4. Results and discussion

As shown in the map presented in Figure 1, the distribution of equivalised household income at municipal level in the Nordic Region was relatively unbalanced in 2022. Two spatial patterns stand out:

Figure 1: Mean equivalised household income in Nordic municipalities, 2022
(EUR, PPP)
Source: Own estimates based on NSI data.

The patterns observed in Figure 1 reflect important differences in local living standards. However, average equivalised household income values may conceal significant socio-economic disparities within municipalities. The map shown in Figure 2 captures those differences through the Gini coefficient of income inequality in 2022. That index reflects the average difference between all possible pairs of incomes in the distribution, expressed as a proportion of total income.

The spatial distribution of the Gini coefficients suggests that income inequalities tend to be higher in urban areas. Leaving aside some outliers observed in 2022 in rural Finland and Sweden, most of the municipalities with higher Gini coefficients are located in the largest urban agglomerations. The highest Gini coefficients can be found in municipalities such as Frederiksberg, Gentofte, Lyngby-Taarbæk, Hørsholm, Rudersdal, Sottunga, Danderyd, Lidingö and Smedjebacken.

Figure 2: Gini coefficients calculated based on equivalised household income deciles in Nordic municipalities, 2022 (EUR, PPP)
Source: Own estimates based on NSI data.

Figure 3 shows the development of total inequalities according to a set of indices calculated on the basis of equivalised municipal household income by decile for the entire Nordic Region. Those indices exhibit various properties, which are detailed in Annex 2 of this report. Regardless of their specific features, the harmonic evolution of all the inequality indices presented in Figure 3 confirms that the gap between income groups steadily increased in the Nordic Region as a whole during the 2005-2022 period.

Growing inequality in household income is generally linked to several coinciding factors. The literature indicates that the variability in the composition of individual incomes and the increasing weight of capital shares of income towards the top of the income distribution are broadly associated with higher personal income inequality. The Nordic Dual Income Taxation reforms in the early 1990s, which introduced lower progressivity of capital income taxation, are generally considered the triggering factor for that process (Iacono and Palagi, 2020).

Figure 3: Inequality indices calculated based on equivalised household income deciles (Nordic average)
Source: Own calculations based on household income data provided by NSIs.

Figure 4 shows development of the Gini coefficients over recent years in the Nordic countries and self-governing territories and compares those trends with the EU-27 (2020) average. The figure confirms that income inequalities increased in most areas over the 2005 to 2022 period. The largest increase in the Gini coefficient can be observed in Denmark with a 4-point rise. The most positive development is found in Iceland with a 1-point decrease in the Gini coefficient during that period.

While the Gini coefficients in all the Nordic countries and regions are still below the average value observed for the EU-27 (2020) as a whole, the Nordic values are converging towards the EU level. Moreover, Nordic and EU inequalities seem to have been moving in opposite directions in recent years. While inequalities at the EU level declined over the 2005-2022 period, the Gini coefficients based on household income increased steadily in most of the Nordic countries and self-governing territories during those same years.

Another interesting pattern shown by development of the Gini indices plotted in Figure 4 is that, despite greater fluctuations in the less populated regions (Åland, Faroe Islands, Greenland), income inequalities seem to be growing faster in the largest Nordic economies, particularly in Denmark and Sweden – and to some extent also in Finland and Norway – than in the smaller ones (Greenland, Faroe Islands, Iceland).

Trends observed during that period seem to reflect partially overlapping processes. In the case of Sweden, Denmark and Norway, the development of the Gini coefficients presented in Figure 4 suggests a continuation of the long-term trajectory towards increasing income inequalities that began in the 1990s (Pareliussen et al., 2018). In the case of Finland and Iceland, that can mainly be attributed to the economic downturn of the early 2010s (Olafsson, 2011).

Figure 4: Gini coefficients calculated based on equivalised household income deciles (Nordic average)
Source: Own calculations based on household income data provided by NSIs. Data for EU27 retrieved from Eurostat.

An aspect that previous research has not sufficiently addressed is the role of the spatial dimension of income inequality in explaining the widening socio-economic gap observed in the Nordic Region. Figure 5 sheds light on that aspect by showing the absolute change in Gini coefficients based on household income within each municipality

With the exceptions of the Faroe Islands and Iceland, where the territorial divisions correspond to regions and Statistical Output Areas, respectively.

in the Nordic Region between years 2012 and 2022. For better visualisation of results, statistical outliers have been replaced by the maximum or minimum values within 1.5 times the interquartile range.

Figure 5: Variation in Gini coefficients calculated based on equivalised household income deciles in Nordic municipalities (2005-2022)
Source: Own calculations based on household income data provided by NSIs. Data for Iceland refers to Statistical Output Areas.

In general, income inequality increased in most municipalities in Denmark and Sweden between 2005 and 2022. In Sweden, the largest increase took place in Jämtland County in the North, and Kalmar, Växjö, Jönköping, Älvsborg, and Halland counties in the South. In Denmark, the greatest increases in municipal Gini coefficients were seen in the northernmost and southernmost municipalities of Jutland, as well as in the more urban municipalities in Zealand and Fyn. In Norway, most local areas experienced a decrease in Gini coefficients for equivalised household income. However, income inequality grew in several municipalities in the Nordland region. In Finland, the situation is more varied, with some municipalities experiencing widening inequality gaps, while others have seen a decline. Nonetheless, in most cases, the variations in Gini coefficients in Finnish municipalities were small compared to other Nordic countries. Income inequality decreased in the Faroe Is-lands and most Statistical Output Areas in Iceland, except in central Reykjavik. Inequality also in-creased in most major urban agglomerations in the Nordic Region, but not in the largest Finnish cities.

Figure 6 provides a comparison of the distribution of household income at the municipal level for selected geographies and years. The charts display different density plots for the distribution of Gini coefficients, calculated based on equivalised household income deciles in Nordic municipalities. The plots also include the mean (vertical dotted lines) and marginal rug (tick marks placed along the bottom margins).

The vertical dotted lines in Figure 6 indicate the average Gini coefficient on equivalised household income for all the sampled municipalities. These values are similar to those shown in Figure 4 above. However, the match is not perfect, as coefficients in Figure 4 were calculated on the distribution of household income by decile at the national/regional level, while the vertical lines in Figure 6 show the mean value for Gini coefficients calculated at municipal level. In general, the mean values suggest that the distribution of household income has become more dispersed over time, particularly after 2012.

The shape and spread of the shaded areas in Figure 6 reveal several important characteristics of the data distribution. The peaks indicate the most common values in the data. A higher peak signifies a higher density of the Gini coefficients around similar values. Wider plots imply more spread-out values of municipal Gini coefficients on household income, while long tails suggest more skewed distributions of the coefficients between municipalities. Multiple peaks on the density plots indicate multiple modes in the distribution of municipal Gini coefficients.

Comparing the data for various years within each region, the plots suggest more spread distributions of the municipal Gini coefficients on income over time, particularly in Norway, Iceland, Finland, and Sweden, as compared to Finland. Note that a growing dispersion of income inequality levels across municipalities does not necessarily imply greater overall levels of income inequality, as an increase in the num-ber of municipalities where income inequality has grown can be balanced by a similar number of municipalities where income inequalities have declined.

Figure 7 provides a different perspective on development of the spatial dimension of socio-economic inequalities. The lines in this chart reflect the development of spatial inequalities measured through mean equivalised household income values at the municipal level, instead of the more nuanced income distribution by decile group used thus far.

Figure 7: Gini coefficients calculated based on mean equivalised household income at municipal level
Source: Own calculations based on household income data provided by NSIs.

The use of mean equivalised household income values at the municipal level instead of average income values by decile explains the smaller absolute Gini coefficients in Figure 7, compared to those in Figures 3 to 6. Nonetheless, the trajectories of income inequalities shown in the three plots are consistent with one another overall. As income gaps between social groups widen, income differentials between municipalities become larger and vice versa.

The comparison of the trends in Figures 4 and 7 allows us to understand how social and territorial inequalities are co-evolving. The Gini coefficients calculated on the basis of average equivalised municipal income (Figure 7) are increasing in Denmark, Sweden, Iceland and Åland (as are social inequalities; see Figures 3 and 4). However, the social and spatial gaps seem to be developing in the opposite direction in Norway, the Faroe Islands and Finland. In those areas, the Gini coefficients calculated on the basis of average household income at the municipal level (Figure 7) appear to be more stable – like in the Faroe Islands – or even declining over time – like in Finland and Norway – while social inequalities are still growing (Figures 3 and 4).

Figure 8 provides yet another perspective on the co-evolution of income inequalities between municipalities. The charts presented in this figure show the development of Gini coefficients calculated on the basis of income distributions within the same income levels across municipalities in Denmark, Norway and Sweden. Each line on the plot reflects the development of Gini coefficients calculated on the basis of mean equivalised income of households within each income decile. The variability reflected by the lines thus indicates the diversity of income levels among households in the same income deciles but located in different municipalities within each country. That enables a better understanding of the development of spatial inequalities within the same income groups.

Figure 8: Gini coefficients calculated based on the equivalised income of households within the same income deciles across municipalities in the Nordic Region
Source: Own calculations based on household income data provided by NSIs.
D1 = poorest households, D10 = richest households

The data required to produce the plots in Figure 8 are only available for Denmark, Norway and Sweden. Moreover, since the data generation process differs from country to country, the absolute Gini coefficients are not directly comparable. However, the charts in Figure 8 provide interesting insights into the development of socio-economic inequalities between households in different income deciles.

In general, income differentials between households living in different areas tend to be greater for the deciles at both ends of the income distribution, namely the lowest-income households (D1) and, particularly, the highest-income households (D10). That indicates that the variability of average household income at the municipal level tends to be substantially greater for households in extreme income groups than for other categories.

A growing Gini coefficient for the upper income decile in Denmark and Sweden suggests that the spatial gap – i.e. the difference in average income levels for households in this specific income decile – is widening between municipalities in both countries. That pattern contrasts with the development of the Gini coefficients calculated on the basis of the average equivalised household income by municipality for other income groups within both countries, as well as with the development of income differentials between municipalities in Norway. Here, the spatial gap in household income seems to be declining for all income categories.

Looking at the relative trajectories of the inequality lines between different income groups in each country, it can be observed that in some areas and periods – e.g. in Sweden between 2011 and 2022 – the spatial divide increased for the highest income decile (D10), while it converged or stabilised for the remaining groups, particularly after 2018. The same logic holds for Denmark, where the gap within the top-earners category seems to be steadily increasing between municipalities. As mentioned, the situation in Norway differs, as the average levels of municipal household income seem to be converging over time, regardless of the income category.

To delve into the dynamics of changes in household income, it is necessary to examine convergence and catching-up processes. Economic convergence reflects a process where per-capita incomes of poorer economies grow at faster rates than those of richer economies. That logic led to the so-called convergence hypothesis, which has been the subject of heated debates among different schools of thought in economic and regional science since the classical contributions on economic growth and development (Myrdal, 1957; Solow, 1956).

Galor (1996) describes three main convergence scenarios: (1) an absolute convergence hypothesis, according to which national or regional economies tend to converge in the long run, independently of their initial conditions; (2) a conditional convergence hypothesis, according to which only economies that are similar in their structural characteristics – e.g. in terms of specialisations, technologies, rates of population growth, governance etc. – converge in the long run, independently of their initial conditions; and (3) a club convergence hypothesis – polarisation, persistent poverty and clustering – according to which only economies that are similar in their structural characteristics converge in the long run, provided that their initial conditions are also similar.

Each convergence hypothesis is supported by a specific metric to validate it. The first two notions rely on the classic metrics of beta (1) and sigma (2) convergences, respectively. The club convergence (3) hypothesis has its own method to evaluate convergence processes, focusing on clustered trajectories of economic agents.

Other metrics used in regional economics to assess convergence patterns at the territorial level include gamma convergence, which examines changes in country rankings with respect to a particular outcome or policy objective, and delta convergence, which analyses countries’ distance from an exemplary model or group of countries. (see Heichel et al., 2005 for a review of economic convergence metrics).

In the remainder of this section, we will apply those metrics to investigate the development of average municipal household income across the Nordic Region.

The beta-convergence hypothesis was originally formulated by neoclassical growth theory (Solow, 1956). The hypothesis is verified when poor economies tend to grow faster than rich ones, implying that laggards catch up to leaders. When calculated on the basis of average municipal household income, the beta-convergence metric can be used to assess if households in municipalities with a lower level of income are catching up with the leading municipalities in this indicator. Table 1 shows the results of the unconditional beta-convergence statistic calculated on the basis of the equivalised mean household income by Nordic municipality for the 2005 to 2022 period.

Given that there is global beta convergence , it is possible to calculate the speed of convergenceand the so-called half-life indicator . The latter represents the time required to reduce disparities by half (Allington and McCombie, 2007). According to that metric, since the beta coefficient is quite small (-0.03), it would take approximately 353 years to halve the observed territorial disparities in average household income at the municipal level in the Nordic Region.

The beta statistic thus indicates that average income levels in Nordic municipalities tend to converge over time. However, that convergence occurs at such a slow pace that its cumulative effects are imperceptible even in the long run. The beta statistic can be represented by the slope of the regression line plotted in Figure 9. Despite the negative beta coefficient, the points appear to be concentrated in the central region of the chart, suggesting that the regression could be significantly influenced by municipalities with extreme values and statistical outliers.

The sigma-convergence hypothesis is defined as a decrease in the variation of economic outcomes, such as household income. Sigma convergence occurs when the dispersion of the variable , calculated as the standard deviation or coefficient of variation of regional economic output or income, lessens over time. Here, we estimate based on the coefficient of variation of the mean equivalised household income at the municipal level for each year in the time series. Figure 10 shows the resulting trend and regression line for the measure of income.

The positive slope of the trend regression in Figure 10 suggests that there is sigma divergence in our data sample. That result seems to contradict the one obtained above for beta convergence. If the regression line in the beta-convergence analysis was free from specification issues, the outcome would imply that, even if poorer municipalities experienced stronger income growth than richer ones (beta convergence), the average household income at the municipal level could still be gradually dispersing (sigma divergence).

Consistent with the theoretical properties of both metrics, our data show that while beta convergence is a necessary condition for sigma convergence, it is not a sufficient condition (Young et al., 2008). That behaviour could be explained by divergent income growth trajectories among households in different income categories, as observed in Figure 6. That necessitates the use of more flexible metrics, such as the club convergence analysis introduced below.

In addition to global measures of convergence, an increasing number of studies delve into the co-evolution of individual spatial trajectories through the club con­vergence analysis. That approach was originally introduced by Baumol (1986) to describe the relative evolution of a subset of national economies and was more recently streamlined by Phillips and Sul (2007). Essentially, the analysis is based on a time-varying factor model that allows for individual and transitional hetero­gen­eity to identify clustered trajectories known as “convergence clubs” (Phillips and Sul, 2009).

The club convergence method has several advantages over traditional metrics. First and foremost, it allows for individual heterogeneity over time, implying that different transitional paths for different groups are possible. Additionally, unlike other approaches where economies are grouped a priori, the club convergence method enables the endogenous, data-driven determination of convergence clusters. Moreover, it does not impose any particular assumption concerning trend stationarity or stochastic non-stationarity of the time series, as it is robust with respect to both properties (Sichera and Pizzuto, 2019).

Here, we have applied the club convergence method to explore the income development trajectories of households in the Nordic Region. Figure 11 shows the results of applying that method to the mean household income dataset for all Nordic municipalities

With the exceptions of the Faroe Islands and Iceland, where the territorial divisions correspond to regions and Statistical Output Areas, respectively.

from 2012 to 2022. The plot describes the relative transition of each municipality against the benchmark of a full cross-sectional average. Each line in the chart represents the time profile of transition for one group of individual municipalities with a similar development, relative to the average behaviour.

The club convergence algorithm identified four convergence clubs. Club 1 groups municipalities that are diverging from above in terms of average household income. In other words, in those municipalities, average household income has increased much faster than the Nordic average. Club 2 consists of municipalities where mean equivalised household income at the municipal level picked up relative to the Nordic average after 2019, but not before. Club 3 is composed of municipalities that performed systematically better than the baseline behaviour for the whole period, but where average household income levels are increasing at a slower rate compared to municipalities in Clubs 1 and 2. Finally, Club 4 clusters municipalities that performed worse than the other clubs during the entire 2012 to 2022 period, particularly since 2016. Table 2 shows how many municipalities are included in each group, alongside a number of descriptive statistics for each cluster.

The statistical significance of convergence clubs is determined by examining the t-values of individual units. Units with t-values greater than -1.65 are part of the core group, indicating within-group convergence.

The map in Figure 12 illustrates the convergence clubs cartographically. It highlights the fact that the greatest variation in club membership is found in Norway. Generally, national economic drive and the degree of urbanisation appear to be the key factors influencing club performance in terms of equivalised mean household income at the municipal level.

Figure 12: Convergence clubs for mean household income in Nordic municipalities (2012-2022)
Source: Own calculations based on NSI data.

Club 1 is composed of nine small municipalities located within or near the largest metropolitan areas in Denmark, Norway and Sweden. The club consists of the municipalities of Gentofte and Rudersdal in Denmark; Nordre Follo, Frogn, Nesodden, Bærum, Asker and Gjerdrum in Norway; and Danderyd in Sweden.

Club 2 is composed of 55 municipalities where average household income levels increased moderately during the period under analysis, relative to other areas, but without completely detaching from the baseline. The club consists of one municipality in Denmark, Hørsholm; one municipality in Sweden, Lidingö; and 53 municipalities in Norway. Most of those municipalities are either located close to the capital cities, like those in Denmark and Sweden or, in the case of Norway, along the Oslo-Bergen corridor and in coastal locations in Rogaland (e.g. Sandnes municipality) and Finnmark (e.g. Alta municipality).

Club 3 gathers 317 municipalities where average household income levels have grown faster than those in Club 4, but not as rapidly as those in Club 1 and Club 2. The group consists of 274 Norwegian municipalities, alongside 20 municipalities in Denmark, 9 municipalities in Finland and 14 municipalities in Sweden. Examples of municipalities in this club include Rebild, Roskilde and Skanderborg in Denmark, Sipoo, Sottunga and Tuusula in Finland, Kvinesdal, Tysnes and Surnadal in Norway and Båstad, Solna and Österåker in Sweden.

Finally, Club 4 gathers 725 Nordic municipalities characterised by a relative decline in average household income during the 2005-2022 period compared to municipalities in the other clubs. Note that such a relative decline does not imply an absolute reduction in average income levels. Municipalities in this club include most municipalities in Iceland, Greenland and the Faroe Islands and most rural municipalities in Denmark, Finland and Sweden. Examples of municipalities in this club include Vejen, Glostrup and Nordfyns in Denmark; Kopavogur: Smarinn and Fifuhvammur, Reykjavik: Hlidar and Reykjavik: Vesturbaer south in Iceland; Kommune Qeqertalik, Avannaata Kommunia and Kommuneqarfik Sermersooq in Greenland; Engerdal, Våler (Innlandet) and Søndre Land in Finland; Rendalen, Dovre and Kongsvinger in Norway; and Helsingborg, Hörby and Tidaholm in Sweden.

A general conclusion from the club convergence analysis is that Nordic municipalities exhibit a relatively divergent trajectory in terms of average equivalised household income, particularly after 2019. That divergence may be related to the varying impact of the COVID-19 pandemic on household income, depending on how it is made up of income from labour, capital or transfers. The analysis suggests (1) a distinctly different trajectory for Norwegian municipalities compared to those in other Nordic territories and (2) a clear divide in the development of household income levels between municipalities in greater urban agglomerations and those in other areas within each territory.

However, more sophisticated interpretations of the temporal and spatial traits revealed by the club convergence analysis are challenging due to the influence of external macroeconomic factors. Specifically, the relative evolution of free-floating currencies and Purchasing Power Parity transformations may affect the outcome. Such factors might explain why national patterns seem to prevail over other spatial considerations.

A relevant aspect in understanding the genesis and evolution of economic inequalities is the relationship between income and inequality levels. The map in Figure 13 simultaneously represents average equivalised household income and inequality levels for all municipalities

With the exceptions of the Faroe Islands and Iceland, where the territorial divisions correspond to regions and Statistical Output Areas, respectively.

in the Nordic Region. The legend uses a bivariate colour scale based on equivalised mean household income and Gini coefficients calculated on the basis of mean income deciles within each municipality.

Figure 13: Bivariate classification of Nordic municipalities based on equivalised mean household income and Gini coefficients (2022)
Administrative boundaries: EuroGeographics. Source: Own based on data from NSIs.

Again, the spatial patterns shown in the map in Figure 13 allow Norway to be singled out as a special case regarding the distribution of household income within and between spatial units. In general, Norwegian municipalities show higher levels of average household income than other Nordic regions. Moreover, the distribution of household income at the municipal level also seems more equal in Norway than in other countries.

Additionally, the map in Figure 13 indicates marked territorial differences between areas. The largest Nordic urban agglomerations show higher income and higher inequality levels. That is true for the urban agglomerations of Copenhagen and Aarhus in Denmark, Helsinki and Tampere in Finland, Reykjavik in Iceland, Oslo and Bergen in Norway and Stockholm and Gothenburg in Sweden.

By contrast, some rural municipalities in predominantly land-locked regions of Norway and Sweden show lower levels of average household income and lower levels of income inequality. That pattern is visible in many cross-border and land-locked municipalities in Sweden, Finland and Norway, suggesting that smaller municipalities with traditional economies based on forestry and agriculture have evolved into low-income but socially cohesive communities.

Many intermediate municipalities and medium-sized towns in all countries and self-governing territories show lower levels of average household income, combined with higher-than-average levels of income inequality. Such municipalities tend to be located in cross-border regions of Finland and Sweden, as well as in Iceland and Greenland.

Figure 14 provides a more detailed view of the relationships between municipal income and inequality levels. The figure includes two scatterplots representing Gini coefficients of income inequality (on the horizontal axis) and mean equivalised household income (on the vertical axis). Each scatterplot refers to one territorial level (TL), where TL0 means national or self-governing territory level and TL5 means municipal level. Each point represents one spatial unit in one year. The colour legend identifies the observations per country or self-governing territory. To ease interpretation, regression lines summarising the relationship between income and inequality levels have also been plotted for each country. That does not necessarily imply that the relationship between both variables is linear.

Figure 14: Relation between income levels and income distribution in the Nordic Region (2005-2022)
Each point represents a spatial unit (country, region or municipality) in a given year. Note that all territorial levels are available for all countries. The lines represent simple linear regression for each country, with shaded 95% confidence intervals.
Source: Data from national statistical offices.

The scatterplots in Figure 14 show a positive correlation between household income and inequality levels across virtually all areas and scales. However, the distribution of the points suggests that the relationship is neither consistent across territorial levels nor between areas within the same territorial level. While a linear relationship between income and inequality levels appears to hold in some cases, in others such a relationship either does not exist or is not linear.

That indicates that the relationships between average municipal household income and inequality levels may be governed by a variety of spatial regimes. The plots in Figure 15 provide further insights into the relationship between those variables by country and degree of urbanisation. Each point on the charts in Figure 15 represents a municipality in Denmark, Finland, Norway and Sweden in 2022. The regression lines have been generated using two linear effects models. The model predictions shown in the chart on the left inclu­de random intercepts, while the one on the right in­cludes both random intercepts and slopes. The grouping factors used in both mo­dels are the countries and degree of urbanisation, as defined by the Nordic urban-rural typology (Nordregio, 2023). Each level is identified with a different colour and shape for the points and different colours and types for the regression lines.

Figure 15: Relationship between income and income inequality, by degree of urbanisation and country (2022)
*Degree of urbanisation based on EU DEGURBA typology.

The charts in Figure 15 confirm that the structural relationships between income and inequality levels vary in strength depending on the country and the degree of urbanisation of each municipality. In general, in more urbanised municipalities, income inequalities tend to grow proportionally with income levels. In those municipalities, the relationship between both variables is stronger than it is for intermediate and, particularly, rural areas, where the link between both dimensions appears to be somewhat weaker. Additionally, the differences in the strength of the relationship between average municipal income and income inequality by degree of urbanisation appear to be greater in Sweden and Finland than in Norway and Denmark.

Single-point inequality measures fail to capture the role of the geographical distribution of income. That is mostly due to the anonymity property of traditional inequality metrics, by which different spatial distributions of household income may yield identical inequality measurements (Panzera and Postiglione, 2020). In general, the lack of decomposable indicators of income inequality prevents the measurement and comparison of income inequality by spatially defined subgroups (Novotný, 2007).

Several inequality decomposition methods are available in the literature to address that issue (Attili, 2021; see e.g. Fields, 2003; Firpo et al., 2009; Mookherjee and Shorrocks, 1982; Panzera and Postiglione, 2020; Patil et al., 2014; Rey and Smith, 2013). Here, we have applied the method proposed by Rey and Smith (2013) to compare the Gini coefficients within each territorial unit with those in the neighbouring and non-neighbouring municipalities within each country or self-governing territory and year. That method, which considers spatial dependence in shaping the geographical distribution of income, yields a global Gini coefficient and the two components of the spatial Gini: (1) the inequality among nearest (geographically) neighbours and (2) the inequality of non-neighbours .

A shortcoming of that method is that it is highly sensitive to the number of nearest neighbours chosen. Since the spatial and administrative configurations of the sampled regions may differ, even at the same territorial level, the number of nearest neighbours selected can lead to biased results. In regions with fewer spatial units (like the Faroe Islands and Greenland) and in countries with very ad-hoc territorial configurations (like Iceland’s Statistical Output Areas), results are particularly sensitive to the number of nearest neighbours chosen. As a consequence, the and estimates should be interpreted with particular care in those regions.

For the sake of comparability, we sampled 1, 5, 10, 20 and 30 percent of nearest neighbours in each area. The results for a selection of 10 percent of nearest neighbours are shown in Figure 16. The chart indicates the extent to which the contribution of is above or below the expected value, in this case, 10 percent of the global inequality . That level is indicated by the horizontal dashed line that crosses the vertical axis at value 1. Points located above that line are those where the spatial component of inequality is greater than expected. Conversely, points below the line are those where the spatial component is smaller than expected, according to the number of nearest neighbours sampled. 

Figure 16: Inequality decomposition of income inequality in spatial and non-spatial components
Source: Own calculations based on household income income data proved by NSIs.
The horizontal line with 1 denotes equal contribution of within and between inequalities to

Looking at the results presented in Figure 16, it may be argued that the contribution of  to  is fairly stable over time in the largest economies and less so in the smaller territories. The relative contribution of spatial inequalities to the global  index ranges from roughly 5 percent in the Faroe Islands to 21 percent in the Faroe Islands.

With the exception of the least populated regions of the Faroe Islands, Greenland and Iceland, the overall contribution of the spatial inequality component remains below the theoretical level in most areas. Nonetheless, the contribution of spatial inequalities to the overall inequality, relative to the social income gap within the municipalities, is on the rise in several regions, particularly in the Faroe Islands. The lines for the larger economies also seem to indicate a slightly upward trajectory of the spatial component, particularly in Finland and Denmark and to a lesser extent in Iceland and Sweden. That emphasises the role of spatial logics of segregation and concentration in shaping income inequalities across the Nordic Region.

Again, the exception to this general rule is Norway, where the contribution of the spatial component of income inequality to the global inequality level  appears to be decreasing over time. That trend also holds when considering alternative numbers of nearest neighbours (not shown) and is consistent with the charts provided in Figures 7 and 8 above. The decomposition analysis hence confirms that Norway has been particularly successful in reducing the income gap between municipalities, even more so than between income groups.

According to Angell et al. (2016), three key factors explain why the regional disparities in disposable income per capita remain relatively small in Norway and have declined over time. The first is the active redistributive role of its powerful welfare state (also confirmed by Aaberge et al., 2018). Public transfers make up 25 percent of disposable income in many municipalities, particularly in those with a large share of people in the pre-work or retiring age groups. A second reason explaining the decline in income levels between urban and rural households in Norway is the development of several high-income industries, such as energy production, salmon farming, marine resources and oil-related activities, in many rural municipalities. A third reason is a well-developed public sector, which has a solid wage base and often dominates the local labour markets in rural and remote regions. However, understanding the role of policies that make Norway stand out as a more cohesive space, relative to other Nordic countries and regions, would require further comparative research.