People or Places that Don’t Matter? Individual and Contextual Determinants of the Geography of Discontent

abstract

The rise of a geography of political discontent in the EU documented in recent studies highlights a strong spatial association between antisystem voting, regional economic decline, and poor occupational opportunities, suggesting that economic disparities within the EU are the origin of some of the most recent and shocking political events like Brexit. This article reexamines this statement by disentangling the effect on individual and political discontent of different socioeconomic disadvantage conditions at the interregional, intraregional, and individual level. Making use of a large data set on the individual and political discontent perceived by EU citizens between 2013 and 2018, our analysis confirms that a geography of discontent exists across EU regions. Nevertheless, our findings also highlight that intraregional inequalities do matter for individual discontent, and individual socioeconomic disadvantage conditions amplify further this negative effect.

Key words:

• geography of discontent
• intraregional inequalities
• interregional inequalities
• EU regions

JEL codes:

• R12
• I31

Acknowledgments

We are grateful to the journal editors and to the three anonymous referees for their precious and detailed suggestions throughout the review process.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ We thank an anonymous reviewer for comments on the conceptual framing and empirical measurement of this issue.

² Distrust in the EU and its institutions is obviously neither a measure of electoral outcomes nor of vote intention, but it is reasonably associated with both.

³ In short, by using Euroskeptic voting data only, there is the risk of overlooking other important concomitant factors that may affect the measurement and the analysis of individual and political discontent.

⁴ It is interesting to note that while regional economic decline is persistently related to antisystem voting, the same does not hold for the overall level of wealth. Alaimo and Solivetti (2019) do not spot any significant association between the support for Brexit and regional gross value added per capita. Dijkstra, Poelman, and Rodríguez-Pose (2020) find a positive and significant effect of gross domestic product (GDP) level on Euroskeptic preferences once the growth rate of GDP is controlled for. This suggests that people in wealthier places are more likely, keeping constant the recent economic trend, to support Euroskeptic parties, in contrast with the general findings of this literature.

⁵ We are aware of the broad debate in the literature about the differences and/or overlap between the notions of happiness, life satisfaction, and subjective well-being. To facilitate the discussion, however, we use the terms interchangeably. See Lenzi and Perucca (2020) for details on this aspect.

⁶ The complete list of references to the EB survey studies used in the present analysis is reported in Table A1 in the online material. Data from these survey studies has been pooled, in order to increase the sample size as much as possible. Malta and Croatia are not included in our analysis due to data unavailability.

⁷ The discontent category groups not very satisfied and not at all satisfied respondents and the content category groups very satisfied and fairly satisfied ones.

⁸ The study by Dijkstra, Poelman, and Rodríguez-Pose (2020) is the only one analyzing the association between Euroskeptic voting and cross-regional disparities using data for all EU countries.

⁹ Among the studies reviewed above, data on electoral districts are used by Goodwin and Health (2016), Becker, Fetzer, and Kovy (2017), Goodwin and Milazzo (2017), Garretsen et al. (2018), Gordon (2018), Dijkstra, Poelman, and Rodríguez-Pose (2020).

¹⁰ The studies by Hobolt (2016), Antonucci et al. (2017), Lee, Morris, and Kemeny (2018) can be included in this category.

¹¹ In their analysis of populist parties in Italy and the Netherlands, Pirro and Van Kessel (2018, 329) write that “in times of crisis, populist parties can use Eurosceptic frames as ‘interpretative schemata’ to single out an existing social condition or aspect of life, and define it as unjust and deserving of corrective action.”

¹² Typically, parties are classified in categories based on the intensity of their anti-EU sentiment. However, the criteria of such classifications are not unambiguously defined. An example is the distinction between “soft” and “hard” Euroskeptic parties (Treib 2014), and the uncertain boundaries of these two categories. This issue poses serious criticism in studies on the relationship between discontent and Euroskeptic voting across countries and over time.

¹³ The largely unpredicted and unexpected outcome of the referendum on Brexit is exemplary in this respect.

¹⁴ For example, Sweden and Finland constantly score above the EU average in terms of happiness and self-reported well-being (Inglehart et al. 2008) but, at the same time, they are also characterized by a much higher than average support for Euroskeptic parties (Dijkstra, Poelman, and Rodríguez-Pose 2020).

¹⁵ Data is sourced from the EB surveys reporting both the question on individual and political discontent. The four categories of individual discontent shown in Figure 1 correspond to the four alternative answers to the question on life satisfaction, from very satisfied (low discontent) to not at all satisfied (high discontent).

¹⁶ Germany and the UK are exceptions. For these countries information on respondents’ residence is available only at the NUTS1 level. The empirical analysis presented here includes Germany and the UK, and the regional characteristics are measured at the NUTS1 level. Our results are robust to the exclusion of these two countries. Evidence on this is reported in the online material, Table A3.

¹⁷ Alternative measures of interregional inequalities (namely, employment growth rate, variation of the share of employment in manufacturing, population growth rate) have been also tested in order to check the robustness of the results presented here. More details are provided in the online material, Table A7.

¹⁸ Alternative measures of intraregional inequalities (namely, the 80/20 disposable income quintile ratio and the poverty rate) have been also tested in order to check the robustness of the results presented here. More details are provided in the online material, Tables A8–A9.

¹⁹ More precisely the question wording is the following one: “During the last twelve months, would you say you had difficulties to pay your bills at the end of the month … ? Almost never/ From time to time/ Most of the time.”

²⁰ Equation[1] does not include country-level characteristics. This choice has two main motivations. First, individual discontent is assumed to be associated with a number of observable contextual factors, like economic development, demographic change, migration, etc. All these factors are measured at the regional level, thus enabling a much more precise estimate of the phenomena were the same variables measured at the country level. Simultaneously, there are other unobservable contextual factors that are likely to influence self-reported discontent. Among these factors, cultural traits, behavioral modes, etc. probably play a role, at both the country and regional level. For instance, Danes may tend to underrate their discontent compared with Italians. In turn, southern Italians may be less optimistic (other things being constant) than northern Italians. In order to account for these unobserved effects, a three-level hierarchical structure has been used, without adding any regressor at the country level.

²¹ This association was proved to be strongly significant in the case of the US (Cramer 2016), but empirical evidence on Europe provided contrasting results. While a positive relationship between rurality and Euroskeptic voting has been found by some studies (Essletzbichler, Disslbacher, and Moser 2018), its intensity (and even sign) is highly sensitive to the definition of Euroskepticism adopted (Dijkstra, Poelman, and Rodríguez-Pose 2020).

²² A correlation matrix of the contextual factors included in the analysis is reported in Table A2 in the online material.

²³ Empirical results on this are not reported in the article due to space limitations, but they are available from the authors upon request.

²⁴ Further evidence on this is reported in the online material. For more details, Tables A4–A6 replicate the main analysis of the article using different categorization of the dependent variable (four-level vs. binary coding) and statistical techniques (linear models vs. models for categorical variables).

²⁵ We are aware that this result departs from what is generally found in the literature on the geography of political discontent. This aspect is discussed in the comment to Table 3.

²⁶ See the section “Measuring Discontent” for further details on the question wording used in EB surveys. The number of observations is lower compared with the estimates of Table 2, since the question on the tendency to trust the EU is asked in a subsample of the survey studies listed in the online material (Table A1). As a robustness check, we rerun the estimates reported in Table 2 on the subsample used in Table 3, obtaining comparable results (see Table A12 in the online material).

²⁷ For a parallel approach, see Royuela and López-Bazo (2020).

²⁸ For a parallel approach, see Castells-Quintana Royuela, and Veneri (2020).