Abstract Brexit, the wider populist surge in Europe and Trumpism all seem to involve interesting geographies that have been taken as clues to the worrying puzzle facing a political/academic establishment about what is driving the surge and how it might be abated. One major theme has been that of the places left behind economically by an opening up to competition from cheap (migrant or overseas) labour—counterpointed by the idea that specific types of people have been left behind culturally. This article attempts a less reductive approach, starting with examination of oddities in the Brexit geography and then investigating how populist support across European regions is influenced by the interaction of economic/demographic change with varying cosmopolitan/localist influences. Introduction The aim of this article is to advance understanding of the geography of recent populist movements in Europe in broadly political economy terms, positing an important role for the interaction between political and economic processes at sub-national (local and/or regional) scales. These have come to prominence in a range of European countries in various forms, often associated with ultra-right nationalism, though also sometimes with a radical left economic agenda. The particular UK shock from which this article takes off, involving a less extreme form of nationalism, stood out as the first of this wave to secure majority support from a national electorate, via the crucial 2016 referendum on EU membership (rather than a parliamentary election). This particular success (achieved with a modest margin) has brought extraordinary uncertainties both for the national economy and personally for EU migrants. Beyond these, the broad scale of support sustained by the campaign from its outset seems to have let a more purely populist genie out of the bottle, with expectations that are bound to be unsatisfied. The UK case, though distinct from those of (most) other ‘populist’ movements in Europe, thus presents a particularly strong motive for trying to understand potential commonalities in the contexts from which a populist impulse emerges, and what sorts of geographies are relevant to this. In retrospect, it is both extraordinary and symptomatic of a continuing ‘resistance’ that (despite polling evidence) the possibility of a Leave victory was not taken seriously enough by the UK’s political, policy and business elites to warrant effective contingency planning. Part of the story is that—until very late on in the campaign—there was little awareness that positions on this single issue not only cut across the parliamentary parties but reflected a much broader societal cleavage, with the Leave campaign expressing the alienation of many from mainstream/establishment national politics and forms of policy debate. Credible evidence of this link had been presented by Conservative researchers when the UK Independence Party (UKIP) first started to achieve success. In particular, Ashcroft (2012) suggested that its electoral threat was not really based on concerns about ‘Europe’—or even immigration, though that was clearly more salient. Instead, focus group evidence highlighted a basic concern among UKIP supporters ‘with the way country is going’, and a sense that other parties were scared of ‘saying what needs to be said’. These feelings were reported to be acquiring particular force as economic recession and austerity sharpened resentments about migrants’ impact on the character of local areas. Later, survey evidence showed UKIP supporters to be distinctively authoritarian, distrustful of government, rating experience over academic qualification and disbelievers in climate change (NatCen, 2014). Even the much broader group of Leave voters on referendum day overwhelmingly shared the view that multi-culturalism, immigration, social liberalism, the green movement and feminism were forces for ill (LA Polls, 2016). Upsurges of electoral support for marginal parties such as UKIP and their predecessors had been commonly (and reasonably) understood as ‘protest votes’ of almost no real significance. As such they ought not to be reproduced on a large scale in a referendum specifically focused on a question of strategic national economic importance. Significantly, however, populist discourse came to directly challenge the authority of evidence offered on that issue, with a senior minister on the Leave side announcing, three weeks before referendum day, that ‘people in this country have had enough of experts’ (FT, 2016).1 This statement resonated with polling analyses showing a more or less even overall division of support, with distinct age/educational profiles for the two sides, and little apparent effect from a lengthy public debate about the pros and cons of withdrawal (Curtice, 2016). A belated awareness of the strong populist base harnessed by the Leave campaign2 was reflected in Curtice’s invocation of ‘a nation at unease with itself’, with the referendum “playing out a social division between the ‘winners’ and ‘losers’ from globalization”, and Leave support coming from those ‘left behind’ by this process.3 The implication that the root of populist discontent lies in processes of economic liberalisation on an inter-continental scale is ironic (at least), given that many senior backers of the campaign saw EU-exit as a means for the UK to pursue a less constrained global trading role. Whether correct or not, this diagnosis made a clearer connection with a buoyant Trump campaign in the USA that was explicitly playing on the negative impacts of globalisation for particular communities. For subsequent attempts to understand how such a ‘leaving behind’ actually connected with support for the Leave campaign, and other populist movements in Europe, the relative importance of global versus continental scales of integration is less of an issue than whether the hurt expressed by those who have lost out is simply a material matter or primarily a cultural one. The bulk of the commentary so far (whether analytic or political) has focused on sources of economic loss, particularly in terms of jobs—whether off-shored or taken by immigrants. But for supporters of European populist parties at least, Inglehart and Norris (2016) found very strong evidence from the European Social Survey (ESS) that the key drivers were cultural and attitudinal. In relation to the UK case also, Kaufmann (2016) has argued forcefully—against analyses suggesting that ‘poverty, low skills and lack of opportunities’ explained the Brexit vote (Goodwin and Heath, 2016b)—that ‘it’s not the economy stupid (but) personal values’. Geography figured in much of the early reportage and commentary, in a way that has grown familiar as administrative data have become readily mappable. The problem, in this case as in others, is one of interpreting map patterns when the available stereotypes are often clichéd (North-South, London vs provinces, Scotland-England) and not always appropriate. Beyond this, there is the usual ambiguity as to whether such a pattern is read as reflecting the spatial distribution of different types of people whose propensities are unaffected by location, or the effects (on all or most) of exogenous spatial factors, such as accessibility, industrial histories or political inattention in this case. Statistical analyses of the UK referendum have started to look at the relative importance of these, but scarcely so far at how individual and situational factors might interact to shape different patterns of political response in different places (perhaps only in Goodwin and Heath’s (2016b) exploration of British Election Survey data, and now in Lee et al., 2017). This article tries to address two challenges in making sense of the evidence of spatial unevenness of populist movements within European countries. The first involves the question of how far geographic differences in local population mix ‘matter’ politically, in terms of their impacts on the attitudes and behaviour of individuals (the ecological question). The other concerns what relationship there may be between expressed attitudes conveying a sense of cultural/political marginalisation and the spatially/occupationally uneven impacts of internationalisation (the political economy question). Each represents an attempt to get beyond an initial round of ‘reductionist’ approaches ascribing spatial variations in populist movements to either: differing degrees of exposure to new economic threats from sources of cheaper labour; or differing mixes of people from demographic/educational groups with distinctive attitudes and values. To explore these issues, the remainder of the article is organised in three discrete sections. The first addresses the question of how we might understand the uneven geography of populist voters ‘getting left behind’. The second reports on a simple (early) analysis of the spatial pattern of Leave voting in the UK referendum, across the districts of Great Britain and the parliamentary constituencies of Northern Ireland,4 suggesting that neither analyses based on economic inequalities nor ones simply in terms of compositional differences are adequate to the task. And the third follows these up with a (less simple) spatial analysis of micro-level data on support for populist parties across Europe, allowing for interactions between external shocks, different kinds of individual response and group strength. Populism and the people/places which have been ‘left behind’ Populism is a rather unclear kind of concept: neither really a belief system/programme nor simply an expression of individual discontent/apathy: more a phenomenon than either of these. As an insurgent movement, the essence of populism is a rejection of the moral/intellectual authority of a socio-political establishment in favour of the authentic, commonsensical responses of ordinary people(s). But functionally, its appearance in particular contexts seems to depend on a combination of two distinct elements, plus an enabling environment (or market opportunity). The two required elements are: a substantial body of potential supporters, belonging to an ‘imagined community’, though maybe only sharing a thin ideology, involving beliefs both in individuals/micro-communities as the most reliable judges of what should be done—as against elites who are out of touch and corruptly self-serving—and in the need for a strong, orderly state; and entrepreneurial politicians who spot an opportunity to mobilise this rejectionism, in order to secure power and pursue some ends of their own, by supplying the kind of leadership required in the absence of either established organisation or substantive political agenda—aggregating supports, securing access to finance/media, and providing rhetorical/charismatic cover for gaps in manifestos. To take off, however, it also needs the emergence of a significant gap in the effective market coverage of established parties and/or labour movements, as when: their capacity to aggregate support around a more purposive agenda or class identity has been fragmented by unresolved conflicts, and/or eroded by failures to satisfy supporters in terms of basic socio-economic and security expectations. The behavioural and circumstantial aspects of the second and third of these conditions entail a substantial element of randomness in when and where significant populist movements actually surface, what agenda they promote and how resilient the movement is. Thus, in the case of the UK’s Brexit movement, it was the decision (as a party management tactic) to hold a referendum, with some government ministers becoming temporary leaders of a populist campaign, that took this from the domain of a marginal and unstable party to majority support. Economic dissatisfactions, including those linked to internationalisation, may come into this picture in several ways via: specific issues related to job losses, or downgrading of wages in those activities recruiting migrants—maybe leading to loss of faith in the genuineness of expressed concerns for the activities and areas involved (for example old industrial regions outside the South East);5 a more generalised loss of trust in the competence and integrity of government, in relation, for example, to the genesis and aftermath of the financial crisis; or a perceived breakdown in the trade-off, for white working class citizens in particular, between: accepting an attrition, via liberal/internationalist social and migration policies, in the values of established ways of life; and counting on sustained improvements in overall living standards—in a period when those improvements were being reversed. Two cautionary inferences may be drawn for efforts to understand the relation between populism and the geography of who has been left behind (in what terms). One is that only the first of these paths may have any clear spatial expression. The other is that, in each case, translation of dissatisfaction into some form of populism might well go with complaints about cultural change or untrustworthy elites, rather than either economic issues or particular personal insecurity. From the most direct examination of these issues so far, Inglehart and Norris (2016) convincingly show that voting for populist parties in Europe has been strongly associated with attitudinal positions on a range of cultural values, and only weakly related to economic insecurity, through personal unemployment. Their set of cultural values covers attitudes to immigrants, authoritarianism and distrust of governance (both national and global), though the overall effect is dominated by a self-assessed left–right political indicator. The interpretation placed on these findings is that current populism represents a cultural backlash by those who lost out in the triumph of post-materialist values from 1970 on. Cumulative value change is seen as having passed a tipping point, notably for the old, but also for ‘the less educated groups left behind by progressive cultural tides’ (my emphasis). Essentially this updates a classic narrative of the impact of post-industrialism (Inglehart, 1971). Where post-nationalism enters the story is primarily via the impacts of burgeoning migration on culturally valued ways of life, but also of mobility more generally on the sense of living in a homogeneous nation-state and on confidence in its enforcement of national values. In value terms then, cosmopolitanism is recognised as one of the antitheses of populism, with Inglehart and Norris (2016) making a link to Merton’s (1957) classic comparison between local and cosmopolitan influentials. But his analysis also has a bearing on more material issues with a likely relevance to the geography of populism. Specifically, the observation that Merton made was not just that there were people who had a particular sentimental/cultural stake in a local system, but that they had economic and political ones too, which others did not. For one group local reputations and contacts (which might as well mean ones local to a firm or even industry as much as a place) were treated as crucial, while for others, it was those of peers in wider networks of expertise or influence which counted. To push this rather further, it could be said that (with more or less freedom) people make strategic life choice between acquiring sets of human and social capital, which are weighted more to the global/universal or the local/particular. This is partly a matter of the balance struck between intensive engagement in networks local to a place or organisation, and extensive efforts to access ones with a wider scope, offering routes to higher-grade knowledge/opportunities, or means of escaping a dependence on a potentially vulnerable locale. And, taking a step still further, this idea can be translated to the temporal domain—with people making related choices between investing effort in acquiring intensive but non-transferable types of knowledge about current/specific markets, technology or politics versus seeking (extensively) some more durable/flexible/universal understanding of processes, their potential variability and methods of researching these.6 One rough indicator of this choice is the pursuit of more vocational versus academic types of qualification. The cultural values transmitted along with these different forms of learning and networking will tend to be respectively more localist/populist or more cosmopolitan in their orientation (as Goodhart, 2017 well documents in terms of higher education and his ‘Somewhere/Anywhere’ divide). But it is also likely that those with personal assets biased towards one or the other will have quite different material stakes in sustaining established locally-specific modes of operation. And hence also, rather different responses to the globalising trends that (over the past 40 years or so) have shifted the general balance of advantage away from personal asset portfolios that are localised (in spatial, organisational, sectoral or temporal terms). An implication of this line of argument is that heightened populism may reflect not simply a culturally conservative reaction, on the part of the elderly particularly, to progressive loss of their ability to control, or simply cope with, a changing local environment, but one which has a strongly economic aspect in so far as many of the (cultural) losers will be people who got tied into an inflexible occupational niche. Many of these will also be relatively old, since the balance in terms of qualifications has been swinging strongly towards academically based ones during this era of globalisation, particularly in higher education. But they are also likely to be ‘localised’ in particular occupations, sectors and places where vocational qualifications, and other less cosmopolitan assets, remain important. Who these economic/political localists and cosmopolitans might actually be, in terms of more conventional survey categories, is unclear, though the petty bourgeois surely figure prominently among the former and the higher ranks of professionals (if not managers) among the latter.7 There is no expectation, however, of a simple correlation between cosmopolitan orientations and social grade, and a likelihood that specific occupational groups within classes will display distinctive biases among their members. Translation of these different spatial orientations into real geographies involves a couple of steps. The first relates to an increasingly functional division of labour on (primarily) occupational rather than industrial lines, with comparative advantage for particular places being established on the basis of particular labour qualities available in different places, as well as the price of these (Massey, 1984). A likely result is an increased differentiation of places in these terms, but in any case, substantial variations in the mix of localists and cosmopolitans, maybe reinforced by culturally selective migration. Mixes (at least) will vary between places. The second step, however, is the (re)formation of labour qualities, but also of local cultures and politics in the light of the shifting mix of peoples, business/personnel strategies and labour organisation—places being made, as Massey (1984), proposed through the sequence of roles occupied in the changing spatial division of labour. In terms of support for populist movements then, the expectation would be not simply of a compositional association with the local occupational mix, but of non-linear relations reflecting local political processes in which: some key occupations contribute disproportionately to political dispositions, shaped partly by the play of local interactions and partly by external norms for the groups involved; and the ways in which these dispositions are affected by potential sources of stress, including substantial immigration of cultural outsiders, also reflect the balance between localists and cosmopolitans. In addition to this occupationally related basis for localist/cosmopolitan orientations (and the related mover/stayer divide highlighted by Lee et al., 2017), membership of some kinds of trans-local civil society organisations may also serve to attenuate a disposition to pure/defensive localism. Specifically, we might expect this to be the case for those religious groups and industrial unions cultivating strong senses of identity cutting across those of place, nationality or ethnicity. And, for any simple erosion of these (without replacement), in the face of economic and social change, to increase the likelihood of a populist response to these changes—as Gordon Brown (2014) suggested for the separatist surge8 in Scotland, where the Kirk and factory/pithead union meetings had lost much of their historic role as nurseries of social solidarity. These loosely formulated hypotheses are explored empirically in the two sections that follow. First, a simple aggregate analysis of district-level vote shares in the 2016 UK referendum is used to see whether their pattern bears out any of the expectations from this perspective, in terms of: the significance of occupational categories, as well as age and qualification levels; deviations from a simply monotonic relation with social/educational level; and evidence of ecological, rather than purely compositional effects. Then, some clearer hypotheses about the interaction between individual attributes, aggregate characteristics and economic trends are investigated more directly, using a regionalised version of the ESS survey data on support for populist parties in European countries. What can we learn from the map of UK referendum results? Despite its single-issue focus, the UK’s 2016 EU referendum is an interesting case in which to explore the geography of populist support, because it was not tied either to election of particular representatives fighting personal campaigns or endorsement of a position of one or other of the two parties with a long-term organisational presence (of varying effectiveness) in local constituencies. For an initial exploration of the geography of support for Leaving, a simple regression analysis was undertaken—soon after the referendum, in parallel with quite a few others (including Clarke and Whittaker, 2016; Goodwin and Heath, 2016a; Harris and Charlton, 2016; Langella and Manning, 2016), using the published vote share figures for each local authority district,9 together with readily available indicators of conditions/population characteristics, mostly from the 2011 Census. The example reported here differs from these others in two related ways: first, it had an eye to the kind of cosmopolitan/localist ideas sketched in the last section, and second, it looked in a less aggregated way at patterns of association with qualification types, and two-digit occupational categories—rather than assume that the relevant contrasts would be between the top and bottom of educational or socio-economic scales. Other factors included in the estimated regression model, notably the age and ethnic mix of the resident population and changing rates of migrant arrivals—plus a few additional factors, which were also examined but not finally used, including housing tenure and industrial mix—were approached in a similar exploratory way, though serving more or less as contextual/control variables. Three related regression models are reported in Table 1, involving different combinations of qualification and occupational measures. All involve simple cross-sectional analyses for the full set of districts in mainland Britain, using the Leave share of the referendum vote as the dependent variable and with sets of (Census-based) independent variables, all of which represent rates/proportions of the adult population of these districts. And each confirms that (as originally indicated at micro-level by polling data) age and qualification differences really matter, accounting for a large part of the variation, though not explaining a substantially lower Leave vote share across Scotland. To allow for (if not explain) this Scottish factor, a dummy variable for Scotland was always included, with dummies for other regions (and some sub-regions that showed up in maps of residuals) also being tried. Table 1. Regressions of Leave vote share on age, education and occupation composition. Variable (1) (2) (3) Constant 0.533 (6.9)*** 0.323 (4.4)*** 0.784 (9.5)*** % in age group 30–59 0.305 (3.2)*** −0.296 (3.3)*** −0.385 (4.1)*** 60+ 0.323 (4.3)*** 0.224 (3.5)*** −0.015 (0.2) % in ethnic group Black + mixed multi-ethnic −0.399 (7.6)*** −0.134 (2.2)* −0.170 (3.4)*** % migrants in population Arrived pre-1991 −0.336 (2.7)** −0.176 (1.4)*** −0.315 (3.0)** Arrived 1991+ 0.376 (4.9)*** −0.023 (0.3) 0.262 (4.2)*** % with highest qualification at Level 2 0.648 (3.9)*** 0.218 (1.4) Level 3 −0.508 (3.4)*** −0.702 (5.5)*** Level 4 −1.018 (31.5)*** −0.762 (11.7)*** % in occupation (two digits) 11: corporate managers 0.605 (3.3)*** 1.281 (8.0)*** 23: education professionals −1.770 (7.0)*** −0.354 (1.7) 33: protective services: associate professors 0.495 (3.0)** 0.248 (1.7) 34: culture, media, sport occupations −0.969 (3.7)*** −1.026 (4.5)*** 42: secretarial 2.619 (6.4)*** 1.519 (4.0)*** 52: skilled metal/electrical 2.544 (8.9)*** 1.708 (7.0)*** 72: customer service −0.372 (1.0) −0.973 (3.1)** 82: transport driver/operators 3.225 (11.5)*** 1.021 (3.5)*** 92: elementary admin service occupations 0.701 (4.3)*** 0.277 (2.0) Spatial location Merseyside −0.105 (8.7)*** −0.084 (7.2)*** −0.085 (8.9)*** Wales −0.052 (7.1)*** −0.031 (4.3)*** −0.036 (6.0)*** Scotland −0.139 (11.6)*** −0.154 (13.9)*** −0.137 (13.2)*** Central Scotland (additional) −0.076 (6.1)*** −0.035 (2.9)** −0.053 (5.2)*** Adjusted R2 0.918 0.927 0.954 N 380 380 380 Variable (1) (2) (3) Constant 0.533 (6.9)*** 0.323 (4.4)*** 0.784 (9.5)*** % in age group 30–59 0.305 (3.2)*** −0.296 (3.3)*** −0.385 (4.1)*** 60+ 0.323 (4.3)*** 0.224 (3.5)*** −0.015 (0.2) % in ethnic group Black + mixed multi-ethnic −0.399 (7.6)*** −0.134 (2.2)* −0.170 (3.4)*** % migrants in population Arrived pre-1991 −0.336 (2.7)** −0.176 (1.4)*** −0.315 (3.0)** Arrived 1991+ 0.376 (4.9)*** −0.023 (0.3) 0.262 (4.2)*** % with highest qualification at Level 2 0.648 (3.9)*** 0.218 (1.4) Level 3 −0.508 (3.4)*** −0.702 (5.5)*** Level 4 −1.018 (31.5)*** −0.762 (11.7)*** % in occupation (two digits) 11: corporate managers 0.605 (3.3)*** 1.281 (8.0)*** 23: education professionals −1.770 (7.0)*** −0.354 (1.7) 33: protective services: associate professors 0.495 (3.0)** 0.248 (1.7) 34: culture, media, sport occupations −0.969 (3.7)*** −1.026 (4.5)*** 42: secretarial 2.619 (6.4)*** 1.519 (4.0)*** 52: skilled metal/electrical 2.544 (8.9)*** 1.708 (7.0)*** 72: customer service −0.372 (1.0) −0.973 (3.1)** 82: transport driver/operators 3.225 (11.5)*** 1.021 (3.5)*** 92: elementary admin service occupations 0.701 (4.3)*** 0.277 (2.0) Spatial location Merseyside −0.105 (8.7)*** −0.084 (7.2)*** −0.085 (8.9)*** Wales −0.052 (7.1)*** −0.031 (4.3)*** −0.036 (6.0)*** Scotland −0.139 (11.6)*** −0.154 (13.9)*** −0.137 (13.2)*** Central Scotland (additional) −0.076 (6.1)*** −0.035 (2.9)** −0.053 (5.2)*** Adjusted R2 0.918 0.927 0.954 N 380 380 380 Notes: (i) Observations are weighted by the square root of population. Adjusted R2 values are about 0.007 lower in unweighted versions of the regressions; (ii) all variables are defined relative to the adult population; (ii) omitted categories thus include 18–29 years olds; long-term unemployed and never worked; all other ethnic groups; the UK born; people with other qualifications (including apprenticeships), no qualifications or level 1 only; and the rest of England. Stars indicate levels of statistical significance: * = 5%, **= 1%, *** = 0.1%. Sources: Published Referendum counts for Districts; Census 2001 for all other variables. View Large A rather coarse age grouping was used, with just three age categories (18–20; 30–59; and 60+), capturing adequately for these aggregate data the observations which had been made about the divergent voting behaviour of young adults and older people. The basic premise has been that younger people (like the more educated) were generally less likely to vote Leave, though at the individual level these two factors clearly interact because younger cohorts have had substantially more education. In the aggregate analyses reported in Table 1, introducing qualifications and/or occupational mixes yields three radically different profiles for the implied age factor, none of which involves a monotonic relation with age/youth. This is not of direct interest here, however, and the point should be repeated that these aggregate relationships do not necessarily reproduce the pattern observed at the individual level. (For example, because those in any one of these three age brackets may be more inclined to live in areas with particular sets of other characteristics that make their residents more likely to vote Leave/Remain—or hold to populist values.) A somewhat similar point should be made about the ethnic group variable, with the additional proviso that the ethnic mix of local population may be relevant for two very different reasons. On the one hand, as micro-evidence indicates, white voters were rather more prone to vote Leave than members of black and minority ethnic groups. But on the other hand, that part of the white population that is sensitive to the presence of (some or all) minorities in their area might be more likely to vote Leave in areas where these minorities represented a larger share of the population. For ethnic as for other variables in these analyses, a rather full set of disaggregated categories was originally tried, and then tested down, with just the statistically significant categories/contrasts being included.10 The inclusion of a single Black and Mixed Multi-Ethnic variable in Table 1, attracting negative coefficients, thus reflects the fact that the variables relating to the white population and to various Asian origin groups attracted coefficients that were similar (and more positive)—though maybe for a different mix of reasons (the white group perhaps because it included more members unsympathetic to culturally different migrants; and the Asian groups perhaps because they were the most liable to be categorised as ‘culturally different’). The migrant flow variables were based entirely on dates of arrival into the UK reported in the 2011 Census by people born abroad—not actual records of flows. Again, the choice of a (1991) breakpoint is the outcome of experiment rather than preconception. But what the results show (with a negative coefficient for those who had arrived in the earlier period and a positive one for the later years) is, like several of the other aggregate referendum analyses, that it is either increases in the rate of inflow relative to a past norm, or maybe just the presence of recent/less fully integrated migrants, that seems to have raised the Leave vote—not the simple fact of having a significant migrant population. The variables that are of much greater statistical importance—and of central relevance to the hypotheses sketched in the last section—are those relating to the proportions of the adult population with different qualification levels and occupational types. Either set can account for a considerable part of the variance, raising the R2 value from about 0.5 when both are omitted from the models reported in Table 1 to over 0.9 when either is included and 0.95 when both are included. As this implies, there is a substantial (but not complete) overlap between the two sets of variables. The three points of particular interest in the results reported for these variables in the three regressions are: 1. The relation with qualification levels is not monotonic. Having more people with qualifications at Level 3 (A levels) or above is—as all other analyses have suggested—associated with having a lower Leave vote. And, in the model (1) results at least, where occupations are omitted, this effect is even stronger in relation to those with Level 4 (degree equivalent) qualifications. But, having more people with Level 2 qualifications, rather than Level 1 or no/unrecognised qualifications, is found to be very clearly associated with a higher Leave vote; 2. In relation to occupations, the absence of a monotonic relationship between voting Leave and lower social status—of the kind generally implied—is even more striking. Concentrations of corporate managers (as well as of others in lower level, generic ‘office jobs’) seem to be associated with higher propensities to vote Leave, as with drivers and skilled metal/electrical workers—whereas most types of operative and elementary worker (and the long-term unemployed) are not. On the other side, having more educational professionals, culture/media/sports workers and people in customer service (not sales) roles was linked with a larger Remain vote share. The occupational structure seems to be very important, in ways that are not reducible to qualifications level or social class position; 3. The estimated occupational effects are much too strong to be simply an expression of inter-occupational differences in individuals’ propensity to vote one way or the other. If all that were involved were mix effects of that kind, then the rule of thumb is that the maximum achievable variation between any pair of occupations (one with 100% Leavers among its members and the other 100% Remainers) would involve a difference between their coefficients of 1.0.11 The model (2) estimates, however, involve a whole series of cases in excess of that, including one pair with a difference of at least 4.25 (at a 95% confidence limit). In model (3) also, this limit is very clearly breached, though the numbers look less extreme because the qualifications factor absorbs some of the excess. The effects are hard to separate, but where qualifications are included on their own, one of the differences in coefficients (of 1.5 at the 95% limit) is also too large to represent simply a mix effect. These represent very clear signals that some particular aspects of occupational mix (in particular) are playing an important ecological/contextual role in influencing local voters’ propensity to vote one way or another—and quite probably the attitudes underlying this behaviour. In addition to these substantive factors, there is at least one clear territorial variation in voting patterns, which is not going to be explained in these kind of terms, namely the much stronger propensity of the Scottish electorate to vote Remain. Trials with regional dummies and inspection of residuals: confirmed the strength of this Scottish effect, while also suggesting that a weaker version of it operated in Wales (significantly reducing what would otherwise have been a clearer Leave majority); and that there were two or three identifiable sub-regional clusters of areas where Remain voting was higher than expected: Merseyside (as a whole), Central Scotland and maybe Mid/West Glamorgan.12 Notably, neither these exceptions nor scrutiny of the map of residuals (included in the Supplementary Material) lends support to the idea that there is an association between Leave voting and relative economic decline or deprivation at a sub-national level. The Merseyside case is an interesting one in its own right, but is given a particular irony by Rodríguez-Pose’s (2017) use of an economic expert’s Liverpool lecture about the productive superiority of the South East to epitomise why revanchist voters in places that ‘don’t matter’ economically would turn naturally turn to populism (and the Leave campaign).13 This is not to say that uneven economic development has nothing to do with the spatial pattern of voting in the referendum. The association, which Langella and Manning (2016) report between higher rates of Leave voting (after controls for age, qualifications and migration) and the scale of attrition of jobs in heavy industry and public sector activity, over the very long run, gives the lie to that. But the connection is not a simple one in terms of current levels of deprivation, and is very likely to be mediated by, and picked up here, the occupational structures, which have evolved in more/less affected areas. That is a question which cannot be resolved by any simple cross-sectional regression analysis. But what this one does show is that—except for a strong Scottish bias towards Remaining—the pattern of spatial variation in Leave voting can be very well accounted for statistically in terms of structural differences in age, qualification and occupation—plus some influence from past international migration. However, two features of the occupational results indicate that underlying processes are more complex than this kind of analysis can handle since estimated effects are too strong to be purely compositional, while the pattern of variation cuts across conventional socio-economic classes, in a way that might be more consistent with the hypothesised cosmopolitan/localist divide. Regionalising the analysis of populist voting in Europe To look more directly at how individual- and area-level processes may interact, the second empirical analysis returned to the rich ESS data source used by Inglehart and Norris (2016), but with the emphasis shifted to geographical variations in circumstances and responses, rather than purely individual variation within a pan-European population. The data set used for this analysis pooled the first seven biennial waves (from 2002 to 2014), from the consolidated file provided by ESS (2014). For this analysis, the priority is less one of investigating the relative importance of distinct economic and cultural factors than: to gain an understanding of the difference that location may make, in terms of both aggregate economic/demographic trends and ecological effects of population mix; and to use information on types of qualification (and affiliation) to test the hypothesis that those with more ‘localist’ forms of personal capital (and fewer potentially cosmopolitan attachments) are more likely to adopt a populist response to challenges posed by rapid internationalisation. Area effects and local stimuli were investigated at the level of regions (NUTS 1 or 2, depending on availability), within varying national and temporal contexts controlled through fixed effects. A consequence was to restrict coverage of the sample to include only countries with multiple regions, and waves in which a national sub-sample recorded votes/support for at least one populist party. The effect was to reduce national representation from a possible 33 countries to 18—within a more compact version of Europe (now closer to that of the European Economic Area than that of Eurovision). On average, there were nine regional units and four useable waves (from a possible seven) per country. Populist parties were identified in the same way as by Inglehart and Norris (2016), though respondents’ support for them was determined on the basis of which party they identified as closest to their position, rather than the one for which they had last recorded a vote.14 Individual information on types (as distinct from level) of qualification was only available for the last three waves but, following analyses of these, it was determined that the prevalence of vocational qualifications in a respondent’s occupation was as effective an indicator of their orientation as their own particular qualification—allowing all available waves to be used.15 The analysis was in two stages: the first applies an adapted version of the Inglehart and Norris’s (2016) model to this restricted data set, examining the effects both of adding regionalised variables to it and then of removing the attitudinal variables; and the second applies this adapted, regional model to testing specific hypotheses from section 2, about: reactions to potential impacts of internationalisation; differences in responses to cultural aspects of these between people with different types of personal capital and organisational affiliation; and possible ecological/booster effects from differential concentrations of these within a region’s population. Exclusion of attitudinal variables from the main body of the analysis—despite their close relation with individuals’ political disposition to support populist parties—is on the basis that they serve to channel a combination of other influences, potentially including economic geographical ones of direct concern here. Apart from a set of background variables, the primary focus here is on the role of qualifications, potential exposure to migrational/competitive shocks and the (regional level) interactions between these. Introducing regional measures into a baseline model The baseline model was loosely adapted from Inglehart and Norris’s (2016) original, retaining two of their scales, but dropping the self-defined left–right political identity as too broad to be helpful, deconstructing two other scales after experimentation and adding some more indicators of labour market position, relevant to the hypotheses of section 2. Variables included in the version reported here have been winnowed down from the large array available in ESS (2014), and ones (including economic indicators for which Inglehart and Norris reported negative results) that might have been salient but proved not to be so in practice have been silently omitted. For the basic model and each of the variants, a large set of fixed effects for combinations of country and survey year have been included—allowing for national variations in both economic and political cycles as well as in the general level of support for populist movements. Even after winnowing, the baseline model included 23 independent variables—almost all of clear statistical significance, though a few marginal ones were retained because of their relevance to the second stage work. These variables were grouped into three categories—personal background factors, attitudinal indicators and those relating to labour market position—and composites of these are used (in Table 2) for comparison of variant models.16 A full set of estimates for the baseline version is available in the Supplementary Material. Table 2. Logistic regressions of Leave votes on summary individual and regional factors. (1) Fixed effects only (country × year) (2) Background, attitudinal and labour market effects (3) Individual factors and regional fixed effects (4) Individual factors + regional means (5) Individual factors + regional means and employment change (6) Removing the attitude effects Individuals Personal background (PB) 1.00 (24.8)*** 1.00 (25.0)*** 0.99 (24.6)*** 0.99 (24.6)*** 0.87 (23.0) Personal attitudes (PA) 1.00 (51.2)*** 1.00 (50.7)*** 1.00 (50.9)*** 1.00 (50.6)*** Labour market position (LM) 1.00 (8.4) *** 1.00 (8.1) *** 0.95 (7.9)*** 0.95 (7.0)*** 1.96 (19.5) Regions PB mean −0.61 (1.8) −0.78 (2.0)* −0.37 (1.0) PA mean 0.04 (0.3) 0.02 (0.1) LM mean 5.27 (4.4)*** 5.13 (4.2)*** 5.71 (5.0)*** Total employment change % 0.004 (0.4) 0.001 (0.0) Industrial employment change % 0.013 (0.8) 0.015 (0.9) Urban population % −0.001 (0.9) −0.004 (2.8)** Fixed effects Countries × years X X X X X X Regions X R2 equivalent (Nagelkerke) 0.273 0.398 0.415 0.398 0.398 0.311 N 49,661 49,661 49,661 49,661 49,661 49,661 (1) Fixed effects only (country × year) (2) Background, attitudinal and labour market effects (3) Individual factors and regional fixed effects (4) Individual factors + regional means (5) Individual factors + regional means and employment change (6) Removing the attitude effects Individuals Personal background (PB) 1.00 (24.8)*** 1.00 (25.0)*** 0.99 (24.6)*** 0.99 (24.6)*** 0.87 (23.0) Personal attitudes (PA) 1.00 (51.2)*** 1.00 (50.7)*** 1.00 (50.9)*** 1.00 (50.6)*** Labour market position (LM) 1.00 (8.4) *** 1.00 (8.1) *** 0.95 (7.9)*** 0.95 (7.0)*** 1.96 (19.5) Regions PB mean −0.61 (1.8) −0.78 (2.0)* −0.37 (1.0) PA mean 0.04 (0.3) 0.02 (0.1) LM mean 5.27 (4.4)*** 5.13 (4.2)*** 5.71 (5.0)*** Total employment change % 0.004 (0.4) 0.001 (0.0) Industrial employment change % 0.013 (0.8) 0.015 (0.9) Urban population % −0.001 (0.9) −0.004 (2.8)** Fixed effects Countries × years X X X X X X Regions X R2 equivalent (Nagelkerke) 0.273 0.398 0.415 0.398 0.398 0.311 N 49,661 49,661 49,661 49,661 49,661 49,661 Notes: (i) A version of model (1) using separate country and time fixed effects, rather than interactions between them has an R2 equivalent of 0.211, and indicates greater support for populist parties from 2008 on; (ii) in model (2), coefficients on the three factors (composites of several variables) are 1.0 by construction, with variable weights based on coefficients from a preliminary regression with the full set of variables (Supplementary Table A1); (iii) in model (5), employment changes relate to the 3 years prior to the survey and are each relative to total employment in the base year. Stars indicate levels of statistical significance: * = 5%, **= 1%, *** = 0.1%. Sources: European Social Survey Database: waves 1–7, with employment data from the European Labour Force Survey (via Eurostat Regional database). View Large In relation to Personal Background (PB), the characteristics identified as of particular significance to support for populist parties were (in order of significance): • (Younger) Age • less than Tertiary Education • male gender • highly religious Catholicism/Orthodox Christianity • membership of the ethnic majority • not being Muslim • living in a non-urban area (countryside or village). The sign of the age effect is the reverse of that reported for UK referendum voting, but consistent for most countries in our sub-sample, with varying degrees of significance; it remains, but with less impact, in analyses based on actual voting. Among indicators of Personal Attitudes (PA) recorded by the survey, by far the strongest (as was the case in Inglehart and Norris’s (2016) ‘cultural values’) is a self-rating of left–right political position; after excluding this, it still left 11 significant attitudinal variables (two of which were themselves composites created by Inglehart and Norris): • anti-immigrant views (Inglehart and Norris’s scale, combining perceptions of negative effects on cultural life, the economy and the country as a place to live) • opposition to further migrants from a different ethnic background • distrust of the European Parliament • distrust and dissatisfaction with national government and politicians (scale from Inglehart and Norris) • satisfaction with the state of the economy • opposition to homosexual equality • support for more migrants from own ethnic background • valuing a strong state • valuing conformity with traditions/customs • belonging to some (unidentified) group subject to discrimination17 • seeing understanding of different people as unimportant. The oddity in this set is that personal dissatisfaction with the state of the economy (relative to the national average at the time) made support for populism less likely, rather than more. It should also be noted that greater trust in the national government did not always mitigate populist support; in Hungary, at least it increased it. The third set of salient indicators, relating to labour market position (LM) comprised (in order, again): • employment in an occupation where vocational qualifications were the norm • experience of an unemployment spell of over 3 months • working in one of the occupations particularly open to migrant workers (personal service, construction and cleaning) • working in manufacturing • employment in an establishment with fewer than 10 workers. Table 2 compares results from a regression including the composite variable for each of these three sets (model 2) with those from one with fixed effects only (model 1), three which introduce regional variables, in different ways (models 3–5) and a final one suppressing the attitudinal factor (model 6). In summary, what these show is that: • very strong fixed effects operate, varying jointly between places and periods; • but strong effects are still evident in relation to all three personal factors, most strongly (as in Inglehart and Norris) for attitudes and least for labour market position; • adding regional fixed effects shows that there is also a fairly important sub-national variation, over and above what can be accounted for by identified personal factors (operating in a consistent way); • allowing for possible aggregative effects from the three types of personal factor (over and above the purely compositional) accounts for hardly any of this implied regional effect—and that entirely through the labour market factor, which seems to exert a much stronger effect at the aggregate level; • for the personal background factor, its aggregated regional effect seems the reverse of that found at the individual level—perhaps because members of ethnic/religious minorities who individually oppose populism may by the size of their presence collectively encourage others to support it; • for the attitudinal factor, it seems that a concentration of people holding similar views does not of itself affect their translation into party support; • aggregate measures of employment growth make no contribution to explaining the regional effect; • excluding the attitudinal factor (as will be done throughout the next sub-section) considerably reduces the explanatory power of the regression, but gives a substantially stronger role to labour market position and makes the urban effect significant, suggesting that some labour and urban factors contribute to regional variations in attitudes of salience to populist politics. Testing the localist/cosmopolitan hypothesis about the geography of populism This section attempts to put some flesh on the bones of these schematic regressions by drawing on ideas from the second section above. In particular, it tries to build a more substantively interesting explanation of the uneven spatial map of populism within countries by testing three ideas: 1. That internationalisation has imposed three sorts of shock or stress on particular communities, and especially on some groups who may be particularly exposed to them: a. a risk of economic hardship via shifting of local jobs to competitors or sub-contracting facilities located overseas, most obviously relevant to manufacturing workers (though perhaps also for clerical workers)—with a risk factor related to local employment trends in that sector; b. a similar risk of hardship as a result of being displaced from some kinds of (remaining) local job by migrant workers with lower reservation wages, most obviously in relation to the three job types (personal service, construction and cleaning) where recent migrants are most over-represented, but also for those working in small establishments—with a risk factor related to the proportion of the local population who had arrived in (say) the last decade; and c. a risk to quality of life and sense of identity for those local residents with least adaptability and hence the strongest stake in a local status quo—which would presumably include the elderly and those restricted to areas of cheaper housing cost—with a risk factor here related to the significance of longer-term change in terms of migrants (say over the last two decades) and the size of the Muslim population; 2. That those more at risk of converting shock into support for a populist movement will be: a. people who have pursued a more active strategy of engagement with a particular place, business etc—signalled by their working in occupations characterised by vocational qualifications—rather than the cosmopolitan possibilities of university education; and/or b. those not affiliated to civil society organisations (for example unions or churches) involving strong trans-local forms of identity; and 3. That there will be spillover effects on support for populist parties from a stronger local representation of workers in occupations with a higher incidence of vocational qualifications (positively) and of graduates from tertiary education (negatively). Detailed results from testing these hypotheses sequentially are presented as models 1–3, alongside a baseline model 0 in Supplementary Table A.2. This is rather large, because it embodies quite a few (significant) interaction and ecological effects, of the kind we should expect if real geographical processes intervened between the attributes of their population members and political mobilisation. The main results can be summarised fairly simply in relation to the preceding hypotheses, however. To start with the first hypothesis, about groups who might be reacting to particular stresses imposed on them through internationalisation, only one is clearly supported by the evidence. Two of the sets of workers identified as vulnerable (in manufacturing and in a trio of occupations with easy entry) were more supportive of populism than their peers, but in neither case was this significantly related to the expected source of stress (industrial job loss and recent migration, respectively). There was, however, a possibly significant relationship between recent migration rates and populist support among workers in small establishments. Similarly, in relation to the idea that older residents might be particularly likely to react against a cultural shock from long-term changes in the population mix, there was a possibly significant positive relation between their support for populist parties and the scale of in-migration (though not with that of the Muslim population). For other locals, relations between populist support and different potential stressors were actually clearer. Two of the three demographic factors showed significant effects, though in (unexpectedly) different directions. On the one hand, the size of the Muslim population was positively related to populist support, but, on the other hand, the scale of long-term in-migration (over the previous 20 years) was negatively related to it. Shorter-term (10-year) migration apparently had no overall effect. On the economic side, the relation with recent industrial job loss was not statistically significant, but consistently negative (against what might have been expected). In terms of hard economic connections then, looking at the regional level seems to add nothing substantial to the cases either that populist support comes from people who have been materially left behind—or from regions with recently weak growth performance. In terms of the second hypothesis—about the significance of factors likely to be associated with localist rather than cosmopolitan orientations (types of qualification and/or organisational affiliation)—the evidence does provide some much clearer support. In relation to qualifications, the simple finding is that, while individuals with tertiary academic ones are much less likely to support populist parties than those without, those in occupations where vocational qualifications are the norm seem substantially more likely to do so than others (including the formally unqualified in other kinds of job). Moreover, the vocationally oriented groups are substantially more likely to support populist parties in those places with large migrant and Muslim populations, while graduates clearly show the opposite response, at least in relation to Muslims. If we can interpret these two qualification-related groups as standing more broadly for those with cosmopolitan or localist strategies of personal investment (in human and social capital), then it is the latter, ‘left behind’ group who are particularly associated with populism and especially its correlation with growing migrant/Muslim communities. In relation to two associational variables, current trade union membership and highly religious Catholic/Orthodox adherents,18 each seen as representing potential supports for non-localist attitudes, sources of identity and reference points, there was similar evidence that they did actually represent cosmopolitan effects. Specifically, members of these groups were significantly less likely to support populist parties in places with a higher proportion of long-term migrants (in both cases) and/or of Muslims (though only for the religious group). The degree of similarity here is notable, given that the base level of populist support was well above average for the highly religious Catholic/Orthodox group and well below for trade unionists. These interaction effects between likely cosmopolitans/localists and local migration levels in shaping populist support are akin to that reported by Lee et al. (2017) for non-movers in relation to Leave voting in the UK referendum. The third hypothesis had suggested that local attitudes and responses to populism were likely to be more than simply an aggregation of individuals acting out parts determined by their backgrounds, but also to be affected by social interactions in which the size of particular groups (and hence the strength of their institutions) made a difference. This was tested in relation to the same pairs of indicators of cosmopolitan/localist influences as for the examination of interactions under the second hypothesis, that is tertiary/vocational qualifications and union/religious associations, now aggregated to average values for observations in each NUTS1/2 region. An initial expectation was that (as with early ideas about neighbourhood effects in voting19), there might be positive spillovers from locally strong groupings, leading to bandwagon effects. The pattern that emerged was rather more complex, however. For the two qualifications variables, that relating to tertiary education showed no significant aggregate effects at all, while that for vocational qualifications showed a powerful positively reinforcing effect—maybe the more cosmopolitan group is simply less locally embedded. The two associational variables showed two other distinct patterns. The trade union density measure showed a moderate (if not quite significant) negative effect, reinforcing the tendency for the average union members—though this seems to have no real force outside the context of migration reception areas. The highly religious Catholic/Orthodox indicator shows a very strong negative one, which is very clearly against the pattern of populist support among this group, outside those areas. That might reflect them holding more than one kind of value deviating from the norm, with a salience that varies between areas; for example both a general social conservatism shared by many populists, and an inclination to charity which pulls the other way in areas with disadvantaged migrant communities. Such effects are evidently not simple in their geographies, but (with the probable exception of the graduate group) it seems that this set of variables, linked to the cosmopolitan/localist distinction, deliver their (geographic) impacts at least as much through aggregate and interactive effects as simply via the relative voting power of more/less populist-inclined kinds of people in areas with differing population mixes. Discussion of the geographies of populist politics has tended so far to one or other of a pair of reductionist explanations of its spatial patterning: one in terms of different local mixes of population groups with political preferences which are distinct but independent of location, and the other in terms of a homogeneous electorate responding in consistent ways to objectively different local economic situations. Neither of these seems at all consistent with a regional analysis of the ESS data, though evidence for the economistic version is much harder to find—the real geographies are substantially more complex, and deserve unpicking to get a handle on what has been happening. Conclusions Within the UK, one recurring line of argument since the Brexit Referendum has been in terms of ‘the chickens’ of regional inequality (eventually) ‘coming home to roost’. In other words, that, as long forecast, persistent patterns of uneven economic development would eventually lead to major political breakdown of some kind. One problem with this simplistic idea was, and still is, that the political and economic geographies do not actually fit at all well. The same is true of more micro-level conjectures that link an anti-establishment revolt with economic vulnerability or working class status. The micro-evidence is much more supportive of the idea that powerful (individual) attitudinal factors, associated in some ways with different groups, have spatially uneven effects simply because local populations differ in occupational and educational terms. This interpretation fits both the British (referendum) and the European (populist party) evidence very much better than the economistic story. But regional analyses provide evidence that it too is over-reductive, in its assumption of a single underlying national pattern. In the British referendum analysis, the effects attributed to particular occupational (and probably educational) groups are just too strong to reflect simple mixity. And in the European analysis of the support for populist parties, evidence was found of radically different responses by several groups to the same sources of cultural shock, as well as of ecological effects from their relative size. In particular, this was found to apply to groups with different types of qualification, reflecting a localist/cosmopolitan divide associated with different kinds of stake in stability, but also to ones with differing trans-local connections. The balance between these is not, however, just an arbitrary fact about particular places but (we should assume) outcomes of the sequence of economic roles they have played (as Massey, 1984, proposed). This might perhaps explain why, when recent economic change seems fairly irrelevant to the local strength of populism, very long-term structural changes do seem to have a bearing (Langella and Manning, 2016). Supplementary Material Supplementary material is available at Cambridge Journal of Regions, Economy and Society online. Acknowledgements I am grateful to Murray Low and participants in an informal post-referendum LSE seminar, which initiated this article; Michael Storper for discussion and helpful suggestions during its development: to all at the CJRES ‘Globalisation in Crisis’ conference sessions where a first version was presented; and to Tony Champion for pointing out a significant omission in its final stage. References Ashcroft, M. ( 2012) The UKIP threat is not about Europe. Available online at: http://lordashcroftpolls.com[Accessed 18 September 2017]. Benda, J. ( 1927) La Trahison des Clercs . Paris: Ed. Grasset. Bernstein, B. ( 1971) Class, Codes and Control: Theoretical Studies Towards Sociology of Language . London: Routledge & Kegan Paul. Brown, G. ( 2014) My Scotland, Our Britain: A Future Worth Sharing . London: Simon and Schuster. Clarke, S. and Whittaker, M. ( 2016) The Importance of Place: Explaining the Characteristics Underpinning the Brexit Vote Across Different Parts of the UK . London: Resolution Foundation. Cox, K. R. ( 1969) The voting decision in a spatial context, Progress in Geography , 1: 81– 117. Curtice, J. ( 2016) A Nation at Unease With Itself? Britain on the Eve of the EU Referendum . Academy of Social Sciences, 14 June. Available online at: https://acss.org.uk/news/eu-referendum-polling-disparities-expose-demographic-divisions-across-uk/. European Social Survey( 2014) ESS1–7, ESS Cumulative File Rounds 1–7, Data File Edition 1.0 . ESS, NSD – Norwegian Centre for Research Data. Norway: Data Archive and Distributor of ESS data for ESS ERIC. Financial Times( 2016) Britain has had enough of experts, says Gove, Financial Times , 3 June. Available online at: https://www.ft.com/content/3be49734-29cb-11e6-83e4-abc22d5d108c?mhq5j=e3. Goodhart, D. ( 2017) The Road to Somewhere: The Populist Revolt and the Future of Politics . London: Hurst. Goodwin, M. and Heath, O. ( 2016a) The 2016 referendum, Brexit and the left behind: an aggregate-level analysis of the result, Political Quarterly , 87: 323– 332. Google Scholar CrossRef Search ADS Goodwin, M. and Heath, O. ( 2016b) Brexit vote explained: poverty, low skills and lack of opportunities . Joseph Rowntree Foundation. Available online at: https://www.jrf.org.uk/report/brexit-vote-explained-poverty-low-skills-and-lack-opportunities?gclid=EAIaIQobChMIqeyL7ICY1gIVhhobCh0IrwaBEAAYASAAEgIhdvD_BwE [Accessed 18 September 2017]. Gordon, I. R. ( 2016) Less certainty and more choice: still waiting for a credible metropolitan strategy, Town and Country Planning , August, 317– 319. Harris, R. and Charlton, M. ( 2016) Voting out of the European Union: exploring the geography of Leave, Environment and Planning A , 48: 2116– 2128. Google Scholar CrossRef Search ADS Inglehart, R. F. ( 1971) The silent revolution in Europe: inter-generational change in post-industrial societies, American Political Science Review , 65: 991– 1017. Google Scholar CrossRef Search ADS Inglehart, R. F. and Norris, P.(August 2016) Trump, Brexit and the Rise of Populism: Economic Have-Nots and Cultural Backlash , Faculty Research Working Paper 16-026. Kennedy School of Government, Harvard University. Kaufmann, E. ( 2016) It’s not the economy stupid: Brexit as a story of personal values. British Politics and Policy Blog, LSE. Available online at: http://blogs.lse.ac.uk/politicsandpolicy/personal-values-brexit-vote/ [Accessed 18 September 2017]. Kimball, R. ( 1992) The treason of the intellectuals and the “undoing of thought”, The New Criterion , 11: 10– 22. Langella, M. and Manning, A. ( 2016) Who voted leave?, Centrepiece: Magazine of the Centre for Economic Performance , Autumn, 6– 11. Lee, N., Morris, K. and Kemeny, T. ( 2017) Immobility and the Brexit vote, Cambridge Journal of Regions, Economy and Society , 11: 143– 163. Lord Ashcroft Polls( 2016) How the UK voted: full tables. Available online at: http://lordashcroftpolls.com [Accessed 18 September 2017]. Massey, D. B. ( 1984) Spatial Divisions of Labour: Social Structures and Geography of Production . Basingstoke: Palgrave Macmillan. Merton, R. K. ( 1957) Patterns of influence: local and cosmopolitan influentials. In Social Theory and Social Structure , rev. edn. New York, NY: Free Press. Mudde, C. ( 2004) The populist zeitgeist. Government and Opposition , 39: 542– 563. Google Scholar CrossRef Search ADS Mudde, C. ( 2007) The populist radical right: a pathological normalcy. West European Politics , 33: 1167– 1186. Google Scholar CrossRef Search ADS NatCen( 2014) British Social Attitudes 31st Report . London: NatCen; data-file from UK Data Archive, University of Essex. Rodríguez-Pose, A. ( 2017) The revenge of the places that don’t matter (and what to do about it), Cambridge Journal of Regions, Economy and Society , 11: 189– 209. Endnotes 1 This quote is better seen as combining personal impatience with populist rhetoric than as an expression of an authentically populist sentiment, given the double role of its author Michael Gove (along with Boris Johnson) as a scribbler/politician with some intellectual credentials. He would thus definitely count as a ‘clerc’ in Benda’s (1927) terms (covering ‘academics and journalists, pundits, moralists and pontificators of all varieties’; Kimball, 1992), though Gove’s target is more the authority of (social) science than that of Enlightenment values, and reflective more of opportunism than the nihilism that Benda had seen. 2 My own sense of late discovery is reflected in a piece written just before the referendum (Gordon, 2016). 3 For example, by Goodwin and Heath (2016a), though this language soon came to be taken up by government figures, including Mark Carney, from the Bank of England, and then Theresa May (in her November 2016 Lord Mayor’s Banquet speech), adding liberalism to globalisation as sources of this division. 4 These are the areas for which results were officially published. 5 These are intertwined themes in Rodríguez-Pose’s (2017) analysis. 6 There are connections here to Bernstein’s (1971) socio-linguistic distinction between restricted and elaborated codes, with much of the meaning in the former being conveyed implicitly between people operating with a set of shared assumptions and expectations, whereas the latter (operating beyond these) involves more verbally explicit indications of intent. 7 Among the five social classes included among Inglehart and Norris’s (2016) economic indicators, the two which showed significant deviations (after controlling for education) were the petite bourgeoisie, with a substantially higher propensity to support populist parties, and professionals/managers with a considerably lower one. 8 Which clearly had a populist dimension to it. 9 Or, in Northern Ireland, parliamentary constituencies. 10 Subject to using the same set in each of the relevant regressions for consistency, even if a category ceased to be significant in one of the three reported models. 11 This is just a rule of thumb because other factors are also involved which may not simply cancel out—especially if they are correlated, as in the occupation/qualification case referred to below. 12 The Northern Ireland case is different, in that religion (and hence nationalist/unionist sentiments) was clearly the major explanatory factor behind marked spatial variations in voting patterns—though a very strong association between the Leave vote and the share of Protestants in a constituency was markedly moderated in areas with higher proportions of graduates in the population. 13 There are a series of distinctive local institutional factors (including Catholicism, strong labourism and a maritime history), which might plausibly be invoked. But it is also notable that national leaders and media backers of the Leave campaign (notably Boris Johnson and the Sun newspaper) are prominent among those seen to have treated the Merseyside community with notable disrespect over the years. Being seen ‘not to matter’ is not simply about hard-nosed economic judgements, nor is (nationalistic) populism always going to be the natural political response to mistreatment by outsiders. 14 The ‘closeness’ measure was preferred despite a rather lower response rate, to avoid the contingencies both of when relevant elections had occurred and of electoral arrangements with different implications for the likelihood of a populist vote being registered by potential supporters. 15 Further information about data and methods is provided in the Supplementary Material. 16 At this stage, the trade union membership variable (used in the second stage) was not included, reflecting ambiguity about whether it was part of personal background or of labour market position. 17 That is, some group not based on ethnicity/race, nationality, religion, gender age, sexuality or disability. 18 As explained in the Supplementary Material, the focus on the Catholic/Orthodox sub-group of the highly religious does not involve any judgement about what might be the case for other sub-groups (for example Protestants or Jews). 19 Notably Cox (1969). © The Author(s) 2018. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved. For permissions, please email: firstname.lastname@example.org
Cambridge Journal of Regions, Economy and Society – Oxford University Press
Published: Mar 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera