Becoming Poor in Belgium and Britain: the Impact of Demographic and Labour Market Events
by Caroline Dewilde
University of Antwerp
Sociological Research Online, Volume 11, Issue 1,
Received: 8 Sep 2005 Accepted: 13 Jan 2006 Published: 31 Mar 2006
In this paper, we model the impact on the entry into poverty of a range of demographic and labour market events. Our analyses are based on longitudinal panel data from two countries belonging to a different regime type – Belgium and Britain. The results show that while in Belgium the impact of most life events is relatively small, in Britain most demographic and labour market events significantly raise the chances of becoming poor. We link the observed poverty entry patterns to the ways in which economic welfare in Belgium and Britain is distributed between the three main systems of resources distribution: the welfare state, the labour market and the family. We furthermore find that the combined influence of the interrelated parts of the welfare regime on the role of women in the household economy is a potentially important explanatory factor.
Keywords: Life Events; Longitudinal Research; Poverty Dynamics; Welfare Regimes
Introduction1.1 During the last decade, the increasing availability of long-running and representative household panel data has resulted in both theoretical and methodological advances. At the theoretical level, by tracing households and individuals over time, household panel studies offer the possibility to start disentangling the complicated links between large-scale social and economic change and the micro-level life courses of families and individuals (Rose, 2000). At the methodological level, the incorporation of the variable ‘time’ into a complex data structure has resulted in more dynamic ways of thinking about social problems, in innovative approaches concerning data manipulation and in the application of more sophisticated methods, such as event history analysis (Ruspini, 2002).
1.2 Poverty research has perhaps been the main ‘beneficiary’ of this new research tradition. The gradual shift from cross-sectional poverty indicators to longitudinal trajectories has engendered several new and important insights, both from a theoretical, methodological and social policy point of view. For instance, early studies have shown that mobility into and out of poverty is much larger than expected on the basis of traditional stereotypes about ‘the poor’ (Duncan, 1984; Jarvis and Jenkins, 1995). Furthermore, in a much cited article, Bane and Ellwood (1986) demonstrated that, while most respondents entering poverty are poor for only a short time, a large proportion of the panel members identified as poor at one particular moment in time are in the middle of a long poverty spell. This finding has important consequences, not only in terms of the way researchers and society in general think about poverty, but also for the development of more effective social policy and anti-poverty measures. In the next phase, attention was mainly focused on the identification of different longitudinal poverty ‘patterns’ or ‘profiles’ and their correlates (Ashworth, et al., 2000; Walker, 1998) and on the determinants of entering/exiting poverty spells (Jenkins, 2000).
1.3 Although our knowledge of the processes that lead people into and out of poverty has been greatly enhanced, we can nevertheless formulate some general considerations and possible avenues for future research. Firstly, most studies are of a rather ‘empiricist’ nature. Usually, a range of life events is related to the transition into and out of poverty, but this alone does not allow us to explain why only certain individuals and families become poor at given times. This problem is partly caused by the way in which life events are usually measured. Due to low numbers, studies often use broad categories such as ‘changes in needs’ or ‘changes in the number of workers’, which refer to several life events at once and cannot be related to a specific life stage.
1.4 Secondly, in recent years a lot of researchers have embarked upon cross-national comparisons of poverty dynamics, trying to link the ‘macro-level’ literature on welfare regimes to the micro-level processes observed at the household or individual level. However, the welfare regime often figures as a ‘black box’ variable in the background, ‘explaining’ the identified cross-national variations. This of course follows from the conceptual and methodological problems researchers are confronted with when trying to link institutional processes to individual outcomes.
1.5 As a result of the above, few studies attempt to link specific institutional arrangements on a certain domain of welfare to the entry into and out of poverty following specific life events. However, if the goal of this type of research is to formulate policy recommendations or at least to suggest avenues for more policy-oriented research, we need more detailed analyses.
1.6 Apart from these conceptual problems, some additional methodological considerations can be made. One often gets the impression that the operationalisation of the relationship between life events and poverty mobility is rather tautological. For instance, in a recent paper Layte and Whelan (2002: 19) assert that ‘changes in the level of income are of vital and prime importance’ when explaining the transition into and out of income poverty. In other studies as well, labour market events are often conceptualised as ‘income events’, resulting in a model in which both the independent variable (a change in labour market income) and the dependent variable (falling under the income poverty line) are, in fact, based on the same information. Furthermore, since recent research has shown that a significant proportion of observed income poverty mobility is due to measurement error (Breen and Moisio, 2003), it is not impossible that measurement error on the independent variable is related to measurement error on the dependent variable.
1.7 In this article, we try to take some steps in the right direction by analysing the impact of a range of demographic and labour market events on the entry into income poverty in Belgium and Britain, two countries characterised by a different welfare regime. Although non-monetary indicators reveal an important part of poverty reality, the analyses in this article are based on the ‘traditional’ relative income poverty measure. We use data from the Panel Study on Belgian Households (PSBH), 1992-1998 and the British Household Panel Survey (BHPS), 1991-1999.
1.8 The outline of this article is as follows. In the first section we devote some attention to the problem of linking institutions to individual lives and clarify our approach. We then proceed with a discussion on the conceptualisation and measurement of poverty. We conclude with a review of the welfare regimes in Belgium and Britain, and the ways in which the poverty risks following several life events are dependent on the interrelations between the different parts of the welfare regime. In the third section, the data and methods are discussed, followed by the results of the statistical analysis. In the final section we summarise our findings and formulate some general conclusions.
Linking institutions to individual lives2.1 While over their life-course, most people are confronted with most of the life events stated to influence poverty dynamics, so far little attention has been devoted to explaining why only certain individuals and families enter (exit) poverty following these events. In an earlier publication (Dewilde, 2003), we evaluated to what extent different sociological perspectives on the life-course can be integrated into a general framework for the study of poverty dynamics. Next to other influences such as the earlier life-course trajectories and the interrelationships within the family, the prevailing welfare regime was identified as an important variable explaining cross-national variations.
2.2 In this article, our main interest lies with the ways in which the impact of specific demographic and labour market events on the entry into poverty is related to the welfare regime in Belgium and Britain. As stated in the introduction, the impact of the institutional context is not easily measured. A first issue concerns the conceptualisation of the notion ‘welfare regime’. As it seems, most researchers still take the ‘original’ welfare state typology – or some adaptation of this typology – derived by Esping-Andersen (1990) in his influential publication The Three Worlds of Welfare Capitalism as their point of departure. Concluding that an approach solely based on social expenditure is unsatisfying, Esping-Andersen developed a welfare state typology based on three dimensions: the relationship between the state and the market in the deliverance of welfare, the stratifying effects of the welfare state, and the quality of social rights. Central to this last dimension is the extent of de-commodification or independence of market participation. Considering cross-national variations along these dimensions, the author identified his well-known ‘Three Worlds of Welfare’: the social-democratic welfare state, the liberal welfare state and the conservative-corporatist welfare state.
2.3 The social-democratic welfare state is characterised by universal social rights, extensive protection against social risks and generous flat-rate benefits. The state strives to guarantee full employment and comprehensive public services. The result is a high level of de-commodification. We find this model in the Scandinavian countries. The liberal welfare state is, first of all, residual: eligibility is confined, universal benefits are low and means-tested social assistance is preferred. Market-solutions for social risks are encouraged. This model exists primarily in the Anglo-Saxon countries. It is characterised by a low level of de-commodification. The conservative-corporatist countries are characterised by a medium level of de-commodification. The distinctive nature of this model concerns the differentiation of social rights according to class and status, and a preference for the traditional male breadwinner family. This model is dominant in most continental European countries.
2.4 The ‘Three Worlds’-typology of Esping-Andersen engendered a storm of reactions. Most comments relate to the way in which Esping-Andersen operationalises the concept of de-commodification (e.g. Bonoli, 1997; Castles and Mitchell, 1993), to the position of the family and women in the welfare state (e.g. O’Connor, 1996; Orloff, 1996; Sainsbury, 1994) and to the identification of additional clusters (e.g. Ferrera, 1996).
2.5 Without elaborating on the details, most authors conclude that the Southern European countries – Italy, Greece, Spain and Portugal – can be grouped into a separate cluster (Andreotti, et al., 2001). These rudimentary welfare states are underdeveloped and residual. The family, as the corner stone of society, carries the burden of unprotected risks. Ferrera (1996) justifies the identification of a fourth cluster by pointing out a number of common traits: a highly fragmented and corporatist system of income maintenance, a health care system based on universalistic principles, a low degree of state influence in the welfare sector and a highly collusive mix between public and private actors and institutions, and finally, the persistence of clientelism and patronage, resulting in the selective distribution of cash benefits.
2.6 Though useful, this typology should not be considered as ‘definite’: in most countries, the institutional set-up deviates more or less from the ideal-type and important differences exist between countries belonging to the same welfare regime. Therefore, different authors sometimes classify the same country as belonging to a different welfare regime. The Netherlands, Italy and – to a lesser extent – Ireland are generally known as doubtful cases. Furthermore, the accumulation of policy changes over a longer period of time is possibly of such a nature that for some countries we can identify a ‘regime change’.
2.7 During the last decade, all this research activity has resulted in a gradual widening of the concept of ‘welfare state’ to ‘welfare regime’, in his later work defined by Esping-Andersen as ‘the combined, interdependent way in which welfare is produced and allocated between state, market and family’ (Esping-Andersen, 1999: 35). Concerning the topic of this article, the entry into income poverty following specific life events, we are particularly interested in the interrelations between the family and the other systems of resources distribution: being poor is generally considered as a condition of households, and the chances of becoming poor are influenced by the positions, limitations and opportunities on several life domains of all household members. Therefore, in our description of the welfare regime in Belgium and Britain and the interpretation of our results, we are particularly interested in the role of women in the household economy.
2.8 A second issue when trying to link welfare regimes to individual life courses concerns the methodological pitfall of how to measure the impact of institutions. In the empirical literature we roughly find three ways of accounting for the institutional context. In the more policy-oriented research tradition, the impact of institutions is often gauged through simulation. For instance, the poverty-reducing effects of welfare state policies are ‘calculated’ by comparing pre- and post-transfer poverty (e.g. Fouarge, 2002; Goodin, et al., 2000). The problem with this approach is however that we are dealing with a so-called counterfactual situation, meaning that ‘pre-transfer poverty’ is a theoretical concept: we do not know how individuals would act if the welfare state did not exist. A further problem with this approach is, as we stated before, that being poor or not poor not only follows from welfare policies alone, but rather results from a complex set of influences, which are difficult to simulate.
2.9 A second way to ‘formally’ estimate the impact of institutions is by including dummy variables referring to ‘welfare regime types’ in a statistical model (e.g. Muffels and Fouarge, 2004). However, as the authors recognise themselves, it is entirely possible that so-called ‘regime effects’ cannot be primarily attributed to the prevailing welfare regime, but arise from the whole range of economic, social, political and physical assets a society possesses. Thus, with this approach as well it is difficult to entangle how the interrelations between the different parts of the welfare regime influence individual lives.
2.10 In this article we therefore use the more ‘deductive-explorative’ comparative case-study approach recently advocated by Mills and Blossfeld (2004). In their work on the impact of globalisation on the individual life course, these authors develop a ‘multi-level’ framework in which empirical expectations are deduced from theory based on the assumption that the influence of macro-level changes on individual lives is ‘institutionally filtered’ and thus impacts differently on different social groups in different countries or welfare regimes. In this article, we follow a similar strategy: starting from a description of the institutional context in both countries, we formulate some general hypotheses concerning the link between the interrelating parts of the welfare regime in Belgium and Britain and the chances of becoming poor following specific life events.
Poverty in the welfare state
2.11 Much has been written on the conceptualisation and measurement of poverty. The waning of post-war optimism and the realisation that even an extended welfare state cannot prevent many people being confronted with shorter or longer periods of poverty during their life course has contributed to the regular resurfacing of the concept of poverty, often accompanied by ‘new’ qualifications such as ‘(relative) deprivation’, ‘social exclusion’ or ‘underclass’. Furthermore, poverty is essentially both a political and normative concept, implicating that all poverty measures are in one way or another based on value judgements (Engbersen, 1998). Therefore, poverty researchers should make explicit the choices they make and ensure that their results have been obtained through a verifiable procedure (Andreß, 1998).
2.12 Notwithstanding the lack of a general conceptual framework, most poverty researchers do agree on some points. Firstly, poverty should be understood as a social and relative concept, depending on the general standard of living in a society. Many authors still revert to the poverty definition introduced in the 1970s by Townsend (1979: 31): ‘Individuals, families and groups in the population can be said to be in poverty when they lack the resources to obtain the type of diet, participate in the activities and have the living conditions and amenities which are customary, or at least widely encouraged, or approved, in the societies to which they belong. They are, in effect, excluded from ordinary living patterns, customs and activities’.
2.13 Secondly, most researchers agree that poverty is a multidimensional concept. We can distinguish between the multi-aspectual nature of poverty and the longitudinal dimension (Dewilde, 2004). Multi-aspectuality refers to the height, the width and the depth dimensions. Poverty can manifest itself on several life domains (width) and in different gradations of severity (height). The depth dimension refers to the accumulation of problematic situations on several life domains. This conceptualisation takes account of the reality that ‘the poor’ are in fact a heterogeneous group of people faced with one or several problems and who ended up in poverty following different events and/or trajectories. Furthermore, longitudinal data have exposed the transitory nature of poverty: most households who fall below the poverty line actually manage to escape relatively quickly from this situation (Bane and Ellwood, 1986).
2.14 Despite the general agreement on the multidimensional nature of poverty, the literature has been dominated by discussions on its operationalisation. The most important debate has taken place between those who favour a direct measurement and those who call for an indirect approach (Ringen, 1988). In the first case, poverty is measured directly using information on living standards or consumption. In the latter case, poverty is measured indirectly based on the resources people dispose of, with income as the usual and only indicator. This discussion has been animated further by the repeated finding that different methods classify different population groups as ‘poor’ (Kangas and Ritakallio, 1998; Whelan, et al., 2002).
2.15 Although non-monetary indicators provide essential information on the everyday reality of living in poverty, in this article we focus on the monetary dimension. In a society where most goods and services are commodified, income is by far the most important determinant of living standards. We use a relative income poverty line, identifying respondents as ‘poor’ when their household income is lower than a certain proportion (60%) of median income. The main advantage of this poverty measure is that it is relative ‘by definition’. This relativity applies in several ways. Firstly, the financial situation of the respondent is compared to ordinary living standards at a certain place and time. Secondly, the relative poverty line is self-updating, as changes in living standards over time are taken into account (Atkinson, et al., 2002).
Poverty Risks in the Belgian and British Welfare Regimes
2.16 In this section we briefly consider the impact of the following systems of resources distribution on poverty entry risks: the welfare state, the labour market and the family.
2.17 In the aftermath of the Second World War, British social policy was based on the same Beveridgean principles as the early Scandinavian model (Korpi and Palme, 2003): universal basic benefits, a universal health care system and a political commitment to full employment. However, from the 1970s onwards, the United Kingdom started to follow a more liberal course (Esping-Andersen, 1999). Basic benefits were never upgraded to an adequate replacement level, increasingly pressing the middle classes to resort to private insurance. This change of course was reinforced by the succession of conservative governments since 1979: in a recent study Korpi and Palme (2003) find that 18 years of conservative domination have brought some British welfare programs back to a pre-Beveridgean level of social protection. The electoral victory of New Labour in 1997, leading to a change of power, did not result in a major reversal of existing social policies.
2.18 Though generally considered as a liberal welfare state, social expenditure levels in the United Kingdom are not much below the levels reached by ‘high-spending’ European countries such as France or Belgium and are much higher than the figure for the United States. Therefore, Castles and Mitchell (1993) consider the United Kingdom as a ‘radical’ liberal welfare state, aimed at the targeted redistribution of resources in order to prevent poverty. Although replacement levels of most social benefits are low, a whole range of means-tested additional measures are targeted at the low-income groups, Housing Benefit being the most important one. Nevertheless, income poverty and income inequality in the United Kingdom are among the highest in Europe (EU-15) (Dennis and Guio, 2003). Furthermore, this type of welfare state produces a dual society, leaving the lower social groups to rely on less generous transfers, while the middle and higher classes ensure their welfare through private channels.
2.19 A particular case in point is the provision of childcare services, which is mainly left to the market. The number of public childcare places is accordingly very low. This gap is however not filled by the market: the private childcare sector in Britain is both relatively small and expensive (Esping-Andersen, 1999). This has important consequences for the labour market trajectories of British women, which are often interrupted because of child rearing-responsibilities. There is however an important difference by education. Based on their simulation of life-time income, Joshi and Davies (2002: 124) distinguish two groups of women: ‘On the one hand are women who cannot afford market childcare, and on the other hand are those who cannot afford not to use it’.
2.20 The Belgian welfare state is situated on the boundaries of the conservative and social-democratic regime. On the one hand, social and fiscal policy is based on the traditional family model with a male breadwinner. In a comparative perspective, female labour market participation is rather low, resulting in a relatively high number of single-earner households. For the younger generations, this traditional family pattern is mainly limited to the low-skilled. Benefits are furthermore differentiated according to professional status and related to the employment record. Additionally, the so-called ‘social partners’ – union and employers – play a central role in the administration and management of social security services (De Lathouwer, et al., 1999).
2.21 On the other hand, the Belgian welfare state guarantees an exceptionally high level of de-commodification, comparable to the Scandinavian countries (Esping-Andersen, 1990). Although social rights are linked to employment, over the decades there has been a substantial loosening of this link. A particular case in point is the unemployment insurance: although in a comparative perspective replacement rates are quite low, coverage is very high, duration is unlimited and amounts are less strongly dependent on the former wage (De Lathouwer, 1996; De Lathouwer, et al., 2003). Combined with a strong ‘modularisation’ of benefits according to family type, this uncoupling between employment record and social rights results in relatively low income poverty figures, both for dual-earner and single-earner households (Cantillon, 1999), be it to a lesser extent for the latter group.
2.22 The Belgian welfare state furthermore takes on a significant part of family care responsibilities: social services provision, in particular relating to childcare, is relatively high, facilitating the combination of work and family for women. Although the percentage of women aged 18 to 60 with a full-time job is higher in Britain (43,8%) than in Belgium (38,2%), the number of full-time working mothers with young children (<6 years) is much higher in Belgium (41,6% compared to 19,4% in Britain) (own calculations PSBH and BHPS, 1995).
2.23 In Britain, the deregulation of the labour market and the introduction of more flexibility in the employment relationship has resulted in a stagnation of real wages and growing income inequality (Esping-Andersen, 1999). Although (long-term) unemployment among the young and the low-skilled is relatively low, they are often employed in lower-quality service sector jobs and have to rely on an income below the poverty line, hence the term ‘working poor’. Also, women are more likely to hold small part-time jobs, with no social rights attached (Cook, et al., 2001).
2.24 Despite the introduction of some flexibility regarding working times and employment protection during the 1980s and 1990s, the Belgian labour market is relatively regulated (Delsen, 1997). Regulation concerning fixed-term labour contracts is fairly strict, and small part-time jobs are virtually non-existent (De Lathouwer and Marx, 2002). Long-term unemployment in Belgium is rather high and mainly caused by the low probabilities of exiting unemployment (De Lathouwer, et al., 1999). Furthermore, many older people losing their jobs never return to the labour market. For this group of discouraged workers, collectively negotiated early retirement schemes often provide in a generous replacement income.
2.25 Concerning the desinstitutionalisation of the family, Western-Europe seems to be divided in a ‘progressive’ North, with the Scandinavian countries, closely followed by Britain, assuming a pioneering role, while the countries belonging to the Mediterranean welfare regime remain strongly wedded to the traditional family patterns (Ditch, et al., 1998; Iacovou, 1998). The so-called ‘conservative-continental’ countries form an intermediate group. For instance, while the crude marriage and divorce rates in Belgian and Britain are roughly equal, the number of people living in a consensual union, the number of births out of wedlock and the number of lone parent families is larger in Britain.
2.26 This process of desinstitutionalisation can be interpreted as the result of both cultural and institutional factors. Concerning the cultural influences, religion turns out to be an important determinant of public policies such as divorce laws (Castles, 1994; Castles and Flood, 1993). Examples of institutional factors are for instance social policies for families with young children and the resulting female labour market participation patterns.
2.27 An important cultural difference between Belgium and Britain is the way in which parental responsibilities for adult children are defined. According to Gallie and Paugam (2000), in Belgium the relative autonomy model dominates: children live with their parents as long as their financial situation does not permit them to set up their own household. Co-residing is considered as a temporary situation, and parents expect the young adult to actively look for a job and thus, independence. In Britain the advanced autonomy model is advocated: self-realisation is considered to be impossible without achieving some degree of autonomy. Thus, although youngsters can to some extent be financially dependent on their parents, they are expected to establish their own household, often with friends or other non-related adults.
2.28 Based on this literature review, we formulate the following general expectations:
- The impact of both demographic and labour market events on the entry into poverty is stronger in Britain than in Belgium, where both the family and the welfare state take on a greater responsibility for negative life-course risks.
- The higher provision of welfare state services such as childcare offers more possibilities for Belgian women, especially mothers, to adapt their labour market participation in order to shield themselves and their families from the poverty risks associated with different types of demographic and labour market events.
Data and Methods
Data3.1 We use data from the Panel Study on Belgian Households (PSBH), 1992-1998 and from the British Household Panel Survey (BHPS), 1991-1999. Both are household panel studies, tracing an initial sample of households and individuals over time, gathering information on a yearly basis using standardised questionnaires. Data on different life domains are collected on the household and individual level.
3.2 Both the Belgian and the British panel are representative for the population in private households. The PSBH started in 1992; in the first wave 4.438 households and 11.332 adults and children were interviewed. In 1998, 859 additional Flemish households were included in the sample. The BHPS started in 1991. The initial sample contained 5.511 households; 10.265 adult interviews were completed. Since the start of the BHPS, several subsamples have been added to the survey. In this article, the ‘original’ 1991-panel is analysed. For both panels, weighting coefficients are calculated to minimise the impact of different sources of bias, in particular selective attrition (Dewilde, et al., 2000; Taylor, et al., 2001).
3.3 In this article we are interested in the impact of different types of life events occurring between wave(t) and wave(t+1) on the probability of entering income poverty in wave(t+1). Since the number of observed life events between each pair of consecutive waves is typically very small, longitudinal analysis of panel data is often based on a pooled dataset, collecting the available information from each comparison of waves (Sandefur and Tuma, 1987). In this way, cell frequencies are maximised and more robust estimates can be obtained.
3.4 Following common practice in poverty research, our pooled file is defined at the individual level and contains all respondents living in a co-operating household. Because of the central importance of the gender dimension in the study of social exclusion and poverty, we also estimate the impact of different demographic and labour market events for men and women separately. Both subsamples are limited to those individuals who are classified as either the household reference person or his/her partner in wave(t) or wave(t+1).
Demographic and Labour Market Events
3.5 The following demographic events are considered:
- partnership dissolution (marriage or cohabitation)
- partnership formation
- birth of a child
- formation of an independent household by an adult child (≥16 years)
- adult child (≥16 years) leaving the parental household
- ‘other’ change in the composition of the household
- transition from employment to ‘forced’ inactivity (unemployment, disability)
- transition from employment to ‘unforced’ inactivity (temporarily interrupted, education, inactivity, other)
- transition from full-time to part-time employment (<30 hours per week)
- (early) retirement
3.7 In each model, the age, sex and educational level of the household reference person, and the number of children (<16 years) in wave(t) are included as control variables. Concerning the models for the subsamples of men and women, we distinguish between the respondent’s own labour market transitions and his or her partner’s.
3.8 In both panels several measures of ‘net disposable household income’ (post-tax post-transfer) are available. In the PSBH the current income measure is based on the estimation of total household income by the household reference person. This measure is available for all waves; the response rate amounts to 90% or more for each wave. Due to changes in the questionnaire, the annual income measure, referring to the sum of all incomes from all sources for all household members during the last calendar year, is only available from wave3 onwards. Response rates for this ‘calculated’ income fluctuate around 75%. In the BHPS, both the current and annual net incomes are calculated from the raw data by ESRC-staff and provided as an addition to the main BHPS-dataset (Bardasi, et al., 2001). Response rates for both measures fluctuate from 80 to 84%.
3.9 Since the results for both types of income measures are largely similar, we only report the figures based on the current income measure. To adjust for differences in the size and composition of households, we use the modified OECD-equivalence scale. Our relative income poverty line is set at 60% of median population income.
3.10 Since the dependent variable only has two categories, ‘poor’ and ‘not poor’, the basic assumptions on which the general linear model is based, more specifically normality and homoscedasticity of the error terms, are violated. Furthermore, applying the linear model to a binary dependent variable often results in predicted values lower than 0 or higher than 1, where the true probabilities are restricted to the [0,1] interval. We can avoid these problems by using a logit model. In this model, the independent variables are related to the natural logarithm of the odds of the dependent variable. The parameters of the logit model are estimated by means of the maximum likelihood procedure (Menard, 1995; Pampel, 2000).
3.11 A further problem is caused by the fact that, due to our complex data structure, not all observations are independent: individuals are clustered within households, while most individuals contribute more than one ‘comparison of waves’ to the dataset. This clustering of observations generally results in an underestimation of standard errors (Allison, 1999). Therefore, standard errors reported in this article are robust, taking into account the clustering of observations at the level of the household and over time.
3.12 For each model we report the global x2-test, evaluating the null hypothesis that the coefficients for all independent variables equal 0. We also report the generalised R2 (Allison, 1999). This measure varies from 0 to 1 and is based on the same rationale as the conventional R2 in linear regression. However, it cannot be interpreted as an indication of the explained variance.
3.13 The models predicting the poverty entry rates are limited to the relevant population ‘at risk’: only the respondents who are classified as ‘not poor’ in wave(t) are at risk of an entry into poverty between wave(t) and wave(t+1).
Results4.1 The results for the logit models predicting the entry into poverty for the total sample are presented in Table 1. Table 2 (Belgium) and Table 3 (Britain) contain the results for the subsamples of men and women.
4.2 Concerning the effects of our control variables, in both Belgium and Britain the respondents living in a household with a female or lower educated reference person have a higher chance of entering income poverty. While the effect of living in a female-headed household is stronger in Britain, the effect of education is stronger in Belgium. This is possibly linked to the fact that the coupling between education and the labour market is stronger in Belgium, where the educational system is more stratified. Additional dependent children result in higher entry probabilities in both countries, although the effect of this variable is stronger in Britain. Compared to the respondents living in a household with an older reference person (>49 years), the respondents with a reference person aged 26 to 49 have a significantly lower chance of becoming poor.
4.3 The effects of widowhood on the chances of entering poverty are mainly limited to Britain, and negatively impact on the financial situation of both men and women. In Belgium, where most pensioners are dependent on an earnings-related social security pension, there is some evidence that widowhood has a negative impact on the economic situation of widows, but not of widowers: none of the men at risk enter poverty upon widowhood. Further analyses of the British data show that in particular those respondents who are dependent on a state pension are vulnerable to an entry into poverty. Both before and after this event, respondents who enter poverty following widowhood are to a larger extent dependent on state transfers compared to those who do not become poor. This dependency furthermore increases upon widowhood, with 81,3% of income based on state transfers and only 7,0% on occupational or private pensions. The drop in private pension resources from 19,9% to 7,0% suggests that, in spite of widow protection legislation, not all private arrangements adequately provide for the remaining spouse.
4.4 In both countries partnership dissolution has a positive and strongly significant effect on the probability of becoming poor. However, in line with other research, the effects are gender-specific: the economic burden of partnership dissolution mainly falls on the shoulders of women – and the children who live with them. The ‘positive’ effect for men in both countries might be overstated, since the alimony and maintenance payments are not deducted from total household income. However, other research for Britain (Jarvis and Jenkins, 1999) shows that this does not change the general conclusion. Contrary to our second hypothesis, the negative effect of partnership dissolution on the financial situation of women seems somewhat stronger in Belgium than in Britain. Although the effect of partnership dissolution is comparable for both countries, the absolute level of income poverty following this event is much larger in Britain. For the Belgian women experiencing partnership dissolution, poverty rates increase from 8,1% to 20,3% between wave(t) and wave(t+1). The corresponding percentages for Britain are 17,9 and 35,7.
|Table 1. Results from the pooled logit model predicting the entry into income poverty (60%-poverty line, individuals at risk, weighted)|
4.5 The relatively strong effect for the Belgian women can be related to their higher propensity to set up an ‘independent’ household (87,5% compared to 78,8% in Britain). In spite of the fact that the British family model stresses autonomy to a higher extent, more British women (13,7% versus 6,2% in Belgium) find shelter in the parental household or some ‘other’ household arrangement. This finding is in line with our second hypothesis and suggests that Belgian women seem to have more opportunities to stay independent after divorce. It also shows that not only the economic consequences of life events vary cross-nationally, but also the choices people make within a given opportunity structure and the influence of these choices on their economic situation. The number of women living with a new partner in wave(t+1) is roughly equal in both countries: 6,2% in Belgium and 7,4% in Britain.
4.6 Additional analyses (results not reported) show that in both countries having a full-time job offers adequate protection against entering poverty upon partnership dissolution. In Britain especially, non-active and part-time working women are more vulnerable. In line with our second hypothesis, labour market entrance upon relationship dissolution is a more realistic option for Belgian women: 31,2% of non-working Belgian women younger than 60 enters the labour market between wave(t) and wave(t+1), compared to 22,3% of British women.
4.7 Although the pooling of resources is generally considered to be a route out of poverty, partnership formation also seems to have a positive effect on the poverty entry chances for Belgian women. Further analyses suggest that this effect mainly involves women who are heading a lone parent family in wave(t) and who start living with an economically inactive partner (mostly unemployed) between wave(t) wave(t+1).
4.8 The birth of a child has a positive and highly significant effect on the poverty entry probabilities in Britain. The coefficients for the Belgian models are positive, but not significant. In part this difference can be linked to the higher number of transitions into lone parenthood in Britain (13,8% compared to 6,8% in Belgium). Nevertheless, both for the male and female British subsamples, the birth of a child leads to higher poverty entry risks. We can relate the difference between both countries to the value of the ‘child benefit package’. This indicator composed by Ditch et al. (1998) calculates the extent to which welfare states assume responsibility for the cost of raising children. Compared to the European mean, the value for the Belgian ‘child benefit package’ is 137% larger. The corresponding value for Britain is -9%. Furthermore, in Britain the conditional poverty entry probabilities following the birth of a child increase with each additional child from 12,7% for the first child to 19,4% for the third child, reflecting the fact that child benefit for the second and later children is lower compared to the first child. In Belgium on the other hand, the amount of child benefit progressively increases. The poverty entry probabilities following the birth of the first and third child are consequently roughly equal (4,1% and 3,5%).
4.9 Leaving the parental home in order to establish a new household, alone or with a partner or friends, has a strong positive effect on the entry probabilities into income poverty for the British respondents. In the Belgian models, the only positive and significant effect concerns the entry probability for the male subsample. To some extent the difference between both countries can be explained by the fact that in the BHPS, students tend to be treated as members of their term-time household, while most Belgian students are officially registered with their parents, even when they are living in some form of student accommodation. Nevertheless, even when we exclude the student households from the analyses, the conditional poverty entry probabilities for the British respondents are about twice as high compared to the Belgian numbers (16,3% versus 8,5%). This difference cannot be explained by the labour market position: in both countries about 80% of respondents in these newly formed households belong to a household with a working reference person. Maybe the explanation is to be found in the quality of the labour market participation. In Britain, the educational system is less stratified and less aimed at the transmission of specific vocational skills. Therefore, job mobility during the first years after leaving full-time education tends to be rather high, and youngsters often spend some time in so-called ‘stop gap’ jobs, waiting for a better opportunity to come along (Mills and Blossfeld, 2004). Another explanation can be traced to the fact that British respondents have a higher propensity to form a single-person (47,8%) or lone parent household (7,1%), compared to 40,0% and 2,0% respectively for the Belgian respondents.
4.10 In contrast to the situation in Rowntree’s days, the departure of an adult child has no effect on the poverty entry probabilities for the respondents in the parental household. The only positive and significant effect is found for the subsample of British women. Slightly less than half of British women who become poor following this event are heading a lone parent family in wave(t). This finding again suggests, in line with our second hypothesis, a greater economic dependency on the part of British women, in this case on their adult children.
4.11 Our last demographic event refers to ‘other’ changes in the household composition, such as a grandparent who joins the nuclear unit or a lone mother who moves in with siblings. This event is more often recorded in the British panel (involving 5,3% of respondents compared to 1,8% in Belgium). The impact of ‘other’ household changes on the poverty entry probabilities is negligible for the British respondents. In the Belgian panel on the other hand the logistic regression coefficients are positive and significant.
|Table 2. Results from the pooled logit model predicting the entry into income poverty in Belgium (60%-poverty line, men and women at risk, weighted)|
4.12 Next we consider the impact of the different types of labour market events on the risk of entering income poverty. Regarding the results for the total sample, in both countries the household reference person becoming unemployed or disabled has a positive and highly significant effect on the poverty entry probabilities. The poverty entry risk is lower for the respondents belonging to a household where both partners have a job in wave(t). In Belgium, 42,3% of respondents belonging to a single-earner household enter income poverty upon the ‘forced’ inactivity of the reference person, compared to 8,6% for the respondents in a dual-earner household. This difference is however much smaller in Britain, with 45,1% and 34,2% respectively. As it turns out, the labour market participation of the partner does not offer much protection against falling into poverty for the British respondents: 63,7% entering income poverty upon the unemployment or disability of the household reference person belongs to a dual-earner household in wave(t).
4.13 Although the effect concerning the ‘forced’ transition into inactivity for the ‘other’ household members is positive and significant, the partner becoming unemployed or disabled does not lead to significantly higher poverty entry rates for the Belgian respondents. For the British respondents, the transition into ‘forced’ inactivity for all so-called ‘secondary earners’ leads to significantly higher poverty entry risks. Furthermore, a sizeable minority experiences more than one ‘negative’ labour market transition: for 36,3% the household reference person has also left the labour market between time(t) and time(t+1). We can link this finding to the poverty traps built into the British unemployment benefit system. Contribution-Based Jobseeker’s Allowance can be claimed for only 26 weeks. After that – or for those not qualifying –, the unemployed have to rely on Income-Based Jobseeker’s Allowance, which is means-tested and dependent on the labour market participation of the other household members.
4.14 This finding suggests that any evidence confirming the additional worker hypothesis (de Graaf and Ultee, 2000) should be stronger in Belgium, where the ‘penalties’ of entering the labour market for the non-working household members are much smaller. To test this hypothesis, we analyse the labour market behaviour of the partners of the household reference persons becoming unemployed or disabled between wave(t) and wave(t+1). In the Belgian panel, 17,0% of non-working partners has entered the labour market by time(t+1), while 47,3% of part-time working partners made the transition to a full-time job. The corresponding numbers for Britain are 15,2% and 20,9%. Furthermore, more British partners leave the labour market between wave(t) and wave(t+1): 14,5% compared to 8,9% for Belgium.
4.15 In Belgium, only the ‘unforced’ inactivity for the household reference person has a limited positive effect on the poverty entry risks. In Britain on the other hand, the exit from the labour market for both the household reference person and his/her partner results in higher poverty entry risks. Further analyses show that two thirds of respondents becoming poor upon the labour market exit by the partner or one or more ‘other’ household members belong to a household with a working reference person. Most of the British women leaving the labour market on their own initiative have children and low educational qualifications. This suggests that these ‘unforced’ exits are related to the high costs of childcare in Britain, and may not be so voluntarily as we might think (Joshi and Davies, 2002).
4.16 In both countries, the transition from a full-time to a part-time job does not lead to higher poverty entry probabilities. The negative estimates for the partner’s transitions point to a selection effect: only the households that can afford it opt to reduce the partner’s labour market participation.
4.17 While the (early) retirement of the household reference person has a limited positive effect on the poverty entry risks for the Belgian respondents, for the British respondents this event clearly results in a higher risk of entering poverty. We also find a limited effect concerning the (early) retirement by the partner. In Britain, poverty entry risks are clearly related to the social class of the household reference person in wave(t), with the highest risks occurring for the unskilled labourers (results not reported). In Belgium, risks are relatively low for all social classes, with the exception of the self-employed and the farmers. We can link this to the fact that these professional groups reside under a less generous social security scheme. On the other hand, this particular group might be prone to the underreporting of certain types of income, such as income from investments or savings. Further analyses (results not reported) indicate that the British respondents who become poor upon retirement more often report having no access to any private pension income, having followed an irregular labour market trajectory and having experience with long-term unemployment.
|Table 3. Results from the pooled logit model predicting the entry into income poverty in Britain (60%-poverty line, men and women at risk, weighted)|
4.18 The results for the separate subsamples of men and women confirm the stronger dependency of British respondents on the income of the so-called secondary earners. While for the Belgian men, only the effects concerning their own labour market exits are significant, for the British men both their own and their partner’s transitions to inactivity have a positive and significant effect on the poverty entry risks. For the Belgian women, only the respondent’s own transition to unemployment/disability leads to a higher poverty entry chance, as well as the different types of labour market exit by their partner. The British women on the other hand are vulnerable to an entry into income poverty following both their own exit from the labour market (‘forced’, ‘unforced’ and retirement) as well as their partner’s ‘forced’ exit and retirement.
Conclusion5.1 In this article, we have modelled the impact on entering income poverty of a range of specific demographic and labour market events. Our analyses are based on longitudinal panel data from two countries – Belgium and Britain – belonging to a different welfare regime. We have tried to link the observed poverty entry patterns to the ways in which financial welfare in Belgium and Britain is distributed between the three main systems of resources distribution: the welfare state, the labour market and the family. To this end, specific institutional arrangements on certain welfare domains were linked to the impact of specific life events on the poverty entry rates.
5.2 In line with our first hypothesis, the impact of both demographic and labour market events on the entry into income poverty is stronger in Britain than in Belgium. In Belgium, the main events associated with an entry into poverty are partnership dissolution (for women only) and the ‘forced’ inactivity due to unemployment or disability for the household reference person. In Britain on the other hand, most demographic events have a positive and significant effect on the poverty entry probabilities, especially widowhood, partnership dissolution, the birth of a child, and leaving the parental home in order to set up one’s own household. Regarding the impact of the different types of labour market events, we conclude that in Britain both the ‘forced’ and ‘unforced’ labour market exits of all household members, including retirement, are related to the poverty entry risk. The stronger dependency of British respondents on the labour incomes of the so-called secondary earners can be linked to the fact that for the more vulnerable segments of the population, one labour income does not suffice to stay above the income poverty line. Furthermore, British poverty risks are intensified by the lack of affordable childcare and the structure of the benefit system, both of which result in a withdrawal from the labour market by the working household members, and prevent the non-working members from entering the labour market.
5.3 According to our second hypothesis, women in Belgium are thought to have more opportunities to engage in paid labour if needed, in order to protect themselves and their families against the adverse effects of different demographic and labour market events, especially during the child-rearing years. Most of the evidence supporting this hypothesis is rather indirect, and relates to the available strategies to avoid falling into poverty. For instance, more non-working Belgian women enter the labour market upon partnership dissolution, while less British women are able to set up their own independent household. Furthermore, faced with the unemployment/disability of their partners, Belgian women seem to have more possibilities to enter the labour market or increase their work hours. This can be linked to the lack of affordable childcare in Britain on the one hand and the specific structure of the British benefit system on the other hand, penalising welfare recipients for the labour market participation of their household members. A final illustration of our hypothesis that British women have less opportunities to shield their families from income poverty is provided by the fact that the ‘voluntary’ exit from the labour market for low-educated British women with children results in a significantly higher poverty entry risk.
5.4 Generally speaking, we conclude that Esping-Andersen’s ‘amended’ welfare regime typology, i.e. taking into account the family, is a useful heuristic tool to predict poverty risks following different life events. Our results do however suggest that the work of feminist researchers, pointing out the importance of the available possibilities for women to achieve economic independence, might have been somewhat biased towards the female sex. As it turns out, women play an important role in the family economy and especially in countries where dual-earnership is a necessary condition for the lower-educated in order to achieve a decent living standard, their contribution is of vital importance for the economic well-being of all family members. Stated differently, different welfare regimes produce different patterns of dependence within families, with different consequences for their poverty trajectories.
5.5 Although the empirical evidence we presented is in line with the expectations deduced from our review of the welfare regimes in Belgium and Britain, some issues remain unresolved. Firstly, our research strategy does not allow for a ‘formal’ test of the impact of the institutional context. However, as explained before, this is a very complex problem which might not be resolved with the available data and/or methods of analysis. Secondly, although the level of detail of our analysis allows us to formulate some interesting avenues for further more policy-oriented research – especially with regard to enhancing the opportunities for British women to play a more pro-active role in the household economy – it is rather difficult to formulate more specific policy recommendations. In part this can be linked to the fact that even in large-scale representative surveys, the number of observations at the subgroup-level quickly becomes to small to pursue any further analysis. However, we also have to recognise that analyses specifically aimed at evaluating existing policies and preparing new ones may require a different approach. Perhaps administrative data could be useful for this purpose. Using existing information, one could for instance select a large enough sample aimed at specific population groups and research questions, in order to evaluate, in a prospective way, the impact of certain policy measures.
Notes1 A preliminary draft of this article was presented at the RC28 Spring Meeting on ‘Social Stratification, Mobility and Exclusion’, 7-9 May 2004, Neuchâtel (Switzerland).
4 % of children aged 0 to 2 in public childcare: 20; % of children aged 3 to school age in public childcare: 95 (Gornick, et al., 1997).
6 Since we are interested in the impact of demographic life events, which often involve households containing temporary or non-original sample members, the analyses in this article are weighted with the rescaled cross-sectional household weights rather than with the longitudinal base weights.
7 Because of small numbers, the labour market events for the ‘other’ household members are limited to ‘forced’ inactivity and ‘unforced’ inactivity.
8The results based on the annual income measure are available from the author.
9 This equivalence scale attributes a weight of 1 to the first adult in the household, each additional adult is given a weight of 0,5 and each child younger than 14 years of age is attributed a weight of 0,3.
10 The regression coefficient is positive but not significant for the poverty measure based on the annual income.
11 The received alimony and maintenance payments on the other hand are included in the total income.
12 Less than 95% of the ‘available’ working life spent in employment or any equivalent status (maternity leave, military service…).
AcknowledgementsThe author would like to thank three anonymous referees for their useful comments and suggestions.
ReferencesALLISON, P. D. (1999) Logistic Regression Using The SAS® System: Theory and Application. Cary, NC: Sas Institute Inc.
ANDREOTTI, A., GARCIA, S. M., GOMEZ, A., HESPANHA, P., KAZEPOV, Y. and MINGIONE, E. (2001) ‘Does a Southern European Model Exist?’, Journal of European Area Studies, Vol. 9, No. 1, pp. 43-62.
ANDREß, H.-J. (1998) ‘Empirical Poverty Research in a Comparative Perspective: Basic Orientations and Outline of the Book’, in H.-J. Andreß (editor) Empirical Poverty Research in a Comparative Perspective. Aldershot: Ashgate.
ASHWORTH, K., HILL, M. S. and WALKER, R. (2000) ‘A New Approach to Poverty Dynamics’, in D. Rose (editor) Researching Social and Economic Change. The Uses of Household Panel Studies. London: Routledge.
ATKINSON, T., CANTILLON, B., MARLIER, E. and NOLAN, B. (2002) Social Indicators. The EU and Social Inclusion. Oxford: Oxford University Press.
BANE, M. J. and ELLWOOD, D. T. (1986) ‘Slipping into and out of Poverty: The Dynamics of Spells’, The Journal of Human Resources, Vol. 21, No. 1, pp. 1-23.
BARDASI, E., JENKINS, S. P. and RIGG, J. A. (2001) Documentation for Derived Current and Annual Net Household Income Variables, BHPS waves 1-9, Colchester: Institute for Social and Economic Research.
BONOLI, G. (1997) ‘Classifying Welfare States: A Two-Dimension Approach’, Journal of Social Policy, Vol. 26, No. 3, pp. 351-72.
BREEN, R. and MOISIO, P. (2003) Poverty Dynamics Corrected for Measurement Error, Working Paper 17. Colchester: Institute for Social and Economic Research.
CANTILLON, B. (1999) ‘Samenvatting en beleidsconclusies: naar een herstel van het evenwicht tussen arbeid, gezin en verzorgingsstaat’, in B. Cantillon (editor) De welvaartsstaat in de kering. Kapellen: Pelckmans.
CASTLES, F. G. (1994) ‘On Religion and Public Policy: Does Catholicism make a Difference?’ European Journal of Political Research, Vol. 25, No. 1, pp. 19-40.
CASTLES, F. G. and FLOOD, M. (1993) ‘Why Divorce Rates Differ: Law, Religious Belief and Modernity’, in F. G. Castles (editor) Families of Nations. Patterns of Public Policy in Western Democracies. Aldershot: Darthmouth.
CASTLES, F. G. and MITCHELL, D. (1993) ‘Worlds of Welfare and Families of Nations’, in F. G. Castles (editor) Families of Nations. Patterns of Public Policy in Western Democracies. Aldershot: Darthmouth.
COOK, J., ROCHE, M., WILLAMS, C. C. and WINDEBANK, J. (2001) ‘The Evolution of Active Welfare Policies as a Solution to Social Exclusion in Britain’, Journal of European Area Studies, Vol. 9, No. 1, pp. 13-26.
DE GRAAF, P. M. and ULTEE, W. C. (2000) ‘United in Employment, United in Unemployment? Employment and Unemployment of Couples in the European Union in 1994’, in D. Gallie and S. Paugam (editors), Welfare Regimes and the Experience of Unemployment in Europe. Oxford: Oxford University Press.
DE LATHOUWER, L. (1996) Twintig jaar beleidsontwikkelingen in de Belgische werkloosheidsverzekering, CSB-Berichten. Antwerpen: Centrum voor Sociaal Beleid.
DE LATHOUWER, L., BOGAERTS, K. and VAN DEN BOSCH , K. (2003) Schorsing artikel 80 gewikt en gewogen. Een evaluatie vanuit herintrede, behoefte en verzekeringsperspectief, CSB-Berichten. Antwerpen: Centrum voor Sociaal Beleid.
DE LATHOUWER, L., CANTILLON, B., ANDRIES, M. and MARX, I. (1999) ‘De Belgische route in perspectief: markt, gezin en sociale zekerheid in de OESO-welvaartsstaten’, in B. Cantillon (editor) De welvaartsstaat in de kering. Kapellen: Pelckmans.
DE LATHOUWER, L. and MARX, I. (2002) ‘Werkgelegenheid en armoede. De prestaties van België en Nederland in vergelijkend perspectief’, Tijdschrift voor Arbeidsvraagstukken, Vol. 18, No. 4, pp. 335-50.
DELSEN, L. (1997) ‘Flexibilisering van de arbeid in Europa’, Tijdschrift voor Arbeidsvraagstukken, Vol. 13, No. 1, pp. 23-36.
DENNIS, I. and GUIO, A.-C. (2003) ‘Poverty and Social Exclusion in the EU after Laeken - Part 1’, Statistics in Focus, Theme 3 - 8/2003.
DEWILDE, C. (2003) ‘A Life-Course Perspective on Social Exclusion and Poverty’, The British Journal of Sociology, Vol. 54, No. 1, pp. 109-28.
DEWILDE, C. (2004) Vormen en trajecten van armoede in het Belgische en Britse welvaartsregime. Multidimensionele armoededynamieken bestudeerd vanuit de sociologie van de levensloop, Proefschrift voorgelegd tot het behalen van de graad van Doctor in de Politieke en Sociale Wetenschappen aan de Universiteit Antwerpen.
DEWILDE, C., BAUWENS, A., MARYNISSEN, R. and LAUWERS, J. (2000) Methodebericht golf 6 (1997), Antwerpen: Steunpunt Gezinsdemografisch Panel.
DITCH, J., BARNES, H., BRADSHAW, J. and KILKEY, M. (1998) A Synthesis of National Family Policies 1996. York: European Observatory on National Family Policies (European Commission).
DUNCAN, G. J. (editor) (1984) Years of Poverty, Years of Plenty: The Changing Economic Fortunes of American Workers and Families. Michigan: Institute for Social Research.
ENGBERSEN, G. (1998) ‘De taal van de armoede’, in G. Engbersen, J. C. Vrooman and E. Snel (editors), Arm Nederland. Effecten van armoede en sociale uitsluiting. Derde jaarrapport armoede en sociale uitsluiting. Amsterdam: Amsterdam University Press.
ESPING-ANDERSEN, G. (1990) The Three Worlds of Welfare Capitalism. Cambridge: Polity Press.
ESPING-ANDERSEN, G. (1999) Social Foundations of Postindustrial Economies. Oxford: Oxford University Press.
ESRC RESEARCH CENTRE ON MICRO-SOCIAL CHANGE (2002) British Household Panel Survey. Data Files and Associated Documentation. Colchester: The Data Archive.
FERRERA, M. (1996) ‘The ‘Southern Model’ of Welfare in Social Europe’, Journal of European Social Policy, Vol. 6, No. 1, pp. 17-37.
FOUARGE, D. (2002) Minimal Protection and Poverty in Europe. An Economic Analysis of the Subsidiarity Principle within EU Social Policy, Proefschrift ter verkrijging van de graad van doctor aan de Katholieke Universiteit Brabant.
GALLIE, D. and PAUGAM, S. (2000) ‘The Experience of Unemployment in Europe: The Debate’, in D. Gallie and S. Paugam (editors), Welfare Regimes and the Experience of Unemployment in Europe. Oxford: Oxford University Press.
GOODIN, R. E., HEADEY, B., MUFFELS, R. and DIRVEN, H.-J. (2000) The Real Worlds of Welfare Capitalism. Cambridge: Cambridge University Press.
GORNICK, J. C., MEYERS, M. K. and ROSS, K. E. (1997) ‘Supporting the Employment of Mothers: Policy Variation Across Fourteen Welfare States’, Journal of European Social Policy, Vol. 7, No. 1, pp. 45-70.
IACOVOU, M. (1998) Young People in Europe: Two Models of Household Formation, Working Paper 6. Colchester: European Panel Analysis Group.
JARVIS, S. and JENKINS, S. P. (1995) Do the Poor Stay Poor? New Evidence about Income Dynamics from the British Household Panel Survey, Occasional Paper 95-2. Colchester: ESRC Research Centre on Micro-Social Change.
JARVIS, S. and JENKINS, S. P. (1999) ‘Marital Splits and Income Changes: Evidence from the British Household Panel Survey’, Population Studies, Vol. 53, pp. 237-54.
JENKINS, S. P. (2000) ‘Modeling Household Income Dynamics’, Journal of Population Economics, Vol. 13, No. 4, pp. 529-67.
JOSHI, H. and DAVIES, H. (2002) ‘Women’s Incomes over A Synthetic Lifetime’, in E. Ruspini and A. Dale (editors), The Gender Dimension of Social Change. The Contribution of Dynamic Research to the Study of Women’s Life Courses. Bristol: The Policy Press.
KANGAS, O. and RITAKALLIO, V.-M. (1998) ‘Different Methods - Different Results? Approaches to Multidimensional Poverty’, in H.-J. Andreß (editor) Empirical Poverty Research in a Comparative Perspective. Aldershot: Ashgate.
KORPI, W. and PALME, J. (2003) ‘New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 countries, 1975-95’, American Political Science Review, Vol. 97, No. 3, pp. 1-22.
LAYTE, R. and WHELAN, C. T. (2002) Moving in and out of Poverty: The Impact of Welfare Regimes on Poverty Dynamics in the EU, Working Paper 30. Colchester: European Panel Analysis Group.
MENARD, S. (1995) Applied Logistic Regression Analysis. Newbury Park: Sage Publications.
MILLS, M. and BLOSSFELD, H.-P. (2004) ‘Globalization, Uncertainty and the Early Life Course: A Theoretical Framework’, in H.-P. Blossfeld, E. Klijzing, K. Kurz and M. Mills (editors), Globalization, Uncertainty and Youth in Society. London: Routledge.
MUFFELS, R. and FOUARGE, D. (2004) ‘The Role of European Welfare States in Explaining Resources Deprivation’, Social Indicators Research, Vol. 68, No. 3, pp. 299-330.
O’CONNOR, J. S. (1996) ‘From Women in the Welfare State to Gendering Welfare State Regimes’, Current Sociology, Vol. 44, No. 2, pp. 1-130.
ORLOFF, A. S. (1996) ‘Gender in the Welfare State’, Annual Review of Sociology, Vol. 22, pp. 51-78.
PAMPEL, F. C. (2000) Logistic Regression. A Primer. Thousand Oaks: Sage.
RINGEN, S. (1988) ‘Direct and Indirect Measures of Poverty’, Journal of Social Policy, Vol. 17, No. 3, pp. 351-65.
ROSE, D. (editor) (2000) Researching Social and Economic Change. The Uses of Household Panel Studies. London: Routledge.
RUSPINI, E. (2002) Introduction to Longitudinal Research. London: Routledge.
SAINSBURY, D. (editor) (1994) Gendering Welfare States. London: Sage Publications.
SANDEFUR, G. D. and TUMA, N. B. (1987) ‘How Data Type Affects Conclusions about Individual Mobility’, Social Science Research, Vol. 16, pp. 301-28.
TAYLOR, M. F., BRICE, J., BUCK, N. and PRENTICE-LANE, E. (2001) British Household Panel Survey User Manual Volume A: Introduction, Technical Reports and Appendices, Colchester: University of Essex.
TOWNSEND, P. (1979) Poverty in the United Kingdom. A Survey of Household Resources and Standards of Living. Harmondsworth: Penguin Books.
WALKER, R. (1998) ‘Rethinking Poverty in a Dynamic Perspective’, in H.-J. Andreß (editor) Empirical Poverty Research in a Comparative Perspective. Aldershot: Ashgate.
WHELAN, C. T., LAYTE, R. and MAÎTRE, B. (2002) ‘Multiple Deprivation and Persistent Poverty in the European Union’, Journal of European Social Policy, Vol. 12, No. 2, pp. 91-105.