Does regional decline trigger discontent? Unravelling regional development in the Netherlands

ABSTRACT

A new geography of regional decline is signalled as a prominent source of discontent. This study considers regional divides in the Netherlands, and examines what trajectories can be detected in regional development and how these relate to recent regional discontent. Based on quantitative analyses of longitudinal data at the NUTS-3 level, we found a persistence of social and economic inequalities over the last decade(s). Our main result is that evidence for economic decline at the regional scale is limited, and unlikely to account for regional discontent in the Netherlands. Pathways of regional divergence and decline were mostly found in demographic change. Anti-establishment attitudes across Dutch regions were strongly associated with lower income, lower population change and ageing.

KEYWORDS:

• regional development
• geography of discontent
• regional inequality
• centre–periphery
• urban–rural divide
• peripheralisation

JEL:

• R1

1. INTRODUCTION

From the 1990s onward, a new phase of industrial transformation, state restructuring and shifting policy preferences rearranged the economic landscapes of subnational regions in the Global North (Horner et al., 2018; Martin et al., 2018b; Odendahl et al., 2019). Some regions bloom in prosperity while others dwindle and are earmarked as places that are ‘left behind’ or ‘don’t matter’ (Gordon, 2018; Rodríguez-Pose, 2018; Wuthnow, 2018; MacKinnon et al., 2021). The populist surge brought attention to these regional divides, which are signalled as prominent sources of discontent directed at a political establishment accused of ignoring the interests of rural and old industrial communities (Cramer, 2016; Hochschild, 2017; Gordon, 2018; Mamonova & Franquesa, 2019; Rodríguez-Pose, 2020).

The rise of the geography of discontent is predominantly explained by socio-economic decline (Dijkstra et al., 2020; Essletzbichler et al., 2018; Guilluy, 2019; Rodríguez-Pose, 2018). Accordingly, regions are left behind in industrial transformation and innovation (Dawley et al., 2010; De Ruyter et al., 2021; Rodríguez-Pose, 2018). Recent studies contested to what extent the fertile grounds for regional discontent are economically or actually culturally determined (Carolan, 2019; Iocco et al., 2020; Florida, 2021), if geographical cleavages are determined by a changing composition of people or changing context of places (Maxwell, 2019; Koeppen et al., 2021), and whether resentment is coming from rurality or peripherality (Harteveld et al., 2019; Ziblatt et al., 2020). To date we do not have a complete understanding of the sources of regional discontent and their recent socio-economic history. We argue that it is important to consider paths of regional inequalities over time (see also Odendahl et al., 2019; Carrascal-Incera et al., 2020; Evenhuis et al., 2021; Diemer et al., 2022), and to critically examine them for trends of decline and divergence with an in-depth understanding of geographical divides within a national context.

This paper takes the Netherlands as a case study. The motivation for investigating the Netherlands comes from the fact that it is a least-likely case, being a proportional democracy without territorial seats in parliament (Van der Meer et al., 2019), and being one of the smallest and most densely populated countries in Europe with a much smaller retraction of rural population (Eurostat, 2019a). Yet, similar regional trends to other European countries are observed. There are economic frontiers while others are lagging behind (Iammarino et al., 2018; Raspe & Van der Berge, 2017), divergent cosmopolitan–nationalist attitudes across urban and rural areas (Huijsmans et al., 2021), and disproportionally more supporters for right-wing populism on the periphery (Van Leeuwen et al., 2021; De Lange et al., 2022). Yet how these regional cleavages differ or complement each other, whether regional inequalities are widening over time, and if and how regional development fuels discontent requires further investigation.

This study exposes trajectories of socio-economic development across regions and over time in the Netherlands, and examines how these relates to recent regional discontent. In doing so, we explicitly examine inequality trajectories along the centre–periphery and urban–rural divides over the last decade(s). We formulate the following research questions:

• What trajectories (of decline and divergence) can be detected in regional development over the last decade(s) across regional divides?

• How are attitudes of regional discontent distributed across regional divides?

• How do regional inequalities relate to attitudes of regional discontent?

To address this we undertake two analyses. First, a descriptive analysis endeavours to provide a detailed and multifaceted description of regional development trajectories (also referred to as paths, patterns or trends) in the Netherlands, based on longitudinal data on NUTS-3 regions. We engage in two types of disadvantages of regional inequalities that might trigger discontent: economic disparities and social disparities. The former grasps the distribution of income, wealth and unemployment; the latter the access to public services and demographic change. Specifically we look for patterns of convergence and divergence, growth and decline. Subsequently we capture feelings of discontent through three different attitudes in survey data at NUTS-3, and describe how these are distributed across regions. Second, with a correlation analysis we explore how variables of economic disparities and social disparities relate to discontent. Before that, we will first ground our theoretical understanding of regional inequality and discontent, and specify our chosen methods and selected statistical datasets.

2. THEORETICAL UNDERSTANDING AND EVIDENCE

2.1. Understanding regional development

Evidence in the literature of regional development highlights trajectories of long-term economic divergence across regions within countries in the post-industrial era (Butkus et al., 2018; Iammarino et al., 2018; Blažek et al., 2019; Evenhuis et al., 2021; Diemer et al., 2022), including the Netherlands. Rural and old industrial heartlands in Europe suffered from relative – and in some cases, absolute – productivity downturn (Odendahl et al., 2019). We aim to understand economic disparities beyond productivity and sort out place-bound economic hardship for households and less for industries – even though the two are interdependent. In a study on right-wing populist support across the Netherlands, Harteveld et al. (2021) found that unemployment is particularly relevant in rural areas, while low incomes matter especially in urban areas. Further, there has been a rapid growth of house prices in large cities (Odendahl et al., 2019), and we also know that the people-based wealth gap is larger than the income gap in the Netherlands (Chancel et al., 2022). For those reasons, we also examine the development of regional wealth of households (financial assets minus liabilities), in addition to income and unemployment.

We go beyond economic context when considering regional cleavages, and include disparities in demography and quality of life in communities, typically brought forward in peripheralisation literature (Kühn, 2015). Scholars exposed an uneven distribution of essential public services across regions, in which rural areas suffer the burden (Lang, 2015; Bock, 2016; Dax & Fischer, 2018; Humer, 2018). Many sparsely populated regions are afflicted by population decline and ageing (Eurostat, 2019a), including the Dutch countryside. As a result public services retract such as public transport, health facilities and educational services (Wirth et al., 2016; Kühn et al., 2017). Ageing, however, particularly requires adequate access to public services (Scharf et al., 2016). Moreover, the loss of a village supermarket, for example, can be a source for grievances and for a decrease in social attachment in rural Netherlands (Gieling et al., 2019; Haartsen & Gieling, 2021). The pressures on communities’ standards of living are indicated in this study by population change, access to public services and ageing.

2.2. Understanding regional discontent

In most studies discontent is measured by populist voting (Rodríguez-Pose, 2018; De Ruyter et al., 2021; Van Leeuwen et al., 2021). Yet we argue here it is important also to use survey data. Of course data from elections results are, in most liberal democracies, very accessible, reliable and detailed. Plus every election serves a new batch of data. Yet election data are also limited. They lack a full understanding of what people are discontented about. Also, it is contested which political parties can be labelled as populist (Rooduijn & Pauwels, 2011; Norris, 2020), and whether their voters share the same resentment towards the same establishment (Rooduijn, 2018). Discontent can comprise different sets of feelings, which are not rooted in a single issue. Moreover, discontent may not always be expressed in a populist vote, or even through the ballot box, which therefore may overlook resentment in less populist regions. To better understand regional discontent, we should not only unravel multiple causes but also we should consider multiple forms of discontent. Therefore, this paper chooses to unfold discontent as feelings inhabitants might share regardless of their voting behaviour.

We pose three attitudes here to explore different forms of regional discontent in the Netherlands. First, regional discontent can exist of the perception of unfair redistribution of regional investments, and feeling neglected by the established government. In rural or peripheral areas people often claim that their community does not receive a fair share of public investments (Cramer, 2016; Huijsmans, 2022; Van Vulpen, 2022). Second, regional discontent can stem from dissatisfaction with the local environment in which people live. Wuthnow’s (2018) study of people in places that are left behind shows, among others, that a lack of jobs, population decline and brain drain affect a local community as a whole. Third, regional discontent can be rooted in anti-establishment attitudes and distrust towards a distant political power. Several studies conducted in different countries found significant lower levels of institutional and political trust in regions characterised by poor economic development (Ashwood, 2018; Hobolt, 2016; Lipps & Schraff, 2020; McKay et al., 2021; Mitsch et al., 2021). With that, we extend the set of attitudes for regional discontent in comparison with the recent research of De Lange et al. (2022).

What shapes the geography of discontent, as well as other political attitudes, can be distinguished into contextual and compositional effects (Maxwell, 2019; Koeppen et al., 2021). Contextual effects are place-based circumstances that might cause people to be discontented. For instance a higher unemployment rate, population decline or low access to public services. This is typically assumed in the literature of geography of discontent (e.g., De Ruyter et al., 2021; Essletzbichler et al., 2018; Rodríguez-Pose, 2018). In contrast, compositional effects examine the geographical clustering of people sharing the same individual characteristics or lifestyle, also known as geographical sorting (see, for instance, the ‘rural consciousness’ of Cramer, 2016). In this view sources of discontent are explained by the sociodemographic composition of an area. Both views are used to discuss our results.

2.3. The Netherlands: centre–periphery and urban–rural divide?

There are two unremitting geographical divides that mark a place-bound ‘us’ versus ‘them’: centre–periphery and urban–rural (De Souza, 2017). Both are powerful socio-spatial imaginations that are invoked in public and political debate. See, for instance, the farmers’ upheaval from the Dutch countryside (Van der Ploeg, 2020) and the protests against the extraction of natural gas in northern Netherlands (Van der Voort & Vanclay, 2015). As these two divides overlap, their borders can become blurry and are commonly mixed up. It should be explicitly noted here that centres and peripheries are first and foremost produced by applying this spatially oriented metaphor itself. The centre–periphery cleavage is still useful as an analytical approach to highlight the power–hierarchical relations or structures between spaces.

In this study we refer to the centre as the Randstad area, a polycentric metropolitan area (Zonneveld & Nadin, 2021). The Randstad comprises a cluster of the four largest cities in western Netherlands, and all (agricultural) land in between. These cities were also historically wealthy trade cities in the (pre-)modern era (Brand, 2021). In the post-war period the Randstad was first designed for decentralisation of industrial activity and for clustering suburbanisation; since the 1990s it has been planned as an economic powerhouse to compete at an international scale (Zonneveld & Nadin, 2021). The periphery, on the other hand, is generally considered as hinterland, close to the borders of Germany and Belgium. Dutch peripherality is characterised by a sparse population, stronger regional identities and more social cohesion.

Due to the relatively small size and dense population of the Netherlands, the urban–rural divide is different than in most other countries. According to European standards the Netherlands does not have any rural areas at a regional scale, only at a local scale (Eurostat, 2019a). Yet about 54% of land is used for agriculture, 34% of land consists of natural terrain and of inland and open water, and only 13% of land is used for buildings and traffic (CBS, 2020b). Rooted in a long agrarian history, citizens also hold strong beliefs about the Dutch countryside. On the one hand it is imagined as a rural idyll, a romanticised view of the countryside (Van Dam et al., 2002; Haartsen et al., 2000). On the other hand, rural areas can be challenged to shake off the territorial stigmas imprinted on them by outsiders, such as the former agrarian pauper colonies in north-eastern Netherlands (Meij et al., 2020).

2.4. Dutch regional typology

In this section we formulate a new regional typology to capture the centre–periphery and urban–rural divide for NUTS-3 regions in the Netherlands. First, to define the centre–periphery we work with a classification of Randstad and non-Randstad regions. This makes the following definitions:

• Centre: regions comprising the four largest cities plus all the regions within this ‘ring’.

• Periphery: all other regions, which are generally considered as national hinterland.

Second, we classify regions according to a new urban–rural typology. Eurostat (2019b) defined an urban–rural typology in NUTS-3 regions in Europe based on population density, which is extremely helpful for making international comparisons. Yet it is not particularly sensitive to applying it to the Dutch context only due to a high population density across the country. To make a more detailed and context-specific distinction between urbanity and rurality at a regional scale in the Netherlands, we choose to work with the absolute number of people living in urban areas. Taking the population size per degree of urbanity (very strong, strong, slightly, little, not), the following urban–rural typology is used:

• Large urban (LU) regions: more than 100,000 inhabitants living in very strong urban areas.

• Mid-urban (MU) regions: between 10,000 and 100,000 inhabitants living in very strong urban areas, or more than 100,000 inhabitants living in very urban areas.

• Rural (RU) regions: between 0 and 10,000 inhabitants living in very strong urban regions.

In our analysis we sorted LU, MU and RU regions in centre and periphery, making five regions (accordingly there were no rural regions in the centre). Appendix A in the supplemental data online shows an extra descriptive overview of the size of the population and urbanity level across the regions we composed, but also the population density and the share of population. Figure 1 maps our regional typology.

Figure 1. Map of regional typology of the Netherlands.

Note: Readers of the print article can view the figures in colour online at https://doi.org/10.1080/00343404.2023.2231031

Sources: Authors and CBS Statline.

Urbanity levels in this map are from 2017 as measured by CBS Staline.

3. DATA AND METHODS

3.1. Measuring regional development trajectories

All data used in this study were s obtained at the NUTS-3 level. Indeed, the level of scale at which the analysis is conducted both reveals and conceals things. The data at the NUTS-3 level is most detailed and complete for analysing regional development trajectories in relation to discontent. Most importantly, the regional scale makes it possible to explore correlations with attitude data since only few nationwide survey studies are representative at the regional scale, let alone at the local scale. In terms of development, the regional scale also allows us to go back further in time and rule out missing data for local areas. Moreover, the data are more valid as data variations due to municipal mergers, of which there have been many throughout the years in the Netherlands, have far less impact than on the local scale. Certainly, the NUTS-3 level does not show the within-region differences at the local scale. As a result, this study is limited in tracing disparities in local development. To compensate this, we included an overview of all social and economic variables at the municipality for 2019 (see Appendix D in the supplemental data online). These descriptive statistics at the municipal level show no large discrepancies with our main analysis at the regional scale. Also, we present an overview of income and wealth change at the municipality level in Appendix C online. For employment and public services the data were missing or incomplete.

The trajectories of regional development are examined with a descriptive analysis of region fixed variables. Economic disparities are examined by taking three indicators: mean standardised household income, median household wealth and unemployment rate. Social disparities are measured with three indicators: population change, distance to public services and the share of elderly population (65+ years). These six indicators are also used as independent variables in a correlation analysis examining for both longitudinal and cross-sectional effects (see also Martin et al., 2018a). The data used in this study consist mainly of CBS Statline, a large dataset of Statistics Netherlands.

The timeframes of our data are restricted. When possible, the data measure the regional trajectories from 1995 to 2019. However, the NUTS-3 level data of some variables were not available over the full period. Therefore, we took the longest period possible within that timeframe. With regards to multicollinearity, we consider the correlation between the independent variables robust. Variance inflation factor (VIF) values for regional inequality in 2019 are < 6.0 and for long-term regional development are < 2.0. Table 1 provides an overview with the details of all the indicators used in the analysis of long-term regional development, including the units of measurement, the exact periods and the data sources.

Table 1. Indicators for regional development trajectories.

	Indicators	Unit of measurement	Period
Economic disparities	Income	Mean standardised income private households (excluding students)	2011–19
	Wealth	Median wealth private households (excluding students)	2011–19
	Unemployment	Unemployment rate of workforce	2003–18
Social disparities	Population change	Number of in- or out-migration per 1000 inhabitants	1995–2019
	Accessibility public services	Average distance to facilities related to:	2008–19
		Health and well-being (general practitioner; hospital)
		Daily groceries (large supermarket)
		Catering industry (restaurant)
		Education (primary school; secondary school)
		Traffic and transport (main road; railway station)
	Ageing	Average share of inhabitants aged 65 and over	1995–2019

3.2. Examining regional discontent

This study conducts a correlation analysis to explore whether regional inequality and/or regional change affected feelings of regional discontent. Since our data samples are limited to regional averages at the NUTS-3 level, we choose a correlation analysis and test for the strength of linear associations between two variables, and not a regression analysis. This may limit statements about the multivariate relationships between variables, but it increases the reliability of our linear associations. Regional discontent is captured by dissatisfaction with regional investments, satisfaction with the living environment and institutional trust (Table 2). These dependent variables are based on existing statistics and secondary data analysis at the NUTS-3 level. The survey data about dissatisfaction with regional investments are from the SCoRE project, an international collaboration of universities that studies the subnational context and radical right support in Western Europe. This latter indicator, focused on the perceived insufficiency of regional investments by central government, is one of three items that together are constituted as regional resentment in the recent study of De Lange et al. (2022). In addition, we used survey statistics from two other indicators retrieved from CBS Statline: institutional trust and satisfaction with the living environment.

Table 2. Indicators for regional discontent.

Indicators	Unit of measurement	Period	Source
Dissatisfaction with regional investments	Percentage of private households that believe the government has done too little to improve one’s regional economy	2017	SCoRE
Satisfaction with the living environment	Percentage of private households very satisfied or satisfied with the current living environment	2018	CBS Statline
Institutional trust	Percentage of private households that trust the House of Representatives, the police and the judiciary	2019	CBS Statline

4. RESULTS

4.1. Economic disparities

In this section we highlight several economic development trajectories across regional divides in the Netherlands. Figure 2 demonstrates the development of standardised household income, median wealth per household and unemployment rate. The graphs on the left side show the averages per year of the regions classified according to the five regional typologies. On the right side we plotted the z-scores across most recent year (y-axes) and oldest year (x-axes), to show the spread of all 40 regions.

Figure 2. Economic development trajectories.

Source: CBS Statline.

Development of standardised household income, median wealth per household, and unemployment rate. The graphs on the left side show the averages per year of the regions classified according to the five regional typologies. The plots on the right side show the z-scores across most recent year (y-axes) and oldest year (x-axes), at the NUTS-3 level.

Between 2011 and 2019, there was an overall increase of standardised household income across regions in the Netherlands. There was no decline, merely growth at the regional scale. There was, however, a stable centre–periphery cleavage (Figure 1). In the centre regions the average income was significantly higher than the peripheral regions. Even though disproportionally large shares of both poor and rich people are located in the large cities (Buitelaar et al., 2016; Hoff et al., 2019), the average income was highest in both large urban and mid-urban regions in the centre. Mean income was lowest in rural regions on the periphery. Figure 2 shows that these income inequalities across centre–periphery divide stayed more or less the same. The z-scores along the diagonal also indicate no con- or divergence of regional incomes. At the municipality level the income change presents a similar pattern, one of a persistence of inequalities and on average a slightly smaller growth in the rural periphery (see Appendix C in the supplemental data online).

Looking at the regional development of median wealth per household, one can clearly see the economic shock that came from 2008’s financial crisis. From 2011 to 2013 there was an overall downfall in regional wealth, while in 2011 the median wealth of households was highest in rural regions on the periphery, it was also here where recovery from the economic shock was slower. The Randstad regions caught up with a more rapid recovery. In doing so, on average regions surpassed their urban–rural counterparts on the periphery. The rural regions lost their leading position to the mid-urban regions in the centre and only saw a small increase compared with 2011. Other regions, especially in the Randstad, significantly improved their wealth. When looking at the z-scores, one can see that the large urban regions are very much spread with three regions which remained at the bottom and three regions which had a relatively large increase in median wealth. At the municipality level the wealth change looks a bit different (see Appendix C in the supplemental data online). A substantial part of the municipalities on the periphery had a decline in wealth, though on average wealth increased.

Between 2003 and 2018 there was an almost unchanged path of the highest regional averages of unemployment rates in LU regions (Figure 2). The LU regions in both centre and periphery show compatible rates over time, making them the least well-off in terms of unemployment. The development of unemployment rate in rural regions somewhat changed over time. Before the 2008 financial crisis, RU regions had a relatively high unemployment rate that was comparable with the LU regions. From 2010 on, however, the unemployment rate became lower than of the LU regions. The MU regions were at all time better off than other types of regions, especially the MU regions in the centre. Overall, there was no clear divergence or convergence. Rather, we see an accordion pattern: in times of economic regression, at the peaks of national unemployment, regional divides were stretched. When unemployment rates were lower regional differences converged.

4.2. Social disparities

In this section we focus on the development trajectories of several social disparities between regions. Figure 3 presents the population change per 1000 inhabitants between 1995 and 2019, the distance (km) to public services between 2008 and 2019, and the shares of elderly population between 1995 and 2019. Looking at the relative population change across regional divides, one can see a divergence from 2005 on. Between 1995 and 2005, trends across regional divides were quite similar. Yet, from 2005 on there were periods of population decline in the RU regions while other regions continuously experienced population growth. Since 2016 there was a resurgence in population growth in rural regions. The z-scores present that especially the LU regions in the centre are above the diagonal.

Figure 3. Social development trajectories.

Source: CBS Statline.

Development of the population change per 1000 inhabitants, distance to public services in km, and shares of elderly population. The graphs on the left side show the averages per year of the regions classified according to the five regional typologies. The plots on the right side show the z-scores across most recent year (y-axes) and oldest year (x-axes), at the NUTS-3 level.

The accessibility of public services is characterised by a sharp divide between centre and periphery. Public services are much closer to people living in the centre regions, compared with those on the periphery. Especially inhabitants of the rural regions have to travel further to reach public services. Over the years there was an overall growth in distance, be it very small at the regional scale. Disparities at the local scale are likely to be larger. In MU and RU regions the distance increased a bit more than in the LU regions. A very small though undeniably present divergence between regions can be seen across the urban–rural cleavage.

There was consistent divergence across the regional divides in ageing. Since 1995 the distribution of the share of elderly population across regions became more uneven. There was a general growth in elderly population, which over time developed into an urban–rural divide. In 1995 the regional inequality with regards to the elderly population was much smaller compared with the situation in 2019. From 2003 on, the smallest share of people aged 65 or over can be found in the LU regions. Especially the rural regions increasingly lodged more elderly who enjoy their pension in a quiet environment.

4.3. Regional ranking

Figure 4 shows scatter plots of the relative regional change of economic and social disparities. It represents the change in mean ranking position of the three variables, and compares the oldest year with the most recent year. Regions were ranked from most positive score (1) to most negative score (40). The dots below the diagonal indicate regions that improved their ranking, and above the diagonal line which regions’ ranking depreciated.

Figure 4. Old and new ranking of economic and social marginalisation disparities.

Source: CBS Statline.

Ranking is calculated as average ranking of the three variables of economic disparities (standardised household income, median wealth per household, and unemployment rate) and of social disparities (population change per 1000 inhabitants, distance to public services in km, and shares of elderly population).

Figure 4 shows that regions are scattered around the diagonal, especially in the lower rankings. This indicates that there have not been any substantial changes in the economic mobility across regions (see Appendix B in the supplemental data online for a more detailed overview of rankings per region). Regions that rank lowest in 2019 were already in a more or less low ranking previously. No distinct patterns of regional divides in terms of economic disparities were observed in our rankings. The scatter plot reveals a greater dispersion in the ranking of social disparities, suggesting that there were larger relative changes in this aspect. Particularly, the LU regions in the Randstad demonstrated significant improvement in their position, while rankings of RU and MU regions on the periphery mostly depreciated.

4.4. Regional discontent

Our descriptive findings on regional discontent reveal consistent disparities between central and peripheral regions in the Netherlands. Regional discontent is measured here by respondents’ agreement with propositions about perceived insufficiency of regional investments, satisfaction with living environment and institutional trust (Figure 5). Among the three indicators, dissatisfaction with regional investments exhibits the widest regional variation. People living in regions on the periphery generally feel more dissatisfied with the central government’s investments in their region than those in the centre. This accounts especially for rural regions, although there is quite some variation between them. In addition, Figure 5 shows that inhabitants of peripheral regions also have less institutional trust than those in centre regions. The regional divide in trust corresponds with international studies that found a similar pattern of higher levels of institutional distrust in economically disadvantaged regions (Lipps & Schraff, 2020; McKay et al., 2021).

Figure 5. Attitudes of discontent.

Sources: CBS Statline and SCoRE.

Calculated as the share of population that in 2019 (somewhat to strongly) trusted state institutions – taking the average of the House of Representatives, the police, and judges (institutional trust), the share of population that in 2017 (somewhat to strongly) agreed that the government has done too little to improve one’s regional economy (dissatisfaction regional investments), and the share of private households that in 2018 was (strongly) satisfied with their living environment (satisfaction living environment), at the NUTS-3 level.

Compared with other regions, inhabitants of peripheral regions are, however, generally more satisfied with their living environment. Even though the differences between the four regional categories are small, it is an interesting result. The fact that residents of large urban regions in the centre are less likely to be satisfied with their living environment compared with regions on the periphery could come from the disadvantages of urban overdevelopment, such as a lack of affordable housing, high levels of criminal offenses, high levels of feeling unsafe, less green spaces and high levels of air pollution (CBS, 2020a). This resonates with many other EU member states in which the subjective well-being is relatively high in rural areas than it is in urban areas (De Dominicis et al., 2020). Even though regional differences for most accounts are not large, these descriptive results suggest that regional discontent is essentially political discontent strongly present on the periphery.

4.5. Correlation results

In order to investigate the causes of regional discontent in the Netherlands, we conducted an exploratory analysis of correlations for both recent regional inequality (most recent year) and regional change (the value of most recent year minus oldest year). In this type of analysis, coefficients are not controlled for other variables. Indeed a more extensive analysis with a larger number of observations would provide a more comprehensive understanding of the effects on discontent, our data only permit us to explore potential associations with discontent. Knowing that some regional inequalities remained relatively stable over time, such as income, we still included them in the analysis. See the z-scores in Figures 3 and 4: if values were clustered along the diagonal then are not much changes between regions. For an interpretation of the strength of the correlation coefficients, we follow Hinkle et al. (2003): negligible (0.0 to 0.3); low or weak (0.3 to 0.5); moderate or strong (0.5 to 0.7); high or very strong (0.7 to 0.9); very high or extremely strong (0.9 to 1.0), likewise for negative correlations. Table 3 presents the correlation coefficients and probabilities for each of the indicators with each of the three indicators for regional discontent.

Table 3. Correlations by regional inequality and regional change.

Correlation probability	Dissatisfaction with regional investments		Satisfaction with living environment		Institutional trust
	Regional inequality	Regional change	Regional inequality	Regional change	Regional inequality	Regional change
	2019	2011–19	2019	2011–19	2019	2011–19
Income	−0.619** 0.000	−0.587** 0.000	0.047 0.772	0.118 0.468	0.620** 0.000	0.567** 0.000
	2019	2011–19	2019	2011–19	2019	2011–19
Wealth	−0.295 0.064	−0.455* 0.003	0.580** 0.000	−0.072 0.660	0.405 0.010	0.457* 0.003
	2018	2003–18	2018	2003–18	2018	2003–18
Unemployment	0.381 0.015	−0.351 0.027	−0.445* 0.004	−0.067 0.682	−0.303 0.057	0.379 0.016
	2019	1995–2019	2019	1995–2019	2019	1995–2019
Population change	−0.699** 0.000	−0.209 0.195	−0.226 0.161	−0.207 0.199	0.333 0.036	0.407* 0.009
	2019	2008–19	2019	2008–19	2019	2008–19
Accessibility public services	0.467* 0.002	0.490* 0.001	0.286 0.074	−0.053 0.745	−0.196 0.226	−0.256 0.110
	2019	1995–2019	2019	1995–2019	2019	1995–2019
Ageing	0.529** 0.000	0.324 0.041	0.476* 0.002	0.471* 0.002	−0.339 0.032	−0.359 0.023

Note: N = 40. **p < 0.001; *p < 0.01.

The correlations between income and dissatisfaction with regional investments demonstrate significant strong positive associations, and income shows negative associations with institutional trust. These results reveal that people living in regions with relatively low average household incomes are more likely to be discontented with regional redistribution policies by government and to be distrusting towards institutions. The findings for wealth show significant correlations between regional change and dissatisfaction with regional investments and institutional trust, but the associations were weak. Moreover, we found a significant and strong positive association between wealth inequality and satisfaction with living environment. This suggests that inhabitants in more wealthy regions are generally more satisfied with their residential area. For unemployment rates we only found one significant yet weak correlation, which indicates that higher unemployment levels have a negative relationship with residents’ satisfaction with their living environment.

The strongest coefficient in our correlation analysis was found between population change in 2019 and the perceived insufficiency of regional investments. This suggests that at the regional scale the best predictor for belief in government failing to sufficiently invest in one’s regional economy is population change per 1000 inhabitants, followed by mean household income. Accessibility of public services has no strong correlations with attitudes: it correlates mostly with dissatisfaction with regional investments, indicating that in places with a longer distance to public services people are a bit more likely to be discontented with regional redistribution policies. Lastly, Table 3 shows that ageing in 2019 strongly correlates with ideas about insufficient government efforts for regional development. The larger the share of elderly population, the higher the discontent. Also, highly ageing regions have a weak association with being happier about living environment and a weak association with being more distrusting towards institutions.

5. DISCUSSION

In this section, we delve deeper into how the findings of our study on economic disparities and social disparities align with the existing literature on the geography of discontent. First, in correspondence with several studies on discontent (Van Gent et al., 2014; Dijkstra et al., 2020; De Lange et al., 2022), our results reveal that regional income inequalities in the Netherlands are strongly associated with both feelings of insufficient government support and institutional distrust. In contrast to earlier studies conducted at a local level (Harteveld et al., 2021; Huijsmans, 2023), our study at a regional scale detected no strong link between unemployment and discontent. The findings from our descriptive analysis provide insights into the economic development of regions, which among others indicate that there were no relative changes in income mobility of regions characterised by lower average income. From a contextual perspective, lower income regions, mostly situated in the Dutch periphery, correspond to the notion of being ‘trapped’ in the lower tiers of economic development within the country (Diemer et al., 2022), and less so to the notion of ‘economic decline’. However, it is essential to also consider the effects of composition (see for example Maxwell, 2019). Several researches of the Netherlands showed that a higher concentration of lower incomes also signifies the geographical clustering of a lower socioeconomic class that tends to harbor greater grievances against a political establishment (Van Gent et al., 2014; Harteveld et al., 2021; Noordzij et al., 2021). Similarly, the strong link between high concentrations of elderly population and regional discontent in our study aligns with research suggesting that elderly tend to harbor more negative sentiments towards the established political order (Dijkstra et al., 2020; Harteveld et al., 2021). An economic development trajectory characterised by lower income job opportunities and a lack of relative upward mobility, perpetuates the concentration of low-income groups in these regions. The presence of these compositional effects prompt further research into spatial sorting and its relationship to place-based discontent in the Netherlands.

Rural communities in the Netherlands, particularly on the periphery, face challenges from a vicious cycle of depopulation and decreasing public services (see also Gieling et al., 2019; Ubels et al., 2019b). In line with recent studies of the Netherlands (Harteveld et al., 2021; Van Leeuwen et al., 2021), we found that the distance to public services is a weak predictor for feelings of discontent. Possibly rural residents in the Netherlands perceive the centralisation of public facilities as a minor inconvenience and shrug their shoulders for a slightly longer travel, after all the Dutch situation is much less pronounced compared with countries such as Spain or Italy (Eurostat, 2019a). Yet, qualitative studies illustrated a sense of loss and decreased social attachment in peripheral areas in the Netherlands coping with retraction of public services such as a supermarket (Gieling et al., 2019; Haartsen & Gieling, 2021). Possibly there is, what we term here, a rural service paradox. Our study indicates that population change is strongly associated with people’s dissatisfaction with regional investments. Considering that population change emerges as the primary catalyst for the decline in local public services, it is plausible that the strong correlation between population change and dissatisfaction with government investments comes from a worsening regional context. Many rural residents engage in social innovation to sustain public services (Bock, 2016; Ubels et al., 2019a), for instance, through initiating a cooperative that gives the local supermarket a new lease life. Local initiatives help alleviate the drawbacks of public services, but the lack of government support simultaneously evoke feelings of being left behind (see also Bolet, 2021). It presents a paradox because, on one hand, the local initiatives have a positive impact on the local quality of life in terms of service provision and social attachment, but, on the other, the absence of government support undermines the overall relationship between rural citizens and the government, potentially evoking negative sentiments of being left behind or neglected. Further exploration of this dynamic is necessary to comprehensively understand the relationship between population change, public service accessibility and discontent.

6. CONCLUSIONS

The geography discontent causes much concern, and numerous scholars are puzzling with pinpointing the underlying causes. Frequently, discontent is associated with long-term economic decline in regions that are left behind in industrial transformation. For instance, in parts of peripheral France, peripheral England, southern Italy and eastern Germany the populist surge is (partly) explained by deindustrialisation and employment decline (Essletzbichler et al., 2018; Diemer et al., 2022; Greve et al., 2022; Rodríguez-Pose, 2020). These are, however, more likely cases for finding regional discontent, since these large countries are heavily industrialised and have more severe economic inequality between regions. From this least-likely case study we conclude that in the Netherlands there has not really been any economic divergence or decline across households at the regional level. Our findings show that socio-economic distribution in the Netherlands is regionally unbalanced and mostly to the detriment of the rural periphery. However, we did not detect strong decline or divergence in recent economic development for households across regions. There were, however, pathways of decline and divergence in demographic change. The Netherlands exemplifies that recent changes in economic development are not likely to explain feelings of political discontent in all cases.

Building on studies about uneven regional development in a post-industrial era (e.g., Butkus et al., 2018; Iammarino et al., 2018; Blažek et al., 2019; Evenhuis et al., 2021; Diemer et al., 2022), we unravelled several other regional development trajectories within the Dutch context. Even though average household income in the rural periphery remained lowest, there was also structural income growth for all regions in the Netherlands. Moreover, large urban regions were afflicted by lower median wealth and higher employment peaks. Starting from 1995 a nationwide trend of ageing emerged, leading to a large increase of elderly population in the rural regions on the periphery. Additionally, these same regions experienced a period of population decline between 2010 and 2014.

This study showed that discontent on the periphery was more strongly expressed in anti-establishment attitudes, such as the perception of one’s region not receiving their fair share of government investments and institutional distrust. These differences correspond with international findings on higher shares of political distrust in economic disadvantaged and rural areas (Hobolt, 2016; Lipps & Schraff, 2020; McKay et al., 2021; Ashwood, 2018). The correlation analysis showed that regional discontent, in terms of anti-establishment attitudes, is best predicted by household income inequality, population change and ageing. This study indicated that regional discontent is likely to be explained at a regional scale by consistent lower average income, social challenges from population decline, and larger shares of elderly population. Our results also show that people in the less urban peripheral regions are generally happier with the place they live in than people in other parts of the country, which resonates with findings of higher subjective well-being in the countryside of Western Europe (De Dominicis et al., 2020). Dissatisfaction with residential area, strongly present in large urban regions in the Dutch Randstad, can be considered as a specific form of discontent from urban overdevelopment (see also Florida, 2021).

This study yields several new findings, in addition to the literature on the geography of discontent in the Netherlands, and more specifically to the work of De Lange et al. (2022) on regional resentment in the Netherlands. First, by providing a detailed overview of inequality trajectories of the explanatory variables, this study helped to create a more complete understanding of recent regional developments that are associated with regional discontent. Second, our findings showed that political distrust is stronger on the periphery – especially in the rural periphery – but that the differences are substantially smaller than for dissatisfaction with regional investments. This suggests that place-based sentiments towards politics (such as perceived insufficiency of government investments in one’s region) tend to be more geographically divided than non-place-based discontent towards politics (such as political distrust). Third, this study contributed by revealing that people in the peripheral regions tend to be more satisfied with their living environment, and so it is likely that many of these residents are very much attached to their place of residence. Fourth, we confirmed that also at a regional scale, and not only the local scale as analysed in De Lange et al. (2022), political discontent cuts somewhat more strongly along the centre–periphery than the urban–rural divide in the Netherlands.

This least likely study thus confirms the persistence of interregional inequalities and the presence of regional discontent in the peripheral Netherlands. Yet we also need to state the limits. First, since we did not conduct a multivariate regression analysis, there was no control for any indicators that might better explain the strong associations we found, for instance, education levels. Second, since we work with only 40 NUTS-3 level regions, our approach unfortunately does not allow us to work out a more detailed correlation analyses and compare different predictors across the regional cleavages. It would be very relevant also to study development within regions rather than between regions. Yet, our explorative analysis of municipalities (see Appendix C in the supplemental data online) showed that for household income it is unlikely that it would distil different patterns of divergence and decline across geographical divides, but for wealth it might. Third, it requires longitudinal panel survey data to find out how specific conditions and compositions develop parallel to attitudes over time.

ACKNOWLEDGEMENTS

We are very grateful to two anonymous reviewers for their encouragement and critical support throughout the reviewing process. In addition, we thank Eelco Harteveld, Sarah de Lange and Twan Huijsmans for providing data and sharing valuable insights.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.