The State of the Question

Roughly two different points of view may be identified on the question of the evolution of income inequality in advanced economies. The most accepted one is that income inequality has been rising and the income share of the middle class has been declining as a result of the same general forces working everywhere. These forces are in general referred to as globalization, and more particularly as skill-biased and routine-biased technological change (Milanovic Reference Milanovic2016; Goos et al. Reference Goos, Manning and Salomons2009; Fernández-Macías Reference Fernández-Macías2012). Global market forces erode low-skilled occupations, pushed away by cheap manufactured imports from emerging economies such as China, and also middle-skilled occupations, particularly upper manual and routine non-manual ones, likewise replaced by services outsourced to low-wage countries. These changes in the composition of the labour force are detrimental to the middle class and increase inequality of individual earnings. Moreover, in Europe, these forces could have been reinforced by the process of European integration, which weakens trade unions, enhances exchange and, in setting limits to budget deficits, reduces resources for welfare and redistribution (Beckfield Reference Beckfield2006).

The view that the forces behind recent trends in the inequality and polarization of disposable income mainly lie with the world of work, has been widely accepted by international organizations (Dabla-Norris et al. Reference Dabla-Norris, Kochhar, Suphaphiphat, Ricka and Tsounta2015; ILO 2015). However, this point of view could be seriously flawed because of the long chain of weak links between occupational classes and disposable income. To begin with, there is quite a difference between changes in occupational classes and changes in wage inequality and individual earnings distribution (Mishel et al. Reference Mishel, Schmitt and Ierhol2013; Reinhold Reference Reinhold2016). Manual workers and routine non-manual employees may face less favourable labour markets, so that they and their incomes grow less than professionals and their incomes; but this does not preclude that individual market incomes become more equal (if only more people work, for instance). And even if it could reasonably be argued that globalization and technological change have a un-equalizing effect on individual market income, there is still a long array of factors influencing its translation to household income and individual equivalent income, such as household size and composition (Carabaña and Salido Reference Carabaña, Salido and Moreno2000; McCall and Percheski Reference McCall and Percheski2011; Esping-Andersen Reference Esping-Andersen, Giddens and Diamond2005, Reference Esping-Andersen2007). It might be, as suggested by the OECD’s in-depth analysis, that demographic factors and political redistribution fell short of neutralizing these forces at the household level (OECD 2008, 2011), but the contrary could also happen, and in fact has happened in some countries. Briefly stated, different income groups, including the middle class, form their income package from all kinds of sources and form and manage households in many different ways, so that developments in labour and industrial relations cannot be expected to have a direct and explicit translation to the distribution of household equivalent income.

The second point of view, held by a minority of scholars, is just that all these factors are highly country-specific, so that country-level differences are relevant to understanding the ‘big picture’ of changing inequality and that national policies would be a key element in explaining recent trends in the European middle classes (Pressman Reference Pressman2007, Reference Pressman2010, Reference Pressman2015).

Contrary to the United States (US), where in accordance with the prevailing point of view, income inequality seems actually to have grown and the middle class actually to have decreased, the situation in Europe is rather paradoxical. On the one hand, official bodies insist upon the predominant narrative about income inequality and a waning middle class (see, for instance, Schmid-Drüner, Reference Schmid-Drüner2016) while on the other hand, statistical data produced by Eurostat and the empirical literature based on it are far from supporting this position, but rather the alternative paradigm of country-specific developments without a general trend.

Let us look at the empirical literature. Research shows a decline in middle occupations in some European countries (Oesch and Rodríguez Menés Reference Oesch and Rodriguez Menés2011; Fernández-Macías Reference Fernández-Macías2012); in Europe as a whole there has been a decline between 1993 and 2010 in middle-income occupations, as predicted by the Routine-Biased Technological Change (RBTC) theory, although not linked to globalization or offshorability (Goos et al. Reference Goos, Manning and Salomons2009, Reference Goos, Manning and Salomons2014). But Eurostat data show that Gini coefficients (measures of statistical dispersion) of income inequality, both before and after redistribution, hardly changed in Europe, at least since the mid-1990s, contrary to what happened in many other OECD countries, and more particularly in the US (Table 1). And certainly, in the EU-15, inequality increased in some countries and decreased in some others; since these changes compensated each other, the overall Gini index remained quite steady. Not only do Eurostat data raise doubts on the presumed increase of inequality in Europe, but also data from other sources (the Luxembourg Income Study (LIS), the OECD, the World Income Inequality Database (WIID), and the Review of Economic Dynamics (RED)) hardly support the view of inequality having increased in the EU-15 in the last decades, be it that inequality is calculated as the average of single country scores (Thewissen et al. Reference Thewissen, Kenworthy, Nolan, Roser and Smeeding2015; Nolan et al. Reference Nolan, Salverda, Checchi, Marx, McKnight, Tóth and van de Werfhorst2014; Ballarino et al. Reference Ballarino, Bogliacino, Braga, Bratti, Checchi, Filippin, Maestri, Meschi and Scervini2012), be it that it is directly estimated for a set of several countries taken as a whole (Darvas Reference Darvas2016; Troitiño Reference Troitiño2007; Brandolini and Smeeding Reference Brandolini and Smeeding2007; Salverda Reference Salverda2015; Bonesmo Reference Bonesmo2012; Bönke and Schröder Reference Bönke and Schröder2015; Wang et al. Reference Wang, Caminada, Goudswaard and Wang2015).

Table 1.

Changes in income inequality in the EU-15 since the 1980s (Gini indexes, sources as indicated).

(1) 2000 figures from Brandolini (Reference Brandolini2007); 2005, 2012 figures from Wang et al. (Reference Wang, Caminada, Goudswaard and Wang2015).

(2) Source: Salverda (Reference Salverda2015), from OECD-IDD data.

(3) Source: OECD (2015), In it together, Annex, Inequality Indicators.

In such a situation of income inequality remaining rather steady at least since the 1990s, a hypothesized squeeze in the income share of the middle class becomes in itself not less, but more intriguing. In a situation of unchanged inequality like this, Atkinson and Brandolini’s advice of giving special attention to the middle classes seems particularly relevant. For how could the income share of the middle class decrease without harming equality? Most measures of inequality grow when income is transferred from middle to higher income deciles, lessen when the transfer is to lower deciles, and are hardly or not at all affected when income is redistributed equally from the middle to the extremes. What are the research findings about the income share of middle and extreme economic classes?

Indeed, the only paper directly raising the issue of ‘squeezing the middle’ is the one co-authored by Atkinson and Brandolini (Reference Atkinson, Brandolini, Gornick and Jäntti2013). Based on LIS data, it covers the period from the 1980s to the mid-2000s and includes the following 15 countries: the US, Canada, Mexico, Taiwan, Poland, Germany, Austria, Luxembourg, Norway, Sweden, Denmark, Finland, France, the UK, and Italy. Eleven out of these are European countries, and nine belong to the EU-15. In these nine countries, the decrease in the income share of the middle 60% ranges between 3 and 4 percentage points in Austria, Finland, and the UK, and between 2 and 3 points in Germany, Luxembourg, and Sweden, mostly to the benefit of the top 20%; on the other hand, loss among the French middle classes has only been of one point and to the benefit of the bottom 20%, while the middle class share remains stable in Denmark and Italy (Atkinson and Brandolini Reference Atkinson, Brandolini, Gornick and Jäntti2013). These figures support the hypothesis of a general tendency towards the ‘squeezing’ of the middle class by the upper class, although affecting some countries more than others. The doubt remains, notwithstanding, about how this trend affects the six countries absent from the sample and the EU-15 as a whole, and moreover, about the developments since 2014, not yet covered by the study.

More or less at the same time as Atkinson and Brandolini, Gabriel Palma proposed his now well-known inequality ratio, based on the empirical finding that in most countries deciles 5 to 9 earn a proportion of income of around 50%. The empirical justification for this inequality index puts Atkinson and Brandolini’s findings in a wider frame, that now appear as a deviation of what according to Palma (Reference Palma2011) is a kind of a ‘universal constant,’ thus reinforcing concerns about the risks of generalizing to the whole of the EU-15 what happened in only six of its member countries. At the same time, this enhances the interest of the conjecture that Europe could be a deviant case of such a well-established regularity.

In summary, conventional wisdom (assumed and promoted not only by international organizations such as the IMF and the OECD, but even by official bodies of the EU itself) about a recent increase in inequality in Europe, boosted by exogenous market forces (i.e. globalization and technological change) and not compensated by internal policies or social developments, is not supported by the available empirical evidence. Such evidence, which is for the most part produced by the EU itself, disproves the existence of any upward trend in inequality after the 1980s in the EU-15, the only space consistently exposed to globalization and market forces over the last decades. In such a context of steady inequality, the income share of the middle class can only have decreased if extreme classes have benefited equally from this decline. This is precisely the hypothesis we will test hereafter.

Methods, Data and Variables

Classes are defined in terms of individuals, so that ‘income’ refers to individual equivalent disposable income, i.e. the quotient between household disposable income and the number of consumption units in the household, according to the OECD modified equivalence scale. Following Atkinson and Brandolini, we will define the middle class, in terms of population, as the middle 60% of the income distribution. The middle class will also be defined according to Palma as the deciles 5 to 9, reducing the upper class to the richest 10% and broadening the lower class to the poorest 40%.

This study spans a relatively large period of time, from the mid-1990s, when the longest expansion phase of the economic cycle began in most developed countries, up to the most recent crisis years. Our analysis is based on Eurostat data from the European Community Household Panel (ECHP, 1995–2001) and the European Union Statistics on Income and Living Conditions (EU-SILC, 2005–2016). These surveys provide the most comparable income information available for the different EU-15 countries, homogenized in accordance with Eurostat guidelines, for a relatively long period of time and on a large set of variables.Footnote ¹ They cover a period of growth, between 1994 and 2007, one of economic downturn, from 2007 until 2013, and still another one, though a short one, of recovery from 2013 onwards, allowing us to examine variations across different stages of the economic cycle. The information about income does not include Sweden and Denmark until 2004, but this hardly affects the main results.

Only the first 15 member countries of the EU (EU-15) are included in our analysis, on the understanding that they make up an economic zone of similar importance to that of the United States (although less homogeneous) and one that has also remained relatively stable over the last decades. The analysis will help shed some light on possible regularities or differences across the EU-15 countries according to institutional or geographical variables. One added value of our work is that, as far as we know, it is the first time that an empirical study of the ‘squeezed middle’ is addressed taking the EU-15 as a whole as well as all of its member countries individually, although this has only been done from 2004 onwards.

Is the Income Share of the Middle Classes getting Smaller in the EU-15?

If we look at the whole economic cycle, it appears that between 1994 and 2013 the mean income share held by the middle class, as defined by Atkinson and Brandolini, remained virtually unchanged in the EU-15,Footnote ² with only barely noticeable variations from 53% to 54% (Table 2). The same happens when following Palma’s definition of the middle classes that includes deciles 5 to 9: it merely raises the share of the middle classes by one point, from 54% to 55%.

Table 2.

Changes in the income share (in %) of deciles and three big income classes, EU-15, 1994–2013.

Source: Eurostat, Table ilc_di01 (ECHP and EU-SILC). Retrieved February, 2018.

Likewise, there is hardly any variation across the growth and crisis periods, except for an increase to 55% in 1999. Furthermore, the income share of the top and bottom quintiles shows no variation either across the business cycle: that of the poorest 20% went from 8% in 1994 to 7.7% in 2013, with only a deviation to 9%, again in 1999; the same happened with the income share of the richest 20%, which went from 39% in 1994 to 38.7% in 2013. Thus, if we were to take 1999 as the base year, we might deduce that there has been a drop in the income share of the middle class along with increased inequality between bottom and top tiers. But if we consider 1994 as the starting point, then the only conclusion that can be drawn is that of constancy, both during the expansion period and the recession period. Mutatis mutandis, this also holds for the extreme classes as defined by Palma.

Virtually the same result holds when, instead of the EU-15 country average provided by Eurostat, we take the average obtained from our own estimates for the EU-15 countries, in this case from 2004 to 2015 (Table 3). Following Atkinson and Brandolini’s definition, the middle class earns between 53.5% and 54% of the disposable income, with trendless fluctuations; using Palma’s definition, the income share of the middle class slightly increased from 54.4% in 2004 to 55.1% in 2012, a change that could be interpreted as a trend between 2004 and 2011 were it not for the 2007 figure of 54.4%.

Table 3.

Changes in the income share of deciles and three big income classes, EU-15, 2004–2015. (Calculated as the average for the 15 member countries(*).)

(*) Since 2013 Eurostat stopped presenting the data for the EU-15.

In summary, various developments from Eurostat data support the conclusion that there is no evidence of change, neither before nor after the recent economic crisis, consistent with a steadiness of inequality. And if there is no loss of income by the middle class, then the causes of the purported ‘squeezing’ cannot be investigated.

A Look at the Single Countries

Contrary to what happened at the global EU-15 level, there were variations in the income share of the middle class in some countries over the period (Table 4). Namely, over the last two decades, the income share of the middle class has risen by 1.5 percentage points or more in Belgium, Ireland, Luxembourg, the Netherlands, Portugal, and Spain, and has fallen in about the same proportion in Denmark and France (see Figure 1, panel A). In the remaining countries variations are below 1 point, and therefore may be considered as due to chance.Footnote ³ According to these data, we cannot establish a general trend toward the economic downturn of the middle class for the countries that make up the EU-15.

Table 4.

Income shares of the different economic classes in the EU-15 (1994–2013)^(*).

Source: Eurostat database, Table ilc_di01 (retrieved on 25 January 2017).

(*) Labels refer to the year previous to the survey.

(**) Data start in 2002 for Denmark, and in 2003 for Sweden. Series for these countries are calculated taking the earliest reference time available.

(***) 1994 data not available for Finland, so the series begins in 1995.

Figure 1.

Changes in the income share of the different economic classes (in percentage points) in the EU-15 across different phases of the economic cycle. To view this figure in colour please see the online version of this journal. (Source: Authors, using data from Eurostat, Table ilc_di01.)

A further question is whether the constancy of the middle class across the EU-15 is due to a consistent pattern across the economic cycle or to varying patterns cancelling each other out. A rough look at the data shows that there is no evident pattern of change associated with the business cycle, with singular variations across boom and crisis periods in the different countries (Figure 1, panels B and C). If any, the pattern of change seems to be countercyclical, with a larger number of countries where the middle class has improved its economic position during the downturn than during the previous period of economic growth. During the expansion period, the only significant increases in the income share of the middle class took place in Ireland and Spain, while during the crisis four countries experienced such increases, namely, Belgium, Portugal, the Netherlands, and the UK. As for decreases by more than one point, only two countries were concerned during the growth period (France and Germany), and only one during the crisis (Denmark).

Furthermore, the economic upsurge of the middle class during the expansion period was reversed during the crisis in most of the countries under study, with only the exceptions of Belgium and Luxembourg, giving way to a rise of the top income group and a decline of the lowest and middle income quintiles. Ireland, Italy, and Spain suffered a strong economic shock during the crisis that seemed to primarily affect the lowest quintile, followed by the middle class, while the top 20% increased their relative income with regard to the rest.

Looking now at the extremes of the distribution, we find that in four out of the five countries where the income share of the middle class increased by more than one point from the mid-1990s up to 2013, it did so mainly by reducing the share of the top 20%. Spain stands as the only country where the bottom quintile was the loser. In the two countries where the middle class ended up ‘squeezed’, it did so to the benefit of the lowest quintile in France and to the highest one in Denmark. Of the 15 countries that make up the EU-15, therefore, only Denmark fits the hypothesis of a ‘squeezed’ middle class in favour of the richer deciles.

These changes in income distribution point to the existence of varying country-specific patterns throughout the economic cycle, patterns that also imply a different balance between the extreme groups of the income distribution. We can therefore conclude that our data do not support the idea of ‘squeezing the middle’ as a general trend for the EU-15 countries, neither jointly nor separately, and neither before nor after the Great Recession. This finding about income shares is at odds with the results of Atkinson and Brandolini. But it should be noted that EU-SILC data do not go back to the mid-1980s and that only seven countries can be compared with both sources from the 1990s on.

Now, even if the disposable income share of the middle class has not changed in the whole of the EU-15 and in most countries, it could be that country-level changes were produced either by general forces affecting in diverging ways groups of countries similarly exposed to them, or by singular forces common to some of them. Thus, it might be that the income share of the middle class in the EU-15 as a whole has not changed because of divergent patterns in different groups of countries according to geography, welfare regimes, division of labour, economic specialization inside the EU or the impact of the recent financial crisis.

Here it should be recalled that following our findings (Table 4), over the last two decades the income share of the middle class has risen by 1.5 percentage points or more in Belgium, Ireland, Luxembourg, the Netherlands, Portugal and Spain; that it has fallen in about the same proportion in Denmark and France; and that it remained constant in the other countries (Austria, Finland, Germany, Greece, Italy, Sweden, the UK). Do these differences correspond to any common features between the countries under study? It seems evident that no clear geographical pattern can be established. Pressman (Reference Pressman2007) and Dallinger (Reference Dallinger2013) found that the changing size of the middle class to some extent matched Esping-Andersen’s (Reference Esping-Andersen1990) typology of welfare regimes – which otherwise responds to clear geographical patterns – but there is no way to find any association of this typology with the trends described in Table 4. For instance, among Mediterranean countries, the income share of the middle class increased in Spain and Portugal, while it remained stable in Italy and Greece; among Scandinavian countries, this income share decreased in Denmark, and remained constant in Finland and Sweden, etc.Footnote ⁴An influence of the business cycle on the size of the middle class was observed earlier by Horrigan and Haugen in the US (Horrigan and Haugen Reference Horrigan and Haugen1988), but not by Pressman in European countries (Pressman Reference Pressman2007). Changes in the income share of the middle class do not seem to parallel the intensity of the last economic cycle: during the growth years it grew in Spain and Ireland but not in Greece, and during the crisis it did not decrease, and even grew in Portugal.Footnote ⁵ It might be suggested that the division of labour within the EU-15 could have displaced middle-paying occupations to Central, industrial Europe but, according to Table 4, the Benelux, Ireland, Portugal, and Spain would have attracted these jobs from Denmark and France, something quite unlikely. Without entirely excluding this hypothesis – in particular, some demographics patterns could emerge – it must be acknowledged that the search for regular patterns or sociological regularities does not look very promising (see Salido and Carabaña Reference Salido and Carabaña2019).