, however, we use multiple waves to conduct a longitudinal analysis to demonstrate the robustness of our findings. Appendix 4 presents analyses using a pooled sample of all BESIP waves containing questions about perceived group representation (2016–2021) to demonstrate that our models that do not include the ethnocentrism control are fully replicable when using a broader sample.
The absence of ethnocentrism items in most British social science surveys has been bemoaned (Sobolewska and Ford Reference Sobolewska and Ford2020, 338). While immigration items are common, they are not ideal. Firstly, immigration preferences are also affected by non-racial considerations about economic ramifications (Dancygier Reference Dancygier2010). Secondly, preferences for restricting new arrivals may themselves be driven by perceptions of oneself as relatively deprived compared to the existing population. Conversely, ethnocentrism is usually conceived of as a stable, underlying personality trait leading one to prefer in-group members to out-groups, independent of short-term factors (Motyl Reference Motyl2001, 153–4; Sobolewska and Ford Reference Sobolewska and Ford2020, 36–41).
Fortunately, in Wave 11 (April-May 2017), the BESIP respondents were asked to ‘rate [their] feeling towards people from different groups born in Britain’ – white people, Asian people, and black people – on a scale of 0 (‘very cold’) to 100 (‘very warm’), following similar longstanding items in the American National Election Study. Following Hainmueller and Hopkins (Reference Hainmueller and Hopkins2015, 542) and Kinder and Kam (Reference Kinder and Kam2009, 42–61), we averaged the respondents' feelings towards blacks and Asians and subtracted this figure from the score given to whites to capture both positive in-group and negative out-group sentiment. Dividing this product by 10 produced a final variable ranging from −10 to +10, where higher values indicate greater closeness to whites relative to non-whites. If the association of Labour with minorities substantially repels WWC voters, even when controlling for ethnocentrism, this suggests that relative political deprivation may be important. We provide evidence for the construct validity of our measure of ethnocentrism in the Appendix and demonstrate its broader utility as a predictor of British political behaviour.Footnote 6 Kinder and Kam (Reference Kinder and Kam2009, 66–8) also provide evidence of this measure's within-person over-time stability in the United States. Appendixes 12–17 demonstrate the robustness of all our results to controlling for attitudes to immigration or the European Union instead.
Again, our analysis distinguishes between white (occupationally) working class, white graduates, ethnic minorities, and ‘other white’ respondents.Footnote 7 All our models control for the respondents' age group (18–25; 26–44; 45–64; 65+), gender, and country (England; Wales; Scotland).Footnote 8 Where possible, we also present models controlling for respondents' values on validated economic left-right and social liberal-authoritarianism measures (Evans, Heath, and Lalljee Reference Evans, Heath and Lalljee1996). These are two additive scales of five items that capture attitudes towards inequality and class conflict, and traditional values and maintaining order, respectively. Both are predictive of vote choice, linked to class and educational attainment and, in the latter case, plausibly related to ethnocentrism (Evans and Tilley Reference Evans and Tilley2017; Kinder and Kam Reference Kinder and Kam2009; Sobolewska and Ford Reference Sobolewska and Ford2020). Appendix 23 details the construction of all variables.
In what follows, we first present descriptive evidence that the WWC are more likely to believe that Labour does a better (worse) job at representing black and Asian people (the working class) than graduates and ethnic minorities. Second, we run OLS regressions to demonstrate the association between perceptions of how well Labour represents minorities and the working class. Third, we use logistic regressions to examine whether these perceptions predict Labour support among different voters. We then examine the consequences of Labour's perceived focus on ethnic minorities and their relative focus on this group versus the working class.
Results
Do WWC respondents consider themselves relatively deprived of representation by Labour? Based on OLS regression models presented in Appendix 24, which controlled for age, gender, UK country, and survey year, Fig. 2a shows the mean predicted evaluation of how closely Labour represents working class, middle class, and black and Asian interests for WWC, white graduate, and non-white respondents, respectively, using pooled data from Waves 10, 19, 20, and 21 of the BESIP (December 2016 to May 2021). Crucially, while all three types of respondents believe that Labour represents the middle class least closely, the WWC evaluate Labour's working class representation less favourably than do minorities and graduates, and they are unique in believing that Labour most closely represents black and Asian Britons.
Figure 2b clarifies this picture. The bars here present the predicted values from an almost identical regression model in which the dependent variable was a measure of Labour's perceived relative focus on the working class or non-whites. This was calculated by subtracting evaluations of Labour's working class representation from evaluations of the party's black and Asian representation, which varied from −3 (very close to the working class, not at all close to minorities) to +3 (the reverse). In all other respects, the model specification was identical to Fig. 2a. Crucially, the WWC see Labour as relatively more ‘pro-minority’ than ‘pro-working class’; white graduates and ethnic minorities say the opposite. In accordance with H1, the WWC does, on average, see themselves as relatively deprived politically. Importantly, using data from Waves 10 and 11 (December 2016 – April 2017) of the BESIP, Appendix 25 demonstrates that the greater tendency of WWC respondents to judge Labour as closer to minorities than the working class is not, primarily, driven by higher white ethnocentrism or different left-right or libertarian-authoritarian values. Variation in all three variables can explain about 40 per cent of the gap between the WWC's ratings and those of white graduates and about 36 per cent of the difference between the ratings of the WWC and those of ethnic minority respondents.
To what extent does the WWC view Labour's social group representation in zero-sum terms? Do they think focusing on minorities harms Labour's ability to represent itself? Table 1 presents a series of OLS regression models testing this. Ratings of Labour's working class representation (‘Labour WC’) are regressed on ratings of Labour's black and Asian (‘Labour BME’) representation interacted with the respondent's social background. If WWC people are more likely to perceive a trade-off between representing either group, this interaction effect should be positive for all other groups when the WWC is the baseline. This is indeed what Model A shows. For WWC respondents, the net marginal ‘effect’ of a one-unit improvement in evaluations of Labour's BME representation is only associated with a small increase (0.11 units) in positivity about Labour's working class representation, but this relationship is 3 and 3.5 times larger for white graduates and minority respondents, respectively. Importantly, these interaction effects are robust to controlling for underlying group differences in ethnocentrism (Model B), as well as left-right and libertarian-authoritarian values and the strength of Labour partisanship (Model C). While the impact of evaluations of Labour's minority representation on its perceived working class representation is not negative for WWC respondents (which would indicate entirely zero-sum attitudes to representation), the relationship between evaluations is unusually modest.
Source: BESIP Wave 10 (Nov-Dec 2016) for all variables except ethnocentrism, which was recorded in BESIP Wave 11 (April 2017). Data are weighted using the standard BESIP Wave 10 survey weight. * p-value <0.05; ** p-value <0.01.
Note: Each model presents the results from a different OLS regression predicting evaluations of how closely the Labour Party ‘look[s] after working class interests’ (0–3). Perceived Labour BME representation is measured on the same scale.
Demographic Controls: Age, Gender, Country.
Using data from Table 1 (Model C), Fig. 3 visualizes the association between evaluations of Labour's representation of different groups for the WWC (red), white graduates (blue), and ethnic minority respondents (black), holding all else constant. Our model predicts that WWC, white graduate, and ethnic minority respondents who assigned Labour a 1/3 for how closely they represent non-whites (that is ‘not very closely’) – the first score for which we have many observations – would all be predicted to give the party around 1.5/3 for working class representation. However, if the same respondents gave the party a maximum 3/3 score for BME representation, the model would predict that white graduates and minorities would score Labour's working class representation 2.1/3 and the WWC just 1.7. Contrary to H2, therefore, it is not the case that the WWC believe that Labour cannot represent both the working class and minorities simultaneously; however, they are less likely to perceive a positive sum game than other parts of Labour's coalition. Furthermore, as per H4, this is not reducible to ethnocentrism.Footnote 9
Do people factor Labour's proximity to different social groups into their vote choice? Table 2 presents a series of logistic regression models where the dependent variable is support for Labour (1) versus any other party (0), according to vote intention in BESIP Wave 10.Footnote 10 This is regressed onto perceptions of the party's representation of blacks and Asians and the middle and working classes (Model A). All ratings are interacted with the respondent's social background, holding constant age, gender, and country. Subsequently, we add controls for ethnocentrism (Model B) and left-right and libertarian-authoritarian values (Model C).
Source: BESIP Wave 10 (November-December 2016) for all variables except ethnocentrism, which was recorded in BESIP Wave 11 (April 2017). Data are weighted using the standard BESIP Wave 10 survey weight. *p-value <0.05; **p-value <0.01.
Note: Each model presents the results from different logistic regressions predicting support for Labour (1) v. any other party (0).
Demographic Controls: Age, Gender, Country.
The models corroborate H3a: WWC respondents (and, to a lesser extent, other non-graduate whites) are less likely to support Labour when they associate the party with non-white interests; however, graduates and ethnic minorities are more likely to do so. Importantly, the differences between the WWC and other groups are barely altered when controlling for ethnocentrism and other potential attitudinal confounders. Therefore, prejudice alone is unlikely to drive the WWC's aversion to the Labour-minority nexus. Unsurprisingly, the strongest predictor of WWC support for Labour is the perception of how well the party looks after the working class. However, it is notable that, unlike minority representation, perceptions of how well Labour looks after middle class interests are also positively associated with WWC support for the party.Footnote 11 Based on the data here, it is the ‘new’ dilemma of electoral socialism and the difficulty of reconciling WWC supporters to minority representation that is more pressing for Labour than the ‘old’ electoral dilemma of appealing to both the working and middle classes (cf. Przeworski and Sprague Reference Przeworski and Sprague1986).
How large are these effects? Based on data from Table 2 (Model C), Fig. 4 presents the predicted probability of WWC, white graduate, and ethnic minority respondents supporting Labour given a particular evaluation of the party's representation of the working class (top left panel), black and Asians (top right panel), and the middle class (bottom left panel). The figure demonstrates that there is something unique about minority representation as opposed to other types of social constituencies. Moving evaluations of both working and middle class representation from their most negative to their most positive increases the likelihood of a WWC voter choosing Labour by around 49 percentage points; however, equivalent improvements in the ratings of how well Labour represents minorities produce a 44 percentage point decrease in the likelihood of WWC support for the party. The equivalent figures for white graduates are + 25, + 43, and −3 percentage points, respectively, and + 1, + 29, and + 6 percentage points, respectively, for non-white respondents.Footnote 12 Therefore, the most reliably strong predictor of respondents' proclivity to support Labour is their perception of how well the party represents middle class interests. It is plausible that this is because some respondents now view ‘middle class’ as synonymous with ‘average person’. The net positive ‘effect’ of working class representation is strong for the WWC and, to a lesser extent, white graduates. However, holding all else constant, it plays a small role in the average vote calculation of non-whites.
The generally weak associations of social group representation with non-white voting behaviour are worth noting. Previous research in Britain and America has indicated that non-white support for the centre-left tends to be very high regardless of economic status, given that minorities tend to hold the issue of non-discrimination salient above all other considerations (Dawson Reference Dawson1994; Heath et al. Reference Heath2013). It may also be that, in a majoritarian system such as that in the UK, there are few credible alternative ‘pro-minority’ parties for a non-white voter to defect to even when they deem the major social democratic party to be doing poorly in representing their interests.
In general, however, these models clearly corroborate H3a and H4. WWC respondents are most likely to perceive Labour as closer to minority interests than working class ones. But what are the electoral consequences of such beliefs? Using the same BESIP Wave 10 dataset, Models A and B of Table 3 contain logistic regressions of Labour vote intention on perceptions of the extent to which respondents think that the party more closely represents non-white interests than working class interests (−3 to + 3), where higher values denote perceived pro-minority bias. The variable is again interacted with respondents' social background. As per H3b and H4, such beliefs exert a greater negative effect on the probability of the WWC supporting Labour than they do for other groups, regardless of controls.Footnote 13 In Model C, we replicated this finding, albeit for a much smaller sample, using evaluations of Labour's relative emphasis on minority representation and vote choice in the December 2019 General Election (BESIP Wave 19). WWC respondents who believed Labour were more ‘pro-minority’ than ‘pro-working class’ were less likely to support Labour in that contest and were also less likely to do so than white graduates who believed the same thing. In Model D, we add our control for ethnocentrism, which was measured only in Wave 11 (April 2017) of the BESIP. Consequently, we further reduced our sample by half and introduced a panel weight to take account of non-random patterns of attrition.Footnote 14 That said, despite the different time periods and a great reduction in statistical power, our key finding holds up. Where both consider Labour to be closer to minorities than the working class, a WWC voter is more likely to reject Labour than a white graduate. In fact, according to this model, a perceived pro-minority bias has essentially no effect on white graduate support for the party. Given the control for ethnocentrism, this seems to be a partial corroboration for our thesis that a sense of relative political deprivation is pushing WWC supporters away from Labour.
Source: Models A-B: BESIP Wave 10 (Nov-Dec 2016) for all variables except ethnocentrism, which was recorded in BESIP Wave 11 (April 2017). Models C-F: BESIP Wave 19 (Dec 2019) for all variables except ethnocentrism, which was recorded in BESIP Wave 11 (April 2017), and the lagged dependent variable, Labour Vote (2017), which was measured in BESIP Wave 13 (June 2017). Models A and B are weighted using the standard Wave 10 BESIP survey weight, Model C with the standard Wave 19 weight, and Models D-F with the Wave 13 to Wave 19 panel weight. *p-value <0.05; **p-value <0.01.
Note: Each model presents the results from different logistic regressions predicting support for Labour (1) versus any other party (0).
Demographic Controls: Age, Gender, Country.
Using the panel element of the BESIP, Models E and F further demonstrate the robustness of this relationship by adding a lagged dependent variable to the model. That is, we regress support for Labour in 2019 (Wave 19) on the extent to which the party is perceived to represent non-whites more closely than the working class, controlling for whether the respondent supported Labour in the 2017 General Election (Wave 13). This means that we effectively predict loyalty to Labour versus defection. This has two advantages. Firstly, it might (partly) assuage fears that unmeasured longstanding affection for Labour is driving both support for the party and ratings of its performance on a variety of topics, including working class representation. Secondly, by focusing on the 2017–19 period, when the defection of working class supporters was widely highlighted (Fieldhouse et al. Reference Fieldhouse2023; Goes Reference Goes2020), we can also show that feelings of relative political deprivation are associated with real world changes in support for Labour, rather than (potentially) just longstanding opposition. We adopt panel weights to adjust for non-random patterns of survey attrition, and Model F also incorporates the same ethnocentrism control as before.
Examining Models E and F reveals that (even net of ethnocentrism) WWC respondents who believed that Labour better represented minorities than the working class were indeed more likely to defect from the party between the 2017 and 2019 contests. By contrast, belief in a pro-minority bias had essentially no effect at all on white graduate support and, if anything, is positively associated with loyalty from non-white voters themselves. Thus, these results largely corroborate H3b and H4. The substantive size of the association between perceived relative group representation and Labour support is visualized in Fig. 5. Holding all other Model F variables constant, a WWC 2017 Labour voter who perceives the party's relative group representation to be slightly biased (−1) towards the working class is predicted to have a 1 in 10 chance of defection to another party in 2019. By contrast, the same voter has just over a 1 in 3 chance of abandoning the party if they think it is slightly biased towards representing non-whites (+1). In contrast, the chance of defection by a white graduate 2017 Labour supporter barely changes across that range of values, and the chance of defection from an ethnic minority supporter decreases from roughly 1 in 5 to just 1 in 10. On average, loyalty across all three groups is highest when Labour is perceived to represent both groups equally or when they are slightly biased towards representing the working class rather than blacks and Asians. Perceptions of Labour being disproportionately pro-minority quickly polarize the party's 2017 coalition. Given the controls for ethnocentrism and the lagged dependent variable, this finding does not seem to be an artefact of WWC racial animus or confounding longstanding Labour (non-) support.Footnote 15
Discussion and Conclusion
Are the white working classes really rejecting social democrats because they associate these parties with ethnic minority interests? Despite widespread concern about the impact of ‘identity politics’ on centre-left party support (Berman and Snegovaya Reference Berman and Snegovaya2019; Ehsan and Stokes Reference Ehsan and Stokes2021; Fukuyama Reference Fukuyama2018; Lilla Reference Lilla2016; O'Neill Reference O'Neill2021), few have sought to test this hypothesis empirically or the corollary proposition that ‘identity politics’ appeals might solidify social democrat support among the growing white graduate and ethnic minority electorates (Abou-Chadi, Mitteregger, and Mudde Reference Abou-Chadi, Mitteregger and Mudde2021; Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2020; Häusermann et al. Reference Häusermann2021; Sobolewska and Ford Reference Sobolewska and Ford2020). Nor have efforts been undertaken to ascertain why this might occur. Is it simply because of underlying hostility to the well-being of racial outgroups (ethnocentrism), or does it instead stem from the WWC's widely discussed (Evans and Tilley Reference Evans and Tilley2017; Gest Reference Gest2016, 39; Hochschild Reference Hochschild2016, 137–9) feelings of displacement by other groups at the forefront of politician's priorities and their fears about falling yet further down the agenda (relative political deprivation)?
In this paper, we examined these questions using recent BESIP data measuring perceptions of how well Labour represents particular social groups. First, we showed that the WWC generally believe that Labour more closely looks after non-white interests than working class interests per se, particularly in comparison to other members of the party's new electoral coalition (graduates and minorities themselves). Second, while the WWC do not generally view working class and minority representation in stark zero-sum terms, it is true that members of this community view the link between representing working class and ethnic minority interests as weaker than others do. Third, while the WWC are more supportive of Labour when they associate the party with both working and middle class interests, positive evaluations of Labour's minority interest representation are a negative predictor of this group's support. Finally, the belief that Labour represents minorities more closely than the working class is robustly associated with WWC voters shunning the party. The same does not appear to be true for white graduates or ethnic minorities, who are either indifferent or slightly positively inclined, respectively, to a disproportionate focus on non-white interests. Importantly, all of these results are robust to controlling for underlying levels of white ethnocentrism, suggesting that it is not just the WWC's greater distance from racial ‘outgroups’ that is driving these effects; relative political deprivation is also politically consequential.Footnote 16 Rather than the dilemma of appealing to the middle class without losing working class support (Evans and Tilley Reference Evans and Tilley2017; Przeworski and Sprague Reference Przeworski and Sprague1986), the evidence here suggests that the most important conundrum facing social democrats is reconciling their remaining white working class voters with ethnic minorities and more liberal white graduates (Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2020; Harris Reference Harris2022; Kaufmann Reference Kaufmann2018; Sobolewska and Ford Reference Sobolewska and Ford2020).
While various studies show that the demographics of political representatives influence attitudes towards parties (Campbell and Heath Reference Campbell and Heath2021; Dancygier Reference Dancygier2017; Evans and Tilley Reference Evans and Tilley2017; Heath Reference Heath2015), we further demonstrate that overall evaluations of parties' abilities to represent different social groups may affect the size and composition of social democratic electorates. Our study thus upholds the relevance of copious American research to European political systems (Ahler and Sood Reference Ahler and Sood2018; Kane, Mason, and Wronski Reference Kane, Mason and Wronski2021; Mason, Wronski, and Kane Reference Mason, Wronski and Kane2021; Robison and Moskowitz Reference Robison and Moskowitz2019; Zhirkov and Valentino Reference Zhirkov and Valentino2022). In addition, our application of established American measures of ethnocentrism (Hainmueller and Hopkins Reference Hainmueller and Hopkins2015, 542; Kinder and Kam Reference Kinder and Kam2009, 42–61) to the study of British political behaviour is a notable secondary contribution (Sobolewska and Ford Reference Sobolewska and Ford2020, 338). The results presented here and in our online appendices indicate that ethnocentrism does seem to play a role in the rejection of the Labour Party by segments of the WWC. However, as we have shown, this is far from the only factor at play in this regard.
There are clearly ways that our study could be improved and extended. First, it would be prudent to ascertain which factors influence working class and ethnic minorities perceptions of representation. If certain policies or campaign strategies increase subjective feelings of representation for both groups, this will reduce concerns about potential trade-offs in appealing to these communities. Second, it would be useful to gather data on whether white respondents think of ‘white working class’ people specifically when judging the quality of Labour's representation or if they are open to a more inclusive definition that also encompasses the growing non-white working class population. Current studies of the factors that lead respondents to sort others into different class categories have not studied the role of ethnicity (Stubager et al. Reference Stubager2018), but reconciling minority and working class interests is trickier if the latter group really is treated as a placeholder for white working class interests more specifically (Sandhu Reference Sandhu2018). We lacked the survey items to fully test perceptions of the intersectionality of race and class in political representation in this regard.
It is worth reflecting on the provisional implications of our study for Labour. The finding that WWC respondents can react negatively when they perceive that parties are disproportionately dominated by ethnic minority interests demands careful consideration. While some have argued that a retreat from ‘identity liberalism’ may be unnecessary given the rapidly expanding graduate and minority electorates (Sobolewska and Ford Reference Sobolewska and Ford2020, 338–44; see also Abou-Chadi, Mitteregger, and Mudde Reference Abou-Chadi, Mitteregger and Mudde2021), we are more sceptical. First, our findings (particularly Figs. 4 and 5) suggest that the additional average marginal impact of ethnic minority representation on boosting non-white and graduate support for the party is fairly weak and outweighed by the comparatively large WWC counter-reaction. Second, even if white graduates and minorities did react more positively, one cannot completely overlook the WWC electorate. Under Britain's majoritarian electoral system, their votes are likely to be more electorally consequential than those cast by non-white Britons and graduates concentrated in densely-populated urban constituencies that Labour already monopolize (Jennings and Stoker Reference Jennings and Stoker2017; Thompson et al. Reference Thompson2022). Third, if the mobilization of racial ‘identity politics’ is not an obviously successful strategy for the American Democratic Party (Fukuyama Reference Fukuyama2018; Lilla Reference Lilla2016) in a country that is, as of 2017, around 40 per cent non-white (Poston Jr and Sáenz Reference Poston and Sáenz2020), it is likely even less viable in a country such as Britain that is not predicted to reach this level of diversity till at least the 2060s (Lomax et al. Reference Lomax2020). This is particularly likely if, as under Boris Johnson, the Conservatives can make successful populist appeals to WWC voters (Evans, de Geus, and Green Reference Evans, de Geus and Green2023).
Elsewhere in Europe, the story is more complicated. On the one hand, many countries have smaller non-white populations and more restrictive enfranchisement legislation (Dancygier Reference Dancygier2010; Dancygier Reference Dancygier2017; Ford and Jennings Reference Ford and Jennings2020), reducing the likelihood that identity politics can benefit social democrats in much of the continent. Furthermore, the widespread use of proportional representation (PR) further reduces the extent to which spatially concentrated and less well-integrated ethnic minorities become pivotal players in parties' seat-maximizing strategies (Dancygier Reference Dancygier2017). It is also conceivable that, under PR, competition for WWC voters could be even more pronounced given the larger radical right parties (Rydgren Reference Rydgren2013). Conversely, PR systems allow social democrats to be challenged from the left if they backtrack too hard on minority representation. Indeed, very permissive electoral systems may gradually witness ethnic minorities developing their own vehicles for representation – for instance, the Netherlands’ DENK party (Siegel Reference Siegel2016). In addition, while we have uncovered no evidence to suggest white graduates in Britain reject the centre-left when they associate these parties with weaker representation of minorities, it is possible that the story could be different where there is a larger and more electorally credible left-liberal alternative to the social democrats, such as the Green Party (Abou-Chadi, Mitteregger, and Mudde Reference Abou-Chadi, Mitteregger and Mudde2021; Oesch and Rennwald Reference Oesch and Rennwald2018). Of course, it may still prove easier for (more socially conservative) social democrats to form a post-election coalition with these other parties rather than trying to unite all these voters under one party pre-election, as Labour has attempted. More research will be needed to understand how things play out in multiparty European systems.
Our findings should not be considered a justification for ignoring the well-being of ethnic minority communities, especially as the WWC does not appear to view representation of black and Asian and working class interests in strict zero-sum terms. Rather, working class whites must be assured that any increase in the political attention and resources devoted to the minority community is met with a renewed focus on the issues facing their own social group, and not at their own expense. Where possible, centre-left parties might attempt to establish positive-sum gains. For instance, nominating more ethnic minority candidates from unambiguously working class backgrounds might help improve both ethnic minority and working class perceptions of representation (Quilter-Pinner et al. Reference Quilter-Pinner2022). Elsewhere, perceptions that a party is primarily trying to represent minorities rather than the working class could be blunted if care was taken to frame and justify policies principally benefiting minorities in race-neutral terms. This means avoiding the impression that economically underprivileged non-white communities are being targeted for assistance because they are non-white rather than because they belong to a wider class of economically disadvantaged communities (Abbott et al. Reference Abbott2021; English and Kalla Reference English and Kalla2021; Sniderman et al. Reference Sniderman1996).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007123423000601.
Data availability statement
Replication data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/FWCKSE
Acknowledgements
We want to thank the anonymous reviewers of this article for their valuable insights and suggestions, in addition to Rune Stubager and Chitralekha Basu, who provided very useful comments on earlier drafts. Previous versions of this manuscript were presented at the 2021 Elections, Public Opinion, and Parties (EPOP) annual conference, the 2022 European Political Science Association (EPSA), and the European Consortium for Political Research (ECPR) general conferences, and at separate talks at the Nuffield Early Career Researchers' Workshop and the Juan March Institute (both in 2022). We are very grateful for the thoughtful comments of all those who attended these events. Any remaining errors are our own.
Financial support
None.
Competing interests
None.