“The grass is greener on the other side”: The relationship between the Brexit referendum results and spatial inequalities at the local level
2021; Elsevier BV; Volume: 100; Issue: 6 Linguagem: Inglês
10.1111/pirs.12630
ISSN1435-5957
AutoresDiana Gutiérrez Posada, María Plotnikova, Fernando Rubiera Morollón,
Tópico(s)Urban, Neighborhood, and Segregation Studies
ResumoDespite seven decades of development of the European Union project, on 23 June 2016, the United Kingdom, Europe and the rest of the world were surprised when the Leave campaign won the Brexit referendum, offering an extraordinary case study for researchers. We spatially disaggregate the vote share data, which allows us to explore where anti-European sentiment took root in the UK and why. In this paper, a spatial dependence model is applied to clarify and quantify the relevance of the different dimensions—demographic, cultural/educational and economic—that play a role in explaining the rise of support for the Leave campaign. The analysis is conducted at the local level, using local authorities (LAs) as the spatial unit of analysis due to the combination of official datasets with newly generated data in the context of an EU H2020 project. A new indicator capturing the affluence of each local area relative to its close neighbours is proposed and included in the model. In general, we observe that most of the main conclusions obtained by large regions or at the national level also hold at the local scale. However, it is particularly interesting that inequalities by LAs are clearly significant, indicating a marked influence on voters' decisions that have thus far been unaccounted for. This result provides further support for the existence of, to use Andrés Rodriguez-Pose's terminology, an even more intense "revenge of the places that do not matter" at the local scale. A pesar de siete décadas de desarrollo del proyecto de la Unión Europea, el 23 de junio de 2016, el Reino Unido, Europa y el resto del mundo se vieron sorprendidos cuando la campaña Leave ganó el referéndum de Brexit, lo que ofreció un estudio de caso extraordinario para la investigación. En este artículo se desagregaron espacialmente los datos de distribución del voto, lo que permitió examinar en dónde arraigó el sentimiento antieuropeo en el Reino Unido y por qué. Se aplicó un modelo de dependencia espacial para aclarar y cuantificar la relevancia de las diferentes dimensiones (demográfica, cultural/educativa y económica) que intervienen en la explicación del aumento del apoyo a la campaña Leave. El análisis se realizó a nivel local, utilizando las autoridades locales (AL) como unidad espacial de análisis debido a la combinación de conjuntos de datos oficiales con datos recién generados en el contexto de un proyecto Horizonte 2020 de la UE. Se propone un nuevo indicador que capta la prosperidad de cada área local en relación con sus vecinas cercanas, que se incluyó en el modelo. En general, se observó que la mayoría de las conclusiones principales obtenidas por las grandes regiones o a nivel nacional aplican también a escala local. Sin embargo, es especialmente interesante que las desigualdades a nivel de AL son claramente significativas, lo que indica una marcada influencia en las decisiones de los votantes que hasta ahora no se han tenido en cuenta. Este resultado proporciona apoyo adicional a la existencia de, según la terminología de Andrés Rodríguez-Pose, una "venganza de los sitios que no importan" aún más intensa a escala local. 欧州連合(EU)のプロジェクトが70年にわたって進展してきたにもかかわらず、2016年6月23日に英国のEU離脱の是非を問う国民投票では離脱運動が勝利したが、英国をはじめ、欧州、その他の国々はその結果に驚き、研究者らに特別なケーススタディを提供することとなった。我々は、得票率のデータを空間的に分析することにより、英国のどの地域に反EU感情が根付いたのか、またその理由を探索する。本稿では、空間依存モデルを用いて、離脱運動の支持率上昇の原因となる様々な側面(人口統計学的、文化的/教育的および経済的)の重要性を明らかにし、定量化する。地方当局(local authority:LA)を、EUのホライズン2020プロジェクトのなかで新たに生成されたデータと公式データセットの組み合わせによる分析の空間単位として用いて、地域レベルでの分析を実施する。隣接する地域と比較した各地域の豊かさを測る新しい指標を提案し、分析モデルに含めた。概して、大規模な地域や国レベルで得られた主な結論のほとんどは、地域規模にも当てはまると考えられる。しかし、LA間での不平等は明らかに重大であり、これまで明らかにされなかった有権者の意思決定に大きな影響を与えていることを示しており、非常に興味深い。この結果は、Andrés Rodriguez-Poseの言葉を借りれば、地域規模ではさらに激しい「revenge of the places that do not matter(重要ではない場所の復讐)」が存在することをより強く裏付けるものである。 The success of the campaign for the exit of the United Kingdom from the European Union in the referendum of 23 June 2016, generated a shockwave not only in the United Kingdom but also throughout Europe and the world. After seven decades of European Union construction focused on economic and social improvements, the citizens of the United Kingdom and the rest of the world witnessed the surprising triumph of Brexit, opening the door to an uncertain and complex future. However, seen in perspective, Brexit can be considered a chronicle of an event foretold: after the outbreak of the severe international economic crisis of 2007–2008, the insufficient response of European Union institutions generated a climate of rapidly spreading general dissatisfaction with the political system. This displeasure has not stopped growing since, with Brexit being a clear materialization of this wave of discontent, albeit neither the only nor the last one. To prevent similar situations from occurring in other EU countries in the imminent future, it is crucial to understand the elements that came together to generate the discomfort and disaffection with European Union institutions in the UK. From the UK's perspective, it is important to understand all the factors that led to such a dramatic decision affecting its future social and economic developments, especially in the local domain (Billing et al., 2019). Additionally, the Brexit referendum gives social science researchers valuable data on the location of anti-EU sentiment-data that can be combined with other location-specific information to test hypotheses on the determinants of vote outcomes. Consequently, a large and ever-growing number of studies have used this UK case to understand the general phenomenon of political discontent, leaving behind a much clearer knowledge of how, where and why anti-immigration, anti-globalization and anti-Europe narratives succeed. This literature reviews how the emergence of discontent channelled into anti-European voting has been driven by multidimensional factors, with demographic, cultural/educational and socio-economic characteristics all coming together to play a part. Several works also point to the existence of a clear geographical element insofar as there was a spatial concentration of the pro-Leave vote. Among the studies focusing on the geographical dimension, two in particular draw attention to the role of regional inequality in shaping voting outcomes. Rodríguez-Pose (2018) contributes the very provocative idea that gives the title to his paper: the revenge of the places that do not matter. This phenomenon describes how deprived areas, as the main enclaves of support for populist political alternatives and the usual bearers of anti-Europe, anti-immigration or anti-globalization sentiment, vote against institutions that are considered to have forgotten about an areas' needs and failed to provide solutions to their concerns. McCann (2016) highlights the common misconception that the metropolitan elites of London are the only ones who have benefited and that other regions have not received anything from the EU. As McCann points out, although many of these weaker regions have certainly seen the downside of internationalization, belonging to the European Union actually mitigated those perverse effects, although this fact did not reach the population as the disparities between wealthier and poorer regions grew steadily larger. Both Rodríguez-Pose (2018) and McCann (2016) agree that territorial inequality is what matters, which is not to say that interpersonal inequality is not important but rather that the challenge to the system has come from this neglected source of inequality. Previous literature has developed analyses on a regional scale; individual decisions are aggregated by areal units at the spatial level of large regions. When insufficient spatial disaggregation is used, the local heterogeneity that may exist within areal units may be cancelled out in the aggregation, losing nuances in the analysis or reaching biased conclusions. In this paper, our main goal is to contribute to the previous literature and, in particular, to the ideas of Rodríguez-Pose (2018), Los et al. (2017) and McCann (2016), reproducing the analysis of the spatial behaviour of the pro-Leave vote in the Brexit referendum but using a more disaggregated unit of spatial analysis: local authorities (LAs hereinafter). Working at a local scale, we can observe whether the spatial patterns of discontent remain and identify the nuances derived from more spatially detailed data. In addition, our study contributes to the wide existing literature in the following ways: first, by generating local income data that are consistent with the more aggregated spatial scales (as part of the deliverables for the EU H2020 IMAJINE Project); second, by using these new data as a key element to analyse the spatial pattern of the Leave vote; and third, by creating a novel indicator reflecting an area's average income difference with its neighbours to test how local inequalities influenced Brexit. Finally, high spatial disaggregation requires the use of sophisticated spatial econometric techniques. In this paper, the econometric specification carefully takes into account the spatial dimension of the data. Our results indicate that local spatial inequality matters for the Leave vote. The relative income position of each locality within a group of neighbours matters in the sense that after controlling for a host of location characteristics, local authorities that are poorer than their neighbours tend to have a higher Leave vote share, and vice versa. In light of the existing literature on the Brexit vote, we interpret the results as supporting the argument of the discontent of "left-behind" places channelling into a punishment vote against the status quo of UK membership in the EU (McCann, 2016; Rodríguez-Pose, 2018). The paper is organized as follows. In the next section (Section 2), we include a brief review of the extensive literature on the geography of anti-European sentiment. This review helps identify the dimensions that we must consider in our analysis. Section 3 introduces the empirical setting of the study: the spatial unit used (local authority) (subsection 3.1), our response variable (subsection 3.2), and the set of explanatory factors we intend to use to quantify the dimensions of the phenomenon (subsection 3.4), with special attention to our measure of local inequalities (subsection 3.3). Section 4 gathers all the information provided previously into the chosen specification of spatial dependence. The results are presented and discussed in Section 5. The paper concludes with a final summary and policy recommendations (Section 6). Throughout the last decade, we have witnessed a general feeling of disaffection with the system and institutions, a climate of political discontent. This idea of discontent expresses different types of reactions that can be grouped into categories such as: (i) the rise of populist parties and movements with an anti-European stance in many EU countries (Algan et al., 2017; Vasilopoulou, 2018); (ii) reduced trust in the European Union and national institutions, which has been referred to as a trust crisis (European Commission, 2015); (iii) reduced support for the European project in general, as well as reduced levels of identification with the European Union (Flesher, 2017); and (iv) a reduction in social engagement and participation among the citizens of EU countries (Magni, 2017). Furthermore, as noted in the literature, these dimensions are localized in the sense that they are more prevalent in some places than in others. More boldly speaking, there is a geography of discontent (Los et al., 2017) such that the rise of populist movements, Euroscepticism, the loss of trust in the European Union and national institutions and the reduction of social engagement exhibit particular spatial patterns (Georgiadou et al., 2018; Oesch, 2008). In the context of this growing literature concerned with understanding the drivers of the discontent in general and the patterns across space in particular, Brexit has been a case study of enormous relevance, attracting substantial academic attention and extensive analysis from the social sciences. Thanks to this literature, we are starting to understand what the drivers of Brexit were, why it happened, who Leave voters are and where the Brexit campaign was most successful. Through this particular case, we can better understand the phenomena of institutional disaffection, anti-Europeanism or general political discontent. The percentage of votes in favour of Brexit increased monotonically with age, going from only 27% of the group aged 18–24 to 60% of those aged 65+, while conversely, it decreased monotonically with education (Crescenzi et al., 2017). Academic studies have rapidly confirmed the effects of age and education on the probability of voting to leave the EU (see Arnorrsson & Zoega, 2016; Clarke, Whiteley, et al., 2016; Harris & Charlton, 2017; and Manley et al., 2017, among others). However, in addition, there was a clear relationship between voting Leave, concern over immigration and the slippery concept of identity. With different approaches and techniques, authors such as Hobolt and Wratil (2016), Hobolt (2016), Arnorrsson and Zoega (2016), Clarke et al. (2016), Langella and Manning (2016) and Goodwin and Heath (2016) coincide in finding that voters who express concerns about immigration and multi-culturalism voted Leave. What about economic variables such as income, deprivation or unemployment? Kaufmann (2016, p. 1) summarizes the main conclusion of most of the previous studies: "Age, education, national identity and ethnicity are more important than income and occupation," However, as noted by Los et al. (2017), conceptual perspectives motivating Brexit exclusively based on cultural issues (identity, national sovereignty, etc.) have been regarded by some as inadequate to thoroughly describe the geography of the Brexit vote. This alternative perspective claims that variables accounting for the economic conditions of citizens and the economic geography of UK regions are at least as important as culture and identity in determining individual attitudes towards the EU and voting patterns in the 2016 referendum (Crescenzi et al., 2017). Indeed, empirical analyses performing comprehensive investigations of the Brexit vote and considering not only demographic and political variables but also proxy variables for local economic structure and "economic exposure" to the rest of the European Union all seem to suggest that economic factors played a significant role (see Arnorrsson & Zoega, 2016; Becker et al., 2017; or Hobolt, 2016). Additionally, Darvas (2016) claims that wage inequality and poverty were two crucial drivers of Brexit. Clarke et al. (2017) demonstrate that economic cost–benefit evaluations are at least as influential as any sense of identity. Indeed, Curtice (2017) claims that the perceived impact of leaving the EU on the economy is the variable most strongly related to how people voted. Clarke et al. (2016) show how labour market conditions are crucial in conditioning voters' choices. Higher employment levels are associated with a lower propensity to vote Leave, suggesting that unemployed people were more prone to support Brexit than those with safe salaries and jobs (see Becker et al., 2017 or Alabrese et al., 2019; Clarke et al., 2017; Goodwin & Heath, 2016; Goodwin & Milazzo, 2017). Los et al. (2017, p. 788) summarize the conclusions of the empirical work to date, stating that econometric studies "all suggest that local economic conditions were the single most important factor driving the pattern of voting, interacting with the characteristics of the individuals making up that locality." Harris and Charlton (2017, p. 2127) read the general context very well: "Ultimately, the story is perhaps less about the EU itself but one of industrial decline and growing social and economic inequality, overlapping with nationalism and political beliefs." Focusing on the geographical perspective, it is clear that within England, there were marked geographical differences in voting patterns. Remain votes dominated in London and many parts of the home counties—the western arc around London from Cambridge to Oxford and down to Surrey—along with some of Britain's major cities such as Leeds, Manchester, Cardiff, Leicester, Bristol, Liverpool, Edinburgh and Glasgow. In addition, Remain voter preferences in both Scotland and Northern Ireland displayed markedly different patterns from those in localities that were perceived to have benefited most from globalization (Coyle, 2016). As noted by Goodwin and Heath (2016), the geography of deprivation and prosperity both interacted with and overlaid each of the individual-specific explanatory variables. In the context of this previous literature on the spatial patterns of support for Leave, Rodríguez-Pose (2018) pitches a stirring idea: the revenge of the places that do not matter, whereby those areas specialized in declining activities and located on the periphery voted down a system that they perceived to have quelled their potential and driven them down a road in which the future offers no opportunities, no jobs and no hope. In line with this idea of the generation of a feeling of spatial revenge, Los et al. (2017) highlight that people in less prosperous regions who sensed that they had suffered under modern globalization were much more likely to vote Leave; this holds even after personal characteristics are controlled for. Ironically, the regions that voted Leave also tended to be more dependent on Europe than their counterparts that voted Remain (Los et al., 2017). These Leave-voting regions tended to be more dependent on EU markets for their trade and prosperity, and many of them had benefited significantly from regional development funding from the EU Cohesion Policy over many years. Additionally, as McCann (2016) emphasizes, while many of these weaker regions have suffered under globalization, they actually benefited from trade integration under the EU, with the latter process partially mitigating the effects of the former. However, the public did not understand this. According to McCann (2018, p. 4), "In the UK, an important pro-Leave narrative was that the 'metropolitan elites' of London were the only real beneficiaries of EU membership, while other regions had not enjoyed the benefits of European economic integration. In contrast, empirically, it is now clear that this metropolitan elite argument was completely incorrect and that the regions that most benefited from the EU markets for their viability were largely the non-core weaker regions of the UK." In terms of research design, the studies can be divided along the lines of the type of unit of analysis they are using, individuals or area units. Individual responses form the variable of interest in individual-level studies; in the latter case, responses are aggregated by areal units, such as local authorities or regions. Most of the previous literature on the geography of discontent developed analyses on a regional scale. When using regions as units of analysis, one must refrain from the ecological fallacy of assuming that conclusions about individuals based on population-level or "ecological" data will hold. In this case, the analysis becomes about places and their characteristics, and the research question becomes, "What is it about places that resulted in certain vote outcomes?" In addition, there are many studies on the determinants of the Brexit vote using individual-level data on voting intention. The Understanding Society Survey includes questions about the Brexit referendum such as "Should the United Kingdom remain a member of the European Union or leave the European Union?" In a study using voting intent as a dependent variable, Lee, Morris, and Kemeny (2018) place individual mobility as the central indicator of the social division that delivered Brexit. The argument is based on Goodhart's (2017) book on Brexit distinguishing between "Anywheres2 and "Somewheres." Goodhart's Anywheres are educated elites whose identities are not tied to any particular local community, or even Britain. These elites are said to have overwhelmingly voted to remain in the European Union, while Somewheres, with strongly place-bound identities and associations, voted to leave it. "The immobile are, by definition, more tied to their local area and thus more exposed to external change, so local change may have influenced them more than other groups, with the Brexit vote being a way of protesting against changes to their local environment" (Lee, Morris, & Kemeny, 2018, p. 155). The main explanatory variable here is whether a respondent was living in the same local area in which they were born. The authors suggest that immobility increased Leave votes in counties experiencing growth in their non-white population. Growth in the average local wage lessens the effect of immobility. Similar to other individual-level studies, this study uses regional controls and area-aggregated variables to account for spatial effects. Despite the interest it arouses, this widely used approach may not reflect the complexity of spatial effects. There are notable exceptions, such as the contribution made by Harris and Charlton (2017), who use a multilevel model with local authorities nested in regions to account for spatial effects in voting outcomes, or Abreu and Öner (2020), who explore the influence of the interaction between individual features and geographical context on voting behaviour (Table 1). Analysis at the individual level and the large spatial scale of NUTS regions may miss important determinants acting at the intermediate, smaller (spatial) scale. A perusal of the area-as-unit-of-analysis literature shows that analysis at different, particularly finer, spatial scales may yield different results; taking the politically contentious issue of in-migration, Colantone and Stanig cited in Lee et al. (2018, p. 145) find that stocks and flows of immigrants in NUTS 3 regions are unrelated to voting intention, "a finding that stands in direct contrast to that found by Goodwin and Milazzo (2017), who use more disaggregated parliamentary constituencies." Additionally, the contrast between the voting outcomes of large cities and rural areas or between the centre and periphery within regions suggests that the analysis should be carried out using smaller spatial units, such as local authorities. There is a dearth of studies on voting outcomes using meso or local area units that could potentially add to our understanding of the geographical determinants of votes filling the middle ground between studies using highly aggregated regional-scale and individual-level studies. A major obstacle to performing analysis at the local or medium scale is the absence of data on income or poverty at levels of disaggregation lower than the regional level. Supported by the European Union H2020 IMAJINE project's development of a database of socio-economic variables at a local scale, in this paper, we propose to analyse how socio-economic variables and spatial disparities can contribute to explaining the Brexit vote, and we test the Rodríguez-Pose (2018) idea of the revenge of places that do not matter at this level of disaggregation. The objective of this work is to study the drivers of Brexit success with a local-level analysis, placing special emphasis on understanding the role of local spatial inequalities. We should therefore use a highly disaggregated spatial unit. In this study, we propose to perform the analysis at the level of local administrative units, LAU level 1 in the EU classification. This is formerly the NUTS 4 level of municipalities or equivalent. These units closely align with the level of local government in the UK, providing local public services in the area. A wide range of socio-economic data from the 2011 Census is available at the LAU 1 level for the UK, making it possible to create variables that have been used in previous studies on determinants of the Brexit vote, namely, demographics, education, labour market condition variables. Therefore, we consider this to be the most appropriate local spatial unit to carry out our research. We refer to these units as local authorities (LAs). However, at the same time, we are strongly limited by the existence of official statistical information. In this study, we resort to several sources of data, particularly the 2011 census safeguarded microdata (a sample of 5% of the local population), containing individual data coded at the LA level. The original count of LAU 1 areas (excluding Northern Ireland) was 406. However, the census microdata aggregate localities with less than 120,000 inhabitants to a neighbouring area, making the number of units used in this study 265. Our dependent variable is the Leave vote share by LAs recorded by the Electoral Commission of the UK. The spatial distribution of Leave vote support by LAs is plotted in Figure 1, which shows a heterogeneous landscape where some areas, such as London and its neighbouring areas in the west (Oxfordshire and Hampshire) and the south (Surrey, West Sussex and East Sussex), exhibit a share of Leave votes below 45%. On the other hand, some LAs in western Britain, namely, the coasts of Lincolnshire and Norfolk, show higher support for the Leave option, which garnered approximately 60% to 75% of the local votes. Source: Own elaboration using data from the Electoral Commission of the UK Apart from using information at the local level, by LAs, of demographic or educational characteristics and the economic circumstances and expectations that have been considered in previous studies, in this research we are particularly interested in the role of spatial economic inequality at the local scale (by LA). This objective, however, is also affected by the absence of income or production data at a local level (by LA). Estimation of this type of information has been one of the objectives of the European Union H2020 IMAJINE project, an attempt to generate reliable information at the local level that is coherent with official sources at more aggregated (regional and national) levels. This project aims to study spatial differences of various natures across EU territories using several socio-economic indicators and explicitly considering the role played by the spatial scale. More specifically, one part of the project consists of disaggregating the income and wellbeing indicators that are available for several EU countries at an aggregated regional level (NUTS 1 or NUTS 2 regions) to produce analogous measures at the subregional or local scale (NUTS 3 or lower). The result of this disaggregation will enable the quantification of potential inequalities between territories that could be masked as a consequence of data aggregation (e.g., urban–rural gaps within regions).1 As further explained below, the main advantage of the estimated data used in this analysis is its coherence with regional and national aggregates. As in any analysis relying on estimated variables, this one is critically affected by the existing gap between the estimated and the actual value of the variable in question. The lack of information is the fundamental reason behind the use of an estimated variable, but we believe that the procedure summarized next (especially the second stage) makes the estimation more reliable, as it allows for the imputation of a value under high uncertainty that can satisfy conditions derived from observable data. The procedure followed to produce these spatially disaggregated data has two stages, as explained in Fernández-Vázquez et al. (2018). In the first stage, the imputation technique proposed in Elbers et al. (2003) and Tarozzi and Deaton (2009) is applied. Although extensive explanations can be found in the mentioned paper, in summary, the procedure combines information from the European Union Social Indicators and Living Conditions (EU-SILC) survey and the population census (PC) for a given economy. The EU-SILC contains detailed socio-economic information about the households surveyed but no details about their geographical location beneath the NUTS 1 level. Consequently, estimates based on the EU-SILC typically do not allow the inference of income figures for subregional units such as municipalities or cities. On the other hand, microdata from the PCs contain geographical information on the individuals surveyed at a disaggregated scale, but economic indicators, and more specifically income figures, are not generally available. The strategy suggested in Elbers et al. (2003) and in Tarozzi and Deaton (2009) consists of estimating regression models of the indicator of interest (y) on a set of regressors (Z) that are observable in both the EU-SILC and the PC. Once these models are estimated at an aggregated spatial scale, the results for the parameters are projected over the households surveyed in the PC. Since the PC has the detailed geographical location of the households, this technique enables the estimation of y at the same disaggregated spatial scale presented in the PC. The second stage seeks to guarantee consistency between the estimates for the subregional units and the regional aggregates from the official dataset (i.e., the EU-SILC), since the sum of the estimates for the households (i = 1,…,D) in a region may be larger or smaller than the regional figure. Fernández-Vázquez et al. (2018) propose a solution to this inconsistency based on generalized maximum entropy (GME) to adjust the estimates and make them consistent with the official aggregates. Using GME, the authors turn the estimations from the first stage into an optimization problem, constrained by the linear relationships to be satisfied (higher-level aggregates). The final value responds to the idea that each possible value that the variable of interest may take has an associ
Referência(s)