Session Number: 6A
Session Title: "Measures of poverty and social exclusion"
Paper number: 5
Session Organiser: Stephen P. Jenkins
Discussant: Holly Sutherland 

Paper prepared for the 26th General Conference of
The International Association for Research in Income and Wealth
Cracow, Poland, 27 August to 2 September

"INCOME, MULTIPLE DEPRIVATION AND POVERTY:
AN EMPIRICAL ANALYSIS USING SPANISH DATA"

Authors:

Rosa Martínez
Jesús Ruiz-Huerta

For additional information please contact:
Rosa Martínez
Departamento de Economía Aplicada IV
Facultad de Derecho - Universidad Complutense
Ciudad Universitaria s/n
28040 Madrid, Spain
E-mail: rosamar@idecnet.com
FAX: (34) 91 394 56 91

Jesús Ruiz-Huerta
Departamento de Economía Aplicada IV
Facultad de Derecho - Universidad Complutense
Ciudad Universitaria s/n
28040 Madrid, Spain
E-mail: jrhuerta@eucmax.sim.ucm.es
FAX: (34) 91 394 56 91

This paper is placed on the following websites: www.stat.gov.pl
                                                                       www.econ.nyu.edu/dept/iariw

 ABSTRACT¹ 

This paper analyses the relationship between income and non-monetary indicators of poverty in Spain, using information contained in the European Community Household Panel (ECHP). Along the lines followed by other studies, we use direct standard of living indicators to construct deprivation indices and examine to what extent such measures could be combined with income to better identify the poor (defined as those having a low level of living due to a lack of resources). After an introduction, the first part of the paper briefly reviews the main questions raised by the use of direct indicators in the measurement of poverty and explains the methodology employed to deal with the information contained in the ECHP and calculate deprivation indices for Spain. In the last part, we examine the degree of overlap between low-income and high-deprivation groups and explore low income and deprivation profiles. Finally, we review our main results and make some concluding comments.

 1. Introduction

Currently, the notion of relative poverty is widely accepted, although frequently criticised, in the developed countries. According to the most commonly employed definition of the term within the European Union, the poor are those individuals and families whose resources are so limited that they are excluded from the patterns of consumption and activities that make up the minimum acceptable standard of living of the society they live in. Thus, poverty should not be understood as the mere inability to satisfy the necessities that ensure physical survival, but rather as the involuntary lack of goods and exclusion from activities considered as necessary by society as a whole.
The concept of poverty as exclusion from a minimum acceptable standard of living is certainly appealing. Nonetheless, it also presents some important measurement difficulties (How should the standard of living be measured? What does an “acceptable minimum standard” mean? When can we say that a person falls below this minimum?). Most of these difficulties have often been side-stepped in empirical studies of poverty, which normally adopt an approach that defines the poor as those people whose income is located at a certain distance from the average or median income of society as a whole (60% of median income in the latest Eurostat reports). This approach is satisfactory in many ways, since it provides a clear and easily comparable series of common procedures which allow estimations of poverty to be constructed in different countries and time periods. However, it also entails a series of limitations which academic studies have examined in detail.
Apart from the accepted arbitrariness of purely relative poverty thresholds, an important debate on the indicators that should be employed to identify the poor has been continuing for some time. This article will concentrate upon a specific strand of this debate i.e. the use of direct and (usually) non-monetary indicators in poverty analysis. Although the idea of incorporating non-monetary variables into the measurement of poverty or inequality is not new², the number of studies including such criteria has grown rapidly only in recent years.

One reason for this increasing interest in direct non-monetary indicators is to be found in the controversies generated by the well-known criticisms of Ringen regarding the conventional approach to measuring poverty³. According to him, income (an indirect one-dimensional indicator) does not permit the correct identification of those individuals who have a low standard of living due to a lack of resources. He suggested that it was necessary to simultaneously employ both direct (or standard of living) and indirect (or resource level) indicators, in order for measurement procedures to be consistent with the theoretical definition. Regardless of the validity of the indicators proposed by Ringen, which have been subject to much discussion, his article has had the undeniable merit of generating a series of research studies focusing on the relationship between income and direct indicators of the standard of living.

Another important reason for the increasing interest in non-monetary indicators of poverty is the growing acceptance of the notion of social exclusion in the official discourse of the European Union. There has been much discussion in recent years on the reasons behind the change of focus from poverty to social exclusion as well as their differential traits and the relationship between both concepts. Though clear operative criteria for the measurement of social exclusion are still to be formulated, the multidimensional nature of the problems involved, the inclusion of non-economic aspects of welfare (such as health or social relationships) and the emphasis on explanatory factors other than economic resources are basic features of the concept which clearly call for the use of non-monetary indicators⁴.

Thus, both more accurate identification of the poor and the development of the concept of social exclusion are arguments in favour of an increasing use of non-monetary indicators. Several pieces of research using this kind of information have been published in the nineties, yielding a set of significant conclusions⁵. Nevertheless, in comparison to conventional methodology, methods that incorporate direct indicators into the measurement of poverty are by no means, as yet, adequately developed. Although some traditional results can be easily generalised to the case of multidimensional and non-monetary variables, the search for operative measures in this area poses new problems, without generally accepted solutions.

An important additional obstacle to the increased use of non-monetary indicators has been the lack of comparability of the results obtained by different studies, due to the absence of comparable data sources and indicators for the various countries. Fortunately, the creation of the European Community Household Panel (ECHP) in the mid-nineties offers new opportunities to overcome this problem.

This paper analyses the relationship between income and non-monetary indicators of poverty in Spain, using information contained in the ECHP. Along the lines followed by other studies, we use direct standard of living indicators to construct deprivation indices and examine to what extent such measures could be combined with income to better identify the poor (defined as those having a low level of living due to a lack of resources).

Empirically, this study provides additional empirical results to those already obtained for other countries such as the UK, the US, Ireland or Sweden. The interest of the results obtained is two-fold. On the one hand, they can be used to check to what extent the main conclusions reached by previous studies concerning the relationship between income and multiple deprivation are relevant for Spain, a country in which this sort of study is almost non-existent⁶. On the other hand, they can be employed to ascertain whether the use of direct indicators helps to better identify the poor. This can be especially useful in countries such as Spain, and others in Southern Europe, whose surveys (mainly household budget surveys) have tended to gather unreliable, low-quality information on income.

Although the empirical analysis refers to cross-section data for a single country, we have paid special attention to the comparability, over time and across countries, of the indices proposed. We firmly believe that the search for a common strategy is vital in this field, in order to produce in the near future much more comprehensive poverty statistics based on standard of living indicators in the EU countries. In this search for valid common procedures, a balance must be sought between the methodological refinement and practical applicability of the measures proposed. This paper tries to offer a pragmatic approach, though without neglecting the important conceptual and methodological issues involved.

In the first part of the paper we briefly review the main questions raised by the use of direct indicators in the measurement of poverty. Subsequently, we explain the methodology we employ to deal with the information regarding standard of living contained in the European Community Household Panel and calculate deprivation indices for Spain. In the following sections, we examine the degree of overlap between low-income and high-deprivation groups and explore low income and deprivation profiles. Finally, we review our main results and make some concluding comments.

Mainstream analysis of poverty, measured as low income (or low expenditure), has proved capable of developing an important body of theoretical results and an elegant set of refined methodological instruments. As is well-known, significant advances have been made during the last few decades in fields like the ethical content of different poverty indices, equivalence scales or sensitivity analysis of poverty estimates when faced with changes in underlying assumptions.

Given this state of affairs, it is permissible to ask why direct indicators should be included at all in the measurement of poverty. From an economist’s point of view, any possible advantages of moving away from the neat and well-structured world of one-dimensional monetary indicators to the rather complex and much less tractable field of standard of living indicators (typically non-monetary and multidimensional) would require convincing justification.

The employment of standard of living indicators in poverty analysis can be defended from different perspectives, which differ in their degree of departure from the conventional income-based methodology involved:

a) On purely descriptive grounds, direct indicators can provide a richer picture of what being income-poor means. We can accept poverty thresholds based on indirect indicators such as monetary income but, at the same time, be interested in the different kinds of deprivation suffered by families classified as poor. This information can help to complement the view offered by a purely relative and one-dimensional poverty approach (e.g. how do people with incomes below the poverty line really live in Spain, or Germany?), as well as to provide better guidance for the construction of social policies (e.g. what mix of social transfers, housing, education or labour market measures should be applied to enhance the standard of living of different groups?).

b) Secondly, direct indicators of standard of living could also serve to improve the process of identifying the poor, without completely leaving aside the income-based methodology. If it is believed that a strong relationship between income and the standard of living exists, direct indicators may still however be useful to investigate the precise level of income that would constitute a reasonable threshold (Townsend, 1979). Alternatively, this hypothetical relationship may be rejected to claim that a combination of income and standard of living indicators is needed to correctly identify poor people. The short time periods the income data refer to, the general exclusion of non-monetary components of the standard of living, the omission of wealth, the difficulties in dealing with household differences in terms of needs and the underreporting of resources are just some of the reasons which could explain the lack of a perfect relationship between income and standard of living.

c) Finally, direct indicators could in themselves constitute an alternative basis for the measurement of poverty. This could be justified if we defend an “standard of living approach” (following the distinction made by Atkinson, 1989). According to this view, poverty essentially consists of an insufficient quantity of food, housing, domestic appliances, leisure activities or any other area considered to be relevant for a particular society at any moment in time. The task of identifying the population groups suffering poverty would therefore require the elaboration of indicators which directly represent the standard of living, without taking into account the amount of resources enjoyed.

Generally speaking, the results produced by studies which compare direct and indirect indicators tend to show that the overlap between low-income and low standard of living groups is far from perfect. Nevertheless, some important questions remain unresolved when trying to define a reasonable strategy to incorporate direct indicators into the measurement process.

On the one hand, direct measurement of poverty poses difficult problems concerning the management of multidimensionality and non-monetary variables. This is a field in which, despite important recent advances, most economists may still feel uncomfortable. Obviously, standard of living indicators can be analysed one by one, without trying to aggregate the different items into a single indicator. In fact, such a method would provide a valuable description of data and be a useful preliminary step. Nevertheless, this non-aggregative strategy has the drawback of not yielding clear operative criteria to order people according to their overall level of deprivation. Thus, it would be useful to do a purely descriptive exercise, but not so much to identify the poor, a task that generally call for some kind of aggregation procedure. On the other hand, it is far from clear how income should be combined with (or even substituted by) direct indicators in identifying the poor. These questions are developed further in the next section.

One of the most common strategies to deal with multidimensional information on the standard of living is to construct deprivation indices. A deprivation index provides an aggregation of the unfulfilled needs experienced by each individual or household with respect to a series of items. These aggregated indices have the advantage of producing a complete order of individuals according to their overall deprivation score⁷, a useful step when attempting to use standard of living indicators in order to substitute or complement income data in poverty measurement.

Obtaining deprivation indices, however, poses a series of specific methodological difficulties. The main areas of debate are: 1) how to select the concrete set of indicators that should be considered; 2) how to evaluate item deprivation i.e. the position of a given household with regard to each individual indicator; 3) how to define a weighting structure, which determines to what extent each item contributes to overall deprivation; 4) how to aggregate indicators, using some kind of function, into one or several deprivation indices; and 5) how to define a threshold which separates the deprived and the non-deprived.

Note that steps 1) and 2) are always necessary in a strategy based upon direct indicators, while 3) and 4) are crucial if we want to aggregate disparate elements to obtain a value capable of synthesising the overall level of deprivation of each household. Finally, decisions regarding step 5) must implicitly or explicitly be adopted whenever we wish to use direct indicators (whether isolated or in combination with income data) to identify the poor. Although we do not try to perform an exhaustive analysis of these issues, we shall briefly consider each of them in the remaining of this section.

1) Selecting indicators

The selection of appropriate indicators poses a series of problems that range from delimiting the areas to be studied to deciding on the concrete variables to be utilised. The first question that must be tackled is that of whether indicators should be restricted to “material” living conditions⁸ or, on the contrary, they should also include variables concerning position in the labour market, education, health and other non material elements of wellbeing.

Of course, the underlying concept of poverty adopted shapes the answer to this question. But once the theoretical concept is agreed upon, we believe it is necessary to clearly differentiate between the constituent elements of poverty and other elements correlated to it yet not intrinsic to situations of poverty. Being unemployed or handicapped, for instance, can be causes or consequences of poverty, but they should not form part of a deprivation index according to the most widely accepted definitions of this concept⁹.

Secondly, there is no overall consensus on whether a wide set of variables representing material standard of living should be included in the list of indicators (a point of view which we shall call the “life-style approach”) or whether these should be restricted to those reflecting the coverage of necessities, in whatever way these may be defined (the “necessities approach”).

In the “necessities approach”, information concerning non-necessary goods coverage would be irrelevant, since the aim of the index is to identify people suffering deprivation in a set of explicit basic aspects of life. According to this view, whether a household lacks or not a “non-necessary” good does not affect the deprivation index. This is the approach followed, among others, by Mack and Lansley in 1985, when they decided to include only those items regarded as necessities by more than 50% of the population. Although this strategy would appear to be more clearly concerned with measuring poverty, its validity relies heavily on the fairness of the criteria applied to define necessities, a question particularly difficult to deal with when measuring relative poverty¹⁰.

The “life-style approach”, on the contrary, takes into account the household’s position concerning a wider series of variables which could also include items not generally regarded as necessities. In this case, the latent variable would be closer to the standard of living, instead of being multiple deprivation in necessary components of minimum living conditions¹¹. This approach has been defended by authors such as Halleröd, who argues that it eludes the difficult debate on exactly where to place the dividing line between “necessities” and “non-necessities”. Moreover, it avoids to a greater degree inconsistencies in the treatment given to differing structures of individual preferences (for example, relatively penalising those individuals whose preference ordering deviates from the average)¹². When making international comparisons, another advantage of the wider set of variables strategy could be the greater facility to establish a common series of indicators, while allowing some variation in the weights given to each item in the different countries.

Whatever the approach adopted, it must be admitted that the selection of concrete indicators to be included in deprivation indices is usually somewhat arbitrary. This is true even if a “consensual” approach is adopted (as in Mack and Lansley’s study¹³). Researchers must decide beforehand which are the questions to be formulated and the precise way they are to be posed. Obviously, procedures exist that allow one to minimise one’s own biases. Discussion groups can be a useful tool in this regard and have been successfully used in some recent research projects¹⁴.

Frequently, a different (and difficult to avoid) source of arbitrariness emerges from data restrictions. In practice, most measurements have to be made using indicators taken from surveys designed for purposes other than the analysis of poverty or social exclusion. These data restrictions are especially important when international and time comparisons are involved¹⁵. Here, it would be necessary to develop methods that permit robust conclusions to be reached from a set of incomplete and imperfect indicators of deprivation¹⁶.

2) Evaluating item-deprivation: dealing with differences in tastes

Evaluating the experience of each individual or household concerning the indicators selected also poses difficulties. In most empirical studies to date, the indicators are dichotomous variables which represent the possession or lack of a given good, or the participation in or exclusion from a given activity. Clearly, this binary classification offers only a rough approximation to the wellbeing resulting from different situations, since it neglects aspects such as the number of items possessed, their quality or their state of repair. This limitation is less important, however, if we consider that our objective is to detect states of cumulative disadvantage or deprivation, rather than to obtain a complete and accurate picture of the distribution of economic welfare.

An important problem raised by this approach is that of the need to determine when the lack of a certain good or the non-participation in a particular activity really implies deprivation. As is widely accepted, personal preferences and life-styles can influence the perception of needs and the set of goods and activities chosen using the resources available. Many of the studies performed take as a starting-point Mack and Lansley’s definition, which states that deprivation indices must reflect the enforced lack of socially-perceived necessities. This approach requires analysing why individuals lack such goods, meaning that only those who could not afford items classified as necessities would be counted as deprived.

Putting the approach described above into practice is not always possible, since in many cases information regarding the voluntary nature of such shortages is lacking. Even if this information were available, difficulties may surface concerning apparent inconsistencies between what people say they can afford and what in fact they do have. As Piachaud points out in his critique of Mack and Lansley’s work, it is not clear how the situation of a household should be evaluated if it affirms that it lacks resources to buy necessities but, nevertheless, enjoys access to goods that are not considered to be essential. At the other extreme, it is also possible to find people with a low level of resources who claim they do not need many of the basic goods they do not possess¹⁷, due to the lowering of expectations that tends to accompany prolonged situations of poverty.

The preceding paragraph touches on the debate about the validity of subjective evaluations in constructing deprivation indices. In our view, it is difficult to support the total exclusion of such subjective perceptions when we try to measure concepts like poverty, which are concerned, after all, with an insufficient level of individual economic welfare. In any case, much more research is still needed on the relationship between objective situations and subjective perceptions. While Mack and Lansley’s strategy is possibly the best one in the present state of affairs, it only represents, in our view, a provisional solution.

3) Weighting deprivation indicators

Once a concrete set of deprivation indicators has been selected, and procedures to evaluate the position of individuals within each of them have been established, another important step is to determine an adequate weighting structure. Given the different nature of the indicators normally used, do all of them have to carry the same weight when determining the level of deprivation? Is not having a refrigerator as important as having access to damp-free housing or being engaged in a hobby? Weights try, at last, to determine the importance we assign to the different shortages in the general deprivation state we want to evaluate. Additionally, they reflect the assumed trade-off among the diverse items in the overall deprivation level.

Studies such as Townsend (1979), Mack and Lansley (1985), Nolan and Whelan (1996), Gordon and Pantazis (1997) or Mayer and Jencks (1989) implicitly assign equal weighting to each element. Although this strategy has the advantage of avoiding the introduction of difficult value judgements on the relative importance of the various dimensions, there is, however, no clear theoretical justification for equal weighting. This is especially true when the list includes a wide range of indicators reflecting items whose degree of generalized ownership differs.

If we reject the validity of equal weighting, there remain several options, from the expert judgement of the researcher to the use of statistical procedures which attempt to derive weights from the available data. One of the most common strategies is to apply weighting systems which give more importance to the lack of goods considered necessary by larger groups of the population (for example, Halleröd (1994, 1995)) or, alternatively, to the goods which are most widely owned in a society (for example, Desai and Shah (1988))¹⁸. The first approach could be seen as an extension of the consensual method proposed by Mack and Lansley to cover not only the list of indicators, but also the evaluation of its relative importance in overall deprivation. The second would take instead the common patterns of observed consumption in society as the point of departure.

Although both strategies could be theoretically justifiable within the context of the relative approach to poverty, the consensual methodology has the advantage of more closely corresponding to social views on what a decent minimum standard of living really means. It also offers a more stable weighting structure (since the perception of needs will probably change more slowly than real consumption patterns). Nevertheless, the information required to know what items are socially perceived as necessities is neither always available nor easy to interpret. In particular, this strategy is not applicable to ECHP data, as we will see below.

4) Aggregating indicators: one or several deprivation indices?

Deprivation indices are elaborated by aggregating items using a suitable functional form, typically some kind of weighted sum of the individual item-deprivation scores obtained by the household.

One important question which has not been sufficiently discussed in previous literature is whether the aggregation of all the indicators into a single overall deprivation index is justified or, alternatively, if we must differentiate between various dimensions of the problem which have to be researched independently. Here it is clear that a choice must be made between the possible loss of information that aggregating heterogeneous aspects into an overall index implies and the more partial view produced by maintaining a separate analysis of the different dimensions of poverty.

Nolan and Whelan (1996) have convincingly defended the convenience of separating various dimensions (namely, basic, secondary and housing deprivation) in the analysis of poverty. They argued that those households suffering from problems in different areas are not necessarily the same in each case and that the processes generating each kind of exclusion depend on different conditioning factors. Bourguignon and Chakravarty (1998a, 1998b) have also emphasised the need to apply methods which allow a truly multi-dimensional poverty measure to be obtained. As they state, aggregating different indicators into an overall index could produce a mere redefinition of the conventional notion of poverty while maintaining its one-dimensional nature.

In our view, the problems regarding the best way to deal with the different dimensions of poverty are crucial when trying to implement a measurement strategy based on standard of living indicators¹⁹. Apart from detecting existing dimensions and the best way to measure deprivation in each field, we must decide how much they contribute to overall deprivation or how they should be included in the identification process. In identifying the poor, Nolan and Whelan (1996) used only “basic deprivation”, defined by a set of indicators reflecting some basic needs coverage and current financial difficulties, but their exclusion of certain items (e.g. durable goods possession or housing conditions) requires further and deeper discussion.

5) Defining a threshold

Finally, the use of deprivation indices to identify the poor poses enormous theoretical and empirical difficulties. The different strategies that have been applied to date can be categorised as follows:

a) Employing only resource-level data in the identification process, once the income level below which multiple deprivation increases markedly is established. This was the approach adopted by Townsend in 1979, but his main results have not been free of criticism. If there is no strong relationship between income and the standard of living, the task of identifying an “objective” income poverty line is clearly a difficult one.

b) Employing only standard of living data when identifying population groups suffering from poverty. This would require setting a level for the deprivation index or indices that could be considered as a critical point, a task hard to put into in practice. A given percentage of the population with the highest scores on deprivation could of course be taken, or, as in Mack and Lansley’s work, those people who lacked more than a certain number of items could be defined as poor. However, if we accept that the overall level of deprivation tends to increase gradually, decisions concerning how much deprivation is required for someone to be considered poor are difficult to make without some degree of arbitrariness²⁰.

c) Combining data on income and standard of living to identify the groups suffering from poverty, a method proposed by Ringen (1987, 1988) and applied to identify the “truly poor” by Halleröd (1995) and the “consistently poor” by Nolan and Whelan (1996). This is the approach that we will tentatively adopt in this paper, since it is coherent with the definition of poverty as a low standard of living due to the lack of resources. It is important to underline, however, that the criteria for establishing a combined income and deprivation threshold are far from clear. Moreover, if we recognise that resources and/or multiple deprivation could be not perfectly measured by the indicators employed, any identification of the “truly” poor by combining both kinds of information must be treated with caution.

In this section we will summarise the methodology applied and some of the results obtained by using the data on living conditions contained in the second wave of the European Community Household Panel Survey (ECHP). In accordance with the theoretical and methodological questions raised in the previous sections, we will try to study the possibilities and limitations of using the ECHP to elaborate deprivation indices which are valid for the measurement of poverty. Although the empirical analysis deals with a cross-section for a single country, we will attempt to evaluate the comparability across space and time of the indices obtained.

The data utilised have been taken from the European Community Household Panel (ECHP), the longitudinal survey compiled by Eurostat for European Union countries since 1994. The survey’s Spanish sample includes some 7,000 households. This data source possesses certain advantages over other sources used in previous studies, such as the Household Budget Surveys. These advantages include, basically: 1) more comprehensive research on the income perceived by households and their members; 2) more information on living conditions, including potentially valid indicators to analyse economic poverty as well as some elements of social exclusion, 3) the possibility of analysing income and living conditions from a dynamic viewpoint, due to the survey’s panel structure; and 4) the survey’s homogeneity, which opens up interesting prospects for elaborating new approaches to poverty in the EU countries.

When compared to other, more traditional, surveys, the ECHP has the disadvantage of not compiling data on consumption expenditure, a variable that may help to understand the relationship between income and the standard of living. Furthermore, the information on living conditions, despite being wide-ranging, is not always complete nor is it collected in the most appropriate way for the elaboration of deprivation indices, as we shall discuss below.

Next we summarise the methodology applied to the analysis of multiple deprivation using ECHP data and attempt to indicate both the restrictions and possibilities inherent in this source of information.

1. The list of potentially valid indicators to construct deprivation indices included in the ECHP is long. The variables contained in the survey range from aspects such as specific housing problems, the difficulties encountered in paying ordinary bills and housing costs, the possibility of affording certain goods and activities with current income, as well as the subjective perception the households have of their own situation. The general criteria summarised below have been employed when selecting the variables to be used in order to construct a deprivation index:

a) Most of the indicators selected concern goods or activities that are most probably considered to be basic necessities by the vast majority of the Spanish population today. Nevertheless, information on some variables representing goods or activities that only a half or less of the population have access to have also been included, and indices including and excluding that items have been constructed.

b) Variables regarding unemployment, health, social relations, or subjective feelings of unease (all of which have been studied in specific sections of the survey) have not been included in the indices. As we have argued above, it is felt that such indicators should not form part of a poverty index when poverty is understood to be a low standard of living due to a lack of resources.

c) As a rule, variables concerning the problems household members subjectively perceive have not been included either, especially when the perception of such problems seems to be determined more by the interviewees’ awareness of diverse quality of life issues than by the objective existence of such problems. For instance, these aspects could include feelings of unease about environmental problems such as noise and pollution, or vandalism in the neighbourhood.

d) Likewise, aspects whose connection to a household’s general economic situation is conceptually weak have been excluded. These elements include, for example, the lack of a terrace, patio or garden in the house, or the lack of natural light in some or all of its rooms.

2. When evaluating item-deprivation experienced by each household, we have followed as far as possible the “enforced lack” criterion discussed above, in order to consider a lack as deprivation only if it seems to be caused by insufficient resources. This kind of information is available for the consumer durable goods (colour televisions, telephones, cars, etc) a household has and the activities it can or cannot afford. The survey, however, does not provide enough information on other indicators to determine whether an enforced lack exists. In these cases, deprivation has been accepted as existing whenever a shortage has been detected. We consider that this decision does not distort the results in any fundamental way, as the lack of these goods or activities is unlikely to be voluntary in Spanish society today.

3. We essentially agree with Nolan and Whelan (1996) that it is useful to differentiate between various dimensions when constructing deprivation indices. Independently of what index or combination of indices is considered the most suitable for identifying the poor, we believe that it is convenient, at least initially, to consider separately those aspects which may be conditioned by different factors. Having undertaken a preliminary analysis of the internal correlation among the various indicators, we felt it was useful to classify the variables into the following four groups (see Table 1):

a) Difficulty in meeting ordinary needs (maintenance dimension). These include: not being able to afford a meat or fish meal every second day, not being able to afford new clothes, not being able to invite family or friends to the house for a drink or meal at least once a month, great difficulties in making ends meet at the end of the month and not being able to avoid arrears in at least two kinds of ordinary bills during the past year.

1. See comments on the text above. (2) Arrears in none or only one of the following kinds of expenses: a) rent or mortgage, b) standing orders, c) credit card payments. (3) Having none or only one of the following problems: a) leaks, b) damp, c) rotting floors or windows’ sills.

 b) Enforced lack of commonly owned durable goods (durables dimension). These are defined as those goods included in the survey and which the vast majority of the population owns (namely, colour televisions, videos, telephones and cars).

c) Poor housing conditions, reflecting problems which affect only a small minority of the population (housing dimension). These include the lack of an inside toilet with running water, the lack of an independent kitchen, the lack of a bath or shower with hot and cold water or not having a dry damp-free dwelling.

d) Enforced lack of goods or activities enjoyed by only a half or less of the population (life-style dimension). These include not being able to afford at least a week’s holiday away from home, not being able to afford new furniture, not being able to save some money, not being able to afford a dishwasher, microwave oven or adequate central heating.

Items included in a) aims to represent what we could call “maintenance poverty”, since it reflects a type of deprivation which is more likely to be related to current resources, whereas b) y c) could perhaps be better explained by reference to access to resources over a longer period. The usefulness of separately analysing the variables grouped as consumer durables and housing conditions is justified by the different amounts of financial resources necessary to accede to each group of goods, as well as by the possible specificity other studies have discovered in the population groups suffering basic housing problems²¹. Lastly, d) gathers together the items whose availability is probably most clearly correlated to income levels. We would naturally expect that those groups reaching high levels of deprivation in basic maintenance, durable goods and housing aspects to obtain high deprivation levels in these secondary “life-style” items as well.

4. Aggregated partial deprivation indices have been constructed for each of the four partial dimensions (maintenance, durables, housing and life-style). These indices represent the weighted and normalised sum of the indicators included in each group. Thus, for a specific household i (i=1,2, ..., n) and a specific dimension of deprivation m (m=1,2, ..., M), represented by a set of items, the partial index of deprivation would be:

=

Where represents the level of deprivation suffered by household i with regard to item j (j=1,2, ..., J^m) of dimension m and is the weight corresponding to this item. In accordance with the type of information collected in the survey, is defined as a binary variable that takes the value 1 when a household is suffering deprivation concerning the jth item and 0 when it is not. The normalised weight given to the jth item is given the following expression:

=

Where is the proportion of people not lacking item j of dimension m²². Thus, the weights attached to each item are functions of the spread of the good or activity among the whole population, compared to the spread of the other goods or activities considered. As we have argued above, this criterion seems preferable to the equal weighting applied in other studies, since it incorporates a trade-off between goods which reflects the consumption patterns of society as a whole.

The indices vary between 0 and 100 by construction. Each specific value can be interpreted as the percentage of deprivation experienced by a household in relation to the theoretical maximum value, which would only be reached if the household were to suffer simultaneous deprivation in all the items included. This normalisation enjoys certain advantages over the indices used in other studies, since it prevents the index’s absolute values from being affected by the number of variables included and it facilitates direct comparisons of the results obtained by different studies.

It is important to highlight that we have used the observed frequency of the different situations, and not the social perception of what constitute necessary goods, as the basic criterion for selecting and weighting indicators. Though the social perception of needs criterion could have certain advantages (as discussed above), this strategy is not applicable to the ECHP data. 

The studies carried out for other countries suggest that there is a strong correlation between the generalised access to goods, services or activities and their perception as necessities. There are, however, some significant exceptions²³. In this respect, we believe it would be both useful and feasible to carry out a periodic survey to gather such information within the framework of a poverty and social exclusion analysis within the European Union.

5. An aggregated overall deprivation index has been elaborated in order to measure the global deprivation suffered by each household. This index is constructed as a weighted sum of the partial indices under consideration, using weights that are functions of the average coefficients applied in constructing the partial indices. For a household i (i=1,2, ..., n), the overall deprivation index would be:

=

=

where is the average of the coefficients attached to the items of dimension m, m=(1, 2,..., M).

In this aggregation procedure, the percentage contribution of a given dimension to overall deprivation depends on the average spread across society of the items included in it, compared to the average spread of the items included in the other dimensions. Thus, the weighting structure mirrors the one applied to construct partial deprivation indices, except in that the dimensions are now treated as single goods. We believe this strategy is preferable to the simple aggregation of the items considered in Table 1, since the relative importance attributed to each dimension would then be affected by the number of indicators included in it, which could be merely the result of data availability.

As explained above, two versions of the aggregate overall deprivation index have been constructed, including (Total Index) and excluding (Basic Index) the items lacked by more than 50% of the population. Although most of the results are presented for both cases, the total index has been given priority when the sake for briefness suggested only one to be chosen, since no significant differences between them have been detected in the profile of high deprivation groups.

Table 2 shows how the population is distributed by segments of the different partial and overall indices obtained in the previous section. As can be observed, most of the population (over 70%) has a score of zero on the basic maintenance, durables and housing dimensions, and around 60% scores zero at the same time in the three dimensions (BI). Housing deprivation is the one least spread among the population, since only 12 in 100 individuals live in households that have some problem in this area. On the other hand, the above is only true for a minority (around 16%) with regard to the life-style and overall indices. Finally, the life-style deprivation index is the only one in which a significant portion of the population scores highly, a result consistent with the nature of the variables used to construct it.

Note: MI: Maintenance Index; DI: Durables Index; HI: Housing Index; LSI: Life-style index; BI: Basic Index; TI: Total Index. Source: Own research using ECHP data.

As was to be expected, a preliminary analysis of the data suggests there is a positive association among the different dimensions. This association, however, is not very strong, with correlation coefficients that range from 0.4 to 0.5 when comparing MI, DI and LS. It is interesting to note that the housing index shows the weakest correlation with the other dimensions (r =0.20 with maintenance and life-style; r =0.27 with durable goods).

 Note: Deprivation deciles are constructed separately for each deprivation index.

When studying the 10% of the population suffering from the highest levels of deprivation in each of the areas (Table 3)²⁴, the existence of different socio-economic profiles becomes clear. Thus, households headed by the unemployed constitute around 18% of each decile with the highest MI, DI and LSI, but are only 10% of households with the greatest HI. A similar pattern can be observed regarding households with three or more children and households which have to bear heavy housing costs. On the other hand, households living predominantly on pensions make up more than 30% of the highest DI and HI deciles, but only 24% of the highest MI decile. Also those who have reasonable housing costs contribute less to the maintenance deprivation dimension than to the other areas. We believe that these differences justify a separate analysis of each of the dimensions included in the index even when the aggregated indices are used to identify the population at greatest risk of suffering multiple deprivation.

Finally, there seems to be a remarkable degree of agreement between the two overall indices obtained regarding the profile of the highest overall deprivation groups. This profile is however somewhat different to the one obtained when examining the composition of the first income decile, especially concerning the position of elderly people (much more deprived than income poor) and the households bearing reasonable housing costs (which show the opposite situation). We must point out that the result obtained for the first group is partially conditioned by the low income threshold used in this comparison, being not so apparent when the two extreme deciles of income ad deprivation are compared.

Table 4 shows the correlation between the deprivation indices and the different household disposable income definitions. The basic income definition adopted is disposable yearly income adjusted according to the modified OECD scale. This scale was chosen because of our interest in comparing the results obtained with the poverty definition applied by the European Union.

As was to be expected, the correlation coefficients in Table 4 show a negative relationship between each index and household income, although the correlation coefficients’ absolute value is not especially high. These results should not be entirely attributed to the different way households are ordered according to the two criteria, since the deprivation indices are discrete variables that can take only a limited number of values. In any case, significant differences are detected between the LSI, the one most strongly correlated with income (r =-0.52), and the HI, which shows the weakest correlation (r =-0.13). These findings are consistent with the results obtained in previous studies performed in other countries²⁵.

Changing the definition of income used does not significantly alter the results. We may thus reject the idea that the low correlation obtained could be due to the use of a particularly unsatisfactory adjustment of income. It is worth noting that the correlation coefficients increase most when using income logarithms. This suggests the existence of a non-linear relationship between income and deprivation, an idea which has some theoretical consistence.

Notes: (*) Parametric equivalence scales (Buhmann et al., 1988)
Source: Own research using ECHP data.

Studying the way the indices vary throughout the income scale allows us to answer the question of whether it is possible to determine a level of income below which the deprivation indices increase in a disproportionate manner. Before examining the results, we should note that the multi-dimensional nature of poverty and the complexity of the factors leading to low standards of living in diverse fields make it extremely improbable that we shall find an income threshold which clearly separates the poor from the non-poor.

Graphs 1 and 2 show the average deprivation indices for the different income deciles. The indices are expressed as coefficients of the overall average for each index. The MI and, to a lesser degree, the DI, are those which increases most at the lower end of the income distribution scale. The HI shows, on the contrary, little variation at the three first deciles, what seems to confirm that a difference exists in the groups affected by these form of basic deprivation. Lastly, we can see that the life-style index decreases moderately but continuously as income increases.

Graph 1
Average deprivation coefficients by income deciles (MI, DI, HI, LSI)

Observing the overall indices (Graph 2), it is apparent that the basic and the total indices show a similar relationship with income, although the BI has a steeper slope at low income levels. Nonetheless, an exponential increase below a specific level of income cannot be detected.

All of the above is confirmed when the behaviour of the deprivation indices are studied at around the income level used as a relative poverty threshold by European Union reports (60% of median income). As can be seen in Table 5, the average values of the indices of the households located slightly above and below the poverty threshold are strikingly similar.

Source: Own research using ECHP data.

Thus, although all the indices increase as income diminishes, we cannot establish a specific level of income which represents an objective threshold to separate households suffering from multiple deprivation from those which are not. From another viewpoint, this implies that many households classified as poor by the income criterion could have deprivation indices that are not particularly high. Exactly the opposite could happen to households located above the poverty line. In conclusion, the Spanish data also seem to confirm the idea that a perfect correlation between income levels and standard of living (insofar as we are able to define and measure both concepts) does not exist. Income could therefore be considered as an insufficient indicator to identify people excluded from minimum standard of living because of a lack of resources. We explore this question in more detail below.

As various studies have pointed out, combining income and standard of living indicators could be a useful tool to more accurately identify the poor. Nonetheless, a clear consensus does not yet exist regarding the most appropriate way of using both criteria jointly. The number and characteristics of the “truly poor”, to use Halleröd’s (1995) expression, depend on the specific definition of income used (the equivalence scale employed, for instance), the level at which the poverty threshold is set, and the variables as well as the aggregation methods employed to obtain the deprivation indices. Before deciding on the identification criteria, we firmly believe that a better understanding of the factors conditioning the relationship between income and standard of living is necessary. This is also true for the properties and limitations of the indicators of standard of living to be employed.

Source: Own research using ECHP data.

We have systematically looked into the degree of overlap between the low income and high deprivation criteria combining equally sized groups making up 10%, 20% and 30% of the population respectively. The results are shown in Table 6. As in other studies, we can observe that a substantial part of those classified as poor according to one of the criteria would not be considered as such by the other. Nevertheless, we can, interestingly, conclude from the results that the level of overlap increases as less restrictive low income and high deprivation criteria are used. Thus, only 18% of those belonging to the first income decile or to the last deprivation decile simultaneously belong to both. The percentage, however, increases to 40% when we consider the three extreme deciles. These calculations have been carried out taking the overall Total Index into account. Using the index that excludes life-style indicators provides almost identical results.

Table 7
Average Income values and deprivation indices for each group
(National average=100)

Note: Total Index deciles ´ Income deciles
Source: Own research using ECHP data.

Table 7 shows the average income and average partial and global indices coefficients for the four groups at the three levels of overlap under study. In all these cases, the poor according to both the income and standard of living criteria are clearly differentiated from those who are neither income nor deprivation poor. As opposed to the “consistently poor”, those who have low income but not low standard of living obtain, as expected, fairly low deprivation indices, despite they have average income levels only marginally above the first group. Some of the characteristics that can explain these differences will be looked at in the next section.

Lastly, it is of interest to note that households only suffering from deprivation have standard of living indicators close to the level of those classified as poor by both criteria. Although our study does not explicitly take this group into account²⁶, we believe it would be necessary to investigate more deeply the characteristics of such households in the future. Similarly, the behaviour of the housing deprivation index clearly requires more in-deep consideration.

What are the characteristics which differentiate the poor (as identified by both criteria) from those who have low incomes but do not have high levels of deprivation? As we shall see, jointly using income and deprivation criteria does not uncover new groups at risk. However, it changes to a certain degree the composition of the population considered to be poor.

Only a limited group of variables are considered in the analysis. This group of variables does not exhaust the possibilities offered by the data contained in the ECHP. This first approach does not aim to analyse the general profile of poverty in Spain. Instead, it attempts to reveal some of the consequences of jointly using direct and indirect indicators to identify the most deprived groups of the population.

The first income quintile is taken as a reference point in this section, since it is a similarly-sized group to the one obtained by setting the poverty threshold at 60% of median income (18.9% of the population). The composition of this group will be compared to the group simultaneously suffering from low income and deprivation using relative incidence (RI) figures, defined as the coefficient between the proportion of each group in poverty (in terms of income, or income and deprivation) and the proportion of the poor among the whole population. Furthermore, the average values of the four partial indices will be examined, in order to investigate (in a rather preliminary way) the relationship between specific deprivation areas and socioeconomic characteristics of the households.

The profile of the poor according to the main breadwinner’s sex and age is not altered appreciably when using standard of living indicators. The greatest difference is found in households headed by women over the age of 65. This kind of household makes up a larger percentage of the poor identified by both criteria (LI+D poor) than of the low income population (LI poor). The average values of the partial deprivation indices allow us to verify that this group clearly suffers from greater problems in all areas, and especially in housing and durable goods. Households headed by men of the same age also find themselves in an unfavourable situation, but less so.

The second conclusion that can be drawn from the examination of the sociodemographic variables concerns the greater presence of single-parent families and households with three or more minors among the groups suffering from low income and deprivation. They are a high risk group according to the income criterion, but the probability of them being poor increases if we simultaneously consider their standard of living. The average values of the different deprivation indices shows that these households have high levels of maintenance and durable goods poverty, something that suggests that the equivalence scale used could underestimate to some extent the cost associated with bringing up children, especially in households headed by a single mother. This result has already been pointed out by other studies²⁷.

Note: LI: Low income. LI+D: Low income and deprivation. (1) Main breadwinner defined as the person who contributes most to the household income.
Source: Own research using ECHP data.

b) Labour force status and sources of income

The household’s sources of income and the labour force status of the main breadwinner are typically variables that have a great impact on families’ economic situation. Table 9 shows the results obtained when using the income criterion or combining income and standard of living indicators to define the poor.

If we focus on the sources of income of households, we can observe that households whose incomes depend on different kinds of social benefits or other unspecified sources of income are more common among the LI+D groups than among the income poor, and the opposite is true for households which obtain their income from self-employment, capital or property.

With regard to households whose income comes from self-employment, it is interesting to highlight that 17,4% have incomes situated in the fifth income quintile, but only 5% simultaneously suffer from high levels of deprivation. This mean that around 71% of the households identified as poor by the income criterion do not belong to the LI+D group. Something similar can be detected in households which receive their main source of income from capital or property, although these constitute a much smaller group (approximately 0.6% of the population), and thus the impact on the overall composition of poverty is negligible. Observing the average deprivation indices of both groups also tends to confirm the idea that they do not have a worse standard of living than households with wages as the main source of income.

There are two possible reasons for the results outlined above. Firstly, income obtained from self-employment and capital tends to be more variable over time than other sources of income such as salaries or pensions. This could give rise to temporary increases and decreases that do not necessarily have a repercussion on the standard of living. Secondly, these kinds of income tend to be especially underestimated in income surveys, probably due to a greater reluctance on the part of interviewees to declare their incomes as well as the difficulties they encounter in remembering the amount of income received during a specific period. Thus, the current income declared by these groups could be an inaccurate reflection of their true economic situation and lead to an overestimation of their levels of poverty.

Note: (1) Main breadwinner defined as the person who contributes most to the household income. (2) The percentages of population do not sum up to 100, because a 0.6 of households lacking information on this characteristic have been excluded.
Source: Own research using ECHP data.

Households whose main source of income is derived from social benefits are relatively more frequent among the poor, as identified by both criteria, than among the low income groups. Households dependent on pensions are not clearly over-represented in the LI+D group. Nonetheless, they obtain higher than average deprivation levels, particularly with regard to housing and durable goods. The highest indices correspond, however, to households depending on unemployment benefits, other social benefits or unidentified income.

The employment status of the principal breadwinner is one of the main factors determining that household’s economic situation, as all studies agree. A traditional limitation on the available data sources is that they generally gather information on the previous year whilst looking into the interviewees’ employment situation for a much shorter period (the previous week, in most surveys). This makes it difficult to analyse any improvement or deterioration in their economic situation in connection with work and unemployment.

Fortunately, the ECHP allows us to largely resolve this difficulty, as it determines interviewees’ current work situation as well as their predominant work situation in the previous year. This makes it possible to distinguish the various categories shown in Table 9. About 65% of the population live in a household whose main breadwinner is employed at the moment of the interview, although this has been preceded by a period of unemployment during the previous year in some cases. 8% of the population belongs to a household whose main breadwinner is currently unemployed, and almost half of this group were also unemployed during the previous year.

Analysing the data collected in the table clearly shows the importance of the interviewees’ recent employment history with regard to income poverty rates and the risk of belonging to the group of the “consistently poor”. The permanently employed tend to have lower than average poverty rates and an even lower presence in the groups suffering from low income and deprivation. Exactly the opposite is true for the unemployed and, particularly, the long-term unemployed. The chances of this last group having incomes situated in the first quintile is 3.3 times higher than the average. Additionally, the probability of the long-term unemployed belonging to the low income and low standard of living group is 4.3 times greater than the average. The average deprivation indices of this group are also the highest, only exceeded by the group housewives’s headed households in the area of housing conditions.

A second interesting observation is that changes in an individual’s employment situation lead to changes in the levels of current income which, however, do not have an immediate effect on the standard of living. Table 10 allows us to examine this in more detail, as it shows the different groups’ income indices and poverty rates for the previous year and month. 

Note: Poverty line = 60% of median equivalent income. Source: Own research using ECHP data.

As can be seen, the employed who were previously unemployed have monthly incomes and short-term poverty rates that indicate an improvement compared with the results obtained for the previous year. Despite this, they still find themselves far below the national average in the income scale and have also over average deprivation indices in all areas. This group probably includes a large number of people working with temporary fixed-term contracts who fluctuate between situations of employment and unemployment. Though some households in this group may be emerging from poverty, for many others the mere fact of being employed at the moment of the survey does not necessarily constitute a reliable way to improve their standard of living.

To a large extent, the situation of the unemployed who worked in the previous year reflects an inverse process, which is however not entirely symmetrical. Comparing their current monthly income to the previous year’s shows a relative decrease, although the repercussion on the standard of living is moderate (especially so in durable goods and housing conditions). This group’s poverty rate in the previous year is at the national average (though higher than the poverty rate of the group of workers in stable employment), while its short-term poverty rate is, on the contrary, clearly below the national rate. However, this group risks decreasing living standards if unemployment continues.

In conclusion, we should note that deprivation levels are not only caused by individuals’ current employment situation, but also by longer-term employment history. This is particularly true for indicators concerning medium-term processes of accumulation such as housing conditions and durable goods.

c) Housing

Unlike other European countries, the large majority of people in Spain own their own houses. The ECHP data shows that more than 80% of the population has bought or is in the process of buying the house they live in. Purchasing a house is undoubtedly the greatest single investment people make in their lives and also the most widespread form of wealth.

One of the limitations of annual monetary income is that it generally does not permit us to adequately incorporate variables such as housing or other assets when determining the level of economic well-being which can be reached with a specific level of income. It is to be expected that households that do not have to meet heavy housing expenses (because they have paid off their mortgages or because they pay relatively low mortgage or rental costs in relation to their income) enjoy a higher standard of living than those that do.

The ECHP includes data that allows us to evaluate this question. Along with house ownership, the survey looks into housing costs (mortgage costs, rental costs, communal housing expenses, property tax, etc.) and whether they constitute a heavy or reasonable financial burden for the households interviewed. It is possible to identify five different groups when we combine both kinds of information, which are shown in Table 11.

Comparing the pattern of poverty according to income with the pattern derived from jointly using the income and standard of living criteria offers results that are consistent with the hypotheses mentioned previously. The half of the population which owns dwellings with reasonable housing costs has lower poverty rates than the national average in both cases, but the risk they run of having low income (RI=69) doubles that of simultaneously having low income and suffering deprivation (RI=35). The group of people (much less numerous) who rent housing at a reasonable cost is also under-represented among the poor as identified by both criteria.

Source: Own research using ECHP data.

Exactly the opposite is true for those finding it difficult to meet their housing expenses, for both house owners and (especially) renters. The chances of simultaneously having a low income and suffering multiple deprivation increases clearly for the latter. The lesser ability of this group to transform a given level of income into a fair standard of living can perhaps be better appreciated if we note that 47% of those not considered to be poor by the income criteria suffer still from high levels of deprivation. This only occurs with 6% of non income-poor house owners who pay reasonable costs.

Thus, housing occupancy situation and, in particular, the financial costs associated with it are important variables in determining whether a household is “consistently” poor or not. We can conclude that the income criterion overestimates the economic position of households that have to meet large housing costs, and underestimates the position of households that do not have to meet such heavy costs (especially in the case of owners).

Lastly, examining the deprivation profile of these groups leads us to some additional conclusions. Firstly, heavy housing costs considerably increase the chances of suffering from maintenance poverty, the inability to meet basic current expenses. Secondly, deprivation concerning housing is much more frequent in the groups that either rent houses or have been ceded houses. These two groups have averages values of this index that are well above those of homeowners.

This paper has examined the possibilities for the analysis of poverty offered by the information on income and living conditions contained in the ECHP. Poverty is understood to be a low standard of living brought about by insufficient resources. Deprivation indices have been elaborated, based on a series of direct indicators for which the survey offers information. The classification of the indicators into various different areas (maintenance, durable goods, housing and life-style) is justified by the idea that poverty is a complex phenomenon comprised of several dimensions, each of which may be influenced by specific sets of factors.

The construction of the deprivation indices has been based on previous work carried out by other authors (Townsend, Desai and Shah, Mack and Lansley, Halleröd, and Nolan and Whelan, among others), though it departs in specific aspects from the methodology employed in each of these studies. The list of indicators chosen includes some items possessed by less than 50% of the population, like in Halleröd (1995), but the weighting structure is closer to Desai and Shah’s strategy, given the lack of information about socially perceived necessities in the ECHP. When aggregating item-deprivation indicators into partial and overall deprivation indices, we use a normalisation procedure that improves the comparability of the values obtained by different studies. We have also stressed the need to avoid the number of items included in a specific dimension influencing the weight attributed to this dimension in the overall deprivation index.

Our results generally confirm those obtained for other countries, with regard to the correlation between income and living conditions, as well as the imperfect overlap that exists between low income groups and groups suffering from high levels of deprivation. We have observed, however, that the degree of overlap between the two measures varies according to the criteria applied to delimit low income or high deprivation groups, with a greater overlap when less restrictive definitions are used.

The analysis of the poor identified by both criteria has been performed by cross-referencing groups of equal size, comprised of the 20% with lowest incomes and the 20% showing the highest overall deprivation level. These groups are of roughly the same size as the European Union’s estimate of relative poverty in Spain. Among the factors which can explain the differences observed, we have highlighted the irregular nature and the underestimation of some sources of income, the importance of individuals’ recent working experience in the process of improving or declining standards of living, the difficulties encountered in correctly estimating the needs of differently sized households and the role played by housing occupancy situations and its associated costs.

Combining income and standard of living indicators does not generate a new profile of poverty. However, it does tend to modify the relative risks attributed to the different groups, and shows that a given level of income (including equivalent income) does not always have the same effects upon standard of living. In addition to the inclusion of additional explanatory variables and the analysis of their interaction, further research on this topic might comprise its repetition with larger or smaller groups of the population. Similarly, the “inconsistent” groups clearly require greater research before being excluded from the group of the “truly poor”. It must be remembered that the indicators employed are imperfect measures of both the standard of living and the level of access to resources. These limitations should warn us against excessive optimism about the results obtained when combining both kinds of variables.

It should be noted that the conclusions obtained in this and other studies suggest ways of better assessing income that could solve some of its traditional limitations. Although our results support the views of researchers like Ringen who have argued that the level of income is an imperfect indicator when identifying the poor, we believe that they by no means demonstrate that income should be left aside in poverty analysis. This is specially true if we recognise that the construction of multiple deprivation indices still faces numerous theoretical and empirical unresolved problems, as we have tried to stress in our paper. These problems include issues such as the question of which indicators to employ and how to weight them; the relationship between the lack and enforced lack of the different items; the distinction of various dimensions of deprivation and the best way to employ these areas to identify the poor; or the comparability of deprivation indices over time and across countries.

Finally, we would like to emphasise the fact that this paper has focused on poverty, rather than on the broader concept of social exclusion. Some viewpoints suggest that there is an implicit critique to the traditional ways of measuring poverty embedded within the idea of social exclusion. Poverty measures would offer a snapshot of the low-income population (according to a rather arbitrary criterion), without looking into the implication low income has for the standard of living and social integration. Though this can be to some degree true regarding main-stream income poverty estimates, it is necessary to highlight that research into poverty has been far more rich and extensive.

We firmly believe that the most commonly accepted definition of poverty (exclusion from an acceptable standard of living due to a lack of resources) necessarily constitutes a central element in the notion of social exclusion, however it may be defined, and that both concepts should be developed within a common theoretical and empirical framework. In this task, the important body of poverty studies on aspects such as the relationship between income and the standard of living, subjective poverty or the dynamics of low income can play an important role in the years to come. 

Atkinson, A.B. (1989), Poverty and Social Security, Harvester Wheatsheaf, Londres.

Atkinson, A.B. (1998): “Social Exclusion, Poverty and Unemployment”, en Atkinson, A.B. y J. Hills [eds.]: Exclusion, Employment and Opportunity, Centre for Analysis of Social Exclusion, CASE Paper, CASE/4.

Bradshaw, J.; S. Middleton; P. Townsend; D. Gordon (1998), “Perceptions of poverty and social exclusion. Final report on preparatory stages of research”. Mimeo. Universidad de York.

Bradshaw, J. (2000), “Child poverty” Paper presented at the Conference “Políticas sociales contra la pobreza”, Bilbao, 7-9 of June.

Bourguignon, F.; S.R. Chakravarty (1998): “A Family of Multidimensional Poverty Measures”, Document nş 98-03, DELTA, Centre National de la Recherche Scientifique, Ecole des Hautes Etudes en Sciences Sociales, Ecole Normale Superieure, Paris.

Bourguignon, F.; S.R. Chakavarty (1998): “The Measurement of Multidimensional Poverty”, Document nş 98-12, DELTA, Centre National de la Recherche Scientifique, Ecole des Hautes Etudes en Sciences Sociales, Ecole Normale Superieure, Paris.

Buhmann, B.; L. Rainwater; L. Schmaus; T. Smeeding (1988), "Equivalence scales, well-being, inequality, and poverty: sensitivity estimates across ten countries using the Luxembourg Income Study (LIS) database", Review of Income and Wealth 34, 115-42.

Callan, T., B. Nolan, B.; C.T. Whelan (1993), "Resources, Deprivation and the Measurement of Poverty", Journal of Social Policy, 22(2), 141-172.

Davies, H.; H. Joshni; L. Clarke (1997): “Is it Cash that the Deprived are Short of?”, Journal of the Royal Statistical Society, 160(1), pp. 107-126.

Desai, M.; A. Shah (1988), "An econometric approach to the measurement of poverty", Oxford Economic Papers, 40(3), pp. 505-522.

Dirven, H-J; G. Linden; B. Mikulic; J. Schiepers; C. Siermann; W. De Wreede (2000): Income Poverty and Social Exclusion in the European Member States, Eurostat, March 2000.

Federman, M. et al. (1996): “What does it mean to be poor in America?”, Monthly Labor Review, May, pp. 3-17.

Gordon, D; C. Pantazis (1997): Breadline Britain in the 1990s, Aldershot: Ashgate.

Halleröd, B. (1994): Poverty in Sweden: a New Approach to the Direct Measurement of Consensual Poverty, Umea Studies in Sociology, nş 106.

Halleröd, B. (1995), “The truly poor: diret and indirect consensual measurement of poverty in Sweden”, Journal of European Social Policy, 5(2), pp. 111-129.

Hutton, S. (1991), "Measuring living standards using existing Data Sets", Journal of Social Policy, 17(2), 237-57.

Mack, J.; M. Lansley (1985), Poor Britain, Londres: Allen and Unwin.

Martínez, R. (1997), “Pobreza monetaria y condiciones de vida en Espańa: algunos resultados basados en la Encuesta de Presupuestos Familiares”, IV Encuentro de Economía Pública, Universidad Pública de Navarra.

Martínez, R.; J. Ruiz-Huerta (1999): “Algunas reflexiones sobre la medición de la pobreza. Una aplicación al caso espańol”, in Maravall, J.M. [ed]: Dimensiones de la desigualdad. Madrid: Fundación Argentaria – Visor.

Mayer, S.; C. Jencks (1989): “Poverty and the Distribution of Material Hardship”, The Journal of Human Resources, 24(1), pp. 88-113.

McGregor, P.P.L.; V.K. Borooah (1992), "Is Low Spending or Low Income a Better Indicator of Wether or Not a Household is Poor: Some Results From the 1985 Family Expenditure Survey", Journal of Social Policy, 21(1): 53-69.

Mercader-Prats, M. (1998), “Identifying Low Standards of Living: Evidence from Spain”, Research on Economic Inequality, 8, 155-173.

Nolan, B.; C.T.Whelan (1996), Resources, Deprivation and Poverty, Oxford: Clarendon Press.

Piachaud, D. (1987): "Problems in the Definition and Measurement of Poverty", Journal of Social Policy 16(2), 147-164.

Paugam, S. (1996): “Poverty and social disqualification: A comparative analysis of cumulative social disadvantage in Europe”, Journal of European Social Policy, 6(4), pp. 287-303.

Ringen, S. (1987): The Possibility of Politics, Oxford: Clarendon Press.

Ringen, S. (1988), "Direct and Indirect Measures of Poverty", Journal of Social Policy, 17(3), 351-66.

Room, G. [ Ed.] (1995), Beyond the Threshold: The Measurement and Analysis of Social Exclusion, The Policy Press.

Ruiz-Huerta, J.; R. Martínez (1994), "La pobreza en Espańa: żQué nos muestran las Encuestas de Presupuestos Familiares?", Documentación Social, 96, 15-109.

Townsend, P. (1979), Poverty in the United Kingdom, Penguin Books, Harmondsworth.

Townsend, P. (1987), “Deprivation”, Journal of Social Policy, 16(2), 125-146.

-----------------------------------------------

1. This study has been partly financed by the funding received from the Science and Technology Inter-ministerial Commission (Comisión Interministerial de Ciencia y Tecnología) to the Project SEC 98-1090.

2. See for example Townsend (1979).

3. See Ringen (1987, 1988).

4. Relativity, agency and the importance of the local environment are other frequently mentioned characteristics of the concept of social exclusion. See Room (1995) and Atkinson (1998) for a more detailed discussion.

5. See for example Callan, Nolan and Whelan (1993), Nolan and Whelan (1996), Federman et al. (1996), Halleröd (1994, 1995), Gordon and Pantazis (1997), Hutton (1991), McGregor and Boroaah (1992).

6. Martínez (1997) and Martínez and Ruiz-Huerta (1999) provide some results based on deprivation indices for Spain. See also Mercader-Prats (1998), where a combination of income and consumption is used to identify the poor.

7. Obviously, the reliability of this order is always dependent upon the correctness of the underlying hypothesis applied in the construction of the deprivation index.

8. Even if indicators with a non strictly material dimension are usually included, such as the capability to participate in some social or cultural activities.

9. Note that they could, however, be considered as constituent elements of social exclusion if we consider that employment and good health are necessary for integration into society.

10. Who, and on what grounds, should define what is necessary and what is not? Should necessities be defined according to consumption patterns, expert judgement or social consensus? What exactly does social consensus on this issue mean? Should the list of necessities vary according to the socio-demographic characteristics of individuals?

11. This is not to say that a deprivation index constructed on this basis could be used to measure economic wellbeing. Obviously, deprivation indices could never be able to reflect the enviable complexity of being disgustingly rich.

12. For a more detailed discussion of this issue see Halleröd (1994).

13. See Mack and Lansley (1985).

14. For example, the preparatory work done for the third edition of the survey originally designed by Mack and Lansley in the 1980s. See Bradshaw, Middleton, Townsend and Gordon (1998).

15. In this respect, see Dirven et al. (2000)’s comments on the problems posed for the selection of a common set of non-monetary indicators of social exclusion for EU countries.

16. In the same line as methodologies developed to deal with “dirty data” in conventional poverty measurement.

17. Piachaud (1987), p. 150.

18. In both cases, the weights assigned to each item could be the same for the whole population or differ across socio-demographic groups.

19. Note that these problems are even more significant if poverty analysis is understood as a part of the analysis of a more wide-ranging concept of social exclusion, in which poverty in itself can be seen as an important dimension of social exclusion.

20. Bourguignon and Chakravarty have recently proposed a family of multidimensional poverty measures based on standard of living information. To calculate these measures, a threshold level must be determined for each of the attributes under consideration. These thresholds are intended to represent the subsistence levels for the different indicators, so that the notion of poverty adopted is of absolute type. See Bourguignon and Chakravarty (1998a, 1998b).
21. See for example Nolan and Whelan (1996).
22. Note that this group does not always coincide with the non-deprived, since some people lacking specific goods by choice are not counted as deprived.
23. Articles such as color TVs, for instance, do not tend to be regarded as basic necessities despite being widely owned. Being able to save money, on the other hand, is considered to be a necessity though only a minority of the population is able to do it.
24. For this analysis, only a limited set containing the most significant variables have been employed.
25. Nonetheless, the correlation coefficients obtained are somewhat greater than the ones obtained by Nolan and Whelan (1996), except in the case of the housing index.
26. The results contained in the following section indirectly point towards some of the possible explanations.
27. See, for example Bradshaw (2000).