RELATIVE TO WHAT?: CROSS-NATIONAL PICTURE OF EUROPEAN POVERTY MEASURED BY REGIONAL, NATIONAL AND EUROPEAN STANDARDS

The idea of the relativity of poverty and the social character of needs dates back to the writings of Adam Smith. Smith in his The Wealth of Nations (1981, 869–870 [originally published in 1776]): ‘By necessity I understand, not only the commodities which are indispensably necessary for the support of life, but whatever the custom of the country renders it indecent for creditable people, even of the lowest order, to be without’.² Thus, he argues, needs are highly socially determined. Hence, needs and the content of poverty vary in time and place. He nicely continues: ‘Custom … has rendered leather shoes a necessity of life in England. The poorest respectable person of either sex would be ashamed to appear in public without them. In Scotland, custom has rendered them a necessity of life for the lowest order of men; but not to the same order for women … In France they are neither a necessity for men nor for women …’

As a rule all international comparisons of poverty have been based on the relativist research strategy sketched by Smith: nation states being used as units of analysis. Poverty has been defined and operationalized within national boundaries and the criteria of relative poverty vis-à-vis income distribution have been determined separately for each country. The concept of poverty is thus by definition tied to the context of a nation state.

This nation-based relativistic research strategy has both its advantages and disadvantages. The advantages are obvious, as it has been the sovereign state and a specific nation-based political history that has traditionally produced social welfare for its citizens. Moreover, in spite of the growing impact of globalization and all associated aspects, nation states are still the locus of decision-making: they decide how to cope with changing global circumstances. Thus, various political and policy related considerations speak in favor of using nation states as units of research even in the globalization era.

There are also practical motivations for using the state as the unit of analysis: data is compiled using increasingly uniform national criteria, producing a fairly reliable picture of the development of inequality and poverty within a country when following changes over time.

Choosing a nation-bound relativistic approach is theoretically based on the reference group theory, which is derived from the notion that deprivation always has to be defined contextually (Halleröd 2004; Merton 1959; Runciman 1966). Tastes and preferences are context-bound and, therefore, poverty equates to lack of resources necessary for participation in the normal way of life of the surrounding society (Gordon and Pantazis 1987; Gordon and Townsend 2000; Townsend 1979). But, what is the right context in which to apply the reference group theory? How far should the idea of relative poverty be extended in international comparisons of poverty?

Thus, there are growing problems associated with nation-centered comparative strategies (for discussion, see e.g., Atkinson 1998; Berthoud 2004; Hagenaars and Vos 1988; Heidenreich 2003; Jesuit et al.2002; Rainwater et al.2001; Steward 2003). The relative poverty figures based on the income distribution of each country are, as the name suggests, highly relative, i.e., dependent on the shape of the income distribution of each country, not on absolute income. If the shape of the income distribution for any two countries is the same, the relative degree of poverty will also be the same when determined by, say, 50 or 60 percent of the median income. Yet one of the countries could be significantly more prosperous in terms of per capita national product, and the poor people in one country may even be classified as rich in the other. To take an example, the relative poverty rates in the US and Estonia may be the same but those classified as poor in the former were, in absolute terms, prosperous in the latter. Thus the relative poverty rate is very much dependent on the form of income distribution and therefore the relative method easily loses sight of the connection between poverty and actual subsistence levels. (For criticism of the relative poverty measure, see Ringen 1987; Sen 1981.)

One of the ways this problem of excessive relativization can be alleviated is by composing comparative units that are larger than individual countries. We can, for example, set a common poverty line for all the Nordic countries (as in Kangas and Ritakallio 2000) or merge European nations together and apply a common standard for the constructed ‘Euroland’ (as in Beblo and Knaus 2000, see also Atkinson 1998) and then study whether the depiction of poverty is different compared to that produced by poverty lines at the national level. The extension of poverty lines to the European Union in particular is pertinent today, as nation states are losing their decision-making powers to Union-level supra-national bodies, and the comparative perspectives of a growing number of people are extending beyond national boundaries. It can also be argued that Western mass media has built a more uniform reference base for Western people; at least, the Europeans are more explicitly comparing themselves and their standard of living with levels of living in the other member states of the European Union. This kind of intra-European comparison is facilitated by the common European production of income and poverty statistics (e.g., European social statistics compiled by the Eurostat). In sum, the deepening European integration, common currency and intensifying cultural globalization create uniform standards of comparison and common European yard-sticks for measurement of poverty and standard of living in different areas of Europe. National standards will eventually be replaced at least partially by EU standards.

In a similar vein, while globalization can justify the use of reference units which are larger than nation states (such as Scandinavia, Europe as a whole, or the entire OECD area), it is also tenable to look for analytical units smaller than that of the nation state. Visions of a Europe of Regions, i.e., Europe that has been divided in terms of regional affiliations, for example, would motivate the use of regional units in analysis, such as European capitals or European peripheries. This approach is also politically well-motivated: the whole idea of structural funds in the European Union aims at eradicating or mitigating regional disparities in the Union. We could go a step further. Various occupational groups are without a doubt comparing their income levels with incomes earned by the same occupational groups in other European countries more frequently rather than with the average wage levels of their own country.³ Thus, there may be substantial occupational and regional differences that are concealed in national level inspections. In this study we are going to touch upon income differences between regions, nations and the European Union, or ‘Euroland’ if you prefer. The interesting question as to what different social groups compare their living standards deserves its own separate article (see for example Halleröd 2005; Kapteyn et al.1985, 1988). This analysis, and the inspection of other poverty measures, are outside the scope of this paper, which shall concentrate only on the monetary side of poverty (for different operationalizations of poverty and their empirical implications, see for example, Gordon and Pantzatis 1997; Hagenaars and Vos 1988; Kangas and Ritakallio 1998; Maître 2004; Whelan and Maître 2005).

The aim of this study is to play with different kind of relativization and establish what will happen if we, instead of using conventional nation state-based poverty lines, apply a common European poverty line or regional lines. Our starting point is the conventional approach where we will compare poverty rates in 13 member states of the European Union. Thereafter, in order to analyze poverty at the international level, we pool these 13 national data-sets together to form a larger supranational Euroland data-set. Here, we will shed some light on the incidence of overall European poverty. As one of the central goals of the European Union policy making is to bridge the gap between advantaged and disadvantaged regions, we will divide Europe into privileged areas – which are often capital cities or the most prosperous areas in a country – and poor peripheries. Thereafter, we will calculate separate poverty lines for these smaller supranational entities to observe both the within-country and between-country variation in poverty. Separate national sub-areas are thus compared nationally and then internationally. This kind of methodological exercise, we believe, will give a more nuanced picture of poverty as a regional, national and international problem. So far, only a handful of studies of this nature have been conducted. This approach also allows us to evaluate in more detail the relevance of national median-based poverty measures. Our approach is depicted in Figure 1.⁴

The standard method of measurement is represented by the fifth box: both the poverty line and the unit of measurement are national state. Our analysis will concentrate on all other boxes, except the fourth and seventh, once. In principle it had been possible to calculate regional poverty rates for all sub-regions within a nation and, on the basis of these figures, to construct a new national poverty rate (box 4). Similarly, it had been possible to follow the same procedure and calculate European poverty rates on the basis of regional poverty measures. However, since we shall only deal with the richest and poorest regions in the country, as described later on, we skip these analyses.⁵

The structure of the paper is as follows: in the subsequent section, we briefly describe the database used and explain our methodological choices. Thereafter, we discuss regional, national and international poverty lines which constitute the basis for our analyses on poverty levels. The penultimate section, in short, assesses who the European poor are. The final section discusses the findings at a more general level and addresses some problems connected to the enlargement of the European Union.

The conditions for carrying out comparative research on income distribution have improved greatly with the development of the Luxembourg Income Study (LIS) project. The most pertinent achievement has been the databank made available to the research community (see Smeeding 2002; Smeeding and Vlemincks 2001; Smeeding et al.1990) that contains commensurate information for 29 countries. Each country's data-set includes information on two to fifty thousands households’ income and income formation, i.e., how much of their income consists of salaries, capital or business income and various kinds of received and paid redistributive sources. Also, for each household, information is available on the essential structural features such as the type of household, age of provider, number of children, and numbers of wage earners or recipients of other income, as well as the educational attainment, profession and social group of the provider. The national variables have been ‘lis-ified’, i.e., they have been homogenized by the LIS experts to guarantee the best comparability as possible.

Despite efforts to make the different variables as uniform and commensurate as possible, the LIS data are by no means unproblematic. In the Swedish data, for example, a problem is that all people (children) over 18 years of age who live in their parents’ homes have been accounted for as separate households. As these kinds of households are often without personal income, the Swedish data overestimates the extent of Swedish poverty. In this study, we have corrected this by excluding the data on all persons under 30 years of age who live alone and whose income level is below 30 percent of the median income of the population (half of our poverty threshold) which is well below the statutory minimum income security level for a person living alone. This adjustment removed 2 percent of the cases, and consequently, the poverty situation in Sweden appears less dire (e.g., at the 60 percent poverty line, the Swedish poverty rate without the adjustment is 9.1 percent; with the aforementioned adjustment it is 7.1 percent), and the Swedish data becomes more commensurate with those of the other countries.

For most countries, there is also a prodigious amount of cross-sectional data (for the United Kingdom, for example, there are cross-sectional data from the years 1969, 1974, 1979, 1986, 1991, 1995 and 1999). Here we will use the fourth wave of LIS which is for the mid to late 1990s. For some countries – such as Finland, Germany, Italy, Luxembourg, The Netherlands, Sweden and the UK – data are also available for around 2000, whereas for other countries – such as Austria, Belgium, Denmark, France, and Ireland – the latest data are for the mid 1990s (for Spain the data is for 1990). We decided to use the latest ‘complete’ wave. As previous studies (see for example Fritzell and Ritakallio, 2004) display that poverty figures are rather robust over time, the time gap between national waves does not severely contaminate our results. Data for Greece and Portugal are missing from the LIS and hence the number of countries included in our exercise is 13 instead of the entire 15 “old” EU countries.

Data-sets cover all persons except those living in institutions. The flexibility of data allows us to manipulate and re-group data at national, regional and international levels. Thus, the micro-level data available in the LIS databank makes it possible to compare income distribution, poverty, and income equalizing effects of socio-political schemes between countries included in the database flexibly and accurately. It is worthy of recognition that here we only use income-based poverty lines and do not employ consumption-based poverty measures – which, strictly speaking, the ‘Smithian’ approach demands – or subjective feelings of poverty. Such practices have been exercised in other studies (see for example Berthoud 2003; Gordon and Townsend 2000; Halleröd 1997; Kangas and Ritakallio 1998). In this article we do not have the possibility to touch upon the issue of subjective poverty; instead we concentrate solely on income.

As a rule, the LIS data contain weights with which the samples of each country can be ‘raised’ to the level of the total population. In the case of Belgium, Denmark, Germany, Italy, Luxembourg, The Netherlands, Spain, and the UK there are no such weights, so we constructed the weights ourselves by dividing the total population by the sample size. When pooling the national data-sets into the common European data-set, each country has been weighted by the size of its population. Germany, the most populous country with its 82 million inhabitants, has the biggest weight and Luxembourg, with 0.4 million inhabitants, has the smallest. The other countries lie between these two extremes (see Table 1). By applying this methodology we increased the original LIS sample size of 387,746 cases to represent 347 million inhabitants in the “old” EU countries.

When constructing the common European poverty line, each national income data were, when necessary, deflated to 1995 values by using national consumer price indices, whereafter data were converted to a single currency, i.e., to Euros by using purchasing power parities.⁶

The income concept applied here includes wages/salaries plus income from self-employment and capital income (for all members of the household), which together comprise factor income for the household. Factor income plus all transfers paid to the household form the gross income of the household and the concept disposable income is gross income minus taxes and other transfers the household must pay. Our concept of disposable income does not include the value of social services, which are difficult to evaluate (for a closer discussion, see for example Saunders et al.1992). The LIS-data for households were converted into data for individuals by entering the material for each household into the data as many times as the number of household members. So, the research unit used here is the individual and his/her income is the sum income of the household divided by the number of consumption units in that household (for a closer description of the methods, see Gustafsson and Uusitalo 1990). The equivalence scale used here is the so-called old OECD scale (in fact the scale used does not have such a great impact upon our results).

After the above-mentioned adjustments to the data, it was possible to apply the common methods of poverty research. For the sake of simplicity and space considerations, we restricted ourselves to the most commonly-used definition of poverty, defining as poor as those persons whose disposable income falls below a certain level of the median equivalent income (using the OECD equivalence scale) of the population in any area being investigated. By utilizing these common methods we also share the advantages and disadvantages of such approaches (for a closer description, see Mitchell, 1991; Ringen 1987; Saunders, 1994). In order to test the sensitivity of our results, we measured poverty rates by using two poverty thresholds, i.e., 50 and 60 percent of median income. In principle there were only minor differences between the thresholds. Therefore, to make the tables more reader-friendly, we only display the 60 percent results, which is also in line with the Eurostat procedure⁷ (Atkinson et al.2001).

In the previous sections the hypothesis presented was that the proceeding European integration will lead to a common European frame of reference in terms of the standard of living; therefore, the use of some kind of common European poverty line is warranted. However, there are also other processes that may lead us in another direction. We can think that Europeanization will gradually dissolute the powers of nation states and regional units will grow in importance. If this ‘Europe of Regions’ materializes, then instead of using a nation state as a reference group, it is conceivable that people living in the capital area will compare themselves with others living in the same area, not with those living in national peripheries. Furthermore, people living in capitals in different European countries are surely more prone to compare themselves with those resident in other capital areas in Europe. Correspondingly, the people living in less prosperous regions may use their neighboring areas as a frame of reference.

In Table 2, we played with these kinds of relativizations and present national, regional and European poverty lines and European-level rankings. For each country we first calculated the conventional national poverty line (the upper number printed in bold in the second column from the left) and then – providing that a regional variable was available in LIS – we counted separate poverty lines for the richest (often the capital city) and the poorest areas (measured by LIS income per capita). Thus the table also gives some hints how much within-country variation there is in income levels. Finally the national and regional poverty lines were related to the weighted overall European poverty line (the third column from the left) to indicatively see how sensitive countries are to shifts in poverty thresholds. Let us take a look at Austria as an example of how to interpret the figures in Table 2. The national poverty line in Austria is €6,456 or 106 percent of the weighted European common poverty line (€6,081). The Austrian poverty line is the seventh (7) highest in Europe. The regional poverty line for Vienna, which represents the rich area, is €6,939 which is the ninth (9) highest of all regional poverty lines. For the poorer Tirol region the corresponding indicators are €5,650 and [17]. Differences in rank-orders can be used as preliminary indicators of income disparities within a nation. The greater the gap in rank between the rich and the poor areas, the greater the within nation differences are likely to be.

The national poverty lines can be interpreted as indicators of the overall level of income in a country. At the national level the Luxembourgers in particular seem to enjoy an extremely high level of income followed by the Danes. The other rich countries are much closer to the European mean. The regional figures reveal interesting stories. There are only two exceptions – Ireland and Spain – where all the poverty lines, i.e., the national and regional lines are below the European level, whereas in all other countries, at the very minimum, the richest area exceeds the European mean value. In four countries (Belgium, Denmark, Luxembourg, and Sweden) all indicators are higher than the EU benchmark.

As can be seen, there are substantial differences in poverty lines between countries (cross-national differences) as well as between regions (within-nation differences). The range of variation between nations spans from the minimum value in Spain, 65 percent of the EU mean, to the maximum Luxembourg value of 173 percent. Thus the between-country range is as much as 108 percentage points. The national poverty line for Luxembourg is 2.7 times higher than that of Spain. The between nation variation is largely contaminated by the extreme Luxembourg figures; if we exclude that case, the national differences diminish remarkably and the richest versus poorest country ratio is 1.8. The figure for the variation between richest European regions is at the same magnitude (Roskilde versus Catalonia 1.7), whereas the ratio for the poorest areas is a slightly higher (Viborg to Sicily 2.2).

In some cases we find substantial within-nation variation. The within country variation between the rich and the poor areas is the biggest in Italy (Milan to Sicily ratio = 2.2) followed by Spain (1.5), France (1.5), Ireland and Germany and the UK (1.4). Regional differences are smallest in Belgium and Sweden (1.1) followed by Finland (1.2) and Denmark and Austria (1.3). Among this latter group of countries all poverty lines hover around the European mean which indicates that these cases are the most robust and they are not as sensitive as the other countries to the choice of relativization.

The situation outlined above can be seen in condensed form using the coefficients of variation (CV) presented in Table 2. The CVs are, to some extent, sensitive to tiny Luxembourg: the variation both between nations and between rich areas decreases (from 0.24 to 0.16) if we omit this outlier. According to the CVs, relative differences are bigger within nations than between nations, i.e., cross-national differences are smaller than national differences. This appears to substantiate our previous criticism of the use of nation states as the only point of reference. The picture is however rather more nuanced as displayed in Figure 2.⁸ Our interpretation is dependent on the way we compare countries. If we apply the absolute comparative strategy, the differences are greatest between the rich areas and between nations. The smallest absolute differences are found between nations. The relativist way of comparing (the right-hand side of the figure) reverses the picture. If we standardize differences by dividing them by the respective means, our conclusion is the same as that given by the coefficients of variation: differences within nations are more pronounced than between nations.

The crucial question is which comparison is the correct one? Are relative or absolute differences the ‘real’ ones? How should comparisons be made? On the basis of the LIS it is impossible to answer to this question; however, tentatively, one could present a couple of hypotheses. Firstly, when it comes to comparisons within a nation, people usually use relative comparisons. The Swedes living in Northern Sweden compare themselves with the Swedes living in the southern part of the country. Secondly, when it comes to the international comparisons we usually use absolute measures. Thirdly, we could assume that people tolerate bigger differences on the cross-national level than on the within country level. People living in Sicily are perhaps not that upset about the wealth of the Luxembourgers yet are much more sensitive to the income gaps in comparison to the rest of Italy.

Poverty is always attached to a lack of money. Therefore, it is interesting to see to what extent regional differences in income are attached to the general income level of the country in question. In Figure 3, we have used national poverty lines as proxies for national prosperity and plotted regional differences (in percentages) against this measure. Among the European countries (the extreme case, Luxembourg, excluded) there is negative correlation (r= − 0.52) between the income level of the country and the depth of regional disparities within a country: the poorer the country, the bigger the regional differences. In this inspection, Italy is a clear outlier with its extremely high regional differences. The two other low-income countries, Spain and Ireland, display lower degree of disparity compared to Italy. At the other end of the continuum we find Denmark, Sweden and Belgium with high income levels and relatively small differences between regions. The exclusion of Italy, Spain and Ireland will decrease the correlation (−0.30); nevertheless, there is a tendency that in rich countries regional disparities are smaller than in poorer countries. The same negative relationship is also evident when it comes to the overall income differences within countries and national prosperity. The correlation between national Gini coefficients (Table 2) and national poverty lines is significantly negative (−0.60). In poorer countries the overall income inequalities tend to be larger.

In Table 3 we present results from measurements based on different poverty lines. First we calculated national poverty rates (presented in the second column from the left) based on national 60 percent poverty lines. The third column displays national poverty rates when the measurement is based on the European poverty line. The remaining columns depict regional figures. In the fourth column, poverty lines are regional, i.e., for rich areas and poor areas there are separate thresholds, and consequently, the poverty rates shown in the column are regionally relative. In the fifth column, we apply the national poverty lines to regional units and finally, in the last column, the figures indicate what the regional poverty rates were if we applied the common European poverty threshold for regional analyses.⁹

According to national poverty lines, the weighted European average is 15.5 percent (Table 3, last row, second column from the left). The incidence of poverty varies from the 7.1 percent in Sweden to 20 percent in the UK Countries could be loosely merged into three groups: in Finland and Sweden the poverty rates are clearly below 10 percent. In the second group of countries the poverty rates vary between 10 and 15 percent (as in Luxembourg, Belgium, the Netherlands, Denmark, France and Germany), whereas in the rest of the countries the rates are higher: about 16 percent in Spain and about 18 percent in Ireland and Italy, and 20 percent in the UK.

The large differences in poverty lines presented in Table 2 are naturally mirrored in poverty rates based on the European median income (the third column in Table 3). Some countries are more sensitive than others. In poorer countries the shift from the national standard to the European measure increases poverty. The shift from the national poverty lines to the common European poverty line raises the overall weighted European poverty rate from 15.5 to 18.2 percent (Table 3, last row). In Spain, the use of the European threshold would almost treble the poverty rate. In Italy poverty would increase by 9 percentage points and in Ireland by 14.6 percent points. Belgium, Denmark, Finland, Germany, Luxembourg, and Sweden display patterns to the contrary. In these countries the shift to the European measurement decreases the poverty rates. For the rest of the countries the change does not have great consequences. Perhaps the best pair of countries to compare in order to highlight the importance of the poverty line used is Austria and Spain. If we apply the national poverty lines, these two countries are precisely the same, whereas the use of the European threshold leads to tremendously higher poverty rates in Spain and not that much happens in Austria.

We have applied three different poverty thresholds to regional analyses, each of them telling a somewhat different political story. The national poverty line will tell the national policy makers if there are huge disparities between the rich and poor regions in a country. The inspection based on the European poverty line indicates whether there are some regions that are lagging behind in EU development and whether they are in need of structural subsidies. Finally, the third approach is purely regional in the sense that we apply different thresholds for the prosperous and poor regions.

On the basis of national poverty thresholds we can see that regional discrepancies in the prevalence of poverty are widest in Italy. In Milan district the poverty rate is only 7.3 percent, whereas almost half of the Sicilians are classified as poor. The gap widens further if we use the European standard: almost two-thirds of Sicilians are poor. The same kind of phenomenon is visible in Ireland and Spain. Sweden is the other extreme. There are no substantial differences between the Swedish regions, whether we use the national or EU poverty thresholds. To some extent the same holds true for Austria, Germany and Denmark. In Austria and Sweden the overall national figures are higher than corresponding indicators for the two extreme areas. In all other countries poverty rates in the prosperous regions are lower than the national average.

Different relativizations lead to somewhat different conclusions on income gaps between rich and poor regions as indicated in Figure 4. If we use national poverty lines, the dispersion of poverty rates between the poor areas is bigger than between rich areas (and even greater if we use European poverty lines). Perhaps not surprisingly, the use of national poverty lines reveals the poverty rate to be higher in peripheries than in rich areas (left hand panel in Figure 4).

The regional relativizations (poverty rates are calculated separately for each area on the basis of regional poverty thresholds) in the right-hand panel in Figure 4 (or the fourth column in Table 3) indicate an interesting and perhaps a slightly surprising result: relatively speaking there is less poverty in the poorer areas than in capitals. The poverty rate for the peripheries is lower in Austria, Denmark, France, Germany, Ireland, Sweden, and the UK – a factor which indicates that incomes in poorer areas in these countries are generally speaking more evenly distributed than incomes in the most prosperous regions.

Our inspection above has some ramifications for comparative welfare state studies. Consequences for welfare regime-based interpretations given in Table 3 are pretty much the same as in Table 2. Results for the Scandinavian regime are the most robust against all kinds of data-manipulation: regional differences are negligible and none of the shifts from regional, to national, or even further to the European poverty line, affect the results. Only in the case of the three Nordic countries (fortified by Luxembourg) are the results based on national averages representative, while in the case of Italy and Spain the results are for the most part contaminated. In Italy, for example, the average poverty rate at the 60 percent level is about 21 percent, whereas the very same measure gives 7.9 percent for Milan, and as much as 49 percent for Sicily. Thus at least in the Italian and Spanish cases we can wonder what the national mean is actually good for. The European level comparison yields even greater discrepancies.

The gaps between the smallest and largest poverty measures given in Table 3 vary significantly between countries. On one hand, we have Sweden (with a gap of 4.7 percent points), Austria (6.4), Finland (7.3), Denmark (8.0) and Germany (9.6) which display robust poverty rates regardless of the measurement used. On the other hand, we have Spain (51.4) and Italy (39.3) and Ireland (29.8) where results are highly sensitive to the measurement, which is of course due to the lower income levels in these countries. Luxembourg is also an unstable case but for the opposite reason to the three former countries. The shift from the national poverty level to the European level would totally eradicate poverty in rich Luxemburg. When it comes to the welfare state regimes, the testimonies given by our two ways of comparing are pretty much the same for the Scandinavian and central European models, while more dramatic changes occur in the Mediterranean regime. The poverty rates for the three Nordic countries are among the lowest regardless the relativization used. The same goes for the Central European countries.

In the beginning, we posed the question as to how warranted it really is to use nation states as research units. Our inspection above gives some tentative answers. Some countries are more homogeneous than some others. One way to evaluate the homogeneity of countries is to calculate dissimilarity indices for our regions and see whether different areas within a country are more similar than different areas in other countries. The results from distance correlation (between cases and Euclidean distances; variables used in the analysis were the poverty rates according to regional, national and EU thresholds and regional poverty lines in Euros) analyses are presented in Table 4 where the correlations are rescaled from 0 to 100. The former indicates the closest relationship and the latter the greatest distance. In a sense the table can be interpreted as a sociomatrix.

As a rule, in all cases the closest case is not from the same country. For example, in the Austrian case the distance between Vienna and Tirol is 24, whereas Vienna displays a much closer correlation with Belgian regions, Viborg, Helsinki, Milan and Stockholm. Tirol, for its part, has similarities to Calais, Saxony, Dublin, Lapland and south-east England. On the basis of dissimilarity indices, there are three cases that deviate significantly from the rest: Sicily and Andalusia for their poverty, and Roskilde for its wealth.

Conventional studies on poverty usually operate at two different levels. First, they inspect which cases, be they socio-economic groups, family types and so on, are the most exposed to poverty. The aim is to identify the most poverty prone groups in society. Second, since some groups may be extremely poverty prone (e.g., the unemployed) yet small in number, their total contribution to the overall poverty rate is marginal; meanwhile, some bigger groups (e.g., families with children), among which the incidence of poverty is much lower, may contribute to the overall poverty rate to a greater extent. The issue is socio-politically important: should we give priority to measurements which help the most exposed group even though this would bring the overall poverty rate down significantly, or should we develop policy measures to help the larger group (who might be better-off) in order to reduce poverty rates? In principle we have followed the same procedure in our analyses above and in principle the policy implications at the European level are as described above.

We applied the first approach in Table 3 where we established which countries were the most poverty prone. We also conducted an analysis of the second type by attempting to see which countries are the most responsible for the prevalence of poverty in the European Union. Results from this exercise are displayed in Table 5.

Here again the results are a slightly different if we use national or European relativizations. According to the national 60 percent poverty lines, there are approximately 54 million poor people in Europe. The use of the European poverty line increases the number of destitute people to 63 million. When it comes to the contributions of individual countries, the biggest countries, the UK, Germany, Italy, France and Spain, contribute most to the overall European poverty (12, 11, 10, 9, and 6 milion, respectively). If instead of using national poverty lines we use the EU poverty line, the very same countries ‘explain’ the incidence of the European poverty but now – not that surprisingly – the impact of our two Southern European countries is of the most importance: they comprise as much as 52 percent of the total European poverty, 33 million. By national standards the number of poor people in these two countries was half that, 17 million. This is visible evidence that the way of conducting the relativization matters. It is important to recognize this from the point of view of European social policy making. It makes a huge political difference if the biggest contributors to European poverty are the UK and Germany (as they are by using conventional standards) or Spain and Italy (as they are by using Common European Standard). Who should the European Union help? The issue is even more accentuated presently when the EU is facing various challenges with the new member countries.

When it comes to social policy decision-making, the story told in Table 3 gives some leads both to domestic and European politicians. Relative national poverty lines help detect the groups in society who are most in need of state intervention. Helping those groups is within the power of national politics. When it comes to the EU, the use of European standards depicts the regions in Europe that perhaps are most in need of help using EU funds.

We began with a quotation from Adam Smith who eloquently spoke of the relativity of needs. In this study we were not afforded the opportunity to go into detail about the social and subjective determination of human needs. Instead, we followed the path paved by Adam Smith by shedding light on to what extent our picture of poverty will change if we accept a very relative concept of poverty. The first problem we encountered was the selection of the benchmark. If needs and poverty are relative, which standards should we then apply? In our play with data, we selected a couple of alternative ways to conduct relativizations. First, we applied the conventional poverty approach. The poor were those whose income remained below 60 percent of the national equivalent disposable income. Second, we collapsed European nations together into one data pool and calculated a common poverty line for the EU. This EU line was then applied in subsequent analyses. Here we tried to see if the Britons have their leather shoes on and if the French are walking in their wooden shoes. Finally, in order to see to what extent the Scottish and the Britons differ, as argued by Smith, we decomposed nation states into smaller units representing the poorest and richest areas in the respective countries.

Our substantial findings fortified the wisdom gained from previous research: the Scandinavian countries display the lowest poverty rates, followed by the Central European nations. The prevalence of poverty in the Mediterranean area is much higher than in the two other groups of nations. However, if we apply the conventional ways of operationalizing poverty, the cross-national variation of poverty is not that big. According to the national poverty lines and 60 percent poverty thresholds, the poverty rate varies from 7.1 percent in Sweden and to 20.5 percent in Italy. A shift to the common European poverty line widens that gap. The variation is from 0.7 percent in Luxembourg to 43.1 percent in Spain.

Numerically and methodologically the most interesting issues are revealed when we compare regional, national and EU level relativizations. There are substantial regional disparities in Italy and Spain, while regional differences are much smaller in Scandinavia. Methodologically, we must ask what we are comparing, de facto, if we compare nations. Our exercise indicates that within-nation differences are often more pronounced than differences between nations. Therefore, very often national means tend to obscure more than they reveal. The seriousness of the problem varies between groups of countries. In the egalitarian Nordic countries, incomes between regions as well as between individuals are more evenly distributed and consequently, the national means are more representative for these countries. Moreover, the Scandinavian cluster is more or less robust against the mode of comparison. The low poverty rates in the Nordic countries essentially do not change even if we change from national to regional or cross-national poverty lines.

The change in the method of relativization does not alter our understanding of Scandinavian poverty, but it has a substantial impact upon our picture of the Mediterranean countries. The use of the European poverty line leads to two to three times higher poverty rates than analyses based on purely national data. In addition, the regional variation in these countries is the widest. Therefore, conclusions based on national means may in some cases be severely misleading. The results also have some bearing on our use of purchasing power parities. In societies with large socio-economic and regional variation in income, and consequently in consumption capacities, purchasing power parities implicitly assuming homogeneous consumption patterns within society may give a distorted picture of the price levels in a country in question.

When it comes to the Central European countries, to some extent the same story as was told in the Scandinavian case is valid. The countries are not that sensitive to changes in the calibration of the measurement instruments. The results for the UK are also fairly robust, but the main difference between the UK and central Europe is that the poverty rate is about 10 percentage points higher in the former. In comparison to the French, the Britons seem, nowadays, to have problems with their shoes.

The play with different relativizations is not just for fun and academic acrobatics – it has important policy implications as well. The conventional nation-based relative measures have an important story to tell for the national policy makers: Which are the groups in society most exposed to low incomes? What should be done to help those in need? From the national perspective regional analyses are useful especially in cases where substantial price differences exist between regions (cf. Simiski and Saunders 2004). A lower income level in cheap areas may in some cases lead to higher living standards than a higher income in a more expensive area. Therefore, the inspection solely of low incomes is not sufficient to give a reliable picture of the well-being of citizens.

Analyses based on absolute European-level poverty lines provide data on the areas most in need of subsidies from the EU. In our analyses, with the exception of Ireland and Spain, the poverty line for the richest regions exceeded the European poverty line and, in many countries, so did the poorest region. The enlargement of the European Union has totally change this picture. The EU has gained a substantial number of new countries (and a substantial number of people) where both the rich and poor regions clearly lag behind the European mean values. Researchers will have a number of additional questions on relativizations to answer and Eurocrats will have a couple of relative and absolute political problems to deal with.

Olli Kangas holds a Ph.D. in sociology and is currently Research Professor at the Danish National Institute of Social Research in Copenhagen. His research is focused on comparative studies on causes and consequences of social policy institutions. Among his current publications is Social Policy and Economic Development in the Nordic Countries (eds with Joakim Palme). Palgrave, London (2005).

Veli-Matti Ritakallio is professor at the Department of Social Policy, University of Turku, Finland. His major research interest has been crossnational comparisons of social policy and poverty.