Abstract
The German government decided to use Amartya Sen's capability approach as the conceptual framework for the national ‘Poverty and Wealth Reports’ but concluded at the same time that the purely income-based at-risk-of-poverty rate (AROPR) is a satisfactory instrument to operationalise the capability approach. This decision made the latter the official measure to analyse poverty in Germany. This paper studies the question whether this conclusion is indeed justified by introducing two different multidimensional poverty measures to operationalise the capability approach. A thorough empirical analysis compares the poverty evaluations of the three poverty measures over time. It reveals that they differ considerably with regard to poverty trends, the identification of the most deprived and the impact of location, especially regarding West and East Germany, which may have considerable implications for targeting and demonstrates that there is indeed an urgent need for multidimensional poverty measures that complement the traditional AROPR.
1. Introduction
Until the election in 1998, no official poverty assessment existed in Germany. When the new German government finally decided to implement regular ‘Poverty and Wealth Reports’, it assigned the Institute for Applied Economic Research (IAW) the task to develop a conceptual framework for these reports. The responsible research team recommended using Sen's (Citation1979, Citation1985, Citation1987, Citation1992, Citation1997, Citation2009) capability approach to conceptualise poverty and wealth in the German context (Arndt & Volkert, Citation2007; Volkert et al., Citation2004).
The capability approach makes capabilities and functionings, i.e. what a person is actually able to do and be, the subject of analyses rather than economic resources. Poverty is then defined as a restricted set of essential capabilities that are needed to pursue whatever one has reason to value in life. This way, the approach places freedom of choice at the core of any poverty analysis and can do without the rather restrictive assumptions of other poverty approaches.
The German government decided to follow the advice of the IAW research team and to use the capability approach as the conceptual framework for its ‘Poverty and Wealth Reports’ (Arndt & Volkert, Citation2007). However, when it came to the operationalisation of the capability approach, the government concluded that the purely income-based at-the-risk-of-poverty rate (AROPR)Footnote1 is an indirect indicator for social inclusion and as such an appropriate instrument for the operationalisation of the capability approach in Germany (Deutscher Bundestag, Citation2005, p. 12). This statement made the AROPR the only official poverty measure in Germany.
However, two very restrictive assumptions have to be fulfilled if the AROPR is indeed to be an adequate instrument for the operationalisation of the capability approach. Every person in a society must have the same access to perfectly functioning markets and face the same individual conversion factors as anyone else. In other words, every person in a society must have the same ability to transform economic resources in whatever he or she has reason to value in life. These assumptions disregard, in particular, the existence of public goods and the whole range of personal heterogeneities that influence the ability of individuals to convert economic resources into whatever they may need and want (Sen, Citation2009, p. 255).
Thus, it is not at all straightforward to conclude that the AROPR is a satisfactory instrument to operationalise the capability approach and it is the purpose of this paper to analyse if such a conclusion is indeed justified. In order to do so, the paper develops two multidimensional poverty indices in order to operationalise the capability approach in the German context. The German Correlation Sensitive Poverty Index (GCSPI) measures poverty in Germany according to objective standards, while the Subjective Correlation Sensitive Poverty Index (SCSPI) is based on the self-assessment of respondents.
A thorough poverty analysis is conducted for the period 2002–2010, including dimensional and regional decompositions, in order to compare the differences in the application of the AROPR, the GCSPI and the SCSPI. The empirical analysis demonstrates that the different poverty measures arrive at significantly different assessments of poverty and poverty trends in Germany. The results provide empirical evidence that the assumption that the AROPR alone is a satisfactory instrument to operationalise the capability approach in the German context is unjustified and that high additional value is generated in case the AROPR is complemented by other, multidimensional poverty indices.
The paper proceeds as follows. Section 2 introduces the two multidimensional poverty measures GCSPI and SCSPI, in particular the choice of dimensions, indicators, thresholds and weights. The empirical application is presented in Section 3, Section 4 concludes.
2. Two Multidimensional Poverty Measures
2.1 Theoretical Background
The methodology that provides the foundation for GCSPI and SCSPI has been developed by Rippin (Citation2012, Citation2013). The author introduced a new identification method that builds on a sort of fuzzy approach to identify the poor. In particular, a multi-step function differentiates between different degrees of poverty severity. This function is always non-decreasing in the number of deprivations; however, the marginal increase in poverty severity is lower the higher the substitutability between attributes is. As Rippin (Citation2012, Citation2013) demonstrated, the new identification method can be utilised to derive a class of multidimensional poverty indices that is unique in the sense that it is fully decomposable according to dimensions (factor decomposability) and subgroups (subgroup decomposability) and yet sensitive to inequality increasing switches:
In other words, any person who is deprived is considered to be poor. However, the degree to which persons are considered poor depends on the number of dimensions in which they are deprived as well as the correlation between these dimensions. In the following, I will differentiate between three categories of poverty degrees:
Category 1: Deprivation affected. Individuals are classified as deprivation affected whenever the weighted sum of their deprivations is one-third or less. It is important from a policy perspective to have them on the radar in order to ensure that their situation does not further deteriorate. However, no action has to be taken at that level. Thus, whenever I will compare different headcounts in the following empirical analysis, I will only concentrate on the headcount of those who belong either to Category 2 or 3. | |||||
Category 2: Poor. Individuals are classified as poor whenever the weighted sum of their deprivations is higher than one-third but not higher than two-thirds. | |||||
Category 3: Severely poor. Individuals are classified as severely poor whenever the weighted sum of their deprivations is higher than two-thirds. These are the poorest of the poor whose capability set is limited in such a way that it is almost impossible for them to pursue their goals in life. They are the ones who should be high on the political agenda. | |||||
For the empirical analysis, I will utilise the following simple variant of the class of indices characterised in Equation (1):
The final form of the poverty measure that is calculated in two subsequent steps is the square of weighted deprivations. In other words, the calculations are based on the assumption that (i) poverty dimensions are (weak) substitutes and (ii) inequality is an important aspect of poverty, i.e. persons with a higher number of deprivations receive a higher weight in the calculations. Thus, the index is in a way an equivalent to the poverty gap ratio in the one-dimensional case in the sense that the former accounts for inequality between different dimensions and the latter for inequality within a single dimension.
Having thus introduced the mathematic formula for the poverty calculations, the following section deals with the selection of dimensions, indicators and weights.
2.2 Choice of Poverty Dimensions
I will utilise the German Socio-Economic Panel (GSOEP) in order to identify poverty dimensions specifically for the German context. The GSOEP is a representative longitudinal panel data-set collecting socio-economic information at the household level in Germany since 1984. After the German reunification in 1990, the data-set has been expanded in order to cover the former German Democratic Republic. The survey is repeated annually with every adult in a household aged 17 years or older being surveyed (Wagner, Frick, & Schupp, Citation2007). The latter property provides the rare opportunity to measure poverty at the individual and not only at the household level. Thus, only the responses for adults are utilised for the calculations. At the same time, the existence of missing values in one or more of the chosen indicators was countered with the removal of the whole observation. This treatment led to a considerable reduction in the final sample size.Footnote3
In order to identify central functionings that are necessary to pursue whatever one has reason to value, I start with the theoretical approach of Nussbaum (Citation2003) which is typically considered to be the most influential and thorough operationalisation of the capability approach developed so far. The philosopher draws heavily on the work of Aristotle in proposing the following list of 10 ‘central human capabilities’ (Nussbaum, Citation2003, pp. 41–42): life; bodily health; bodily integrity; senses, imagination and thought; emotions; practical reason; affiliation; other species; play; control over one's environment.
Such a theoretically derived list can of course never have universal validity and can only serve as a useful guide. Nevertheless, it receives additional legitimacy by the fact that the same list served as a basis for the roundtable discussions of public advisors and scientific experts involved in the development of the German Poverty and Wealth Report (Arndt & Volkert, Citation2007). Thus, when selecting dimensions and indicators, I will refer to these capabilities to the extent that available data allows.Footnote4
Regarding the respective thresholds, I follow two different approaches. For the GCSPI, I will utilise absolute thresholds, i.e. certain minimum standards that are derived from laws, regulations and alike. For the SCSPI, I will utilise subjective thresholds that are based on questions regarding individual satisfaction. The latter comes closest to Aristotle's concept ‘Eudaimonia’ (Bartlett & Collins, Citation2011, p. x), i.e. long-term happiness that is considered the most adequate for evaluation.
The first suggested dimension is health and captures mainly capability number two, i.e. ‘bodily health’. However, it influences a lot of other capabilities as well. Suffering from bad health limits a person's capability to participate in social life, negatively influences emotions and might even prevent the person to practise his or her occupation. Also, those concerned would typically need more money than their fellow citizens as they are often forced to invest considerable amounts of money in medical treatment.
Two indicators are used to capture this dimension, bad health condition and severe health impairments. Admittedly, these indicators are subjective rather than objective indicators. The reason why I made an exception from my own rule to use solely objective indicators is due to a new research direction in anthropology that argues that health is first and foremost a matter of self-evaluation. For if an individual claims to feel bad, i.e. to suffer impairments, to feel pain, etc., who can by any means claim this self-evaluation to be wrong?Footnote5 Thus, bad health condition is based on the self-evaluation of respondents on a scale ranging from one to five (i.e. ‘very good’, ‘good’, ‘satisfactory’, ‘poor’, ‘bad’). Anyone considering his or her health status to be either ‘poor’ or ‘bad’ is considered to be deprived according to this indicator. Severe health impairments is also based on a self-evaluation of respondents according to the following five impairments: (i) has trouble climbing stairs, (ii) health limits vigorous activities, (iii) achieved less due to physical health condition, (iv) achieved less due to mental health condition and (v) reduced social contacts due to health problems. Anyone who reports to suffer from at least four of these impairments is considered to be deprived according to this indicator.
The ‘subjective’ indicator is a self-evaluation of the health condition of the respondents on a scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied). The chosen threshold level for this (and all other subjective dimensions) is 3, i.e. considerably dissatisfied.
The second suggested dimension is education and captures mainly capability number four, i.e. ‘senses, imagination and thought’, but has a distinct influence on other capabilities as well, such as occupational choice and future income, but also on emotional issues like self-confidence and the ability to engage in social interaction. Education surely correlates with household income, but also with the quality of the education system as regulated by the individual Bundeslaender, the availability of training positions, parental support and promotion, personal effort and others. Thus, income alone does not seem to be an appropriate indicator to capture this dimension.
Two objective indicators are used to capture this dimension. The first one is school drop out, the respective threshold is based on compulsory schooling in Germany—which is either 9 or 10 years of schooling, depending on the respective Bundesland. Thus, any person who dropped out of school with < 9 years of schooling is considered deprived according to this indicator. The second indicator is no graduation or training qualification. That indicator captures the aspect that a person might have spent 9 years in school but never graduated. Or, even in case a person was able to complete secondary education, he or she might not have received any further training qualification that is part of the German dual education system and would be needed in order to enter the labour market. Thus, any person who left school without graduation and/or training qualification will be considered deprived according to this indicator.
Unfortunately, the GSOEP does not include a self-evaluation with regard to education; thus this dimension cannot be included in the SCSPI.
The third suggested dimension is employment and directly captures capabilities number 7, 9 and 10, i.e. ‘affiliation’, ‘play’ and ‘control over one's environment’. Employment is a lot more than just a source of income. Indeed, a recent study analysing poverty in Europe finds ‘evidences that income sources and socio-economic endowments, and not only income level, matter for the individual well-being’ (Figari, Citation2012, p. 416). This finding is strongly connected with issues such as self-respect. Though there are of course those who seemingly enjoy the fact that they do not need to work, there are also many persons who would even sacrifice money in order to be able to claim that they earned what they have. These and related arguments are the reason why income alone does not seem to be an appropriate indicator to capture this dimension. Three indicators are utilised to capture this dimension: (i) ‘main personal activity status unemployed’, (ii) ‘working poor’ and (iii) ‘time poor’.
First, I consider all those as deprived who are registered as seeking employment but whose main personal activity status over the year has been unemployed. However, to consider only unemployment within the employment dimension falls way too short. For instance, such a minimalist approach would mask working poverty. Working poverty is of increasing importance for Germany, due to a rapid increase of short-time work (‘Kurzarbeit’) and temporary work (‘Leiharbeit’ or ‘Zeitarbeit’) in the course of the economic crisis (Faik, Citation2012, p. 6). This is a precarious situation as short-term and temporary workers earn considerably less for the same type of work than ordinary staff, leaving many dependent on additional social security benefits (Burmeister, Citation2012, p. 4). Not only is this detrimental to the aforementioned principle of self-respect; in effect, the state subsidises the low wages, thereby contributing to poverty in old age. I will utilise a minimum wage in order to capture this problem. One might argue that a minimum wage is too closely related to the minimum income level that is used as another dimension in GCSPI and SCSPI (see pp. 13–14). However, the correlation between the two indicators is much lower than one might expect at first sight, with a Kendall τb value of merely 0.2065. There are a couple of reasons for this, the main ones are that the minimum income level refers to the household level and as such does not only include the wages of other household members but also social security benefits and debt payments. In other words, the income dimension measures how well a household is able to get along financially whereas the minimum wage captures the precarious situation of those whose work is not appropriately remunerated.
As Germany does not have an official minimum wage, I draw on a suggestion of the Hans Böckler Stiftung and utilise the limit of exemption from execution in order to derive an appropriate minimum wage (Böckler Impuls, Citation2006, p. 1). Until 1 July 2011, this limit was 989.99 EUR, an amount that can be easily translated into a minimum wage of 8.29 EUR per hour.Footnote6 In order to ensure the comparability of this amount over time, the value is indexed by the consumer price index (CPI) (base year 2010).
The third component of the objective deprivation indicator is ‘time poverty’. Especially from a capability perspective, the ability to have a sufficient amount of leisure time is an elementary precondition for the ability to participate in social life. The Monitor Familienleben (BMFSFJ, Citation2012) highlights the importance of the topic in the German context. The majority of parents with children under the age of 16 expressed their desire to be able to spend more (45%) or even much more (28%) time with their families. Only 23% of the respondents declared to have sufficient time for their families. I account for this fact in the following way.
The so-called working hour tension captures the disparity between actual and preferred working hours and explicitly accounts for income changes induced by changes in working hours (Merz, Citation2002). The indicator is based on the following questions in the GSOEP: (i) actual weekly working hours and (ii) preferred weekly working hours, taking the fact into account that income changes with working hours. Working hour tension is then calculated as the difference between actual and preferred working hours. Individuals are considered time poor if their working hour tension is 10 hours or more.
The subjective indicator is a self-evaluation of the working condition of the respondents on a scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied). The chosen threshold level is 3, i.e. considerably dissatisfied.
The fourth dimension is housing, directly capturing capabilities number two, ‘bodily health’ and seven, ‘affiliation’. Of course the requirement to have adequate shelter is especially important from the perspective of the ‘affiliation’ capability. In fact, it is one of the perspectives in which the capability approach is most convincing. In order to be able to participate in the social life of the community, a certain minimum standard of living is absolutely necessary, what Smith (Citation1776, pp. 466–467) described as the ability to appear in public without shame. In order to capture this aspect, I follow the suggestions of Arndt and Volkert (Citation2007) and use the following three objective indicators: ‘unacceptable housing’, ‘lack of socially necessary amenities’ and ‘overcrowding’.
Anyone whose housing is characterised as either ‘in urgent need of complete renovation’ or ‘in danger of breaking down’ is considered deprived according to ‘unacceptable housing’ (Arndt & Volkert, Citation2007, p. 28). Persons are identified as deprived according to a ‘lack of socially necessary amenities’, if they lack either of the following: ‘in-house bath/shower’, ‘in-house toilet’, ‘warm water’, ‘central heating’ (Arndt & Volkert Citation2007, p. 28). Persons are considered to suffer from ‘overcrowding’ if their living space is below what was, at least until 2010, the appropriate living space for those receiving welfare payments under the so-called Hartz-IV-scheme: 45 m2 for the first and 15 m2 for every additional person (infants below 2 years of age excluded).
The subjective indicator is a self-evaluation of the housing condition of the respondents on a scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied). The chosen threshold level is 3, i.e. considerably dissatisfied.
The fifth dimension is mobility, capturing basically capabilities three, ‘bodily integrity’, and seven, ‘affiliation’, as mobility is a precondition for the ability to participate in social life. Mobility can be restricted due to (i) limited access to transportation and/ or (ii) an insecure environment.Footnote7 The first aspect is captured by the inability to afford a much-needed car. In cities, mobility is usually ensured by public transportation systems; however, this system might not be very well developed outside of towns. Thus, persons are considered deprived if they (i) live in a household without a car and (ii) public transport is more than 20 min away.
The second aspect is captured by the indicator insecure environment. Persons are deprived according to this indicator if they live in an environment that is classified as either ‘insecure’ or ‘dangerous’.Footnote8
Unfortunately, the GSOEP does not include a self-evaluation with regard to this dimension, thus it cannot be included in the SCSPI.
The sixth and final suggested dimension is income, capturing directly capability number 10, ‘control over one's environment’. Though income is definitely not the only indicator for poverty measurement, it is obviously an important one. Persons are income deprived if their disposable (i.e. after debt service) household income is below the official German breadline as defined in the seventh Existenzminimumbericht (breadline report) for 2010, i.e. below 638 EUR for single persons; 1,083 EUR for couples and 322 EUR for each child. As the issue of additional persons in the household is not captured by the report, I utilise the difference in the amount allowed to single persons and to couples, i.e. 356 EUR, for each additional adult in the household. In order to ensure the comparability of the breadline over time, the value is indexed by the CPI (base year 2010). Please note that the breadline figures are considerably below the threshold levels of the AROPR.
The subjective indicator for this dimension is a self-evaluation of the income situation of the respondents on a scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied). The chosen threshold level is 3, i.e. considerably dissatisfied.
Table provides an overview of the different dimensions, indicators and thresholds that are used for the empirical analysis of multidimensional poverty in Germany.
Table 1 Dimensions, Indicators and Thresholds for Germany.
Once dimensions, indicators and threshold levels have been chosen, the next exercise concerns the choice of weights for dimensions and indicators. Several options can be applied in order to choose the weights. As far as the dimensions are concerned, I utilise equal weights—i.e. each dimension contributes to overall deprivation in the same way—in the light that these dimensions are all directly derived from Martha Nussbaum's list of central human capabilities. Thus it seems somewhat inappropriate to utilise different weights for them—at least as long as no participatory approach is available that would provide a convincing basis for a deviation. I apply the same approach when it comes to the choice of weights for the indicators, i.e. indicators of the same dimension receive the same weight.
However, in order to test the robustness of the results, all results were also calculated with the prevalence or frequency-based weighting approach, which is becoming increasingly popular (Figari, Citation2012, p. 8).Footnote9 Spearman rank correlation coefficients range from 0.9984 (2006 and 2010) to 0.9986 (2004) considering all observations, and from 0.9933 (2010) to 0.9947 (2004) considering only those observations for which the index is larger than zero. Also, the respective ranking of the Bundeslaender is the same for both weighting methods with only very few minor exceptions.Footnote10 All results are available from the author upon request.
In the following section, a thorough poverty analysis is conducted with the three poverty measures AROPR, GCSPI and SCSPI. The analysis shows a positive correlation between the income level of the AROPR and all other deprivation levels; however, it is far less stronger than might be expected. In addition, the three poverty measures differ considerably in the impact of location, especially regarding Western and Eastern Germany, which may have considerable implications for targeting. Thus, the empirical analysis highlights the value that is added if the traditional income-based AROPR is complemented by multidimensional poverty measures.
3. Empirical Application
The following analysis addresses the question whether and, if so, in how far differences in the way we measure poverty do indeed make a difference ‘on the ground’. In other words, when applied to real data, do the evaluations of poverty and poverty trends really differ? Is there indeed a need to have other poverty measures in addition to the AROPR?
For a start, the following table provides the results of the statistical correlations between the poverty dimensions of GCSPI, SCSPI and AROPR based on the respective Kendall τb correlations for 2010 (number of observations: 8,555) (Table ).
Table 2 Kendall τb Correlation All Dimensions.
All variables show positive values indicating that they all measure the same, i.e. poverty. A significant positive value of Kendall τb is a sign for a positive correlation between two variables. With one exception, all of them are statistically significant; only the correlation between housing and dissatisfaction with one's health condition has a p-value of >0.05 and is thus statistically insignificant. The results reveal the connection between the different aspects of poverty. At the same time, Kendall τb is considerably lower than 0.80 in all cases except for the correlation between the breadline and AROPR, already indicating that each dimension measures a distinctively different aspect of poverty. This supports the fact that, though income is of course correlated with other poverty dimensions, it is not an equally good proxy for all of them.
It is interesting to note that the strongest correlation exists between the different dimensions of dissatisfaction. It might point to a problem with subjective evaluation: a person that is dissatisfied in one dimension is rather likely to transfer this dissatisfaction to other dimensions as well. Imagine, for instance, a person suffering from a bad health condition or a bad mental condition. It seems rather likely to suggest that the overall bad feeling of the person is not only reflected in questions directly related to health but as well with regard to any other dimension. This would explain why the correlation between the two dimensions health and housing is rather low when measured in absolute terms, as would be expected, whereas the correlation between dissatisfaction with health and dissatisfaction with housing is comparatively strong.
The question that inevitably comes to mind considering the weak correlation between the dimensions of the three poverty measures is whether and, if so, how far this affects poverty measurement. In order to get a more dynamic impression of how the different approaches to measurement may differ, the following figure compares them over time and within the context of the development of the most important figures for the German economy. Please note that in this figure the absolute values of GCSPI and SCSPI are illustrated and not just their headcounts (Figure ).Footnote11
The first thing to mention is the obvious volatility of the SCSPI over time, just as one might expect in the case of subjective poverty measures. It seems that the SCSPI is better placed to reflect short-term changes than overall trends. The other two poverty measures on the other hand, i.e. AROPR and GCSPI, seem to be better placed to provide overall trends. However, the trends they indicate diverge during an interesting time period, i.e. between 2002 and 2004, which captures the crisis of the German economy in 2003.
As Faik (Citation2012, p. 8) points out, income inequality decreases during times of crisis, followed by an increase as soon as the economy recovers. It is a typical outcome of a social welfare state whose social security system cushions the effect of the economic crises for the poorest parts of the population whereas the wealthiest parts typically experience its full force. In order to get an impression of the development of income inequality over time, I included the development of Theil's enthropy measure over time in the figure:
As described by Faik (Citation2012, p. 8), income inequality decreased during the economic crises in 2003 and 2009 and increased immediately afterwards as the economy recovered. However, despite the same trends in income inequality during the time periods 2002–2004 and 2008–2010, there are two important differences. During the first recession in 2003, the percentage change of the social budget was reduced (from 3.5% in 2002 to 2.1% in 2003) whereas it was significantly increased during the second recession in 2009 (from 2.3% in 2008 to 7.1% in 2009).Footnote12 Also, the unemployment rate happened to be higher after the first crisis, increasing from 9.8% in 2002 to 10.5% in 2004, whereas it was slightly lower after the second crisis, decreasing from 7.8% in 2008 to 7.7% in 2010. Whatever induced these differences, they provide a rather unique opportunity to compare the behaviour of the AROPR and the GCSPI over time.
During the first economic crisis (2002–2004), rising unemployment and the decline in the percentage rise in social benefits, in addition to other implications of the crisis, provide a more than convincing explanation for the noticeable increase in the GCSPI, which includes indicators for unemployment as well as income. The AROPR, however, fell slightly during the same time period. This can only be explained if the trend in inequality is taken into account: the loss of the wealthier parts of society has been stronger than the loss of the poorer parts, causing a reduction in the income inequality of the society as a whole. To put it more plainly, the AROPR decreases not because of an improvement in the living conditions of the poor but because the deterioration in the living conditions of the poor was weaker than the deterioration in the living conditions of the wealthy (Faik Citation2012).
In the case of the second, much more severe, economic recession—and completely different from the first—the AROPR increases, reflecting again the development of overall inequality in society, which is in 2010 almost as high as in 2008. The increase of the GCSPI, however, is hardly noticeable, capturing the compensational effect of a slightly decreasing unemployment rate between 2008 and 2010 and a strong increase in the percentage change in the social budget in 2009.Footnote13
I will commence the analysis with an illustration of the usefulness of the regional decomposability of the three poverty measures. The question is whether there are inequalities across regions.Footnote14
Figure provides the deprivation headcounts (Categories 2 and 3) for GCSPI, SCSPI and AROPR for the German Bundeslaender. All observations have been dropped that did not provide enough information to calculate all three poverty measures, thus all three calculations are based on the same sample of 8,555 individuals.
The first thing that immediately strikes the eye is the considerable discrepancy between West and East Germany. All three poverty measures yield comparably higher values for the Eastern Bundeslaender but their relevance for the different parts of Germany is rather different. The AROPR plays a much stronger role in the Eastern Bundeslaender and, while GCSPI and SCSPI are almost equal in the Western Bundeslaender, the SCSPI plays a minor role in the Eastern Bundeslaender. In fact, the figure almost acts like a mirror reflecting these two different trends. The fact that the SCSPI is almost as high in the Western as in the Eastern Bundeslaender despite the higher values of GCSPI and AROPR in the latter shows once more the need to be cautious in interpreting subjective poverty measures. It is rather likely that the comparatively high values of the SCSPI in the Western Bundeslaender are due to adaptive preferences as people in these regions always enjoyed higher living standards than in the Eastern Bundeslaender.
Are these differences stable or did they evolve over time? The following figures provide poverty maps for Germany with regard to the AROPR, the GCSPI, and the SCSPI from 2002 to 2010 (Figures ).
All poverty indices tell the same story: though there has only been a slight increase in overall poverty over time, a concentration process seems to have taken place that makes poverty more and more a problem of East Germany.
Considering the overall trend that the regional decomposition made visible, the first question that vies for attention is whether a similar overall trend can be detected with regard to dimensional decompositions. As has been pointed out before, the GCSPI as well as the SCSPI can be decomposed according to poverty dimensions—despite the fact that they are sensitive to inequality. Figure illustrates the development of the dimensional decompositions of GCSPI and SCSPI over time.
The high overall volatility of the SCSPI is reflected in the dimensional decomposition. Clear trends in the contribution of the different dimensions are often not recognisable. One exception is the contribution of the dimension dissatisfaction with employment to overall poverty that shows a continuous decrease from 29% in 2002 to 27% in 2010. The GCSPI, on the other hand, shows some clear trends. There exists a slight increase in the contribution of income, from 13% in 2002 to 15% in 2010, and a decrease in the contribution of education, from 23% in 2002 to 20% in 2010. In addition, the contribution of employment shows a strong increase in the years 2008 and 2010, from 26% in 2002 to 31% and 30%, respectively. These trends gain additional significance if they are compared with a dimensional decomposition according to regions as illustrated by Figure .
The figure indicates two regional peculiarities. It seems that in Eastern Germany income deprivation is a much stronger contributor to overall deprivation compared to Western Germany, as already indicated by the stronger importance of the AROPR for this region (Figure ). In Western Germany, on the other hand, education seems to contribute more to overall deprivation than in Eastern Germany. This corresponds to the two different types of support programmes of the European Social Funds in Germany. The Eastern Bundeslaender—with the exception of Berlin—belong to the ‘Convergence Regions’, i.e. regions that are characterised by weak economic performance. The objective of the respective support programs is to speed up gross domestic product growth in those regions so that they are able to catch up with the rest of Germany. The Western Bundeslaender belong—without exception—to the ‘Competitiveness and Employment Regions’. The objective of the support programmes in those regions is to promote life-long learning, training and to improve the reconciliation of work and family life.
If this regional peculiarity is associated with the trends in the dimensional decomposition it reveals two important sources for the pronounced drift apart in East and West German poverty levels: the declining significance of educational deprivation, which has been a great contributor to poverty in West Germany, combined with the increasing significance of income deprivation, which has been a great contributor to poverty in East Germany.
By far not all stories have been told, the attempt would have gone well beyond the scope of this paper. However, the stories that have been told all get to the same conclusion: the operationalisation of the capability approach by means of GCSPI and SCSPI seems to be worth the effort. Both poverty measures are able to capture aspects and trends of German poverty that the AROPR is unable to capture.
4. Conclusion
With the publication of the second ‘Poverty and Wealth Report’ the German government adopted the decision to define poverty in Germany on the basis of the capability approach. However, in the same document the government concluded that the purely income-based AROPR is an indirect indicator for social inclusion and as such a satisfactory instrument for the operationalisation of the capability approach in Germany.
This paper studies the question whether this conclusion is indeed justified. In order to do this, the paper introduced two multidimensional poverty measures to operationalise the capability approach: the GCSPI and the SCSPI.
The analysis revealed that, though a positive correlation exists between the AROPR and all other deprivation levels, it is nevertheless far less distinct than might be expected. This observation is backed by a thorough empirical analysis that compares the poverty evaluations of the three poverty measures over time. The three measures differ considerably with regard to poverty trends, the identification of the most deprived and the impact of location, especially regarding West and East Germany, which may have considerable implications for targeting.
All results point in the same direction, namely that the conclusion that the AROPR suffices to analyse poverty as defined by the capability approach is false. Other, multidimensional poverty measures are needed in order to complement the AROPR as they are able to provide information about poverty that traditional income-based poverty measures are unable to discover. Thus, further research in this area could provide very valuable insights with regard to poverty in Germany; all the more so as specific capability-related questions will be available in the GSOEP for the first time ever.
Acknowledgements
I am very grateful to Stephan Klasen, Milorad Kovacevic, Peter Krause, Sebastian Vollmer and an anonymous referee as well as the research staff at the German Institute for Economic Research (DIW) for valuable comments and suggestions and thought-provoking discussions. Thanks as well to the participants of the Human Development and Capability Association annual conference 2012 (Jakarta) for helpful comments and suggestions.
Notes
1 The rate is defined as the percentage of the population with a net equivalence income below 60% of the median. The concept of the net equivalence income accounts for the fact that bigger households have saving opportunities through the joint use of household items. Therefore, the new scale of the organisation for economic co-operation and development attributes a weight of 1 to the first adult, a weight of 0.5 for every additional person aged 15 or over and a weight of 0.3 for persons below the age of 15. Thus, the net equivalence income is the household's net income divided by the weighted sum of household members.
2 A function has a non-decreasing marginal if
whenever
.
3 Tables A1 and A2 in the Appendix provide a detailed overview of the missing values for the different dimensions and indicators of the GCSPI and SCSPI. Note that this reduction in the sample size is one of the main reasons for the discrepancy between the AROPR as calculated in this paper and those that are officially reported in the German poverty reports. The other main reason being the fact that official calculations utilise the German Microcensus instead of the GSOEP.
4 For instance, though the operationalisation of capability one would be desirable and ‘life expectancy’ would be a good indictor to capture it, the GSOEP does not provide enough information to calculate such an indicator.
5 There has been a lot of discussion about the best way to capture individual health conditions that is due to a new research direction in anthropology initiated by Arthur Kleinman and others (Kleinman, Eisenberg, & Good, Citation1978; Kleinman, Citation1988; Sen, Citation2009). The experts strongly criticise the traditional way of utilising health statistics to evaluate health in a society. Instead, Kleinman (Citation1988, p. 3) defines illness as ‘the innately human experiences of symptoms and suffering’ that has to be captured by patient interviews. The questions he proposes for this self-evaluation have become known as Kleinman's Questions. Considering the strength of arguments and the fact that this is the current state-of-the-art approach in anthropology, I decided to make an exception and use subjective instead of objective indicators.
6 This is a monthly wage of 1,370 EUR (based on a 38-h work week) which is higher than the official minimum wage of the UK (1,202 EUR), but considerably lower than the official minimum wages of France (1,398 EUR), Belgium (1,444 EUR), Netherlands (1,447 EUR), Ireland (1,462 EUR) and Luxembourg (1,802 EUR).
7 Note that a bad health condition could also limit a person's mobility and could therefore be included as a third indicator. I nevertheless refrain to do so in order to prevent double-counting as a bad health condition is already included in the health dimension.
8 Please note that a disadvantage of this indicator is that the GSOEP retrieves this information only every 5 years. Its inclusion thus creates missing values whenever there is no information about a person's whereabouts in the years for which the respective question was not included in the survey.
9 The prevalence or frequency-based weighting approach weights each indicator in dependence of the proportion of the individuals in the population who are not deprived in that indicator at each point in time. The higher the proportion of those who are not deprived in a given indicator, the higher is the weight assigned to it. The reasoning behind this approach is that the lower the likelihood that a person is deprived in an indicator, the more he or she has reason to feel deprived. Thus, the higher weight acknowledges the stronger indicative nature of this specific indicator with regard to deprivation. Moreover, as prevalence weights are calculated for each point in time, this weighting approach is able to account for a situation in which the condition of a person does not change while the condition of the society as a whole improves.
10 When changing from the equal weighting approach to prevalence weighting, there are four minor rank changes, one in 2002 (Rhineland–Palatinate, initially rank 6, switches places with Berlin, initially rank 7); two in 2004 (Schleswig–Holstein, initially rank 1, switches places with Hamburg, initially rank 2; Saarland, initially rank 14, switches places with Berlin, initially rank 15) and one in 2010 (Saxony–Anhalt, initially rank 13, switches places with Saxony, initially rank 14).
11 I refrained from using only the headcounts of the GCSPI and the SCSPI in order to make use of the interesting properties of the two indices which are a product of poverty incidence (as measured by the headcount), poverty intensity (as measured by the aggregate deprivation count ratio) and inequality among the poor (as measured by the Generalised Entropy of deprivation counts) (Rippin, Citation2012, Citation2013).
12 The percentage reduction in the social budget in the course of the economic recession of 2003 is unusual but might be due to the fact that there had been a rather strong increase in the previous year when the social budget was raised from almost 662 billion in 2001 to more than 685 billion in 2002—maybe an election gift.
13 Other interesting stories could be told from the figure, for instance with regard to the steep increase in income inequality from 2005 to 2006, the year in which a new set of rules for the long-term unemployed and social welfare assistance was introduced, the Hartz IV regulations, only to name one. To tell them all, however, would go beyond the scope of this paper.
14 Note that a comparison by Bundesland is impossible due to the fact that the GSOEP is not representative at Laender level. A comparison of East and West Germany is, however, possible and provides interesting results.
References
- Arndt, C., & Volkert, J. (2007). A capability approach for official German Poverty and Wealth reports: conceptual background and first empirical results. Institute for Applied Economic Research (IAW) Working Paper No. 27. Tübingen: University of Tübingen.
- Bartlett, R. C., & Collins, S. D. (2011). Aristotle's Nicomachean ethics: a new translation. Chicago, IL: The University of Chicago Press.
- BMFSFJ. (2012). Monitor Familienleben 2012 [Monitor family life 2012]. Berlin: Author.
- Böckler Impuls. (2006). Mindestlohn: die Linie nach unten [Minimum wage: the bottom line]. Böckler Impuls, 11, 3.
- Burmeister, K. (2012). Leiharbeit 2012: Zwischen gewerkschaftlichem Erfolg, unternehmerischen Ausweichmanövern und politischer Verantwortung [Temporary agency work 2012: between union successes, entrepreneurial evasive manoeuvres and political responsibility]. WISO direkt September 2012 (1–4). Bonn: Friedrich Ebert Foundation.
- Deutscher Bundestag. (2005). Lebenslagen in Deutschland: Der 2. Armuts- und Reichtumsbericht der Bundesregierung [The German Federal Government's 2nd report on poverty and wealth]. Berlin: Federal Ministry of Labour and Social Affairs (BMAS).
- Faik, J. (2012). Income inequality and poverty in front of and during the economic crisis—an empirical investigation for Germany 2002–2010. The Society for the Study of Economic Inequality (ECINEQ) Working Paper No. 255. Verona: University of Verona.
- Figari, F. (2012). Cross-national differences in determinants of multiple deprivation in Europe. Journal of Economic Inequality, 10, 397–418.
- Kleinman, A. (1988). The illness narratives: Suffering, healing, and the human condition. New York, NY: Basic Books.
- Kleinman, A., Eisenberg, L., & Good, B. (1978). Culture, illness, and care: Clinical lessons from anthropologic and cross-cultural research. Annals of Internal Medicine, 88, 251–258.
- Merz, J. (2002). Time and economic well-being—a panel analysis of desired versus actual working hours. Review of Income and Wealth, 48, 317–346.
- Nussbaum, M. (2003). Capabilities as fundamental entitlements: Sen and social justice. Feminist Economics, 9, 33–59.
- Rippin, N. (2012). Distributional justice and efficiency: Integrating inequality within and between dimensions in additive poverty indices. Courant Research Centre (CRC) Discussion Paper No. 128. Göttingen: University of Göttingen.
- Rippin, N. (2013). Considerations of efficiency and distributive justice in multidimensional poverty measurement (Ph.D. dissertation, Department of Development Economics, University of Göttingen).
- Sen, A. (1979). Equality of what?In The Tanner lectures on Human Values (pp. 195–220). Stanford: Stanford University.
- Sen, A. (1985). Commodities and capabilities. Oxford: Oxford University Press.
- Sen, A. (1987). The standard of living. Cambridge: Cambridge University Press.
- Sen, A. (1992). Inequality reexamined. Oxford: Oxford University Press.
- Sen, A. (1997). On economic inequality. Oxford: Oxford University Press.
- Sen, A. (2009). The idea of justice. Cambridge, MA: Harvard University Press.
- Smith, A. (1776). An inquiry into the nature and causes of the wealth of nations (Vol. II. Oxford: Clarendon Press. (1869).
- Volkert, J., Klee, G., Kleimann, R., Scheurle, U., & Schneider, F. (2004). Operationalisierung der Armuts- und Reichtumsmessung [Operationalising the measurement of poverty and wealth]. Final report for the Federal Ministry for Health and Social Security. Tübingen: Institute for Applied Economic Research (IAW), University of Tübingen.
- Wagner, G., Frick, J., & Schupp, J. (2007). The german socio-economic panel study (SOEP) – scope, evolution and enhancements. Schmollers Jahrbuch, 127, 139–170.
![Supercharge Your Next Research Paper: Research and write your next paper with Jenni AI.]({"type":"image" , "src" : "/sda/112445/Skyscraper-econ.png"})