Thesis submitted for the degree of Doctor of Social and Economic Sciences at the University of and Doctor of Social Sciences at the KU

The interplay between parental employment, social background and targeted cash support

Julie Vinck

Supervisors: Prof. dr. Wim Van Lancker - KU Leuven - Faculty of Social Sciences - Centre for Sociological Research Prof. dr. Bea Cantillon - University of Antwerp - Faculty of Social Sciences - Department of Sociology

Joint PhD University of Antwerp and KU Leuven

The poverty puzzle among children with a disability The interplay between parental employment, social background and targeted cash support

Thesis submitted for the degree of Doctor of Social and Economic Sciences at the University of Antwerp and Doctor of Social Sciences at the KU Leuven to be defended by Julie Vinck

Supervisors

Prof. dr. Wim Van Lancker ‒ KU Leuven Faculty of Social Sciences ‒ Centre for Sociological Research

Prof. dr. Bea Cantillon ‒ University of Antwerp Faculty of Social Sciences ‒ Department of Sociology Antwerp, 2020

Doctoral jury

Prof. dr. Wim Van Lancker ‒ supervisor ‒ KU Leuven Prof. dr. Bea Cantillon ‒ supervisor ‒ University of Antwerp Prof. dr. Sarah Van de Velde ‒ chair ‒ University of Antwerp Prof. dr. Jo Lebeer ‒ University of Antwerp Prof. dr. Griet Roets ‒ University of Prof. dr. Kitty Stewart ‒ London School of Economics and Political Science Prof. dr. Wim van Oorschot ‒ KU Leuven Prof. dr. Frank Vandenbroucke ‒ University of Amsterdam, University of Antwerp

ACKNOWLEDGEMENTS

During the five-year journey of writing this thesis I was supported, shaped and encouraged by a fantastic group of people. I owe a huge thank you to

Wim, for being smart, critical, constructive, approachable, funny, relaxed, simply the best guide and motivator I could have wished for.

Bea, for creating a stimulating research environment and inspiring anyone concerned about social issues.

Frank, for guiding me into academia in the first place.

Idunn and Jo, for the enriching collaborations along the way.

The members of the jury, for devoting the time to read and review this thesis.

The Research Foundation ‒ , for funding my fellowship, in Belgium as well as in Norway.

The Datawarehouse, Census and FAPD, for providing the administrative data.

The interviewed experts and parents, for giving insight behind the numbers.

Joyce, for designing this happy book cover.

My (former) colleagues, for compensating for the individual side of PhD life with enlightening inputs, pep talks, distracting chats and fantastic companionship.

My friends, for filling my life with beauty, joy, enthusiasm and craziness.

My families, for the opportunities, unconditional support and sweet concerns.

And most importantly Jorn, for everything.

TABLE OF CONTENTS

Introduction ...... 1 1. The issue of child poverty ...... 2 2. Why a focus on childhood disability is needed ...... 6 3. Belgium as a case study...... 10 3.1. The Belgian policy measures for families of children with a disability 13 3.1.1. Policies (previously) organised at the federal level ...... 13 3.1.2. Policies organised at the regional level ...... 15 4. Data ...... 16 5. Definitions and operationalisations ...... 21 5.1. Childhood disability ...... 21 5.2. Child poverty ...... 24 5.3. Parental employment and social background ...... 26 6. Overview of the thesis ...... 28 PART I Child poverty ...... 33 CHAPTER 1 Belgium: creeping vulnerability of children ...... 35 1.1. Introduction...... 36 1.2. The Great Recession and its aftermath ...... 37 1.3. Child poverty before, during and after the crisis ...... 41 1.3.1. Trends in living conditions of children in Belgium ...... 41 1.3.2. The role of the labour market ...... 46 1.3.3. Gauging the impact of the crisis ...... 48 1.3.4. Sociodemographic characteristics ...... 50 1.3.5. Work intensity and sociodemographic characteristics ...... 53 1.4. Policies ...... 56 1.4.1. Policies before the crisis ...... 56

i

1.4.2. Policies during the crisis ...... 59 1.4.3. Policy discourse ...... 61 1.4.4. The way ...... 62 1.5. Conclusion ...... 66 PART II Childhood disability ...... 69 CHAPTER 2 Non-take-up of the supplemental child benefit for children with a disability in Belgium: a mixed-method approach ...... 71 2.1. Introduction ...... 72 2.2. Understanding non-take-up: a dynamic multilevel model of claiming benefits ...... 73 2.3. Policies for children with a disability in Belgium: multiple recognition levels ...... 75 2.4. Methods and data ...... 78 2.5. Results ...... 82 2.5.1. Non-take-up in the supplemental child benefit ...... 82 2.5.2. Characteristics of children with a disability ...... 85 2.5.3. Determinants of non-take-up ...... 87 2.5.3.1. Threshold stage ...... 88 2.5.3.2. Trade-off stage ...... 91 2.5.3.3. Application stage ...... 94 2.6. Discussion and conclusion ...... 97 2.7. Appendix Chapter 2 ...... 100 2.7.1. Appendix 2.1 Supplemental child benefit: pillars, subscales, points and benefit amounts ...... 100 2.7.2. Appendix 2.2 Topic questionnaires ...... 101 2.7.2.1. Experts ...... 101 2.7.2.2. Parents ...... 102 2.7.3. Appendix 2.3 Questionnaires ...... 104

ii

CHAPTER 3 An intersectional approach towards parental employment in families with a child with a disability: the case of Belgium ...... 117 3.1. Introduction...... 118 3.2. Theoretical framework and previous research ...... 120 3.3. Research questions and hypotheses ...... 124 3.4. Data, methods and variables ...... 125 3.5. Results ...... 128 3.5.1. How does childhood disability overlap with social disadvantages? 128 3.5.2. Parental employment gap: childhood disabilities or social disadvantages? ...... 131 3.6. Discussion and conclusion ...... 136 3.7. Appendix Chapter 3 ...... 141 3.7.1. Appendix 3.1 Descriptive information on dependent, independent and control variables for children with and without a disability in Belgium, 2010 ...... 141 3.7.2. Appendix 3.2 Sensitivity check: regression results without applying the population weight ...... 142 3.7.3. Appendix 3.3 Sensitivity check: stepwise multivariate regression on work intensity, excluding extended families, Belgium, 2010 ...... 146 3.7.4. Appendix 3.4 Sensitivity check: stepwise multivariate regression for mothers and fathers separately, Belgium, 2010 ...... 148 3.7.5. Appendix 3.5 Further investigation of migration background ...... 152 3.7.6. Appendix 3.6 Social category and disability effect for children with - 1 standard deviation and +1 standard deviation from the mean severity score ...... 154 CHAPTER 4 Gender and education inequalities in parental employment and earnings when having a child with increased care needs: Belgium versus Norway ...... 157 4.1. Introduction...... 158 4.2. Theoretical framework, previous research and hypotheses ...... 160 4.3. Data, variables and methods ...... 166

iii

4.4. Results ...... 169 4.5. Discussion ...... 174 4.6. Conclusion ...... 177 4.7. Appendix Chapter 4 ...... 179 4.7.1. Appendix 4.1 The family policy packages in Belgium and Norway ...... 179 4.7.2. Appendix 4.2 Overview variables Belgian and Norwegian sample 184 4.7.3. Appendix 4.3 Descriptive information Belgian and Norwegian sample ...... 185 4.7.4. Appendix 4.4 Employment and wage gaps between parents with and without children with increased care needs...... 187 PART III Childhood disability and child poverty...... 193 CHAPTER 5 Income poverty among children with a disability in Belgium: the interplay between parental employment, social background and targeted cash support...... 195 5.1. Introduction ...... 196 5.2. Previous research and mechanisms ...... 198 5.3. Research questions and contributions to the literature ...... 202 5.4. Data, variables and methods...... 203 5.5. Results ...... 209 5.5.1. Income poverty estimates and determinants among children with and without a disability ...... 209 5.5.2. Poverty reducing impact of cash support systems ...... 215 5.6. Discussion and conclusion ...... 222 5.7. Appendix Chapter 5 ...... 226 5.7.1. Appendix 5.1 Equivalised net disposable household income simulation ...... 226 5.7.1.1. Simulation strategy ...... 226 5.7.1.2. Income distribution of households with children, DWH LM&SP versus BE-SILC, 2010...... 230

iv

5.7.1.3. Children with a disability versus children without a disability ... 232 5.7.2. Appendix 5.2 Descriptive information on variables of interest ...... 236 5.7.3. Appendix 5.3 Sensitivity check: applying the 50% and 70% at-risk- of-poverty threshold...... 238 5.7.3.1. 50% at-risk-of-poverty threshold ...... 238 5.7.3.2. 70% at-risk-of-poverty threshold ...... 241 5.7.4. Appendix 5.4 Sensitivity check: logistic results without applying the population weight ...... 244 5.7.5. Appendix 5.5 Further exploration of regular child benefit, Belgium, 2010 ...... 247 Conclusion ...... 249 1. What have we learned … ...... 250 1.1. … about the social background of children with a disability? ...... 250 1.2. … about the labour market participation of parents with children with a disability? ...... 250 1.3. … about the income poverty risk of children with a disability?...... 252 2. What is new? ...... 253 3. What remains to be done? ...... 254 4. What can policy makers do? ...... 261 References ...... 265 Contributions per chapter ...... 293 Nederlandstalige samenvatting...... 297 1. Definities van de centrale concepten ...... 299 2. Achtergrond: kinderarmoede in België ...... 302 3. Wat leert deze doctoraatsthesis ons over… ...... 303 3.1. … de sociale achtergrond van kinderen met een handicap? ...... 303 3.2. … de arbeidsmarktparticipatie van ouders van kinderen met een handicap? ...... 303 3.3. … het inkomensarmoederisico van kinderen met een handicap? ...... 304

v

4. Wat niet onderzocht kon worden in deze doctoraatsthesis ...... 306 5. Wat beleidsmakers kunnen doen ...... 309

vi

INTRODUCTION

The Belgian system of targeted cash support for children with a disability succeeds in reducing their risk of income poverty. As demonstrated in this thesis, even though their parents often work less, earn less, and have a disadvantaged social background compared to parents without this type of increased care burden, the income poverty risk is lower for children with a disability than for children without a disability. However, an income-based poverty indicator is not necessarily a good representation of the standard of living for these families as they face higher medical and care costs incurred by the child's disability. Moreover, the targeted cash support suffers from non-take-up, jeopardising its full poverty-reducing potential.

That is the take-home message of this thesis. My main aim is to unravel the poverty puzzle among children with a disability. In doing so, I scrutinise how the interplay between childhood disability, parental employment, social background and the receipt of targeted cash support affects the poverty risk of children with a disability. In the remainder of this introduction, I justify why research on this topic is needed and discuss my contributions to the literature. First, I briefly give an overview of the issue of child poverty in the social policy literature. I then clarify why an explicit focus on children with a disability is necessary. I further explain why Belgium is an interesting case study. Subsequently, I describe the data and definitions used throughout this thesis. Finally, I outline what will follow in the next chapters.

1

1. The issue of child poverty

“There can be no keener revelation of a society’s soul than the way in which it treats its children.” ‒ Nelson Mandela, 1995

At the risk of stating the obvious, welfare states should worry about child poverty. The reason why is that child poverty measures the temperature of the entire society. It is a persistent multidimensional problem and a lead indicator for future social problems as poverty is often passed down from generation to generation (Bellani & Bia, 2017; Corak, 2006; UNICEF, 2007; Wagmiller & Adelman, 2009). Growing up poor has immediate and long-lasting detrimental repercussions for individuals themselves and society as a whole (Brooks-Gunn & Duncan, 1997; Cooper & Stewart, 2013; 2017; Duncan, Morris, & Rodrigues, 2011; Duncan, Ziol-Guest, & Kalil, 2010; Evans, 2016; Griggs & Walker, 2008; Lai et al., 2019; Lesner, 2018; Main, 2014; Milligan & Stabile, 2011; Ridge, 2002; 2011)1. Poor children have poorer outcomes in terms of health, educational attainment, cognitive development, social development, behavioural development, and subjective well-being which, in turn, reduces their future employment opportunities, earnings, and health, and increases their future reliance on benefits. The longer children experience poverty, the greater the adverse impact on their outcomes will be, though even short episodes of poverty can have detrimental effects. The negative consequences of childhood poverty are reproduced at the societal level. As children who grow up poor are more likely to depend on benefits and be in poor health, and less likely to be employed and earning (high) wages when they are adults, it is expected that the welfare state’s social and health

1 This section briefly summarises the consequences of growing up poor, see Van Lancker and Vinck (2020) for an elaborated overview in a global perspective.

2 expenditures will rise and its revenues from consumption and taxation will diminish. Additionally, poorer outcomes in terms of health, educational attainment and employment can reduce productivity and competitiveness, increasing the societal costs of child poverty even further.

The fact that the thermometer shows an increasing temperature in most of the OECD countries should add to this worry (Thévenon, Manfredi, Govind, & Klauzner, 2018). Over the past decades, the number of children growing up poor has been on the rise, a trend that has been intensified by the Great Recession in some countries (Cantillon, Chzhen, Handa, & Nolan, 2017; Jenkins, Brandolini, Micklewright, & Nolan, 2013). In fact, in many welfare states, an intergenerational shift in the poverty risk has taken place, which has been accentuated by the Great Recession as well (Bradshaw et al., 2012; Förster & Mira D’Ercole, 2005; OECD, 2008; 2014b; UNICEF, 2014). In general, the elderly, a perpetually vulnerable population subgroup, experienced a steady decline in their poverty risk, whereas the opposite occurred among the young. Today, 20.3% of children (younger than 18) live at risk of income poverty in the European Union on average, while the corresponding shares are 16.5% among the working-age population (18-64) and 15.9% among the elderly (65 or older) (EUROSTAT, 2020a).

An ample amount of the social policy literature on child poverty shows that the risk of growing up poor is unequally distributed within as well as across societies and that it changes over time (Bradshaw et al., 2012; Chen & Corak, 2008; Galloway, Gustafsson, Pedersen, & Österberg, 2015; Gornick & Jäntti, 2012; Rainwater & Smeeding, 2003; Thévenon et al., 2018; Vandenbroucke & Vinck, 2013). To understand within-country differences, parental employment and social background matter a great deal. Income-poor children often live with parents holding little or no attachment to the labour market and with parents belonging to

3 disadvantaged social categories such as single parents, parents with lower educational qualifications and parents with a migration background. Yet, for understanding between-country differences and child poverty trends over time, the social background of the household that children grow up in is not truly enlightening, it only moderately explains these differences and trends. Rather, differences in the institutional set-up and outcomes of labour markets and tax- benefit systems, and how these interact, offer more insight. Research shows for instance that child poverty rates are negatively correlated with child-related benefits and child poverty reduction is better achieved in countries with more targeting towards lower-income families with children, though redistributive policies not directly targeted towards families with children, such as housing benefits, social assistance and even pensions, matter as well (Cooper & Stewart, 2013; 2017; Diris, Vandenbroucke, & Verbist, 2017; Gornick & Nell, 2017; Thévenon et al., 2018; Van Lancker & Van Mechelen, 2015).

The fight against child poverty has been high on the policy agenda. To give some examples, the United Nations (2015) incorporated a target related to child poverty in their Sustainable Development Goals stating that by 2030, the share of poor children must be halved, including in high-income countries. The OECD (2016) emphasised that improving (young) children’s living conditions is indispensable in its Inclusive Growth strategy. In Europe, the Commission (2013) formulated a Recommendation on Investing in children: breaking the cycle of disadvantage, providing countries with guidelines and strategies on how to combat child poverty and foster the social integration and well-being of its children. Moreover, more recently, the European Pillar of Social Rights included a right to protection from poverty for children in its Principle 11 (European Parliament, Council of the European Union, & European Commission, 2017).

4

Policy strategies to reduce child poverty often focus on integrating parents into the labour market to increase family income, while simultaneously investing in the future potential of children to break the cycle of disadvantage (Esping-Andersen, Gallie, Hemerijck, & Myles, 2002). High-quality early childhood education and care services are promoted to accomplish this. These services are expected to achieve a dual goal: on the one hand, they should foster maternal employment as the care for young children can (partly) be outsourced, while they also aim to improve children’s human capital by offering them a stimulating learning environment on the other hand. This should particularly benefit children from a disadvantaged social background and hence fight child poverty by reducing inequalities at an early point in life. As a precondition, disadvantaged families should have access to these services, which is not always the case (e.g. Gambaro, Stewart, & Waldfogel, 2014; Van Lancker, 2013).

These policy strategies are rooted in the social investment paradigm (Morel, Palier, & Palme, 2012). The idea at the heart of social investment is that welfare states should recalibrate their policies to help their citizens prepare for new social risks and needs, rather than repair them once setbacks occur (Hemerijck, 2018; Vandenbroucke, Hemerijck, & Palier, 2011). This should be achieved through human capital investments beginning in early childhood, equipping citizens with the necessary skills to grab employment opportunities themselves instead of merely granting passive cash support to protect them from poverty when needed. The advocates of social investment argue, however, that rather than replacing social protection policies completely, social investment policies should be mutually reinforcing (Esping-Andersen et al., 2002). Nevertheless, it is assumed that income protection and social integration are best safeguarded through participation in paid employment. Indeed, many welfare states reformed their

5 policies with a stronger emphasis on activation which was often accompanied by cuts in cash support to avoid inactivity traps (Bonoli, 2012). The flipside of this is that nowadays minimum income protection schemes are often incapable of keeping non-working people at active age out of poverty (Marchal, 2017). On top of this, even groups that have previously been spared, such as single mothers, people with a disability and people who provide care, are increasingly included in the activation policies (Burkhauser, Daly, & Ziebarth, 2016; Good Gingrich, 2008; Hvinden, 2003; Lindsay, Greve, Cabras, Ellison, & Kellet, 2015; Marin, Prinz, & Queisser, 2004; Roets et al., 2012).

2. Why a focus on childhood disability is needed

In a context where families are pressured to work, it is crucial to understand how parental employment can be improved in families where the care needs of children with a disability exceed those of families without this type of care burden. The social investment paradigm does not offer much guidance in this regard as the literature does not specifically address how social policies should deal with the issue of caregiving in a context heavily affected by the “work-first imperative” (Cantillon & Van Lancker, 2013)2. This is unfortunate, as previous research within the disability and child development literature has shown that there is a clear link between childhood disability and child poverty where the causality can run in both directions (e.g. Emerson, Shahtahmasebi, Lancaster, & Berridge, 2010; Houtrow, Larson, Olson, Newacheck, & Halfon, 2014; Lustig & Strauser, 2007; Mont, 2014). Accordingly, children with a disability are more prone to the detrimental

2 A recent EU-funded project (SPRINT 2015-2018) examined the issue of long-term care for the elderly in the context of the EU’s Social Investment Package, yet the main focus of the project is a social cost-benefit analysis of long-term care provision.

6 consequences of growing up poor than children without a disability. However, the issue of childhood disability is often overlooked in the social policy literature on child poverty with a few notable exceptions that focus on out-of-pocket costs related to the child’s disability (Meyers, Lukemeyer, & Smeeding, 1998) and the family’s social background (Shahtahmasebi, Emerson, Berridge, & Lancaster, 2011). Thus far, it is still unclear what precise role parental employment patterns play in explaining the poverty risk of children with a disability and to what extent targeted cash support for these children reduces their poverty risk.

One should know that for children with a disability, poverty reduction through parental employment is prima facie problematic for two reasons (Cantillon & Van Lancker, 2013). First, parents of children with a disability need to provide more care than the typical parental care (Owen, Gordon, Frederico, & Cooper, 2003) and the time required for this hampers their labour market participation, especially for mothers (Brown & Clark, 2017; Stabile & Allin, 2012). The more severe the child’s disability is, the more difficult it is for their parents to engage in paid employment (Chou, Kröger, & Pu, 2018; Crettenden, Wright, & Skinner, 2014; DeRigne, 2012; Gordon, Rosenman, & Cuskelly, 2007; Hauge et al., 2013; Leiter, Wyngaarden Krauss, Anderson, & Wells, 2004; Lu & Zuo, 2010; Wasi, van den Berg, & Buchmueller, 2012). Having a child with a disability can also influence parental employment indirectly through poorer parental (mental) health (Brekke, Albertini Früh, Kvarme, & Holmstrøm, 2017; Brekke & Nadim, 2016; Singer & Floyd, 2006).

Second, the poverty risk and employment patterns of families with children with a disability are strongly related to processes of social stratification (Brown & Clark, 2017; Rothwell, Gariépy, Elgar, & Lach, 2019; Shahtahmasebi et al., 2011; Stabile & Allin, 2012). Children with a disability are more likely to live with single

7 parents, parents with lower educational qualifications, and other household members with a disability (Bauman, Silver, & Stein, 2006; Blackburn, Spencer, & Read, 2010; Clarke & McKay, 2008; Emerson & Hatton, 2007; Fujiura & Yamaki, 2000; OECD, 2010; Powers, 2003; Reichman, Corman, & Noonan, 2008; Risdal & Singer, 2004; RKW, 2013; Sebrechts & Breda, 2012; Van Landeghem, Breda, & Mestdagh, 2007; Wasi et al., 2012). These disadvantaged social background characteristics might affect the family’s employment opportunities and poverty risk independently of having a child with a disability (Debacker, 2008; Gornick & Jäntti, 2012; Grammenos, 2018; Konietzka & Kreyenfeld, 2010; Nieuwenhuis & Maldonado, 2018; Nys, Meeusen, & Corluy, 2016; Stahl & Schober, 2018; Vornholt et al., 2018).

However, the intersection between childhood disability and social disadvantage has been hitherto overlooked in the literature on parental employment in families of children with a disability. This will be my first contribution to the literature. Moreover, studies investigating parental employment in these families are still lacking a cross-country perspective. My second contribution will, therefore, consist of a cross-country comparison between Belgium and Norway. The role of parental employment in the explanatory framework of poverty among children with a disability also remains unexplored. My third contribution to the literature will be to take the interplay into account between childhood disability, parental employment and the social background of the family.

In addition to parental employment and social background, the tax-benefit system is also crucial to understand a child’s poverty risk. So far, however, the social policy literature lacks studies that investigate the role of targeted cash support in mitigating poverty for families of children with a disability. To the best of my knowledge, there are only a few studies in the academic and grey literature that

8 look into this. These studies either assess the poverty-reducing impact of the “Supplemental Security Income” in the United States, a cash programme that is both disability-tested as well as means-tested (Boat & Wu, 2015; Luca & Sevak, 2019; Romig, 2017; Stegman Bailey & Hemmeter, 2014). Alternatively, they highlight that by including benefits designed to (partly) cover the costs of the child’s disability into the household’s income, the actual poverty risk is underestimated (Byrne, 2014; Monteith, Casement, Lloyd, & McKee, 2009 for the United Kingdom’s “Disability Living Allowance”). For Belgium, Van Landeghem et al. (2007) hint at the poverty alleviation achieved by the tax-benefit system by comparing incomes before and after taxes and benefits between families with and without a child with a disability. However, they do not provide evidence to pinpoint which component of the tax-benefit system causes the income difference between families with and without a child with a disability to be smaller after taxes have been paid and benefits received than before. That will be my fourth contribution to the literature: by allowing for the interplay between childhood disability, parental employment and social background, this thesis scrutinises how the receipt of different targeted cash support systems reduces the poverty risk of children with a disability. Yet, the actual effectiveness of targeted cash support could be jeopardised by the issue of non-take-up. How receiving targeted cash support for children with a disability is related to their family’s labour market participation and social background remains to be revealed. Hence, in my final contribution to the literature, I investigate the magnitude, characteristics and determinants of non-take-up of cash support targeted at children with a disability.

9

3. Belgium as a case study

Belgium is an interesting case study for investigating child poverty as the country is confronted with a comparatively high child poverty risk and a creeping vulnerability of children (Vandenbroucke & Vinck, 2013; Vinck, Van Lancker, & Cantillon, 2017 or Chapter 1). Today, 20.6% of children below 18 are living at risk of income poverty, a similar record as the European Union on average, whereas 12 years ago this was 15.3% (Figure 1). On top of that, the income poverty risk for children is higher than for the rest of the population. This is a disappointing and worrying result for a country with such a long-standing and highly developed welfare state.

To a great extent, the reason why Belgium performs so poorly in a European comparative perspective can be found in the employment patterns of the households that children are living in (Vandenbroucke & Vinck, 2013). In fact, the country is characterised by a “dual polarisation”: a high share of children lives in a household where nobody works and, at the same time, these children are confronted with a high poverty risk. Put differently, there is an enormous difference between the “haves” and the “have not’s”, both in terms of participation in employment and in terms of the impact of this participation on poverty. Today, almost 50% of income-poor children live in a household where nobody works, only Ireland performs worse (54%) (EUROSTAT, 2020a, 2020b, 2020d)3.

3 Over the past 12 years, Belgium almost consistently occupies the second place after Ireland, except in 2007 when it is also preceded by Bulgaria, and in 2008 when the United Kingdom, Germany, the Czech Republic and Ireland perform worse.

10

Figure 1. Relative income poverty risk in Belgium by age group, income years 2005 ‒ 2017

25

20

15

10

5

Relative income risk (%) income poverty Relative 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Income year

Total population Children (0-17) Active age (18-64) Elderly (65+)

Source: EUROSTAT (2020a). Note: this is the updated version of Figure 1.4 in Chapter 1.

Belgium is also an interesting case study regarding childhood disability for the reason that the research hitherto conducted on this topic here remains limited. There is a cross-country comparative study in the demography literature that includes Belgium and explores the familial consequences of having a child with a disability (Di Giulio, Philipov, & Jaschinski, 2014). Relying on the Gender and Generation Surveys, the authors present descriptive evidence on the quality and stability of the households’ partnerships, fertility decisions, participation in work and care, parents’ self-perceived health and well-being, and subjective poverty situations. Yet, these analyses rely on very small samples of children with a disability: for Belgium, just 59 out of 2,280 families are identified as having a child with a disability in the household, corresponding to 2.5% of all families with at least one child under the age of 19.

11

Moreover, up until 2017, neither the European Union Statistics on Income and Living Conditions (EU-SILC, see Section 4 for more information on this data source) nor the Labour Force Survey offered information that allowed for an investigation into childhood disabilities, as health-related questions were restricted to people of at least 15 or 16 years old. In 2017, an ad-hoc module on children’s health and activity limitations due to health problems was included in EU-SILC, making it possible to study health-related issues among children in a comparative perspective4. However, the data only became recently available for research purposes and the sample sizes remain limited as well: in Belgium, only 139 out of almost 2,600 children minus 16 years old experience moderate to severe activity limitations because of health problems, corresponding to 5.0% (EUROSTAT, 2020c).

Within the social policy literature, a few descriptive studies build on the Families and Care Survey (see Section 4 for more information on this data source) and have given us insight into the need for and use of care concerning children with a disability in the northern region of Belgium, Flanders. The survey further offers information into these families’ social backgrounds, participation in employment and poverty situations (Debacker, 2007; Sebrechts & Breda, 2012; Van Landeghem et al., 2007). This work has been influential in shaping the research questions of this thesis, but the data employed are insufficient for a thorough investigation.

To sum up, clarification is needed on the extent to which the poverty risk of children with a disability is affected by the interplay between childhood disability, parental employment, social background and the receipt of targeted cash support.

4 It should be noted that as EU-SILC only includes private households, children who live in institutions are left out of the sample.

12

3.1. The Belgian policy measures for families of children with a disability

Belgium is a federal state where responsibilities for social policies are (partly) decentralised. Consequently, different levels of government are in charge of the financing, legislation and organisation of various policy measures. The policy measures targeted at families of children with a disability are also divided between the federal (central) and regional (decentral) level. Until recently, cash support measures for families of children with a disability were centrally regulated but since 2014, they have mainly been decentralised to the regional level. Care support measures have been organised at the regional level for quite some time. Social and tax advantages are rights derived from the recognitions needed for the cash support or care support measures. Below, I will describe the most important cash and care support measures5 available for children with a disability as they were in place in 2010 (the year of data collection, see Section 4). When the measure is decentralised, the focus is on Flanders.

3.1.1. Policies (previously) organised at the federal level

The main cash support measure that is targeted at families with children, disabled and non-disabled alike, is the child benefit system (hereafter referred to as the “regular child benefit”). It is characterised by “targeting within universalism”, though the lion’s share of the budget is universally allocated (Van Lancker & Van Mechelen, 2015). Children below the age of 18 and students between 18 and 24 years old are eligible. The system is composed of a universal amount with age and rank supplements, and a number of income-tested supplements for vulnerable

5 Appendix 4.1 additionally includes the characteristics of leave provisions and childcare services.

13 groups, in particular the social assistance recipients, the long-term unemployed, the long-term sick, and single parents.

For children with increased care needs, including children with a disability, a supplement within the regular child benefit system exists. To be entitled to this “supplemental child benefit”, children must meet three criteria simultaneously: they need to be eligible for the regular child benefit, they need to be younger than 21 years old, and their disability needs to be recognised by doctors of the Federal Public Service (FPS) for Social Security. Therefore, medical control doctors assess children’s care needs related to the disability and assign them a score on a 36-point scale for which they use a standardised criteria list. The scale is made up of three pillars that represent the severity of the child’s disability in terms of (1) the physical and mental consequences of the disability (maximum 6 points), (2) the consequences for the child’s participation in daily life (maximum 12 points), and (3) the consequences for the family (maximum 18 points). The supplement is not conditioned on an income test, but the size of the benefit depends on the score the child has on the 36-point scale, ranging from € 80 up to more than € 500 per month. For more information on the benefit and the recognition procedure, see Chapter 2 of this thesis6.

A final targeted cash support measure that is important to consider here is the “refundable tax credit for dependent children”. For each dependent child in the household, the tax allowance (i.e. the amount that is exempt from personal income taxes) of one parent is increased. If the parent’s taxable income is lower than the

6 So far, the recent decentralisation of the child benefit system has not resulted in changes to this cash support measure for children with a disability, except in that Flemish children must now turn to another agency to get the recognition (nothing changed in the other regions). For more information on the decentralisation, see Béland and Lecours (2018).

14 sum of the tax allowance and tax credits7, the unused part of the increased tax allowance becomes refundable, limited to € 390 per dependent child per year. Children who receive the supplemental child benefit and who have at least four points on the first pillar (corresponding to a disability of at least 66%) are counted as dependent twice.

3.1.2. Policies organised at the regional level

To qualify for subsidised in-kind care support (such as residential, semi-residential or ambulatory care services), or to apply for additional financial support (i.e. a budget to buy personal assistance, assistive technology for communication or mobility, or to make adaptations to the home), a recognition from the Flemish Agency for Persons with a Disability (FAPD) is needed8. To obtain this recognition, a multidisciplinary team composed of a medical doctor, a psychologist or pedagogue, and a social worker or social nurse, assesses whether children are substantially limited long-term in their social participation due to their disability. After children acquired the recognition of the FAPD, they can use care services or apply for additional financial support if the availability of places and budget permits. Again, more information on this recognition procedure can be found in Chapter 2.

7 Tax credits are tax advantages given according to the composition of the fiscal households, see Appendix 5.1 for more information. 8 Today, all services for children have been grouped together in the intersectoral administrative agency of the , known as “Integral Youth Help”, that encompasses a wide range of intensive and less intensive forms of support, both directly and non-directly accessible. For an updated overview of the available in-kind care and support services and how they are organised for children with a disability, see Lebeer, Vinck and Farah (2018). For a critical assessment of the current discourse of disability policy and practice in Flanders, see Roets et al. (2020).

15

Another important source of care support for children with a disability in Belgium is organised within the educational system. If children wish to receive support in an inclusive educational setting or want access to special education, recognition from the Pupil Guidance Centre is required. Since 2015, priority is given to inclusive education, though the most recent figures indicate that the vast majority (85%) of children with this official recognition are still enrolled in special education schools (EASIE, 2018).

In short, different recognition procedures apply at the different governmental levels and they generally operate independently from each other. A faster recognition procedure at the FAPD is only possible if children obtain at least 18 on the 36-point scale of the supplemental child benefit9.

4. Data

Research examining the link between childhood disability and child poverty has been impeded by the lack of sufficient, reliable and comparable data. The cross- country comparative Gender and Generations Surveys suffer from small sample sizes regarding childhood disability. In a total of 11 countries10, there are barely 771 out of almost 40,000 interviewed families reporting that they have a child with a disability below the age of 19, corresponding to 1.9% (Di Giulio et al., 2014). In EU-SILC 2017, the number of children experiencing activity limitations because of health problems also remain fairly small: 4,376 out of 90,832 children under 16

9 With the decentralisation of the child benefit system, steps were taken towards more coherence in the policy landscape in Flanders. From 2019 onwards, if children apply for non-directly accessible care support at the intersectoral administrative agency Integral Youth Help, their eligibility for the supplemental child benefit is simultaneously evaluated. 10 Austria, Belgium, Bulgaria, France, Georgia, Hungary, Italy, Lithuania, Poland, Romania and Russia.

16 years old in the 28 countries that belonged to the European Union in 2017, corresponding to 4.7% (EUROSTAT, 2020c).

Within Belgium, only one source of quantitative data has been gathered that allows for an analysis of children with a disability and their families in Flanders: the Flemish Families and Care Survey. The data includes socioeconomic and care information on households with children below the age of 16, collected via interviews with almost 2,800 families over the course of the 2004 – 2005 school year. Children with a disability are oversampled in the data through a random selection (n = 458) from the customer database of the FAPD, the agency responsible for in-kind care support and additional financial support. Yet, this dataset is also not sufficient for my intended research.

Therefore, I employ unique and large-scale administrative data that have never been disclosed for research purposes before. The main data source employed throughout this thesis (Chapters 2 to 5) is drawn from the Datawarehouse Labour Market and Social Protection (DWH LM&SP), managed by the Crossroads Bank for Social Security. The DWH LM&SP brings together socioeconomic information from Belgian social security agencies for people living or working in Belgium as well as personal and household information from other administrative agencies such as the National Register. A random sample is obtained from the DWH LM&SP of 50% of children below 21 who were recognised by the FPS Social Security for the supplemental child benefit (see Section 3.1.1. for more information) on the 31st of December 2010 and who were living in Belgium (n = 25,717). An equally sized randomly sampled control group is also acquired, consisting of children younger than 21 who do not receive the supplemental child benefit (n = 25,057 after removing children who have siblings who get the supplemental child benefit). A population weight is constructed based on the

17 child’s age, gender and the type of household they are living in to represent the full Belgian population of children below 21, with or without a (recognised) disability. Information is added on persons who are living at the same address according to the National Register (parents, siblings and other household members)11. Parents are identified as those who take up the parental role, meaning that they are either the reference person or their partner in a household where at least one child, stepchild, grandchild, great-grandchild, child-in-law or a third- degree cousin is present. Siblings are all non-sampled children below 21. Other household members are the remaining persons living at the same address.

The data contain information on the place of residence, gender, date and country of birth, household composition, disability status, labour market participation, gross taxable income from current and past employment, and non-taxable income from disability benefits, social assistance and (simulated) child benefits. Concerning the disability information, when children who are recognised for the supplemental child benefit have multiple non-identical records regarding this recognition, a stepwise selection strategy is applied. For recognitions where the severity of the disability is the same and no change occurred in 2010, one record is randomly picked (step 1). For recognitions where the severity of the disability is the same but a change occurred in 2010, the record with the longest recognition period in 2010 is taken (step 2). For recognitions with a different severity score but where no change occurred in 2010, the record corresponding to the highest benefit amount is selected (step 3), and if that did not result in a unique record, the one with the longest recognition period is chosen (step 4). Finally, when

11 Children who are recognised for the supplemental child benefit and who live in institutions are also included in the sample. When they have their official residence in the institution, no parent can be assigned to them. However, this only concerns a limited amount of children.

18 recognitions have different severity scores and changes occurred in 2010, the information from these recognitions is combined into one record (step 5).

To the DWH LM&SP microdata, information on the highest educational qualification is added from the latest Belgian Census of January 1st, 2011. For children living in Flanders, information about the recognition at the FAPD and the disability type as identified by the FAPD is linked to the microdata as well.

This administrative microdata is supplemented with four other data sources. First, because the microdata does not allow for the identification of children with a disability who are not recognised for the supplemental child benefit, an administrative dataset of all children who are only recognised by the FAPD is acquired as well (n = 8,968). This data includes the children’s disability type as recognised by the FAPD, their gender and date of birth, and their parents’ country of birth. Unfortunately, the data lack information on the household composition, educational qualifications, labour market participation, income, and disability status of other household members. Aggregate figures on the number of children below 21 living in Flanders who are only recognised for the supplemental child benefit, or who are recognised for both the supplemental child benefit and for the subsidised services of the FAPD are also obtained.

Second, qualitative information from 22 semi-structured interviews is used to gain more insight into why children with a disability are not necessarily recognised for the supplemental child benefit. Between February 2017 and March 2018, interviews were conducted with nine experts working in different organisations involved in implementing policy for children with a disability and with 13 parents of children with autism spectrum disorders or behavioural disorders. The information from the interviews, together with the first additional information source are employed in Chapter 2.

19

Third, comparable information from Norwegian administrative data is used in a cross-country comparative study on the link between childhood disability and parental employment (Chapter 4). The data consists of all children born in Norway between 2000 and 2005. Their mothers and fathers are followed over time, with the last available observation point in 2008. In accordance with the Belgian administrative data, children with a disability are identified as those who receive a non-means-tested cash benefit designed to financially compensate for the extra private care (i.e. the “attendance benefit”). The eligibility criteria to receive this Norwegian cash benefit are comparable to those applied in the Belgian supplemental child benefit: both include a (certain) degree of incapacity and take into account how the increased care needs impact different facets of the child’s daily life, and how providing care affects the caregiver’s or family’s life.

Finally, I work with the Belgian cross-sectional sample of the EU-SILC in Chapter 1 (using survey years 2006 – 2014) and Chapter 5 (using survey year 2011). The latest wave of the EU-SILC cross-sectional microdata contains representative samples of private households and their members for all countries of the European Union, the United Kingdom, Iceland, Norway, Switzerland, Montenegro, North Macedonia, Serbia and Turkey12. The data are collected yearly and are the primary reference for monitoring poverty and social exclusion in Europe. The Belgian sample consists of approximately 6,000 households each year. In this thesis, information on children and their households is used to analyse income poverty among children in general and to examine its trends and determinants.

12 Quality reports are accessible through EUROSTAT (2020e).

20

5. Definitions and operationalisations

This section defines this thesis’ central concepts of childhood disability and child poverty along with the main determinants of interest for studying the link between the two: parental employment and social background. For more information on the third component that is key to disentangle the child poverty puzzle, namely the targeted cash support, see Section 3.1.1.

5.1. Childhood disability

Childhood disability can be theoretically conceptualised in different ways. In general, the medical and social model dominate the field (Haegele & Hodge, 2016). The advancements in medical knowledge and science (largely) put an end to the theological perspective that disability is an act of God and shaped the development of the medical model (Stiker, 1999). In the medical view, disabilities originate within the mind or body of individuals. It is an internal impairment that limits the individual’s functioning and this situation is perceived as being inferior; the only solution is to treat the individuals. Alternatively, the social model that came to the fore in the 70s, regards disability as socially constructed and forced onto individuals with impairments as they experience barriers to fully participate in society. Removing those barriers should, therefore, be a societal rather than individual responsibility. Recently, the biopsychosocial view on disability gained popularity, even though it was developed more or less simultaneously with the social model as a critique of the medical view (Engel, 1977). The biopsychosocial view seeks to integrate a medical and social perspective on disability, alongside a psychological one (i.e. how is the disability experienced by the individual). It is used as a basis for the World Health Organisation’s (WHO) International

21

Classification of Functioning, Disability and Health (ICF) and for the children’s and youth’s specific classification (ICF-CY) (Wade & Halligan, 2017; WHO, 2001; 2007). Within the ICF, disability is conceptualised as “an umbrella term for any impairment, activity limitation or participation restriction which limits functioning within contextual (personal and environmental) factors” (Palmer & Harley, 2012: 359).

Throughout this thesis, children with a disability are defined as children who receive the supplemental child benefit (see Section 3.1.1 for more information). Hence, childhood disability is operationalised through an administrative recognition of the disability as evaluated by control doctors of the FPS Social Security. The supplemental child benefit combines a medical and social view on disability in their assessment. The medical view is reflected in the first pillar which assigns a disability percentage to capture the physical and mental consequences of the child’s disability. The social view is generally reflected in the second and third pillar, though mainly medical criteria are used to gauge the impact on the child’s participation in daily life (pillar 2) and on the family (pillar 3). The severity of the child’s disability is measured by the score the child has on the 36-point scale.

Figure 2 presents the share of children who receive the supplemental child benefit among all dependent children below the age of 25, from its introduction in 1964 until today. The most important events in the history of its existence are indicated in the figure. Three phases can be discerned. The first phase (from 1964 to 1974) is characterised by a strong increase in the share of children receiving the supplemental child benefit. In the early years, only a medical certificate of the child’s GP was needed to receive the supplement. The requirement to undergo a medical examination at an administrative agency marked the beginning of the second phase (from 1975 to 1999), which is characterised by a steady decrease in

22 the share of children receiving the supplemental child benefit. The downward trend perpetuated after the supplement was restricted to children below 21 years old in 1987. Since the turn of the century, the downward trend was reversed again with the introduction of a new evaluation system in 2003 (third phase). Among other things, it was no longer necessary for the child to have a disability of at least 66% to be eligible for the supplement. In 2018, 2.3% of children younger than 25 received the supplemental child benefit.

Figure 2. Evolution of the share of children receiving the supplemental child benefit among dependent children under 25, 1964-2018

3.0% 2003: 2nd reform: 1987: 1st reform: new evaluation benefit limited to system (incl. no 2.5% children < 21 longer 66% disabled) with 6 2.0% benefit amounts

1.5% 1975: medical examination by 2006: further 1.0% administrative differentiation in agency 1991: benefit amounts differentiation in (7 amounts) 0.5% benefit amount (3 amounts)

0.0%

benefit among benefit among dependent children < 25

1980 2008 1964 1966 1968 1970 1972 1974 1976 1978 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2010 2012 2014 2016 2018 Share of children with Share children of with supplemental child Phase 1: increase Phase 2: decrease Phase 3: increase 1964-1974 1975-1999 2000-2018

Source: RKW (2013) for 1964-2012 figures, FAMIFED (2019a, 2019b) for 2013-2014 figures, and FamiStat (2019a, 2019b) for 2015-2018 figures. Notes: Dependent children are children who receive the regular child benefits. Before July 1st 2014, child benefits differed (to some extent) by the occupational regime of the parents: there were three socio-professional child benefit systems (for employees, the self-employed and civil servants) and one social assistance child benefit system. The three socio-professional systems were unified from July 1st 2014 onwards. Today, Belgium has four territorial systems in place due to the decentralisation of the child benefits (not included in the figure). Children eligible within the self- employed system are only included from 2014 onwards.

23

Certain issues follow from this administrative operationalisation of childhood disability. To start with, the supplemental child benefit is not only granted to children with a disability but to a broader group of children with increased care needs: long-term or seriously ill children (e.g. young cancer patients) could also be eligible. Moreover, this type of disability recognition presumably does not cover all children with a disability in Belgium (as shown in Chapter 2). Therefore, whenever possible, previous research based on the Flemish Families and Care Survey (see Section 4 for more information) is included in the discussion of the results to extend the conclusions to children with a disability who are not (necessarily) recognised for the supplemental child benefit. Besides, an administrative recognition of the disability does not automatically imply that parents consider their child to have a disability: according to the Flemish Families and Care Survey, 11% of children who are recognised by the FAPD are not identified as having increased care needs by their parents (Van Landeghem et al., 2007).

5.2. Child poverty

Poverty can also be conceptualised and operationalised in different ways (Greve, 2020). Child poverty is often seen as a lack of resources and operationalised through a monetary or non-monetary approach (Minujin, Delamonica, Davidziuk, & Gonzalez, 2006; Roelen & Gassmann, 2008). A monetary approach measures child poverty as the share of children living in a low-income household. This is the prevailing approach to child poverty in high-income countries (Cantillon et al., 2017), where a poverty line is determined to differentiate children who are poor from those who are not (Ravallion, 1994). It looks at child poverty indirectly, as it captures the living conditions of the household the children are living in, rather

24 than those of the children themselves. By contrast, a non-monetary approach operationalises child poverty as a lack of goods and services that are considered indispensable for child development and the realisation of their basic human rights and capabilities (Biggeri & Mehrotra, 2011). This can be considered a more direct approach to child poverty, as it aims to grasp what children need from their perspective.

To identify whether a child is living in poverty, I use the European headline at- risk-of-poverty indicator in this thesis, the predominant touchstone to gauge income poverty in Europe. The indicator measures child poverty as the share of children younger than 18 living in a household with an equivalised net disposable household income below a poverty line set at 60% of the national median equivalised net disposable household income. The OECD-modified equivalence scale is used to adjust the income to the household’s size and composition (Hagenaars, de Vos, & Zaidi, 1994). It is an indirect monetary approach to child poverty, as it supposes that child poverty is directly related to parental income. Moreover, the poverty measure is a relative indicator, as it compares the household’s income to the living standard considered to be necessary to live a minimally acceptable way of life (UNICEF, 2005). Throughout this thesis, income poverty is used to refer to the 60% at-risk-of-poverty indicator.

The indicator, however, is subject to a number of concerns, two of which are worth mentioning here (Decancq, Goedemé, Van den Bosch, & Vanhille, 2014). First, this measurement of child poverty assumes that all the resources within the household are shared equally among its members, including children, which is not necessarily true. Second, and potentially more important in the context of this thesis, by using the OECD-modified equivalence scale to make incomes comparable across households, it is indirectly presumed that households with the

25 same equivalised income can convert that income into the same standard of living. Yet, families of children with a disability face higher out-of-pocket costs to pay for the children’s additional medical and care needs (e.g. Stabile & Allin, 2012), which places an extra burden on the household budget. Accordingly, a focus on monetary resources potentially masks important aspects of children’s living conditions. Unfortunately, neither previous research nor the data at hand allow for taking these increased out-of-pocket costs into account, but their presence should be kept in mind throughout this thesis. Hence, no claims are made about the household’s standard of living.

5.3. Parental employment and social background

Parental employment is jointly measured by another European lead indicator labelled the “household work intensity” in Chapters 1, 3 and 5. It captures the combined degree to which all working-age household members (individuals aged 18-59, excluding students aged 18-24) participate in paid employment. It is defined as the ratio between their total number of months worked (in full-time equivalents) and the total number of months that they could have worked in theory (Ward & Özdemir, 2013). The ratio ranges from zero to one, with zero meaning that none of the working-age household members participated in paid employment in that year, while one indicates that they all worked full-time for the full year. It should be noted that the indicator can have a different meaning depending on the size of the household. For instance, a household work intensity of 0.8 could represent a single parent who works four days out of five, or it could refer to a couple with children where one parent works full-time and the other three days per week. In terms of household income, the couple’s income will most likely be higher than the single parent’s income. In terms of costs, the single parent is also

26 more likely than the couple to be confronted with higher childcare costs to balance work and family responsibilities. Moreover, with this joint labour market participation indicator, it is not possible to discern which parent works (and to what extent). Therefore, in Chapter 4, the labour market position of mothers and fathers is investigated separately, using a less detailed employment status (yes or no) to analyse the gender inequalities in employment participation in two-parent households with a child who has a disability.

Finally, the social background of the household that children belong to is measured by four different indicators in this thesis: the household type, the parents’ highest educational level, the parents’ country of birth, and the presence of other household members with a disability. Regarding the household type, single parents are contrasted with two-parent households. The parents’ educational qualifications are operationalised using the International Classification of Education (ISCED), differentiating between low-skilled (ISCED 0-2: lower secondary or less), medium-skilled (ISCED 3-4: secondary education) and high-skilled parents (ISCED 5-6: tertiary education). The educational level of the parent with the highest qualifications is taken as the household value (i.e. a dominance criterion is applied). For the country of birth, a distinction is made between parents born in Belgium, in other countries belonging to the European Union in 2010 (EU27) and in non-EU27 countries. When one of the parents is born in Belgium or in another EU27 country, the household is considered to have a Belgian or EU27 migration background respectively (i.e. a closeness criterion is applied). When the parents are born outside of the EU27, the household is assigned a non-EU27 migration background. Lastly, consistent with the operationalisation of childhood disability, the disability of other household members is identified by the receipt of a disability-specific benefit. A dummy variable indicates whether at least one other

27 household member receives either the supplemental child benefit (for minus 21- year olds) or a disability benefit for individuals aged over 21. An administrative recognition of the disability is needed for the latter too but, contrary to the supplemental child benefit, the receipt of the benefit is also conditioned on a means test.

6. Overview of the thesis

The thesis is structured in three parts and five chapters. The first part focusses on income poverty among children in general and examines its trends and determinants in Belgium. It is the background against which the remainder of this thesis should be read. Chapter 1 shows that Belgium puts down a poor, disappointing and worrying performance in terms of poverty among children in general. The analyses demonstrate that the rising vulnerability of children started in the good years before the Great Recession. The steady rise is related to the relatively high and growing share of children in households where no adult at active age participates in the labour market and to the decreasing redistributive impact of social policies on poverty rates. In addition to jobless households, children living with single parents, parents with a migration background and parents with lower educational qualifications are at an increased risk of living in poverty.

The second part of the thesis is centred around childhood disability. Chapter 2 discusses the uptake of available cash and care policies for children with a disability and investigates their overlap and mismatch. The chapter focuses on the question: what are the magnitude, characteristics and determinants of non-take- up in the Belgian supplemental child benefit? In doing so, the differentiated Belgian policy landscape is exploited (see Section 3.1 for more information). A

28 mixed-method approach is applied, revealing that (1) at least 10% of children with a disability in Flanders do not take up the supplemental child benefit, (2) children with autism spectrum disorders, intellectual and psychological disorders are the main non-take-up group, and (3) insufficient provision of information, the complexity of the application process and the construction of the scale to assess eligibility are important non-take-up drivers.

The remainder of the second part examines the relationship between childhood disability and parental employment. Chapter 3 questions whether parental employment among families with a child with a disability can be explained by the child’s disability, the family’s social background, or both. An intersectional approach is adopted to explore whether childhood disability overlaps with disadvantaged social background characteristics and whether they reinforce each other in impeding parental employment. The results show that children with a disability are overrepresented in households with a disadvantaged social background: they more frequently live together with parents holding lower educational qualifications, single parents, and other household members with a disability. Moreover, parents of children with a disability work less than parents of children without a disability, a gap that is only partially explained by their disadvantaged social background. However, the extent of the gap differs according to the family’s social background: for single parents, parents with low and medium educational qualifications and parents who have multiple children with a disability, childhood disability and social background reinforce each other. This reinforcing effect is not found for parents with a migration background, parents who have a disability themselves or who live together with other adults with a disability.

29

Chapter 4 adds a gender perspective as well as a comparative perspective to the parental employment gap studied in the previous chapter. Employing comparable administrative data for Belgium and Norway, this chapter examines whether gender inequalities, education inequalities and country differences exist in parental employment and earnings among families with a child with a disability. The chapter demonstrates that parents of children with a disability work and earn less than parents of children without a disability in both countries, this is especially the case for mothers and lower skilled parents. For parental employment, this is driven by (1) lower employment probabilities for mothers and lower skilled parents in general, as well as (2) an intensification of these inequalities among the parents of children with a disability. Additionally, the parental employment gap depends on the country of residence, as significantly larger inequalities are reported in Belgium compared to Norway. The story is different for the parental wage gaps, however. The significantly lower earnings for mothers and lower skilled parents of children with a disability are largely the result of gender and education inequalities that exist among parents in general and are not the consequence of intensified inequalities among parents with children with a disability. These inequalities in the wage gap are not significantly larger in Belgium than in Norway.

In the third part, childhood disability and child poverty are brought together. Chapter 5 investigates whether, and if so why, childhood disability coincides with child poverty in Belgium. This chapter connects the insights from all previous chapters and thus examines the poverty risk of children with a disability while the interplay between the family’s labour market participation, social background and receipt of targeted cash support is taken into account. The results show that children with a disability have a lower income poverty risk than children without

30 a disability, even when their family’s labour market participation and social background are controlled for. This can be explained by the targeted cash support these families receive: the supplemental child benefit has a strong poverty- reducing impact for children with a disability, as the families who receive this benefit are more likely to live in income poverty in the first place: i.e. families with lower levels of labour market participation and families with a disadvantaged social background.

The concluding chapter of this thesis summarises the three key findings that can be derived from the previous chapters and highlights its contributions to the literature. The main limitations and avenues for future research are then discussed. Finally, three interrelated recommendations are formulated to guide policymakers in making (further) progress in fighting poverty among children with a disability.

31

PART I

CHILD POVERTY

33

CHAPTER 1 BELGIUM: CREEPING VULNERABILITY OF CHILDREN

Published as Vinck, J., Van Lancker, W., & Cantillon, B. (2017). Belgium: Creeping vulnerability of children. In B. Cantillon, Y. Chzhen, S. Handa & B. Nolan (Eds.), Children of austerity: The impact of the Great Recession on child poverty in rich countries (pp. 30-55). Oxford, United Kingdom: The United Nations Children’s Fund and Oxford University Press.

Abstract

Belgium has been plagued by comparatively high levels of child poverty, and by a creeping, yet significant, increase that started in the good years before the crisis. This is related to the relatively high share of households with an extremely low attachment to the labour market, the extremely high and increasing poverty risk of children growing up in these households, and benefits that are inadequate to shield jobless families with children from poverty. Although the impact of the Great Recession was limited in Belgium, the crisis seems to have had an impact on child poverty, by increasing the number of children living in households with no or only a marginal attachment to the labour market. Although the Belgian welfare state had an important cushioning impact, its poverty reducing capacity was less strong than it used to be in the past. The most important lesson from the crisis is that in order to make further headway in reducing child poverty, not only activation but also social protection should be improved.

35

1.1. Introduction

Belgium is one of the most advanced welfare states in the world. It has a highly developed social security system, high levels of social spending, and low levels of income inequality, which have remained fairly stable over the past few decades. Moreover, the country fared pretty well during the Great Recession that followed the worldwide 2007 – 2008 financial crisis (Jenkins et al., 2013). The impact in terms of job loss and gross domestic product (GDP) contraction was rather modest, and the Belgian economy quickly regained its pre-crisis levels. The Organisation for Economic Co-operation and Development (OECD) and the European Commission have both hailed the role of the Belgian safety net in safeguarding the welfare of the majority of citizens (European Commission, 2010; OECD, 2011).

Despite these achievements, however, the Belgian welfare state has been plagued by comparatively high levels of child poverty, and by a creeping, yet significant, increase in child poverty that started in the good years before the crisis. This suggests structural shortcomings in the way that children are protected against the risk of living in poverty. In Belgium, the share of jobless households is relatively high; meanwhile the redistributive impact of social policies on poverty rates has declined since the 1990s, in consequence of a declining benefit adequacy for vulnerable families and a shift from working-age cash spending to pensions, healthcare, and working-age in-kind spending (Cantillon & Vandenbroucke, 2014).

In this chapter we discuss how children fared before, during, and after the Great Recession in Belgium. We focus on the nature of the crisis in Belgium, discuss trends in child poverty and its determinants, and examine whether the crisis has

36 had an impact on pre-existing child poverty trends, and if so, how. Finally, we discuss how social and fiscal policies should be recalibrated in order to improve the living conditions of Belgian children.

1.2. The Great Recession and its aftermath

In the pre-crisis period between 2004 and 2008, Belgium experienced a spell of weak but stable economic growth of 0.7% on average. When the crisis struck in the third quarter of 2008, GDP fell by about 4% (Figure 1.1), and GDP growth was negative for four successive quarters. In the third quarter of 2009 GDP growth was again positive, and by mid-2010 GDP was back to its pre-crisis levels. The global financial and economic crisis hit Belgium mainly through two channels. First, with its small, open economy, Belgium suffered from the drop in global trade, which affected consumption and economic activity (European Commission, 2010). Second, the sustainability of Belgian banks was highly dependent on the international financial system, and the collapse of Lehman Brothers triggered government intervention to sustain all major banks, affecting public debt ratios (IMF, 2009). From 2011 to 2013 Belgium experienced a period of stagnation, including a brief period of negative growth (a so-called “double dip”). From the second quarter of 2013 onward, economic growth was restored, albeit at a slow rate of less than 0.5%. In this chapter, we focus on the first phase of the crisis.

Although the impact of the crisis in terms of economic growth was rather modest, the decline in GDP was the strongest since the Second World War (De Mulder & Druant, 2011).

37

Figure 1.1. Quarterly GDP growth rate (left axis) and GDP relative to 2010 level (right axis), Belgium, 2000 – 2015

2.5% 120 115 1.5% 110

0.5% 105 100 -0.5% 95 90 -1.5% 85

-2.5% 80

2006Q1 2000Q1 2000Q3 2001Q1 2001Q3 2002Q1 2002Q3 2003Q1 2003Q3 2004Q1 2004Q3 2005Q1 2005Q3 2006Q3 2007Q1 2007Q3 2008Q1 2008Q3 2009Q1 2009Q3 2010Q1 2010Q3 2011Q1 2011Q3 2012Q1 2012Q3 2013Q1 2013Q3 2014Q1 2014Q3 2015Q1 2015Q3

quarterly growth rate in real GDP GDP (2010=100)

Source: EUROSTAT (2016b), National Bank of Belgium (2016). Notes: GDP adjusted for seasonal cycles and adjusted by working days.

The initial response of the Belgian government to the crisis was to prevent job loss as much as possible. In doing so, on the one hand it encouraged employers to make use of the existing scheme of so-called ‘temporary unemployment’ for blue-collar workers, while on the other hand it introduced a number of crisis measures for white-collar workers (see Section 1.4). Figure 1.2 shows the unemployment rate over time for Belgium and three neighbouring countries: Germany, the Netherlands, and France. During the pre-crisis period of economic growth, which was almost identical to the French experience, the unemployment rate declined from 8.6% in 2005 to 6.3% in the second quarter of 2008, at which point the crisis kicked in. The unemployment rate peaked again at 8.7% in the third quarter of 2010. In contrast to what happened in France and the Netherlands, Belgian unemployment rates declined again to 6.6% in 2011. The German labour market

38 was hardly affected by the crisis (Bahle & Krause, 2017). In the fourth quarter of 2015, unemployment reached a peak of 8.7%, still 1.9 percentage points (pp) lower than France, and 2.6pp lower than the average for the Eurozone.

Figure 1.2. Quarterly unemployment rate, Belgium and neighbouring countries, 2005 – 2015

14

12

10

8

6

4

Unemployment Unemployment rate(%) 2

0

2005Q1 2005Q3 2006Q1 2006Q3 2007Q1 2007Q3 2008Q1 2008Q3 2009Q1 2009Q3 2010Q1 2010Q3 2011Q1 2011Q3 2012Q1 2012Q3 2013Q1 2013Q3 2014Q1 2014Q3 2015Q1 2015Q3

Belgium Germany France Netherlands

Source: EUROSTAT (2016c). Notes: Unemployment according to ILO definition: persons aged 15-74 who are not employed for at least one hour and are not temporarily absent from work during the reference week, are available for work and are actively seeking work.

In sum, compared to what happened in France and the Netherlands, the Belgian labour market recovered rather quickly in the first phase of the crisis, and employment losses were limited (see Figure 1.3). Although initial employment rates were low (62.4% versus 64.9% in France, 70.1% in Germany, and 77.2% in the Netherlands), the total employment rate declined by only 0.4pp. In Germany, employment rates continued to grow, while in France (-0.9pp) and in particular in the Netherlands (-2.5pp) employment losses were much more tangible.

39

Figure 1.3. Percentage point change in employment rates by age and education,

Belgium and neighbouring countries, 2008 − 2010 total

high

medium education low

55-64

age 25-54

15-24

-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 Percentage point change

Netherlands France Germany Belgium

Source: EUROSTAT (2016a). Notes: Employment according to ILO definition: persons aged 15 or more who are employed for at least one hour or are temporarily absent from work during the reference week.

However, the burden of the crisis was not shared equally. As in the neighbouring countries, young workers were hit hardest: while the employment rate for older workers (aged 55-64) increased by almost 3pp, the employment rate of young workers (15-24) declined by more than 2pp. In particular, industry, construction, the financial sector, administrative and support services, and the information and communication sector suffered job losses (Bulté & Struyven, 2013). As a consequence, and in contrast to the neighbouring countries, it was not the low- skilled, but rather the high- and medium-skilled groups that experienced the biggest setbacks in employment rates.

40

1.3. Child poverty before, during and after the crisis

1.3.1. Trends in living conditions of children in Belgium

In this section we analyse child poverty trends from 2005 to 2013, drawing on the European Union (EU) portfolio of primary indicators of social inclusion. Figures 1.4 and 1.5 show poverty trends for different age groups before, during, and after the crisis, using both a relative (Figure 1.4) and a more absolute poverty threshold “anchored” in 2008 (Figure 1.5). The poverty threshold corresponds to 60% of median equivalised household income. These monetary indicators are complemented by a measure of severe material deprivation (SMD) in Figure 1.6. SMD represents the share of the population who report that they cannot afford at least four out of nine items or activities deemed important for maintaining a decent standard of living. See Nolan and Whelan (2011) for more information.

Unlike those countries hit hard by the recession, in Belgium median incomes continued to grow throughout the period, and the relative poverty risk of the population remained stable at around 15%. Figure 1.4 shows, however, that this pattern of stability disguises a shift in the relative poverty risk from the elderly to children (and the active-age population to a lesser extent). This intergenerational shift is observable in many other countries (see Bradshaw et al., 2012; Förster & Mira D’Ercole, 2005; OECD, 2008; 2014b; UNICEF, 2014) though with important cross-country and temporal variations. In 2005, the relative poverty rate of children (aged under 18) was 15%, this is below the European average, but well above the neighbouring countries of France, Germany, and the Netherlands in terms of child poverty risk. Figure 1.4 illustrates the fact that child poverty went up to 17% in 2008, and increased further to 19% during the crisis period. Afterwards the rate fluctuated somewhat, but was back at 19% in 2013. The rate

41 for active-age adults remained stable at about 12% until 2009, and rose to 14% in 2013. The poverty rate for the elderly (aged 65 and over) fell sharply over the whole period, from 23% in 2005 to 16% in 2013.

Figure 1.4. Relative poverty risk in Belgium, before, during, and after the crisis, income years 2005 – 2013

25

20

15

10

5 Relative poverty risks Relativepoverty risks (%)

0 2005 2006 2007 2008 2009 2010 2011 2012 2013

Total population Children (0-17) Active age (18-64) Elderly (65+)

Source: authors’ calculations using the EU Statistics on Income and Living Conditions (EU-SILC, 2006 – 2014). Notes: income years instead of survey years. The two vertical black lines mark the first phase of the crisis, in which the Belgian labour market was affected.

The anchored poverty measure in Figure 1.5 shows similar (albeit less marked) developments. Child poverty fluctuated around 15%, went up to 17% during the crisis, and fluctuated around 16% after the crisis. Elderly poverty declined strongly, while the poverty rate for active-age adults went up during the crisis.

42

Figure 1.5. Anchored poverty risk in Belgium, before, during, and after the crisis, income years 2005 – 2013

25

20

15

10

5 Anchored povertyrisks (%) 0 2005 2006 2007 2008 2009 2010 2011 2012 2013

Total population Children (0-17) Active age (18-64) Elderly (65+)

Source: authors’ calculations using EU-SILC (2006 – 2014). Notes: see Figure 1.4. The fixed threshold anchors the poverty threshold at the 2008 (income year 2007) level.

Figure 1.6 shows the evolution of SMD rates for different age groups in Belgium. It is clear that the deprivation rate for children has been consistently higher than for pensioners. The rate of deprivation dropped from 10% to about 7% before the crisis, went up again at the peak of the crisis, after which it fluctuated around 8%. The SMD for children dropped sharply in 2013. Corroborating the decline in monetary poverty, SMD rates among the elderly are low and exhibit a downward trend. For other age groups, trends in SMD (if any) have been much less prominent.

43

Figure 1.6. Severe material deprivation in Belgium, before, during, and after the Crisis, survey years 2006 – 2014

10 9 8 7 6 5 4 3 2 1 0

2006 2007 2008 2009 2010 2011 2012 2013 2014 Severe material deprivationrates (%) Total population Children (0-17) Active age (18-64) Elderly (65+)

Source: authors’ calculations using EU-SILC (2006 – 2014). Notes: the horizontal axis represents the EU-SILC survey years 2006 – 2014. The two vertical black lines show the crisis years.

The SMD indicator captures material deprivation of the household in which children live, rather than material deprivation of children themselves. An EU- SILC ad-hoc module carried out in 2009 and from 2013 onwards includes a series of questions that refer directly to the lives of children by including child-relevant items and activities on the list. This allows us to gain more insight into the living conditions of children and how these have changed over time. Although it is not possible to estimate the impact of the crisis, the data provide detailed information on how children’s living standards evolved in the post-crisis period. Table 1.1 shows the share of children (aged under 18) and the share of poor children (aged under 18) living in households that were not able to afford these child-specific items in 2009, and again in 2014.

44

Table 1.1. Share of children and share of poor children living in a household that cannot afford child-specific material deprivation items, Belgium, 2009 and 2014

Share of poor Children <18 Share of children children Child-specific material 2009 2014 t-test 2009 2014 t-test deprivation items Some new clothes 6.2% 8.5% * 18.7% 27.8% * Two pairs of properly fitting shoes 3.7% 3.8% 7.5% 14.4% * Fresh fruit and vegetables once a 1.7% 2.2% 6.8% 8.6% day One meal with meat, chicken, fish, or vegetarian equivalent at least 3.1% 2.7% 13.5% 8.7% once a day Books at home suitable for their 3.9% 4.3% 10.3% 16.4% age Outdoor leisure equipment 3.8% 4.2% 14.2% 14.4% Indoor games 2.1% 2.5% 9.8% 9.0% Regular leisure activity 7.8% 8.7% 28.0% 30.7% Celebrations on special occasions 3.1% 5.7% *** 10.4% 16.7% * Invite friends round to play and eat 2.9% 6.1% *** 12.7% 23.4% ** from time to time Participate in school trips and 3.0% 4.2% 7.9% 14.2% * school events that cost money Suitable place to study or do 7.9% 11.6% ** 21.1% 32.1% ** homework Lacks at least one item 18.6% 19.3% 53.2% 54.8% Source: authors’ calculations using EU-SILC (2009 and 2014). Notes: the child-specific material deprivation items are asked to the household head for children aged 1–15, though the proportions in this table are for (poor) children aged below 18. * = significant at 0.10 level; ** = significant at 0.05 level; *** = significant at 0.01 level.

In Belgium, 19% of children lack at least one of the child-specific material deprivation items. This is strongly related to income poverty: 54% of poor children lack at least one item. Items or activities that children lack the most are having a suitable place in which to do homework, being able to participate in regular leisure activities, and getting some new clothes. One child in ten has no suitable place to study, a proportion that for poor children rises to one-third. In addition, Table 1.1

45 shows a significant increase over the past few years in the number of poor children unable to participate in school trips, to invite friends, or to celebrate on special occasions, and who lack two pairs of properly fitting shoes. It is not the case that more poor children became materially deprived between 2009 and 2014, but those that were deprived of at least one item saw their situation deteriorate even further.

Finally, it is important to acknowledge that marked differences in child poverty rates and living standards prevail across the three Belgian regions (Guio, Vandenbroucke & Vinck, 2015; Vandenbroucke & Vinck, 2013). In the mid- 2000s, 10% of Flemish children lived in relative poverty, compared with 20% in Wallonia and even, in , around 30%. During the crisis, these regional differences were reinforced: the child poverty rate increased to 25% in Wallonia and to almost 40% in Brussels, while remaining stable in Flanders. After the crisis, child poverty rates tended to converge somewhat: they declined in Wallonia (- 5pp), remained stable in Brussels (-1pp), and increased in Flanders (+3pp) (Guio, 2016).

1.3.2. The role of the labour market

Since relative income poverty is measured at the household level, it is necessary to take the labour market attainment of parents into account, in order to appreciate the impact of the recession on child poverty. The extent to which parents are able to earn income from paid labour largely determines the poverty risk of their children. As a corollary, job loss might be the driving force underlying the increase in the child poverty rate during the recession. In this section, we examine how child poverty is related to the work intensity of parents; in the next, we discuss whether the impact of the crisis on the labour market can account for the increase in child poverty during that period.

46

Parental labour market attainment is measured by the household work intensity indicator, representing the ratio between the total number of months worked by working-age household members and the total number of months that they could, in theory, have worked. In very low work intensity (VLWI) households, parents work for less than 20% of their potential and have only loose ties to the labour market, the typical example being jobless households with no or only little income from paid work. In contrast, in high work intensity (HWI) or very high work intensity (VHWI) households parents work for more than 55% of their potential, and have strong ties to the labour market, a typical example being a two-earner family. Low work intensity (LWI) and medium work intensity (MWI), households work for between 20% and 55% of their potential, for instance breadwinner families or single parents working part time (Corluy & Vandenbroucke, 2014).

Figure 1.7 shows that the poverty risk of children is basically an inverted picture of the work intensity of the household in which they live. Merely 5% of children living in HWI-VHWI households are at risk of being poor in Belgium, while for VLWI households the figure is 77%. This is significantly higher than the poverty risk of children living in VLWI households in the neighbouring countries. Moreover, in 2014, 13% of children and 53% of poor children lived in VLWI households, only crisis-struck Ireland reported higher levels. Hence, one piece of the Belgian child poverty puzzle: a high share of children live in VLWI households, and those children run a particularly great risk of being poor (see Vandenbroucke & Vinck, 2013, for further reading).

47

Figure 1.7. Child poverty risk by household work intensity, income year 2013, Belgium and neighbouring countries

100 90

80 17] (%) 17]

- 70 60 50 40 30 20

10 Child Child povertyrisk [0 0 VLWI (0-20%) LWI-MWI (20-55%) HWI-VHWI (55-100%) Household work intensity

BE Neighbouring countries (DE, FR, NL)

Source: authors’ calculation using EU-SILC (2014). Notes: The dashed line refers to the unweighted average of the three main neighbouring countries (Germany, France, and the Netherlands). Working-age household members are defined as all individuals between 18 and 59, except students between 18 and 24. The difference in the poverty risk between VLWI households in Belgium and VLWI households in neighbouring countries is statistically significant (p < 0.05).

1.3.3. Gauging the impact of the crisis

Figure 1.4 showed that child poverty in Belgium started increasing well before the onset of the crisis, but that during the crisis period child poverty rates increased at an even faster rate. The question now is whether this pattern can be attributed (at least partially) to the impact of the crisis on the labour market. This will be answered by means of decomposition analysis (Corluy & Vandenbroucke, 2014; Vandenbroucke & Vinck, 2013). Such analysis allows us to estimate the extent to which (1) changes in the share of children living in households with different work intensity levels and (2) changes in the poverty risk of those children contribute to

48 changes in the child poverty rate, and how the role of these factors changed before, during, and after the crisis.

Since the recession in Belgium translated into job losses but not into a fall in median incomes, one impact of the crisis would be that changes in the child poverty rates during that period are explained by changes in the share of children living in VLWI households, and not by changes in the poverty risk of these children.

Table 1.2 shows that before the crisis, the increase in child poverty of 1.3pp can be explained by an increase in the poverty risk of children living in VLWI households (+0.9pp) and of children living in households with higher work intensity levels (+1.9pp), while a declining share of children living in VLWI households mitigated these trends (-1.4pp). During the crisis period, child poverty rates increased by 2.0pp. This can be wholly attributed to an increase in the share of children living in VLWI households (+2.0pp), while the poverty risk of children living in VLWI (+0.2pp) and in other households (-0.3pp) remained fairly stable. After the crisis, the child poverty rate persisted at about the same level, because the effect of a declining share of children living in VLWI households (-0.5pp) and a decreasing poverty risk of these children (-0.6pp) was undone by an increase in the poverty risk of children living in other households (+1.5pp).

These results suggest, first, that the crisis was indeed a catalyst for increasing child poverty rates, due to its impact on the labour market; and second, that the employment recovery after the crisis has not translated into lower poverty rates, suggesting that labour market participation as such has become less of a protection against poverty. Even though the impact of the crisis was short-lived in Belgium, previous research suggests that even short spells of living in poverty have a

49 scarring effect on the future life opportunities of children, in that they increase the likelihood of future poverty spells (Fouarge & Layte, 2005).

Table 1.2. Decomposition of changes in Belgian child poverty rates on the basis of household work intensity, before, during, and after the crisis

Explanatory factors Change explained by Child Child Child Child Child Child Share of poverty Share of poverty poverty poverty poverty poverty children risk children risk risk change risk VLWI other VLWI other VLWI VLWI WI WI 2005 15.3% 13.1% 72.2% 6.7% 2008 16.6% 11.0% 79.5% 8.8% 1.3pp -1.4pp 0.9pp 1.9pp 2010 18.7% 13.8% 81.0% 8.5% 2.0pp 2.0pp 0.2pp -0.3pp 2013 18.8% 13.0% 76.6% 10.1% 0.2pp -0.5pp -0.6pp 1.5pp Source: authors’ calculations using EU-SILC (2006, 2009, 2011, and 2014). Notes: 2005, 2008, 2010, and 2013 refer to income years. Percentage point (pp) change and relative to previous row in the table.

1.3.4. Sociodemographic characteristics

The risk of being poor is not randomly distributed among children, but is tied to the sociodemographic characteristics of the household in which poor children live: characteristics such as parental education, family structure, and migration status. In turn, these characteristics impact on the opportunities for parents to earn sufficient income in the labour market. Figure 1.8 shows the share of children and the share of poor children by sociodemographic characteristics of the household.

50

Figure 1.8. Share of children and share of poor children by family characteristics: single parents, non-EU households, and highest educational level of the parents, 2014

100 90 80 70 60 50 40 30

20 Share children of (%) 10 0 Share of all Share of Share of all Share of Share of all Share of children poor children children poor children children poor children Single parents Non-EU HHs Parental education

Low Medium High

Source: authors’ calculations using EU-SILC (2014). Notes: non-EU households are defined as households where at least one parent was born outside the EU; the highest educational level observed among the parents is taken as the educational level of the household (dominance criterion).

It is clear that poor children are overrepresented in single parent families (34% of poor children), in families with at least one parent born outside the EU (45% of poor children), and in families with poorly educated parents (34% of poor children). Poor children are also slightly overrepresented in large families (i.e. three or more children), but this is less prominent than the aforementioned characteristics. Therefore, living in a large family is not included as a separate characteristic in the forthcoming analyses. While having low qualifications is an important determinant of living in poverty and a strong predictor of occupational success (Gesthuizen & Scheepers, 2010; Troger & Verwiebe, 2015), it is important to note that 45% of poor children live in a household with at least one medium-

51 skilled parent, and 20% live in a household where at least one parent is highly educated. Child poverty is clearly not to be reduced to lack of educational qualifications. In the same vein, while family structure and migration status are important factors underlying child poverty, a fair share of poor children live in families that do not share these characteristics.

This is confirmed by Table 1.3, in which the overlap between family structure, migration status, and level of education for poor children is shown. First of all, 32% of poor children live in ‘other’ households, and less than a third of these children have low-skilled parents. Second, the overlap between single parents and non-EU families is limited to 10% of poor children. Third, while the children of highly skilled single parents are hardly at risk of being poor, for children living in families with a non-EU migration background that risk is much more tangible. The relationship between parental educational qualifications and child poverty differs for different types of household.

Table 1.3. Overlap between family characteristics for poor children, Belgium, income year 2013

Low Medium Highly Poor children Total educated educated educated Single parent 8.6% 12.8% 2.4% 23.8% Non-EU HH 10.7% 13.3% 10.5% 34.5% Both single parent & non-EU 6.0% 3.0% 1.2% 10.2% Other 8.9% 15.9% 6.8% 31.6% Total 34.1% 45.0% 20.8% 100% Source: authors’ calculations using EU-SILC (2014). Notes: non-EU households are defined as households where at least one parent is born outside the EU. The highest educational level obtained by the parents is taken as the household educational level.

All in all, the sociodemographic characteristics of poor children’s families reveal a second piece of the Belgian puzzle. About 68% of poor children grow up in a

52 single parent family and/or a family with a non-EU migration background. Together with Austria and Sweden, Belgium has the highest share of any European country (analyses not shown). Yet, while poor children are overrepresented in families with low-skilled parents, the majority of the parents of poor children living in a single parent family and/or a family with a non-EU migration background outside the EU are not low-skilled.

1.3.5. Work intensity and sociodemographic characteristics

In the previous sections, it was shown that child poverty in Belgium is related to the work intensity of the household on the one hand, and to sociodemographic characteristics on the other. In this section, these two pieces of the puzzle will be fitted together. Figure 1.9 shows the relationship between child poverty, household work intensity, and the two types of households in which poor children are overrepresented: single parents (left panel) and households with a non-EU migration background (right panel).

The relationship between child poverty and household work intensity differs between single parents and families with a non-EU migration background. The distribution of children living in single parent families over work intensity takes the form of a U-shape: single parents tend to be either strongly or weakly attached to the labour market. The share of children living in families with a non-EU migration background is much more equally distributed over work intensity categories. Importantly, the figure provides further evidence that attaining paid work is not a sufficient condition for combating child poverty: 22% of children living in HWI-VHWI single parent families and 13% of children living in HWI- VHWI families with a non-EU migration background are at risk of being poor, versus only 6% of all children. Interestingly enough, 67% of children living in

53

VLWI single parent families and 86% of VLWI families with a non-EU migration background are living in poverty, versus 77% of all children. In sum, non-working single parents are somewhat better protected against poverty than are non-working families with a non-EU migration background, while working families with a non- EU migration background are somewhat better protected against poverty than working single parents. This is related to the structure of the household: single parents are single earners (by definition), and it is increasingly difficult to maintain a sufficient living standard on only one income (Daly, 2005; Van Lancker, Ghysels, & Cantillon, 2015).

Figure 1.9. Children living in single parent households (left panel) and non-EU households (right panel): share and poverty risks over household work intensity, Belgium, income year 2013

Single parents Non-EU households 60 100 40 100 50 35 80 30 80 40 60 25 60 30 20 20 40 15 40 20 10 20 10 5

0 0 0 0

Share children of (%) Share children of (%)

Child Child povertyrisk (%) Child povertyrisk (%)

Share children (0-17) living in single-parent Share children (0-17) living in non-EU HHs HHs over WI groups over WI groups Child poverty risk (0-17) in single-parent Child poverty risk (0-17) in non-EU HHs HHs over WI groups over WI groups Child poverty risk (0-17) in other HHs over Child poverty risk (0-17) in EU HHs over WI groups WI groups Source: authors’ calculations using EU-SILC (2014). Note: non-EU households are defined as households where at least one parent is born outside the EU.

54

Figure 1.10, finally, shows the relationship between child poverty, parental education, and work intensity of the household. Three observations can be made. First, there is almost no difference in the child poverty risk over work intensity between low- and medium-educated households. Being highly skilled, however, is consistently associated with a lower child poverty risk. Second, here, too, the results show that having paid employment as such is not enough: respectively 9% and 17% of children living with low- and medium-skilled parents with high levels of work intensity are poor. Since the majority of children with low- and medium- skilled parents live in HWI-VHWI households, these are no trivial numbers. Third, the picture confirms that attaining a high level of education is an insufficient condition to avoid poverty as well: 15% of children with highly educated parents live in VLWI or LWI-MWI households with child poverty rates of 62% and 18%, respectively.

In sum, educational policies will not help to combat child poverty if high qualifications do not lead to high levels of work intensity, while policies that focus on employment will not deliver, if even high levels of work intensity are inadequate to steer clear of poverty. This calls for a renewed emphasis on income protection, for jobless and working families with children alike.

55

Figure 1.10. Children living in households with low-, medium-, or high-skilled parents: share and poverty risk over work intensity, Belgium, income year 2013

80 100

60 80 60 40 40 20 20 0 0

Share children of (%) VLWI LWI-MWI HWI-VHWI Child Child povertyrisk (%)

Share children (0-17) living in low-educated HHs over WI groups Share children (0-17) living in medium-educated HHs over WI groups Share children (0-17) living in high-educated HHs over WI groups Child poverty risk (0-17) in low-educated HHs over WI groups Child poverty risk (0-17) in medium-educated HHs over WI groups Child poverty risk (0-17) in high-educated HHs over WI groups

Source: authors’ calculations using EU-SILC (2014). Note: the highest educational level obtained by the parents is taken as the household educational level.

1.4. Policies

1.4.1. Policies before the crisis

Belgium’s policy response to the great economic, social, and demographic transitions of the 1970s and 1980s has been depicted as a prime example of “welfare without work” (Esping-Andersen, 1996). In that period, Belgium was characterised by low employment rates, a high level of social expenditure, a strong social safety net, and low (and stable) poverty rates (Cantillon, 1999). Since the second half of the 1990s, haunted by historically high levels of public debt (reaching a peak of 135% of GDP in 1993) and pressured by the obligation to

56 reduce public debt under the Maastricht Treaty, successive Belgian governments slowly contributed to a process of policy reorientation, in order to achieve budgetary restraint, employment growth, and equality.

In hindsight, that process of reorientation can be interpreted as a “social investment turn” in policy making (Hemerijck, 2012). The ambition was for higher employment rates to translate into lower poverty rates and reduced social spending on “passive” cash transfers, thus creating budgetary room to invest in more “productive” activation policies and human capital policies (i.e. childcare, higher education). This in turn was projected as supporting higher levels of labour market participation. On the face of it, the strategy seems to have delivered. Employment rates started to rise; and while overall social spending remained steady at a high level, there was a shift in the nature of spending (Figure 1.11): spending on elderly care, childcare, career breaks, education, and active labour market policies increased from 1.3% of GDP in 1985 to 5.1% of GDP in 2008, while spending on cash benefits for the active-age population declined from 9.6% of GDP in 1985 to 7.4% in 2008 (Cantillon & Vandenbroucke, 2014).

However, the “social investment turn” largely failed to deliver on its promise to reduce poverty. Although the design of many individual tax-benefit policies were pro-poor (Decoster, Perelman, Vandelannoote, Vanheukelom, & Verbist, 2015), below the radar there was a creeping vulnerability of children (Vandenbroucke & Vinck, 2013). The analyses of child poverty presented in this chapter have demonstrated that a policy paradigm focusing on human capital investment, including expansion of childcare and higher education, and labour market activation, falls short of reducing child poverty. Four reasons can be discerned: first, employment growth has benefited VLWI households much less than it has benefited households with higher levels of work intensity (Corluy &

57

Vandenbroucke, 2014). Second, as a consequence, government investment in policies that are grafted onto labour market participation (such as childcare services) have also mainly benefited work-rich households (Van Lancker, 2013). Third, access to higher education remains socially stratified (OECD, 2014a). Fourth, despite continuous upwards adjustments of social benefits in the period before the crisis, cash benefits for the active aged (i.e. minimum unemployment benefits, social assistance, and child benefits) became less adequate in protecting these families from poverty (see below).

Figure 1.11. Social spending on persons aged less than 65 as a percentage of GDP, functional distribution, Belgium, 1985 – 2011

16 14 12 10 8 6 % % GDP of 4 2

0

2002 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2003 2004 2005 2006 2007 2008 2009 2010 2011

Cash benefits Parental leave Elderly care Childcare ALMP Primary and secondary education Tertiary education

Source: calculations by Linde Buysse and Annemie Nys, based on Meeusen and Nys (2014). Note: spending on cash benefits includes spending on unemployment benefits, sickness and disability benefits, and child benefits.

58

1.4.2. Policies during the crisis

During the Great Recession, the Belgian government implemented a number of crisis measures to prevent job losses. First, before the crisis struck, a temporary lay-off scheme that allowed employers to partially or fully suspend the employment contract of blue-collar workers in case of severe economic or natural circumstances (i.e. bad weather, strikes, economic downturns) was already in place. This “temporary unemployment” scheme grants affected workers (partial) unemployment benefit, even though they retain a contractual bond with their employer; while employers save the costs associated with dismissal (Van Gyes, 2010). Employers resorted to this measure in high numbers: in the first quarter of 2009, some 279,643 workers were temporarily unemployed for at least a day, a 75% increase over the first quarter of 2008.

Second, in April 2009 the government and the social partners agreed on crisis measures for white-collar workers, with the explicit aim of keeping job losses and dismissals at bay. Three provisions were implemented: a so-called short-time work arrangement scheme, meant to facilitate a temporary reduction in working time at the company level; the extension of the temporary unemployment scheme to white-collar workers; and a specific crisis-related extension of the existing time credit scheme at the individual level (Van Gyes, 2009).

Finally, as part of the crisis measures, the Belgian government increased the levels of social assistance benefits and levels of temporary unemployment benefits by 2% above price indexation (Marchal, Marx, & Van Mechelen, 2014). This seems to be in line with initial European Commission recommendations to increase transfers to the poor, since these are likely to be spent immediately, and hence will contribute to improving aggregate demand (European Commission, 2008).

59

As a consequence of the increase in unemployment, the compensation for reduction in working hours, and the slight increase in some benefits for the unemployed, spending on unemployment benefits went up by 0.5pp of GDP. Belgium’s public debt ratio again increased to 97% of GDP in 2010, and surpassed 100% of GDP in 2012. Successive Belgian governments had reduced the public debt ratio quite successfully to 84% in 2007, from its peak in the 1990s. The crisis hence halted two decades of falling public debt ratio (IMF, 2011).

From 2012 onwards, following a general trend among European welfare states, the Belgian government started following a more austerity-guided policy agenda, including a focus on fiscal consolidation through the reduction of social expenditure and the implementation of cost-cutting measures. This has been partially fuelled by fears of increasing public debt ratios and a looming impact of ageing, two evolutions that have been designated as potential threats to the long- term fiscal sustainability of Belgium by the European Commission. At the federal level, these measures (decided upon or implemented by the Di Rupo government from 2011 until 2014, and by the Michel government thereafter) included imposing stricter conditions on unemployment benefit entitlement, a sharper decrease in levels of unemployment benefit over time (while not imposing a formal time limit on benefit entitlement), and first limiting and subsequently scrapping an activation allowance (i.e. “wachtuitkering”) covering unemployed people who did not qualify for regular unemployment benefits (i.e. mostly higher education graduates who had not yet found a job). Other austerity measures included an across-the-board cut of 10% in the public wage bill and cost-reducing measures in healthcare. In an attempt to reduce the tax burden on wages still further, in 2014 it was decided to have an across-the-board reduction in labour costs (by reducing employers’ social security contributions) and a one-off

60

“skipping” of automatic wage indexation in 2015. At the same time, and under European impetus, the federal and regional governments implemented a Youth Guarantee Plan that focused on integrating (early) school leavers into the labour market. At the time of writing, no evaluation studies have been carried out.

Austerity measures have also been taken at the regional level. For example, in Flanders the Bourgeois government (2014 – 2019) has decided on various measures that will raise the prices paid by families for water and electricity: the free quotas of water and electricity are being abolished, an energy tax (i.e. “Turteltaks”) is being introduced, and the VAT rate on electricity is again being increased to 21% (having been reduced to 6% by the second Peeters administration (2011 – 2014)). Moreover, childcare becomes more expensive: the minimum price is raised from €1.56 to €5 per day, and the child discount (applicable to the second and any subsequent child) is limited to children under the age of 12.

1.4.3. Policy discourse

It should be noted that before, during, and after the crisis, poverty has always been high on the policy agenda in Belgium and its regions. Various “action plans” to eradicate poverty have been adopted, often with a clear focus on child poverty. At the federal level, the most recent plan was adopted in 2012 (and a new plan was recently announced); at the Flemish level in 2015; in Brussels in 2012; and in 2015 for the first time also in Wallonia. In Flanders, moreover, a clear target to halve the child poverty rate by 2020 was adopted in 2010 (Pact 2020 Strategy). However, these plans lack real bite and are focused on supporting local initiatives and small projects (e.g. cheap or free meals, crowdfunding initiatives, parental support programmes), without defining long-term policy initiatives to ensure a structural improvement in the income and labour market position of poor families.

61

In line with other countries, and encouraged by the European Commission and international organisations such as the World Bank and the OECD, the policy focus of anti-poverty programmes has shifted towards early childhood over the past decade (Vandenbroeck, Roets, & Roose, 2012). The result, however, is that policy makers increasingly focus on children aged 0-3, emphasising childcare use and parental support programmes, not necessarily the kind of public measures needed to strengthen the social safety net. A focus on the earliest years to combat child poverty carries with it the risk that structural policies that actually matter will be disregarded.

1.4.4. The way forward

Despite the introduction of tax deductions on low wages (i.e. “werkbonus”) and the uprating of many benefits in the late 2000s before the crisis, at the structural level the net minimum wage, social assistance, and unemployment benefit levels are well below the poverty threshold for families with children (Van Mechelen & Marchal, 2013; Cantillon & Marchal, 2016). Added to that, the policy focus on employment as a fast track to social inclusion (which Belgium shares with other European countries) and the post-crisis austerity measures in the unemployment insurance system mentioned above have made unemployed families with children even more vulnerable to living in poverty. The stricter conditions for entering unemployment insurance that have been imposed on school leavers, for instance, have led to a surge in young people claiming social assistance benefits (POD MI, 2014), while microsimulation studies have demonstrated that the changes in the benefit levels for the longer-term unemployed will lead to a significant surge in the poverty risk among these families (Centrale Raad voor het Bedrijfsleven, 2014).

62

In the previous section, we showed that child poverty in Belgium is concentrated in single parent families on the one hand, and in non-EU families on the other. These families are more often low skilled than other families: the prevalence of lone parenthood is related to social class and educational attainment (Pintelon, Cantillon, Van den Bosch, & Whelan, 2013). Corluy and Verbist (2014) have shown that lower educational levels, larger families, and region they are living in explain, at least in part, the low labour market attachment of non-EU-born immigrants. Corluy (2014) has documented extensively how Belgium has the worst track record of the EU-15 countries when it comes to the labour market integration of non-EU migrants. This means that policies must focus on enabling non-working single parents and non-EU migrants to get a foothold in the labour market.

However, many non-EU families with a migration background and lone parents are connected to the labour market, yet still face above-average poverty risks. This means that strengthening the safety net should not be limited to non-working families. In that respect, child benefits could play an important role. Research has shown that child cash benefits are a potentially powerful weapon to combat child poverty (Van Lancker & Van Mechelen, 2015) and single mother poverty (Maldonado & Nieuwenhuis, 2015; Van Lancker et al., 2015).

Figure 1.12 shows the impact of child cash benefits and other cash transfers on reducing child poverty among VLWI households across European welfare states. Belgium stands out as having only a small impact due to transfers. In absolute terms, all cash transfers taken together reduce child poverty by 21pp, of which child benefits account for 12.5pp.

The Belgian child benefit system is universal in nature, meaning that child benefits are for the most part allocated to families with children, irrespective of the income

63 level of the parents (Van Lancker & Van Mechelen, 2015). Only a marginal share of the child benefit budget is selectively allocated to certain types of low-income families, such as the long-term unemployed and single parents. Microsimulation exercises have shown that redressing the balance between selectivity and universality would yield significant gains in terms of child poverty reduction (Hufkens, Vandelannoote, Van Lancker, & Verbist, 2013), and hence would contribute to strengthening the safety net for vulnerable families with children, even if they are engaged in paid employment.

Figure 1.12. Absolute child poverty reduction by transfers (excluding pensions), VLWI households, income year 2009

80 70 60 50 40 30 20

Child Child povertyreduction(%) 10

0

SI FI IS

IT IE

ES PT SE PL

EE LT FR

BE SK CZ

LV DE AT NL LU

BG RO CY GR

NO UK HU DK

child benefits other transfers

Source: authors’ calculations using EU-SILC (2010).

Belgium has been rather slow to adopt the ‘activation turn’ in social policy, but since 2004 activation policies have been implemented for the unemployed, starting with the youngest age groups. Cockx, Dejemeppe and Van der Linden (2011) and Cockx and Baert (2015) show that the activation turn has been quite successful,

64 though there are limits to what activation can achieve, since much depends on the profile of the unemployed and the availability of jobs. As the majority of low- educated parents have no or only marginal attachment to the labour market, it is crucial to create jobs at the lower end of the labour market. Yet, to be effective in reducing child poverty, these jobs have to be adequately paid. Given the profound transitions in the labour market currently faced by developed economies (Goos, Manning & Salomons, 2009), with declining opportunities for low-skilled persons to attain decent jobs (Salverda, 2016), this will require greater government involvement through subsidies for lower-end jobs, public employment, and tax credits (Atkinson, 2015).

A precondition for successfully integrating families with children into the labour market is to make sure that formal childcare facilities are available, accessible, and affordable. Here, too, progress can be made. In Belgium, childcare has been the responsibility of the regions since the 1980 state reform. Responsibility for monitoring care for the under-3s is entrusted to a public organisation, while the actual provision of services is performed by semi-private organisations. Those childcare centres and childminders that are subsidised by the regional government have income-related parental fees and have to give some priority to vulnerable groups, but given the supply shortage, in practice priority is often given to working parents. It has to be noted here that there is no legal entitlement to a place in childcare. Earlier research has clearly shown that the lack of places in childcare is detrimental for low-skilled and low-income parents. Disadvantaged families are hardly able to plan their childcare needs well in advance, since they often find themselves in precarious and flexible forms of employment, if they are employed at all (Vandenbroeck, De Visscher, Van Nuffel, & Ferla, 2008). As a consequence,

65 childcare use is highly biased in favour of work-rich and higher-income families (Van Lancker & Ghysels, 2012).

1.5. Conclusion

Before the crisis, the Belgian welfare state was plagued by comparatively high levels of child poverty, and by a creeping, yet significant, increase in child poverty. This is related to the relatively high share of jobless households and a decline in the redistributive impact of social policies on poverty rates. Although the impact of the crisis in terms of job loss and GDP contraction was rather modest, and although unemployment affected those with a low level of education less than those with a high level, our analyses show that it has had an impact on the child poverty risk, by increasing the number of children living in VLWI households. This has translated into deteriorating material living conditions among children since the Great Recession.

In Belgium, the majority of poor children grow up in a single parent family and/or a family with a non-EU migration background. Yet, it is important to acknowledge that while poor children are overrepresented in families with low-skilled parents, the majority of the parents of poor children living in single parent and/or families with a non-EU migration background are not low skilled. Moreover, while a high share of children live in households with no or only a marginal attachment to the labour market, and while those children run a particularly high risk of being poor, many poor children are living in households with higher levels of work intensity. Child poverty is not simply a problem of being poorly qualified, and nor is it simply a problem of having work.

66

In policy terms, first it is crucial to integrate workless families with children into the labour market. To achieve this, successful activation, making work pay for the low-skilled, and structural labour market reforms are needed, while childcare services should be more readily available, accessible, and affordable. Second, though expensive, it is not impossible to guarantee a decent minimum income in general, and adequate child benefit packages in particular, provided the policy design is efficient. Child benefit reforms are a potentially powerful weapon in this respect. In general, in Belgium the poverty-reducing capacity of social spending should be reinforced.

Although the Belgian safety net in general is still going strong, with high levels of social spending, the now prevailing policy paradigm, with its focus on human capital investment and employment, has failed to address the creeping vulnerability of children. The crisis has had an amplifying effect on these evolutions, but did not cause them. The main unresolved issues of the Belgian welfare state are the high numbers of jobless households (especially among families with a non-EU migration background and single parents), the extremely high poverty risk of children growing up in these households, and benefits that are inadequate to shield families with children from poverty, whether or not they work. A focus on education and work will not be sufficient to solve these structural problems.

67

PART II

CHILDHOOD DISABILITY

69

CHAPTER 2 NON-TAKE-UP OF THE SUPPLEMENTAL CHILD BENEFIT FOR CHILDREN WITH A DISABILITY IN BELGIUM: A MIXED-METHOD APPROACH

Published as Vinck, J., Lebeer, J., & Van Lancker, W. (2019). Non-take up of the supplemental child benefit for children with a disability in Belgium: A mixed- method approach. Social Policy and Administration, 53(3), 357-384.

Abstract

Families of children with a disability run a greater risk of being poor, and although policies providing poor families with financial benefits should be effective in reducing poverty, the actual effectiveness is often jeopardised by the issue of non- take-up (NTU). Yet, NTU of benefits aimed at children with a disability is for the most part uncharted territory. In this chapter, we fill this gap using a mixed- methods approach to (1) estimate the magnitude and characteristics of NTU in the Belgian “supplemental child benefit” by drawing on a large-scale administrative dataset on childhood disabilities; and (2) explore the determinants of NTU by means of semi-structured interviews with experts and parents. We estimate a NTU rate of at least 10%, a substantial figure given that the benefit is not income-tested. This mainly concerns children with “less visible disabilities” (i.e. autism spectrum disorder, intellectual disorder and psychological disorder) and results from insufficient information provision about the benefit’s existence and eligibility criteria, process costs, for instance the complexity of the procedure, and the way the scale to assess a child’s disability is constructed.

71

2.1. Introduction

Families of children with a disability run a greater risk of being poor, and poor families are more likely to have children with a disability. The overlap between childhood disability and child poverty has been documented extensively (e.g. Emerson & Hatton, 2007). Although policies providing poor families with financial benefits should be effective in reducing poverty, the actual effectiveness is often jeopardised by non-take-up (NTU), the observation that people who are legally entitled to benefits do not receive them (Hernanz, Malherbet, & Pellizzari, 2004). Yet, NTU of benefits for children with a disability is for the most part uncharted territory. In this chapter, we use a mixed-methods approach to (1) estimate the NTU magnitude and characteristics; and (2) explore the NTU determinants, for the “supplemental child benefit” for children with a disability in Belgium.

Recent reviews demonstrate that NTU is a common problem across EU and OECD countries. In particular for means-tested benefits, NTU frequently affects more than half of the eligible population (EUROFOUND, 2015). For example, NTU estimates in social assistance benefits across OECD countries range from 40% to 80% (Hernanz et al., 2004). Recent estimates for social assistance NTU for Belgium are in the same ballpark, ranging from 57% to 76% (Bouckaert & Schokkaert, 2011). For benefits targeted at persons with a disability, estimates are scarce. For the United Kingdom, it was estimated in the 1990s that between 30% and 70% of eligible working-age persons received (part of the) Disability Living Allowance (Craig & Greenslade, 1998). To our knowledge, studies regarding NTU of Belgian disability benefits do not exist and never has anyone attempted to estimate the take-up of benefits targeted at children with a disability. That will be

72 our first contribution to the literature. We exploit the fragmented Belgian policy system to detect NTU.

Policies that are aimed to reduce poverty or designed to alleviate the increased healthcare costs owing to a disability, are missing their very own purpose if they fail to reach those most in need. Yet, in order to improve the effectiveness of such benefit schemes, it is indispensable to unravel the complexities of NTU. In doing so, we draw on van Oorschot’s (1996) dynamic model of benefit receipt as an explanatory framework to explore the determinants of NTU. We regard NTU as an (undesired) outcome of multilevel actor-behaviour, located at three different levels (client, administration, and benefit scheme) involving three different actors (clients, administrators, and policy makers). Moreover, NTU at the client level does not occur at one specific moment in the procedure, but is the product of clients’ experiences while going through three consecutive stages: threshold, trade-off, and application stage. In each stage, the administrators’ behaviour and the design of the benefit scheme might induce NTU as well. That will be our second contribution to the literature: using semi-structured interviews with experts and parents, we get better purchase on the NTU determinants.

2.2. Understanding non-take-up: a dynamic multilevel model of claiming benefits

More than two decades ago, van Oorschot (1996) put forward a powerful critique on the majority of studies at the time that tried to explain NTU by focusing solely on the client level. Many economists, for instance, tended to see NTU as the result of utility-maximising decisions of rational actors (e.g. reviews in Craig, 1991; Currie, 2004). According to such logic, potential clients weigh the benefits and costs associated with claiming, and act accordingly. In a famous example, Moffitt

73

(1983) models individuals’ NTU behaviour as the result of welfare stigma, a social and psychological “cost” associated with benefit claiming. Other costs identified in the literature include information costs, i.e. lack of information or misinformation on benefits, and process costs such as queuing, filling in complex forms, and the uncertainty of the outcome (Van Mechelen & Janssens, 2017, provide an overview). Benefits include the level and duration of the benefit. Only if potential clients regard the benefit level worth the trouble of going through administrative hassle, for instance, they will claim the benefit.

Although such focus on costs and benefits is helpful to shed light on individual behaviour and responsibility in explaining NTU, van Oorschot (and others, see e.g. Craig, 1991) emphasised that claiming is not simply a matter of balancing costs and benefits at one point in time. It is rather a dynamic process in which clients go through consecutive stages where costs can outweigh benefits. Van Oorschot identified three such stages.

In the threshold stage, individuals need to overcome several barriers before they actually claim a benefit. Here, information costs are relevant: people need to be aware of the benefit’s existence, and if so, they have to consider themselves eligible. Once the first stage is crossed, they enter the trade-off stage. In this stage, individuals trade off “claim inhibiting factors” and “claim stimulating factors” (van Oorschot, 1996, p. 16). Here, costs, especially process, social and psychological costs, and benefits come into play again. Is the benefit worthy of going through all the trouble? What will other people think? If benefits outweigh costs, people will claim the benefit and enter the application stage. This stage can result in receiving the benefit, or the application can be rejected. Here too, costs are relevant. If individuals are not aware of, for instance, all the information necessary for successfully applying for a benefit, their claim can be rejected.

74

Moreover, clients can drop out of the process at each stage, and re-enter the process at a later time.

An important part of the explanatory framework is that NTU is regarded as the result of a multilevel process involving different actors: client behaviour is influenced by administrators and by the design of the benefit scheme. At the administrative level, influencing factors include the quality and quantity of information provision, the simplicity of the application procedure, the internal (e.g. stigmatising communication) and external (e.g. collaboration between stakeholders) organisation of the responsible agencies (Van Mechelen & Janssens, 2017; van Oorschot, 1996). At the benefit scheme level, factors include the degree of selectivity, the associated selection criteria, and the discretionary power built into the system (Van Mechelen & Janssens, 2017). The multilevel perspective is important at each stage of the claiming process. For instance, administrators can or cannot provide sufficient information at the threshold stage. At the trade-off stage, stigma might stem from conditions associated with claiming, such as being obliged to do volunteer work or engage in mandatory training programs (Dwyer & Wright, 2014). At the application stage administrators can wrongfully reject applications, or decide for or against granting a benefit depending on their discretionary power. In the subsequent analyses, this dynamic and multilevel explanatory framework will be applied to gain insight into the NTU of the supplemental child benefit for children with a disability in Belgium.

2.3. Policies for children with a disability in Belgium: multiple recognition levels

Belgium is a federacy in which responsibilities for person-related matters are largely regionalised. As a result, the “policy package” targeted at children with a

75 disability is fragmented. Various income supplements, social and fiscal benefits, and in-kind support measures are available at the federal and regional level. Here, we focus on the Belgian region of Flanders.

First, children with higher care needs, including children with a disability, might be entitled to supplemental child benefits at the federal level. This is a top-up of the regular child benefit. To claim said supplement, children need to receive the regular child benefit first, and their disability needs to be assessed. Doctors of the Federal Public Service for Social Security (FPS) assess the severity of the disability and score the child on a 36-point scale for which they make use of standardised criteria. The scale consists of three pillars which gauge the (1) physical and mental consequences of the child’s disability (maximum 6 points), (2) consequences for the child’s participation in daily life (maximum 12 points), and (3) consequences for the family (maximum 18 points). Although the scale includes non-medical criteria, the assessment is called a “medical examination”. The higher the score a child receives on the total scale, the higher the alleged impact on the family’s care burden and the higher the supplemental child benefit will be. The supplement ranges from € 80 up to more than € 500 per month (see Appendix 2.1 for an overview). Of all Belgian children under the age of 21 in 2015, 2.37% are recognised as disabled at the federal level and hence receive supplemental child benefits (FAMIFED, 2016).

Second, if children with a disability want to make use of subsidised care services (e.g. residential, semi-residential or ambulatory care) or apply for additional financial support (a personal assistance budget (PAB), assistive technology for communication or mobility, or for adaptations to the home), a recognition at the

76 regional level is needed13. To acquire this recognition in Flanders, a multidisciplinary team of the Flemish Agency for Persons with a Disability (FAPD) assesses whether children are substantially and long-term limited in their social participation due to their disability. Such team consists of a doctor, a psychologist or pedagogue, and a social worker or social nurse. This is not a medical examination per se, but an assessment of the child’s needs in relation to the care request, taking account of the child’s medical, psychological and social situation. Once the recognition is obtained the child can make use of care services, depending on the availability of places, apply for support measures or financial help to purchase devices. Whether the child receives a PAB also depends on the availability of funding. Currently, the average waiting period for children who are entitled to a PAB is six years.

It should be noted that children with a disability in Flanders can also be enrolled in special or inclusive education, for which yet another recognition is necessary. As a matter of fact, 4% of children in primary education are enrolled in special education, by far the highest percentage in Europe (EASIE, 2017). Obviously this can be an important source of care support for their families as well.

In sum, at both policy levels, different recognition procedures are in place, and in general, they operate separately from each other. Only when children have at least 18 out of 36 points for the supplemental child benefit, an accelerated application procedure at the FAPD is possible. Due to a recent state reform, the regions will gain responsibility for regulating child benefits from 2020 onwards (Béland &

13 Since March 1, 2014, a distinction is made between directly and non-directly accessible care services in Flanders based on the frequency of care use. Only when individuals deplete their directly accessible quantity, they need a recognition of the Flemish Agency for Persons with a Disability (FAPD). In this chapter, we discuss the situation prior to this reform.

77

Lecours, 2018). However, at the time of writing none of the Belgian regions plan to change the supplemental child benefit scheme.

In the next sections, we will exploit these different recognitions to estimate the NTU rate of the supplemental child benefit by means of administrative data.

2.4. Methods and data

We draw on quantitative and qualitative data to estimate NTU of the supplemental child benefit as well as to gain insight into its determinants. For estimating NTU, we use two administrative datasets. First, microdata from the Datawarehouse Labour Market and Social Protection (DWH LM&SP) is linked with FAPD and census data. The DWH LM&SP compiles administrative socioeconomic information from Belgian social security agencies, and personal and household information from the National Register, including place of residence, gender, date and place of birth of all people living in Belgium. We obtained a random sample of 50% of children below 21 years old with a recognised disability at the federal level living in Belgium in 2010 (n = 25,717), including their score on the 36-point scale. To this dataset, information on the recognition at the FAPD and the type of disability as recognised by the FAPD is added. Information on parental education is added from the 2011 Census. Additionally, we obtained a randomly drawn control group of children below 21 years old without a recognised FPS disability from the DWH LM&SP, of equal size (n = 25,057, after removing children having siblings with disabilities).

Because this dataset does not allow to identify children with a disability who are only recognised at the Flemish level, we complement this with an administrative dataset including basic personal and household characteristics of all children with

78 a disability who are recognised by the FAPD only (n = 8,968). The data include the same information on disability type as mentioned above but unfortunately do not include information on parental education or household type. Finally, we also obtained aggregated figures on the number of all children below 21 recognised at the Flemish level, federal level, or both. The data includes neither enrolment in special or inclusive education nor information on private expenditures for unsubsidised care services or support.

In order to estimate NTU in the supplemental child benefit we exploit the differences between the recognitions at the two levels on the basis of the aggregated data. Afterwards, we discuss differences in personal, household and disability characteristics, drawing on the microdata.

These results are complemented by qualitative analyses to gain better understanding of the NTU determinants. Therefore, we conducted semi-structured interviews with nine experts working in different organisations involved in carrying out policies for children with a disability at both the Flemish and federal level (Table 2.1), and with 13 parents of children with autism spectrum disorder (ASD) or behavioural disorders (Table 2.2). The experts were recruited in two ways, either via contacts the authors already had at the institutions or by an internet search for the responsible person within a specific organisation. At least one expert was interviewed from each key organisation. The initial contact was made by email in which we explained the twofold purpose of the interview. First, we wanted to get a clear understanding of the specific role played by the organisation. Second, the respondents were asked to identify potential NTU problems they experienced in their specific setting.

79

Table 2.1. Interviews experts

Number Institution Who When Duration E1 FPS 3 tenured control doctors 3 February 2017 02:43:10 E2 FPS 1 social worker 2 March 2017 01:29:14 E3 Children’s hospital 1 social worker 9 March 2017 01:40:57 E4 Health Insurance 1 head of social service 30 March 2017 01:45:27 Fund E5 FAPD 2 employees of Team Policy 4 April 2017 02:36:14 and Organisation E6 Pupil Guidance 1 employee responsible for 12 May 2017 01:12:27 Center special education schools E7 Center for 1 coordinator 16 October 2017 02:45:24 Developmental Disorders E8 Special education 1 orthopedagogue 27 October 2017 02:50:32 school E9 Center for 2 doctors 30 March 2018 00:59:25 Developmental Disorders

The parents were recruited via multiple channels: Facebook groups, self-help groups, schools, pupil guidance centres, hospitals, services for persons with a disability, and care centres. The first contact was made via a Facebook post or a letter asking about their experience with the supplemental child benefit. In total, the study includes 16 children with ASD or behavioural disorders. Five children receive the supplemental child benefit, two children used to receive it, two children were rejected at their first application, one child’s application is still ongoing at the time of the interview, and the remaining six never applied for it.

80

Table 2.2. Interviews parents

Number Respondent Children with a Recognised When Duration disability supplemental child benefit P1 Mother Boy, 12y, ASD, dyslexia, Not after re- 23 00:59:53 dyscalculia examination February Boy, 10y, nuclear autism Yes 2018 P2 Mother Boy, 15y, ASD, ADHD, Yes 1 March 01:01:13 highly intelligent, 2018 dysorthography, P3 Mother Boy, 15y, ASD Rejected 2 March 00:48:42 Boy, 9y, ASD, highly No 2018 intelligent P4 Mother & Girl, 25y, ASD, ADHD Used to be 3 March 01:18:24 Father recognised 2018 when < 21y Girl, 17y, ASD, ADHD Yes P5 Mother Boy, 16y, ASH, ADHD, Yes 8 March 01:17:14 Gilles de la Tourette 2018 P6 Father Girl, 2y, rare genetic No 9 March 00:45:43 disorder 2018 P7 Mother Boy, 9y, ASD, dyslexia, No 9 March 00:36:23 dyscalculia, ADHD 2018 P8 Mother Boy, 4y, suspected ASD, No 15 00:39:23 obsessive compulsive March disorder, oppositional 2018 defiant disorder, GLUT1 deficiency syndrome P9 Mother & Boy, 8y, suspected ASD, In process 15 00:48:18 Father ADHD March 2018 P10 Mother Girl, 19y, ASD, highly No 17 00:56:17 intelligent March 2018 P11 Mother Boy, 8y, ASD, ADHD Rejected 21 00:41:18 March 2018 P12 Mother Boy, 7y, ASD Yes 23 00:34:43 March 2018 P13 Mother Boy, 17y, ASD, No 24 00:58:30 developmental March coordination disorder, 2018 highly intelligent

81

The interviews were structured with a topic questionnaire (see Appendix 2.2). The respondents and the researcher could give their own input to the interview. The experts received the interview guide approximately one week in advance, parents not. All interviews took place between February 2017 and March 2018. The expert interviews were conducted at the respondent’s office, the parent interviews at their homes. The duration varied from 34 up to 170 minutes. All interviews were recorded, transcribed in a verbatim way and analysed with NVivo. We applied an initial node structure based on the theoretical framework explained in Section 2.2. Subsequently, potential NTU determinants were identified. In what follows, we refer to the respondents by their chronological number (see first columns of Tables 2.1 and 2.2).

2.5. Results

2.5.1. Non-take-up in the supplemental child benefit

Policies for children with a disability are located at both the federal and Flemish level and the data reveal large discrepancies between the two. The first rows of Table 2.3 show that of all Flemish children with a recognised disability at either level, 42% are only recognised at the federal level (and only receive supplemental child benefit), 21% are only recognised at the Flemish level (and only applied for subsidised care services or support), while only 37% are recognised at both.

It is perhaps not surprising that a substantial group of children with a disability receive cash benefits but do not make use of subsidised care support (42%). Parents may prefer to provide home care for their children with a disability, may choose to purchase care with a non-subsidised provider, or their child is at school during the day. It is surprising, however, that 21% of children with a disability are

82 only recognised at the Flemish level, forgoing the supplemental child benefit. Why would parents forgo (sometimes substantial) cash support that is tailored to their child’s disability, if that same disability is recognised at the Flemish level anyway?

One possible reason is that these children have applied but were rightfully rejected at the application stage, at least according to the standardised criteria. Official FPS statistics for 2010 indeed show that 14.16% of valid applications are rejected because the child is awarded too few points to be recognised (FPS, personal communication, January 3, 2018). Another possible reason is that parents started the application but dropped out on the way. This applies to 2.18% of the applications where parents did not show up on the medical examination, chose to revoke their claim or did not send the necessary medical reports (FPS, personal communication, January 3, 2018). Whereas the former percentage can indicate an indirect form of NTU (we discuss the role of the benefit scheme in Section 2.5.3.3), the latter is NTU in its purest form: parents do not take-up the benefit because they struggle with the application process. Correcting the initial figure of 21% for these rejections (deducted from initial figure) and drop-outs (added to initial figure) results in an adjusted lower-bound NTU estimation of 10%, assuming that all children rejected at the federal level are recognised at the Flemish level (see Table 2.3 for calculations and assumptions). Relaxing that condition to the initial overlap of 37% results in an upper-bound estimation of 19%. So, at least one out of ten children with a recognised disability in Flanders do not receive the supplemental child benefit because they did not apply or dropped out during the process. Since NTU of the regular child benefit in Belgium is estimated to be extremely low (0.25-0.49%, FAMIFED, 2017b), and given that the supplemental child benefit is not subjected to any other income test besides the disability recognition at the federal level, this is a rather substantial NTU rate. It is telling

83 that all interviewed experts were surprised by its magnitude. Moreover, it is likely to be an underestimation since only children with formally recognised disabilities are included. Some children have not (yet) undergone medical examination, or families have not yet accepted the disability as being an issue, while other children are enrolled in special education without a formal recognition of their disability although they could qualify.

Table 2.3. Raw and adjusted NTU estimations

Steps Federal Federal + Flemish Total only Flemish only 0. Initial mismatch Recognitions 17,279 15,070 8,781 41,130 (a) (b) (c) (d) % 42% 37% 21% 100% (a) / (d) (b) / (d) (c) / (d) 1. Estimate federal level applications: FGS (+ FAPD) increase recognitions with rejection rate Recognitions 32,349 8,781 41,130 (a) + (b) = (e) (c) (d) Applications: (e)*1,1416 37,685 (f) Rejected 5,336 (f) – (e) = (g) 2. Assume rejected are recognised at Flemish level (2.1) Lower bound estimate: 3,445 all: (c) – (g) NTU % of (d) 8% (h) (2.2) Upper bound estimate: 6,826 initial overlap only (37%): (c) – 0.37*(g) NTU % of (d) 17% (i) 3. Increase with dropout rate (3.1) Lower bound NTU estimate: 10% (h) + 2.18% (3.2) Upper bound NTU estimate: 19% (i) + 2.18% Source: compiled by the authors based on personal communication with FPS (January 3, 2018).

84

2.5.2. Characteristics of children with a disability

Let us now turn to the characteristics of children with recognised disabilities. Table 2.4 shows results for children with a disability, subdivided by recognition level, as well as for the control group of children without any recognised disabilities. Given the combination of two datasets (see Section 2.4), not all information is available for all groups.

In line with previous research, the results show that children with a disability are more likely to (1) be older (e.g. Blackburn et al., 2010); (2) be male (e.g. Emerson & Hatton, 2007); (3) live with single parents (e.g. Clarke & McKay, 2008); (4) have parents with low or medium educational qualifications (e.g. Sebrechts & Breda, 2012); and (5) live together with other household members with a disability (e.g. Blackburn et al., 2010).

Table 2.4. Personal, family and disability characteristics of Flemish children, 2010

Federal Federal + Flemish Non-disabled only Flemish only children Personal characteristics Age 0-5 17% 18% 13% 29% 6-11 39% 39% 30% 27% 12-17 37% 37% 43% 28% 18-20 7% 7% 14% 16% Gender Boys 62% 67% 70% 51% Girls 38% 33% 30% 49% Family characteristics Country of birth parents Belgium 83% 90% 90% 86% EU27 4% 2% 3% 4% Non-EU27 14% 8% 6% 10%

85

Federal Federal + Flemish Non-disabled only Flemish only children Parental education (highest level) Low-skilled 23% 21% No info 15% Medium-skilled 45% 42% No info 36% High-skilled 32% 37% No info 49% Household type Couples with children 80% 78% No info 84% Single parents 19% 21% No info 15% Other 1% 1% No info 1% Other household members with a disability Yes, at least one 17% 19% No info 2% Disability characteristics Severity of disability (points) 1-5 0% 0% No info / 6-10 63% 51% No info / 11-15 23% 27% No info / 16-20 9% 12% No info / 20+ 5% 10% No info / Single disability: type Autism spectrum disorder (ASD) No info 17% 29% / Severe behavioral disorder (SBD) No info 2% 9% / Minor intellectual disability No info 5% 8% / (MID) Other intellectual disability No info 10% 8% / Other psychological disorder No info 1% 7% / Sensory disability No info 4% 3% / Physical disability No info 5% 2% / Suspected retardation No info 0.4% 0.5% / Multiple disabilities: types 2 or more: ASD, SBD, MID No info 5% 3% / ASD, SBD and/or MID with other No info 29% 21% / disabilities 2 or more other disabilities No info 21% 9% / Source: own calculations based on DWH LM& SP (2010), FAPD (2010), and Census (2011). Notes: country of birth: at least one parent born in Belgium/EU27 or two parents born outside EU27.

The scores on the 36-point scale suggest that those with more severe disabilities are more likely to combine cash and care, indicating that children with a disability

86 posing less of a care burden are less likely to apply for care support at the Flemish level. Finally, the share of children with a disability with both parents born outside the EU27 is lower among those who are recognised at the Flemish level only whereas it is higher among those who are recognised at the federal level only, compared to children without a disability. This suggests that parents with a migration background are more likely to apply for cash benefits and less likely to apply for care services or support. Although there is some evidence that non-EU migrants in Belgium are more likely to provide home care for their children (Kil, Neels, Wood, & de Valk, 2017), at this point it remains an open question how the underrepresentation of children with a disability with a migration background at the Flemish level can be explained.

If we shift our focus to the “NTU-group” of interest, i.e. children with a disability recognised at the Flemish level only, and compare them to children with a disability that are recognised at both levels, one observation clearly stands out. There is an overrepresentation among the NTU-group of ASD, intellectual and psychological disorders. Roughly one third of children in the NTU-group have ASD, while another third consists of children with intellectual or psychological disorders. Four experts (5, 6, 7, 9) stated that this was no surprise to them. Expert 7 raised that “these are actually the children for whom we do not actively inform the parents that they might be eligible”, and according to expert 5 “there has always been criticism of the FPS that disabilities such as autism that are not sufficiently visible … are not sufficiently recognised”.

2.5.3. Determinants of non-take-up

In order to receive the supplemental child benefit, parents have to go through nine different steps (Figure 2.1). At each step, NTU can occur. To get a better grasp of

87 its underlying determinants, we interpret the findings from the semi-structured interviews drawing on van Oorschot’s explanatory framework of NTU as a dynamic, multilevel process (see Section 2.2). We focus on the level of the administration and benefit scheme throughout the three, consecutive stages of benefit claiming.

Figure 2.1. Supplemental child benefit application procedure, 2010

1. Request to start the procedure at the child benefit fund

2. FPS sends acknowledgement of receipt to parents

3. FPS sends two questionnaires to parents: part A (psychosocial and family information) and part B (medical information)

4. Parents fill in part A and go to child’s doctor with part B

5. Parents send part A + B to FPS within six weeks from step 3

6. FPS control doctors conduct medical examination

7. FPS notifies parents and child benefit fund about decision

8. Child benefit fund communicates amount to be received to parents

9. The child benefit fund pays supplemental child benefit Source: compiled by the authors.

2.5.3.1. Threshold stage

The first stage can only be crossed if potential clients know about the benefit’s existence and consider themselves eligible. The majority of experts raised concerns that parents are often unaware that their children might be eligible for

88 the supplemental child benefit. In fact, two interviewed parents only found out about its existence through our study (P6, P9). Throughout the interviews, multiple channels through which parents can be notified about the benefit were identified, including interest groups, the parents’ social network, social services of hospitals or health insurance funds, special education schools, pupil guidance centres, diagnostic centres, rehabilitation centres, doctors (both general practitioners (GPs) and specialists), the Flemish Agency for Child and Family Welfare, Facebook groups and the internet. However, there is much diversity in the way they actually inform parents about the benefit. For instance, five experts (E3, E6, E7, E8, E9) provide parents with information regarding the supplemental child benefit only when they believe children will be actually eligible, based on their own experience and knowledge of the benefit criteria. Expert 8 indicates that the special education school provides information to parents upon registration, but only for children with ASD, intellectual or physical disabilities and not when the child has “minor” intellectual disabilities or severe learning difficulties. Moreover, this happens on the school’s own initiative, there is no legal obligation to do so and some schools do not provide any information at all (E1, E6, E8). Expert 7 admits that the doctors in the Centre for Developmental Disorders only provide information to parents of children with “visible” physical disabilities. Both experts argue that they do not want to falsely raise parents’ expectations about receiving supplemental benefits. However, if parents have questions about the benefit, their organisation provides them with detailed information including a warning that they might be ineligible. Even if children are being treated in a rehabilitation centre, their parents are not always informed that they could also apply for supplemental child benefits. Caregivers in these centres are often focused on providing good care while losing sight of the bigger picture (E5). More generally, in none of the surveyed

89 organisations, basic information about the supplemental child benefit is available (e.g. no flyers in waiting rooms).

Differences in the way parents are informed also exist among GPs and specialists. There is no guarantee that the doctor is fully aware of the benefit and its eligibility criteria. Although it is the duty of social workers of the FPS to inform their partners and frontline organisations, they admit that it is very difficult to disseminate information to the almost 9,000 GPs working in Flanders (E2). In addition, two respondents (E1, E7) highlight that some doctors are reluctant to provide information because it entails more work for them without getting any remuneration. Doctors have to fill in part B (see Appendix 2.3, only in Dutch) of a medical information questionnaire, and this is necessary before the assessment by the FPS doctors can take place.

An important issue in the threshold stage is whether parents consider their child as eligible for the benefit. Some respondents (E1, E3, E4, P9, P13) raise concerns about the benefit’s name. Although it is called the “supplemental child benefit for children with a disability or disorder” (i.e. “Toeslag voor kinderen met een handicap of aandoening”), long-term or seriously ill children (e.g. young cancer patients) can also be entitled. In fact, a more accurate name would be something like the “supplemental child benefit for children with higher care needs”. One parent states that this is the main reason why she never applied: “I still feel that we are not going to get it because he is not really disabled. He has autism that does not seem a disability.” (P13). Moreover, the 36-point scale combines a medical with a social view. The medical view is reflected in pillar 1 wherein a disability percentage is assigned to children, the social view is generally reflected in pillars 2 and 3 in which the impact on self-reliance and the family’s care burden is gauged. It is the combination of points on the three pillars that determines

90 eligibility. This is not always clear for the majority of parents, however. Experts indicate that parents sometimes believe that their child is ineligible because it does not have a disability, only to realise they do qualify after they have been informed that the assessment is not purely medical (E2, E3, E4, E8).

2.5.3.2. Trade-off stage

Once potential clients have crossed the threshold stage, they enter the trade-off stage in which they weigh perceived benefits against perceived costs. The perceived benefits include both the level and duration of the benefit. However, both elements are not predefined when parents consider to apply for the supplemental child benefit. The amount varies from € 80 up to more than € 500 per month while the duration of the benefit can go from six months up to the moment the child turns 21. When the benefit expires, children need to be re- examined to extend the benefit. Usually the granted duration is adjusted to the typical school transition ages, meaning that children are re-examined at the ages of three, six, and twelve, but the youngest children often have to be reassessed every two years (E1). Sometimes benefits are granted immediately up to age 21 without any additional examination, for instance in the case of diabetes type 1 (E1).

Almost half the respondents (E6, E8, P3, P6, P7, P8, P9, P10, P11, P12) understand that there should be some kind of follow-up, but wonder why this should be so frequent and why parents must go through the entire administrative procedure again. “A moderate intellectual disability or ASD is not gone after two years, why do they already have to show the same things? The way in which ASD is expressed does not have the same impact at every stage of your life, but every two years is very fast.” (E6). For some parents, this re-examination is puzzling as their child

91 will never “heal” of its disability (P1, P2, P11, P13): “Actually, I have to say the same thing every time, because it does not change. The diagnosis remains the same, it is not that he is cured at one time.” (P2). Other respondents state that frequent reassessment is necessary precisely because it is not purely a medical examination but also about the consequences for the family. And these consequences can change (E2, E3, P4). In any case, for parents it is unclear from the beginning for how long the benefit will be granted and how generous it will be.

Regarding the perceived costs, both process costs and social and psychological costs are relevant. One clear administrative obstruction parents run into when applying is a waiting period. The FPS website documents the average waiting period varying from one to three months depending on the region where you live. This is however a gross underestimation of the actual waiting period (E1, E3, E4). If everything goes smoothly, the minimum time needed is about three months but usually it takes six to eight months (E1, E3). This often leads to frustration amongst parents, particularly among those facing financial difficulties while having to cope with medical expenses (E3, E4).

Parents themselves sometimes contribute to the long waiting period: applications can be incomplete, or parents do not show up for the medical examination (2.18% of the applications, see Section 2.5.1). This is related to the complexity of the procedure. Parents have to gather medical records, meet specific deadlines, and be able to physically go to the assessment by the FPS doctors (E6, E7, P2, P3, P4). Apparently it is a common complaint amongst parents that they feel pushed from pillar to post during the application procedure (E3, E6, E8, P1, P10). Yet, not all respondents buy into the argument that the procedure is complex (E4, P5): “The administration is not too bad. You have the request, the questionnaire, the medical

92 part and then it is actually waiting for the doctor’s examination and waiting for the decision.” (E4).

Moreover, eleven respondents indicate that parents can get help to reduce these process cost (E2, E3, E4, E7, E8, E9, P1, P3, P4, P11, P12). This includes information provision, assistance with filling in the questionnaire (Part A, see Appendix 2.3, only in Dutch), and preparing parents for the assessment (e.g. advice on what to emphasise). Still, it is mentioned multiple times that the administrative language and terminology used in the communication of the FPS is not comprehensible for all parents and regularly needs clarification (E2, E3, E4, P2, P11, P12).

Social and psychological cost such as stigma might prevent parents from applying as well. Some parents are still grappling with the fact that their child has a disability. Having to go through the whole procedure can be traumatising in such cases (E2, E4, E7, P1, P13). Moreover, even if parents have accepted the disability, the fact that they have to go through “another examination” can be frustrating (E4, P4, P10, P13): “so many examinations have already happened, I already have so many reports from specialists, is that not enough?” (E4). Likewise, children have to be present during the medical examination and parents often need to focus on their shortcomings. This is perceived as detrimental for the child’s self-image (P1, P2, P3, P8, P13): “They may know everything about me, I want to be open about it, but I do not want to confront my children with it, they do not have to be there. That impacts on your child's self-image.” (P3).

93

2.5.3.3. Application stage

In the final stage, clients start the application which can subsequently be approved or declined. It also happens that parents drop out at this stage. We focus on problems that could arise in two steps: filing the questionnaires, and the assessment by the FPS doctors (steps 4 and 6 in Figure 2.1).

Parents have to complete a questionnaire concerning psychosocial and family information (part A, see Appendix 2.3). Five parents mention that the questions are open for interpretation (P2, P4, P5, P8, P9). Moreover, it is important that they respond to the questions in an elaborate way as this has repercussions for the number of points their child will be granted:

For example, if you look at toilet, there is no question on whether a child is continent or incontinent, it says “independence in washing and hygiene.” … As a parent you should in fact say “my child of seven years old is actually incontinent and she is wearing diapers”, but if you do not get that question explicitly and you have not thought about it, then that is certainly not written down by the control doctor, while it leads to a quotation and points. (E3)

Besides, parents have to find a doctor willing to fill in the medical questionnaire (part B, see Appendix 2.3). Experts 1 and 2 underline the importance of adding medical and school performance reports to the application to better assess how the disability impacts on the child’s life. Since these reports are not always centralised, for instance with the GP, parents have to gather these reports themselves. As one parent puts it: “The hardest part is that you always have to go to psychologists and psychiatrists for all the reports.” (P2). This can be burdensome for more vulnerable

94 parents who are less able to navigate administrative systems or for parents who do not go see a GP in the first place.

After filing the questionnaires, the child has to be examined by a FPS doctor. For that, parents receive an invitation to go to FPS offices at a proposed date. These offices are located in Belgian/Flemish central cities. Home visits are rarely an option and consequently parents may encounter logistic problems in organising this trip. Moreover, if the proposed date does not suit their agenda, parents have to contact the FPS and motivate why they want to change. In total they get two chances (E1, E4). Some children, however, do not have to come to the medical examination as they are automatically granted points. In 2010, this accounted for 20% of all applications, though this mainly concerns renewal applications and priority cases (E1). First applications and particularly those of young children have to be examined in person.

Control doctors typically approach the child first during the examination and they ask questions directly to them to verify the information provided in the questionnaires. If they are finished with talking to the child, they already have a clear idea about the number of points they are going to grant (E1). Only when parents ask, they can have a moment alone with the control doctor to complete and clarify what their child said during the examination. Yet, almost all parents who went to the medical examination feel unheard by the control doctor (P1, P3, P4, P5, P11). Preparation by parents is key, which may create an imbalance between stronger and more vulnerable families (E3, E4, E8, E9). Furthermore, the examination lasts approximately 20 to 30 minutes, which might be too short to capture all of the child’s developmental delays and behavioural problems (E5, E8, E9, P1, P2, P3, P5, P11, P12). Finally, all parents report some unpleasant experience with the control doctor: he or she did not make enough time, did not

95 seem to know much about their child’s disability, did not take a look at the reports, or was unfriendly or not empathic: “At a certain moment the doctor said 'it is clear now, it will not improve any more'. That was so confronting, so negative that I have cried all the way home.” (P1).

The perception lives that parents need to justify that behavioural disorders are also a disability (P2, P3, P4, P7, P8, P9, P10, P11, P12) and that control doctors more easily recognise a physical or visual disability than a less visible impairment such as ASD (E5, E6). That is what we find in the microdata as well (see Section 2.5.2). This is presumably partly the result of how the 36-point scale is constructed. Despite integrating a medical and social perspective, the emphasis is put on the former. Pillar 1 assigns a disability percentage. The scoring relies on a list of paediatric disorders as well as on the official Belgian scale to determine the degree of disability14. However, the latter is outdated and not adjusted to the specificities of childhood disability nor to intellectual disorders as it was developed right after the Second World War to capture the reduction in earnings capacity of war victims. The list of paediatric disorders does comprise intellectual and psychological disorders, but to assign a disability score for these disabilities much emphasis is put on IQ test results while aspects such as social adaptability are lacking. Only when children with ASD or ADHD have an IQ below 60, they are assigned a disability percentage between 66 and 79%, corresponding to four points on pillar 1 (see Appendix 2.1). However, this is hardly “achieved” by these children (E9). Control doctors state these tests are not sanctifying, for instance when the child has to take medication like Ritalin in case of ADHD or when a child is from a different cultural background (E1). For children with ASD and a

14 For the full decree (in Dutch or French), see http://www.ejustice.just.fgov.be/cgi_loi/change_lg.pl?language=nl&la=N&cn=2006020831&tabl e_name=wet.

96 normal intelligence, then, it is almost necessary to explicitly specify (preferably in an additional medical or school performance report) the limitations they experience due to ASD (E6). In sum, “for ASD or ADHD, assigning a percentage to the disability is really difficult, it is guesswork” (E1).

An additional problem for these “less visible disabilities” is that much weight is attached to the score on pillar 1 in calculating the benefit amount. If the child scores six to eight points in total but less than four points on pillar 1, the benefit amount will be four times less than when at least four points would have been awarded (see Appendix 2.1). Three experts would prefer for the percentage to be dropped from the classification system or at least to receive less weight in defining the benefit amount (E1, E2, E5). Pillars 2 and 3 try to capture the consequences for the child’s self-reliance and for the family, in essence a social perspective on disability. Yet, here too, medical criteria are used to assess this (E9). For example, two out of three subscales of pillar 3 look at the type and frequency of treatment and medical supervision needed for the child, inside or outside their home. Finally, some respondents address that the examination is not tailored to the child’s disability (E9, P1, P3, P4): “They ask a few things to the child but things that they obviously can do, not the things where the problems actually are. Writing and math is all okay, but if you cannot make contact with others, if you cannot help yourself on the playground, that is much worse for a child.” (P4).

2.6. Discussion and conclusion

In this paper we examine the extent and determinants of NTU of the supplemental child benefit for children with a disability in Belgium. We exploit differences in disability recognitions at the federal and Flemish policy level, drawing on a unique and large-scale administrative dataset. We find that at least 10% of children with

97 a recognised disability in Flanders do not receive the supplemental child benefit because they did not apply or dropped out during the process. This is a rather substantial rate, given that the NTU estimate in the regular child benefits is extremely low and that the benefit is not means-tested, only subjected to a disability recognition at the federal level. We find that the disability type is of major importance to understand this NTU rate: two thirds of children missing out on the supplemental child benefit have ASD, intellectual or psychological disorders.

To better understand the underlying NTU determinants, we conduct semi- structured interviews with experts and parents of children with ASD or behavioural disorders, and interpret the findings drawing on van Oorschot’s (1996) dynamic multilevel framework of claiming benefits. The results point out the role of costs and benefits at each stage of the application process. First, NTU results from insufficient information provision about the benefit by frontline organisations and doctors. More generally, it is confusing and difficult for parents since different kinds of support measures are located at different policy levels, all applying their own recognition and application procedures. Second, parents face process costs: they have to gather medical reports, meet specific deadlines and be able to physically go to the assessment by the FPS doctors. There is a minimum waiting period of three months. On top of that, neither the benefit level nor the duration are predefined when parents consider to apply and are hard to estimate beforehand.

Finally, NTU is probably partly the result of how the benefit scale is constructed. Despite integrating a medical and social perspective, the emphasis is still put on the former. Assigning a disability percentage to the less visible disabilities like ASD, ADHD and other mental disorders is not straightforward (Boat & Wu, 2015

98 report similar difficulties). Yet the disability percentage has important repercussions for the amount awarded. Even the parts of the scale meant to gauge the child’s self-reliance and the family’s care burden still strongly reflect a medical perspective.

Our analysis hints at three policy implications. First, more effort needs to be put into providing frontline organisations, doctors and parents with correct information about the benefit’s existence and eligibility criteria. Although automatic benefit entitlement is difficult if not impossible to implement, given its reliance on an assessment of the child’s disability, we believe there is much to gain in terms of proactive information provision: when a child is enrolled in special education, hospitalised for a long time, recognised at the Flemish level, examined by a GP, or diagnosed by a specialist, information on the supplemental child benefit should be provided by default. Proactive information provision by those who diagnose, or automatically starting the procedure in a more extreme version, is also put forward by all but two parents (P6 and P8). Related to that, simply removing the word “disability” out of the benefit’s name and changing it to something like “higher care needs” might circumvent the stigma associated with disability and might be more telling for parents.

Second, a revision of the benefit scale seems warranted. The benefit criteria still puts much emphasis on the medical perspective of the child’s disability which is detrimental for children with less visible disabilities. Validation or reliability studies have never been conducted before. A study using a vignettes set-up could be a next step to pinpoint the role of how the benefit criteria treat these less visible disabilities and whether, and if so how, they are assessed differently by different FPS doctors.

99

Finally, our results point to the need for coherence in the disability policy package. Six parents raise the issue that improved communication between different administrations would help not having to provide the same information multiple times (P1, P2, P6, P8, P11, P12). In the same vein, making a multidisciplinary team responsible for the recognition of the supplemental child benefit, in line with the FAPD procedure, is desirable for all experts and the majority of parents (P1, P2, P5, P6, P7, P11, P13). The recent transfer of responsibility for child benefits from the federal to the Flemish level to be concluded in 2020 provides a unique opportunity to align the recognition procedures for both cash and in-kind support to children with a disability. First steps have been taken in this respect as the current Flemish minister of well-being, health and family recently announced a fusion of the different agencies, to be finalised in 2019.

2.7. Appendix Chapter 2

2.7.1. Appendix 2.1 Supplemental child benefit: pillars, subscales, points and benefit amounts

Table A2.1.1. Supplemental child benefit: pillars, subscales and points, 2018

Pillar Subscale Points 1. Degree of incapacity 0-24% 0 25-49% 1 50-65% 2 66-79% 4 80-100% 6 Total P1 max 6 2. Activity and Learning, education and max 3 participation social integration Communication max 3 Mobility and movement max 3 Self-care max 3 Total P2 max 12

100

Pillar Subscale Points 3. Family burden: highest Follow-up of the treatment at max 3 score doubled home Leaving the home for max 3 medical supervision and treatment Adaptations to way of living max 3 Total P3 max 18 Source: personal communication with FPS (March 30, 2017).

Table A2.1.2. Supplemental child benefit: benefit amounts according to number of points, Belgium, 2018

Total points Points on P1 Benefit amount €/month < 6 ≥ 4 80.75 6-8 < 4 107.55 6-8 ≥ 4 414.28 9-11 < 4 250.97 9-11 ≥ 4 414.28 12-14 n/a 414.28 15-17 n/a 471.07 18-20 n/a 504.71 > 20 n/a 538.36 Source: FAMIFED (2018).

2.7.2. Appendix 2.2 Topic questionnaires

2.7.2.1. Experts

Role of the organisation

- Which children do you help? Age, disability, socioeconomic background - What role do you play in the disability policy landscape in general, and for the supplemental child benefit in particular? - Which information do you provide on the different policies for children with a disability?

101

Supplemental child benefit

- Are parents aware of the benefit’s existence? - If yes: applied/rejected? - If no: why not applied? - Do parents complain about the procedure? If yes, which complaints? - Do you see possible barriers in the application process? - What can be improved in the application process? - Do you think the benefit criteria strike a good balance between a medical and social view on disability? - What can be improved in the legislation/assessment? - Waiting period - Benefit duration - Control doctors as “gatekeepers” - What can be improved in the medical examination?

NTU

- Are you surprised with the magnitude of the mismatch? - Do you have an idea which children make up this missed group? Age, disability, socioeconomic background?

2.7.2.2. Parents

Child information

- Age child - Medical background child - Education and profession parents - Number of children (with a disability) in the household

102

- Disability recognised by FAPD - If yes: use of FAPD care services? Experience application procedure? - Medical point of contact

Supplemental child benefit

- Knowledge of the benefit - If yes: information channel? Applied? - If no: why not applied? - Client level - Lack of information? - How can information provision be improved? - Received help with application? If yes, from whom? - Shared experiences with friends, family, acquaintances? - Did their experience influence you? - Do you compare your child with other children with a disability? - Stigma about your child’s disability? - Does this influence the receipt of the supplemental child benefit? - What can be improved so that you apply for the benefit? - Policy level - Questionnaire clear or help needed? - What feeling did you have when completing this questionnaire? - What do you know about the legislation? - How is your child’s disability judged? - Familiar with the three pillars? - What do you think about the application process? - What can be improved in the legislation/assessment?

103

- Administrative level - Experience control doctor - Waiting period - Limited benefit duration - What can be improved in the medical examination? - General - Experienced/perceived difficulties - Complaints

2.7.3. Appendix 2.3 Questionnaires

Questionnaire for parents on psychosocial and family information (Part A), see pages 105 – 108.

Questionnaire for child’s doctor on medical information (Part B), see pages 109 – 115.

104

105

106

107

108

109

110

111

112

113

114

115

CHAPTER 3 AN INTERSECTIONAL APPROACH TOWARDS PARENTAL EMPLOYMENT IN FAMILIES WITH A CHILD WITH A DISABILITY: THE CASE OF BELGIUM

Published as Vinck, J., & Van Lancker, W. (2020). An intersectional approach towards parental employment in families with a child with a disability: The case of Belgium. Work, Employment and Society, 34(2), 228-261.

Abstract

For parents of children with a disability labour market participation is difficult since these children require care that exceeds typical parental care. At the same time, children with a disability often live in families who belong to social categories that are associated with lower employment probabilities. However, the intersection between disability and social categories is hitherto overlooked in the literature. Drawing on a case study of Belgium, this chapter empirically examines to what extent parental employment is explained by the child’s disability and/or the family’s social disadvantages. For this, unique and large-scale register data is used. The results show that (1) childhood disability overlapped with social disadvantages; (2) childhood disability inhibited parental employment; but (3) the relationship differed by social category: for single parents, parents with low educational qualifications, and parents having multiple children with a disability, disability and social disadvantage reinforced each other.

117

3.1. Introduction

Drawing on a case study of Belgium, this chapter empirically examines how having a child with a disability overlaps with different social disadvantages and how such an intersection relates to the risk of low parental employment.

Over the past decades, welfare states were reoriented to support and encourage paid employment for all. In many countries the recalibration of welfare policies was guided by the idea that income security and social integration are best ensured by paid labour (Burkhauser et al., 2016; Hemerijck, 2017). Consequently, policy reforms in many welfare states were characterised by a stronger emphasis on activation, often accompanied with the curtailment of cash benefits to avoid inactivity traps (Bonoli, 2012). In Belgium, activation was supported by simultaneously fostering job demand (mainly through reducing employers’ social security contributions) and job supply (by lowering low wage employees’ social security contributions and by increased monitoring and sanctioning of the unemployed) (Hemerijck & Marx, 2010). Moreover, activation policies increasingly expanded to cover people who were exempted before, such as single mothers, persons with a disability, and caregivers (Lindsay et al., 2015; Roets et al., 2012). In fact, disability policies were mainstreamed into regular labour market programmes throughout European welfare states (Burkhauser et al., 2016; Hvinden, 2003; Marin et al., 2004).

For families of children with a disability, however, parental employment is prima facie problematic (Cantillon & Van Lancker, 2013). Children with a disability generally require an amount of care that exceeds typical parental care and the time required to provide such care impedes parents’, especially mothers’, engagement in the labour market (Brown & Clark, 2017; Stabile & Allin, 2012). On top of that,

118 they also face higher, direct costs to pay for medical and care needs. As a corollary, families with children with a disability usually have to get by on lower incomes. Given the aforementioned changes in the welfare settlement, it becomes increasingly difficult to sustain a decent living standard on only one or even no income from paid employment. As a result, these families face a high risk of poverty (Shahtahmasebi et al., 2011).

In a context in which families with children with a disability are pressured to work, it is crucial to understand the mechanisms that inhibit the full realisation of their work potential. This is important since the child’s disability might be only half the story. Parents of children with a disability often belong to social categories that might impede their employment opportunities in their own right: they are more likely to be lower educated, single, and disabled.

So, then, can parental employment be explained by the child’s disability, the family’s social disadvantages, or both? That is the central question of this chapter. Using an intersectional approach to capture combinations of multiple social disadvantages (see Zuccotti & O’Reilly, 2019, for a similar approach), the aim of this chapter is to further our understanding of whether childhood disability overlaps with other social disadvantages in Belgium and how this intersection correlates with parental employment. To test this, large-scale register data in which information on children’s disabilities is linked with information on socioeconomic characteristics and employment is used. The dataset also includes a control group of children without a disability. In the empirical analyses, first, the overlap between the child’s disability and the family’s social categories (lower versus higher educated parents, single versus two parents, parents with versus without a migration background, and households with versus without multiple members with a disability) is investigated. Second, the relationship between

119 having a child with a disability and parental employment is examined. Finally, it is tested to what extent the employment gap differs by the aforementioned social categories.

3.2. Theoretical framework and previous research

It is theorised that an intersectional approach is needed to understand the employment gap between families with and without children with a disability. This approach assumes that multiple social disadvantages not only independently affect one’s employment chances but might also reinforce one another (McBride, Hebson, & Holgate, 2015; Mooney, 2016; Tapia & Alberti, 2019). Therefore, their association with employment should be analysed jointly instead of separately.

There is a wealth of theoretical and empirical literature on intersectionality in different research fields (Bilge, 2010; Cho, Crenshaw, & McCall, 2013; Choo & Ferree, 2010; Collins, 2015; Dhamoon, 2011; Hancock, 2007; McCall, 2005; Walby, 2007; Walby, Armstrong, & Strid, 2012; Yuval-Davis, 2006). In this chapter, an inter-categorical approach towards intersectionality is adopted. According to McCall (2005), this approach allows to study the complexities of multiple sources of inequalities through the analysis of “existing analytical categories to document relationships of inequality among social groups and changing configurations of inequality among multiple and conflicting dimensions” (2005, p. 1773). In doing so, all dimensions of the social categories are considered, meaning that both advantages and disadvantages are simultaneously analysed. Such an approach allows to gain analytical purchase on how the category of “childhood disability” overlaps with and relates to other social disadvantages such as single parenthood (versus two-parent households), lower educational qualifications (versus higher educational qualifications), migration

120 background (versus native parents), and the presence of other family members with a disability (parent(s), other child(ren), other adult(s) versus none). As such, a child with a disability living with a low-skilled single mother with a migration background is placed at the intersection of childhood disability with the disadvantage of low parental educational qualifications, single motherhood and parental migration background. The analytical assumption is that the experiences at this intersection will differ from those at other intersections, such as a child with a disability living in a native-born couple family.

It is clear from the literature that children with a disability require additional care that exceeds typical parental care (Owen et al., 2003). Specifically, maternal labour supply is sensitive to the level of caregiving responsibilities. As such, there is a direct employment effect of having a child with a disability since mothers (and less often fathers) will reduce their working hours or leave the labour market altogether to cope with the care burden (Brown & Clark, 2017; Stabile & Allin, 2012). This pattern is found in Australia (Crettenden et al., 2014; Gordon et al., 2007; Zhu, 2016), the United States (DeRigne, 2012; Porterfield, 2002; Powers, 2001, 2003; Wasi et al. 2012), Norway (Hauge et al., 2013), Sweden (Olsson & Hwang, 2006), Taiwan (Chou et al., 2018) and Belgium (Debacker, 2007; Van Landeghem et al., 2007). In contrast, Brekke and Nadim (2016), also for Norway, find that employment probabilities of parents with and without children with a disability are comparable, which they attribute to gender-egalitarian policies. They do find a strong, negative effect on average earnings, however. Additionally, having a child with a disability can affect parental employment indirectly through deteriorated parental health (Brekke et al., 2017; Brekke & Nadim, 2016).

The size of the employment gap between families with and without children with a disability not only depends on the institutional context but also on the household

121 type, type and severity of the disability, and age of the child. Some authors find that the negative employment impact is larger for single than partnered mothers (Powers, 2001, 2003; Wasi et al., 2012) while others find the opposite (Gordon et al., 2007; Porterfield, 2002) or find that it mainly depends on the disability type (Lu & Zuo, 2010). Regarding the latter, DeRigne (2012) presents evidence that the employment impact of having a child with a disability is strongest when children have mental or developmental disabilities whereas Lu and Zuo (2010) and Wasi et al. (2012) find the strongest impact in families of children with physically disabilities. Nonetheless, the single most consistent finding in the literature is that the more severe the child’s disability, the higher the care burden, the more difficult it is to sustain paid employment (Chou et al., 2018; Crettenden et al., 2014; DeRigne, 2012; Gordon et al., 2007; Hauge et al., 2013; Leiter et al., 2004; Lu and Zuo, 2010; Wasi et al., 2012). Only Powers (2003) does not find a substantial difference by the severity of the child’s disability in the US. Finally, for children without a disability, the care burden tends to lessen when a child ages, in particular when a child reaches schooling age. This frequently coincides with parents, mainly mothers, resuming or increasing employment. For parents of children with a disability, however, the opposite may hold since the care burden often increases or does not decrease as much when their child gets older (Crettenden et al., 2014; Gordon et al., 2007; Haveman, van Berkum, Reijnders, & Heller, 1997).

In order to fully understand the employment gap, it is indispensable to investigate the intersection between the child’s disability and other social categories that increase a family’s labour market vulnerability. There is a persistent link between disadvantaged socioeconomic status and poor health (Case, Lubotsky, & Paxon, 2002; Goldman, 2001; Warren, 2009), and children with a disability are

122 overrepresented in households embodying social disadvantages that jeopardise parental employment. Parents of children with a disability are more likely to have lower educational levels (Bauman et al., 2006; OECD 2010; Powers, 2003; Sebrechts & Breda, 2012; Van Landeghem et al., 2007; Wasi et al., 2012), are more frequently single parents (Bauman et al., 2006; Blackburn et al., 2010; Clarke & McKay, 2008; Emerson & Hatton, 2007; Fujiura & Yamaki, 2000; Powers, 2003; Reichman et al., 2008; Risdal & Singer, 2004; Sebrechts & Breda, 2012; Wasi et al., 2012), and more often have a disability themselves (Blackburn et al., 2010; RKW, 2013; Sebrechts & Breda, 2012; Van Landeghem et al., 2007). The results are less straightforward for the parents’ migration background. Some authors find no difference with respect to migration status (Bauman et al., 2006; Emerson & Hatton, 2007; Fujiura & Yamaki, 2000), while others show that children with a migration background are more (Kawa et al., 2016; Wasi et al., 2012) or less likely (Blackburn et al., 2010; Emerson, 2012; Singh & Lin, 2013) to have a disability, depending on the disability type and countries of origin investigated.

These social disadvantages clearly affect one’s employment potential independent of having a child with a disability. It is for instance well established that strong educational discrepancies in work-care arrangements still exists (Debacker, 2008; Konietzka & Kreyenfeld, 2010; Stahl & Schober, 2018), with highly educated mothers being much more likely to work compared with low-skilled mothers. Furthermore, adults with disabilities are largely marginalised from the labour markets in the majority of welfare states (Grammenos, 2013; Vornholt et al., 2018) and often live in work-poor households (Nys et al., 2016). Balancing paid work with childcare is a major challenge for single parents, and strongly related to welfare state policies such as leave policies and childcare (Nieuwenhuis &

123

Maldonado, 2018). Finally, it is well established that families with a migration background have lower probabilities to attain paid employment in many Western welfare states, including Belgium (Corluy, Haemels, Marx, & Verbist, 2015; EUROSTAT, 2011; OECD, 2012).

So, while it is clear that (1) having a child with a disability is associated with reduced parental employment; and (2) children with a disability often live in families who belong to social categories that are associated with lower employment probabilities, there is a lack of research on the link between the two: can parental employment be explained by the child’s disability, the family’s social disadvantages, or both?

3.3. Research questions and hypotheses

This chapter is structured around three research questions. First, how does the category of childhood disability overlap with other social disadvantages (RQ1)? Second, do parents of children with a disability face an employment gap compared to parents of children without a disability (RQ2)? Finally, to what extent is the parental employment gap stronger at the intersection of the child’s disability and the family’s other social disadvantages (RQ3)?

Regarding RQ1, it is expected that having a child with a disability overlaps with having lower educational qualification (H1.1), being a single parent (H1.2), and living with other household members with a disability (H1.3). Previous literature remains inconclusive on the overlap with migration background, therefore, this chapter examines the nature of this relationship (H1.4). For RQ2, the expectation is that parents of children with a disability work less than parents of children without a disability (H2). Finally, with respect to RQ3, the literature has hitherto

124 not applied an intersectional approach. This chapter investigates whether the experience of multiple social disadvantages adds up, reinforces, or offsets each other in relation to parental employment (H3).

3.4. Data, methods and variables

A unique and large-scale register dataset is used to gain insight into the intersection of the child’s disability and the family’s social disadvantages, and to estimate their reciprocal relationship with parental employment. Microdata from the Datawarehouse Labour Market and Social Protection (DWH LM&SP) is linked with the latest Belgian Census. The DWH LM&SP compiles administrative data from Belgian social security agencies as well as information on personal and household characteristics from the National Register. From the 2011 Census, a snapshot of the total Belgian population on January 1, 2011, information on the parental educational level is extracted. The data consists of a random sample of 50% of children below 21 who lived in Belgium and received the supplemental child benefit on 31 December 2010 and an equally large randomly drawn control group of children below 21 who did not receive the supplemental child benefit.

To obtain the supplemental child benefit, doctors of the Federal Public Service for Social Security (FPS) assess the care needs owing to the disability and score the child on a 36-point scale making use of a standardised criteria list. The purpose of these criteria is to gauge the consequence of the child’s disability for the child’s (1) physical and mental health; (2) self-reliance; and (3) family. There is no income test, but the benefit amount is related to the score on the 36-point scale (for more information on the recognition procedure, see Vinck, Lebeer, & Van Lancker, 2019 or Chapter 2 of this thesis). The benefit varies between € 80 and more than € 500 per month. In 2015, 2.37% of children under 21 received the

125 supplemental child benefit (FAMIFED, 2016). Thus, the category of childhood disability that is applied in this chapter, is an administrative recognition of the disability as evaluated by control doctors of the FPS.

After deletion of children having siblings with a disability from the control group, children whose parents’ employment status cannot be determined with certainty15, children who have missing information on either variable included in the analyses, and selecting the youngest child (with a disability) per household, there are 17,481 children with a disability and 17,934 children without a disability in the unweighted sample.

Logistic and linear regressions are applied to answer the research questions. Regarding RQ1, the overlap between the child’s disability and the family’s social disadvantages is tested through a logistic regression. The dependent variable is whether a child receives the supplemental child benefit (yes/no). As independent variables, the child’s sex and age (categorised into 0-3, 4-5, 6-11, 12-17, or 18-20 years old) are included, together with the social categories the child belongs to (single parent household (yes/no); parental educational qualification; parental country of birth; and living together with other family members with a disability (yes/no)). The parents’ educational qualification is operationalised using the International Standard Classification of Education (ISCED), differentiating between low-skilled (ISCED 0-2: lower secondary education or less), medium- skilled (ISCED 3-4: secondary education) and high-skilled parents (ISCED 5-6:

15 The parents’ employment participation builds upon an administrative record indicating in which branch of the social security system one is registered. If parents do not occur in any social security record, they are assigned to the ‘other’ category and assumed not to be working. However, there is no information to assume this and hence the ‘other’ category is rather diverse, including e.g. housewives, rentiers, outbound frontier workers, and international officials and diplomats. Therefore, parents belonging to this ‘other’ group at any point in 2010 are excluded from the analyses.

126 tertiary education). A dominance criterion is applied, meaning that the highest education level of one of the parents is assigned to the household. For country of birth, a distinction is made between Belgium, other European Union countries (EU27), and non-EU27 countries. A “closeness criterion” is applied, meaning that if at least one parent is born in Belgium or in another EU27 country, the household is considered to have a Belgian or EU27 migration background. Only when both parents are born outside the EU27, the household is considered to have a non- EU27 migration background.

With respect to RQ2 and RQ3, three stepwise linear regression models are employed to estimate the employment gap between families with and without children with a disability and to examine its explanatory factors (i.e. the child’s disability and/or the family’s social disadvantages). The dependent variable measures parental employment jointly by an indicator labelled “household work intensity”. It captures the degree to which all working-age household members together (individuals aged 18-59 years, excluding students 18-24 years) participate on the labour market16. It is defined as the ratio between their total number of months worked (expressed in full-time equivalents) and the total number of months that they could, in theory, have worked. The ratio goes from zero to one, with zero indicating that no working-age household member participated in the labour market in that year, whereas a score of one implies that they all worked full-time all year. In case the actual employment status differs from the contractual situation, the household work intensity variable was recoded to zero17. Model 1 includes the severity of the child’s disability as measured by their score on the 36-

16 Descriptive analyses show that 9% and 10% of children with and without a disability respectively live in an extended family (at least one other working-age household member). 17 3% of children with a disability and 2% of those without had parents who were not working during 2010 but had a work intensity larger than zero.

127 point scale, and five control variables: region of residence; age and sex of the child; age of the mother or if the mother is not present, age of the father (centred around the mean); and the number of siblings. Model 2 adds variables on the aforementioned social categories (family type; parental educational qualification; parental country of birth; and presence of other family members with a disability). To depict the presence of other family members with a disability now three dummy variables are included: one for the parents, one for the other children, and one for the other adults. Finally, in Model 3, interaction effects between the severity of the child’s disability and the social categories are introduced to disentangle the disability and social category effects.

A population weight is applied to both samples to represent the full Belgian population of children with and without a disability. Descriptive information on all variables of interest is presented in Appendix 3.1: in 2010, 2% of Belgian children below 21 have a recognised disability. As sensitivity checks, the analyses were performed without the weight (Appendix 3.2), excluding the extended families (Appendix 3.3), and for mothers and fathers separately (Appendix 3.4, this is elaborated upon in the discussion section as well as in Chapter 4 of this thesis) with no impact on the interpretation of the results.

3.5. Results

3.5.1. How does childhood disability overlap with social disadvantages?

The results regarding RQ1 showed that several social disadvantages overlapped with childhood disability in Belgium (Table 3.1). Children with a disability more frequently lived with low- and medium-skilled parents (H1.1), single parents (H1.2), and other household members with a disability (H1.3). Yet, parents born

128 abroad (in EU27 or non-EU27 countries) were significantly less likely to have a child with a recognised disability compared to native parents (H1.4).

Since immigrants in Belgium are more often lower educated than in other countries (Corluy & Verbist, 2014), it is useful to further investigate this (Appendix 3.5). An interaction between parental education and country of birth revealed no significant educational gradient in the likelihood of having a child with a disability for EU27 and non-EU27 parents (Figure A3.5.1). Yet, low- and medium-skilled EU27 parents and low-skilled non-EU27 parents were less likely to have a child with a recognised disability than natives, while there was no significant difference with natives for high-skilled EU27 parents, and medium- and high-skilled non-EU27 parents. Corroborating the contradictory findings in the literature, in Belgium too the association of childhood disability with migration background depends on the region of origin. If anything, parents with a migration background were less likely to have a child with a disability compared with natives if they have low educational qualifications.

Finally, children with a disability were more often older and boys compared to children without a disability which is in line with previous research (Table 3.1, e.g. Blackburn et al., 2010; Emerson & Hatton, 2007; Van Landeghem et al., 2007).

129

Table 3.1. Logistic regression on having a child with a disability, odds ratios, Belgium, 2010

Children <21 Odds ratios Constant 0.003*** (0.000) Social categories Single parent 1.184*** (0.037) Country of birth parents (Belgium ref.) EU27 0.670*** (0.060) Non-EU27 0.848** (0.051) Parental education (High-skilled ref.) Medium-skilled 1.477*** (0.041) Low-skilled 1.703*** (0.067) Other household member(s) with a disability (None ref.) Yes, at least one 5.743*** (0.317) Child characteristics Age (0-3y ref.) 4-5y 2.437*** (0.120) 6-11y 4.300*** (0.159) 12-17y 4.231*** (0.162) 18-20y 1.727*** (0.102) Male 1.645*** (0.043)

Model fit Log pseudolikelihood -160356.71 Pseudo R² 0.0733 Prob > Chi² 0.000 N 35415 Source: own calculations on DWH LM&SP (2010) and Census (2011). Notes: no disability is base outcome. Robust standard errors in parentheses. * p<0.05; ** p<0.01; *** p<0.001. Odds ratio’s < 1 indicate a lower risk to have a disability, odds ratio’s > 1 indicate a higher risk, compared to the reference category.

130

3.5.2. Parental employment gap: childhood disabilities or social disadvantages?

There was a significant negative relationship between the child’s disability and parental employment in each model (H2, Table 3.2). Moreover, higher scores on the administrative severity scale used for the supplemental child benefit put more pressure on parental employment. Model 1 showed that the “disability effect” ranged from a parental employment gap of 1.3 percentage point (pp) for the lowest number of points (6) to a gap of 35.6 pp for the highest severity score observed (32)18.

Table 3.2. Stepwise multivariate regression on household work intensity, Belgium, 2010

Children <21 Model 1 Model 2 Model 3 0.805*** 0.907*** 0.906*** Constant (0.007) (0.006) (0.006) Disability -0.013*** -0.008*** -0.003** Severity of disability (0.000) (0.000) (0.001) Social categories -0.223*** -0.222*** Single parent (0.006) (0.007) Country of birth parents (Belgium ref.) -0.034* -0.034* EU27 (0.015) (0.016) -0.145*** -0.145*** Non-EU27 (0.011) (0.011) Parental education (High-skilled ref.) -0.111*** -0.110*** Medium-skilled (0.004) (0.004) -0.247*** -0.247*** Low-skilled (0.007) (0.007)

18 As children with 1 to 5 points on the severity scale are not observed in the data, this variable is rescaled by subtracting 5 points of each individual’s total. Only 0.13% of Belgian children receiving supplemental child benefits in 2010 have less than 6 points on the severity scale (FAMIFED, 2017a).

131

Children <21 Model 1 Model 2 Model 3 Other HH members with a disability (excluding siblings) Parent(s) -0.267*** -0.267*** (0.015) (0.016) Other adult(s) -0.166*** -0.168*** (0.029) (0.030) Interaction x Severity of disability Single parent x Severity of disability -0.006*** (0.001) Country of birth parents (Belgium ref.) EU27 x Severity of disability -0.001 (0.003) Non-EU27 x Severity of disability 0.001 (0.002) Parental education (High-skilled ref.) Medium-skilled x Severity of disability -0.004*** (0.001) Low-skilled x Severity of disability -0.002* (0.001) Other HH members with a disability Other adult(s) x Severity of disability 0.007 (0.004) Parent(s) x Severity of disability 0.000 (0.002) Sibling(s) x Severity of disability -0.004** -0.003** (0.001) (0.001) Age of the child (0-3y ref.) 4-5y x Severity of disability -0.003* (0.001) 6-11y x Severity of disability -0.003** (0.001) 12-17y x Severity of disability -0.001 (0.001) 18-20y x Severity of disability 0.002 (0.002) Control variables Age of the child (0-3y ref.) 4-5y 0.056*** 0.063*** 0.064*** (0.008) (0.006) (0.006) 6-11y 0.077*** 0.091*** 0.092*** (0.007) (0.006) (0.006) 12-17y 0.099*** 0.119*** 0.119*** (0.008) (0.007) (0.007) 18-20y 0.060*** 0.098*** 0.098*** (0.012) (0.010) (0.011) Male -0.001 0.001 0.001 (0.004) (0.004) (0.004)

132

Children <21 Model 1 Model 2 Model 3 Region of residence (Flanders ref.) Brussels -0.188*** -0.073*** -0.073*** (0.011) (0.009) (0.009) Wallonia -0.123*** -0.073*** -0.073*** (0.005) (0.004) (0.004) Age mother (centred around the mean) -0.006*** -0.004*** -0.004*** (0.001) (0.000) (0.000) Number of siblings (<21) -0.026*** -0.025*** -0.025*** (0.003) (0.002) (0.002)

Model fit Log likelihood -6805.48 -285.30 -280.60 R² 0.0835 0.3658 0.3660 Likelihood-ratio test compared to M-1 13040.37 9.39 Prob > Chi² compared to M-1 0.000 0.586 N 35415 35415 35415 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: robust standard errors in parentheses. * p<0.05, ** p<0.01, *** p<0.001.

When the different social category variables were included in Model 2, the magnitude of the disability effect was reduced by 39%. Single parenthood, having a migration background, being low- or medium-skilled, or the presence of other household members with a disability had a negative impact on a household’s work intensity.

In Model 3, the association between childhood disability and parental employment was tested at the intersection with other social categories by means of interactions between the severity of the child’s disability and these social categories. The results showed that the employment gap experienced by parents of a child with a disability persisted, meaning that the negative association between having a child with a disability and parental employment could not be solely explained by these social disadvantages. The household work intensity was lower in all families of children with a disability. Yet, the significant interaction effects suggest that the association differed by social category. For single parents, parents with low and medium educational qualifications and parents having multiple children with a

133 disability, the risk of low parental employment is reinforced at the intersection (H3). In fact, for these groups, the disability effect was about one to two times as large compared to partnered parents, parents with high educational qualifications and parents without other children with a disability respectively. For parents with a migration background, parents who have a disability themselves or with other adults with a disability in their household, no significant interaction effect was found. These social disadvantages did not reinforce the negative association between childhood disability and parental employment (H3).

Figure 3.1 visualises these results. It shows the combined effect of childhood disability and belonging to a particular social category for children with a disability with a mean severity score (see Appendix 3.6 for –1 and +1 standard deviation from the mean). The figure distinguishes between the effect at the intersection of childhood disability and the dimension of the social category (i.e. disability effect plus interaction effect, cross-hatched part of the bar) and the effect due to belonging to that particular dimension of the social category (i.e. white part of the bar). For instance, single parents with a child with a disability had a lower labour market participation because they are single parents (i.e. social category effect, white part of the bar), but also because the effect of childhood disability is stronger at the intersection (i.e. black and white cross-hatched part of the bar), compared to two-parent families the association between childhood disability and parental employment was twice as large.

To sum up, all parents with a child with a disability were less likely to work compared to parents without children with a disability. This association was stronger for higher severity of the disability. The employment gap was only in part explained by the family’s social disadvantages. Importantly, the disability effect was stronger at the intersection of childhood disability with single parenthood,

134 lower parental educational qualifications, and the presence of other children with a disability in the household. The effect was not stronger, however, at the intersection with migration background, parental disabilities or the presence of other adults with a disability in the household.

Figure 3.1. Marginal effects of having a child with a disability with the mean severity score on household work intensity by social category, for average values of other variables in Model 3 from Table 3.2, Belgium, 2010

0.05

0.00

-0.05

-0.10

-0.15

-0.20

-0.25

-0.30

Impact Impact onhousehold work intensity -0.35

No No No No

BE

Yes Yes Yes Yes

Low

High

EU27

Medium Non-EU27 Single Migration Parental Multiple Disabled Other parent background education disabled parent(s) disabled children adult(s)

Social category effect Disability effect (including interaction) Total

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: the figure presents the marginal effect on household work intensity by social category for families with a child with a disability who has a mean severity score (10.84326 (5.843258 on the recoded scale)).

135

3.6. Discussion and conclusion

The key contribution of this chapter is that it advances an intersectionality approach to the analysis of labour market disadvantage associated with childhood disability. Following McBride et al.’s (2015) call to apply an intersectional lens towards issues of work and employment, in this chapter it was found that (1) childhood disability overlapped with other social categories that are associated with lower employment probabilities such as single parenthood, lower educational qualifications, and the presence of other household members with a disability; (2) childhood disability inhibited parental employment; but (3) the relationship differed by social category. At the intersection of childhood disability with single parenthood, lower levels of parental education or the presence of other children with a disability in the household, the negative association between childhood disability and parental employment was stronger. Yet, at the intersection of childhood disability with migration background, parental disability or disability of other adult household members, no reinforcing effect could be discerned. Thus, not all sources of disadvantage strengthen one another. The intersectionality approach uncovers multiple sources of disadvantage faced by families of children with a disability in today’s labour markets.

From the point of view of parents, having a child with a disability is an employment disadvantage that contributes to the experience of inequalities in its own right as well as at the intersection with other social disadvantages where the relationship can be intensified. These results add to what McCall (2005) has called the “complexity of intersectionality”. Although the point of view of children was not studied in this chapter, the analysis shows that sources of parental disadvantage can be derived from the social categories children belong to.

136

Several limitations of this study should be mentioned. First, the definition of childhood disability is based on an administrative measure tied to the allocation of supplemental child benefits in Belgium. Such recognised disability presumably does not capture all children with a disability. In Vinck et al. (2019 or Chapter 2 of this thesis), the non-take-up rate of the supplemental child benefit is estimated to be at least 10%. However, the results reported in this chapter are in line with previous research on Flemish data (Sebrechts & Breda, 2012; Van Landeghem et al., 2007). As these authors apply a broader disability definition, including also children with increased care needs who are not recognised for the supplemental child benefits, it is safe to assume that the findings presented here can be extended to children with a disability who have not (yet) obtained this administrative recognition as well.

Second, the household work intensity indicator measures the labour market participation of both parents taken together. It corresponds with one of the European lead indicators to evaluate the EU2020 at-risk-of-poverty or social exclusion target (Ward & Özdemir, 2013). However, the indicator does not allow to disentangle which parent bears the burden of the child’s increased care needs. When analyses are performed on the less detailed employment status (yes/no) of mothers and fathers separately, the same patterns more or less hold true, though they are stronger for mothers than they are for fathers (Appendix 3.4). In line with previous research, gender inequalities are intensified among families with children with a disability. Future research ought to look more in detail into the gendered employment effects of having a child with a disability in Belgium.

Third, the cross-sectional data only allows to assess associations. To unpack the causal relationships between childhood disabilities and other social disadvantages, and how their intersection affects parental employment, the exploration of

137 longitudinal data is necessary. This is potentially an important avenue for future research.

Fourth, the use of formal and informal childcare arrangements could not be controlled for in the analyses. Both care use dimensions obviously play an indispensable role in reconciling work and family life. Yet, previous research showed that in the majority of EU countries, children aged 0-2 from disadvantaged backgrounds make less use of formal childcare arrangements than their more advantaged counterparts (Ghysels & Van Lancker, 2011; Pavolini & Van Lancker, 2018; Van Lancker, 2013). For some of these countries, including Belgium, this is also true for informal childcare (Van Lancker, 2013). Moreover, the available research for Flanders indicates that formal and informal childcare among children with a disability only plays a limited role in responding to the increased care burden faced by their parents (Van Landeghem et al., 2007). They usually have little access to these provisions due to their disadvantaged socioeconomic position (Sebrechts & Breda, 2011). This again points to the importance of adopting an intersectionality approach towards investigating these issues.

Related to the previous point, (semi-)residential or ambulatory care arrangements and enrolment in a special or inclusive educational setting (for which recognitions from the Flemish Agency for Persons with a Disability (FAPD) and the Pupil Guidance Centre are needed in Flanders) are important sources of care support for families of children with a disability. Yet, only 47% of Flemish children receiving the supplemental child benefit, make use of any FAPD care provision (Vinck et al., 2019 or Chapter 2). Additionally, even in families who combine the two collective types of care arrangements (special education and (semi-)residential care), nearly half the mothers are not participating in the labour market (Van Landeghem et al., 2007). As a corollary, for families who do not make use of any

138 support measure, combining work with the increased care demands could be even more difficult.

Finally, the observation that low-skilled parents with a migration background are less likely to have a child with a recognised disability might be related to issues of non-take-up (van Oorschot, 1996). These parents perhaps lack knowledge on the benefit’s existence and the recognition procedure, they may fear the stigma attached to disabilities, encounter more difficulties in overcoming administrative hurdles, or the doctors conducting the medical examination might be inclined to ascribe the child’s developmental delay to cultural differences and language delays and hence reject their application (Kawa et al., 2016; Vinck et al., 2019 or Chapter 2). It could also be hypothesised that length of stay is an important omitted variable here, since the issues related to non-take-up might be particularly salient to newly arrived immigrants. Available data do not allow us to address this question here, but it remains an open question for future research.

In a context where disability policies are increasingly mainstreamed into regular labour market programmes and families of children with a disability are pressured to work, the analyses presented in this chapter suggest two policy implications that extend beyond the case of Belgium.

First, as families of children with a disability experience an employment gap, irrespective of their social disadvantages, it is clear that parents of children with a disability face an additional challenge to combine work and care. This suggests that the activation of families with a child with a disability will be difficult to achieve without increased support provided to these families. Allowing parents with a child with a disability to resume, retain or reinforce their labour market participation presupposes that the access to high-quality formal care provisions adapted to the child’s care needs is improved for their parents, so (part of) the care

139 can be outsourced. This could help to reduce the disability gap for all families. Moreover, as families of children with a disability in disadvantaged social categories experience an increased disability gap, these care services should be in particular available for these social groups as well. This means that the aforementioned socioeconomic inequalities in care service use should be addressed.

Second, extra care support alone will not suffice for these families. The general activation of individuals in disadvantaged social categories matters as well. This calls for improving the labour market opportunities for the low-skilled, single parents, parents with a migration background and individuals with a disability. If welfare states succeed in activating parents belonging to these social groups, this will be beneficial for families with and without a child with a disability alike. Complementary to this, parents of children with a disability should be provided with greater flexibility in their jobs (e.g. being able to organise their own time) which will offer them more possibilities to combine work and care (e.g. Crettenden et al. 2014). This is probably the most challenging for the jobs in which people belonging to these disadvantaged social categories are employed (Kossek & Lautsch, 2018).

To conclude, in order to properly understand employment patterns of parents with children with a disability, the results presented here suggest that childhood disability is an important social category that should not be investigated in isolation but at the intersection with other social categories that are negatively related to parental employment as well. An inter-categorical intersectional approach is a fruitful way of analysing complex patterns of inequalities in today’s labour market.

140

3.7. Appendix Chapter 3

3.7.1. Appendix 3.1 Descriptive information on dependent, independent and control variables for children with and without a disability in Belgium, 2010

With a Without a Children <21 disability disability Dependent variable Household work intensity Very low work intensity (0-0.2) 19% 10% Low work intensity (0.2-0.45) 5% 3% Medium work intensity (0.45-0.55) 11% 7% High work intensity (0.55-0.85) 23% 21% Very high work intensity (0.85-1) 42% 59% Independent variables Disability prevalence 2% 98% Severity of disability (points) No n/a 1-5 observations 6-10 60% n/a 11-15 24% n/a 16-20 9% n/a 21-25 5% n/a 26-30 1% n/a 31-36 0.1% n/a Household type Two-parent household 70% 78% Single mother household 27% 19% Single father household 4% 3% Country of birth parents At least one parent born in Belgium 91% 91% At least one parent born in other EU27 country 2% 3% Both parents born in non-EU27 countries 7% 6% Parental education (highest level) Low-skilled 24% 15% Medium-skilled 41% 35% High-skilled 35% 50% Other household member(s) with a disability At least one of the siblings (non-exclusive category) 9% n/a At least one of the parents (non-exclusive category) 8% 2% At least one of the other adults (non-exclusive 1% 1% category) None (exclusive category) 83% 97%

141

With a Without a Children <21 disability disability Control variables Age child 0-3y 9% 30% 4-5y 9% 11% 6-11y 40% 28% 12-17y 36% 24% 18-20y 6% 8% Sex child Male 64% 51% Female 36% 49% Region of residence Brussels 6% 8% Flanders 61% 59% Wallonia 33% 33% Age mother (mean) 40.2 39.8 Number of siblings (mean) 1.2 1.1 Source: own calculations on DWH LM & SP (2010) and Census (2011) Note: all proportions differ significantly (p<0.05) between children with and without a disability, except for the share of children living in Wallonia, with at least one parent born in Belgium or in another EU27 country.

3.7.2. Appendix 3.2 Sensitivity check: regression results without applying the population weight

Table A3.2.1. Logistic regressions on having a child with a disability, without population weight, odds ratios, Belgium, 2010

Children <21 Model 1 Model 2 Constant 0.168*** 0.167*** (0.006) (0.006) Social categories Single parent 1.165*** 1.160*** (0.033) (0.033) Country of birth parent (Belgium ref.) EU27 0.728*** 0.832 (0.055) (0.119) Non-EU27 0.994 1.230* (0.048) (0.116)

142

Children <21 Model 1 Model 2 Parental education (High-skilled ref.) Medium-skilled 1.456*** 1.469*** (0.037) (0.039) Low-skilled 1.674*** 1.748*** (0.058) (0.066) Other household member(s) with a disability (None ref.) Yes, at least one 5.819*** 5.795*** (0.298) (0.297) Interaction Country of birth parents x Parental education EU27 x Medium-skilled 0.787 (0.157) EU27 x Low-skilled 0.845 (0.156) Non-EU27 x Medium-skilled 0.871 (0.113) Non-EU27 x Low-skilled 0.681** (0.081) Child characteristics Age (0-3y ref.) 4-5y 2.391*** 2.392*** (0.111) (0.111) 6-11y 4.158*** 4.156*** (0.145) (0.145) 12-17y 4.314*** 4.309*** (0.155) (0.155) 18-20y 1.926*** 1.919*** (0.101) (0.100) Male 1.605*** 1.604*** (0.038) (0.038)

Model fit Log likelihood -21700.03 -21693.61 Pseudo R² 0.1159 0.1162 Prob > Chi² 0.000 0.000 Likelihood-ratio test compared to M-1 12.83 Prob > Chi² compared to M-1 0.012 N 35415 35415 Source: own calculations on DWH LM&SP (2010) and Census (2011). Notes: no disability is base outcome. Standard errors in parentheses. * p<0.05; ** p<0.01; *** p<0.001. Odds ratio’s smaller than one indicate a lower risk to have a disability, odds ratio’s larger than one indicate a higher risk, compared to the reference category.

143

Table A3.2.2. Stepwise multivariate regression on household work intensity, without population weight, Belgium, 2010

Children <21 Model 1 Model 2 Model 3 Constant 0.832*** 0.940*** 0.936*** (0.005) (0.004) (0.005) Disability Severity of disability -0.008*** -0.005*** -0.004*** (0.000) (0.000) (0.001) Social categories Single parent -0.258*** -0.247*** (0.003) (0.004) Country of birth parents (Belgium ref.) EU27 -0.030** -0.024* (0.009) (0.011) Non-EU27 -0.136*** -0.136*** (0.006) (0.007) Parental education (High-skilled ref.) Medium-skilled -0.128*** -0.120*** (0.003) (0.004) Low-skilled -0.261*** -0.261*** (0.004) (0.005) Other HH members with a disability (excluding siblings) Parent(s) -0.274*** -0.284*** (0.006) (0.009) Other adult(s) -0.136*** -0.164*** (0.014) (0.020) Interaction x Severity of disability Single parent x Severity of disability -0.004*** (0.001) Country of birth parents (Belgium ref.) EU27 x Severity of disability -0.002 (0.002) Non-EU27 x Severity of disability 0.000 (0.001) Parental education (High-skilled ref.) Medium-skilled x Severity of disability -0.003*** (0.001) Low-skilled x Severity of disability -0.001 (0.001) Other HH members with a disability Other adult(s) x Severity of disability 0.006* (0.003) Parent(s) x Severity of disability 0.003 (0.002)

144

Children <21 Model 1 Model 2 Model 3 Sibling(s) x Severity of disability -0.002* -0.002* (0.001) (0.001) Age of the child (0-3y ref.) 4-5y x Severity of disability -0.002 (0.001) 6-11y x Severity of disability -0.000 (0.001) 12-17y x Severity of disability 0.001 (0.001) 18-20y x Severity of disability 0.003 (0.001) Control variables Age child (0-3y ref.) 4-5y 0.024** 0.040*** 0.045*** (0.007) (0.006) (0.006) 6-11y 0.012* 0.057*** 0.058*** (0.005) (0.004) (0.005) 12-17y 0.030*** 0.085*** 0.081*** (0.006) (0.005) (0.005) 18-20y 0.007 0.072*** 0.063*** (0.009) (0.007) (0.008) Male -0.014*** -0.006 -0.006 (0.004) (0.003) (0.003) Region of residence (Flanders ref.) Brussels -0.200*** -0.074*** -0.073*** (0.007) (0.006) (0.006) Wallonia -0.149*** -0.080*** -0.080*** (0.004) (0.003) (0.003) Age mother (centred around the mean) -0.005*** -0.003*** -0.032*** (0.000) (0.000) (0.001) Number of siblings (< 21y) -0.034*** -0.031*** -0.032*** (0.002) (0.001) (0.001)

Model fit Log likelihood -10476.16 -2806.44 -2774.34 R² 0.0899 0.4100 0.4110 Likelihood-ratio test compared to M-1 15339.44 64.19 Prob > Chi² compared to M-1 0.000 0.000 N 35415 35415 35415 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: standard errors in parentheses. * p<0.05, ** p<0.01, *** p<0.001.

145

3.7.3. Appendix 3.3 Sensitivity check: stepwise multivariate regression on work intensity, excluding extended families, Belgium, 2010

Children <21 Model 1 Model 2 Model 3 Constant 0.884*** 0.949*** 0.949*** (0.008) (0.006) (0.006) Disability Severity of disability -0.012*** -0.007*** -0.003** (0.000) (0.000) (0.001) Social categories Single parent -0.231*** -0.230*** (0.007) (0.007) Country of birth parents (Belgium ref.) EU27 -0.033 -0.033 (0.017) (0.017) Non-EU27 -0.142*** -0.142*** (0.012) (0.012) Parental education (High-skilled ref.) Medium-skilled -0.106*** -0.106*** (0.004) (0.004) Low-skilled -0.244*** -0.243*** (0.008) (0.008) Other HH members with a disability (excluding siblings) Parent(s) -0.287*** -0.287*** (0.017) (0.018) Other adult(s) -0.145** -0.146** (0.049) (0.050) Interaction x Severity of disability Single parent x Severity of disability -0.006*** (0.001) Country of birth parents (Belgium ref.) EU27 x Severity of disability -0.001 (0.003) Non-EU27 x Severity of disability 0.001 (0.002) Parental education (High-skilled ref.) Medium-skilled x Severity of disability -0.004*** (0.001) Low-skilled x Severity of disability -0.003* (0.001) Other HH members with a disability Other adult(s) x Severity of disability 0.005 (0.007) Parent(s) x Severity of disability 0.001 (0.003)

146

Children <21 Model 1 Model 2 Model 3 Sibling(s) x Severity of disability -0.004** -0.003* (0.001) (0.001) Age of the child (0-3y ref.) 4-5y x Severity of disability -0.003* (0.001) 6-11y x Severity of disability -0.002* (0.001) 12-17y x Severity of disability -0.001 (0.001) 18-20y x Severity of disability 0.001 (0.002) Control variables Age child (0-3y ref.) 4-5y 0.029*** 0.052*** 0.052*** (0.008) (0.006) (0.006) 6-11y 0.021** 0.063*** 0.063*** (0.007) (0.006) (0.006) 12-17y 0.005 0.068*** 0.068*** (0.009) (0.008) (0.008) 18-20y -0.071*** 0.024 0.024 (0.015) (0.013) (0.013) Male -0.002 0.001 0.001 (0.004) (0.004) (0.004) Region of residence (Flanders ref.) Brussels -0.194*** -0.075*** -0.075*** (0.011) (0.009) (0.009) Wallonia -0.129*** -0.075*** -0.075*** (0.005) (0.004) (0.004) Age mother (centred around the mean) 0.002** 0.000 0.000 (0.001) (0.000) (0.000) Number of siblings (< 21y) -0.034*** -0.031*** -0.031*** (0.003) (0.002) (0.002)

Model fit Log likelihood -5433.71 468.89 473.32 R² 0.0741 0.3597 0.3599 Likelihood-ratio test compared to M-1 11805.20 8.86 Prob > Chi² compared to M-1 0.000 0.6347 N 31999 31999 31999 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: Robust standard errors are in parentheses. *p<0.05, ** p<0.01, *** p<0.001.

147

3.7.4. Appendix 3.4 Sensitivity check: stepwise multivariate regression for mothers and fathers separately, Belgium, 2010

Table A3.4.1. Stepwise multivariate regression on maternal employment status (yes/no), Belgium, 2010

Children <21 Model 1 Model 2 Model 3 Constant 0.866*** 0.950*** 0.950*** (0.008) (0.008) (0.008) Disability Severity of disability -0.014*** -0.009*** -0.004** (0.001) (0.001) (0.001) Social categories Single parent -0.183*** -0.183*** (0.008) (0.008) Country of birth parents (Belgium ref.) EU27 -0.013 -0.013 (0.020) (0.020) Non-EU27 -0.122*** -0.122*** (0.015) (0.015) Parental education (High-skilled ref.) Medium-skilled -0.106*** -0.105*** (0.005) (0.005) Low-skilled -0.227*** -0.227*** (0.010) (0.010) Other HH members with a disability (excluding siblings) Parent(s) -0.376*** -0.376*** (0.022) (0.023) Other adult(s) -0.111** -0.112** (0.040) (0.042) Interaction x Severity of disability Single parent x Severity of disability -0.002 (0.001) Country of birth parents (Belgium ref.) EU27 x Severity of disability 0.001 (0.004) Non-EU27 x Severity of disability -0.001 (0.002) Parental education (High-skilled ref.) Medium-skilled x Severity of disability -0.006*** (0.001) Low-skilled x Severity of disability -0.007*** (0.002)

148

Children <21 Model 1 Model 2 Model 3 Other HH members with a disability Other adult(s) x Severity of disability 0.003 (0.006) Parent(s) x Severity of disability 0.002 (0.003) Sibling(s) x Severity of disability -0.007*** -0.006*** (0.002) (0.002) Age of the child (0-3y ref.) 4-5y x Severity of disability -0.002 (0.002) 6-11y x Severity of disability -0.002 (0.002) 12-17y x Severity of disability 0.001 (0.002) 18-20y x Severity of disability 0.003 (0.002) Control variables Age child (0-3y ref.) 4-5y 0.076*** 0.083*** 0.083*** (0.009) (0.008) (0.008) 6-11y 0.099*** 0.113*** 0.114*** (0.008) (0.007) (0.007) 12-17y 0.013*** 0.146*** 0.145*** (0.010) (0.009) (0.009) 18-20y 0.105*** 0.142*** 0.142*** (0.015) (0.013) (0.013) Male 0.003 0.005 0.005 (0.005) (0.005) (0.005) Region of residence (Flanders ref.) Brussels -0.208*** -0.110*** -0.110*** (0.013) (0.012) (0.012) Wallonia -0.124*** -0.080*** -0.080*** (0.006) (0.005) (0.005) Age mother (centred around the mean) -0.006*** -0.004*** -0.004*** (0.001) (0.001) (0.001) Number of siblings (< 21y) -0.031*** -0.027*** -0.027*** (0.003) (0.003) (0.003)

Model fit Log likelihood -12186.31 -8788.78 -8785.67 R² 0.0686 0.2364 0.2365 Likelihood-ratio test compared to M-1 6795.05 6.23 Prob > Chi² compared to M-1 0.000 0.8573 N 34210 34210 34210 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: robust standard errors in parentheses. * p<0.05, ** p<0.01, *** p<0.001.

149

Table A3.4.2. Stepwise multivariate regression on paternal employment status (yes/no), Belgium, 2010

Children <21 Model 1 Model 2 Model 3 Constant 0.913*** 0.942*** 0.942*** (0.007) (0.007) (0.007) Disability Severity of disability -0.006*** -0.004*** 0.001 (0.000) (0.000) (0.001) Social categories Single parent -0.091*** -0.089*** (0.015) (0.015) Country of birth parents (Belgium ref.) EU27 -0.003 -0.003 (0.019) (0.019) Non-EU27 -0.100*** -0.100*** (0.015) (0.016) Parental education (High-skilled ref.) Medium-skilled -0.034*** -0.034*** (0.004) (0.004) Low-skilled -0.116*** -0.115*** (0.009) (0.009) Other HH members with a disability (excluding siblings) Parent(s) -0.325*** -0.324*** (0.025) (0.026) Other adult(s) -0.048 -0.048 (0.036) (0.037) Interaction x Severity of disability Single parent x Severity of disability -0.016*** (0.002) Country of birth parents (Belgium ref.) EU27 x Severity of disability -0.001 (0.004) Non-EU27 x Severity of disability -0.001 (0.001) Parental education (High-skilled ref.) Medium-skilled x Severity of disability -0.002** (0.001) Low-skilled x Severity of disability -0.005** (0.002) Other HH members with a disability Other adult(s) x Severity of disability 0.001 (0.006) Parent(s) x Severity of disability -0.001 (0.004)

150

Children <21 Model 1 Model 2 Model 3 Sibling(s) x Severity of disability -0.003 -0.002 (0.001) (0.001) Age of the child (0-3y ref.) 4-5y x Severity of disability -0.001 (0.001) 6-11y x Severity of disability -0.003** (0.001) 12-17y x Severity of disability -0.003** (0.001) 18-20y x Severity of disability -0.004* (0.002) Control variables Age child (0-3y ref.) 4-5y 0.033*** 0.033*** 0.033*** (0.006) (0.006) (0.006) 6-11y 0.051*** 0.051*** 0.051*** (0.006) (0.006) (0.006) 12-17y 0.081*** 0.081*** 0.082*** (0.009) (0.009) (0.009) 18-20y 0.098*** 0.106*** 0.106*** (0.013) (0.013) (0.013) Male -0.001 0.001 0.001 (0.004) (0.003) (0.003) Region of residence (Flanders ref.) Brussels -0.080*** -0.041*** -0.041*** (0.011) (0.010) (0.010) Wallonia -0.040*** -0.030*** -0.030*** (0.004) (0.004) (0.004) Age mother (centred around the mean) -0.008*** -0.007*** -0.007*** (0.001) (0.001) (0.001) Number of siblings (< 21y) 0.005 0.008** 0.008** (0.003) (0.002) (0.002)

Model fit Log likelihood 2675.39 4248.09 4253.34 R² 0.0576 0.1599 0.1602 Likelihood-ratio test compared to M-1 3145.40 10.51 Prob > Chi² compared to M-1 0.000 0.4850 N 27379 27379 27379 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: robust standard errors in parentheses. * p<0.05, ** p<0.01, *** p<0.001.

151

3.7.5. Appendix 3.5 Further investigation of migration background

Table A3.5.1. Logistic regressions on having a child with a disability, odds ratios, Belgium, 2010

Children <21 Model 1 (same as Table 3.1) Model 2 Constant 0.003*** 0.003*** (0.000) (0.000) Social categories Single parent 1.184*** 1.177*** (0.037) (0.037) Country of birth parent (Belgium ref.) EU27 0.670*** 0.821 (0.060) (0.126) Non-EU27 0.848** 1.117 (0.051) (0.122) Parental education (High-skilled ref.) Medium-skilled 1.477*** 1.491*** (0.041) (0.043) Low-skilled 1.703*** 1.784*** (0.067) (0.075) Other HH member(s) with a disability (None ref.) Yes, at least one 5.743*** 5.722*** (0.317) (0.316) Interaction Country of birth parents x Parental education EU27 x Medium-skilled 0.788 (0.176) EU27 x Low-skilled 0.751 (0.155) Non-EU27 x Medium-skilled 0.835 (0.128) Non-EU27 x Low-skilled 0.643** (0.090) Child characteristics Age (0-3y ref.) 4-5y 2.437*** 2.443*** (0.120) (0.120) 6-11y 4.300*** 4.305*** (0.159) (0.159) 12-17y 4.231*** 4.224*** (0.162) (0.162) 18-20y 1.727*** 1.718*** (0.102) (0.101)

152

Children <21 Model 1 (same as Table 3.1) Model 2 Male 1.645*** 1.647*** (0.043) (0.044)

Model fit Log pseudolikelihood -160356.71 -160320.20 Pseudo R² 0.0733 0.0735 Prob > Chi² 0.000 0.000 Likelihood-ratio test compared to M-1 73.02 Prob > Chi² compared to M-1 0.000 N 35415 35415 Source: own calculations on DWH LM&SP (2010) and Census (2011). Notes: no disability is base outcome. Robust standard errors in parentheses. * p<0.05; ** p<0.01; *** p<0.001. Odds ratio’s < 1 indicate a lower risk to have a disability, odds ratio’s > 1 indicate a higher risk, compared to the reference category.

Figure A3.5.1. Marginal effects of country of birth parents and parental education on having a child with a disability, Model 2, Belgium, 2010

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: marginal effects from Model 2, shown at reference values for age and gender of the child, household type, and whether or not there are other household members with a disability.

153

3.7.6. Appendix 3.6 Social category and disability effect for children with -1 standard deviation and +1 standard deviation from the mean severity score

Figure A3.6.1. Marginal effects of having a child with a disability with -1 SD severity score on household work intensity by social category, for average values of other variables in Model 3 from Table 3.2, Belgium, 2010

0.05

0.00

-0.05

-0.10

-0.15

-0.20

-0.25

Impact Impact onhousehold work intensity -0.30

No No No No

BE

Yes Yes Yes Yes

Low

High

EU27

Medium Non-EU27 Single Migration Parental Multiple Disabled Other parent background education disabled parent(s) disabled children adult(s)

Social category effect Disability effect (including interaction) Total

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: the figure presents the marginal effect on household work intensity by social category for families with a child with a disability who has a -1 standard deviation from the mean severity score (5.911886 (0.911884 on the recoded scale)).

154

Figure A3.6.2. Marginal effects of having a child with a disability with +1 SD severity score on household work intensity by social category, for average values of other variables in Model 3 from Table 3.2, Belgium, 2010

0.05 0.00 -0.05 -0.10 -0.15 -0.20 -0.25 -0.30 -0.35

Impact Impact onhousehold work intensity -0.40

No No No No

BE

Yes Yes Yes Yes

Low

High

EU27

Medium Non-EU27 Single Migration Parental Multiple Disabled Other parent background education disabled parent(s) disabled children adult(s)

Social category effect Disability effect (including interaction) Total

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: the figure presents the marginal effect on household work intensity by social category for families with a child with a disability who has a +1 standard deviation from the mean severity score (15.774634 (10.774632 on the recoded scale)).

155

CHAPTER 4 GENDER AND EDUCATION INEQUALITIES IN PARENTAL EMPLOYMENT AND EARNINGS WHEN HAVING A CHILD WITH INCREASED CARE NEEDS: BELGIUM VERSUS NORWAY

Forthcoming as Vinck, J., & Brekke, I. (forthcoming). Gender and education inequalities in parental employment and earnings when having a child with increased care needs: Belgium versus Norway. Journal of European Social Policy, 00(0), 1-14.

Abstract

Caring for children with increased care needs can be demanding and the time required to provide such care hampers parents’ employment participation. Especially, mothers and lower educated parents are affected by the increased care burden and reduce or stop their employment participation. So far, the literature lacks studies investigating the employment impact in a comparative perspective. We fill this gap by comparing Belgium and Norway. We use comparable administrative datasets, identifying children with increased care needs as those receiving a cash benefit designed to financially compensate for the extra private care. The results confirm that gender and education inequalities exist in both countries. Moreover, we find that the negative care burden gap in employment depends on the country of residence, with significantly larger inequalities in Belgium. Our analyses suggest that increased support on multiple fronts is needed for these families.

157

4.1. Introduction

In this chapter, we contrast parental employment and labour earnings between families of children with and without increased care needs. First, we investigate how the employment and wage gaps differ between mothers and fathers. Second, we examine how these gaps vary according to the parents’ educational level. Third, to add to the existing research, we explore whether the employment and wage gaps of parents caring for children with increased care needs differ between welfare states, comparing Belgium and Norway.

Over recent decades, welfare states have increasingly embraced a political commitment to full employment. Nowadays, policy making is dominated by the social investment perspective in Europe, Australia, Canada, and in some less developed welfare states of Asia and Latin America. In addition to investment in human capital from early childhood onwards, social investment places individual responsibility and social inclusion through labour market participation at the forefront (Hemerijck, 2017). Working-age adults are expected to participate in gainful employment and work-facilitating family policies, such as childcare and parental leave, are pushed forward to accomplish this. The European Commission (2013) and the OECD (2006) have adopted the social investment perspective on policy making and emphasised the importance of activation to achieve economic growth and combat poverty and social exclusion.

For the two countries under study, different approaches to activation are taken. In Belgium, activation measures mainly focus on stimulating job demand (by reducing employer’s social security contributions), which, to some extent, are matched by job supply measures (e.g. cutting down the low wage employee’s social insurance contributions and intensified monitoring and sanctioning of the

158 unemployed; Hemerijck & Marx, 2010). In Norway, job supply measures are the core of the activation strategy, mainly by tightening eligibility criteria for welfare benefits and strengthening obligations to participate in activation and training programmes (Halvorsen & Jensen, 2004). Additionally, people who were spared from activation policies before (e.g. single mothers, people with disabilities and people giving care), are nowadays increasingly included throughout European welfare states (Burkhauser et al., 2016; Good Gingrich, 2008; Lindsay et al., 2015; Roets et al., 2012).

In families with children with increased care needs, employment participation is challenging for the parents (Cantillon & Van Lancker, 2013). These children usually require more care, and the time required to provide such care hampers the parents’ employment participation. Previous research has highlighted that gender and education inequalities in this employment impact exist. Especially, mothers are affected by the increased care burden as they, rather than fathers, reduce working hours or retract completely from the labour market (Brown & Clark, 2017; Stabile & Allin, 2012). In fact, gender inequalities in the work-care division are more apparent in families with children with increased care needs than in families with children without increased care needs. Moreover, the effect of having children with increased care needs on parental employment seems to be stronger among less educated parents, signalling the existence of education inequalities (DeRigne & Porterfield, 2017; Lu & Zuo, 2010; Vinck & Van Lancker, 2020 or Chapter 3 of this thesis; Wasi et al., 2012). On top of these indirect costs, parents also face direct costs related to the child’s medical and care needs which impose an additional burden on the household budget (Stabile & Allin, 2012). These direct costs depend on the welfare state settlement, the severity of the increased care needs, the child’s age and household composition (Mitra,

159

Palmer, Kim, Mont, & Groce, 2017). Together, the direct and indirect costs force these families to make ends meet with lower incomes (Larkins et al., 2013). Yet, their poverty risk is also strongly tied to processes of social stratification (Shahtahmasebi et al., 2011): parents have on average lower educational levels; a higher risk of divorce; and are more likely to be disabled themselves (e.g. Blackburn et al., 2010; Sebrechts & Breda, 2012).

The literature on how having children with increased care needs is related to parental employment is short of comparative studies, however. We contribute to the existing research by investigating how mothers and fathers with various educational levels cope differently with the increased care burden in Belgium versus Norway. We use comparable administrative datasets defining children with increased care needs as children who receive a cash benefit that partially compensates the extra care needs they impose on their environment. Comparing Belgium and Norway is interesting as they represent two different welfare regimes. The Norwegian work-family policies promote a dual earner-dual carer family model for all, while in Belgium, more traditional family support policies are combined with a weaker form of dual earner policies which are more socially unequally distributed than in Norway (Ghysels & Van Lancker, 2011; Korpi, 2000; Korpi, Ferrarini, & Englund, 2013). There is a strong connection between these welfare state’s provisions and labour market participation. Therefore, we expect a stronger and more unequal care burden gap in Belgium than in Norway.

4.2. Theoretical framework, previous research and hypotheses

Although gender inequalities in paid employment have substantially decreased in western countries over the last 50 years, mothers still tend to reduce their paid work upon parenthood, even in welfare states with elaborated dual earner policies

160

(Uunk, Kalmijn, & Muffels, 2005). This indicates that gender inequalities in the division of care and work still exist. Especially, when children have increased care needs, mothers are likely to reduce or stop their employment participation (Brown & Clark, 2017; Stabile & Allin, 2012). This pattern is found in Australia (Crettenden et al., 2014; Gordon et al., 2007; Zhu, 2016), Belgium (Debacker, 2007; Van Landeghem et al., 2007), Norway (Brekke & Nadim, 2016; Hauge et al., 2013), Sweden (Olsson & Hwang, 2006), Taiwan (Chou et al., 2018) and the United States (DeRigne & Porterfield, 2010, 2017; Porterfield, 2002; Powers, 2001, 2003; Wasi et al., 2012). This gendered division in paid work can be explained from different angles.

According to the specialisation theory (Becker, 1991), the division of paid and unpaid work is a rational contract between the partners motivated by a utility maximisation. The partner who earns less, often the woman, is expected to do a larger share of the housework and caring tasks, while the partner who earns more, often the man, will specialise in paid employment. According to this perspective, the expectation is that caring for children with increased care needs will mainly be negatively associated to maternal employment and less so to paternal employment.

The gendered work-care division can also be explained from a gender role perspective. The question of how to balance work and parenthood is tied to people’s identities as moral beings and their understanding of “the proper thing to do” in given circumstances (Finch, 1989). It invokes notions of what a good mother or father is, what is best for the children, and what makes for a meaningful life. Gender roles expectations held by others are important in this context. Although women have massively entered into paid employment and men have increasingly taken on household chores and childcare duties, the behaviour typically associated with being a “good mother” still differs from being a “good

161 father”: it is generally expected from mothers to have main caregiving responsibility, while fathers have the main breadwinning responsibility (Duncan, Edwards, Reynolds, & Alldred, 2003). In other words, traditional views on gender roles persist. On this background, we again expect that having children with increased care needs will be negatively related with maternal employment and less to paternal employment.

H1: The negative care burden gap is stronger for mothers than for fathers

Previous research has shown that several factors at the household, organisational and welfare state level influence the employment participation among parents of children with increased care needs. At the household level, the household type, age, number of children, severity and type of increased care needs are found to be important factors in this context, though the results are generally inconclusive (Brown & Clark, 2017; Stabile & Allin, 2012). Only regarding the severity of the child’s increased care needs, previous research consistently reports a positive relationship (except Powers, 2003): the more severe the child’s increased care needs, the more challenging it will be to work for the parents (Chou et al., 2018; Crettenden et al., 2014; DeRigne, 2012; Gordon et al., 2007; Hauge et al., 2013; Leiter et al., 2004; Lu & Zuo, 2010; Vinck & Van Lancker, 2020 or Chapter 3; Wasi et al., 2012). Moreover, organisational level factors such as supervisory support and workplace flexibility as well as welfare states’ policy measures like good quality, available and affordable childcare and paid parental leave, are also essential in understanding the parental employment impact (Brown & Clark, 2017).

Some studies also look into the mitigating role of parents’ educational qualifications on the care burden effect. The results generally show that the effect on parental employment is stronger among less educated parents (DeRigne &

162

Porterfield, 2017; Lu & Zuo, 2010; Vinck & Van Lancker, 2020 or Chapter 3; Wasi et al., 2012), only Leiter et al. (2004) report the opposite. According to human capital theory (Becker, 1985), individuals who invest in their education and training anticipate a return on investment in terms of higher future pay. Hence, parents with high educational qualifications have higher opportunity costs of staying at home. This means that highly educated parents of children with increased care needs have a stronger attachment to the labour market and thus will withdraw to a lesser degree than lower educated parents. Moreover, higher educated individuals hold other types of jobs. They have more choice in how they control their tasks and working time making it easier to combine work and care. On this basis, we suppose that the adverse employment gap of having children with increased care needs will be stronger for lower than for higher skilled parents.

H2: The negative care burden gap is stronger for lower skilled parents

The existing literature remains short of comparative studies on the parental employment impact of having children with increased care needs, however. Yet, one could expect that these patterns differ between welfare states as the level and type of welfare state support influence the parental labour market attachment (Gornick & Meyers, 2003). Welfare states have different histories, normative gender roles expectations, and policy measures that contribute to this employment obligation. In the Nordic welfare states, here represented by Norway, both full employment and gender equality have historically been high on the political agenda (Esping-Andersen, 1990). From the beginning, especially Sweden and Norway, incorporated activation and work-facilitating policy measures into their income maintenance systems to ensure high labour market participation by both men and women (Kautto, Fritzell, Hvinden, Kvist, & Uusitalo, 2001). Norway supports the dual earner-dual carer household that encourages the sharing of care

163 and paid work obligations between the parents (Korpi, 2000). This is exemplified by the right to and high availability of public childcare for the youngest children (Haug & Storø, 2013) and the extensive and generous parental leave scheme, with a substantial number of weeks reserved for fathers. These policies have led to changing gender role perspectives in Norway: mothers are nowadays supposed to work whereas fathers have to take on part of the daily care work when they have young children (Ellingsæter & Gulbrandsen, 2007). Still, we should be careful attributing the comparatively high employment rates in the Nordic countries solely to the provision of work-facilitating policies. Havnes and Mogstad (2011) show that the large expansion of publicly provided childcare during the 1970s in Norway has not resulted into a higher net employment rate as it mainly replaced informal childcare use.

Belgium represents the conservative-corporatist welfare states. It is characterised by a traditional family support model combined with a weak type of a dual earner model (Korpi, 2000). When the conservative-corporatist countries designed their welfare states after the Second World War, they saw the family as the cornerstone of their income maintenance systems (Esping-Andersen, 1990). A division of labour was envisioned by a male breadwinner-female carer household. Men were expected to fully participate in employment, through which they built up social rights for themselves and for their wives who were responsible to care for the young and the old. Only when the family was not able to provide the aid themselves, the welfare state stepped in. This stands in sharp contrast to the social democratic welfare states of Northern Europe that socialised care for children, the elderly and the disabled from the onset (Esping-Andersen, 1990). Since the mid- 1990s, Belgium has made the turn to an “active” welfare state and later to a “social investment state” which implied a stronger emphasis on activation and human

164 capital investment from early childhood onwards instead of solely focussing on passive income protection (Esping-Andersen et al., 2002; Vandenbroucke, 2013). Today, childcare is largely publicly provided and parents pay an income-related fee, though there remains a lack of availability and the use of the existing places is largely socially stratified (Van Lancker, 2013). The parental leave scheme has similar characteristics to the Norwegian system, though it is less extended in duration and pay. Appendix 4.1 overviews the relevant family policy measures in Belgium and Norway.

As combining paid work and increased care responsibilities may be less challenging in Norway, we expect a stronger negative care burden gap in Belgium than in Norway. Specifically, we suppose that the gender and education inequalities are larger in the former country. Regarding the gender inequalities, the Norwegian welfare state is characterised by a stronger gender equality ideology and stronger women-friendly policies than the Belgian welfare state. Korpi et al. (2013) show that dual earner-dual carer family policies have contributed to higher female employment rates and smaller gender inequalities in employment than in countries where family policies are more traditional as they focus on supporting women's unpaid care work. This result mainly applies to women with low and medium educational qualifications. Hence, we expect that both gender and education inequalities are larger in Belgium.

H3.1: The negative care burden gap is more unequal in terms of gender in Belgium than in Norway

H3.2: The negative care burden gap is more unequal in terms of education in Belgium than in Norway

165

4.3. Data, variables and methods

Hitherto, comparative studies on the parental employment gap between families of children with and without increased care needs are scarce due to the lack of sufficient, reliable and comparable data. In fact, to our knowledge no such studies exist. We draw on comparable administrative datasets to investigate this. For Belgium, the microdata consists of a cross-sectional random sample of children below 21 from the Datawarehouse Labour Market and Social Protection (DWH LM&SP) on 31 December 2010. The DWH LM&SP compiles administrative data from Belgian social security agencies as well as personal and household information from the National Register. To this microdata, parental education information is added from the 2011 Census, a snapshot of the Belgian population on January 1, 2011. For Norway, the administrative data are obtained from the Medical Birth Registry of Norway (MBRN), containing information on all births in Norway, and is linked to the National Education Database (NUDB) and Historical Event Database (FD-Trygd) of Statistics Norway. The FD-Trygd panel has information on personal and household characteristics along with employment income. The Norwegian sample consist of all children born in Norway between 2000 and 2005 as well as their mothers and fathers. The last observation point we have is 2008.

Both datasets allow us to compare families of children with and without increased care needs. To do so, we define children with increased care needs as children receiving a non-means-tested cash benefit designed to financially compensate for the extra private care. This corresponds to children receiving the “supplemental child benefit” in Belgium and children receiving the “attendance benefit” in Norway (see Table A4.1.2 in Appendix 4.1 for a detailed description of the

166 entitlement criteria). The control groups are children who do not receive these benefits.

In Belgium, to be entitled to the supplemental child benefit, children need to receive the regular child benefit, should be less than 21 years old and their increased care needs must be assessed by a medical doctor of the Federal Public Service for Social Security. These doctors score the child on a 36-point scale for which they make use of standardised criteria. The scale gauges the impact of the child’s increased care needs in terms of (i) the physical and mental consequences (maximum 6 points), (ii) the consequences for the child’s participation in daily life (maximum 12 points), and (iii) the consequences for the family (maximum 18 points). The higher a child scores on the scale, the higher the impact on the family’s care burden and the higher the supplemental child benefit. The supplement ranges from € 80 for the lowest scores up to more than € 500 per month if the child scores at least 18 points (FAMIFED, 2018). Of all Belgian children below 21 in 2015, 2.37% receive the supplemental child benefit (FAMIFED, 2016).

In Norway, children who need long-term private care and supervision due to a medical condition may be entitled to attendance benefits from the Norwegian Labour and Welfare Administration (NLWA). The application form needs to specify the private care arrangements taken to cope with the child’s increased care needs. To assess the eligibility for attendance benefits at different rates, NLWA considers the degree of physical and mental functional impairment, the amount of help for personal care and supervision needed, the need for stimulation, training and physical activity, and to what extent giving care restricts the care provider. The overall workload of the care provider is the determining factor. The benefit is

167 paid at four different rates, ranging from € 128 up to € 770 per month (NLWA, 2018).

To harmonise both datasets, we focus on children born between 2000 and 2005 in Belgium and Norway respectively, living together with two parents to understand which parent bears the burden of the increased care needs. We randomly select one focal child per household in both the treatment and control group. The sample sizes after deleting observations with missing information on one of the variables of our interest (see Appendix 4.2) are n=3,876 children with and n=4,494 without increased care needs in Belgium19, and n=7,680 and n=231,746 in Norway. Information of other household members is added to the sample and a population weight is applied to the Belgian data to represent the full population of children with and without increased care needs. Appendix 4.3 presents descriptive information for both samples: 2.3% of Belgian children and 3.2% of Norwegian children are identified as children with increased care needs in 2010 and 2008 respectively.

We estimate two linear regression models to examine how and in what way parental employment and earnings are related to having children with increased care needs. For that, we contrast families with children with increased care needs to a control group of families with children without increased care needs. To be able to compare the effect sizes across the two countries and to overcome the problem of unobserved heterogeneity, we follow Mood (2010) and estimate a linear probability model on “parental employment” (0/1) in the first model.

19 The employment status builds upon an administrative record indicating in which branch of the Belgian social security system one is registered. If parents do not occur in any social security record, they are assigned to the ‘other’ category (including housewives, rentiers, outbound frontier workers, and international officials and diplomats) and assumed not to be working. In the analyses, parents belonging to this ‘other’ group are excluded.

168

Logistic and probit regressions are estimated as sensitivity checks yielding comparable results20. In the second model, we run an ordinary least squares (OLS) regression on “parental earnings” (i.e. gross yearly employment income, purchasing power parity (PPP)-adjusted, ln transformed) for employees only. These analyses will enable us to shed light on the existence and extent of an employment and wage gap between parents of children with and without increased care needs.

In both models, we are particularly interested in the gender (H1) and education inequalities (H2) of having a young child with increased care needs in a comparative perspective. For that, we include interactions between having a child with increased care needs on the one hand, and the parent’s gender and educational level on the other. We are aware that other intersections might exist (e.g. Vinck & Van Lancker, 2020 or Chapter 3). We control for the parent’s country of birth, age at the child’s birth, the child’s age and gender, number of siblings, age of the youngest child in the household, employment status of the partner, and the region of residence (Appendix 4.2). To answer H3.1 and H3.2, we test the significance of the difference between Belgium and Norway applying a two sample t-test (Appendix 4.4).

4.4. Results

The predicted employment probabilities and gross labour earnings of parents with and without children with increased care needs are presented in Appendix 4.4. Figures 4.1 and 4.2 visualise these results. They combine information on the marginal main effects (coloured parts of the bars) and interaction effects (cross-

20 These models constrain the predicted outcome to fall within the 0-1 range. Results are available upon request.

169 hatched parts of the bars) of (1) having a child with increased care needs, and (2) being a mother, or (3) being low- or medium skilled, using the mean for all other variables in the model.

Figure 4.1. Marginal effects of having a child with increased care needs on parental employment by gender, educational level and country one is living in, for average values of other variables in Table A4.4.1.

0.00

-0.05

-0.10

-0.15

-0.20

-0.25 Predicted employment Predicted employment gap

-0.30

Father Father

Mother Mother

Low-skilled Low-skilled

High-skilled High-skilled

Medium-skilled Medium-skilled BE NO BE NO Gender Education

CICN gap Gender gap Gender x CICN gap Education gap Education x CICN gap Total

Source: Authors’ calculations on DWH LM&SP (2010) and Census (2011) for Belgium, and on MBRN (2000–05), NUDB and FD-Trygd (2008) for Norway. Notes: For the two countries, marginal effects for gender and education inequalities are calculated separately, at means of the other variables in the models. Parents without a child with increased care needs are the reference; 95% confidence intervals for total employment gap are presented by the black lines.

First, compared to parents of children without increased care needs, negative employment and wage gaps exist for parents of children with increased care needs

170 in Belgium and Norway, indicated by the negative diamonds in Figures 4.1 and 4.2. However, this is not true for all parents. Among Belgian fathers, no significant wage gap is found (Figure 4.2), while there is no significant employment gap for high-skilled fathers in the two countries (Table A4.4.1).

Second, the negative care burden gap that is observed for parents of children with increased care needs differs by the parent’s gender, educational level and country of residence.

Regarding the employment gap (Figure 4.1), Belgian mothers of children with increased care needs have a 17 percentage points (pp) lower employment probability compared to fathers, all else being equal. The corresponding number for Norwegian mothers is 13pp. This is because, on the one hand, mothers have lower employment probabilities than fathers in general (black part), and on the other, because these gender inequalities are intensified among mothers of children with increased care needs (black-white cross-hatched part). Hence, we can accept H1: the gap is stronger among mothers than among fathers of children with increased care needs. Moreover, these gender inequalities are significantly larger in Belgium (-4.3pp, Table A4.4.3), both for children in general (-1.7pp) and for children with increased care needs in particular (-2.6pp). Hence, we find support for H3.1 in case of parental employment.

The care burden gap in employment also differs significantly by the parents’ educational level. In both countries, parents who are lower skilled have a larger employment gap compared to high-skilled parents, supporting H2. In Belgium, low-skilled parents of children with increased care needs have a 22pp lower employment probability compared to their high-skilled counterparts. The corresponding number for Belgian medium-skilled parents is 12pp. In Norway, the difference equals 19pp for low-skilled and 5pp for medium-skilled parents.

171

Again, this is the result of lower employment probabilities for lower skilled parents in general (grey part) and of intensified education inequalities among parents of children with increased care needs (grey-white cross-hatched part). Moreover, the education inequalities differ significantly between the two countries (Table A4.4.3). Low-skilled parents of children with increased care needs have a 2.8pp lower employment probability in Belgium (-3.9pp significant difference for children with increased care needs and +1.1 pp insignificant difference for children in general). For medium-skilled parents, the corresponding number is 6.7pp (-3.9pp significant difference for children in general and -2.9pp significant difference for children with increased care needs). Hence, we can accept H3.2 in case of parental employment.

Similar patterns are true for parental earnings (Figure 4.2). In both countries, employed mothers of children with increased care needs have a larger wage gap than employed fathers, all else being equal. Belgian mothers have a 61% lower wage, whereas the gap equals 64% in Norway. Once more, this gives us support for H1. In both countries, this is largely explained by the wage gap observed for mothers in general (52% for Belgium and 58% for Norway, black part), as the wage gap for mothers of children with increased care needs only marginally adds to this (9% for Belgium and 6% for Norway, black-white cross-hatched part). However, this time we cannot accept H3.1 (Table A4.4.3). The gender inequalities in labour earnings for parents in general are in fact significantly smaller in Belgium than in Norway (6pp), whereas for parents of children with increased care needs the gap in Belgium is not significantly different from the Norwegian gap.

172

Figure 4.2. Marginal effects of having a child with increased care needs on parental labour earnings by gender, educational level and country one is living in, for average values of other variables in Table A4.4.2

10% 0% -10% -20% -30% -40% -50% -60%

Predicted gross wagegap -70%

-80%

Father Father

Mother Mother

Low-skilled Low-skilled

High-skilled High-skilled

Medium-skilled Medium-skilled BE NO BE NO Gender Education

CICN gap Gender gap Gender x CICN gap Education gap Education x CICN gap Total

Source: Authors’ calculations on DWH LM&SP (2010) and Census (2011) for Belgium, and on MBRN (2000–05), NUDB and FD-Trygd (2008) for Norway. Notes: For the two countries, marginal effects for gender and education inequalities are calculated separately, at means of the other variables in the models. Parents without a child with increased care needs are the reference; 95% confidence intervals for total gross wage gap are presented by the black lines.

For the education inequalities, only among Norwegian low-skilled parents, intensified inequalities in labour earnings exist when they have children with increased care needs. These parents earn 49% less than their high-skilled counterparts, because low-skilled parents have a wage gap compared to high-

173 skilled parents in general (grey part) and because this gap is intensified when they have children with increased care needs (-7pp, grey-white cross-hatched part). A wage gap also exists for Norwegian medium-skilled (-24%), Belgian medium- skilled (-43%) and Belgian low-skilled parents (-58%), yet only due to the wage gaps these parents have in general (grey part), not because these gaps are intensified for parents of children with increased care needs (grey-white cross- hatched part). Comparing the total education inequalities between Belgium and Norway, we find significantly larger differences in the former country, due to larger education inequalities for lower skilled parents in Belgium in general, not for parents of children with increased care needs in particular (Table A4.4.3). In fact, for low-skilled parents of children with increased care needs, we find significantly smaller education inequalities in Belgium (9pp), closing their wage gap to 10pp difference between the two countries. As we do not find significantly larger education inequalities in the negative care burden gap in Belgium, we cannot accept H3.2 in case of parental earnings.

4.5. Discussion

We should note that our analyses are constrained by some limitations. First, with the available data, we can test correlations between having children with increased care needs and parental employment or labour earnings, not the causal relationship between them. Parents may have unobserved characteristics affecting their employment and labour earnings as well as the likelihood of having children with increased care needs. For Norway, our results are comparable with the longitudinal Norwegian register study of Brekke and Nadim (2016). In that study, a quasi- experimental difference-in-difference design is used to examine the causal impact

174 of having children with increased care needs on parental labour market participation and earnings, strengthening the robustness of our results.

Second, we only consider children with increased care needs if they are administratively recognised and receive a cash benefit. Country differences may therefore arise if the selected children differ between the two countries. However, the eligibility criteria to receive the cash benefits are comparable (Appendix 4.1): both include (1) a (certain) degree of incapacity, (2) the impact of the increased care needs on different facets of the child’s daily life, and (3) how providing care affects the caregiver’s/family’s life. Yet, the definition used in this study does not represent all children with increased care needs. For Belgium, Vinck et al. (2019 or Chapter 2) estimate the non-take-up rate of the supplemental child benefit to be at least 10%, whereas for Norway, Brekke, Evensen and Kaldager Hart (2020) report a 5% non-take up rate of the attendance benefit for children with Down syndrome. In both countries, children with a migration background are less likely to receive the benefit than their native counterparts (Brekke et al., 2020; Vinck & Van Lancker, 2020 or Chapter 3). Given that (1) the entitlement criteria are comparable, (2) both benefits are prone to non-take-up, and (3) children with a migration background are less likely to receive the benefits, it is safe to assume that both benefits capture similar groups of children with increased care needs in the two countries. Moreover, our findings are consistent with previous studies applying a more extensive definition of children with increased care needs (Albertini Früh, Lidén, Gardsjord, Aden, & Kvarme, 2016 for Norway; Sebrechts & Breda, 2012 for Belgium). Therefore, we believe that our results can be extended to children with increased care needs who are not administratively recognised.

175

Third, the Norwegian data only allows to observe a household’s composition at the child’s birth and we assume this situation still holds true in 2008. This could imply that the Norwegian mothers and fathers in our data are actually single parents facing additional challenges of combining work and family life as they are the sole carers. However, Tøssebro and Wendelborg (2017) report a lower separation risk for families caring for children with intellectual and developmental disabilities in Norway than for families with children in general. Hence, we are confident in the reliability of our results, but this issue could be addressed in future research.

Finally, the use of formal and informal care, both general and disability-specific, could not be taken into account. Without a doubt, using these care services is helpful for parents in combining work and care. Future research should look into whether the gender and education inequalities reported here still hold if the children’s care use is controlled for.

Against a background where everyone is expected to fully participate in employment, our analyses allow us to formulate policy implications that can be informative for other welfare states too. As families with children with increased care needs face an additional challenge in combining work and care, our analyses suggest that increased support on multiple fronts is needed, particularly for mothers and low-skilled parents. First, improved access to and use of high-quality care services could allow parents to partly outsource their child’s care and hence increase their employment participation. Yet, reducing the general gender and education inequalities parents are confronted with, will be crucial too. Integrating mothers and lower skilled parents into the labour market will be helpful for families of children with and without increased care needs alike. In this respect,

176

Belgium as well as other welfare states, can learn from the equality promoting employment policies of Norway.

Second, even if care provisions are improved and parents are integrated in the labour market, this will not suffice. We demonstrate that families with children with increased care needs have to get by on lower incomes because of reduced labour earnings. They are probably also confronted with higher direct costs related to the child’s medical and care needs putting an additional burden on the household budget (Mitra et al., 2017). Extra financial support could be provided to these families to (partly) compensate the income loss they experience and, hence, (partly) offset the increased poverty risk they possibly face.

Finally, workplace support could be crucial too. Equipping parents with increased flexibility in their jobs will provide them with more opportunities to combine work and care (Brown & Clark, 2017). This will probably be the most challenging for jobs occupied by people holding lower educational qualifications (Kossek & Lautsch, 2018).

4.6. Conclusion

In this chapter, we investigate how and in what way parental employment and labour earnings differ between families of children with and without increased care needs, comparing Belgium to Norway. We are interested in how these employment and wage gaps vary by the parent’s gender (H1), educational level (H2) and country of residence (H3.1 and H3.2). To our knowledge, this is the first comparative study of its kind. We draw on comparable administrative datasets.

The results show that parents of children with increased care needs work and earn less than parents with typically developing children. Our analyses confirm that

177 gender and education inequalities exist in the employment and wage gap. Moreover, we find that the negative care burden gap differs by the country of residence. The driving force behind these gaps, however, depends on the outcome variable.

For employment participation among parents of children with increased care needs, we find, in both countries, a stronger care burden gap among mothers than among fathers (supporting H1), as well as among lower skilled parents than among high-skilled parents (supporting H2). This is because mothers and lower skilled parents have lower employment probabilities in general, and these inequalities are intensified for parents of children with increased care needs. Additionally, these gender and education inequalities are stronger in Belgium, for parents in general as well as for parents of children with increased care needs in particular (supporting H3.1 and H3.2).

We find comparable results for labour earnings. Again, gender and education inequalities exist in Belgium and Norway. Yet, this time, the wage gaps are largely the result of gender and education inequalities that exist for parents in general. For parents of children with increased care needs, the inequalities are only marginally (for gender) or insignificantly (for education, except low-skilled parents in Norway) intensified. However, this time, the gender and education inequalities are not significantly larger in Belgium. In fact, the gender inequalities are significantly smaller among Belgian parents in general, whereas there is no significant difference for parents of children with increased care needs in particular. The education inequalities, on the other hand, are significantly larger for Belgian parents in general, but not for parents of children with increased care needs. Actually, among the latter, the gap is significantly smaller for low-skilled parents in Belgium.

178

To conclude, in both Belgium and Norway, parents of children with increased care needs are confronted with additional difficulties in employment and earnings, particularly mothers and lower skilled parents. This suggests that the burden of increased care needs falls mostly on mothers and that highly educated parents, even those who have to take on increased care needs, have a stronger attachment to the labour market than lower educated parents. Yet, the institutional context of the country in which parents live matters. When we look at whether parents are employed or not, the gap is smaller in the Norwegian equality promoting welfare state. A long-standing tradition of full employment and an elaborated policy package to make this work seem to pay off.

4.7. Appendix Chapter 4

4.7.1. Appendix 4.1 The family policy packages in Belgium and Norway

Table A4.1.1. Overview selected Belgian and Norwegian policies

Belgium Norway All children Child benefits  National competence before  National competence 2020a  Age 0-17  Age 0-17 and students <25  Equal amount per child  Age and rank supplement  Not income-tested  Not income-tested universal amount  Income-tested supplement for vulnerable groupsb Single parents Single parent supplement Extended child benefit  Income-tested supplement  Not income-tested child within child benefit system benefit for one additional child Transitional benefit  Age 0-8 in general  Benefit period limited to 3 years

179

Belgium Norway  Work requirements when child is one year or older  Income-adjusted

Infant supplement  Age 0-3  Within child benefit system  When receiving extended child benefit and full transitional benefit Maternal, paternal Maternal leave Parental benefit and parental leave  15 weeks: 1-6 weeks prior to  Prior gainful employment birth, 9-14 weeks after birth  Age 0-2 Paternal leave  Benefit period: 49 weeks  10 days, free to choose 100% or 59 weeks 80%, split within 4 months after birth between parents Parental leave  Mothers: 3 weeks prior to  Prior gainful employment birth + 15 weeks after (6  Age 0-12 or 0-21 if child is weeks reserved immediately ≥66% disabled and receives after birth) supplemental child benefit  Fathers: 15 weeks  Temporarily suspend or  16 or 26 weeks to share reduce work  Paternal quota is transferred  Benefit period: 4 months to mothers if sole carer 100%, 8 months 50% or 20 Lump-sum grant months 20%  When not entitled to parental  Part-time employees can benefit only choose 100% option Care benefits Career break Cash-for-care benefit  Temporarily suspend or  Age 1-2 reduce work  Not attending full-time  100%, 50% or 20% government subsidised  Care-related reasons, 51 kindergarten calendar monthsc: (1) caring  Benefit period limited to 11 for children under 8; (2) months providing palliative care; (3) Childcare benefit caring for severely ill family  Single parents only member; (4) caring for  Help to pay for childminding disabled child under 21; (5) when at work providing assistance or care  Age 0-10 in general to severely ill child under 18  Extended if child needs more  Education-related reason, 36 care or if irregular working calendar monthsc: (6) hours (proof needed) following recognised  Income-tested training Leave for medical assistance

180

Belgium Norway  Temporarily suspend or reduce work to assist severely ill family member  Benefit period: 12 months 100%, 24 months 50% or 20%  Single parents with severely ill child under 12: 24 months 100%, 48 months 50% or 20% Palliative care leave  Temporarily suspend or reduce work to provide palliative care to person suffering from an incurable disease  Maximum 3 months per patient Childcare  Regional competence  Municipal competence  Age < 3 for day care and ≥ 3  Incorporated into national for after-school care education system  No formal right to childcare  Formal right to kindergarten  Income-adjusted fee  Age 1-5 for pre-school care  Prioritisation for specific and 6-10 for after-school groups care  Income-adjusted fee  Prioritisation for specific groups Children with increased care needs Cash benefits Supplemental child benefitd Attendance benefitd  Age 0-20  No age limit for rate 1, age  Top-up of regular child 0-17 for rates 2-4 benefit  Not income-tested  Not income-tested  Norwegian Labour and  Federal Public Service for Welfare Administration Social Security recognition (NLWA) recognition needed needed  Severity-adjusted  Severity-adjusted Basic benefit Personal assistance budget  No age limit  To buy personalised care (at  To cover additional expenses home or in institutions) related to medical condition  Flemish Agency for Persons (excluding medication) with a Disability (FAPD)  NLWA recognition needed recognition needed

181

Belgium Norway Financial support  Adjusted to severity of  To buy devices or do expenses adaptations to the house  FAPD recognition needed Care services Integrated childcare Integrated childcare  Integrated into regular  Integrated into regular childcare system childcare system  No prioritisation solely on  Prioritisation of children the basis of increased care with increased care needs needs over other children  Parents have to ask childcare Other care services provider  Municipal competence Other care services  Duty to organise  Regional competence coordination units  FAPD recognition needed  Municipal NLWA  Subsidised care services recognition needed (residential, semi-residential  Support personnel, relief and or ambulatory care) personal assistance Education  Regional competence  Public special education  Since 2015, priority given to schools closed down in 1992 inclusive education  Inclusive education is  Advice needed from Pupil widespread Guidance Centre for needed  In 2014-2015 school year, support measures in only 0.09% of children 6-11 inclusive educational setting is enrolled in special or access to special education (EASIE, 2018) education  In 2014-2015 school year, 4.6% of children 6-11 enrolled in special (EASIE, 2018) Source: compiled by the authors Notes: (a) The regions will gain competences for regulating child benefits from 2020 onwards (Béland & Lecours, 2018). (b) Social assistance recipients, long-term unemployed, long-term sick and single parents. (c) Throughout the employee’s entire career, non-cumulative. (d) See Table A4.1.2 for more information.

Table A4.1.2. Criteria used to assess eligibility for supplemental child benefit (Belgium) and attendance benefit (Norway)

Supplemental child benefit (Belgium) Attendance benefit (Norway) Children below the age of 21 with a disability, Persons with a disability, injury or illness disorder or illness are scored on a 36-point are assessed according to their need for scale based on 3 pillars: supervision and long-term private care

182

Supplemental child benefit (Belgium) Attendance benefit (Norway)  Pillar 1 (max 6 points): physical and  Taken into consideration: mental consequences captured by the o The degree of physical and mental degree of incapacity: functional impairment o 0-24%: 0 points o The amount of help for personal o 25-49%: 1 point care and supervision needed o 50-65%: 2 points o The need for stimulation, training o 66-79%: 4 points and physical activity o 80-100%: 6 points o To what extent giving care restricts  Pillar 2 (max 12 points): consequences for the care provider (determining child’s participation in daily life in terms factor) of:  The need must be caused by the o Learning, education and social person’s medical condition integration: max 3 points  The care and supervision must be o Communication: max 3 point provided by private individuals o Mobility and movement: max 3 (including parents), for at least 2-2.5 points hours per week o Self-care: max 3 points  The private care needs should be long-  Pillar 3 (max 18 points): consequences for term in nature (2-3 years or more) the family with respect to (highest score  Help needed for practical assistance doubled): and care provided by public services o Follow-up of the treatment at home: are not taken into account max 3 points o Leaving the home for medical supervision and treatment: max 3 points o Adaptations to way of living: max 3 points Monthly benefit amount (2018): Monthly benefit amount (2018): Determined by the number of points: Determined by the amount of care and  <6 total points, ≥4 points on pillar 1: supervision needed: €80.75  Rate 1: €128  6-8 total points, <4 points on pillar 1:  Rate 2: €257 €107.55  Rate 3: €513  6-8 total points, ≥4 points on pillar 1:  Rate 4: €770 €414.28 Everyone with a disability, injury or illness  9-11 total points, <4 points on pillar 1: can apply for rate 1, whereas rates 2-4 are €250.97 restricted to children below the age of 18.  9-11 total points, ≥4 points on pillar 1: €414.28  12-14 total points: €414.28  15-17 total points: €471.07  18-20 total points: €504.71  >20 total points: €538.36 Source: compiled by the authors.

183

4.7.2. Appendix 4.2 Overview variables Belgian and Norwegian sample

Table A4.2.1. Overview variables

Belgium Norway Source DWH LM&SP (2010) and Census MBRN (2000-2005), NUDB and (2011) FD-Trygd (2008) Dependent variables Employed (0/1) 1 = working as an employee or 1= working as an employee or self-employeda (31 March 2010) self-employed in 2008b (n=16740) (n=457675) Employment Simulated gross yearly Gross yearly employment earnings employment income, PPP income, PPP adjustedd, ln adjustedc, ln transformed, transformed, employees only employees only (n=12203) (n=379243) Independent variables Children with Receiving supplemental child Receiving attendance benefit increased care benefit needs (CICN) Gender inequalities Mother Female partner in the household Biological mother (or second male partner) Mother x CICN Does the increased care burden differently affect the employment/earnings of mothers versus fathers? Education inequalities Parental education Highest ISCED level obtained on Highest ISCED level obtained on 1 January 2011 (low (0-2), 1 October 2008 (low (0-2), medium (3-4), high (5-6)) medium (3-4), high (5-6)) Parental education Does the increased care burden differently affect parental x CICN employment/earnings by the educational level of the parent? Controls Age, age² At birth of focal child, centred At birth of focal child, centred around the meane around the meane Age child In 2010, centred around the meane In 2008, centred around the meane Gender child Boy/girl Boy/girl Number of siblings Number of siblings (< 18) living at All children born with the same the same address mother Age youngest child Age in 2010 of youngest child in Age in 2008 of youngest child in the household the household Partner employed 1 = partner worked as an employee 1= partner worked as an (0/1) or self-employeda (31 March employee or self-employedb (1 2010) October 2008) Country of birth BE; EU27 + Iceland, NO; EU27 + Iceland, Liechtenstein, Norway, Liechtenstein, Switzerland; non- Switzerland; non-EU27 EU27

184

Belgium Norway Region of residence Brussels, Flanders, Wallonia Operationalised by controlling for the county unemployment rate, centred around the mean Source: compiled by the authors Notes: (a) parents who have an employment contract on 31 March 2010 but actually did not participate on the labour market (in terms of full-time equivalents) are recoded to 0; (b) parents who are employed but do not have employment income are recoded to 0; (c) 2010 conversion factor = 0.836; (d) 2008 conversion factor = 8.859. Accessed at https://data.oecd.org/conversion/purchasing-power-parities-ppp.htm. (e) Centred around the mean for children with and without increased care needs respectively.

4.7.3. Appendix 4.3 Descriptive information Belgian and Norwegian sample

Table A4.3.1. Descriptive information Belgian data, 2010

Children born in Belgium in 2000-05, living CICN No CICN in two-parent household Child characteristics Age (mean) 7.75 7.44 Gender Boys 66.31% 51.23% Girls 33.69% 48.77% Region of residence Brussels 4.15% 5.67% Flanders 67.20% 62.79% Wallonia 28.65% 31.53% Increased care needs 2.27% 97.73% Household characteristics Number of siblings (mean) 1.35 1.30 Age youngest child (mean) 5.77 5.53 Parental characteristics Mothers Fathers Mothers Fathers Age (mean) 29.64 32.49 29.80 32.32 Country of birth BE 91.51% 89.90% 90.71% 88.93% EU27 2.84% 2.75% 3.30% 3.19% Non-EU27 5.66% 7.35% 5.99% 7.88% Education Low-skilled 19.37% 28.09% 11.93% 19.55% Medium-skilled 43.93% 44.02% 37.82% 40.78% High-skilled 36.70% 27.89% 50.25% 39.67% Partner employed 90.05% 78.18% 95.30% 87.81%

185

Children born in Belgium in 2000-05, living CICN No CICN in two-parent household Outcome variables Employed (2010Q1) 75.26% 88.80% 86.18% 94.47% Gross employment income (mean) 29219.94 45361.75 34399.62 51173.47 Source: authors’ calculations on DWH LM&SP (2010) and Census (2011). Note: CICN = child with increased care needs.

Table A4.3.2. Descriptive information Norwegian data, 2008

Children born in Norway in 2000-05, living CICN No CICN in two-parent household Child characteristics Age (mean) 5.88 5.49 Gender Boys 62.03% 50.93% Girls 37.97% 49.07% Region of residence Unemployment rate county (2008Q1) 2.48 2.47 Increased care needs 3.21% 96.79% Household characteristics Number of siblings (mean) 1.31 1.07 Age youngest child (mean) 5.33 5.15 Parental characteristics Mothers Fathers Mothers Fathers Age (mean) 29.65 32.69 29.74 32.69 Country of birth NO 85.77% 86.46% 84.87% 86.10 % EU27 2.96% 2.90% 4.12% 4.10% Non-EU27 11.28% 10.64% 11.01% 9.80% Education Low-skilled 20.22% 21.13% 16.26% 17.20% Medium-skilled 39.68% 49.63% 37.00% 47.38% High-skilled 40.10% 29.24% 46.73% 35.42% Partner employed 88.70% 77.86% 91.92% 83.88% Outcome variables Employed 77.57% 88.43% 83.64% 91.55% Gross employment income (mean) 32828.00 55664.26 36242.54 60285.86 Source: authors’ calculations on MBRN (2000-05), NUDB and FD-Trygd (2008). Note: CICN = child with increased care needs.

186

4.7.4. Appendix 4.4 Employment and wage gaps between parents with and without children with increased care needs

Table A4.4.1. Linear probability model on parental employment

Employment regression Belgium Norway Constant 1.006*** 0.924*** (0.022) (0.004) Child with increased care needs (CICN) 0.010ns 0.008ns (0.007) (0.005) Gender inequalities Mother -0.104*** -0.088*** (0.006) (0.001) Mother x CICN -0.064*** -0.037*** (0.010) (0.006) Education inequalities Education (high-skilled ref.) Medium-skilled -0.070*** -0.031*** (0.007) (0.001) Low-skilled -0.128*** -0.139*** (0.012) (0.002) Education (high-skilled ref.) x CICN Medium-skilled x CICN -0.051*** -0.023*** (0.010) (0.006) Low-skilled x CICN -0.088*** -0.049*** (0.016) (0.009) Controls Age 0.000ns 0.001*** (0.001) (0.000) Age² -0.001*** -0.000*** (0.000) (0.000) Age child -0.001ns 0.004*** (0.002) (0.000) Gender child (Boy ref.) 0.005ns -0.001ns (0.006) (0.001) Number of siblings -0.031*** -0.011*** (0.005) (0.001) Age youngest child 0.007*** -0.001* (0.002) (0.000) Partner employed 0.038** 0.094*** (0.014) (0.002) Country of birth (BE/NO ref.) EU27 -0.005ns -0.023*** (0.020) (0.003)

187

Employment regression Belgium Norway Non-EU27 -0.094*** -0.143*** (0.019) (0.002) Region of residence (Flanders ref.) Brussels -0.070*** n/a (0.018) Wallonia -0.045*** n/a (0.007) Unemployment rate county n/a -0.015*** (0.001) Source: authors’ calculations on DWH LM&SP (2010) and Census (2011) for Belgium, and on MBRN (2000-05), NUDB and FD-Trygd (2008) for Norway. Notes: *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant. Robust standard errors are in parentheses. R² is 0.1444 for Belgium and 0.0867 for Norway. N is 16740 for Belgium and 457675 for Norway.

Table A4.4.2. OLS regression on gross parental employment income, ln transformed and ppp-adjusted

Earnings regression Belgium Norway Constant 11.054*** 11.156*** (0.045) (0.007) Child with increased care needs (CICN) -0.024ns -0.038*** (0.018) (0.010) Gender inequalities Mother -0.516*** -0.579*** (0.013) (0.002) Mother x CICN -0.092*** -0.060*** (0.019) (0.011) Education inequalities Education (high-skilled ref.) Medium-skilled -0.424*** -0.241*** (0.014) (0.002) Low-skilled -0.605*** -0.422*** (0.022) (0.003) Education (high-skilled ref.) x CICN Medium-skilled x CICN -0.012ns 0.005ns (0.020) (0.011) Low-skilled x CICN 0.020ns -0.070*** (0.029) (0.019) Controls Age 0.016*** 0.016*** (0.002) (0.000) Age² -0.001*** -0.001*** (0.000) (0.000) Age child 0.011* 0.022***

188

Earnings regression Belgium Norway (0.005) (0.001) Gender child (Boy ref.) 0.031* -0.001ns (0.013) (0.002) Number of siblings -0.061*** -0.029*** (0.010) (0.001) Age youngest child -0.004ns -0.005*** (0.004) (0.001) Partner employed 0.077** 0.023*** (0.028) (0.003) Country of birth (BE/NO ref.) EU27 -0.194*** -0.027*** (0.056) (0.005) Non-EU27 -0.266*** -0.200*** (0.036) (0.004) Region of residence (Flanders ref) Brussels 0.048ns n/a (0.036) Wallonia -0.062*** n/a (0.014) Unemployment rate in the county n/a -0.066*** (0.001) Source: authors’ calculations on DWH LM&SP (2010) and Census (2011) for Belgium, and on MBRN (2000-05), NUDB and FD-Trygd (2008) for Norway. Notes: *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant. Robust standard errors are in parentheses. Effects need to be interpreted as percentage differences. R² is 0.3500 for Belgium and 0.2972 for Norway. N is 12203 for Belgium and 379243 for Norway.

Table A4.4.3. Two-sample t-tests of cross-country differences (effects from Tables A4.4.1 and A4.4.2)

Employment regression Belgium Norway DF 16721 DF 457657 Gender inequalities (H3.1) Children in general Mother -0.104 -0.088 (SE 0.006) (SE 0.001) Difference (Belgium – Norway) -0.017 (SE difference 0.006) T-test difference -2.556* Children with increased care needs Mother X CICN -0.064 -0.037 (SE 0.010) (SE 0.006) Difference (Belgium – Norway) -0.026 (SE difference 0.012) T-test difference -2.283*

189

Employment regression Belgium Norway Education inequalities (H3.2) Children in general Medium-skilled -0.070 -0.031 (SE 0.007) (SE 0.001) Difference (Belgium – Norway) -0.039 (SE difference 0.007) T-test difference -5.760*** Low-skilled -0.128 -0.139 (SE 0.012) (SE 0.002) Difference (Belgium – Norway) 0.011 (SE difference 0.012) T-test difference 0.951ns Children with increased care needs Medium-skilled x CICN -0.051 -0.023 (SE 0.010) (SE 0.006) Difference -0.029 (SE difference 0.012) T-test difference -2.458* Low-skilled x CICN -0.088 -0.049 (SE 0.016) (SE 0.009) Difference -0.039 (SE difference 0.042) T-test difference -2.154* Earnings regression Belgium Norway DF 12184 DF 279225 Gender inequalities (H3.1) Children in general Mother -0.516 -0.579 (SE 0.013) (SE 0.002) Difference (Belgium – Norway) 0.063*** (SE difference 0.014) T-test difference 4.682*** Children with increased care needs Mother X CICN -0.092 -0.060 (SE 0.019) (SE 0.011) Difference (Belgium – Norway) -0.032 (SE difference 0.022) T-test difference -1.435ns

190

Earnings regression Belgium Norway Education inequalities (H3.2) Children in general Medium-skilled -0.424 -0.241 (0.014) (0.002) Difference (Belgium – Norway) -0.183 (SE difference 0.015) T-test difference -12.529*** Low-skilled -0.605 -0.422 (0.022) (0.003) Difference (Belgium – Norway) -0.184 (SE difference 0.022) T-test difference -8.181*** Children with increased care needs Medium-skilled x CICN -0.012 0.005 (SE 0.020) (SE 0.011) Difference -0.017 (SE difference 0.023) T-test difference -0.730ns Low-skilled x CICN 0.020 -0.070 (SE 0.029) (SE 0.019) Difference 0.090 (SE difference 0.034) T-test difference 2.634** Source: authors’ calculations on DWH LM&SP (2010) and Census (2011) for Belgium, and on MBRN (2000-05), NUDB and FD-Trygd (2008) for Norway. Notes: CICN = children with increased care needs. DF = degrees of freedom. SE = standard error. *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant.

191

PART III

CHILDHOOD DISABILITY AND CHILD POVERTY

193

CHAPTER 5 INCOME POVERTY AMONG CHILDREN WITH A DISABILITY IN BELGIUM: THE INTERPLAY BETWEEN PARENTAL EMPLOYMENT, SOCIAL BACKGROUND AND TARGETED CASH SUPPORT

Abstract

Previous research has shown a clear link between childhood disability and child poverty. This is related to the fact that their parents (1) need to provide more care which impedes their employment participation, and (2) more often belong to disadvantaged social categories. However, the adverse relationship between childhood disability and child poverty can be cushioned by cash support systems. Hitherto, the literature lacks insight into how the receipt of different cash support systems is related to the parents’ employment participation and social background, and what joint role these three factors play in understanding the poverty risk of these children. To fill this gap, a case study on Belgium is conducted making use of unique and large-scale register data. The results show that children with a disability have a lower income poverty risk than children without a disability, even when their parents’ labour market participation and social background are taken into account. This can be explained by the targeted cash support children with a disability receive. However, a substantial group of children with a disability does not receive this benefit. Hence, more could be achieved if the non-take-up would be addressed, in particular among the most vulnerable children.

195

5.1. Introduction

In this chapter, we build upon the findings from previous chapters to investigate the concurrence of childhood disability and child poverty in Belgium. On the one hand, it is shown in Chapters 3 and 4 that parents of children with a disability work less, earn less, and have a disadvantaged social background compared to parents of children without a disability. On the other hand, Chapter 1 demonstrates that children living in households with lower levels of labour market participation and a disadvantaged social background run an increased risk to live in income poverty, and that cash transfers targeted to families with children alleviate this poverty risk. The question whether, and if so why, childhood disability coincides with child poverty, is thus examined while the interplay between the family’s labour market participation, social background and receipt of cash support is looked at.

Child poverty is a tenacious multifaceted problem that is passed on from generation to generation, also in developed welfare states (UNICEF, 2007). In most of the OECD countries, the number of children growing up in a household with an income considered too low to live a minimal acceptable way of life has been on the rise (OECD, 2014b). This increase was intensified by the Great Recession in some of these countries (Cantillon et al. 2017; Jenkins et al., 2013). Consequently, the fight against child poverty has been high on the policy agenda in European welfare states (European Commission, 2013). It has been shown that child poverty rates are negatively correlated with redistributive policies such as, inter alia, child-related benefits (Diris et al., 2017; Van Lancker & Van Mechelen, 2015). Nowadays, policy strategies to reduce child poverty are often inspired by the social investment paradigm (Hemerijck, 2017; Morel et al., 2012) and primarily focus on labour market integration of the parents while simultaneous investments are made in the future potential of children to stop the vicious circle.

196

For parents of children with a disability, such a work strategy to solve poverty appears to be problematic (Cantillon & Van Lancker, 2013). Previous research has shown a clear link between childhood disability and child poverty, where the causality can run in both directions (Mont, 2014). This is related to the fact that these children generally require additional care that exceeds usual parental care and the time needed for this hampers their parents’ labour market participation (even more), of mothers in particular (Brown & Clark, 2017; Stabile & Allin, 2012). Moreover, their families face an extra burden on the household budget as the child’s disability commonly also entails higher out-of-pocket costs to cover the medical and care needs. Besides, children with a disability more often live in households with a disadvantaged social background which independently increases their poverty risk: their households are frequently comprised of single parents, parents holding lower educational qualification, and other household members with a disability (see Vinck & Van Lancker, 2020 or Chapter 3 for an overview of studies). However, the social policy literature lacks a clear insight into how the receipt of different cash support systems for children with a disability is related to their parents’ labour market participation and social background, and what joint role these three factors play in understanding their poverty risk.

To fill this gap, a case study on Belgium is conducted making use of unique and large-scale register data. Therefore, administrative recognitions on the child’s disability are linked to information on the income, employment and social background of their families. A control group of children without a disability is also included in the dataset. The central questions studied in this chapter are whether children with a disability face an increased poverty risk in Belgium, and what role parental employment, social background and targeted cash support systems jointly play in this story. A three-step analytical strategy is applied to

197 tackle these questions. First, the household’s equivalised net disposable household income and accompanying poverty risk are estimated for children with and without a disability. Second, a multivariate logistic regression is used to examine whether the correlations between child poverty on the one hand, and the family’s labour market participation and social background on the other hand, differ for children with and without a disability. Finally, the poverty reducing impact of the existing cash support systems provided to children with and without a disability in the form of tax advantages and cash benefits is tested. The causal chain of events is disregarded from the analyses.

5.2. Previous research and mechanisms

Previous research conducted in a wide range of countries has found a positive correlation between childhood disability and different indicators of child poverty and deprivation (for reviews on different parts of the world see e.g. Banks, Kuper, & Polack, 2017; Di Giulio et al., 2014; Elwan, 1999). This association has been reported in Australia (Leonard et al., 2005), Belgium (Debacker, 2007), Canada (Nikiéma, Spencer, & Séguin, 2010; Petrenchik, 2008), China (Guo et al., 2019), France (Delobel-Ayoub et al., 2015), the United Kingdom (Blackburn et al., 2010; Emerson, 2003; 2012; Emerson, Graham, & Hatton, 2006; Emerson & Hatton, 2007; Emerson et al., 2010; McKay & Atkinson, 2007; Nikiéma et al., 2010), the United States (Bauman et al., 2006; Fujiura & Yamaki, 2000; Houtrow et al., 2014; Parish, Rose, Grinstein-Weiss, Richman, & Andrews, 2008; Porterfield & Tracey, 2003; Pulcini, Zima, Kelleher, & Houtrow, 2017), and Vietnam (Mont & Cuong, 2011). However, it is not universally true for all disability types. For children with autism spectrum disorder (ASD), the relationship is found to be weaker (Emerson, 2012 for UK), non-existent (Pulcini et al., 2017 for the US), or even reversed (see

198

Durkin et al., 2017 for US). The latter might be attributed to a diagnosis bias: as already shown by Wing in 1980, parents with an advantaged social background more often obtain the correct diagnosis for their child with ASD, though this is more recently not necessarily confirmed in countries with universal health care systems (Rai et al., 2012 for Sweden).

Different mechanisms might explain this relationship between childhood disability and child poverty, where the causality may run in both directions (Boat & Wu, 2015; Elwan, 1999; Lustig & Strauser, 2007; Mont, 2014; Porterfield & Tracey, 2003). The “health selection” explanation states that childhood disability can cause child poverty since the disability entails higher costs, posing a burden on the household budget (see Anderson, Dumont, Jacobs, & Azzaria, 2007; Brown & Clark, 2017; Stabile & Allin, 2012 for reviews of the literature). Families of children with a disability are confronted with higher direct costs to pay for the child’s medical and care needs. Moreover, they face indirect costs in terms of forgone labour earnings as the care for children with a disability requires more time than for children without a disability, impeding their parents’ labour market participation, especially of mothers (for an overview of studies, see Chapter 3). For Belgium too, the presence of both direct and indirect costs has been reported for children with a disability (Van Landeghem et al., 2007; Vinck & Brekke, forthcoming or Chapter 4; Vinck & Van Lancker, 2020 or Chapter 3). If one would account for the additional direct costs, the already greater poverty risk of children with a disability in Flanders would increase substantially (Debacker, 2007). Moreover, it has been shown that the poverty risk of children in general is strongly determined by the level of labour market participation of the household they are living in (Gornick & Jäntti, 2012) and that Belgium occupies an exceptional position in this respect (Vandenbroucke & Vinck, 2013). In a European

199 comparative perspective, Belgium has the second highest relative share of children living in households where nobody works among the income-poor children in 2010 (EUROSTAT, 2020a, 2020b, 2020d)21. Yet, it is still unclear what role the indirect costs precisely plays in the Belgian poverty estimates among families of children with a disability.

In contrast, the “social causation” explanation implies that poverty during childhood or prior to childbirth may cause disability due to the detrimental health effects that may result from limited access to (preventive) health care, poorer health behaviour, growing up in bad or dangerous living environments, or parenting behaviour that is considered to be less assisting, encouraging or consistent (Bolte, Tamburlini, & Kohlhuber, 2010; Kiernan & Huerta, 2008; Maggi, Irwin, Siddiqi, & Hertzman, 2010; Power, Li, & Manor, 2000; Wilkinson & Marmot, 2003). For example, Delobel-Ayoub et al. (2015) show that more deprived areas in south-western France have a higher prevalence of children with ASD or intellectual disabilities than less deprived areas. Or, Blackburn, Spencer and Read (2013) report for the UK that children who grew up in socioeconomic deprived households in their early years are more likely to have disabling chronic conditions later in childhood than their counterparts from more privileged households. In fact, the poverty risk of children with a disability is strongly tied to processes of social stratification (Rothwell et al., 2019; Shahtahmasebi et al., 2011). Children with a disability more often live with parents holding lower educational qualification, single parents, and other household members with a disability (see Vinck & Van Lancker, 2020 or Chapter 3 for an overview of studies). This is no different in Belgium (Vinck & Van Lancker, 2020 or Chapter

21 Over the course of 2007 ‒ 2018, Belgium almost consistently occupies the second place after Ireland, except in 2007 when it is also preceded by Bulgaria, and in 2008 when the United Kingdom, Germany, the Czech Republic and Ireland perform worse.

200

3). These three factors increase the family’s poverty risk independently of having a child with a disability (Gornick & Jäntti, 2012; Grammenos, 2018; Nieuwenhuis & Maldonado, 2018; Vinck et al., 2017 or Chapter 1), but, whether and if so, to what extent, they explain the poverty risk of children with a disability in Belgium still has to be investigated.

Even though it is challenging to disentangle cause from effect in an empirical way, a mutually reinforcing relationship between childhood disability and child poverty might occur, yielding long-lasting negative consequences for the child’s development and opportunities over the life course (Elwan, 1999; McKinley Yoder & Cantrell, 2019; Mont, 2014; Parish & Cloud, 2006). This adverse relationship can be cushioned by welfare programmes that increase the household’s income (Boat & Wu, 2015; Luca & Sevak, 2019; Romig, 2017; Stegman Bailey & Hemmeter, 2014; Van Landeghem et al., 2007), though they have to be evaluated in relation to the costs incurred by the child’s disability (Byrne, 2014; Monteith et al., 2009). For Belgium, Van Landeghem et al. (2007) report that the difference in the equivalised disposable household income between families of children with and without a disability is smaller than when the difference is investigated on the basis of the household’s income from employment only, which, according to them, suggests that the social security system partially eliminates the inequalities. However, they do not provide evidence to pinpoint which component of the social security system succeeds in protecting the income of families with a child with disability. Moreover, if one wants to get insight into the poverty reducing impact of cash support systems among children with a disability, it is crucial to take into account who actually receives the cash support, in particular how the receipt is related to the family’s labour market participation and social background. To our knowledge, this remains to be unravelled.

201

5.3. Research questions and contributions to the literature

The central questions studied in this chapter are whether children with a disability face an increased poverty risk in Belgium (RQ1), and what role parental employment, social background and targeted cash support systems jointly play in this story (RQ2). Our analyses are complementary to previous Belgian research and contribute to the literature in two ways. The first contribution is that this chapter offers more detailed income poverty estimates among children with a disability in Belgium. Until now, the available research only provides descriptive evidence that families of children with a disability run a higher income poverty risk than families of children in general (Debacker, 2007), without looking into the role of potential confounding factors related to the family’s labour market participation and social background. In fact, the population of children with a disability studied in previous Belgian research includes only children who have obtained a disability recognition from the Flemish administrative agency responsible for providing subsidised care services and additional financial support (the Flemish Agency for Persons with a Disability, see Chapter 2 for more information on their recognition procedure). The population of children with a disability considered in this chapter only partially coincides with the one applied in previous Belgian research (see Section 5.4 for our disability definition and Chapter 2 on the overlap between the different recognitions). The second contribution is that this chapter tries to unravel how existing cash support systems for families of children with and without a disability reduce their income poverty risk while controlling for who actually receives this cash support. Drawing on a case study for Belgium, we are able to pinpoint which cash support system(s) provide(s) adequate income protection to families of children with a disability,

202 taking into account how the receipt of these systems is associated with the family’s labour market participation and social background.

The analyses allow to partially shed light on the health selection explanation (by controlling for the family’s labour market participation), the social causation explanation (by controlling for the family’s social background), and the cushioning impact of welfare programmes (by controlling for the receipt of cash support targeted at families of children with and without a disability) described in the previous section. However, the cross-sectional data at hand only allows to test the association between childhood disability and child poverty (see Section 5.4), therefore abstraction is made of the direction of the effect between the two. Moreover, direct costs related to the child’s increased medical and care needs cannot be taken into account. Hence, the analyses presented in this chapter will clarify hitherto unresolved questions on the income issue related to childhood disability, but not on the expenditure issue.

5.4. Data, variables and methods

A unique and large scale administrative dataset is used to answer this chapter’s central questions: do children with a disability face an increased poverty risk in Belgium, and what role do parental employment, social background and targeted cash support jointly play in this story? The dataset links microdata from the Datawarehouse Labour Market and Social Protection (DWH LM&SP) from December 31st, 2010 to the latest Belgian Census of January 1st, 2011. The DWH LM&SP brings together administrative data from Belgian social security agencies and the National Register on individuals’ personal and household characteristics, including their place of residence, income received from different sources, labour market participation, date and place of birth, disability status, and household

203 composition. Information on the educational level is taken from the 2011 Census. A random sample of 50% of children under 21 who lived in Belgium and received the supplemental child benefit on December 31st, 2010 and a randomly drawn control group of children under 21 who did not receive the supplemental child benefit of similar size are acquired.

Children with a disability are operationalised as children receiving the supplemental child benefit, a non-means-tested cash benefit designed for children with increased care needs who are less than 21 years old. It is a top-up of the regular child benefit for which an administrative recognition of the disability is needed. Control doctors of the Federal Public Service (FPS) Social Security evaluate the severity of the disability and award a score on a 36-point scale using a standardised criteria list. The scale intends to capture the consequences of the disability for the child’s (1) physical and mental health (maximum 6 points), (2) self-reliance in daily life (maximum 12 points), and (3) family (maximum 18 points). Higher scores correspond to a higher benefit amount, ranging from € 80 to more than € 500 per month. For more information on the recognition procedure, see Vinck et al. (2019) or Chapter 2. In 2018, 2.34% of all Belgian children minus 18 received the supplemental child benefit (FamiStat, 2019a, 2019b).

Child poverty is defined in accordance with the European headline at-risk-of- poverty indicator, as the share of children below the age of 18 living in a household with an equivalised net disposable household income below a poverty line set at 60% of the national median equivalised net disposable household income. It is an indirect monetary approach to child poverty since it does not directly measure the living conditions of children but of the household the children belong to, which is directly related to parental income. Moreover, the poverty measure is relative in nature as it compares the household’s income to a living standard considered

204 necessary in order to live a minimally acceptable way of life. We draw on the Belgian sample of the EU Statistics on Income and Living Conditions for income year 2010 (BE-SILC 2011) to set the 60% poverty threshold. The OECD-modified equivalence scale is used to make incomes comparable across households with different sizes and compositions (Hagenaars et al., 1994). Yet, this equivalisation does not correct for the costs incurred by families when they are confronted with increased medical and care needs (here when the child has a disability). Unfortunately, neither previous research nor the data at hand allow to take these increased out-of-pocket costs into account, but their potential presence should be kept in mind when interpreting the results. Both a poverty headcount ratio (hereafter income poverty risk) and poverty gap ratio are applied.

A three-phase analytical strategy is performed to investigate the link between child income poverty and childhood disability. First, the household’s equivalised net disposable household income and accompanying income poverty risk are estimated for children with and without a disability. Therefore, the gross taxable income information available in the DWH LM&SP is used. This is the income after social security contributions are paid but before taxes are withdrawn, if applicable22. Information is available for the income received from employment, self-employment, pensions, unemployment benefits, sickness and invalidity benefits, disability benefits, social assistance and child benefits23. To go from an individual’s gross taxable income to its household’s equivalised net disposable income, the withholding tax schedule is employed. These are advance payments to the final personal income taxation, taking only its essential aspects into

22 Not all income components are subject to social security contributions and/or taxes. 23 Child benefits are simulated.

205 consideration (FPS Finance, 2009b). For detailed information on the simulation process, the reader is referred to Appendix 5.1.

Second, a multivariate logistic regression is used to examine whether the correlations between child poverty on the one hand, and the family’s labour market participation and social background on the other hand, differ for children with and without a disability, controlling for the region of residence (Flanders/Brussels/Wallonia). To serve that purpose, interaction effects between the child’s disability status (0/1, one referring to children who receive the supplemental child benefit) and each of the six family characteristics regarding its labour market participation and its social background are included in the model. The family’s labour market participation is measured jointly by an indicator labelled “household work intensity”. It is defined as the ratio between the total number of months worked (expressed in full-time equivalents) by all working-age household members (18-59 year olds, excluding students 18-24 years old) and the total number of months that they could, in theory, have worked. The ratio goes from zero to one, with zero meaning that none of the working-age household members participated in paid employment in that year, while one indicates that all working-age household members worked full-time for the full year24. The work intensity indicator is categorised in line with European practice into very low (work intensity ranges between 0 and 0.2), low (0.2-0.45), medium (0.45-0.55), high (0.55-0.85) and very high (0.85-1) work intensity.

With respect to the family’s social background, the role of household composition, parental education, parental migration background and the presence of other

24 If the actual employment status was different from the contractual situation, the household work intensity variable was recoded to zero. This was the case for 3.9% of children with a disability and 2.6% of children without a disability, minus 18 in the full sample.

206 household members with a recognised disability is investigated. For the household composition, the household type (single parents versus two parents) as well as the number of children under 18 present in the household (less than three versus three or more) are included in the analyses. The International Standard Classification of Education (ISCED) is used to operationalise the educational level of the parents, differentiating between parents with low (ISCED 0-2: lower secondary education or less), medium (ISCED 3-4: secondary education) and high educational qualifications (ISCED 5-6: tertiary education). The highest education level of one of the parents is taken as the household value. For the migration background, a distinction is made between parents born in Belgium, other European Union countries (EU27) and non-EU27 countries. When one of the parents is born in Belgium or in another EU27 country, the household is considered to have a Belgian or EU27 migration background respectively. The household is assigned a non-EU27 migration background if the parents are born in a non-EU27 country. Finally, consistent with the operationalisation of childhood disability, the disability of other household members is identified by the receipt of a disability- specific benefit. A dummy variable indicates whether at least one other household member receives either the supplemental child benefit (for minus 21-year olds) or a disability benefit for individuals aged over 21. An administrative recognition of the disability is needed for the latter too, but, contrary to the supplemental child benefit, the receipt of the benefit is also conditioned on a means test.

The poverty reducing impact of the existing cash support systems for children with and without a disability in the form of tax advantages and cash benefits is tested in a final step. Therefore, the first and second analytical step are repeated for three alternative scenarios when the main cash support systems are, simultaneously and separately, excluded from the net disposable income. Three components of income

207 protection for families of children with or without a disability are considered: the regular child benefit (CB), the supplemental child benefit (SCB) and the refundable tax credit (TC). More information on the characteristics of these cash supplements can be found in Section 3.1.1 of the introduction to this thesis. The first of the three alternative scenarios subtracts all this cash support from the household’s income (‒ CB ‒ SCB ‒ TC), the second includes only the regular child benefit (+ CB ‒ SCB ‒ TC), and the third additionally adds the supplemental child benefit (+ CB + SCB ‒ TC). For all these scenarios, it is examined whether the cash support systems alter the overall income poverty rate and gap as well as the associations that exist between child poverty, childhood disability and the family characteristic (in terms of labour market participation and social background) compared to the baseline scenario when all three cash support systems are included (+ CB + SCB + TC).

The analyses focus on children under 18 in accordance with the child poverty definition. Children whose parents have zero income on any of the income components (excluding child benefits) are disregarded from the analyses as it cannot be assessed with certainty that they truly have no income. They could have no known labour income within the Belgian personal income taxation since they work as outbound workers, are employed at a European or other international organisation, or receive a non-taxable scholarship. Or, they could solely be living off other income sources than the ones related to (current and past) employment (e.g. income from moveable or immovable property). This reduces the sample of children minus 18 with 19%25, resulting in 18,486 children with a disability and

25 The dropped children are mainly living in Brussels, with two parents, low-skilled parents and/or parents born outside Belgium. Mothers have more often zero incomes than fathers. Among fathers with zero income, 80% does not occur in any social security record (and has therefore no known income), the remaining 20% are almost all self-employed. The corresponding numbers for mothers with zero income are 90% (no social security record) and 10% (self-employed).

208

16,961 children without a disability. A population weight is applied to both samples to represent the full Belgian population of children with and without a disability. The total samples are used to estimate the overall income poverty rate and gap in the first and third analytical step as well as for the descriptive analyses on the variables of interest included in Appendix 5.2. The logistic regression model in the second and third analytical step omits children who have missing information on any of the variables included (resulting in 17,677 children with a disability and 16,206 children without a disability). Sensitivity checks with the 50% and 70% poverty thresholds (Appendix 5.3) and without applying the population weight (Appendix 5.4) generally yield the same conclusions.

5.5. Results

5.5.1. Income poverty estimates and determinants among children with and without a disability

The estimated income poverty rate and gap for children below the age of 18 living in Belgium in 2010 are presented in Table 5.1, differentiating between children with and without a disability. Of all children, 22.1% live at risk of income poverty and their average depth of poverty amounts 18.8%26. The results show that children with a disability face both a smaller income poverty risk (15.6%) and gap (14.3%) compared to children without a disability (22.2% and 18.8% respectively). It should be stressed that these estimates, as well as the following, only look at the incomes of the households these children are living in, not at the (extra) costs these households have to make to cover for the child’s (increased)

26 Comparing the DWH LM&SP estimates to the corresponding statistics from BE-SILC 2011, yield a lower poverty headcount ratio (18.7% in BE-SILC) but a higher poverty gap ratio (26.0% in BE-SILC).

209 medical and care needs. Hence, we do not make any inferences about the standard of living of these households.

Table 5.1. 60% income poverty risk and average income poverty gap ratio among the poor, children under 18, with versus without a disability, Belgium, 2010

With a Without a Children <18 Total disability disability 60% poverty risk 22.1% 15.6% 22.2% Average 60% poverty gap ratio among the poor 18.8% 14.3% 18.8% Source: own calculations based on DWH LM&SP (2010).

The multivariate logistic regression confirms that children with a disability have a lower income poverty risk compared to children without a disability (odds ratio smaller than one in Table 5.2), even when their family’s labour market participation and social background are controlled for. Moreover, in line with the social policy literature on child poverty, Table 5.2 shows that the positive correlations that exist between the child’s income poverty risk on the one hand, and lower degrees of labour market participation and various indicators of a disadvantaged social background of the family on the other, are found for both children with and without a disability (odds ratios of the main effects are larger than one). Children who live in households with lower levels of work intensity (compared to households with very high work intensity), headed by a single parent (versus two parents), with three or more children (versus less than three), with low- or medium-skilled parents (versus high-skilled parents), and parents who are born outside Belgium (versus within Belgium), experience a higher income poverty risk than their respective reference groups. Only for the presence of other household members with a disability the reverse is true (odds ratio smaller than one). Yet, for some of these associations, the strength differs between children

210 with and without a disability. Weaker associations are reported for children with a disability with respect to the household’s work intensity, the household type and the non-EU27 migration background of the parents (compared to when at least one parent is born in Belgium). This is apparent from the multiplication of the main effects of the family’s characteristics and their interaction effects with childhood disability. If this multiplication brings the association closer to the value of one, the correlation between the family’s characteristics and child poverty is weaker for children with a disability than for children without a disability. For example, the increased income poverty risk of children living with a single parent (compared to living with two parents) is larger for children without a disability (among them, the odds are 5.334 times higher) than for children with a disability (among them, the odds are only 2.119 times as large (5.334 multiplied by 0.397)). For the number of children living in the household, the parental educational level, the presence of other household members with a disability and when parents have a EU27 migration background (compared to born in Belgium), no significant differences between children with and without a disability are found. Figure 5.1 visualises these results. It presents the relationship between the predicted income poverty risk and each of the family’s characteristics as marginal effects, for children with and without a disability separately, keeping all other variables in the logistic regression model at their mean value.

211

Table 5.2. Logistic regression on 60% income poverty risk, children minus 18, Belgium, 2010

Odds Robust standard Significance Children <18 ratio error level Constant 0.018 0.001 *** Child with a disability (CWD) 0.763 0.081 ** Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) 20.453 1.847 *** Low (0.2-0.45) 25.760 2.741 *** Medium (0.45-0.55) 8.514 0.692 *** High (0.55-0.85) 2.532 0.179 *** Household type (two parents ref.) Single parent 5.334 0.332 *** Number of children (<18) in household (less than three ref.) Three or more 1.278 0.074 *** Parental education (highest level) (high-skilled ref.) Low-skilled 2.669 0.212 *** Medium-skilled 2.738 0.170 *** Country of birth parents (Belgium ref.) EU27 1.590 0.234 ** Non-EU27 2.572 0.238 *** Other household members with a disability (none ref.) Yes, at least one 0.357 0.059 *** Interaction x CWD Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) x CWD 0.478 0.062 *** Low (0.2-0.45) x CWD 0.514 0.075 *** Medium (0.45-0.55) x CWD 0.489 0.061 *** High (0.55-0.85) x CWD 0.728 0.086 ** Household type (two parents ref.) Single parent x CWD 0.397 0.035 *** Number of children (<18) in household (less than three ref.) Three or more x CWD 1.130 0.088 n.s. Parental education (highest level) (high-skilled ref.) Low-skilled x CWD 0.898 0.103 n.s. Medium-skilled x CWD 0.864 0.084 n.s. Country of birth parents (Belgium ref.) EU27 x CWD 0.887 0.179 n.s.

212

Odds Robust standard Significance Children <18 ratio error level Non-EU27 x CWD 0.609 0.071 *** Other household members with a disability (none ref.) Yes, at least one x CWD 1.191 0.213 n.s. Region of residence (Flanders ref.) Brussels 1.525 0.138 *** Wallonia 1.414 0.079 *** Model fit Log pseudolikelihood -499943.92 Pseudo R² 0.4124 Prob > chi² 0.0000 N 33883 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: not living at-risk-of poverty is the baseline. *p<0.05, **p<0.01, ***p<0.001, n.s. not significant.

Figure 5.1. Marginal effects of family characteristics on the predicted 60% income poverty risk, children under 18, with versus without a disability, Belgium, 2010

Household work intensity* Household type* 70% 40% 60% 35% 50% 30% 25% 40% 20% 30% 15% 20% 10%

10% 5%

Predicted child poverty risk childpoverty Predicted risk childpoverty Predicted 0% 0% VLWI LWI MWI HWI VHWI Two parents Single parent

Children without a disability Children without a disability Children with a disability Children with a disability

(Figure continues on the next page)

213

Number of childrenn.s. Parental educationn.s. 16% 20% 14% 12% 15% 10% 8% 10% 6% 4% 5% 2%

Predicted child poverty risk childpoverty Predicted 0% Predicted child poverty risk childpoverty Predicted 1-2 children 3 or more 0% children Low Medium High

Children without a disability Children without a disability Children with a disability Children with a disability

Country of birth parents Other household members with a disabilityn.s. 30% 14% 25% 12% 20% 10% 15% 8% 6% 10% 4%

5% 2% Predicted child poverty risk childpoverty Predicted

0% 0% Predicted child poverty risk childpoverty Predicted Belgium EU27 (n.s.) Non-EU27* None Yes, at least one

Children without a disability Children without a disability Children with a disability Children with a disability

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: marginal effects are shown for each interaction effect of Table 5.2, at mean values for the remaining variables. 95% confidence intervals show that, in each situation, children with a disability have a lower predicted probability to live in income poverty than children without a disability. n.s. the slopes of the lines (w.r.t. the reference) do not differ significantly between children with a disability and children without a disability, * the slopes differ significantly.

214

5.5.2. Poverty reducing impact of cash support systems

Given that parents of children with a disability work less, earn less, and have a disadvantaged social background (see Chapters 3 and 4), the lower income poverty risk of these children compared to children without a disability presented in the previous section is a counterintuitive result. This suggests that the Belgian cash support systems succeed in mitigating the concurrence of childhood disability and child poverty proven in previous case studies in different countries. In this section, we unravel which cash support system(s) provide(s) adequate income protection to families of children with a disability, taking into account who actually receives the support.

Table 5.3. 60% income poverty risk and average income poverty gap ratio among the poor, with and without cash support, children under 18, with versus without a disability, Belgium, 2010

With a Without a Children <18 Total disability disability 60% poverty risk ‒ CB ‒ SCB ‒ TC (scenario 1) 35.3% 46.7% 35.0% + CB ‒ SCB ‒ TC (scenario 2) 23.1% 31.6% 23.0% + CB + SCB ‒ TC (scenario 3) 22.9% 18.7% 23.0% + CB + SCB + TC (baseline, all-in) 22.1% 15.6% 22.2% Average 60% poverty gap ratio among the poor ‒ CB ‒ SCB ‒ TC (scenario 1) 29.7% 31.5% 29.7% + CB ‒ SCB ‒ TC (scenario 2) 20.5% 19.0% 20.6% + CB + SCB ‒ TC (scenario 3) 20.5% 15.3% 20.6% + CB + SCB + TC (baseline, all-in) 18.8% 14.3% 18.8% Source: own calculations based on DWH LM&SP (2010). Notes: CB = regular child benefit, SCB = supplemental child benefit, TC = refundable tax credit for dependent children. The poverty threshold is kept constant over all scenarios. The average poverty gap is calculated among those children who are at-risk-of-poverty in each scenario separately. The baseline scenarios are the estimates presented in Table 5.1.

When the income poverty rate and gap are estimated without including the three main cash support systems families with children receive, the picture changes

215

(Table 5.3). In the scenario where the regular child benefit, the supplemental child benefit and the refundable tax credit for dependent children are simultaneously excluded from the household’s income, the overall income poverty risk among children would increase with 60% (from 22.1% in the baseline to 35.3% in scenario 1), but the increase would be stronger among children with a disability than among children without a disability (+199% versus +58%). As a matter of fact, children with a disability would run a higher risk to live in income poverty than children without a disability when these three cash support systems are subtracted from their household’s income (46.7% versus 35.0%). Furthermore, Table 5.3 shows that each cash supplement has a considerable poverty reducing impact for the children who receive them. The regular child benefit is important for children with and without a disability alike: it decreases their poverty rate with 15.2 and 12.1 percentage points (pp) respectively (comparing scenario 1 and 2). However, if the regular child benefit would be the only cash supplement granted to children, those who have a disability would still run a higher poverty risk compared to those without (31.6% versus 23.0%). Only once the supplemental child benefit is taken into account in the household’s income, the poverty risk of children with a disability drops below the poverty risk of children without a disability (18.7% versus 23.0%). Including the refundable tax credit for dependent children reduces the poverty risk further (comparing scenario 3 to the baseline), more for children with a disability than for those without (-3.1pp versus -0.7pp). The same lessons can be learned from the average poverty gap.

In Table 5.4, the income poverty risks for children with and without a disability are estimated under the different scenarios, controlling for their family’s labour market participation and social background.

216

Table 5.4. Logistic regression on 60% income poverty risk, with and without cash support, children under 18, odds ratios, Belgium, 2010

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Constant 0.030*** 0.018*** 0.018*** 0.018*** (0.002) (0.001) (0.001) (0.001) Child with a disability (CWD) 0.931n.s. 0.987n.s. 0.749** 0.763** (0.074) (0.095) (0.079) (0.081) Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) 13.949*** 23.831*** 23.867*** 20.453*** (1.445) (2.209) (2.213) (1.847) Low (0.2-0.45) 25.087*** 29.435*** 29.462*** 25.760*** (3.140) (3.185) (3.188) (2.741) Medium (0.45-0.55) 9.729*** 8.679*** 8.689*** 8.514*** (0.740) (0.709) (0.710) (0.692) High (0.55-0.85) 3.010*** 2.621*** 2.621*** 2.532*** (0.165) (0.184) (0.184) (0.179) Household type (two parents ref.) Single parent 8.478*** 5.783*** 5.784*** 5.334*** (0.527) (0.363) (0.363) (0.332) Number of children (<18) in household (less than three ref.) Three or more 3.498*** 1.436*** 1.436*** 1.278*** (0.176) (0.082) (0.082) (0.074) Parental education (highest level) (high-skilled ref.) Low-skilled 3.755*** 2.662*** 2.661*** 2.669*** (0.276) (0.212) (0.212) (0.212) Medium-skilled 3.450*** 2.729*** 2.728*** 2.738*** (0.177) (0.169) (0.169) (0.170) Country of birth parents (Belgium ref.) EU27 1.472** 1.663*** 1.665*** 1.590** (0.219) (0.245) (0.245) (0.234) Non-EU27 3.593*** 2.701*** 2.703*** 2.572*** (0.358) (0.252) (0.252) (0.238) Other household members with a disability (none ref.) Yes, at least one 0.503*** 0.386*** 0.387*** 0.357*** (0.079) (0.064) (0.064) (0.059) Interaction x CWD Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) x CWD 1.099n.s. 0.906n.s. 0.480*** 0.478***

217

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC (0.142) (0.107) (0.061) (0.062) Low (0.2-0.45) x CWD 0.839n.s. 0.824n.s. 0.505*** 0.514*** (0.137) (0.117) (0.073) (0.075) Medium (0.45-0.55) x CWD 0.794** 0.698*** 0.478*** 0.489*** (0.079) (0.076) (0.058) (0.061) High (0.55-0.85) x CWD 0.942n.s. 0.928n.s. 0.759** 0.728** (0.072) (0.091) (0.086) (0.086) Household type (two parents ref.) Single parent x CWD 1.081n.s. 0.978n.s. 0.431*** 0.397*** (0.091) (0.080) (0.037) (0.035) Number of children (<18) in household (less than three ref.) Three or more x CWD 1.079n.s. 1.126n.s. 1.266** 1.130n.s. (0.074) (0.084) (0.097) (0.088) Parental education (highest level) (high-skilled ref.) Low-skilled x CWD 1.103n.s. 1.117n.s. 1.012n.s. 0.898n.s. (0.107) (0.117) (0.113) (0.103) Medium-skilled x CWD 1.028n.s. 0.994n.s. 0.922n.s. 0.864n.s. (0.074) (0.084) (0.113) (0.084) Country of birth parents (Belgium ref.) EU27 x CWD 1.398n.s. 1.061n.s. 0.880n.s. 0.887n.s. (0.309) (0.231) (0.178) (0.179) Non-EU27 x CWD 0.774n.s. 0.837n.s. 0.600*** 0.609*** (0.106) (0.101) (0.071) (0.071) Other household members with a disability (none ref.) Yes, at least one x CWD 1.823*** 1.527** 1.161n.s. 1.191n.s. (0.309) (0.273) (0.207) (0.213) Region of residence (Flanders ref.) Brussels 1.514*** 1.551*** 1.533*** 1.525*** (0.126) (0.141) (0.139) (0.138) Wallonia 1.486*** 1.414*** 1.403*** 1.414*** (0.072) (0.079) (0.079) (0.079) Model fit Log pseudolikelihood -618515.13 -496206.19 -495645.20 -499943.92 Pseudo R² 0.4193 0.4319 0.4288 0.4124 Prob > chi² 0.0000 0.0000 0.0000 0.0000 N 33883 33883 33883 33883 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: not living at-risk-of poverty is the baseline. *p<0.05, **p<0.01, ***p<0.001, n.s. not significant. The baseline scenario presents the same results as Table 5.2. Robust standard errors are in parentheses.

218

Once these confounding factors are taken into account, the risk to live in income poverty would no longer differ significantly between children with and without a disability when the three main cash support systems are all excluded from their household’s income (scenario 1) or when only the regular child benefit is included (scenario 2). In other words, their family’s lower levels of labour market participation and disadvantaged social background explain why children with a disability would have a higher income poverty risk than children without a disability in the first and second scenario presented in Table 5.3. Moreover, none of the associations between the family’s characteristics and the income poverty risk vary in a significant way between children with and without a disability in these scenarios (the interaction effects of scenario 1 and 2 in Table 5.4).

The story changes once the remaining cash support systems are taken into account (scenario 3 and baseline). Table 5.4 confirms their strong poverty reducing impact for children with a disability, the supplemental child benefit in particular, when the family’s labour market participation and social background are controlled for. After adding the supplemental child benefit to the household’s income, the income poverty risk among children with a disability is significantly lower than among children without a disability (scenario 3), which is slightly further reduced once the refundable tax credit for dependent children is included too (baseline). This can be explained by the profile of those who receive the supplemental child benefit: they are more likely to live in poverty due to their lower levels of labour market participation and their disadvantaged social background (see also Vinck & Van Lancker, 2020 or Chapter 3). Here it is evident from the interaction effects. Once the supplemental child benefit and the refundable tax credit are integrated into the household’s income the associations that exist between the child’s income poverty risk and it’s family characteristics are mitigated. The correlations with the

219 household’s labour market participation, the household type and the parents’ non- EU27 migration background are now significantly less strong for children with a disability.

Figure 5.2 visualises the poverty reducing impact of the different cash support systems. For each scenario, it presents the relationship between the child’s disability status and the predicted income poverty risk as marginal effects, keeping all other variables in the logistic regression models at their mean value (the black and grey lines, shown on the left axis). Moreover, it shows the weighted average of the equivalised cash support that income-poor households with children receive in each scenario, differentiating between the cash support systems and between children with and without a disability (the bars, shown on the right axis). Three lessons can be learned from Figure 5.2. First, families of children with a disability get, on average, higher regular child benefits than families of children without a disability. This is because they more often (1) qualify for a social supplement, (2) have older children and therefore receive higher age-related supplements, and (3) have multiple children for which they get rank-adjusted universal child benefits (see Appendix 5.5 for the detailed analysis). Second, also the amount of tax credit for dependent children they get refunded is higher as children with a disability are counted twice as dependent in the system (see also Section 5.7.1.1 in Appendix 5.1). Third, the supplemental child benefit for children with a disability is indispensable in protecting their household’s income. In fact, it succeeds to halve the income poverty risk of children with a disability

220

Figure 5.2. Marginal effect of child’s disability status on predicted 60% income poverty risk (left axis) and mean equivalised cash support among the poor (right axis), in different scenario’s with and without cash support, children under 18, Belgium, 2010

30% € 400

€ 350 25% € 300 20% € 250

15% € 200

€ 150 10%

€ 100 (per month) 5% Predicted child povertyrisk € 50

0% € - Children Children Children Children Children Children Children Children without with a without with a without with a without with a a disability a disability a disability a disability

disability disability disability disability Meanequivalised cashsupport among the poor - CB - SCB - TC + CB - SCB - TC + CB + SCB - TC + CB + SCB + TC Scenario 1 Scenario 2 Scenario 3 Baseline

Mean equivalised refundable tax credit at household level Mean equivalised supplemental child benefit at household level Mean equivalised child benefit at household level Income poverty risk children without a disability Income poverty risk children with a disability

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: CB = regular child benefit, SCB = supplemental child benefit, TC = refundable tax credit for dependent children. Predicted child poverty risks are shown as marginal effects of the child’s disability status (left axis), at mean values for the remaining variables of Table 5.4. Mean equivalised cash support is shown for poor children with and without a disability, by component and population weighted (right axis). The population of poor children changes in each scenario.

221

5.6. Discussion and conclusion

This chapter investigates whether childhood disability coincides with child poverty while the interplay between the family’s labour market participation, social background and receipt of targeted cash support is taken into account. To do so, a case study on Belgium is conducted making use of unique and large-scale register data.

It should be noted that the analyses are limited in four ways. First, as mentioned multiple times throughout this chapter, only the incomes of families with children with and without a disability are looked at, not the expenditures they have to make. By using the OECD-modified equivalence scale to make incomes comparable across households, it is indirectly supposed that the same standard of living applies to families with the same equivalised income, regardless of whether they have a child with a disability or not. Yet, it is likely that an equally high income cannot be converted into an equally high standard of living as families of children with a disability face higher out-of-pocket costs to pay for the children’s additional medical and care needs while families of children without a disability do not. Accounting for these increased costs by exploiting information on what households actually spend or (minimally) need when a child with a disability is around, is an important research strand for the future. Previous studies on adults with a disability have attempted to do so and have (1) included the extra costs in poverty estimates and (2) assessed the extent to which these costs are covered by (governmental) support (e.g. Loyalka, Liu, Chen, & Zheng, 2014; Mont & Cuong, 2011; Morris & Zaidi, forthcoming; Palmer, Williams, & McPake, 2019; Saunders, 2007; Zaidi & Burchardt, 2005), though comparable exercises for children remain unexplored territory. Based on this, it can be expected that the

222 poverty estimates provided in this chapter are lower-bound estimates of the real standard of living of families with children with a disability.

The poverty risks presented here are underestimated in another way too. The data at hand allows to consider only children who receive the supplemental child benefit as disabled. However, this benefit is subject to the issue of non-take-up. In Vinck et al. (2019 or Chapter 2) it is estimated that at least 10% of children with a disability are not recognised for the supplemental child benefit and hence do not receive the extra money. Future research should look into the income situation, the corresponding poverty risk and the joint impact of parental employment, social background and receipt of other cash support among children with a disability who are missed by the benefit. The available research for Flanders suggests that for this group the income poverty risk would be higher than for children without a disability (Debacker, 2007), but how it correlates with potential confounding factors and what impact the other cash support systems have on their income poverty risk still has to be investigated.

Third, and related to the previous point, the data does not contain sufficient information to include other cash support systems available to families of children with a disability into the analyses (i.e. mainly personal care budgets). The number of children that actually receive this budget is likely to be limited in the year of data collection though27. Hitherto, the shifting policy logic from funding care institutions (i.e. supply-side financing) towards funding persons with a disability themselves (i.e. demand-side financing) remains an ongoing process for children. At the time of writing, it is unclear when this reform will be finalised. An assessment tool to gauge the care and support needs of children and the

27 Only 1,808 persons with a disability younger than 65 years old living in Flanders received a personal care budget on 1 January 2011 (FAPD, 2010), breakdown by age is however not possible.

223 corresponding budget they are entitled to, still has to be (further) developed. Whether these budgets should be counted as household income and how they impact on the income poverty risk of children with a disability remain open questions that should be examined in future research.

Fourth, given the cross-sectional nature of the data, it is not possible to discern the causal mechanisms between childhood disability and child poverty. Exploring the associations on longitudinal data would allow to get causal purchase on the issue. Moreover, a dataset where families are followed before and after the birth of the child (with a disability) would make it possible to examine whether the targeted cash support measures unintentionally acts as a disincentive to work for the families who receive them.

Keeping these limitations in mind, the results show that children with a disability have a lower income poverty risk than children without a disability, even when their family’s labour market participation and social background are taken into account. This can be explained by the targeted cash support these families receive. When families of children with a disability would only receive the regular child benefit, they would run an equally high income poverty risk than children without a disability once their family’s labour market participation and social background are controlled for. It is especially the supplemental child benefit that has a strong poverty reducing impact for children with a disability. It is indispensable in protecting their household’s income, in fact, it succeeds to halve the income poverty risk of children with a disability. This can be explained by the profile of the children who receive the supplemental child benefit: they initially run an increased poverty risk due to their family’s lower levels of labour market participation and disadvantaged social background (Vinck et al., 2019 or Chapter 2). Hence, the associations between income poverty among children and some of

224 the explanatory factors identified in the literature are significantly weaker for children with a disability than for children without a disability. This is true for children who live in households with lower levels of labour market participation, headed by single parents, and where the parents are born outside the EU27 (versus at least one parent is born in Belgium). For the number of children living in the household, the parental educational level, the presence of other household members with a disability and parents with a EU27 migration background (compared to born in Belgium), no significant differences between children with and without a disability are found.

One important policy implication can be derived from this. It is clear that the supplemental child benefit succeeds in protecting the incomes of families with a child with a disability, even when their lower levels of employment and disadvantaged social background are taken into account. In fact, it mitigates the relationship between the income poverty risk among children and the explanatory factors identified in the literature for the families who receive the benefit. However, a substantial group of children with a disability does not receive the supplemental child benefit. Hence, more could be achieved if the issue of non- take-up would be addressed, in particular among the most vulnerable children.

225

5.7. Appendix Chapter 5

5.7.1. Appendix 5.1 Equivalised net disposable household income simulation

5.7.1.1. Simulation strategy

An equivalised net disposable household income is simulated in six steps using the gross taxable income information available in the DWH LM&SP for income from current and past employment28 and the non-taxable income received from disability benefits, social assistance and child benefits (the latter are simulated).

In a first step, the professional expenses are simulated and subtracted from gross taxable income to get the net taxable income (i.e. the income on which taxes are calculated). Belgium has two systems which can be used to deduct professional expenses: either the actual expenses are proven or a lump sum amount is deducted depending on the level of income. The assumption is made that all employees make use of the lump sum amount as official statistics show that only 3.7% of them reported actual expenses for income year 2010 (FPS Finance, 2012). The income bands and percentages applied on each band to compute the lump sum professional expenses for income year 2010 are presented in Table A5.1.1. For income from self-employment, the administrative data already contain a net taxable income concept, hence no professional expenses need to be simulated. People receiving replacement income can only make use of the actual professional expenses system, which is not simulated. Unfortunately, the data does not have sufficient information to simulate other tax deductions which reduce taxable income. Therefore, the tax advantage granted for inter alia paid maintenance

28 Including income from wage employment, self-employment (available at the net taxable income level), pensions, unemployment benefits, sickness and invalidity benefits.

226 allowances, mortgage repayments and used child care services cannot be taken into account.

Table A5.1.1. Professional expenses: lump sum calculation income year 2010

Minimum gross Maximum gross % applied to Cumulative taxable income limit taxable income limit income band maximum amount (included) to be deducted 0.00 5190.00 28.7% € 1489.53 5190.00 10310.00 10.0% € 2001.53 10310.00 17170.00 5.0% € 2344.53 17170.00 58685.67 3.0% € 3590.00 58685.67 max 0.0% € 3590.00 Source: FPS Finance (2011).

In a second step, fiscal households are constructed. These consists of the head, his or her partner, dependent children and other dependent persons living at the same address. The dependency of children and other persons depends on their personal income. The yearly income from any source excluding child benefits cannot be higher than € 2,830 in 2010. For children, the income limits to be regarded as dependent are higher when they live in a single parent household (€ 4,080) and when they have a recognised disability of at least 66% (€ 5,180). Only children eligible for the child benefit on December 31st, 2010 are taken into account. If other persons are living at the same address, who do not fulfil the conditions to be considered dependent, they are assumed to form a fiscal household on their own. The head is stepwise identified as the individual who has the highest income from any source (excluding child benefits), is the oldest, or is registered as the reference person in the administrative dataset. Both married and cohabiting individuals are considered as partners with no distinction being made between registered and unregistered cohabiting individuals.

227

Table A5.1.2. Withholding tax schedule and final personal income tax schedule, income year 2010

Minimum net taxable income Maximum net taxable income limit % applied on limit (yearly) (yearly, included) income band Part A: withholding tax schedule 0 7900 26.75% 7900 10740 32.10% 10740 15560 42.80% 15560 34360 48.15% 34360 max 53.50% Part B: final personal income tax schedule 0 7900 25.00% 7900 11240 30.00% 11240 18730 40.00% 18730 34330 45.00% 34330 max 50.00% Source: FPS Finance (2010, 2011).

Thirdly, the withholding taxes are simulated on the net taxable income components for each earning individual separately. These withholding taxes are advance payments to the final personal income taxation, taking only its essential aspects into consideration (FPS Finance, 2009b). The withholding tax schedule is applied to the sum of income from employment, self-employment and pensions (see Table A5.1.2 part A), which has in general narrower bands but higher percentages than the final personal income taxation (see part B for comparison). For fiscal households where a partner is present, it is beforehand checked whether the marital quotient applies. When the couple has only one earner, 30% of the income of the earning partner (from employment, self-employment and pensions) is treated as the income of the non-earning partner, limited to € 9,280. Within the withholding tax system, the non-earning partner may not have any source of taxable income. As the Belgian income taxation system is progressive in nature, the application of the marital quotient will tax the income assigned to the non- earning partner at the lowest marginal tax rate rather than at the higher rate of the

228 band in which the income would fall when it was considered as a unity. For the withholding taxes of replacement incomes, a fixed percentage of unemployment benefits (10.09%), and of sickness and invalidity benefits (11.11%) is taken (FPS Finance, 2009a). For singles living entirely of unemployment benefits or couples in which only one partner receives unemployment benefits without any other source of income being present in the fiscal household, no withholding tax is due.

The tax allowance and tax credits are simulated in a fourth step. The tax allowance equals € 1,463.23 of non-taxed income for each partner in 2010, irrespective of whether the partner has earned income. Tax credits are tax advantages given according to the composition of the fiscal households. Depending on the source and level of income of the tax payers, the region they are living in, and the number of dependent children, other dependent persons (distinguishing between individuals aged under or over 65 years old), parents, partners, and disabled individuals present in the fiscal household, a specific amount can be subtracted from the simulated withholding tax. All tax credits are non-refundable with two exceptions. If the simulated withholding tax is smaller than the total sum of tax allowance and tax credits, in the first instance, the not used part of the tax credit for dependent children becomes refundable, bounded to € 390 per dependent child per year29. Thereafter, the tax credit for the self-employed with low income (€ 4,510 up to € 19,580 net taxable per year), is repayable up to € 610 per year30. The sum of the tax allowance and tax credits is subsequently subtracted from the simulated withholding tax on income from employment, self-employment and pensions from step three. For unemployment benefits, sickness and invalidity

29 Children with a recognised disability of at least 66% are counted twice as dependent children. 30 This tax credit (“belastingkrediet op lage activiteitsinkomsten”) also applies to tenured civil servants. Unfortunately, the data does not have enough information to distinguish tenured versus non-tenured civil servants, hence this tax credit can only be simulated for the self-employed.

229 benefits, no tax allowance and tax credits can be deducted within the withholding tax schedule.

Fifthly, the net disposable income at the individual level is simulated as the difference between net taxable income and the remaining withholding tax, augmented with the non-taxable income components one receives (i.e. disability benefits, social assistance and simulated child benefits). Summing it together for all members living at the same address, gives us the total net disposable income at the household level.

Finally, the total net disposable household income is equivalised using the OECD- modified equivalence scale (Hagenaars et al., 1994), assigning a weight of 1 to the first adult, 0.5 to all following adults aged 14 or older, and 0.3 to all children below the age of 14. Based on this, the poverty status is determined.

5.7.1.2. Income distribution of households with children, DWH LM&SP versus BE-SILC, 2010

Figure A5.1.1 compares the distribution of the equivalised net disposable household incomes for children minus 18 estimated using the DWH LM&SP to the one observed in BE-SILC 2011 (income year 2010). In general, the share of children at the bottom and at the top of the income distribution is higher according to BE-SILC than as simulated using the DWH LM&SP, while the reverse is true in the middle. In BE-SILC, 17.1% of children under 18 live in a household with an income up to 60% of the median equivalised net disposable income of households with children in the data, whereas this is only 7.5% according to the DWH LM&SP. 60.4% of children in BE-SILC and as much as 75.9% of children in the DWH LM&SP have a household income between 60% and 135% of their

230 median, while the respective shares equal 22.5% and 16.6% for household income exceeding 135% of the median.

Figure A5.1.1. Income distribution of equivalised net disposable household income for children under 18, relative to the median, DWH LM&SP versus BE-SILC, incomes 2010

7 1 0.9 6 0.8 5 0.7 0.6 4 0.5 3 0.4

2 0.3 0.2 Proportionof children (%) 1 0.1 0 0 5 20 35 50 65 80 95 110125140155170185200215230245260275290305320335350 % of median equivalised net disposable household income (own medians)

DWH all children <18 BE-SILC children <18 median

Source: own calculations based on DWH LM&SP (2010) and BE-SILC (2011). Note: each distribution is compared with its own median equivalised net disposable household income for children minus 18 (DWH LM&SP median = € 16187.57, BE-SILC median = € 19296.15).

Part of the observed differences can be accounted for by the components included in the income concepts of both data sources. The DWH LM&SP contains information on income from current and past employment, supplemented with non-taxable income from disability benefits, social assistance and child benefits. For 19% of children under 18, however, at least one of their parents has no known

231 income on any of these components (see also footnote 25). The BE-SILC is not constrained to (in Belgium) taxable income and additionally takes rental income, movable income, income transfers between households, study allowances, housing allowances as well as second and third pillar pension income into account.

Table A5.1.3 shows the weight of the four taxable income categories (i.e. income from employment (current and past), immovable property, movable property and other sources) in terms of the share of tax returns that declares these income categories as well as the share of the total net taxable income they represent. The figures are for Belgium. Almost all tax returns contain income from employment and 9% of the tax returns (additionally) has income from immovable property. Only a marginal share includes income from movable property (2%) or other sources (1%). Additionally, Table A5.1.3 shows that the lion’s share of net taxable income consists of employment income (98%). The shares of net taxable income from immovable property, movable property or other sources are very small.

Table A5.1.3. Shares of income categories in tax returns and in total net taxable income, incomes 2010, Belgium % of tax % of total net returns taxable income Total net taxable income from employment 99.44% 98.42% Total net taxable income from immovable property 9.09% 1.13% Total net taxable income from movable property 1.56% 0.30% Total net taxable income from other sources 1.25% 0.15% (e.g. income transfers between households) Source: Statistics Belgium (2010).

5.7.1.3. Children with a disability versus children without a disability

Figure A5.1.2 compares the estimated equivalised net disposable household income distributions of children with and without a disability to each other. It shows that children with a disability more often live in the middle of the income

232 distribution whereas children without a disability have a higher share located at the bottom. The differences between the two groups are small at the top of the income distribution.

Figure A5.1.2. Income distribution of equivalised net disposable household income for children under 18, with versus without a disability, relative to the median, Belgium, 2010

8 1

7 0.9 0.8 6 0.7 5 0.6 4 0.5

3 0.4 0.3 2

0.2 Proportionof children (%) 1 0.1 0 0 5 20 35 50 65 80 95 110125140155170185200215230245260275290305320335350 % of median equivalised net disposable household income

Children without a disability <18 Children with a disability <18 Median for all children <18

Source: own calculations based on DWH LM&SP (2010).

This is partially the result of the way childhood disability is measured: only children who receive the supplemental child benefit are identified as children with a disability. Figure A5.1.3 presents the income distributions for children with a disability when the main cash support systems provided to them are excluded, simultaneously and separately, from the net disposable household income (before it is equivalised).

233

Figure A5.1.3. Income distribution for children with a disability, with and without cash support, Belgium, 2010

9 1 8 -23% -10% -1% 0.9 7 0.8 0.7 6 0.6 5 0.5 4 0.4 3 0.3

2 0.2 Proportionof children (%) 1 0.1 0 0 5 20 35 50 65 80 95 110125140155170185200215230245260275290305320335350 % of median equivalised net disposable household income

CWD (<18) + CB + SCB + TC (all-in) CWD (<18) - CB - SCB - TC CWD (<18) + CB - SCB - TC CWD (<18) + CB + SCB - TC Median CWD (<18) + CB + SCB + TC (all-in) Median CWD (<18) - CB - SCB - TC Median CWD (<18) + CB - SCB - TC Median CWD (<18) + CB + SCB - TC

Source: own calculations based on DWH LM&SP (2010). Note: CWD = children with a disability, CB = regular child benefit, SCB = supplemental child benefit, TC = refundable tax credit for dependent children. Distributions are presented as percentages of the median for all children (including children without a disability) in the data (€16,187.57), the medians of each distribution are reported for comparative purposes.

When the regular child benefit, the supplemental child benefit, and the refundable tax credit for dependent children are not taken into account, the share of children with a disability living at the bottom (60% or less) increases from 3.8% to 31.6%, while the shares in the middle (65%-135%) and at the top (more than 135%) are

234 reduced by a quarter and by half respectively. When the cash support systems are stepwise included, it is clear that the regular child benefit (comparing the dotted and dashed line) has a great impact, but also the supplemental child benefit (comparing the dashed and double line) is important for household with children with a disability. The refundable tax credit matters as well (comparing the double to the solid line), especially at the bottom and in the middle of the income distribution.

235

5.7.2. Appendix 5.2 Descriptive information on variables of interest

Table A5.2.1. All children, poor versus non-poor children, Belgium, 2010

Children <18 Total Poor Non-poor Prevalence among total 22.1% 77.9% Household work intensity Very low (0-0.2) 12.4% 42.9% 3.7% Low (0.2-0.45) 4.2% 13.2% 1.7% Medium (0.45-0.55) 8.5% 14.7% 6.7% High (0.55-0.85) 21.6% 14.5% 23.6% Very high (0.85-1) 53.4% 14.7% 64.4% Household type Two parents 77.2% 42.7% 87.0% Single parent 22.2% 55.1% 12.9% Other 0.6% 2.2% 0.1% Number of children (<18) in household Less than three 73.7% 67.6% 75.5% Three or more 26.3% 32.4% 24.5% Parental education (highest level) Low-skilled 15.5% 37.6% 9.7% Medium-skilled 35.3% 44.4% 32.9% High-skilled 49.2% 18.0% 57.3% Country of birth parents At least one parent born in Belgium 88.0% 69.8% 93.0% At least one parent born in other EU27 country 3.5% 7.3% 2.5% Both parents born in non-EU27 country 8.5% 22.9% 4.5% Other household members with a disability None 96.8% 95.7% 97.1% At least one 3.2% 4.3% 2.9% Source: own calculations based on DWH LM&SP (2010) and Census (2011). Note: the shares in the total child population (column 2) are the weighted averages of the shares within the poor and non-poor subgroups of the total child population (columns 3 and 4 weighted by row 2). Bivariate statistics are shown meaning that the sample size can differ for each indicator. Results are comparable when only children with non-missing information on all variables of interest are taken into account, except for the poverty prevalence (20.7%), and the parents country of birth where the shares of foreign-born parents are smaller (among the total child population (BE 90.1%, EU27 2.7%, non-EU27 7.3%), this is more pronounced for the poor subgroup (BE 74.4%, EU27 5.8%, non-EU27 19.8%) than for the non-poor subgroup (BE 94.2%, EU27 1.9%, non-EU27 4.0%)).

236

Table A5.2.2. Children with and without a disability, poor versus non-poor children, Belgium, 2010

Children <18 With a disability Without a disability Non- Non- Subtotal Poor Subtotal Poor poor poor Prevalence among subtotal 15.6% 84.4% 22.2% 77.8% Household work intensity Very low (0-0.2) 22.2% 56.5% 15.8% 12.1% 42.7% 3.4% Low (0.2-0.45) 5.8% 14.6% 4.2% 4.2% 13.2% 1.6% Medium (0.45-0.55) 12.0% 11.4% 12.1% 8.4% 14.8% 6.5% High (0.55-0.85) 22.5% 9.4% 25.0% 21.6% 14.5% 23.6% Very high (0.85-1) 37.5% 8.1% 42.9% 53.7% 14.8% 64.9% Household type Two parents 69.2% 37.5% 75.1% 77.4% 42.7% 87.3% Single parent 30.2% 59.8% 24.8% 22.1% 55.0% 12.6% Other 0.6% 2.8% 0.2% 0.6% 2.2% 0.1% Number of children (<18) in

household Less than three 70.0% 63.0% 71.3% 73.8% 67.7% 75.6% Three or more 30.0% 37.0% 28.7% 26.2% 32.3% 24.4% Parental education (highest

level) Low-skilled 24.5% 45.8% 20.8% 15.3% 37.4% 9.5% Medium-skilled 41.6% 44.1% 41.2% 35.2% 44.4% 32.8% High-skilled 33.9% 10.1% 38.0% 49.5% 18.1% 57.8% Country of birth parents At least one parent born in 88.4% 74.9% 90.8% 88.0% 69.7% 93.1% Belgium At least one parent born in other 2.6% 4.8% 2.2% 3.5% 7.3% 2.5% EU27 country Both parents born in non-EU27 9.0% 20.3% 7.0% 8.5% 22.9% 4.4% country Other household members with

a disability None 82.4% 83.8% 82.2% 97.1% 95.9% 97.5% Yes, at least one 17.6% 16.2% 17.8% 2.9% 4.1% 2.5% Source: own calculations based on DWH LM&SP (2010) and Census (2011). Note: shares in the subtotal populations of children with and without a disability (columns 2 and 5) are the weighted averages of the shares within the poor and non-poor subgroups of these subpopulations (columns 3 and 4 for children with a disability, columns 6 and 7 for children without a disability, weighted by row2). Weighting column 2 and 5 by their prevalence in the total child population (2.1% and 97.9%), will give the total distribution presented in column 2 of Table A5.2.1. Bivariate statistics are shown meaning that the sample size can differ for each risk factor. Results are comparable when only children with non-missing information on all variables of interest are taken into account, except for the poverty prevalence (14.9% and 20.9% among children with and without a disability respectively), and the parents country of birth where the

237 shares of foreign-born parents are smaller (among children with a disability (BE 89.9%, EU27 2.3%, non-EU27 7.8%), this is more pronounced for the poor subgroup (BE 78.2%, EU27 4.3%, non-EU27 17.5%) than for the non-poor subgroup (BE 91.9%, EU27 1.9%, non-EU27 6.1%). Comparable among children without a disability (BE 90.1%, EU27 2.9%, non-EU27 7.3%), poor subgroup (BE 74.3%, EU27 5.8%, non-EU27 19.9%), non-poor subgroup (BE 94.2%, EU27 1.9%, non-EU27 3.9%)).

5.7.3. Appendix 5.3 Sensitivity check: applying the 50% and 70% at-risk-of- poverty threshold

5.7.3.1. 50% at-risk-of-poverty threshold

Table A5.3.1. 50% at-risk-of-poverty headcount ratio and average at-risk-of- poverty gap ratio among the poor, with and without cash support, children under 18, with versus without a disability, Belgium, 2010

With a Without a Children <18 Total disability disability 50% poverty headcount ratio ‒ CB ‒ SCB ‒ TC (scenario 1) 23.6% 33.4% 23.4% + CB ‒ SCB ‒ TC (scenario 2) 11.4% 15.1% 11.3% + CB + SCB ‒ TC (scenario 3) 11.2% 6.5% 11.3% + CB + SCB + TC (baseline, all-in) 9.1% 4.7% 9.2% Average 50% poverty gap ratio among the poor ‒ CB ‒ SCB ‒ TC (scenario 1) 28.5% 28.8% 28.4% + CB ‒ SCB ‒ TC (scenario 2) 19.3% 15.8% 19.4% + CB + SCB ‒ TC (scenario 3) 19.3% 15.6% 19.4% + CB + SCB + TC (baseline, all-in) 20.2% 16.9% 20.2% Source: own calculations based on DWH LM&SP (2010). Notes: CB = regular child benefit, SCB = supplemental child benefit, TC = refundable tax credit for dependent children. The poverty threshold is kept constant over all scenarios. The average poverty gap is calculated among those children who are at-risk-of-poverty in each scenario separately.

238

Table A5.3.2. Logistic regression on 50% child poverty risk, with and without cash support, odds ratios, Belgium, 2010

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Constant 0.012*** 0.009*** 0.009*** 0.011*** (0.001) (0.001) (0.001) (0.001) Child with a disability (CWD) 1.019n.s. 1.144n.s. 0.892n.s. 0.691** (0.103) (0.153) (0.136) (0.114) Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) 19.955*** 23.867*** 23.899*** 19.377*** (1.886) (2.771) (2.776) (2.524) Low (0.2-0.45) 27.567*** 25.740*** 25.758*** 23.148*** (3.037) (3.082) (3.084) (2.919) Medium (0.45-0.55) 8.129*** 6.568*** 6.574*** 6.800*** (0.670) (0.772) (0.773) (0.843) High (0.55-0.85) 2.699*** 2.173*** 2.174*** 2.182*** (0.189) (0.257) (0.257) (0.273) Household type (two parents ref.) Single parent 7.998*** 2.496*** 2.497*** 1.710*** (0.518) (0.204) (0.204) (0.157) Number of children (<18) in household (less than three ref.) Three or more 3.453*** 1.293*** 1.293*** 0.904n.s. (0.197) (0.089) (0.089) (0.065) Parental education (highest level) (high-skilled ref.) Low-skilled 2.674*** 2.062*** 2.062*** 1.866*** (0.217) (0.200) (0.200) (0.188) Medium-skilled 2.771*** 1.927*** 1.927*** 1.676*** (0.173) (0.165) (0.165) (0.149) Country of birth parents (Belgium ref.) EU27 1.294n.s. 1.673*** 1.673*** 1.794*** (0.194) (0.256) (0.256) (0.270) Non-EU27 2.859*** 1.742*** 1.742*** 1.738*** (0.265) (0.167) (0.167) (0.169) Other household members with a disability (none ref.) Yes, at least one 0.505*** 0.332*** 0.332*** 0.381*** (0.082) (0.056) (0.056) (0.381) Interaction x CWD Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) x CWD 0.946n.s. 0.628** 0.452*** 0.524**

239

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC (0.113) (0.097) (0.086) (0.114) Low (0.2-0.45) x CWD 0.720** 0.640** 0.455*** 0.464*** (0.104) (0.103) (0.086) (0.097) Medium (0.45-0.55) x CWD 0.746** 0.639** 0.427*** 0.413*** (0.081) (0.102) (0.083) (0.089) High (0.55-0.85) x CWD 0.861n.s. 0.771n.s. 0.507*** 0.562** (0.083) (0.126) (0.102) (0.124) Household type (two parents ref.) Single parent x CWD 0.924n.s. 1.009n.s. 0.471*** 0.556*** (0.078) (0.106) (0.060) (0.083) Number of children (<18) in household (less than three ref.) Three or more x CWD 1.192** 1.278** 1.135n.s. 1.260** (0.089) (0.112) (0.112) (0.140) Parental education (highest level) (high-skilled ref.) Low-skilled x CWD 1.183n.s. 1.000n.s. 0.889n.s. 0.948n.s. (0.126) (0.130) (0.141) (0.167) Medium-skilled x CWD 1.034n.s. 1.060n.s. 0.985n.s. 1.084n.s. (0.089) (0.124) (0.141) (0.173) Country of birth parents (Belgium ref.) EU27 x CWD 1.081n.s. 0.712n.s. 0.857n.s. 0.881n.s. (0.240) (0.149) (0.200) (0.219) Non-EU27 x CWD 0.833n.s. 0.975n.s. 0.913 n.s. 0.912n.s. (0.102) (0.116) (0.119) (0.128) Other household members with a disability (none ref.) Yes, at least one x CWD 1.555** 1.732** 1.118n.s. 0.844n.s. (0.270) (0.313) (0.219) (0.182) Region of residence (Flanders ref.) Brussels 1.672*** 1.169n.s. 1.165n.s. 1.155n.s. (0.155) (0.125) (0.126) (0.127) Wallonia 1.477*** 1.233** 1.226** 1.196** (0.083) (0.085) (0.085) (0.087) Model fit Log pseudolikelihood -483864.58 -354594.67 -350858.29 -334645.62 Pseudo R² 0.4523 0.3617 0.3611 0.2893 Prob > chi² 0.0000 0.0000 0.0000 0.0000 N 33883 33883 33883 33883 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: not living at-risk-of poverty is the baseline. *p<0.05, **p<0.01, ***p<0.001, n.s. not significant. Robust standard errors are in parentheses.

240

5.7.3.2. 70% at-risk-of-poverty threshold

Table A5.3.3. 70% at-risk-of-poverty headcount ratio and average at-risk-of- poverty gap ratio among the poor, with and without cash support, children minus 18, with versus without a disability, Belgium, 2010

With a Without a Children <18 Total disability disability 70% poverty headcount ratio ‒ CB ‒ SCB ‒ TC (scenario 1) 47.7% 60.1% 47.4% + CB ‒ SCB ‒ TC (scenario 2) 36.0% 46.6% 35.8% + CB + SCB ‒ TC (scenario 3) 35.7% 34.8% 35.8% + CB + SCB + TC (baseline, all-in) 35.4% 32.3% 35.5% Average 70% poverty gap ratio among the poor ‒ CB ‒ SCB ‒ TC (scenario 1) 31.3% 31.2% 33.7% + CB ‒ SCB ‒ TC (scenario 2) 23.0% 23.0% 23.0% + CB + SCB ‒ TC (scenario 3) 22.9% 23.0% 17.9% + CB + SCB + TC (baseline, all-in) 21.6% 21.7% 16.4% Source: own calculations based on DWH LM&SP (2010). Notes: CB = regular child benefit, SCB = supplemental child benefit, TC = refundable tax credit for dependent children. The poverty threshold is kept constant over all scenarios. The average poverty gap is calculated among those children who are at-risk-of-poverty in each scenario separately.

Table A5.3.4. Logistic regression on 70% child poverty risk, with and without cash support, odds ratios, Belgium, 2010

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Constant 0.070*** 0.041*** 0.041*** 0.042*** (0.003) (0.002) (0.002) (0.002) Child with a disability (CWD) 1.045n.s. 0.944n.s. 0.683*** 0.737*** (0.067) (0.070) (0.053) (0.056) Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) 8.968*** 14.862*** 14.876*** 13.972*** (1.077) (1.549) (1.550) (1.423) Low (0.2-0.45) 24.338*** 27.110*** 27.129*** 24.731*** (3.881) (3.449) (3.452) (3.090) Medium (0.45-0.55) 9.761*** 10.593*** 10.600*** 10.249*** (0.804) (0.791) (0.791) (0.761)

241

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC High (0.55-0.85) 3.615*** 3.324*** 3.324*** 3.294*** (0.177) (0.178) (0.791) (0.176) Household type (two parents ref.) Single parent 8.882*** 7.426*** 7.428*** 7.168*** (0.590) (0.454) (0.454) (0.435) Number of children (<18) in household (less than three ref.) Three or more 3.293*** 1.507*** 1.507*** 1.459*** (0.157) (0.076) (0.076) (0.073) Parental education (highest level) (high-skilled ref.) Low-skilled 4.816*** 3.769*** 3.768*** 3.782*** (0.359) (0.273) (0.273) (0.271) Medium-skilled 4.314*** 3.408*** 3.407*** 3.375*** (0.477) (0.168) (0.168) (0.166) Country of birth parents (Belgium ref.) EU27 1.747*** 1.772*** 1.774*** 1.750*** (0.280) (0.259) (0.259) (0.255) Non-EU27 4.137*** 3.428*** 3.431*** 3.304*** (0.117) (0.332) (0.333) (0.317) Other household members with a disability (none ref.) Yes, at least one 0.717** 0.561*** 0.561*** 0.541*** (0.117) (0.090) (0.090) (0.086) Interaction x CWD Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) x CWD 1.089n.s. 1.099n.s. 0.848n.s. 0.677** (0.162) (0.142) (0.105) (0.082) Low (0.2-0.45) x CWD 0.987n.s. 0.883n.s. 0.618** 0.545*** (0.211) (0.145) (0.096) (0.083) Medium (0.45-0.55) x CWD 0.800** 0.745** 0.520*** 0.495*** (0.085) (0.073) (0.051) (0.049) High (0.55-0.85) x CWD 0.849** 0.889n.s. 0.704*** 0.661*** (0.058) (0.067) (0.056) (0.053) Household type (two parents ref.) Single parent x CWD 0.980n.s. 1.037n.s. 0.519*** 0.446*** (0.089) (0.085) (0.041) (0.035) Number of children (<18) in household (less than three ref.) Three or more x CWD 1.118n.s. 1.080n.s. 1.300*** 1.192** (0.075) (0.073) (0.087) (0.079)

242

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Parental education (highest level) (high-skilled ref.) Low-skilled x CWD 1.013n.s. 1.073n.s. 0.884n.s. 0.849n.s. (0.100) (0.103) (0.085) (0.081) Medium-skilled x CWD 1.026n.s. 1.037n.s. 0.900n.s. 0.888n.s. (0.067) (0.072) (0.066) (0.065) Country of birth parents (Belgium ref.) EU27 x CWD 1.141n.s. 1.009n.s. 0.782n.s. 0.834n.s. (0.271) (0.220) (0.164) (0.168) Non-EU27 x CWD 0.727** 0.776n.s. 0.589*** 0.550*** (0.117) (0.102) (0.072) (0.066) Other household members with a disability (none ref.) Yes, at least one x CWD 1.377n.s. 1.476** 0.942n.s. 0.914n.s. (0.241) (0.254) (0.161) (0.155) Region of residence (Flanders ref.) Brussels 1.307** 1.577*** 1.564*** 1.578*** (0.108) (0.131) (0.130) (0.130) Wallonia 1.413*** 1.455*** 1.450*** 1.434*** (0.064) (0.069) (0.069) (0.068) Model fit Log pseudolikelihood -706420.20 -642854.89 -643268.14 -648925.71 Pseudo R² 0.3859 0.4004 0.3985 0.3913 Prob > chi² 0.0000 0.0000 0.0000 0.0000 N 33883 33883 33883 33883 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: not living at-risk-of poverty is the baseline. *p<0.05, **p<0.01, ***p<0.001, n.s. not significant. Robust standard errors are in parentheses.

243

5.7.4. Appendix 5.4 Sensitivity check: logistic results without applying the population weight

Table A5.4.1. Logistic regression on 60% child poverty risk without population weight, with and without cash support, odds ratios, Belgium, 2010

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Constant 0.031*** 0.018*** 0.019*** 0.020*** (0.002) (0.001) (0.001) (0.001) Child with a disability (CWD) 0.929n.s. 0.976n.s. 0.734** 0.749** (0.074) (0.094) (0.077) (0.079) Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) 14.142*** 23.785*** 24.735*** 21.299*** (1.456) (2.189) (2.285) (1.918) Low (0.2-0.45) 25.093*** 29.169*** 29.986*** 26.303*** (3.127) (3.148) (3.238) (2.795) Medium (0.45-0.55) 9.778*** 8.654*** 8.921*** 8.766*** (0.742) (0.706) (0.726) (0.711) High (0.55-0.85) 3.006*** 2.611*** 2.622*** 2.533*** (0.164) (0.183) (0.183) (0.178) Household type (two parents ref.) Single parent 8.454*** 5.750*** 5.803*** 5.354*** (0.523) (0.360) (0.363) (0.333) Number of children (<18) in household (less than three ref.) Three or more 3.499*** 1.436*** 1.437*** 1.279*** (0.176) (0.082) (0.082) (0.074) Parental education (highest level) (high-skilled ref.) Low-skilled 3.776*** 2.652*** 2.640*** 2.648*** (0.276) (0.211) (0.210) (0.210) Medium-skilled 3.450*** 2.703*** 2.680*** 2.690*** (0.176) (0.167) (0.165) (0.166) Country of birth parents (Belgium ref.) EU27 1.453** 1.657*** 1.716*** 1.642*** (0.214) (0.244) (0.255) (0.244) Non-EU27 3.575*** 2.732*** 2.792*** 2.662*** (0.354) (0.252) (0.258) (0.244)

244

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Other household members with a disability (none ref.) Yes, at least one 0.503*** 0.391*** 0.393*** 0.362*** (0.079) (0.065) (0.065) (0.060) Interaction x CWD Household work intensity (very high (0.85-1) ref.) Very low (0-0.2) x CWD 1.096n.s. 0.908n.s. 0.485*** 0.483*** (0.142) (0.107) (0.061) (0.062) Low (0.2-0.45) x CWD 0.841n.s. 0.832n.s. 0.506*** 0.514*** (0.137) (0.118) (0.073) (0.075) Medium (0.45-0.55) x CWD 0.794** 0.700** 0.477*** 0.487*** (0.079) (0.076) (0.058) (0.060) High (0.55-0.85) x CWD 0.942n.s. 0.932n.s. 0.759** 0.728** (0.072) (0.091) (0.085) (0.086) Household type (two parents ref.) Single parent x CWD 1.085n.s. 0.981n.s. 0.429*** 0.396*** (0.091) (0.080) (0.037) (0.035) Number of children (<18) in household (less than three ref.) Three or more x CWD 1.079n.s. 1.128n.s. 1.268** 1.132n.s. (0.074) (0.084) (0.096) (0.087) Parental education (highest level) (high-skilled ref.) Low-skilled x CWD 1.099n.s. 1.123n.s. 1.022n.s. 0.905n.s. (0.107) (0.118) (0.114) (0.104) Medium-skilled x CWD 1.028n.s. 1.005n.s. 0.930n.s. 0.869n.s. (0.074) (0.087) (0.088) (0.085) Country of birth parents (Belgium ref.) EU27 x CWD 1.418n.s. 1.077n.s. 0.870n.s. 0.877n.s. (0.312) (0.235) (0.176) (0.178) Non-EU27 x CWD 0.773n.s. 0.832n.s. 0.599*** 0.607*** (0.105) (0.100) (0.070) (0.071) Other household members with a disability (none ref.) Yes, at least one x CWD 1.818*** 1.503** 1.131n.s. 1.161n.s. (0.307) (0.268) (0.201) (0.206) Region of residence (Flanders ref.) Brussels 1.528*** 1.539*** 1.163** 1.133n.s. (0.098) (0.101) (0.077) (0.076) Wallonia 1.406*** 1.436*** 1.209*** 1.196*** (0.048) (0.054) (0.046) (0.047)

245

Scenario 1 Scenario 2 Scenario 3 Baseline Children <18 ‒ CB ‒ + CB ‒ + CB + + CB + SCB ‒ TC SCB ‒ TC SCB ‒ TC SCB + TC Model fit Log pseudolikelihood -12888.93 -11012.42 -10718.92 -10477.67 Pseudo R² 0.4334 0.4332 0.3603 0.3369 Prob > chi² 0.0000 0.0000 0.0000 0.0000 N 33883 33883 33883 33883 Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: not living at-risk-of poverty is the baseline. *p<0.05, **p<0.01, ***p<0.001, n.s. not significant. Robust standard errors are in parentheses.

246

5.7.5. Appendix 5.5 Further exploration of regular child benefit, Belgium, 2010

Figure A5.5.1. Marginal effect of child’s disability status on predicted 60% income poverty risk (left axis) and mean equivalised cash support among the poor by child benefit component (right axis), in different scenarios with and without cash support, children under 18

30% € 400 € 350 25% € 300 20% € 250 15% € 200 € 150 10% € 100 5%

€ 50 Predicted child povertyrisk

0% € - (per month) Children Children Children Children Children Children Children Children without with a without with a without with a without with a a disability a disability a disability a disability disability disability disability disability - CB - SCB - TC + CB - SCB - TC + CB + SCB - TC + CB + SCB + TC

Scenario 1 Scenario 2 Scenario 3 Baseline Meanequivalised cashsupport among the poor

Mean equivalised refundable tax credit at household level Mean equivalised supplemental child benefit at household level Mean equivalised selective child benefit at household level Mean equivalised age-related child benefit at household level Mean equivalised universal child benefit at household level Income poverty risk children without a disability Income poverty risk children with a disability

Source: own calculations based on DWH LM&SP (2010) and Census (2011). Notes: CB = regular child benefit, SCB = supplemental child benefit, TC = refundable tax credit for dependent children. Predicted child poverty risks are shown as marginal effects of the child’s disability status (left axis), at mean values for the remaining variables of Table 5.4. Mean equivalised cash support is shown for poor children with and without a disability, by component (and by component of the regular child benefit) and population weighted (right axis). The population of poor children changes in each scenario.

247

CONCLUSION

The main purpose of this thesis is to unravel the poverty puzzle among children with a disability by taking the interplay into account between childhood disability, parental employment, social background and the receipt of targeted cash support. To do so, I draw on a case study of Belgium making use of unique and large-scale administrative data.

Previous research conducted within the literature on disability and child development has found a clear link between childhood disability and child poverty in a wide range of countries. However, this strand of the literature is disconnected from the social policy literature on child poverty. Across countries, the currently dominant policy strategies to combat child poverty are rooted in the social investment paradigm and focus on integrating parents into paid employment. But, for families of children with a disability, such a work strategy appears to be problematic. This is related to the fact that these families (1) need to provide more care which directly impedes their employment participation, and (2) often have a disadvantaged social background which affects their employment opportunities and poverty risk in its own right. Until now, we lacked insight into what precise role parental employment patterns play in explaining the poverty risk of children with a disability and to what extent targeted cash support for these children reduces their poverty risk. By scrutinising how the interplay between childhood disability, parental employment, social background and the receipt of targeted cash support affects the poverty risk of children with a disability, I unite two strands of the literature that generally operate independently from each other.

The three main conclusions that can be derived from the previous chapters are set out in the next section. I then briefly summarise the unique contributions of this

249 thesis. I further explain in what way the analyses are limited and what remains to be explored in future research. Finally, I formulate three interrelated recommendations for policy makers to guide (further) poverty reduction among families of children with a disability.

1. What have we learned …

1.1. … about the social background of children with a disability?

The first conclusion that emerges from this thesis is that childhood disability often overlaps with social disadvantages in Belgium (see Chapter 3). This is true for three of the four disadvantaged social background indicators that are taken into consideration. Children with a disability more frequently live together with single parents, parents holding lower educational qualifications, and other households members with a disability than children without a disability. This is in line with research conducted in other countries, see for example Bauman et al. (2006) for the US, and Blackburn et al. (2010) for the UK. For the fourth indicator of a disadvantaged social background that is of interest in the Belgian context, that is the migration background of the parents, no overlap with childhood disability is found, though.

1.2. … about the labour market participation of parents with children with a disability?

The second conclusion that can be drawn from the analyses conducted here is that parents of children with a disability have lower levels of employment than parents of children without a disability (see Chapter 3 and 4). This employment gap is

250 intensified when the severity of the child’s disability increases (see Chapter 3). Moreover, it is found to be stronger for mothers than for fathers (see Chapter 4). This corroborates the findings reported in previous studies (Brown & Clark, 2017; Stabile & Allin, 2012).

Once an intersectional approach is adopted to disentangle the employment gap among families of children with a disability, it becomes clear that the gap is only partially explained by their disadvantaged social background (see Chapter 3). Parental employment participation is lower in all families of children with a disability. But, for some of the disadvantaged social categories, childhood disability and social disadvantage reinforce each other in the risk of lower parental labour market participation. At the intersection of childhood disability with single parenthood, lower parental educational qualifications, and the presence of siblings with a disability, the employment gap is intensified. For parents with a migration background, parents who have a disability themselves or who live together with other adults with a disability, no reinforcement could be discerned. Hence, not all sources of disadvantage strengthen one another.

Moreover, in a cross-country comparative study of Belgium and Norway, it is shown that the size of the employment gap depends on the institutional context of the welfare state (see Chapter 4). In these two countries, an employment gap is found that is stratified in terms of both gender and social background. Interestingly, however, the results show that the employment gap is significantly larger in Belgium than in Norway. Norway’s long-standing tradition of full employment for both men and women, and its family policies that promote a dual earner-dual carer family model for all, seem to pay off.

251

1.3. … about the income poverty risk of children with a disability?

To understand the risk of income poverty among children with a disability, both the labour market participation of the parents as well as the family’s social background offer insight (see Chapter 5). This is in line with general child poverty research (see Chapter 1). But, the third and most surprising conclusion that arises from this thesis is that, even when their parents’ lower levels of employment and disadvantaged social background are taken into account, children with a disability run a lower risk to live in income poverty than children without a disability in Belgium (see Chapter 5). This counterintuitive result can be explained by the targeted cash support families of children with a disability receive: the supplemental child benefit. In fact, when families would only receive the general cash support for children, the regular child benefit, children with a disability would run an equally high income poverty risk than children without a disability once their family’s labour market participation and social background are controlled for. The supplemental child benefit for children with a disability has a strong poverty-reducing impact as it mainly ends up with families who are more likely to live in income poverty in the first place: i.e. families with lower levels of labour market participation and families with a disadvantaged social background. So, also the targeted cash support and who actually receives it are key to disentangle their poverty puzzle.

All in all, a social investment inspired strategy to fight child poverty by integrating parents into paid employment might be little helpful for families of children with a disability, while targeted income protection policies are key. However, it should be stressed that the supplemental child benefit is subject to the issue of non-take- up, which jeopardises its full poverty-reducing potential. This thesis reveals that at least 10% of children with a disability in Flanders do not take-up the

252 supplemental child benefit, mainly children with autism spectrum disorder, intellectual and psychological disorders (see Chapter 2). This non-take-up is the result of insufficient information provision about the benefit by frontline organisations and doctors, the complexity of the application process and the construction of the scale to assess the eligibility. I elaborate on this in the limitations section (see Section 3).

Moreover, it is important to bear in mind that an income-based poverty indicator is not necessarily a good representation of the standard of living for families of children with a disability as they face higher medical and care costs incurred by the child's disability (e.g. Stabile & Allin, 2012). In other words, these families are confronted with an extra burden on the household budget. Hence, it might be the case that a family with a child with a disability and an income above the poverty threshold has in fact a lower standard of living than an income-poor family without a child with a disability. I return to this point in Section 3 as well.

2. What is new?

With a case study on Belgium using large-scale administrative data, this thesis has given us insight into an hitherto overlooked issue in the social policy literature on child poverty: is a work strategy desirable when families face increased care needs incurred by the child’s disability? This set-up proved to be an excellent way to unravel the complex interdependencies between childhood disability, parental employment, social background and the receipt of targeted cash support in studying income poverty among children with a disability.

In doing do, I make five contributions to the literature. First, in order to properly understand parental employment in families of children with a disability,

253 childhood disability should not be investigated in isolation but at the intersection with social disadvantages that are related to the risk of lower parental employment. Second, concerning parental employment in families of children with a disability, the institutional context of the country in which these families live matters as well, demonstrated in this thesis by comparing Belgium to the dual earner-dual carer welfare state of Norway. Third, corroborating the findings in general child poverty research, both parental employment patterns as well as the social background of the family play a crucial role in understanding income poverty among children with a disability. Here too, an intersectional approach is indispensable since the strength of some of these explanatory factors differs between children with and without a disability. Fourth, also the receipt of cash support targeted at children with a disability is essential to disentangle their poverty puzzle. Yet, the poverty- reducing impact varies by the type of cash support system and depends on who actually receives the benefit. Finally, this thesis shows that the poverty-reducing potential is jeopardised by the issue of non-take-up: for the main targeted cash support system, insight is provided into how many, which and why children with a disability miss out on the benefit.

3. What remains to be done?

The conducted research is limited in five ways, indicating important avenues for future research.

First and foremost, the data at hand only allow to operationalise childhood disability as those children who are administratively recognised for the supplemental child benefit. This benefit is not only granted to children with a disability, but to a broader group of children with increased care needs. Yet, distinguishing between different types of increased care needs is not possible in

254 the available data. It might be interesting for future research endeavours to examine whether the same patterns hold for different disability types (e.g. physical disabilities, intellectual disabilities, behavioural disorders, or developmental disorders) or for another group of children with increased care needs such as young cancer patients.

More importantly, the supplemental child benefit suffers from non-take-up. By crossing administrative recognition data for this benefit with the customer database of the Flemish Agency for Persons with a Disability (FAPD), the agency responsible for in-kind care support and additional financial support, the non-take- up rate of the supplemental child benefit is estimated to be at least 10% (see Chapter 2). This ‘at least’ is not without meaning, it could very well be that there is still a group of children with a disability who are missed in either administrative data source. For example, children who have not (yet) started either of the two application procedures, children whose families have not (yet) accepted the disability being an issue, or children who are enrolled in inclusive or special education without an administrative recognition of their disability for the supplemental child benefit nor for the FAPD.

Previous research based on the Flemish Families and Care Survey allows to complement the conclusions drawn from this thesis. This survey was collected in Flanders over the course of the 2004 – 2005 school year and contains socioeconomic and care information of almost 2,800 households with children below the age of 16. It also includes an oversampling of children with a disability by a random selection from the FAPD customer database in that year (n = 458). This subsample permits to apply a broader disability definition as also children with a disability who are not recognised for the supplemental child benefit are included. The results regarding the family’s social background and labour market

255 participation when children have a disability presented in Chapter 3 and 4 are in line with what is reported by Debacker (2007), Sebrechts and Breda (2012), and Van Landeghem et al. (2007). Hence, it is safe to assume that the first and second main conclusion of this thesis can be extended to children with a disability who have not obtained the administrative recognition for the supplemental child benefit: (1) children with a disability often belong to families with a disadvantaged social background, and (2) their parents (mainly mothers) work less than the parents of children without a disability which is not solely due to their disadvantaged social background, however for some of the social disadvantages, childhood disability and social disadvantage strengthen one another in the risk of lower parental employment.

For the income poverty risk, however, a different operationalisation of children with a disability would probably not lead to the same conclusion as it is precisely the supplemental child benefit that is indispensable in protecting their household’s income. In fact, Debacker (2007) shows that, based on the Flemish Families and Care Survey, the income poverty risk is higher for children with a disability than for those without. Also the EU-SILC 2017 ad-hoc module on children’s health and activity limitations indicates this: in the majority of European countries (including Belgium), children experiencing moderate to severe activity limitations are overrepresented among income-poor families (EUROSTAT, 2020c). So, the income poverty risks for children with a disability presented in this thesis are probably lower-bound estimates. What role the family’s labour market participation, social background and receipt of other cash support systems play in explaining the income poverty risk of children with a disability who do not take- up the supplemental child benefit remains to be unravelled.

256

Second, the employed income-based poverty indicator disregards the costs families face. It only takes the family’s income into account. Also, the way in which these incomes are made comparable across households (i.e. by applying the OECD-modified equivalence scale proposed by Hagenaars et al. (1994)) indirectly assumes that households with the same equivalised income can translate that income into the same standard of living. For families of children with a disability, such an income-based poverty indicator probably masks important aspects of children’s living conditions as their families are confronted with increased out-of- pocket costs incurred by the child’s disability. It could very well be that a family with a child with a disability and an income above the poverty threshold has actually a lower standard of living than an income-poor family without a child with a disability as the former family has to cover for the child’s increased medical and care needs. It is likely that this also underestimates the poverty risks presented here. To properly understand the standard of living among families of children with a disability, additional indicators are needed which take the costs incurred by the child’s disability into account. In order to achieve this, information on what these families actually spend (i.e. expenditure-based indicators) or (minimally) need (e.g. reference budget indicators) is required. This is an important future research avenue. Similar attempts have been made for adults with a disability and have (1) included the additional expenditures in the poverty estimates and (2) evaluated to what extent these expenditures are met by (governmental) support (e.g. Loyalka et al., 2014; Mont & Cuong, 2011; Morris & Zaidi, forthcoming; Palmer et al., 2019; Saunders, 2007; Zaidi & Burchardt, 2005), but for children this is yet to be done.

Third, the use of education and care services, both general and disability-specific could not be taken into account in the analyses. These care support services are

257 presumably crucial in balancing work and care obligations for families of children with a disability. With respect to general childcare services, the available research for Flanders points out, however, that its use remains limited among families of children with a disability (Child & Family, 2018; Van Landeghem et al., 2007). This result is mainly driven by the fact that children from a disadvantaged social background make less use of these childcare services and that children with a disability more often live in these disadvantaged households, than by the mere presence of a child with a disability (Sebrechts & Breda, 2011). The EU-SILC 2017 ad-hoc module on children’s health and activity limitations due to health problems can potentially be a useful tool to further explore the social stratification of general childcare use among children who are limited in their activities and how its use is related to parental employment. This data source is subject to small sample sizes though: in Belgium, only 139 out of almost 2,600 children minus 16 years old experience moderate to severe activity limitations, corresponding to 5.0%; in the 28 countries that belonged to the European Union in 2017 taken together, this is the case for 4,376 out of 90,832 children aged 0 to 15, corresponding to 4.7% (EUROSTAT, 2020c).

With respect to the disability-specific care support services, both the (semi-) residential or ambulatory care arrangements as well as the support provided within a special or inclusive educational setting are important for families of children with a disability. For Flanders, however, less than half of the children who are included in this thesis’ disability definition are also recognised at the FAPD and use any of its subsidised care provisions or receive additional financial support (see Chapter 231). Building on Van Landeghem et al. (2007)’s evidence that even in families who combine the two collective types of care arrangements (i.e. special

31 This can be derived from Table 2.3: (b) / (e) *100 or 15,070 / 32,349 *100 = 46.6%

258 education and (semi-)residential care) nearly half the mothers do not work, it cannot necessarily be expected that the use of disability-specific care services allows parents to engage in paid employment. And, for those who do not make use of these services, labour market participation might be even more difficult to achieve. Future research should look into how the family’s social background is related to the uptake of these disability-specific care services, whether its use allows parents to engage in paid employment, and thereby, indirectly, lower their family’s poverty risk. Besides, whether personal care budgets should be counted as household income and, if so, what poverty-reducing effect they have for children with a disability remain open questions that should be examined in future research as well.

These future research directions are especially relevant given the changing policy logic in the Flemish disability support system from a care-centred to a support- centred approach (Lebeer et al., 2018; Roets et al., 2020). This is reflected in a shift from supply-side financing to demand-side financing: persons with a disability are regarded as competent citizens, consumers in fact, who purchase their own care and support through vouchers and personal care budgets at the general and disability-specific care support services that (should) function in a market-oriented way. For children, this transition is not yet finalised32. At the same time, a shift is taking place towards welfare pluralism and subsidiarity: the prime responsibility for care and support provision is transferred (back) to persons with a disability and their informal networks, formal care and support provided by professionals in the general and disability-specific services only step in afterwards. And so, this increasing pressure on the informal networks to take on the care for

32 An evaluation tool to assess the children’s care and support needs, which is required to determine the budget they are entitled to, needs to be (further) developed (see Maes, Maljaars, & Noens, 2017 for some critical reflections).

259 persons with a disability (Dermaut, Schiettecat, Vandevelde, & Roets, 2020), in this case parents of children with a disability, may be at odds with the prevailing solution to child poverty through integrating parents into the labour market. Related to this, it should be acknowledged that some parents, and parents of children with a disability in particular, choose not to (fully) participate in paid employment as they prefer to care for their children themselves. These issues should be addressed in future research.

Fourth, the cross-sectional nature of the data only allows to assess correlations, not causations. To unpack the causal mechanisms between childhood disability, social disadvantages, parental labour market participation and the family’s income poverty risk, and whether, and if so how, the receipt of targeted cash support alters these relationships, the exploration of longitudinal data and the application of methods of causal inference are needed. Longitudinal data would also make it possible to examine trajectories of parental employment, income poverty and care support use among families of children with a disability. These are potentially important research strands for the future.

Finally, the comparative part of the thesis is limited to two countries, hence more cross-country comparative research is desirable, both in terms of parental employment as well as in terms of income poverty among families of children with a disability. In order to do that, sufficient, reliable and comparable data are required. This would allow to investigate whether the same conclusions apply within other countries, whether between-country differences arise, and if so what drives them. It has been established that Belgium performs only mediocre from a European comparative perspective when it comes to the income protection and poverty reduction among families with children (see Chapter 1; Vandenbroucke & Vinck, 2013; Van Lancker & Van Mechelen, 2015). But, how well or poorly

260

Belgium performs for families with children with a disability compared to other countries is still unexplored territory. Perhaps the EU-SILC 2017 ad-hoc module can be instrumental for this, though in-depth analyses such as those carried out in this thesis most likely need larger sample sizes.

4. What can policy makers do?

It should be clear by now that a social investment rooted strategy to combat child poverty by integrating parents into the labour market might be little helpful for families of children with a disability, and that income protection policies play a crucial role. So, what can policy makers do to make (further) progress in the fight against poverty among these families? The analyses presented in this thesis hint at three mutually reinforcing policy implications.

First, as families of children with a disability face an additional challenge to combine work and care, irrespective of their disadvantaged social background, integrating them into the labour market will be hard to realise without providing increased support on multiple fronts. To start with, the access to high-quality formal care services adjusted to the children’s care needs should be improved, allowing their parents to outsource (part of) the care. This could be helpful for all parents of children with a disability to resume, retain or reinforce their labour market participation and, hence, reduce the employment gap they have compared to parents of children without a disability. But, since the employment gap is stronger for some disadvantaged social categories, the care services should especially be accessible for these groups as well. This will require tackling the social inequalities in the use of formal care services (e.g. Gambaro et al., 2014; Van Lancker, 2013). Besides, improved workplace support could be beneficial too. Providing parents with increased working-time flexibility will enable them to

261 better reconcile their work and increased care responsibilities, for example by being able to organise their own working time, having the option to work from home, or taking parental leave (Brown & Clark, 2017; Crettenden et al., 2014).

Second, increased support on its own will not be enough for families of children with a disability. Since childhood disability often overlaps with social disadvantages that largely determine the parents’ labour market participation in their own right, activating individuals in these disadvantaged categories will be crucial too. Hence, labour market opportunities should be improved for single parents, people holding lower educational qualification, people with a migration background and people with a disability. The same applies to mothers. If welfare states achieve this, both families of children with a disability as those without will benefit from it. In this respect, lessons can be learned from the equality promoting employment and family policies of the Norwegian welfare state.

Third, even though the lower levels of parental employment and the disadvantaged social background do not sound promising in terms of poverty outcomes for children with a disability, their family incomes are protected because of the targeted cash support they receive. The supplemental child benefit has a strong poverty-reducing impact for children with a disability, bringing their income poverty risk below the level of children without a disability. However, a substantial group of children with a disability does not receive this benefit. So, further progress could be made if the issue of non-take-up would be addressed, mainly for the most vulnerable children. In order to do this, action must be taken again on multiple fronts. To start with, the information provision about the benefit’s existence and eligibility criteria to frontline organisations, doctors and parents should be improved. This information could even be provided in a proactive way: for example, when a child is enrolled in special education,

262 hospitalised for a long time, recognised at the FAPD for in-kind care support or additional financial support, diagnosed by a specialist, or followed up by a GP, information on the supplemental child benefit should be given by default. Additionally, a revision of the benefit scale seems needed as the criteria used are still to a great extent medical and, on top of that, leave too much room for interpretation by the control doctor. This is unfavourable for children with less “visible” disabilities such as autism spectrum disorder, intellectual and psychological disorders. The International Classification of Functioning, Disability and Health for Children and Youth can and should serve as a framework for this revision. Finally, coherence in the disability policy landscape will without a doubt be beneficial too. This could involve the centralisation of medical reports, provided that they are recent and comprehensive, but also a thoughtful alignment of the application procedures or even of the assessment tools. With the recent decentralisation of the child benefit system to the regional level (see Béland & Lecours, 2018 for more information), steps were taken towards more coherence in the policy landscape in Flanders, but it remains to be seen whether this is indeed a simplification in practice.

Briefly put, despite the lower labour market participation and disadvantaged social background of families with a child with a disability, Belgium performs better in terms of income poverty among these children than among children without a disability through its system of targeted cash support. But, room for improvement remains. Only if action is simultaneously taken in social investment and social protection policies, further progress can be made.

263

REFERENCES

Albertini Früh, E., Lidén, H., Gardsjord, R., Aden, P., & Kvarme, L. G. (2016). Innvandrerfamilier med barn med spesielle behov – mødres tilknytning til arbeidslivet. Søkelys på arbeidslivet, 33(3). Anderson, D., Dumont, S., Jacobs, P., & Azzaria, L. (2007). The personal costs of caring for a child with a disability: A review of the literature. Public Health Reports, 122, 3-16. Atkinson, A.B. (2015). Inequality: What can be done? Cambridge, USA: Harvard University Press. Bahle, T., & Krause, P. (2017). Child poverty during the Recession in Germany. In B. Cantillon, Y. Chzhen, S. Handa & B. Nolan (Eds.), Children of austerity: The impact of the Great Recession on child poverty in rich countries (pp. 56-93). Oxford, UK: The United Nations Children’s Fund and Oxford University Press. Banks, L. M., Kuper, H., & Polack, S. (2017). Poverty and disability in low- and middle-income countries: A systematic review. PLoS ONE, 12(12), 1-19. Bauman, L. J., Silver, E. J., & Stein, R. E. K. (2006). Cumulative social disadvantage and child health. Pediatrics, 117(4), 1321-1328. Becker, G. S. (1985). Human capital, effort, and the sexual division of labor. Journal of Labor Economics, 3(1), S33-S58. Becker, G. S. (1991). A treatise on the family. Cambridge, USA: Harvard University Press. Béland, D., & Lecours, A. (2018). Federalism, policy change, and social security in Belgium: Explaining the decentralization of family allowances in the sixth state reform. Journal of European Social Policy, 28(1), 55-69. Bellani, L., & Bia, M. (2017). The impact of growing up poor in Europe. In A. B. Atkinson, A.-C. Guio, & E. Marlier (Eds.), Monitoring social inclusion in Europe (pp. 449-461). Luxembourg: Publication Office of the European Union. Biggeri, M., & Mehrotra, R. (2011). Child poverty as capability deprivation: how to choose dimensions of child wellbeing and poverty? In M. Biggeri, J.

265

Ballet, F. Comim (Eds.), Children and the Capability Approach (pp. 46- 75). New York, USA: Palgrave Macmillan. Bilge, S. (2010). Recent feminist outlooks on intersectionality. Diogenes 225, 57(1), 58-72. Blackburn, C. M., Spencer, N. J., & Read, J. M. (2010). Prevalence of childhood disability and the characteristics and circumstances of disabled children in the UK: secondary analysis of the Family Resources Survey. BMC Pediatrics, 10(21), 1-12. Blackburn, C. M., Spencer, N., J., & Read, J. M. (2013). Is the onset of disabling chronic conditions in later childhood associated with exposure to social disadvantage in earlier childhood? A prospective cohort study using the ONS longitudinal study for England and Wales. BMC Pediatrics, 13, 101. Boat, T. F., & Wu, J. T. (Eds.). (2015). Mental disorders and disabilities among low-income children. Washington, DC, USA: The National Academies Press. Bolte, G., Tamburlini, G., & Kohlhuber, M. (2010). Environmental inequalities among children in Europe: Evaluation of scientific evidence and policy implications. European Journal of Public Health, 20(1), 14-20. Bonoli, G. (2012). Active labour market policy and social investment: A changing relationship. In N. Morel, B. Palier, & J. Palme (Eds.), Towards a social investment welfare state? Ideas, policies and challenges (pp. 181-204). Bristol, UK: Policy Press. Bouckaert, N., & Schokkaert, E. (2011). A first computation of non-take-up behaviour in the 'leefloon'. Flemosi Discussion Paper 6 (pp. 19). Leuven, Belgium: Flemosi. Bradshaw, J., Chzhen, Y., Main, G., Martorano, B., Menchini, L., & de Neubourg, C. (2012). Relative income poverty among children in rich countries. Innocenti Working Papers No. 2012-01. Florence, Italy: UNICEF, Innocenti Research Centre. Brekke, I., Albertini Früh, E., Kvarme, L. G., & Holmstrøm, H. (2017). Long-time sickness absence among parents of pre-school children with cerebral palsy, spina bifida and down syndrome: a longitudinal study. BMC Pediatrics, 17(26), 1-7.

266

Brekke, I., Evensen, M., & Kaldager Hart, R. (2020). Uptake of attendance benefit for children with a disability in Norway: Impact of socioeconomic status and immigrant background. Scandinavian Journal of Disability Research, 22(1), 127-136. Brekke, I., & Nadim, M. (2016). Gendered effects of intensified care burdens: employment and sickness absence in families with chronically sick or disabled children in Norway. Work, Employment and Society, 31(3), 1-18. Brooks-Gunn, J., & Duncan, G. J. (1997). The effects of poverty on children. The Future of Children, 7(2), 55-71. Brown, T. J., & Clark, C. (2017). Employed parents of children with disabilities and work family life balance: a literature review. Child & Youth Care Forum, 46(6), 857-876. Bulté, S., & Struyven, L. (2013). De Belgische arbeidsmarkt na 5 jaar crisis: nooit eerder werden minder nieuwe jobs gecreëerd. DynaM Analyse December Leuven: HIVA. Burkhauser, R. V., Daly, M. C., & Ziebarth, N. R. (2016). Protecting working-age people with disabilities: experiences of four industrialized nations. Journal for Labour Market Research, 49(4), 367-386. Byrne, B. (2014). Child poverty and disability. In Child Poverty Alliance (Ed.) Beneath the surface. Child poverty in Northern Ireland (pp.35-48). Belfast, UK: Child Poverty Alliance. Cantillon, B. (1999) De welvaartsstaat in kering. Kapellen, Belgium: Pelckmans. Cantillon, B. (2016). De staat van de welvaartsstaat. Leuven, Belgium: Acco. Cantillon, B., Chzhen, Y., Handa, S., & Nolan, B. (Eds.). (2017). Children of austerity: Impact of the great recession on child poverty in rich countries. New York, USA: UNICEF and Oxford University Press. Cantillon, B., & Marchal, S. (2016). Decent income for the poor: Which role for Europe?. CSB Working Paper No.16/01. Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Cantillon, B., & Van Lancker, W. (2013). Three shortcomings of the social investment perspective. Social Policy and Society, 12(4), 553-564.

267

Cantillon, B., & Vandenbroucke, F. (Eds.). (2014). Reconciling work and poverty reduction: How successful are European welfare states? New York, USA: Oxford University Press. Case, A., Lubotsky, D., & Paxson, C. (2002). Economic status and health in childhood: The origins of the gradient. American Economic Review, 92(5), 1308-1334. Centrale Raad voor het Bedrijfsleven (2014). De versterkte degressiviteit van de Belgische werkloosheidsuitkeringen: Effecten op de financiële vallen in de werkloosheid en op de inkomenspositie van de werklozen. Documentatienota, CRB 2014‐0264. Brussels, Belgium: Centrale Raad voor het Bedrijfsleven. Chen, W.-H., & Corak, M. (2008). Child poverty and changes in child poverty. Demography, 45(3), 537-553. Child & Family [Kind en Gezin]. (2018). Het kind in Vlaanderen 2018. Brussels, Belgium: Kind en Gezin. Cho, S., Crenshaw, K. W., & McCall, L. (2013). Toward a field of intersectionality studies: Theory, applications, and praxis. Signs, 38(4), 785-610. Choo, H. Y., & Ferree, M. M. (2010). Practicing intersectionality in sociological research: A critical analysis of inclusions, interactions, and institutions in the study of inequalities. Sociological Theory, 28(2), 129-149. Chou, Y.-C., Kröger, T., & Pu, C.-y. (2018). Underemployment among mothers of children with intellectual disabilities. Journal of Applied Research in Intellectual Disabilities, 31(1), 152-158. Clarke, H., & McKay, S. (2008). Exploring disability, family formation and break- up: reviewing the evidence (pp. 113): Institute of Applied Social Studies, University of Birmingham on behalf of the Department for Work and Pensions. Cockx, B., & Baert, S. (2015). Uitbesteding van verplichte begeleidings- en opleidingstrajecten voor langdurig werklozen: profit, nonprofit, of behoud de overheidsdienst?. Over.Werk, 25(4), 126-136. Cockx, B., Dejemeppe, M., & Van der Linden, B. (2011). L’activation du comportement de recherche d’emploi favorise-t-elle un retour plus rapide à l’emploi?. Regards économiques, 85, 1-14.

268

Collins, P. H. (2015). Intersectionality's definitional dilemmas. Annual Review of Sociology, 41(1), 1-20. Cooper, K., & Stewart, K. (2013). Does money affect children’s outcomes? A systematic review. York, UK: Joseph Rowntree Foundation. Cooper, K., & Stewart, K. (2017). Does money affect children’s outcomes? An update. Centre for Analysis of Social Exclusion Paper No. 203. London, UK: Centre for Analysis of Social Exclusion, London School of Economics. Corak, M. (2006). Do poor children become poor adults? Lessons from a cross country comparison of generational earnings mobility. IZA Discussion Papers No. 1993. Bonn, Germany: Institute for the Study of Labor. Corluy, V. (2014). Labour market outcomes and trajectories of immigrants in Belgium (Doctoral thesis). Antwerp, Belgium: University of Antwerp. Corluy, V., Haemels, J., Marx, I., & Verbist, G. (2015). The labour market position of second-generation immigrants in Belgium. NBB Working Paper No. 285. Brussels, Belgium: National Bank of Belgium. Corluy, V., & Vandenbroucke. F. (2014). Individual employment, household employment and risk of poverty in the EU: A decomposition analysis. In B. Cantillon, & F. Vandenbroucke (Eds.), Reconciling work and poverty reduction: How successful are European welfare states? (pp. 94-130). New York, USA: Oxford University Press. Corluy, V., & Verbist, G. (2014). Can education bridge the gap? Education and the employment position of immigrants in Belgium. ImPRovE Working Paper No. 14/02. Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Craig, P. (1991). Costs and benefits: a review of research on take up of income- related benefits. Journal of Social Policy, 20(4), 537-565. Craig, P., & Greenslade, M. (1998). Preliminary results from the 1997/98 Disability Survey follow on to the Family Resources Survey March 1998. Crettenden, A., Wright, A., & Skinner, N. (2014). Mothers caring for children and young people with developmental disabilities: intent to work, patterns of participation in paid employment and the experience of workplace flexibility. Community, Work & Family, 17(3), 244-267.

269

Currie, J. (2004). The take-up of social benefits Discussion paper No. 1103. Bonn, Germany: IZA. Daly, M. (2005). Gender mainstreaming in theory and practice. Social Politics, 12(3), 433-450. De Mulder, J., & Druant, M. (2011). The Belgian labour market during and after the crisis. Economic Review June 2011, 89-104. Debacker, M. (2007). De socio-economische positie van gezinnen met kinderen. In J. Ghysels & M. Debacker (Eds.), Zorgen voor kinderen in Vlaanderen: een dagelijkse evenwichtsoefening? (pp. 19-43). Leuven, Belgium: Acco. Debacker, M. (2008). Care strategies among high- and low-skilled mothers: a world of difference? Work, Employment and Society, 22(3), 527-545. Decancq, K., Goedemé, T., Van den Bosch, K., & Vanhille, J. (2014). The evolution of poverty in the European Union: Concepts, measurement and data. In B. Cantillon, & F. Vandenbroucke (Eds.), Reconciling work and poverty reduction in Europe: How successful are European welfare states? (pp.60-93). New York, USA: Oxford University Press. Decoster, A., Perelman, S., Vandelannoote, D., Vanheukelom, T., & Verbist, G. (2015). A bird’s eye view on 20 years of tax-benefit reforms in Belgium. KU Leuven Discussion Paper Series DPS15.07. Leuven, Belgium: KU Leuven. Delobel-Ayoub, M., Ehlinger, V., Klapouszczak, D., Maffre, T., Raynaud, J.-P., Delpierre, C., & Arnaud, C. (2015). Socioeconomic disparities and prevalence of autism spectrum disorders and intellectual disability. PLoS ONE, 10(11), 1-13. DeRigne, L. (2012). The employment and financial effects on families raising children with special health care needs: an examination of the evidence. Journal of Pediatric Health Care, 26(4), 283-290. DeRigne, L., & Porterfield, S. L. (2010). Employment change and the role of the medical home for married and single-mother families with children with special health care needs. Social Science & Medicine, 70(4), 631-641. DeRigne, L., & Porterfield, S. L. (2017). Employment change among married parents of children with special health care needs. Journal of Family Issues, 38(5), 579-606.

270

Dermaut, V., Schiettecat, T., Vandevelde, S., & Roets, G. (2020). Citizenship, disability rights and the changing relationship between formal and informal caregivers: it takes three to tango. Disability & Society, 35(2), 280-302. Dhamoon, R. K. (2011). Considerations on mainstreaming intersectionality. Political Research Quarterly, 64(1), 230-243. Di Giulio, P., Philipov, D., & Jaschinski, I. (2014). Families with disabled children in different European countries. Families and Societies Working Paper Series No. 23. Diris, R., Vandenbroucke, F., & Verbist, G. (2017). The impact of pensions, transfers and taxes on child poverty in Europe: the role of size, pro- poorness and child orientation. Socio-Economic Review, 15(4), 745-775. Duncan, G. J., Morris, P. A., & Rodrigues, C. (2011). Does money really matter? Estimating impacts of family income on young children’s achievement with data from random-assignment experiments. Developmental Psychology, 47(5), 1263-1279. Duncan, G. J., Ziol-Guest, K. M., & Kalil, A. (2010). Early-childhood poverty and adult attainment, behavior, and health. Child Development, 81(1), 306- 325. Duncan, S., Edwards, R., Reynolds, T., & Alldred, P. (2003). Motherhood, paid work and partnering: values and theoies. Work, Employment and Society, 17(2), 309-330. Durkin, M. S., Maenner, M., J., Baio, J., Christensen, D., Daniels, J., Fitzgerald, R., … Yeargin-Allsopp, M. (2017). Autism spectrum disorder among US children (2002-2010): Socioeconomic, racial, and ethnic disparities. American Journal of Public Health, 107(11), 1818-1826. Dwyer, P., & Wright, S. (2014). Universal Credit, ubiquitous conditionality and its implications for social citizenship. Journal of Poverty and Social Justice, 22(1), 27-35. EASIE. (2017). European Agency Statistics on Inclusive Education: 2014 dataset cross-country report. In J. Ramberg, A. Lénárt & A. Watkins (Eds.), (pp. 124). Odense, Denmark: EASIE.

271

EASIE. (2018). European Agency Statistics on Inclusive Education: 2016 dataset cross-country report (pp. 124). In: J. Ramberg, A. Lénárt & A. Watkins (Eds). Odense, Denmark: EASIE. Ellingsæter, A. L., & Gulbrandsen, L. (2007). Closing the childcare gap: The interaction of childcare provision and mothers' agency in Norway. Journal of Social Policy, 36(4), 649-669. Elwan, A. (1999). Poverty and disability: A survey of the literature. Social Protection Discussion Paper No. 9932. Social Protection Unit, The World Bank. Emerson, E. (2003). Mothers of children and adolescents with intellectual disability: social and economic situation, mental health status, and the self- assessed social and psychological impact of the child's difficulties. Journal of Intellectual Disability Research, 47, 385-399. Emerson, E. (2012). Deprivation, ethnicity and the prevalence of intellectual and developmental disabilities. Journal of Epidemiology & Community Health, 66(3), 218-224. Emerson, E., Graham, H., & Hatton, C. (2006). Household income and health status in children and adolescents in Britain. European Journal of Public Health, 16(4), 354-360. Emerson, E., & Hatton, C. (2007). The socio‐economic circumstances of children at risk of disability in Britain. Disability & Society, 22(6), 563-580. Emerson, E., Shahtahmasebi, S., Lancaster, G., & Berridge, D. (2010). Poverty transitions among families supporting a child with intellectual disability. Journal of Intellectual & Developmental Disability, 35(4), 224-234. Engel, G. L. (1977). The need for a new medical model: A challenge for biomedicine. Science, 196(4286), 129-136. Esping-Andersen, G. (1990). The three worlds of welfare capitalism. Princeton: Princeton University Press. Esping-Andersen, G. (1996). Welfare states without work: The impasse of labour shedding and familialism in continental European social policy. In G. Esping-Andersen (Ed.), Welfare states in transition: National adaptation in global economy (pp. 66-87) London, UK: Sage. Esping-Andersen, G., Gallie, D., Hemerijck, A., & Myles, J. (2002). Why we need a new welfare state. Oxford: Oxford University Press.

272

EUROFOUND. (2015). Access to social benefits: Reducing non-take-up (pp. 68). Luxembourg: Publications Office of the European Union. European Commission. (2008). Communication from the Commission: From financial crisis to recovery: A European framework for action. COM/2008/0706 final. Brussels, Belgium: Commission of the European Communities. European Commission. (2010). European Economic Forecast: Spring 2010. European Economy 2/2010. Luxembourg: Publications Office of the European Communities. European Commission. (2013). Investing in children: breaking the cycle of disadvantage Commission Recommendation (pp. 21). Brussels: European Commission. European Commission (2016). Country Report Belgium 2016. SWD(2016) 71 final. Brussels, Belgium: European Commission. European Parliament, Council of the European Union, & European Commission. (2017). European pillar of social rights. Luxembourg: Publication Office of the European Union. EUROSTAT. (2011). Migrants in Europe. A statistical portrait of the first and second generation. Luxembourg: Eurostat Statistical Books. EUROSTAT. (2016a). Employment rates by sex, age and educational attainment level (%). Retrieved May 2, 2016, from https://appsso.eurostat.ec.europa.eu/nui/submitViewTableAction.do. EUROSTAT. (2016b). GDP and main components (output, expenditure and income), GDP at market prices, chain linked volumes, index 2010=100, 2000Q1-2015Q4. Retrieved January 29, 2016, from https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=namq_10_gdp &lang=en. EUROSTAT. (2016c). Unemployment rates by sex, age and citizenship (%). Retrieved May 2, 2016, from http://appsso.eurostat.ec.europa.eu/nui/show.do?wai=true&dataset=lfsq_ urgan. EUROSTAT. (2020a). At-risk-of-poverty rate by poverty threshold, age and sex - EU-SILC and ECHP surveys (ilc_li02) 2006-2018. Retrieved February 6, 2020 from

273

https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=ilc_li02&lang= en. EUROSTAT. (2020b). At-risk-of-poverty threshold and work intensity of the household (population aged 0 to 59 years) - EU-SILC survey (ilc_li06) 2006-2018. Retrieved February 6, 2020 from https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=ilc_li06&lang= en. EUROSTAT. (2020c). Children with limitations in activities due to health problems, by income groups, household composition and age (ilc_hch13) 2017. Retrieved February 13, 2020 from https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=ilc_hch13&lan g=en. EUROSTAT. (2020d). Distribution of population by work intensity of the household (population aged 0 to 59 years) - EU-SILC survey (ilc_lvps03) 2006-2018. Retrieved February 6, 2020 from https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=ilc_lvps03&lan g=en. EUROSTAT. (2020e). EU and national quality reports. Retrieved June 6, 2020 from https://ec.europa.eu/eurostat/web/income-and-living- conditions/quality/eu-and-national-quality-reports. Evans, G. W. (2016). Childhood poverty and adult psychological well-being. Proceedings of the National Academy of Sciences, 113(52), 14949-14952. FAMIFED. (2016). Een overzicht per entiteit van de kinderbijslag voor kinderen met een aandoening Focus 2016-3 (pp. 16). Brussels, Belgium: FAMIFED. FAMIFED. (2017a). Aantal kinderen met een aandoening, volgens zelfredzaamheidsgraag en ernst van de gevolgen van de aandoening, op 31 december 2010. Retrieved August 3, 2017 from http://vlaanderen.famifed.be/nl/statistics/222?year=2010. FAMIFED. (2017b). De kinderen in België zonder Belgische kinderbijslag Focus 2017-2 (pp. 59). Brussels, Belgium: FAMIFED. FAMIFED. (2018). Barema van de kinderbijslag. Vlaamse Gemeenschap. Retrieved January 24, 2018, from http://vlaanderen.famifed.be/sites/default/files/uploads/Barema%20Vlaan

274

deren%201-2018%20- %20%20%28aanpassing%20grensbedragen%20inkomens%29.pdf FAMIFED (2019a). Aantal rechtgevende kinderen met een aandoening per leeftijd (algemeen), 2013-2-2014-2. Retrieved November 12, 2019 from https://brussel.famifed.be/nl/statistics/304. FAMIFED (2019b). Aantal rechtgevende kinderen per leeftijd (algemeen), 2013- 2-2014-2. Retrieved November 12, 2019 from https://brussel.famifed.be/nl/statistics/216. FamiStat. (2019a). Aantal rechtgevende kinderen met een aandoening per leeftijdsgroep en per entiteit, 2015-2-2018-2. Retrieved November, 12, 2019 from https://stat.famifed.be/demographic/volet-04c-04.php. FamiStat. (2019b). Aantal rechtgevende kinderen per leeftijdsgroep en per entiteit, 2015-2-2018-2. Retrieved November, 12, 2019 from https://stat.famifed.be/demographic/volet-04-04.php. FAPD. (2010). Jaarverslag 2010. Retrieved from https://www.vaph.be/sites/default/files/documents/jaarverslag-2010- 2010/12/31/jaarverslag_2010.pdf. Finch, J. (1989). Family obligations and social change. Cambridge: Polity Press. Förster, M., & Mira d’Ercole, M. (2005). Income distribution and poverty in OECD countries in the second half of the 1990s. OECD Social, Employment and Migration Working Papers No. 22. Paris, France: OECD Publishing. Fouarge, D., & Layte, R (2005). Welfare regimes and poverty dynamics: The duration and recurrence of poverty spells in Europe. Journal of Social Policy, 34(3), 407-426. FPS Finance. (2009a). Bijlage III van het koninklijk besluit tot uitvoering van het wetboek van de inkomstenbelasting 1992. Schalen en regels die van toepassing zijn om de bedrijfsvoorheffing vast te stellen bij de bron verschuldigd op inkomsten betaald of toegekend vanaf 1 januari 2010. Brussels, Belgium: FPS Finance. FPS Finance. (2009b). Sleutelformule voor het berekenen van de bedrijfsvoorheffing (BV) verschuldigd op bezoldigingen en op in artikel 146, 1°, van het wetboek van de inkomstenbelastingen (WIB 92) vermelde

275

pensioenen of brugpensioenen, betaald vanaf 1 januari 2010. Brussels, Belgium: FPS Finance. FPS Finance. (2010). Tax survey. July 2010 issue. Brussels, Belgium: FPS Finance. FPS Finance. (2011). Tax survey. July 2011 issue. Brussels, Belgium: FPS Finance. FPS Finance. (2012). Cijferverslag 2012. Brussels, Belgium: FPS Finance. Fujiura, G. T., & Yamaki, K. (2000). Trends in demography of childhood poverty and disability. Exceptional Children, 66(2), 187-199. Galloway, T. A., Gustafsson, B., Pedersen, P. J., & Österberg, T. (2015). Immigrant child poverty: The Achilles heel of the Scandinavian welfare state. Measurement of Poverty, Deprivation, and Economic Mobility. Research on Economic Inequality, 23, 185-219. Gambaro, L., Stewart, K., & Waldfogel, J. (Eds.) (2014). An equal start? Providing quality early education and care for disadvantaged children. Bristol, UK: Policy Press. Gesthuizen, M., & Scheepers, P. (2010). Economic vulnerability among low- educated Europeans: Resource, composition, labour market and welfare state influences. Acta Sociologica, 53(3), 247-267. Ghysels, J., & Van Lancker, W. (2011). The unequal benefits of activation: an analysis of the social distribution of family policy among families with young children. Journal of European Social Policy, 21(5), 472-485. Goldman, N. (2001). Social inequalities in health. Disentangling the underlying mechanisms. Annals of the New York Academy of Sciences, 954, 118-139. Good Gingrich, L. (2008). Social exclusion and double jeopardy: The management of lone mothers in the market-state social field. Social Policy and Administration, 42(4), 379-395. Goos, M., Manning, A., & Salomons, A. (2009). Job polarization in Europe. American Economic Review Papers and Proceedings 99(2), 58-63. Gordon, M., Rosenman, L., & Cuskelly, M. (2007). Constrained labour: maternal employment when children have disabilities. Journal of Applied Research in Intellectual Disabilities, 20(3), 236-246.

276

Gornick, J. C., & Jäntti, M. (2012). Child poverty in cross-national perspective: Lessons from the Luxembourg Income Study. Children and Youth Service Review, 34(3), 558-568. Gornick, J. C., & Meyers, M. K. (2003). Families that work: Policies for reconciling parenthood and employment. New York, USA: Russell Sage Foundation. Gornick, J. C., & Nell, E. (2017). Children, poverty, and public policy: A cross- national perspective. LIS Working Paper Series No. 701. Luxembourg: Luxembourg Income Study. Grammenos, S. (2013). European comparative data on Europe 2020 & people with disabilities. GLADNET Collection 12-2013. Ithaca, USA: Cornell University, ILR School. Grammenos, S. (2018). European comparative data on Europe 2020 & people with disabilities: leisure activities & active citizenship of people with disabilities. Report prepared for Academic Network of European Disability Experts. Brussels, Belgium: Centre for European Social and Economic Policy. Greve, B. (Ed.) (2020). Routledge international handbook of poverty. Abingdon, UK: Routledge. Griggs, J., & Walker, R. (2008). The costs of child poverty for individuals and society. A literature review. York, UK: Joseph Rowntree Foundation. Guio, A.C (2016, May). La structure globale des revenus de la population et la lutte contre la pauvreté. In FAMIFED seminar, De socio-economische rol van de kinderbijslag: een instrument in de strijd tegen de armoede. Brussels, Belgium: FAMIFED. Guio, A.-C., Vandenbroucke, F., & Vinck, J. (2015). Kinderarmoede hoger op de politieke agenda plaatsen: enkele cijfers om beter te begrijpen wat er op het spel staat. In I. Pannecoucke, W. Lahaye, J. Vrancken, & R. Van Rossem (Eds.), Armoede in België. Jaarboek 2015 (pp. 113-140). Ghent, Belgium: Academia Press. Guo, C., Luo, Y., Tang, X., Ding, R., Song, X., & Zheng, X. (2019). Poverty and youth disability in China: Results from a large, nationwide, population- based survey. PLoS ONE, 14(4), 1-12.

277

Haegele, J. A., & Hodge, S. (2016). Disability discourse: Overview and critiques of the medical and social models. Quest, 68(2), 193-206. Hagenaars, A. J. M., de Vos, K., & Zaidi, M. A. (1994). Poverty statistics in the late 1980s: Research based on micro-data. Luxembourg: Office for Official Publications of the European Communities. Halvorsen, R., & Jensen P. H. (2004). Activation in Scandinavian welfare policy: Denmark and Norway in a comparative perspective. European Societies 6(4), 461-483. Hancock, A.-M. (2007). When multiplication doesn't equal quick addition: Examining intersectionality as a research paradigm. Perspectives on Politics, 5(1), 63-79. Haug, K. H., & Storø, J. (2013). Kindergarten - a universal right for children in Norway. International journal of child care and education policy, 7(2), 1- 13. Hauge, L. J., Kornstad, T., Nes, R. B., Kristensen, P., Irgens, L. M., Eskedal, L. T., . . . Vollrath, M. E. (2013). The impact of a child's special health care needs on maternal work participation during ealry motherhood. Paediatric and Perinatal Epidemiology, 27(4), 353-360. Haveman, M., van Berkum, G., Reijnders, R., & Heller, T. (1997). Differences in service needs, time demands, and caregiving burden among parents of persons with mental retardation across the life cycle. Family Relations, 46(4), 417-425. Havnes, T., & Mogstad, M. (2011). Money for nothing? Universal child care and maternal employment. Journal of Public Economics, 95(11-12), 1455- 1465. Hemerijck, A. (2012). Changing Welfare States. Oxford, UK: Oxford University Press. Hemerijck, A. (2017). The uses of social investment. Oxford: Oxford University Press. Hemerijck, A. (2018). Social investment as a policy paradigm. Journal of European Public Policy, 25(6), 810-827. Hemerijck, A., & Marx, I. (2010). Continental welfare at a crossroads: The choice between activation and minimum income protection in Belgium and the Netherlands. In Palier B (Ed.) A long goodbye to Bismarck? The politics

278

of welfare reform in continental Europe (pp. 129-155). Amsterdam, the Netherlands: Amsterdam University Press. Hernanz, V., Malherbet, F., & Pellizzari, M. (2004). Take-up of welfare benefits in OECD countries: A review of the evidence Social, Employment and Migration Working Papers (pp. 47). Paris, France: OECD. Houtrow, A. J., Larson, K., Olson, L. M., Newacheck, P. W., Halfon, N. (2014). Changing trends of childhood disability, 2001-2011. Pediatrics, 134(3), 530-538. Hufkens, T., Vandelannoote, D., Van Lancker, W., & Verbist, G. (2013). Hervorming van de Vlaamse kinderbijslag en armoedebestrijding: Een simulatie van alternatieven. CSB Berichten 11. Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Hvinden, B. (2003). The uncertain convergence of disability policies in Western Europe. Social Policy and Administration, 37(6), 609-624. IMF. (2009) Belgium: 2008 Article IV Consultation: Staff Report; Staff Supplement; and Public Information Notice on the Executive Board Discussion. IMF Country Report No. 09/87. Washington, DC, USA: IMF. IMF. (2011). Belgium: Selected issues paper. IMF Country Report No. 11/82 Washington, DC, USA: IMF. Jenkins, S. P., Brandolini, A., Micklewright, J., Nolan, B. (Eds.). (2013). The Great Recession and the distribution of household income. Oxford, UK: Oxford University Press. Kautto, M., Fritzell, J., Hvinden, B., Kvist, J., & Uusitalo, H. (2001). Nordic welfare states in the European context. London and New York: Routledge. Kawa, R., Saemunsen, E., Jónsdóttir, S. L., Hellendoorn, A., Lemcke, S., Canal- Bedia, R., . . . Moilanen, I. (2016). European studies on prevalence and risk of autism spectrum disorders according to immigrant status - a review. The European Journal of Public Health, 27(1), 101-110. Kiernan, K., E., & Huerta, M. C. (2008). Economic deprivation, maternal depression, parenting and children’s cognitive and emotional development in early childhood. The British Journal of Sociology, 59(4), 783-806. Kil, T., Neels, K., Wood, J., & de Valk, H. (2017). Employment after parenthood: women of migrant origin and natives compared. European Journal of Population, 28p.

279

Konietzka, D., & Kreyenfeld, M. (2010). The growing educational divide in mothers’ employment: an investigation based on the German micro- censuses 1976-2004. Work, Employment and Society, 24(2), 260-278. Korpi, W. (2000). Faces of inequality: Gender, class, and patterns of inequalities in different types of welfare states. Social Politics, 7(2), 127-191. Korpi, W., Ferrarini, T., & Englund, S. (2013). Women's opportunities under different family policy constellations: Gender, class, and inequality tradeoffs in Western countries re-examined. Social Politics, 20(1), 1-40. Kossek, E. E., & Lautsch, B. A. (2018). Work-life flexibility for whom? Occupational status and work-life inequality in upper, middle and lower level jobs. Academy of Management Annals, 12(1), 5-36. Lai, E. T. C., Wickham, S., Law, C., Whitehead, M., Barr, B., & Taylor-Robinson, D. (2019). Poverty dynamics and health in late childhood in the UK: Evidence from the Millennium Cohort Study. Archives of Disease in Childhood, 104(11), 1049-1055. Larkins, C., Thomas, N., Judd, D., Lloyd, J., Carter, B., & Farrelly, N. (2013). "We want to help people see things our way": A rights-based analysis of disabled children's experience living with low income. UK: Children's commissioner. Lebeer, J., Vinck, J., & Farah, S. (2018). Belgium fact sheet on social care & support services sector for persons with disabilities. EASPD country fact sheets. Brussels, Belgium: EASPD. Leiter, V., Wyngaarden Krauss, M., Anderson, B., & Wells, N. (2004). The consequences of caring. Effects of mothering a child with special needs. Journal of Family Issues, 25(3), 379-403. Leonard, H., Petterson, B., De Klerk, N., Zubrick, S. R., Glasson, E., Sanders, R., & Bower, C. (2005). Association of sociodemographic characteristics of children with intellectual disability in Western Australia. Social Science & Medicine, 60(7), 1499-1513. Lesner, R. V. (2018). The long-term effect of childhood poverty. Journal of Population Economics, 31(3), 969-1004. Lindsay, C., Greve, B., Cabras, I., Ellison, N., & Kellet, S. (2015). Assessing the evidence base on health, employability and the labour market - Lessons for activation in the UK. Social Policy and Administration, 49(2), 143-150.

280

Loyalka, P., Liu, L., Chen, G., & Zheng, X. (2014). The cost of disability in China. Demography, 51(1), 97-118. Lu, Z.-H., & Zuo, A. (2010). Effects of a child's disability on affected female's labour supply in Australia. Australian Economic Papers, 49(3), 222-240. Luca, D. L., & Sevak, P. (2019). Child disability, maternal labor supply, and household well-being. DRC Working Paper No. 2019-09. Cambridge, USA: Mathematica, Center for Studying Disability Policy. Lustig, D. C., & Strauser, D. R. (2007). Causal relationships between poverty and disability. Rehabilitation Counseling Bulletin, 50(4), 194-202. Maes, B., Maljaars, J., & Noens, I. (2017). Leeftijdsgebonden ankerpunten parameters van zorgzwaarte. Leuven, Belgium: Gezins- en Orthopedagogiek, KU Leuven. Maggi, S., Irwin, L., J., Siddiqi, A., & Hertzman, C. (2010). The social determinants of early development: An overview. Journal of Paediatrics and Child Health, 46(11), 627-635. Main, G. (2014). Child poverty and children’s subjective well-being. Child Indicators Research, 7(3), 451-472. Maldonado, L. C., & Nieuwenhuis, R. (2015). Family policies and single parent poverty in 18 OECD countries, 1978–2008. Community, Work & Family, 18(4), 395-415. Marchal, S. (2017). The social floor: Essays on minimum income protection (Doctoral thesis). University of Antwerp, Belgium. Marchal, S., Marx, I., & Van Mechelen, N. (2014). The great wake-up call? Social citizenship and minimum income provisions in Europe in times of crisis. Journal of Social Policy, 43(2), 247-267. Marin, B., Prinz, C., & Queisser, M. (Eds.) (2004). Transforming disability welfare policies: Towards work and equal opportunities. Farnham, UK: Ashgate. McBride, A., Hebson, G., & Holgate, J. (2015). Intersectionality: are we taking enough notice in the field of work and employment relations? Work, Employment and Society, 29(2), 331-341. McCall, L. (2005). The complexity of intersectionality. Signs, 30(3), 1771-1800.

281

McKay, S., & Atkinson, A. (2007). Disability and caring among families with children: Family employment and poverty characteristics. Department for work and pensions research report No. 460. London, UK: Department for work and pensions. McKinley Yoder, C. L., & Cantrell, M. A. (2019). Childhood disability and educational outcomes: A systematic review. Journal of Pediatric Nursing, 45, 37-50. Meeusen, L., & Nys, A. (2014). The evolution of public social spending 1985- 2009. In B. Cantillon, & F. Vandenbroucke (Eds.), Reconciling work and poverty reduction: How successful are European welfare states? (pp. 325- 381). New York, USA: Oxford University Press. Meyers, M. K., Lukemeyer, A., & Smeeding T. (1998). The cost of caring: Childhood Disability and Poor Families. Social Service Review, 72(2), 209-233. Milligan, K., & Stabile, M. (2011). Do child tax benefits affect the well-being of children? Evidence from Canadian child benefit expansions. American Economic Journal: Economic Policy, 3(3), 175-205. Minujin, A., Delamonica, E., Davidziuk, A., & Gonzalez, E. D. (2006). The definition of child poverty: a discussion of concepts and measurements. Environment & Urbanization, 18(2), 481-500. Mitra, S., Palmer, M., Kim, H., Mont, D., & Groce, N. (2017). Extra costs of living with a disability: A review and agenda for research. Disability and Health Journal, 10(4), 475-484. Moffitt, R. (1983). An economic model of welfare stigma. American Economic Review, 73(5), 1023-1035. Mont, D. (2014). Childhood disability and poverty. Working Paper No. 25. London, UK: Leonard Cheshire Disability and Inclusive Development Centre, University College London. Mont, D., & Cuong, N. V. (2011). Disability and poverty in Vietnam. The World Bank Economic Review, 25(2), 323-359. Monteith, M., Casement, E., Lloyd, K., & McKee, P. (2009). Taking a closer look: Child Poverty and Disability. November Briefing. Belfast, UK: Save the Children/Family Fund

282

Mood, C. (2010). Logistic regression: why we cannot do what we think we can do, and what we can do about it. European Sociological Review, 26(1), 67- 82. Mooney, S. (2016). 'Nimble' intersectionality in employment research: a way to resolve methodological dilemmas. Work, Employment and Society, 30(4), 708-718. Morel, N., Palier, B., & Palme, J. (Eds.) (2012). Towards a social investment welfare state? Ideas, policies and challenges. Bristol, UK: Policy Press. Morris, Z. A., & Zaidi, A. (forthcoming). Estimating the extra costs of disability in European countries: Implications for poverty measurement and disability-related decommodification. Journal of European Social Policy, 00(0), 1-16. National Bank of Belgium (2016). Quarterly and annual aggregates: GDP growth. Chain linked - Q to Q percentage change (reference year 2013), 2000Q1-2015Q4. Retrieved January 29, 2016, from https://stat.nbb.be/index.aspx?queryid=3038. Nieuwenhuis, R., & Maldonado, L. C. (Eds.) (2018). The triple bind of single- parent families: Resources, employment and policies to improve well- being. Bristol, UK: Policy Press. Nikiéma, B., Spencer, N., & Séguin, L. (2010). Poverty and chronic illness in early childhood: A comparison between the United Kingdom and Quebec. Pediatrics, 125(3), e499-507. NLWA. (2018). Rates of basic benefit and attendance benefit (forhøyet) hjelpestønad. Retrieved May 15, 2018 from: https://www.nav.no/no/NAV+og+samfunn/Kontakt+NAV/Utbetalinger/S narveier/satser--380089?kap=380103. Nolan, B., & Whelan, C. T. (2011). Poverty and Deprivation in Europe. Oxford, UK: Oxford University Press. Nys, A., Meeusen, L., & Corluy, V. (2016). Who cares? A counterfactual analysis of household work intensity in households with disabled family members. Social Indicators Research, 128(2), 675-691. OECD. (2006). Starting Strong II: Early childhood education and care. Paris: OECD.

283

OECD. (2008). Growing unequal? Income distribution and poverty in OECD countries. Paris, France: OECD Publishing. OECD. (2010). Sickness, disability and work: Breaking the barriers. A synthesis of findings across OECD countries. Paris, France: OECD Publishing. OECD. (2011). Economic Perspectives Report 2011. Paris, France: OECD Publishing. OECD. (2012). Settling in. OECD indicators of immigrant integration 2012. Paris, France: OECD Publishing. OECD. (2014a). Education at a Glance 2014: OECD Indicators. Paris, France: OECD Publishing. OECD. (2014b). Rising inequality: youth and poor fall further behind Income Inequality Update: OECD. OECD. (2016). Enhancing child well-being to promote inclusive growth. Note by the Secretary General. Meeting of the Council at Ministerial Level, June 1-2, 2016. Retrieved from http://www.oecd.org/officialdocuments/publicdisplaydocumentpdf/?cote =DELSA/ELSA(2016)7/REV1&doclanguage=en. Olsson, M. B., & Hwang, C. P. (2006). Well-being, involvement in paid work and division of child-care in parents of children with intellectual disabilities in Sweden. Journal of Intellectual Disability Research, 50(12), 963-969. Owen, L., Gordon, M., Frederico, M., & Cooper, B. (2003). Listen to us: supporting families with children with disabilities: identifying service responses that impact on the risk of family breakdown. Melbourne, Australia: School of Social Work & Social Policy, La Trobe University and Victorian Government Department of Human Services. Palmer, M., & Harley, D. (2012). Models and measurement in disability: An international review. Health Policy and Planning, 27(5), 357-364. Palmer, M., Williams, J., & McPake, B. (2019). Standard of living and disability in Cambodia. The Journal of Development Studies, 55(11), 2382-2402. Parish, S. L., & Cloud, J. M. (2006). Financial well-being of young children with disabilities and their families. Social Work, 51(3), 223-232.

284

Parish, S. L., Rose, R. A., Grinstein-Weiss, M., Richman, E. L., & Andrews, M. E. (2008). Material hardship in U.S. families raising children with disabilities. Exceptional Children, 75(1), 71-92. Pavolini, E., & Van Lancker, W. (2018). The Matthew effect in childcare use: a matter of policies or preferences? Journal of European Public Policy, 25(6), 878-893. Petrenchik, T. M. (2008). Childhood disability in the context of poverty. Discussion paper prepared for the Ontario Ministery of children and youth services. Ontario, Canada: CanChild Centre for Childhood Disability Research. Pintelon, O., Cantillon, B., Van den Bosch, K., & Whelan, C. (2013). The Social Stratification of Social Risks: The Relevance of Class for Social Investment Strategies. Journal of European Social Policy, 23(1), p. 52-67. doi: doi:10.1177/0958928712463156 POD MI (2014). RVA-sanctie en doorstroom naar de OCMW’s. Focusnota 8. Brussels, Belgium: POD MI. Porterfield, S. L. (2002). Work choices of mothers in families with children with disabilities. Journal of Marriage and Family, 64(4), 972-981. Porterfield, S. L., & Tracey, C. (2003). Disentangling the dynamics of family poverty and child disability: Does disability come first? Center for Social Development Working Paper No. 03-1. St. Louis, MO, USA: Center for Social Development. Power, C., Li, L., & Manor, O. (2000). A prospective study of limiting longstanding illness in early adulthood. International Journal of Epidemiology, 29(1), 131-139. Powers, E. T. (2001). New estimates of the impact of child disability on maternal employment. The American Economic Review, 91(2), 135-139. Powers, E. T. (2003). Children's health and maternal work activity: estimates under alternative disability definitions. The Journal of Human Resources, 38(3), 522-556. Pulcini, C. D., Zima, B. T., Kelleher, K. J., & Houtrow, A. J. (2017). Poverty and trends in 3 common chronic disorders. Pediatrics, 127(3), 1-9. Rai, D., Lewis, G., Lundberg, M., Araya, R., Svensson, A., Dalman, C., … Magnusson, C. (2012). Parental socioeconomic status and risk of offspring

285

autism spectrum disorder in a Swedish population-based study. Journal of the American Academy of Child & Adolescent Psychiatry, 51(5), 467-476. Rainwater, L., & Smeeding, T. M. (2003). Poor kids in a rich country: America’s children in comparative perspective. New York, USA: Russell Sage Foundation. Ravallion, M. (1994). Poverty Comparisons. Chur, Switzerland: Harwood. Reichman, N. E., Corman, H., & Noonan, K. (2008). Impact of child disability on the family. Maternal and Child Health Journal, 12(6), 679-683. Ridge, T. (2002). Childhood poverty and social exclusion: From a child’s perspective. Bristol, United Kingdom: Policy Press. Ridge, T. (2011). The everyday costs of poverty in childhood: a review of qualitative research exploring the lives and experiences of low-income children in the UK. Children & Society, 25(1), 73-84. Risdal, D., & Singer, G. H. S. (2004). Marital adjustments in parents of children with disabilities: a historical review and meta-analysis. Research & Practice for Persons with Severe Disabilities, 29(2), 95-103. RKW. (2013). De kinderbijslag voor kinderen met een aandoening: tien jaar na de hervorming. Focusstudie 2013-1. Brussels, Belgium: RKW. Roelen, K., & Gassmann, F. (2008). Measuring child poverty and well-being: a literature review Maastricht Graduate School of Governance Working Paper Series No. 2008/WP001. Maastricht: Maastricht Graduate School of Governance. Roets, G., Dermaut, V., Benoot, T., Claes, C., Schiettecat, T., Roose, R., … Vandevelde, S. (2020). A critical analysis of disability policy and practice in Flanders: Towards differentiated manifestations of interdependency. Journal of Policy and Practice in Intellectual Disabilities, 17(2), 108-115. Roets, G., Roose, R., Claes, L., Vandekinderen, C., Van Hove, G., & Vanderplasschen, W. (2012). Reinventing the employable citizen: a perspective for social work. British Journal of Social Work, 42(1), 94-110. Romig, K. (2017). SSI: A lifeline for children with disabilities. Policy Futures. Washington, DC, USA: Center on Budget and Policy Priorities. Rothwell, D. W., Gariépy, G., Elgar, F. J., & Lach, L. M. (2019). Trajectories of poverty and economic hardship among American families supporting a

286

child with a neurodisability. Journal of Intellectual Disability Research, 63(10), 1273-1284. Salverda, W. (2016). The tsunamis of educational attainment and part-time employment, and the change of the labour force 1960-2010: What can be learned about self-reinforcing labour-market inequality from the case of the Netherlands, in international comparison?. ImPRovE Working Paper No. 16/04. Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Saunders, P. (2007). The costs of disability and the incidence of poverty. Australian Journal of Social Issues, 42(2), 461-480. Sebrechts, L., & Breda, J. (2011). Kinderen met bijzondere behoeften en hun gezin. De kwetsbaarheid van deze gezinnen binnen het opkomend burgerschapsmodel. CSB Berichten D/2011/6104/04. Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Sebrechts, L., & Breda, J. (2012). Families of children with special needs in Flanders: their vulnerability within the citizenship paradigm (pp. 20). Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Shahtahmasebi, S., Emerson, E., Berridge, D., & Lancaster, G. (2011). Child disability and the dynamics of family poverty, hardship and financial strain: evidence from the UK. Journal of Social Policy, 40(4), 653-673. Singer, G. H. S., & Floyd, F. (2006). Meta-analysis of comparative studies of depression in mothers of children with and without developmental disabilities. American Journal on Mental Retardation, 111(3), 155-169. Singh, G. K., & Lin, S. C. (2013). Marked ethnic, nativity, and socioeconomic disparities in disability and health insurance among US children and adults: The 2008–2010 American Community Survey. BioMed Research International, 1-17. Stabile, M., & Allin, S. (2012). The economic costs of childhood disability. The Future of Children, 22(1), 65-96. Stahl, J. F., & Schober, P. S. (2018). Convergence or divergence? Educational discrepancies in work-care arrangements of mothers with young children in Germany. Work, Employment and Society, 32(4), 629-649. Statistics Belgium. (2010). Verdeling (%) van de hoofdbestanddelen van het totaal netto inkomen per totaal netto belastbaar inkomensklasse van €5000.

287

Retrieved from https://statbel.fgov.be/nl/themas/huishoudens/fiscale- inkomens/plus. Stegman Bailey, M., & Hemmeter, J. (2014). Characteristics of noninstitutionalized DI and SSI program participants, 2010 update. Research and Statistics Note No. 2014-02. Woodlawn, USA: Social Security Administration. Stiker, H.-J. (1999). A history of disability. Michigan, USA: University of Michigan. Tapia, M., & Alberti, G. (2019). Unpacking the category of migrant workers in trade union research: A multi-level approach to migrant intersectionalities. Work, Employment and Society 33(2), 314-325. Thévenon, O., Manfredi, T., Govind, Y., & Klauzner, I. (2018). Child poverty in the OECD: Trends, determinants and policies to tackle it. OECD Social Employment and Migration Working Papers No. 218. Paris, France: OECD Publishing. Tøssebro, J., & Wendelborg, C. (2017). Marriage, separation and beyond: a longitudinal study of families of children with intellectual and developmental disabilities in a Norwegian context. Journal of Applied Research in Intellectual Disabilities, 30(1), 121-131. Troger, T., & Verwiebe, R. (2015). The role of education for poverty risks revisited: Couples, employment and profits from work-family policies. Journal of European Social Policy, 25(3), 286–302. UNICEF. (2005). The state of the world of the children 2005: childhood under threat. New York, USA: UNICEF. UNICEF. (2007). Child poverty in perspective: An overview of child well-being in rich countries Innocenti Report Card 7 (pp. 48). Florence, Italy: UNICEF Innocenti Research Centre. UNICEF. (2014). Children of the Recession. The impact of the economic crisis on child well-being in rich countries. Innocenti Report Card 12. Florence, Italy: Florence, Italy: UNICEF, Innocenti Research Centre. United Nations. (2015). Transforming our world: The 2030 agenda for sustainable development. Retrieved February 3, 2020 from https://sustainabledevelopment.un.org/post2015/transformingourworld.

288

Uunk, W., Kalmijn, M., & Muffels, R. (2005). The impact of young children on women's labour supply. Acta Sociologica, 48(1), 41-62. Van Gyes, G. (2009). Reducing working time as anti-crisis measure. Retrieved from: http://www.eurofound.europa.eu/observatories/eurwork/articles/reducing- working-time-as-anti-crisis-measure. Van Gyes, G. (2010) Crisis in social dialogue due to anti-crisis measures. Retrieved from http://www.eurofound.europa.eu/observatories/eurwork/articles/industrial -relations/crisis-in-social-dialogue-due-to-anti-crisis-measures. Van Lancker, W. (2013). Putting the child-centred investment strategy to the test: evidence for the EU27. European Journal of Social Security, 15(1), 4-27. Van Lancker, W., & Ghysels. J. (2012). Who benefits? The social distribution of subsidized childcare in Sweden and Flanders. Acta Sociologica, 55(2), 125-142. Van Lancker, W., Ghysels, J., & Cantillon, B. (2015). The impact of child benefits on single mother poverty: Exploring the role of targeting in 15 European countries. International Journal of Social Welfare, 24, (3), 210-222. Van Lancker, W., & Van Mechelen, N. (2015). Universalism under siege? Exploring the association between targeting, child benefits and child poverty across 26 countries. Social Science Research, 50(1), 60-75. Van Lancker, W., & Vinck, J. (2020). The consequences of growing up poor. In B. Greve (Ed.), Routledge international handbook of poverty (pp. 96-106). Abingdon, UK: Routledge. Van Landeghem, C., Breda, J., & Mestdagh, K. (2007). Zorgpatronen van kinderen met een handicap. In J. Ghysels & M. Debacker (Eds.), Zorgen voor kinderen in Vlaanderen: een dagelijkse evenwichtsoefening? (pp. 191-235). Leuven, Belgium: Acco. Van Mechelen, N., & Janssens, J. (2017). Who is to blame? An overview of the factors contributing to the non-take-up of social rights Working paper No. 17.08. University of Antwerp, Belgium: Herman Deleeck Centre for Social Policy.

289

Van Mechelen, N., & Marchal, S. (2013). Struggle for life: Social assistance benefits, 1992–2009. In I. Marx, & K. Nelson (Eds.), Minimum income protection in flux (pp. 28-53). Houndmills, UK: Palgrave Macmillan. van Oorschot, W. (1996). Modelling non-take-up: The interactive model of multi- level influences and the dynamic model of benefit receipt. In W. van Oorschot (Ed.), New perspectives on the non-take-up of social security benefits (pp. 7-59). Tilburg, the Netherlands: Tilburg University Press. Vandenbroeck, M., De Visscher, S., Van Nuffel, K., & Ferla, J. (2008). Mothers’ search for infant child care: The dynamic relationship between availability and desirability in a continental European welfare state. Early Childhood Research Quarterly, 23(2), 245-258. Vandenbroeck, M., Roets, G., & Roose, R. (2012). Why the evidence-based paradigm in early childhood education and care is anything but evident. European Early Childhood Education Research Journal, 20(4),537-552. Vandenbroucke, F. (2013). The active welfare state revisited. Brugge, Belgium: Die Keure. Vandenbroucke, F., Hemerijck, A., & Palier, B. (2011). The EU needs a social investment pact. Observatoir Social Européen Opinion Paper No. 5. Brussels, Belgium: European Social Observatory. Vandenbroucke, F., & Vinck, J. (2013). Child poverty risks in Belgium, Wallonia and Flanders: accounting for a worrying performance. In P. Maystadt, E. Cantillon, L. Denayer, P. Pestieau, B. Van der Linden & M. Cattelain (Eds.), Le modèle social belge: quel avenir? (pp. 85-144). Charleroi, Belgium: Presses Interuniversitaires de Charleroi. Vinck, J., & Brekke, I. (forthcoming). Gender and education inequalities in parental employment and earnings when having a child with increased care needs: Belgium versus Norway. Journal of European Social Policy, 00(0), 1-14. Vinck, J., Lebeer, J., & Van Lancker, W. (2019). Non-take up of the supplemental child benefit for children with a disability in Belgium: A mixed method approach. Social Policy and Administration, 53(3), 357-384. Vinck, J., & Van Lancker, W. (2020). An intersectional approach towards parental employment in families with a child with a disability: The case of Belgium. Work, Employment and Society, 34(2), 228-261.

290

Vinck, J., Van Lancker, W., & Cantillon, B. (2017). Belgium: Creeping vulnerability of children. In B. Cantillon, Y. Chzhen, S. Handa & B. Nolan (Eds.), Children of austerity: The impact of the Great Recession on child poverty in rich countries (pp. 30-55). Oxford, UK: The United Nations Children’s Fund and Oxford University Press. Vornholt, K., Villotti, P., Muschalla, B., Bauer, J., Colella, A., Zijlstra, F., … Corbière, M. (2018). Disability and employment–overview and highlights. European Journal of Work and Organizational Psychology, 27(1), 40-55. Wade, D. T., & Halligan, P. W. (2017). The biopsychosocial model of illness: a model whose time has come. Clinical Rehabilitation, 31(8), 995-1004. Wagmiller, R. L., & Adelman, R. M. (2009). Childhood and intergenerational poverty: The long-term consequences of growing up poor. New York, United States: National Center for Children in Poverty, Columbia University, Mailman School of Public Health. Walby, S. (2007). Complexity theory, systems theory, and multiple intersecting social inequalities. Philosophy of the Social Sciences, 37(4), 449-470. Walby, S., Armstrong, J., & Strid, S. (2012). Intersectionality: multiple inequalities in social theory. Sociology, 46(2), 224-240. Ward, T., & Özdemir, E. (2013). Measuring low work intensity - an analysis of the indicator Improve Discussion Paper No. 13/09. Antwerp, Belgium: Herman Deleeck Centre for Social Policy. Warren, J. R. (2009). Socioeconomic status and health across the life course: a test of the social causation and health selection hypotheses. Social Forces, 87(4), 2125-2153. Wasi, N., van den Berg, B., & Buchmueller, T. C. (2012). Heterogeneous effects of child disability on maternal labor supply: evidence from the 2000 US census. Labour Economics, 19(1), 139-154. WHO. (2001). International Classification of Functioning, Disability and Health. Geneva, Switzerland: WHO. WHO. (2007). International Classification of Functioning, Disability and Health: Children & Youth Version. Geneva, Switzerland: WHO. Wilkinson, R., & Marmot, M. (Eds.). (2003). Social determinants of health: The solid facts. 2nd edition. Copenhagen, Denmark: WHO.

291

Wing, L. (1980). Childhood autism and social class: A question of selection?. The British Journal of Psychiatry, 137, 410–417. Yuval-Davis, N. (2006). Intersectionality and feminist politics. European Journal of Women's Studies, 13(3), 193-209. Zaidi, A., & Burchardt, T. (2005). Comparing incomes when needs differ: Equivalization for the extra costs of disability in the U.K.. Review of Income and Wealth, 51(1), 89-114. Zhu, A. (2016). Maternal employment trajectories and caring for an infant or toddler with a disability. Applied Economics, 48(48), 4606-4621. Zuccotti, C. V., & O'Reilly, J. (2019). Ethnicity, gender and household effects on becoming NEET: An intersectional analysis. Work, Employment and Society, 33(3), 1-23.

292

CONTRIBUTIONS PER CHAPTER

The majority of the chapters in this thesis are the result of enriching collaborations with co-authors. In what follows, the contributions of each author and those who indirectly advanced the manuscript are specified.

Chapter 1 Belgium: creeping vulnerability of children

Julie Vinck analysed the data and wrote the section on child poverty trends, its determinants and the impact of the crisis on the child poverty record. Julie incorporated the feedback into the manuscript.

Wim Van Lancker developed the section on the nature of the crisis in Belgium. In collaboration with Bea Cantillon and with help from Julie, Wim drafted the section on the policy discourse.

All authors contributed to final version of the manuscript.

The authors are indebted to the editors and authors of “Children of austerity: impact of the Great Recession on child poverty in rich countries” for their helpful feedback on (earlier versions) of the manuscript.

Chapter 2 Non-take-up of the supplemental child benefit for children with a disability in Belgium: a mixed-method approach

Julie Vinck took the lead in this chapter. Julie did the literature review, developed the policies section, analysed the administrative data, carried out the expert interviews, examined the qualitative information of all interviews, drafted and revised the manuscript.

Jo Lebeer critically analysed the criteria used in the assessment scale to detect potential gaps at the level of the benefit.

293

Wim Van Lancker contributed to the theoretical framework of the study, and provided critical feedback on the qualitative part of the study and on the manuscript as a whole.

The authors are grateful to Dorien Elsen for helping to conduct the parent interviews, and to colleagues at the Centre for Social Policy, the Centre for Sociological Research and two anonymous reviewers for their valuable remarks and ideas.

Chapter 3 An intersectional approach towards parental employment in families with a child with a disability: the case of Belgium

Julie Vinck took the lead in this chapter. Julie did the literature review, analysed the data, drafted and revised the manuscript.

Wim Van Lancker extensively discussed the theoretical framework with Julie and provided critical feedback on the manuscript.

The authors thank Eva Lefevere for pointing out an incorrectness in the administrative data, and Dorien Frans, colleagues at the Centre for Social Policy and three anonymous reviewers for their useful comments and suggestions.

Chapter 4 Gender and education inequalities in parental employment and earnings when having a child with increased care needs: Belgium versus Norway

Julie Vinck took the lead in this chapter. Julie conceived of the presented idea, setup of the method and presentation of the results. Julie did the literature review, drafted and revised the manuscript.

Idunn Brekke provided the Norwegian administrative data and gave critical feedback on the manuscript.

294

Both authors contributed to the theoretical framework, carried out the data analyses in close collaboration and discussed the results.

The authors are grateful to Gerlinde Verbist, Zach Parolin, Eva Lefevere, colleagues at Norwegian Social Research, participants of the FISS 2018 and ESPAnet 2018 conferences, and two anonymous reviewers for their valuable remarks and ideas.

Chapter 5 Income poverty among children with a disability in Belgium: the interplay between parental employment, social background and targeted cash support

Julie Vinck is thankful to Wim Van Lancker and Bea Cantillon for their critical feedback on an earlier version of the manuscript and to the participants of the FISS 2019 conference for their much appreciated comments and suggestions.

295

NEDERLANDSTALIGE SAMENVATTING

Het Belgische systeem van gerichte financiële ondersteuning voor kinderen met een handicap, namelijk de verhoogde kinderbijslag, slaagt erin het risico op inkomensarmoede van deze kinderen te verminderen. Deze doctoraatsthesis toont aan dat het inkomensarmoederisico voor kinderen met een handicap zelfs lager is dan voor kinderen zonder handicap, ook al werken de ouders van kinderen met een handicap vaak minder en hebben ze vaker een kwetsbare sociale achtergrond. Die verminderde tewerkstelling van ouders van kinderen met een handicap wordt trouwens voornamelijk verklaard door hun kwetsbare sociale achtergrond, en minder door de handicap van hun kind. Een armoede-indicator die gebaseerd is op het inkomen is echter niet noodzakelijk een goede weergave van de levensstandaard van deze gezinnen, aangezien de handicap van het kind hogere medische kosten en zorgkosten met zich meebrengt. Bovendien ontvangen niet alle kinderen met een handicap de verhoogde kinderbijslag. Dat ondergebruik (ook non-take-up genoemd) ondermijnt het volledige armoedebestrijdingspotentieel. De armoederisico's die hier worden getoond zijn dus waarschijnlijk onderschat.

Dat is de belangrijkste conclusie van deze doctoraatsthesis over de armoedepuzzel bij kinderen met een handicap in België. Eerder onderzoek in de literatuur over handicaps en de ontwikkeling van kinderen toont aan dat er in verschillende landen een duidelijk verband is tussen handicap en armoede bij kinderen. Maar dat onderdeel van de literatuur staat los van de sociale beleidsliteratuur over kinderarmoede. De huidige beleidsstrategieën ter bestrijding van kinderarmoede zijn vaak geworteld in het sociale investeringsparadigma en focussen zich onder meer op de arbeidsmarktintegratie van de ouders. Dat is in meerdere landen zo. Maar voor gezinnen van kinderen met een handicap lijkt zo’n

297 tewerkstellingsstrategie problematisch (Cantillon & Van Lancker, 2013). Dat heeft te maken met het feit dat deze gezinnen (1) meer zorg moeten bieden, wat hun arbeidsmarktparticipatie direct belemmert, en (2) vaak een kwetsbare sociale achtergrond hebben die op zich al nadelig is voor hun tewerkstellingskansen en armoederisico's (Brown & Clark, 2017; Shahtahmasebi et al., 2011; Stabile & Allin, 2012).

Welke rol spelen de tewerkstellingspatronen van ouders dan bij het verklaren van het risico op inkomensarmoede bij kinderen met een handicap? En in welke mate vermindert gerichte financiële ondersteuning voor deze kinderen hun inkomensarmoederisico? Daarover bestond tot nu toe nog geen duidelijkheid. Deze doctoraatsthesis gaat in op die vragen. Ik onderzoek hoe het samenspel tussen de handicap van het kind, de ouderlijke tewerkstelling, de sociale achtergrond en het ontvangen van gerichte financiële ondersteuning het inkomensarmoederisico van kinderen met een handicap beïnvloedt. Daarmee verenig ik twee onderdelen van de literatuur die over het algemeen van elkaar zijn losgekoppeld. Ik maak gebruik van unieke en grootschalige administratieve gegevens afkomstig van het Datawarehouse Arbeidsmarkt en Sociale Bescherming (steekproef 31 december 2010), aangevuld met gegevens van het Vlaams Agentschap voor Personen met een Handicap (VAPH) en van de Census van 1 januari 201133. De gegevens bevatten ook een controlegroep van kinderen zonder handicap.

33 In Hoofdstuk 2 wordt eveneens gebruikgemaakt van kwalitatieve informatie uit 22 semigestructureerde interviews met experten en ouders van kinderen met een autismespectrumstoornis of gedragsstoornis. In Hoofdstuk 4 worden de Belgische administratieve gegevens vergeleken met soortgelijke Noorse administratieve gegevens. Ten slotte werk ik met de Belgische cross-sectionele steekproef van de EU-SILC in Hoofstuk 1 (enquêtejaren 2006-2014) en Hoofstuk 5 (enquêtejaar 2011).

298

De doctoraatsthesis is opgebouwd uit vijf hoofdstukken waarvan de eerste vier zijn verschenen in internationale wetenschappelijke tijdschriften en als boekhoofdstuk. Het eerste hoofdstuk focust op inkomensarmoede bij kinderen in het algemeen en gaat in op de trends en determinanten ervan in België. Tegen die achtergrond moet de rest van deze doctoraatsthesis worden gelezen. In de overige hoofstukken ligt de focus op kinderen met een handicap. Hoofdstuk 2 richt zich op de vraag of het Belgische systeem van verhoogde kinderbijslag kampt met ondergebruik en wat de omvang, kenmerken en determinanten daarvan zijn. In Hoofdstuk 3 wordt onderzocht of de arbeidsmarktparticipatie van ouders met een kind met een handicap verklaard kan worden door de handicap van het kind, de sociale achtergrond van het gezin of door beide. In Hoofdstuk 4 wordt de tewerkstelling van moeders in tweeoudergezinnen vergeleken met die van vaders, zowel in gezinnen met een kind met een handicap als zonder een kind met een handicap. Daarbij maak ik ook een vergelijking tussen België en Noorwegen aan de hand van vergelijkbare administratieve gegevens. In Hoofdstuk 5 wordt het verband tussen handicap en armoede bij kinderen bekeken: gaat een handicap bij kinderen gepaard met kinderarmoede in België en waarom wel of niet? Dit hoofdstuk brengt de inzichten uit alle voorgaande hoofdstukken samen.

1. Definities van de centrale concepten

Om vast te stellen of een kind in armoede leeft, gebruik ik in deze doctoraatsthesis de 60% armoederisico-indicator op basis van het inkomen, de belangrijkste maatstaf voor het meten van inkomensarmoede in Europa. Deze indicator meet kinderarmoede als het aandeel kinderen jonger dan 18 jaar die in een gezin leven

299 met een equivalent34 netto beschikbaar gezinsinkomen onder een armoedegrens die is vastgesteld op 60% van het nationale mediaan equivalent netto beschikbaar gezinsinkomen. In deze thesis verwijst “inkomensarmoede” naar deze indicator.

Kinderen met een handicap worden gedefinieerd als kinderen die de verhoogde kinderbijslag35 ontvangen. Dat is een toeslag binnen het reguliere kinderbijslagsysteem voor kinderen met verhoogde zorgnoden, onder wie vooral kinderen met een handicap. Om recht te hebben op deze toeslag moeten kinderen aan drie voorwaarden voldoen: ze moeten in aanmerking komen voor de reguliere kinderbijslag, ze moeten jonger zijn dan 21 jaar en hun handicap moet worden erkend door de artsen van de Federale Overheidsdienst (FOD) Sociale Zekerheid. Daarvoor beoordelen de artsen de extra zorgnoden van de kinderen als gevolg van hun handicap en kennen ze hen een score toe op een 36-puntenschaal, waarvoor ze gebruikmaken van gestandaardiseerde criteria. De schaal bestaat uit drie pijlers die de ernst van de handicap van het kind weergeven in termen van (1) de fysieke en mentale gevolgen van de handicap (maximaal 6 punten), (2) de gevolgen voor de deelname van het kind aan het dagelijks leven (maximaal 12 punten) en (3) de gevolgen voor het gezin (maximaal 18 punten). De hoogte van de toeslag is afhankelijk van de score die het kind heeft op de 36-puntenschaal, variërend van € 80 tot meer dan € 500 per maand. In 2015 ontving 2.4% van de kinderen onder de 21 jaar de verhoogde kinderbijslag (FAMIFED, 2016). Handicap bij kinderen wordt dus geoperationaliseerd via een administratieve erkenning van de handicap zoals geëvalueerd door de artsen van de FOD Sociale Zekerheid.

34 Om rekening te houden met de grootte en de samenstelling van het gezin worden de inkomens gestandaardiseerd met de aangepaste OESO-equivalentieschaal (Hagenaars et al., 1994). 35 Sinds de regionalisering werd deze toeslag omgedoopt tot de “zorgtoeslag voor kinderen met een specifieke ondersteuningsbehoefte” binnen het Groeipakket in Vlaanderen. Het beoordelingsinstrument is tot nu toe ongewijzigd gebleven.

300

De arbeidsmarktparticipatie van de ouders wordt in Hoofdstuk 1, 3 en 5 gezamenlijk gemeten met een andere belangrijke Europese indicator, de “huishoudwerkintensiteit”. Deze indicator geeft aan in welke mate alle gezinsleden op actieve leeftijd (18-59 jaar, met uitzondering van studenten 18-24 jaar) samen deelnemen aan betaald werk. Dat wordt gedefinieerd als de verhouding tussen hun totale aantal gewerkte maanden (in voltijdsequivalenten) en het totale aantal maanden dat zij in theorie hadden kunnen werken (Ward & Özdemir, 2013). De verhouding varieert van nul tot één, waarbij nul betekent dat geen van de gezinsleden op actieve leeftijd in het bestudeerde jaar betaald werk heeft verricht, terwijl een score van één aangeeft dat al deze gezinsleden het hele jaar voltijds hebben gewerkt. In Hoofdstuk 4 wordt de arbeidsmarktpositie van moeders en vaders afzonderlijk onderzocht, waarbij de minder gedetailleerde tewerkstellingsstatus (ja of nee) wordt gebruikt.

De sociale achtergrond van het gezin waartoe de kinderen behoren, wordt gemeten aan de hand van vier indicatoren: het gezinstype (eenoudergezin of tweeoudergezin), het hoogste opleidingsniveau van een van de ouders (laaggeschoold, gemiddeld geschoold of hooggeschoold)36, het geboorteland van de ouders (België, EU27 of niet-EU27)37 en de aanwezigheid van andere gezinsleden met een handicap38 (ja of nee).

36 Voor deze opdeling wordt de internationale classificatie van het onderwijsniveau (ISCED) gebruikt. Laaggeschoold is ISCED 0-2 (lager secundair onderwijs of lager), gemiddeld geschoold is ISCED 3-4 (secundair onderwijs), en hooggeschoold is ISCED 5-6 (hoger onderwijs). 37 Als ten minste één ouder in België of in een ander EU27 land werd geboren, krijgt het gezin respectievelijk een Belgische of EU27-migratieachtergrond toegewezen. Wanneer beide ouders buiten de EU27 zijn geboren, krijgt het gezin een niet-EU27-migratieachtergrond toegekend. 38 Opnieuw gedefinieerd aan de hand van het ontvangen van een handicap-specifieke uitkering.

301

2. Achtergrond: kinderarmoede in België

België wordt geconfronteerd met een relatief hoog kinderarmoederisico en een toenemende kwetsbaarheid van kinderen die begon voor de financiële en economische crisis van 2008 (zie Hoofdstuk 1 en Vandenbroucke & Vinck, 2013). Vandaag loopt 20.6% van de min 18-jarigen het risico op inkomensarmoede in België. Dat ligt in lijn met het Europees gemiddelde (20.3%) (EUROSTAT, 2020a). De kinderarmoede neemt bovendien toe: 12 jaar geleden bedroeg het inkomensarmoederisico bij kinderen in België 15.3%. Daarnaast is het risico op inkomensarmoede voor kinderen hoger dan voor de rest van de bevolking. Deze kinderarmoedecijfers en -tendensen zijn teleurstellend én zorgwekkend in een land met zo'n gevestigde en sterk ontwikkelde welvaartsstaat.

Dat België in Europees vergelijkend perspectief zo middelmatig presteert en het kinderarmoederisico gestaag toeneemt, is grotendeels toe te schrijven aan de tewerkstellingspatronen van de gezinnen waartoe kinderen behoren. Ons land wordt namelijk gekenmerkt door een "dubbele polarisatie": een groot deel van de kinderen leeft in een gezin waar niemand werkt en tegelijkertijd worden deze kinderen geconfronteerd met een (erg) hoog inkomensarmoederisico. Vandaag leeft bijna de helft van de inkomensarme kinderen in een gezin waar niemand werkt, alleen Ierland doet het slechter (EUROSTAT, 2020a, 2020b, 2020d). Naast gezinnen zonder werk lopen kinderen van alleenstaande ouders, ouders met een migratieachtergrond en ouders met een lager opleidingsniveau een groter risico op inkomensarmoede. De financiële ondersteuningsmaatregelen voor de bevolking in het algemeen en specifiek voor gezinnen met kinderen doen het kinderarmoederisico wel dalen, maar het herverdelende effect van deze maatregelen is door de jaren heen afgenomen in België.

302

3. Wat leert deze doctoraatsthesis ons over…

3.1. … de sociale achtergrond van kinderen met een handicap?

De eerste conclusie die uit deze doctoraatsthesis naar voren komt, is dat kinderen met een handicap in België vaker in een gezin met een kwetsbare sociale achtergrond wonen (zie Hoofdstuk 3). Dat geldt voor drie van de vier indicatoren die worden bestudeerd. Kinderen met een handicap leven in vergelijking met kinderen zonder handicap vaker samen met alleenstaande ouders, ouders met een laag of gemiddeld opleidingsniveau en andere gezinsleden die ook een handicap hebben. Dat ligt in lijn met onderzoek in andere landen, zie bijvoorbeeld Bauman et al. (2006) voor de Verenigde Staten en Blackburn et al. (2010) voor het Verenigd Koninkrijk. Voor de vierde indicator die in de Belgische context van belang is, namelijk de migratieachtergrond van de ouders, wordt er geen oververtegenwoordiging van kinderen met een handicap gevonden.

3.2. … de arbeidsmarktparticipatie van ouders van kinderen met een handicap?

De tweede conclusie die uit deze doctoraatsthesis kan worden getrokken, is dat ouders van kinderen met een handicap minder werken dan ouders van kinderen zonder handicap (zie Hoofdstuk 3 en 4). Deze tewerkstellingskloof neemt toe met de ernst van de handicap van het kind (zie Hoofdstuk 3). Bovendien speelt dit sterker bij moeders dan bij vaders (zie Hoofdstuk 4). Dat bevestigt de bevindingen uit eerdere studies (Brown & Clark, 2017; Stabile & Allin, 2012).

303

Zodra een intersectionele aanpak39 wordt toegepast om de tewerkstellingskloof bij gezinnen van kinderen met een handicap te ontwarren, wordt het duidelijk dat de kloof slechts gedeeltelijk wordt verklaard door hun kwetsbare sociale achtergrond (zie Hoofdstuk 3). De arbeidsmarktparticipatie van ouders is lager in alle gezinnen van kinderen met een handicap. Maar voor sommige sociaal kwetsbare groepen is de tewerkstellingskloof groter. Bij alleenstaande ouders, ouders met een laag of gemiddeld opleidingsniveau en ouders met meerdere kinderen met een handicap versterken de handicap van het kind en de sociale achtergrond van het gezin elkaar. Voor ouders met een migratieachtergrond, ouders die zelf een handicap hebben of die samenwonen met andere volwassenen met een handicap kon geen versterking worden vastgesteld.

Bovendien blijkt uit een vergelijking met Noorwegen dat de omvang van de tewerkstellingskloof afhangt van de institutionele context van de welvaartsstaat. De kloof is namelijk aanzienlijk groter in België dan in Noorwegen (zie Hoofdstuk 4). Het gendergelijke beleid op het vlak van tewerkstelling en gezin van de Noorse welvaartsstaat lijkt vruchten af te werpen.

3.3. … het inkomensarmoederisico van kinderen met een handicap?

In lijn met het algemene kinderarmoedeonderzoek (zie Hoofdstuk 1) zijn de tewerkstellingspatronen van de ouders en de sociale achtergrond van het gezin belangrijk om het risico op inkomensarmoede bij kinderen met een handicap te begrijpen (zie Hoofdstuk 5). Maar de derde en meest verrassende conclusie die uit deze doctoraatsthesis voortkomt, is dat kinderen met een handicap een lager risico

39 Intersectionaliteit veronderstelt dat meerdere categorieën van ongelijkheid, zoals etniciteit, klasse, geslacht of handicap, niet alleen onafhankelijk van elkaar de tewerkstellingskansen beïnvloeden, maar elkaar ook kunnen versterken.

304 op inkomensarmoede lopen dan kinderen zonder handicap in België, zelfs wanneer er rekening wordt gehouden met de lagere tewerkstelling en de kwetsbare sociale achtergrond van hun ouders (zie Hoofdstuk 5). Deze contra-intuïtieve bevinding kan worden verklaard door de gerichte financiële ondersteuning die gezinnen met kinderen met een handicap ontvangen: de verhoogde kinderbijslag. Als gezinnen alleen de reguliere kinderbijslag zouden ontvangen, dus geen verhoogde kinderbijslag, dan zouden kinderen met een handicap een even groot risico op inkomensarmoede lopen als kinderen zonder handicap, zodra de arbeidsmarktparticipatie en de sociale achtergrond van hun gezin in rekening worden gebracht. De verhoogde kinderbijslag voor kinderen met een handicap heeft een sterk armoedebestrijdend effect, aangezien deze toeslag vooral terechtkomt bij gezinnen die een grotere kans hebben om in inkomensarmoede te leven, namelijk gezinnen met een lagere arbeidsmarktparticipatie en gezinnen met een kwetsbare sociale achtergrond. Samengevat: naast de tewerkstellingspatronen van de ouders en de sociale achtergrond van het gezin, zijn de gerichte financiële ondersteuning en wie die daadwerkelijk ontvangt van groot belang om de armoedepuzzel bij kinderen met een handicap te ontrafelen.

Maar er moet worden benadrukt dat niet alle kinderen met een handicap de verhoogde kinderbijslag ontvangen. Dit ondergebruik ondermijnt het volledige armoedebestrijdingspotentieel van de toeslag. Door de administratieve erkenningen voor de verhoogde kinderbijslag te vergelijken met de administratieve erkenningen bij het VAPH, het agentschap dat verantwoordelijk is voor de gesubsidieerde zorgvoorzieningen en aanvullende financiële ondersteuning, blijkt dat minstens 10% van de kinderen met een erkende handicap in Vlaanderen géén gebruik maakt van de verhoogde kinderbijslag, voornamelijk kinderen met een autismespectrumstoornis, een intellectuele stoornis of een

305 psychische stoornis (zie Hoofdstuk 2). Dit ondergebruik is het gevolg van onvoldoende informatieverstrekking over de toeslag door eerstelijnsorganisaties en artsen, de complexiteit van de aanvraagprocedure en de criteria om te beoordelen of het kind in aanmerking komt.

Bovendien is het belangrijk om in het achterhoofd te houden dat een op inkomen gebaseerde armoede-indicator niet noodzakelijk een goede weergave is van de levensstandaard van gezinnen van kinderen met een handicap, aangezien zij te maken hebben met hogere medische kosten en zorgkosten voor de handicap van het kind (zie bijvoorbeeld Stabile & Allin, 2012). Met andere woorden: deze gezinnen worden geconfronteerd met een extra druk op het huishoudbudget.

4. Wat niet onderzocht kon worden in deze doctoraatsthesis

Het uitgevoerde onderzoek heeft een aantal beperkingen. Ten eerste is de definitie van handicap beperkt tot kinderen met een administratief erkende handicap voor de verhoogde kinderbijslag. Maar deze toeslag heeft zoals eerder vermeld te kampen met ondergebruik. De bevindingen met betrekking tot de sociale achtergrond en de ouderlijke tewerkstelling bij gezinnen met een kind met een handicap uit Hoofdstuk 3 en 4 zijn evenwel in lijn met wat gerapporteerd wordt door Debacker (2007), Sebrechts en Breda (2012) en Van Landeghem et al. (2007) op basis van Vlaamse steekproefgegevens40. Aangezien deze auteurs een bredere handicapdefinitie toepassen die ook kinderen met verhoogde zorgnoden omvat die niet worden erkend voor de verhoogde kinderbijslag, is het dus aannemelijk dat de eerste en tweede conclusie van deze doctoraatsthesis niet noodzakelijk vertekend zijn door het selectieprobleem.

40 De GEzin en ZOrg (GEZO) data uit 2004-2005.

306

Voor de derde conclusie, over het risico op inkomensarmoede, zou een andere operationalisering van kinderen met een handicap waarschijnlijk niet tot hetzelfde resultaat leiden, aangezien juist de verhoogde kinderbijslag onontbeerlijk is voor de inkomensbescherming van deze gezinnen. Sterker nog, op basis van de Vlaamse steekproefgegevens toont Debacker (2007) dat het risico op inkomensarmoede hoger is voor kinderen met een handicap dan voor kinderen zonder handicap. Ook de EU-SILC 2017 ad-hocmodule over gezondheid van kinderen wijst hierop: in de meeste Europese landen (waaronder België) zijn kinderen die matig tot ernstig beperkt zijn in hun activiteiten vanwege gezondheidsproblemen oververtegenwoordigd onder de inkomensarme gezinnen41 (EUROSTAT, 2020c). De risico’s op inkomensarmoede voor kinderen met een handicap uit deze doctoraatsthesis vormen dus waarschijnlijk een ondergrens.

Ten tweede houdt een op inkomen gebaseerde armoede-indicator geen rekening met de kosten die gezinnen moeten maken. Ook de manier waarop inkomens vergelijkbaar worden gemaakt tussen gezinnen gaat er indirect van uit dat gezinnen met hetzelfde equivalent inkomen dit inkomen kunnen vertalen in dezelfde levensstandaard. Voor gezinnen van kinderen met een handicap verbergt dergelijke armoede-indicator waarschijnlijk belangrijke aspecten van de levensomstandigheden van de kinderen, aangezien de handicap van het kind hogere kosten met zich meebrengt. Het kan goed zijn dat een gezin met een kind met een handicap en een inkomen boven de armoedegrens in feite een lagere levensstandaard heeft dan een inkomensarm gezin zonder een kind met een handicap. Waarschijnlijk worden ook daardoor de armoederisico’s uit deze

41 De steekproefomvang is echter beperkt. In België ervaren slechts 139 van de bijna 2,600 kinderen jonger dan 16 jaar matige tot ernstige beperkingen in hun activiteiten, oftewel 5.0%. In de hele Europese Unie is dit het geval voor 4,376 van de 90,832 nul tot 15-jarigen, oftewel 4.7%.

307 doctoraatsthesis onderschat. Om de levensstandaard van gezinnen met kinderen met een handicap goed te begrijpen, zijn aanvullende indicatoren nodig die rekening houden met deze kosten. Daarvoor is informatie nodig over wat deze gezinnen daadwerkelijk uitgeven (d.w.z. op uitgaven gebaseerde indicatoren) of (minimaal) nodig hebben (d.w.z. referentiebudget indicatoren), maar zulke gegevens ontbreken nog grotendeels.

Ten derde kon het gebruik van onderwijs en opvangvoorzieningen, beschikbaar voor de hele kinderbevolking en specifiek voor kinderen met een handicap, niet in de analyses worden meegenomen42. Deze diensten zijn voor gezinnen van kinderen met een handicap vermoedelijk cruciaal om werk en gezinsleven met elkaar te combineren. Wat de kinderopvang betreft, wijst het bestaande onderzoek voor Vlaanderen erop dat gezinnen van kinderen met een handicap er slechts beperkt gebruik van maken (Kind & Gezin, 2018; Van Landeghem et al., 2007). Bij de handicap-specifieke zorgondersteuning zijn zowel de (semi)residentiële of ambulante zorgvoorzieningen als de ondersteuning binnen het buitengewoon of geïntegreerd onderwijs van belang voor de combinatie van werk en gezin. Maar voor Vlaanderen geldt dat minder dan de helft van de kinderen die erkend zijn voor de verhoogde kinderbijslag ook erkend zijn bij het VAPH en dus van gesubsidieerde zorgvoorzieningen gebruikmaken of aanvullende financiële ondersteuning ontvangen (zie Hoofdstuk 2). Van Landeghem et al. (2007) tonen aan dat zelfs in gezinnen die buitengewoon onderwijs en (semi)residentiële zorg

42 Het Vlaamse beleidslandschap voor personen met een handicap wordt gekenmerkt door een veranderende beleidslogica van een zorggerichte naar een ondersteuningsgerichte aanpak (Roets et al., 2020). Dit komt tot uiting in een verschuiving van aanbodfinanciering naar vraagfinanciering (voor kinderen is deze transitie nog niet voltooid) en een verschuiving naar welzijnspluralisme en subsidiariteit. Deze laatste verschuiving, waarbij de hoofdverantwoordelijkheid voor de zorg en ondersteuning (terug) wordt overgedragen aan personen met een handicap en hun informele netwerken, in dit geval ouders van kinderen met een handicap, kan op gespannen voet staan met de heersende oplossing voor kinderarmoede door de integratie van ouders in de arbeidsmarkt.

308 combineren bijna de helft van de moeders niet werkt. Het gebruik van handicap- specifieke zorgvoorzieningen stelt ouders dus niet noodzakelijk in staat om betaald werk te verrichten. Voor de gezinnen die geen beroep doen op deze diensten kan de arbeidsmarktparticipatie nog moeilijker zijn.

Ten vierde maakt het cross-sectionele karakter van de gegevens het alleen mogelijk om verbanden te bestuderen op één moment in de tijd, niet om causale mechanismen uit te pluizen tussen de handicap bij kinderen, kwetsbare sociale achtergrond, arbeidsmarktparticipatie van de ouders en het inkomensarmoederisico.

Ten slotte is het comparatieve deel van deze doctoraatsthesis beperkt tot twee landen. Meer vergelijkend onderzoek tussen landen is dus wenselijk, zowel op het vlak van arbeidsmarktparticipatie van de ouders als op het vlak van inkomensarmoede bij gezinnen van kinderen met een handicap. We weten dat België vanuit Europees vergelijkend perspectief slechts matig presteert op het vlak van inkomensbescherming en armoedebestrijding bij gezinnen met kinderen (zie Hoofdstuk 1; Vandenbroucke & Vinck, 2013; Van Lancker & Van Mechelen, 2015). Maar hoe goed of slecht België presteert voor gezinnen met kinderen met een handicap in vergelijking met andere landen is nog onontgonnen terrein. Daarvoor zijn toereikende, betrouwbare en vergelijkbare gegevens nodig.

5. Wat beleidsmakers kunnen doen

Het zou stilaan duidelijk moeten zijn dat een sociale investeringsstrategie om kinderarmoede te bestrijden door ouders te integreren in de arbeidsmarkt wellicht weinig soelaas biedt voor gezinnen van kinderen met een handicap, terwijl gerichte inkomensbescherming voor hen een cruciale rol speelt. Wat kunnen

309 beleidsmakers dan doen om (verdere) vooruitgang te boeken in de strijd tegen inkomensarmoede bij deze gezinnen? De analyses in deze doctoraatsthesis wijzen op drie elkaar versterkende beleidsaanbevelingen.

Ten eerste, aangezien gezinnen van kinderen met een handicap voor een extra uitdaging staan om werk en zorg te combineren, los van hun kwetsbare sociale achtergrond, zal de arbeidsmarktintegratie van deze gezinnen moeilijk te realiseren zijn zonder hen extra ondersteuning te bieden. Om te beginnen moet de toegang tot kwaliteitsvolle formele opvang die is afgestemd op de zorgbehoeften van de kinderen worden verbeterd, zodat hun ouders (een deel van) de zorg kunnen uitbesteden43. Dat kan alle ouders van kinderen met een handicap helpen om hun arbeidsmarktparticipatie te hervatten, te behouden of te vergroten, en zo de tewerkstellingskloof die zij hebben ten opzichte van ouders van kinderen zonder handicap te verkleinen. Maar aangezien de tewerkstellingskloof voor sommige sociaal kwetsbare groepen groter is, moeten de opvangvoorzieningen zeker voor deze groepen toegankelijk zijn. Dat is vandaag nog niet altijd het geval (zie bijvoorbeeld Gambaro et al., 2014; Van Lancker, 2013). Daarnaast kan ook een betere ondersteuning op de werkvloer nuttig zijn. Door ouders bijvoorbeeld meer flexibiliteit in het organiseren van hun werktijd te bieden, kunnen zij hun werk beter combineren met de grotere zorgnoden van hun kind (Brown & Clark, 2017).

Ten tweede, meer ondersteuning alleen zal niet genoeg zijn voor gezinnen van kinderen met een handicap. Omdat kinderen met een handicap vaak in gezinnen wonen met een kwetsbare sociale achtergrond die de arbeidsmarktparticipatie van de ouders op zich al belemmert, zal het activeren van deze kwetsbare gezinnen

43 Er moet worden opgemerkt dat sommige ouders, en met name ouders van kinderen met een handicap, ervoor kiezen om niet (volledig) deel te nemen aan betaald werk, omdat ze liever zelf voor hun kinderen zorgen.

310 ook van cruciaal belang zijn. België doet het op dat vlak niet erg goed (Cantillon, 2016). Daarom moeten de kansen op de arbeidsmarkt worden verbeterd voor alleenstaande ouders, lager opgeleiden, mensen met een migratieachtergrond en mensen met een handicap. Als het arbeidsmarktbeleid erin slaagt om deze kwetsbare groepen tewerkstellingskansen te bieden, én de omstandigheden te scheppen om die kansen te grijpen, dan zal dat ook gezinnen van kinderen met een handicap ten goede komen.

Ten derde, hoewel de lagere tewerkstelling van ouders en de kwetsbare sociale achtergrond van kinderen met een handicap niet hoopgevend zijn voor hun armoederisico, worden hun gezinsinkomens beschermd door de gerichte financiële ondersteuning die zij ontvangen. De verhoogde kinderbijslag heeft een sterk armoedebestrijdend effect voor kinderen met een handicap, waardoor hun risico op inkomensarmoede onder het niveau van kinderen zonder handicap komt te liggen. Maar een aanzienlijke groep kinderen met een handicap ontvangt deze toeslag niet. Dit ondergebruik aanpakken, is dus cruciaal om verdere vooruitgang te boeken, vooral voor de meest kwetsbare kinderen. Daarvoor moet de informatieverstrekking over deze toeslag en de toekenningscriteria ervan aan eerstelijnsorganisaties, artsen en ouders worden verbeterd. Deze informatie zou zelfs op een proactieve manier kunnen worden verstrekt: bijvoorbeeld standaard wanneer een kind is ingeschreven in het buitengewoon onderwijs, voor lange tijd is opgenomen in het ziekenhuis, bij het VAPH wordt erkend voor zorgondersteuning of aanvullende financiële ondersteuning, wordt gediagnosticeerd door een specialist of wordt opgevolgd door een huisarts. Daarnaast lijkt een herziening van het beoordelingsinstrument nodig, omdat de gehanteerde criteria nog steeds voor een groot deel medisch zijn en bovendien te veel ruimte laten voor interpretatie door de artsen. Dat is nadelig voor kinderen

311 met minder “zichtbare” handicaps zoals een autismespectrumstoornis, een intellectuele stoornis of een psychische stoornis. De International Classification of Functioning, Disability and Health for Children and Youth kan en moet als kader dienen voor deze herziening. Tot slot zal meer coherentie in het beleidslandschap voor personen met een handicap bijdragen aan de bestrijding van het ondergebruik. De centralisatie van medische verslagen zou kunnen helpen, op voorwaarde dat ze recent en volledig zijn, net zoals een doordachte afstemming van de aanvraagprocedures of zelfs van de beoordelingsinstrumenten. Met de recente regionalisering van de kinderbijslag nam Vlaanderen stappen in deze richting, maar het is nog onduidelijk of dit in de praktijk inderdaad een vereenvoudiging voor de gezinnen betekent.

Kortom, alleen als beleidsmakers tegelijkertijd actie ondernemen op het vlak van sociale investeringen en sociale bescherming kan er verdere vooruitgang in de strijd tegen inkomensarmoede bij kinderen met een handicap worden geboekt.

312

THE POVERTY PUZZLE AMONG CHILDREN WITH A DISABILITY

The interplay between parental employment, social background and targeted cash support

Julie Vinck

Currently dominant policy strategies to combat child poverty are rooted in the social investment paradigm and focus on integrating parents into paid employment. Previous research has found a clear link between childhood disability and child poverty, but for families of children with a disability, such a work strategy might be problematic as they (1) need to provide more care which directly impedes their employment participation, and (2) often have a disadvantaged social background which independently affects their employment opportunities and poverty risk. Until now, we lacked insight into how the interplay between childhood disability, parental employment, social background and the receipt of targeted cash support affects the poverty risk of children with a disability. To fill this gap, I draw on a case study of Belgium making use of unique and large- scale administrative data. This thesis demonstrates that the Belgian system of targeted cash support for children with a disability succeeds in reducing their risk of income poverty. Even though their parents often work less and have a disadvantaged social background, the income poverty risk is lower for children with a disability than for children without a disability. However, an income-based poverty indicator is not necessarily a good representation of the standard of living for these families as they face higher medical and care costs incurred by the child’s disability. Moreover, the targeted cash support suffers from non-take- up, jeopardising its full poverty-reducing potential. Only if action is simultaneously taken in social investment and social protection policies, further progress can be made in the fight against income poverty among children with a disability.

Supervisors: prof. dr. Wim Van Lancker & prof. dr. Bea Cantillon