SEARCH

SEARCH BY CITATION

Keywords:

  • poverty;
  • social exclusion;
  • welfare;
  • deprivation;
  • social policy;
  • marginalisation

Abstract

  1. Top of page
  2. Abstract
  3. Poverty and social exclusion
  4. Data and measurements
  5. Results
  6. Conclusions and discussion
  7. References

This article investigates whether, and to what degree, poverty is linked to other types of welfare problems and, in larger perspective, whether the situation can be understood in terms of social exclusion. Two different measures of poverty – income poverty and deprivation poverty – and 17 indicators of welfare problems were used in the analysis. It was shown that income poverty was rather weakly related to other types of welfare problems, i.e. the most commonly used measure of poverty seems to discriminate a section of the population that does not suffer from the kinds of problems we usually assume that poverty causes. Deprivation poverty, identifying those who most often had to forgo consumption of goods and services, did correlate strongly with other types of welfare problems. Hence, people living under poor conditions do suffer from welfare problems even though this section of the population is not always captured by income poverty measures. The final analysis showed that the types of welfare problems that were most likely to cluster were deprivation poverty, economic precariousness, unemployment, psychological strain and health problems. Whether these types of accumulated welfare problems, from a theoretical perspective, can be seen as indicators of social exclusion is more doubtful.

The Amsterdam treaty places the fight against poverty and social exclusion at the centre of the European Union's (EU) social agenda. But what is the relationship between poverty and social exclusion? Can we and should we distinguish these concepts from each other, or is the label just a tautology? Is the fight against poverty one thing and the fight against social exclusion another, or do they essentially constitute a single battle? To answer these questions, we need to conceptually distinguish poverty from its causes and consequences and empirically investigate whether, and how, poverty is linked to other types of welfare problems and, in the end, whether this situation can be understood in terms of social exclusion.

In this article, we will study the consequences of poverty, focusing mainly on how poverty is related to a range of other welfare problems such as unemployment, health, psychological distress, victimisation etc. The purpose is also to analyse whether, and how, different welfare problems are related to not only poverty but also to each other. In a wider perspective, the analysis is linked to the discussion about social exclusion, a phenomenon often understood as accumulated welfare problems, i.e. a situation in which a single individual is suffering from several different welfare problems at the same time (cf. Gallie, Paugam & Jacobs, 2003; Hills, Le Grand & Pichud, 2002).

The article is organised in the following way: the following section deals with the theoretical definition of poverty and how to distinguish poverty from other types of welfare problems and social exclusion. Thereafter, measurements of poverty and welfare problems are discussed. The empirical analyses are presented in the penultimate section, followed by the discussion and conclusions.

Poverty and social exclusion

  1. Top of page
  2. Abstract
  3. Poverty and social exclusion
  4. Data and measurements
  5. Results
  6. Conclusions and discussion
  7. References

Why is it important to identify the poor and to take action against poverty? The commonsensical answer is that poor people suffer from malnutrition, lack of shelter, ill health, exclusion from an ordinary lifestyle in society etc., and that such a situation is unacceptable. In a way, one could say that this broad picture of poverty is correct because if the poor do not suffer from a wide range of problems, why should we bother about poverty? The problem from an analytic perspective is that inclusion of almost every unwanted condition in the theoretical definition of poverty makes it impossible to analyse the causes and consequences of poverty. It also becomes more or less unfeasible to distinguish poverty from other concepts such as social exclusion. We therefore need a theoretical definition of poverty that focuses on the inability to make ends meet. That is, the poor are those who, due to insufficient access to economic resources, have an unacceptably low level of consumption of goods and services. The important consequence of this kind of definition is that, for example, malnutrition per se is not poverty; it is caused by poverty only if it is lack of economic resources that makes it impossible for a person to acquire food. The fact that malnutrition is most often a poverty problem does not mean that it is always a poverty problem. Consider, for example, anorexia, a condition in which malnutrition is not caused by poverty. Similarly, bad health is, in many cases, unrelated to poverty. It is only a consequence of poverty if it is caused by an inability to buy adequate food, provide for shelter or pay for healthcare. The relationship between poverty and other types of welfare problems is further complicated by the fact that welfare problems often cause poverty. There is plenty of evidence to show that unemployment is problematic even if the unemployed person is protected from poverty (cf. Nordenmark, 1999; Strandh, 2000). But it is, of course, also reasonable to argue that unemployment often causes poverty. It is also easy to see that health problems can cause poverty by preventing people from earning a living. Disentangling causes and effects at the individual level is, to put it mildly, empirically complicated, and the process is probably best understood as either a ‘vicious circle’ leading to an accumulation of welfare problems into a situation of social exclusion, or a ‘good circle’ out of social exclusion. An important step in facilitating our understanding of these ‘circles’ is to discover whether, and how, poverty and other welfare problems are interlinked with each other at one point in time. This means that we must first distinguish different types of welfare problems and then empirically check whether they form a cluster of accumulated welfare problems. In this perspective, poverty should be seen as just one welfare problem conceptualised as lack of economic resources.

There are a number of studies showing that welfare problems do cluster (Bask, 2005; Erikson & Tåhlin, 1987; Fløtten, 2005; Halleröd, 1991; Halleröd & Heikkilä, 1999; Tham 1994). Recently, Bradshaw and Finch (2003) examined the degree to which three different measures of poverty were related to different aspects of social exclusion in Britain. In line with earlier findings (cf. Berthoud, Bryan & Bardasi, 2004; Halleröd 1991, 1995, 2000; Kangas & Ritakallio, 1998), they found that the overlap between different poverty measures was fairly limited, i.e. different measures tended to identify different individuals as poor and that different measures of poverty were distributed differently in the population. The unique contribution of Bradshaw and Finch is that they also demonstrated that different measures of poverty relate differently to different indicators of welfare problems. The contribution of this article is that it takes previous analyses of the relationship between welfare problems a step further and studies in detail whether welfare problems are related and, if so, what kind of welfare problems cluster and how such a cluster is related to poverty. Lastly, we will approach the issue of whether an empirically observed cluster of welfare problems can be reasonably understood in terms of social exclusion.

Three hypothetical outcomes regarding the relationship between poverty and other types of welfare problems can be put forward:

  • 1
    Poverty is not related to other welfare problems, a result implying that poverty in today's Sweden, because it apparently does not have any welfare consequences, is more or less a quasi-problem.
  • 2
    We find a cluster of related welfare problems, but the cluster is unrelated to poverty. The conclusion in this case would be that accumulation of welfare problems is an empirical fact clearly distinguished from poverty and that poverty is still to be viewed as a quasi-problem.
  • 3
    Poverty is related to a range of welfare problems. Poverty, thus, can be viewed as a serious problem and keeping people out of poverty would appear to be an important goal for social policy. This outcome also makes it reasonable to view the fight against poverty and the fight against social exclusion as a single battle.

Data and measurements

  1. Top of page
  2. Abstract
  3. Poverty and social exclusion
  4. Data and measurements
  5. Results
  6. Conclusions and discussion
  7. References

The data are from Statistics Sweden's annual Survey of Living Conditions from 1998 (ULF98) and are based on face-to-face interviews with a random sample of the Swedish population aged 16–84 years. The random sample contains 7,400 individuals, and the total working sample amounts to 5,732, which gives a response rate of 77.5 per cent. The working sample is limited to respondents aged 20–74 years, which gives a working sample consisting of 4,941 cases. ULF98 contains a large number of welfare indicators and a unique set of questions regarding consumption of goods and services that facilitates a so-called consensual measurement of poverty introduced by Mack and Lansley back in the early 1980s.

Poverty measures: operationalisation

We will employ two measurements of poverty: income poverty and deprivation poverty. Income poverty is measured in accordance with the conventional EU measurement of relative poverty, i.e. those who live in a household with an equivalent disposable income that is below 60 per cent of the median household income are defined as poor. Income data are gathered from the income register and represent the households’ disposable after-tax income from all registered sources. The disposable income is adjusted to household size using an equivalence scale developed by the Swedish National Board of Health and Welfare (Statistics Sweden [SCB], 2003).1

Deprivation poverty is measured using a weighted deprivation index (WDI) (Halleröd, Gordon, Larsson & Ritakallio, 2006). The index measures inability to consume goods and services in accordance with the general lifestyle in the society, and is a developed version of the consensual measurement of poverty introduced by Mack and Lansley (1985). Information for the measurement of poverty comes from a list of 36 consumption items, see Table 1, and for each item respondents have been asked whether they have the item (or engage in the activity). If the answer is ‘No’, respondents are asked whether this is because they cannot afford the item/activity or whether they are simply not interested in having/engaging in it. Respondents receive a score on the deprivation index for every item they cannot afford. In order to strengthen the connection between the deprivation index and the ordinary lifestyle, the index score is weighted in relation to the fraction of the population that has the item in question. For example, 84 per cent of adults in Sweden state that they have a yearly dental examination. Therefore, not being able to afford such an examination results in a score of 0.841 on the deprivation index. However, these weights are only average. It is likely that consumption patterns are different in different sections of the population. Calculation of the index takes into account two such differences (for a more detailed discussion see Halleröd 1995; Halleröd et al., 2006). First, it is acknowledged that consumption patterns differ across age groups. Second, it is also recognised that families with children have needs that families without children do not have. Construction of the deprivation index implies that:

Table 1.  Proportion of the population that have, do not want and cannot afford consumption items.
Consumption itemHaveDo not wantWould like to have but cannot affordNot relevant
Dishwasher53.535.48.52.2
Washing machine74.815.97.11.9
Freezer98.20.60.80.1
Microwave oven77.417.34.30.9
Vacuum cleaner99.30.20.30.1
Telephone98.40.60.70.2
Mobile telephone65.826.85.91.2
TV98.61.00.30.1
VCR83.611.73.90.6
Stereo equipment89.27.92.20.5
Computer (PC or Mac)58.025.913.02.7
Daily newspaper76.614.28.40.7
First-hand contract or self-owned accommodation97.00.71.50.7
Modern dwelling (bath/shower, WC, central heating, oven and refrigerator)98.60.50.60.2
Balcony or garden92.62.02.82.4
Not more than two persons per bedroom96.01.01.91.0
Comprehensive home insurance96.80.61.30.6
Car84.57.17.31.0
Driving licence86.05.55.92.4
Public transport for one's needs73.617.11.08.1
Clothes that to some degree correspond with fashion87.67.63.80.4
A ‘best outfit’ for special occasions91.93.83.80.2
Buying new clothes, not second-hand92.22.25.10.2
A haircut every third month69.321.46.72.4
A hot meal each day97.31.80.60.2
A special meal once a week72.918.57.50.7
Celebrations on special occasions87.07.54.40.8
Presents for friends and family at least once a year97.21.61.00.1
Friends/family for a meal once a month40.541.911.94.6
One-week holiday away from home once a year (not with friends or relatives)63.513.719.72.6
Access to a summer cottage one week once a year43.236.313.07.1
A night out once a fortnight27.551.317.83.0
Go to the cinema, theatre or a concert once a month22.754.617.64.5
Dental examination once a year84.17.16.91.5
Medical treatment and medicine If necessary96.81.01.60.4
Private pension insurance45.821.619.010.6
Save at least 500 Swedish Crowns every month53.79.634.61.5
  • a)
    People who cannot afford consumption items/activities that most people have/engage in are suffering from more deprivation than are people who cannot afford things/activities that very few people have/engage in.
  • b)
    An elderly person is not suffering from deprivation if he/she cannot afford things/activities that only young people have/engage in, and vice versa.
  • c)
    A person who lives in a household without children does not score on the index if he/she lacks things that, practically, only households with children have.

Formally, the WDI is calculated in the following way:

  • I = Σγi (P = α + βiχ1 + βiχ2)

where γi tells us whether individual i wants but cannot afford item γ, P is the weight for item γ, estimated as the probability for individual i to have γ given χ1 (his/her age), and χ2 (whether or not she/he is living in a household with children). As is the case for most deprivation of this type, the reliability is very high, with a Cronbach alpha score of 0.86. As can be seen in Table 1, 10.4 per cent of the population is counted as income poor. We use this figure to define a likewise arbitrary poverty line for the deprivation index. Counting those with a deprivation score above 4.36 as poor will result in a corresponding poverty figure of 10.4. In the best of all worlds, the same 10.4 per cent is defined as poor by both measures. However, we do not live in such a world, and only 2 per cent are actually income poor and deprivation poor at the same time. Thus, there are good reasons for using both measures in the following analysis.

Welfare problems: indicators of social exclusion

To analyse how different welfare problems relate to each other, in addition to the poverty measures, 17 indicators of different welfare problems will be used. These indicators range from aspects that tap into areas often discussed in relation to social exclusion, such as neighbourhood characteristics, unemployment and lack of political involvement, but more individual characteristics such as health conditions, loneliness and psychological distress are also included.

Lack of political activity is measured in the following way. Respondents are asked whether they have ‘at any time tried to do something about any deficiency or inaccuracy in your municipality’, and whether they in that case had: contacted any civil servant or other representatives, written a letter to the press or an article in a newspaper, signed any appeal, participated in any demonstration or done anything else to express their view? In addition to this, a more general question was also posed: ‘Considering political issues overall, not only here in the municipality, have you at any time: written a letter to the press or an article in a newspaper, signed any appeal or participated in any demonstration’. Those who answered ‘No’ to all eight questions are defined as politically inactive. This indicator is supplemented with a measure that identifies those who did not vote in the latest general election.

We do not have access to information about housing environment as such, i.e. we lack information related to specific areas. However, we do have information from the respondents that, at least to some degree, can be seen as indicators of area features at the same time as they also indicate individual welfare problems. To indicate what we will call ‘socially disorganised areas’, a concept originally used by Shaw and McKay (1969), we use a set of variables that indicate whether the respondent is living in a suburban area in a larger town or city in Sweden. The indicator shows that the respondent is living in a town with at least 90,000 inhabitants, that the accommodation is rented and located in a block of flats with at least three floors and that the respondent states that damage to housing facilities is common. If all four of these criteria are met, the housing area is defined as socially disorganised (Nilsson & Estrada, 2004). An additional housing variable measures whether the respondent thinks that his/her dwelling is too small, adequate or too big. The answer ‘too small’ is used as an, admittedly subjective, indicator of crowded housing.

We also measure victimisation and concerns about crime. One variable measures whether the respondent has been exposed to any type of theft or property damage during the previous 12 months. A second variable measures whether the respondent, during the previous 12 months, has been exposed to violence or threats of violence. A third variable captures whether the respondent, again during the past 12 months, has refrained from going out during the evening due to fears of being assaulted in any way or whether the respondent is worried about theft and damage to his/her own dwelling.

Health, psychological strain and, to some degree, health behaviour are measured using six different indicators. Health problems are measured via a question that asks whether the respondent suffers from any long-standing illness or handicap that negatively impacts on his/her ability to work or perform daily activities. Additional indicators of health condition and psychological distress are based on questions about occurrence, over the previous two weeks, of sleeping problems, recurrent headaches or migraine and anxiety, worry or anguish. Lastly, we also include two important health predictors: smoking and obesity. Smokers are those who smoke every day, and people are defined as obese if their body mass index exceeds 30.

Loneliness is measured using a subjective indicator. Respondents are confronted with five statements and are asked to pick the one that best describes their situation. The statements are:

  • 1)
    I basically never feel lonely.
  • 2)
    Sure, sometimes I feel lonely, but I don't consider it to be a problem.
  • 3)
    Sometimes I feel lonely and I wish that I could see other people more often.
  • 4)
    I very often feel lonely.
  • 5)
    I almost always feel lonely.

People are considered to be lonely if they agree with statement 3, 4 or 5.

Human capital is indicated by education, and a variable is created that discriminates those who have primary compulsory education as their highest level of education. Economic vulnerability is a condition that, in one sense, is almost synonymous with poverty. We will argue, however, that lack of a cash margin in the household economy, measured as the inability to raise 14,000 SEK (Swedish Crowns) within a week, is an additional indicator of economic vulnerability that may or may not coincide with poverty. We will, lastly, measure unemployment. The measure we have chosen captures those who, during the previous five years, have been unemployed for a total period of at least 12 months.

The distribution of welfare problem indicators is shown in Table 2. The most common problems are political inactivity and long-standing illness. Seven of the indicators are experienced by less than 10 per cent of the population, whereas nine of the welfare problems are experienced by between 10 and 40 per cent of the population.

Table 2.  Basic statistics on poverty and social exclusion indicators.
 NPer cent
Income poor51010.4
Deprivation poor49010.4
Income poor and deprivation poor932.0
Not poor3,83781.3
Did not vote49010.6
Politically inactive2,04641.8
Victimisation: crime1,47530.1
Victimisation: violence3336.9
Worried about crime4519.3
Disorganised area2976.2
Crowded housing72515.0
Health problem1,97840.4
Headache74215.3
Anxiety87218.0
Sleep96119.8
Obesity4068.5
Smoking1,02621.1
Loneliness4809.9
Low education1,05721.6
Lack of cash margin75515.6
Unemployed4669.7

Results

  1. Top of page
  2. Abstract
  3. Poverty and social exclusion
  4. Data and measurements
  5. Results
  6. Conclusions and discussion
  7. References

Table 3 shows the proportion among income poor, deprivation poor and non-poor that experience any of the 17 additional welfare problems. Eleven of the 17 welfare problems are significantly more common among the income poor than among the non-poor, while all 17 welfare problems are significantly more common among the deprivation poor than among the non-poor. In 14 cases, there is also a significant difference between the income poor and the deprivation poor. In all these cases, incidences of welfare problems are higher among the deprivation poor. Thus, poverty seems to be closely related to other types of welfare problems, something that is emphasised when we move from the commonly used income measure to a deprivation-based measure of poverty.

Table 3.  Prevalence of welfare problems among income poor, deprivation poor and non-poor (per cent).
 Income poorDeprivation poorNot poor
  • *

    Indicates significance in relation to the non-poor: * p < 0.05; ** p < 0.001; *** p < 0.0001. Bold figures indicate significant difference (p < 0.001) between income poor and deprivation poor.

Did not vote18.5***20.6***8.5
Politically inactive49.3***48.2**40.5
Victimisation: crime39.3***38.6***28.4
Victimisation: violence8.9* 15.1 *** 5.6
Worried by crime7.4 19.0 *** 8.2
Disorganised area9.0*** 17.1 *** 4.6
Crowded housing18.9** 26.6 *** 13.5
Health problem34.4* 49.4 *** 39.1
Headache14.7 28.0 *** 14.1
Anxiety22.6*** 40.1 *** 14.7
Sleeplessness18.3 36.7 *** 17.8
Obesity7.1 14.2 *** 8.1
Smoking23.5 35.9 *** 19.1
Loneliness12.3*** 27.9 *** 7.6
Low education20.2 26.0 ** 20.8
Lack of cash margin28.3*** 60.5 *** 9.1
Unemployed18.4*** 28.9 *** 7.0

Estimations of the bivariate association between different welfare problems are displayed in Table 4. To more clearly see the pattern of associations between the variables, in Figure 1 we have used the coefficients from Table 4 as input values to a multi-dimensional scaling analysis, estimating Euclidian distances in order to construct two-dimensional visualisations of the results. The fit of this two-dimensional model is not particularly good; the so-called stress value is 0.17, indicating that the data have a structure that is too complex to be represented by two dimensions. The map nevertheless gives an informative first overview of how welfare problems are related to each other. Looking first at the poverty measures, we see that income poverty is placed far out to the left of the map. Income poverty is, in fact, one of the variables used here that has the weakest total relationship to other kinds of welfare problems, and it is also one of the variables that are significantly related to the fewest of the other variables (see Table 4). In this analysis, economic poverty shares this peripheral position with lack of political involvement, crowded housing, exposure to property crime and obesity. For these four indicators, we can find convincing arguments as to why they are only weakly related to other types of welfare problems. The way we measure lack of political involvement discriminates everyone who has not made an active effort to act in a political matter. Thus, people might score on this indicator because they are content with political decisions, not because they lack a political voice. Crowded housing is more affected by household composition than by social exclusion; it is mainly households with children that report lack of housing space. Exposure to property crime is a function not only of the vulnerability related to social exclusion, but also of the extent to which people have properties that are worth stealing (Larsson, 2006). Obesity was included as an indicator of health hazards and is also an indicator that relates strongly to long-standing health problems, but, as the results indicate, obesity is only loosely associated with other welfare problems. Therefore, we have few reasons to be concerned about the weak correlation between these four indicators and other welfare problems. It is, of course, more disturbing that the most commonly used poverty measure discriminates a section of the population that is only marginally connected to other welfare problems.

Table 4.  Bivariate associations between welfare problem indicators: Kendall tau_b.
 Income PoorDeprivation poorDid not votePolitical inactiveVictimisation: crimeVictimisation: violenceWorried by crimeDisorganised areaCrowded housingHealth problemHeadacheAnxietySleeplessnessObesitySmokingLonelinessLow educationLack of cash margin
Deprivation poor0.10**                 
Did not vote0.10**0.11**                
Politically inactive0.06**0.05**0.11**               
Victimisation: crime0.07**0.06**0.01−0.08**              
Victimisation: violence0.03*0.11**0.06**−0.07**0.11**             
Worried by crime−0.020.12**0.04**0.06**0.06**0.04**            
Disorganised area0.04**0.16**0.04**0.000.13**0.07**0.11**           
Crowded housing0.04**0.12**0.05**−0.03*0.08**0.05**0.000.09**          
Health problem−0.05**0.07**−0.00−0.02−0.020.010.12**0.05**−0.01         
Headache−0.010.13**0.06**−0.04**0.04**0.06**0.06**0.07**0.05**0.15**        
Anxiety0.04**0.20**0.09**0.010.05**0.11**0.16**0.09**0.05**0.17**0.17**       
Sleeplessness−0.010.15**0.06**0.010.020.09**0.13**0.10**0.020.19**0.22**0.31**      
Obesity−0.020.08**−0.020.01−0.03*−0.010.03*0.03−0.010.13**0.06**0.010.04**     
Smoking0.020.13**0.07**0.03*0.03*0.07**0.09**0.06**0.020.05**0.06**0.07**0.06**0.01    
Loneliness0.030.21**0.07**0.04**0.010.08**0.10**0.07**0.06**0.06**0.10**0.23**0.13**0.020.04**   
Low education−0.010.05**0.05**0.14**−0.12**−0.06**0.11**−0.00−0.10**0.16**−0.010.05**0.06**0.09**0.11**0.04**  
Lack of cash margin0.12**0.43**0.15**0.07**0.020.09**0.12**0.14**0.08**0.08**0.13**0.18**0.12**0.10**0.15**0.15**0.10** 
Unemployed0.10**0.21**0.11**−0.000.010.05**0.020.030.04**0.000.04**0.07**0.04**0.04**0.07**0.07**0.010.18**
image

Figure 1. Relationship between income poverty and welfare problems. Multi-dimensional scaling based on Kendall tau_b estimates.

Download figure to PowerPoint

Poverty measured in terms of deprivation is placed almost exactly in the middle of the map. Further investigation of Table 4 shows that deprivation poverty is the only variable to be positively and significantly correlated with all the other variables, and the correlations are generally high. Around deprivation poverty we see a cluster of welfare problems that are strongly interlinked and also closely connected to deprivation poverty. There is a cluster of welfare problems that tells us that deprivation poverty often coincides with long-standing health problems, anxiety and sleeping problems and, also, loneliness. Deprivation poverty, to a much higher degree than economic poverty, is also associated with lack of an economic buffer. Living in a socially disorganised area is also connected to deprivation poverty, as are concerns about being subjected to crime. The connection between unemployment and deprivation is also relatively strong.

The analysis presented in Table 4 and Figure 1 not only tells us that deprivation poverty seems to be more closely connected to other types of welfare problems than is income poverty, but also gives us a picture of the kinds of problems that most likely make up the core causes and consequences of poverty, i.e. inability to consume in accordance with the ordinary lifestyle, unemployment, living in disorganised areas, suffering from social and psychological strain and also, to some degree, health problems.

Latent class analysis

The analysis above clearly shows that some types of welfare problems are more closely interlinked than others. It also reveals that the data are too complex to be properly represented by two dimensions. To find out whether there are distinct clusters of welfare problems, a series of latent class analyses was conducted. The analysis identifies clusters that group together persons who share common characteristics. The model classifies cases into clusters based on membership probabilities. It offers a wide range of indicators of model fit and also a decomposed measure, making it possible to evaluate the contribution of single indicators. In the first set of analyses, models were specified that included all 19 welfare indicators (including the poverty measures). The number of clusters was allowed to vary from 1 to 10 in order to determine a stage at which an additional increase in the number of clusters did not lead to a substantial improvement of the model fit. The improvement of the model is indicated by the chi-square-based likelihood value (L2). The L2 value for one cluster model tells us how much total residual variance there is left to explain in the model, variance that theoretically can be explained by additional clusters. These analyses showed that when the one-cluster model was relaxed to a three-cluster model, the L2 value decreased by 20 per cent, from 12,431 for the one-cluster model to 9,887 for the three-cluster model. Thereafter, the L2 value decreased only marginally for each extra cluster added to the model. However, the overall performance of the three-cluster model was quite poor, as indicated by the very modest decrease in L2 value. This was due to the fact that several of the manifest welfare problem variables were only weakly associated with the three latent clusters at the same time as these welfare problems were only weakly related to each other, i.e. they did not form separate clusters when the model was relaxed to allow more than three clusters. To find a simpler model containing only variables that were more closely related to each other, we reduced the number of variables in the model. Looking at the residuals for each variable and the Entrypo-R2 indicating to what degree the latent cluster variable predicted the manifest variable outcome, we started to delete from the model the manifest variables that were most weakly associated with the three clusters, beginning with the one with the lowest R2 value (which happened to be income poverty). After the removal of a variable, we checked the model fit, as indicated by the BIC variable (the L2 value adjusted for model complexity), calculated as BICL2 = L2 – In(N)DF, making sure that the removal actually improved the model. In the end, we ended up with eight manifest measures of welfare problems that were all relatively strongly associated with the three clusters: long-standing health problems, headaches, anxiety, sleeping problems, loneliness, lack of cash margin, unemployment and deprivation poverty.

For these eight variables, a series of analyses was conducted in which the number of clusters was allowed to vary between 1 and 5. Allowing for only two clusters resulted in a considerable improvement of the model. The L2 value decreased from 2,445 to 704. Allowing for an additional third cluster decreased the L2 value to 369. A fourth cluster only marginally improved the model, at the same time as making a substantial interpretation of the clusters increasingly difficult. The first row of Table 5 shows the size of the three clusters estimated from the modal values. Here we can see that 71 per cent of the population ended up in the non-problematic group, i.e. they have low probabilities of scoring on any of the eight welfare problems. Almost 20 per cent belong to cluster 2. The risk of belonging to this cluster is basically determined by the health-related indicators, while unemployment, lack of cash margin and deprivation poverty played a minor role. The third cluster was the smallest, and around 10 per cent of the population was estimated to belong to this cluster. The difference between cluster 2 and cluster 3 is that cluster 3 combines high probability of being unemployed and having economic problems with high probabilities of having health impairments, anxiety, sleeping problems and headaches. Hence, cluster 3 groups together people with a high probability of scoring on all eight welfare problems.

Table 5.  Latent class analysis: cluster size, classification probabilities and model fit.
 Cluster 1Cluster 2Cluster 3
Health problem0.300.680.51
Headache0.070.370.30
Anxiety0.050.490.42
Sleeplessness0.060.570.36
Loneliness0.040.180.29
Lack of cash margin0.060.150.72
Unemployment0.060.070.35
Deprivation poverty0.020.050.75
Cluster size (per cent of population)711910
Df229  
L2369  
BIC(L2)−1566  

In a final analysis, we used a multinominal logistic regression model to estimate the risk for people in different sections of the population to end up in cluster 2 or cluster 3. The independent variables are: equivalent disposable income, age (both these variables are supplemented with a quadratic term) and socio-economic class as indicated by Statistic Sweden's socioeconomic code, which in turn closely resembles the well-known EGP schema (Erikson & Goldthorpe, 1993). Seven different classes were distinguished (see Table 6). People currently not employed or self-employed were classified according to their previous labour market position or, as the second option, according to their spouse's labour market position. If neither of these options were available, the categorisation ‘unclassified’ was used. Also included are household type and gender. The purpose of this analysis was not to make a thorough search for the most important determinates of accumulated welfare problems or to test a particular theoretical model. It was simply a test of the degree to which a number of important stratification variables are related to the cluster we obtained and therefore also helps us to interpret the clusters in a meaningful way.

Table 6.  Multinominal logistic regression analysis. Odds-ratio estimates for cluster belongingness.
 Cluster 2 Odds ratioCluster 3 Odds ratio
Equivalent disposable household income0.9960.717***
Quadratic term1.0000.981***
Age1.016***0.984**
Quadratic term1.0000.998***
Class: Higher white collar as ref group
Unskilled blue collar1.521**5.258***
Skilled blue collar1.2394.004***
Lower white collar1.1413.155**
Middle white collar0.8741.814
Self-employed1.0550.805
Others1.3934.824***
Household: Cohabiting with child(ren) as ref group
Cohabiting without children1.264*1.354
Single adult without children1.567***3.639***
Single adult with children1.673*5.158***
Men (women ref group)0.493***0.639***

Looking first at cluster 2, there is no relationship between income and cluster probability. The probability increases with age, but is only weakly related to social class. There is a slight risk increase among those who live in households without children and among single parents compared with couples with children. Lastly, it is revealed that the odds ratio is significantly lower for men. The analysis confirms that cluster 2 is basically a health cluster that is more or less unrelated to economic circumstances. Cluster 3 is different. Here we see a close connection to income: the higher the income, the lower the cluster probability. The significant estimate for the quadratic term also tells us that the risk of belonging to cluster 3 decreases with increasing speed as income increases. This is, of course, not a surprise, as deprivation poverty and lack of cash margin are strongly related to the cluster. However, the result is interesting when seen against the above results regarding income poverty, as it basically indicates that accumulation of welfare problems is a low income problem, but not so much an income poverty problem. Relating this outcome to earlier findings about the relationship between income distribution and deprivation, the most likely explanation is that the prevalence of income measurement errors is largest at the bottom end of the income distribution. Therefore, we find among the income poor a rather large fraction not suffering from deprivation with regard to consumption of goods and services and, likewise, not particularly exposed to other types of welfare problems (Halleröd, 1997; Halleröd et al., 2006). To further underpin this conclusion, in Figure 2 the probability score for cluster 3 is plotted against disposable income. The estimates are derived from a logistic regression model that includes disposable income and the quadratic term of disposable income. Here we can see that the probability is highest among people living in a low-income household and that the probability peaks just below the poverty line. However, among those living in households with extremely low incomes, the probability risk decreases. Because it is unlikely that the extremely poor are better off than people living in households closer to the poverty line, it is difficult to imagine that the result is caused by anything other than measurement problems. A recent comparative analysis indicates that this problem is not exclusive to Sweden, as the same phenomenon has been seen in two other studied countries: Finland and Great Britain (Halleröd et al., 2006).

image

Figure 2. Estimated probability of belonging to cluster 3 by household equivalent income.

Download figure to PowerPoint

The risk of belonging to cluster 3 is highest among young people and decreases with age with increasing speed, which is shown by the negative coefficient for the quadratic age term. There are large differences between different classes, and the odds ratio for unskilled workers is five times greater than that for higher white-collar workers. The same is true of people living in single adult households with children. Also, single adults without children have a substantially higher risk of falling into cluster 3. The risk for inclusion in cluster 3 is also lower for men than for women. Thus, the analysis supports the interpretation that clusters 2 and 3 are essentially different.

Conclusions and discussion

  1. Top of page
  2. Abstract
  3. Poverty and social exclusion
  4. Data and measurements
  5. Results
  6. Conclusions and discussion
  7. References

The questions addressed here concern the relationship between poverty and a wide range of welfare problems and, in a wider perspective, the relationship between poverty and social exclusion. The analyses were conducted in several steps. Our data allowed us to use both an income-based measure of poverty defining those with an income below 60 per cent of the median income as poor, and a deprivation-based measure defining those who most often have to forgo consumption of goods and services as poor. In addition to these two poverty measures, we also distinguished 17 indicators of welfare problems that covered integration in the political process, neighbourhood conditions, housing conditions, health impairments, anxiety and psychological distress, health hazards, social integration, educational marginalisation, unemployment and economic vulnerability. In the first step of the analysis, the interrelationships between poverty and all 17 welfare problems were estimated. The analysis showed that poverty measured as deprivation was significantly related to each of the other 17 welfare problems. However, the income poverty measure was related to only 11 of the welfare problems, and these relationships were generally comparably weak. In fact, the analysis revealed that income poverty was one of the most peripheral of all welfare problems. Thus, knowing that someone has an income below 60 per cent of the median income does not tell us with any certainty that she/he actually suffers from welfare problems, and we can certainly not claim that fighting income poverty, as we are able to identify it, will also contribute significantly to the fight against other forms of welfare problems. These results are in line with much of the previous research in the area (Berthoud et al., 2004; Bradshaw & Finch, 2003; Callan, Nolan & Whelan, 1993; Halleröd, 1995; van den Bosch, 2001). However, we cannot draw the conclusion that poverty in today's Sweden is a trivial problem. Those who were identified as deprivation poor, i.e. those most unable to consume goods and services in accordance with the general lifestyle in contemporary Sweden, suffered greatly from an accumulation of welfare problems.

What our analysis shows is that the traditional way of measuring poverty as low income is problematic because of a series of unsolved measurement problems that are most likely to affect the lower tail of the income distribution. To estimate poverty correctly based on income data, we of course need correct data on people's access to economic resources from the start. We can be relatively certain that income data are problematic from this point of view. Income from the black economy, income in kind, savings, non-monetary resources etc., cloud the connection between income data and consumption (Behrendt, 2002; Halleröd, 2000). Other problems concern the difficulties in identifying the correct household unit and the fact that equivalence scales adjust for household size in a rather crude way. Time is another important factor. A short poverty spell might be mitigated by the use of savings, and the acquisition of clothes, furniture and other kinds of seldom-consumed items can be postponed (Berthoud et al., 2004; Breen & Moisio, 2004; Gordon, 2005; Layte & Whelan, 2003). The assumption of equal sharing within the household has been increasingly questioned during recent decades (Nyman, 2002; Pahl, 1989; Vogler & Pahl, 1994).

In the next step of the analysis, latent class analysis was used to distinguish a set of closely related welfare problems. In the end, a three-cluster model was derived, based on eight welfare problems. These were long-standing health problems, recurring headaches, anxiety, sleeping problems, loneliness, unemployment, lack of cash margin and poverty measured as deprivation. A cluster that comprises about 10 per cent of the population scored high on all eight of these variables. Thus, when people in today's Sweden suffer from a range of welfare problems, these are the types of problems most likely to affect them. The results presented here are based on cross-sectional data and do not prove any causality between welfare problems. Thus, we cannot know whether poverty causes health problems or whether health problems cause poverty. A realistic guess is that causality goes in both directions. In our opinion, further investigation of the causality between welfare problems is of great importance to social policy discussions. However, the analysis does give an indication of what it is like to be poor in Sweden today.

The results presented here indicate that fighting poverty and social exclusion largely constitutes the same battle if we accept two basic assumptions. The first is that we need to measure poverty in a more accurate way than is usually the case, i.e. we cannot rely on available income data only. Second, the approach to social exclusion that we have applied is largely data driven. Based on our analysis, we can argue that the socially excluded in today's Sweden are poor and unemployed and that they also experience health problems, loneliness and psychological distress. The socially excluded, thus, are excluded from the labour market and from ordinary consumption of goods and services, and in addition they suffer from more individual problems such as ill health and psychological distress. However, aspects that are often looked upon as central in the debate on social exclusion – such as exclusion from the political arena and education system, and spatial segregation – seem to be only weakly related to each other and to the core problems identified in this study, making the relation between poverty and social exclusion less straightforward than is usually assumed. Our hope is that the findings presented here can serve as fuel not only for continued empirical analysis of social exclusion, but also for the theoretical debate on the concept of social exclusion.

Regarding further empirical studies, we see two important issues. First, the kind of analysis conducted here needs to be extended to include other countries. The EU-SILC (Survey of Income and Living Conditions) database will hopefully provide the information necessary for such an analysis. Second, we need to understand the social mechanisms that link poverty and different kinds of welfare, i.e. we need to understand the process of social exclusion and, of course, also the processes that lead to social inclusion. This is, on all levels, a challenging endeavour for social science.

Footnotes
  • 1

    The equivalence scale gives a weighting of 1.16 to the first adult, 0.76 to the second adult and 0.96 to any additional adults (18 years or older). Children between 11 and 17 years are given weightings of 0.76, 4–10 years, 0.66, and children below 4 years are given a weighting of 0.56.

References

  1. Top of page
  2. Abstract
  3. Poverty and social exclusion
  4. Data and measurements
  5. Results
  6. Conclusions and discussion
  7. References
  • Bask M (2005). Welfare problems and social exclusion among immigrants in Sweden. European Sociological Review 21: 7389.
  • Behrendt C (2002). At the Margins of the Welfare State. Aldershot, Ashgate.
  • Berthoud R, Bryan M, Bardasi E (2004). The dynamics of deprivation: the relationship between income and material deprivation over time. Report No. 219, Department for Work and Pensions, London.
  • Bradshaw J, Finch N (2003). Overlaps in Dimensions of Poverty. Journal of Social Policy 32: 513525.
  • Breen R, Moisio P (2004). Poverty Dynamics Corrected for Measurement Errors. Journal of Economic Inequality 2: 171191.
  • Callan T, Nolan B, Whelan CT (1993). Resources, Deprivation and the Measurement of Poverty. Journal of Social Policy 22: 141172.
  • Erikson R, Goldthorpe JH (1993). The Constant Flux; A Study of Class Mobility in Industrial Societies. Oxford, Clarendon Press.
  • Erikson R, Tåhlin M (1987). Coexistence of Welfare Problems. In: Erikson R, Åberg R, eds. Welfare in Transition. Oxford, Clarendon Press.
  • Fløtten T (2005). Poverty and Social Exclusion – Two Sides of the Same Coin? A Comparative Study of Norway and Estonia. Oslo, Fafo.
  • Gallie D, Paugam S, Jacobs S (2003). Unemployment, poverty and social exclusion: Is there a vicious circle of social exclusion? European Societies 5: 132.
  • Gordon D (2005). The Concept and Measurement of Poverty. In: Pantazis C, Levitas R, eds. Poverty and Social Exclusion in Britain: The Millennium Survey. Bristol, The Policy Press.
  • Halleröd B (1991). Den svenska fattigdomen[The Swedish Poverty]. Lund, Arkiv.
  • Halleröd B (1995). The Truly Poor: Indirect and Direct Measurement of Consensual Poverty in Sweden. Journal of European Social Policy 5: 111129.
  • Halleröd B (1997). Fattigdom och materiell deprivation: En empirisk analys [Poverty and Deprivation: An Empirical Analysis]. In: Vogel J, Häll L, eds. Välfärd och ojämlikhet i 20-års perspektiv 1975–1995. Örebro, Statistiska Centralbyrån.
  • Halleröd B (2000). Socialbidragstagande och fattigdom [Social Assistance and Poverty]. In: Puide A, ed. Socialbidrag i forskning och i praktik. Göteborg, Gothia.
  • Halleröd B, Gordon D, Larsson D, Ritakallio V-M (2006). Relative Deprivation – A Comparative Analysis of Britain, Finland and Sweden. Journal of European Social Policy 16: 328345.
  • Halleröd B, Heikkilä M (1999). Poverty and Social Exclusion in the Nordic Countries. In: Kautto M, Fritzell J, Hvinden B, Kvist J, Uusitalo H, eds. Nordic Social Policy. London, Routledge.
  • Hills J, Le Grand J, Piachud D (2002). Understanding social exclusion. Oxford, Oxford University Press.
  • Kangas O, Ritakallio V-M (1998). Different Methods – Different Results? Approaches to Multidimensional Poverty. In: Andreß HJ, ed. Empirical Poverty Research in a Comparative Perspective. Aldershot, Ashgate.
  • Larsson D (2006). Exposure to crime as a consequence of poverty. Umeå, Umeå University, Department of Sociology.
  • Layte R, Whelan CT (2003). Moving in and out of poverty – The impact of welfare regimes on poverty dynamics in the EU. European Societies 5: 167191.
  • Mack J, Lansley S (1985). Poor Britain. London, George Allen & Unwin Ltd.
  • Nilsson A, Estrada F (2004). Exposure to threatening and violent behaviour among single mothers – the significance of lifestyle, neighbourhood and welfare situation. British Journal of Criminology 44: 168187.
  • Nordenmark M (1999). Unemployment, Employment Commitment and Well-Being. Umeå, Umeå University, Department of Sociology.
  • Nyman C (2002). Mine, yours or ours. Umeå, Umeå University, Department of Sociology.
  • Pahl J (1989). Money and Marriage. London, Macmillan Education Ltd.
  • SCB (2003). Inkomstfördelningsundersökningen[Income Distribution Survey]. Örebro, Statistics Sweden.
  • Shaw CR, McKay HD (1969). Juvenile delinquency and urban areas; a study of rates of delinquency in relation to differential characteristics of local communities in American cities. Chicago, University of Chicago Press.
  • Strandh M (2000). Varying Unemployment Experience? The Economy and Mental Well-Being. Umeå, Department of Sociology, Umeå University.
  • Tham H (1994). Ökar marginaliseringen i Sverige? [Is Marginalization in Sweden Increasing?]. In: Fritzell J, Lundberg O, eds. Vardagens villkor [Conditions of daily life]. Stockholm, Brombergs.
  • van den Bosch K (2001). Identifying the Poor. Aldershot, Ashgate.
  • Vogler C, Pahl J (1994). Money, Power and Inequality within Marriage. The Sociological Review 42: 263288.