Introduction

Child Poverty and its drivers are heavily researched. The Social Metrics Commission (SMC) (2020) estimated that in the UK in 2018/19, 33% of children were in income poverty (compared to only 22% of working-age adults and 11% of pension-age adults). It is also known that families with children are more likely to be in poverty compared to those who are childless (Valletta, 2006). Research carried out by Carers UK (2019), using data from 2018/19, showed that an estimated 2.6 million people had to stop working due to caring responsibilities, implying a change in income. One such responsibility could be related to taking care of a person with disabilities. For example, the SMC (2020) indicates that 50% of those in income poverty are part of a family that supports at least one person with disabilities. A further breakdown, by the presence of disability, of people touched by income povertyFootnote 1 in the UK in 2018/19 is shown in Table 1.

Table 1 Composition of poverty in the UK (2018/19).
Full size table

The impact of poverty on children and their families can be serious and lasting (Adjei et al., 2022; Lesner, 2018; Notten and Roelen, 2011). For example, poverty can hinder access to education, which can, in turn, have lifelong consequences (Adamson, 2012). Moreover, the presence of poverty during childhood could enforce a cycle of intergenerational harm and prejudice (Blanden et al., 2013; McEwen and McEwen, 2017), meaning the damage created by poverty in childhood can increase the risk of disadvantage in adulthood.

One of the factors associated with families being in poverty is the provision of informal care, also referred to as unpaid care. As noted in the carers’ section on the NHS’s (2021) website, providing some form of care is something the majority of people will experience at some point during their lifetimes. Parenting, in general, requires parents to spend a certain amount of time on childcare-related activities (Dotti Sani and Treas, 2016), which could impact parents’, especially mothers’, capacity for paid work (Ruppanner et al., 2019). Additionally, looking after children with extra needs can increase the pressure on a family’s financial well-being (Arora et al., 2020; Solmi et al., 2018). According to Dotti Sani and Scherer (2018), as children get older, parental childcare decreases, allowing mothers to be more engaged with the labour market. However, in the case of parents supporting a child with disabilities, it is reasonable to hypothesise that: (1) the time spent on childcare does not decrease (as much) as the child ages, and (2) these families are in a worse position in terms of financial well-being compared to those not providing for a child with disabilities (i.e., labour market constraints on parents, extra costs associated with disability, etc.).

To our knowledge, the three-way relationship between unpaid care, child poverty, and child disability has not been researched particularly in the context of the UK. Consequently, this study aims to evaluate this tripartite relationship and whether being in poverty is influenced by different factors when first stratifying data by the disability status of children and then by unpaid care provision by a resident family member. In line with previous research, two relative measures of child poverty will be used: the after housing costs (AHC) income of a family and a material deprivation index for children and their families. With this in mind, the following subsections will briefly discuss past research results related to each of the three components of the tripartite relationship: indicators of (child) poverty in the UK context, disability, and unpaid care. Subsequently, the data sources and analytical strategy used in the paper are presented, followed by a summary of the results. Finally, the analytical outcomes and their implications are discussed.

Indicators of (child) poverty: the UK context

The conceptualisation of poverty is contested, and measurements and definitions have long been debated, as pointed out by Lister (2021) in her book on poverty. Nevertheless, in the UK, Parliamentary acts were passed in which certain indicators are used to measure annual poverty levels. In this way, for example, the government can assess whether its goals to reduce poverty have been achieved and, concomitantly, whether the policies that they have implemented have been effective.

The UK’s Child Poverty Act 2010 set four main targets, using specific poverty indicators, for the government to achieve by 2020. However, the Conservative Government, elected in 2015, passed into law the Welfare Reform and Work Act 2016 and, in doing so, repealed the Child Poverty Act 2010 and all the indicators and targets that came with it. Consequently, the UK government no longer has a legal obligation to reduce income child poverty. Moreover, although reports using the indicators (absolute low income, relative low income, a combined measure of low income, and material deprivation) set in the Child Poverty Act 2010 are still created by the government, they are not presented to parliament. According to the Welfare Reform and Work Act 2016, the new official poverty indicators are the level of education attainment by children and parents ’worklessness’, with statistics on these indicators being presented to parliament. In doing so, an inaccurate representation of the levels of poverty in the UK is presented to those in power, leading to inadequate implementation of policy interventions. However, a paper by Stewart and Roberts (2019) which investigated feedback from a government consultation from 2012/13 on child poverty measurement, showed that only two out of 251 responses fully supported the removal of income from poverty measurement, while the majority believed material deprivation and income are central to poverty measurement. Moreover, becoming employed is no longer a guaranteed way out of poverty (Hick and Lanau, 2018), as in-work poverty has risen by 5% in Britain in the last 25 years (Bourquin et al., 2019). For example, according to a report by the Department of Work and Pensions (DWP) (2021b), in the 2020 financial year, 19% of children were part of a working family but were still in BHC income poverty. So, even if the parents are employed and the child does obtain good grades, that does not necessarily mean the family is not experiencing some type of poverty. Overall, income and/or material deprivation are appropriate and widely used indicators of poverty.

In terms of income poverty indicators, differences in poverty rates are observed between the before housing costs (BHC) indicator and the AHC one. Generally, BHC represents the income of a family, while AHC usually represents “disposable income”—the income a family is left with after housing-related costs (i.e., rent, some types of bills) are deducted from their overall income.

An Office for National Statistics (ONS) (2021) report on household expenditure showed that, between April 2019 and March 2020, in households headed by somebody under 30 years, 41% of spending was on housing and food. Moreover, a report by the Joseph Rowntree Foundation JFR (2020) found that 21.5% of the UK population was AHC poor but only 17.1% BHC poor in 2017/18. Additionally, AHC can be seen as a more reliable indicator of living standards as it accounts for the various types of housing available as well as the housing benefits associated with them (Bourquin et al., 2020). For example, two households might have the same BHC income, but if one of them is renting, their disposable income is going to be different from a household that owns their home, potentially leading those renting to be AHC poor. Depending on the aim of the study, thresholds to classify someone as poor are set at either 50, 60, or 70% of the median income. For example, The Organisation for Economic Cooperation and Development OECD (2017) uses 50 and/or 60%. In the case of the UK, 60% is the most common threshold (Brewer et al., 2007; Child poverty action group, 2021; Dickerson and Popli, 2018; Mack, 2016; McKnight et al., 2017). Also, the 60% threshold is the measure mostly used in official UK statistics on poverty (Francis-Devine, 2021). Overall, this paper uses both BHC and AHC measures as well as material deprivation to assess the poverty levels of families with children.

Child poverty and child disability

Children are considered to be income-poor when the household they reside in is classified as income-poor (Magadi, 2010). However, there might be families who are income-poor, but the children in these families are not necessarily deprived. For example, Main and Bradshaw (2016), using data from the 2012 Poverty and Social Exclusion (PSE) and both household income and individual deprivation indices, found cases in which struggling parents would put their children’s needs before their own, meaning children were not deprived of basic necessities. Additionally, data from the UK Millennium Cohort Study was analysed by Dickerson and Popli (2018), who constructed a composite multidimensional poverty (MP) indicator based on five dimensionsFootnote 2, and showed that in the fourth wave, 9.6% of children were MP but not income-poor and 11.7% were income-poor but not MP poor. These examples illustrate that the way we choose to measure poverty is crucial to how we understand and narrate it, as discussed in the previous section. With this in mind, this paper aims to further investigate differences between poverty measures for specific groups (i.e., children with disabilities and their families).

In the UK in 2017/18, families with a disabled child had a poverty rate of 26% (JFR, 2020) compared to 19% of families in which no disability was present. Disability is no longer defined solely in medical terms but rather through the social model in which social and environmental conditions are considered an essential part of the definition (Union of the Physically Impaired Against Segregation (UPIAS), 1976; World Health Organization, 2007; World Health Organization, 2001). The legal framework for defining disability in the UK takes into consideration the social model, and according to the Equality Act of 2010, a person (P) is classified as having a disability if: “P has a physical or mental impairment, and (b) the impairment has a substantial and long-term adverse effect on P’s ability to carry out normal day-to-day activities”. This definition is used by a majority of UK representative surveys to identify those who consider themselves as having a disability.

Disability is often associated with negative stereotypes and bias (Silverman and Cohen, 2014). Raising a child with disabilities not only has adverse effects on the child’s well-being and health (Emerson and Hatton, 2007) but also on the financial, social well-being, and work-life balance of parents (Yeandle and Valentine, 2013). The direct cost (i.e., medical bills, etc.) of a child’s disability is varied, depending on disability type, needs, location, resources available, etc. (Stabile and Allin, 2012). The UK government provides support to those looking after a child who has “difficulties walking or needs much more looking after than a child of the same age who does not have a disability” (GOV.UK, 2022) through the Disability Living Allowance benefit scheme. However, these benefits do not always cover the full cost of disability. For example, using data from a period between 2004/05 and 2011/12, Solmi et al. (2018) predicted that the family of a child with any mental health condition needed, on average, an additional £49.31 (e.g., medical expenses) per week to have similar living standards to an otherwise equivalent family with no children with disabilities and this amount was found to be even higher (£59.28 per week) for more deprived families (Solmi et al., 2018) (e.g., living in harder to reach locations). The calculation done in the study mentioned above included any disability benefits received by the family.

An important secondary question is whether there are specific sociodemographic characteristics usually associated with poverty. Determining which these are could improve the provision of help to those in need. Education level, job type, ethnicity, and household structure are some of the elements which can typically be associated with an individual’s income and material poverty. For example, Hick and Lanau (2018), using data from the longitudinal survey Understanding Society from 2010–14, showed that the probability of entering in-work income poverty increases for those with an education level below a first degree. Considering ethnicity/race, Iceland and Sakamoto (2022) showed that while the poverty/hardship gap between ethnicities narrowed over 27 years, there still are significant racial disparities. In terms of household structure, a recent econometric analysis by Antonelli and De Bonis (2021) of the Eurostat data showed that as the number of non-traditional families (i.e., extended families, single parents) increases, so does the poverty rate.

According to research by Main and Bradshaw (2016), on the 2012 PSE survey, lone parents were one of the groups most vulnerable to income poverty. A recent report on UK poverty by the Joseph Rowntree Foundation (2021) argues that of all family types, lone parents are still the ones with the highest levels of in-work poverty.

Focusing on child disability and poverty, several sociodemographic characteristics were identified in previous studies. For example, based on US data from 2004 to 2008, Ghosh and Parish (2013) identified an increased likelihood for households with single mothers, adult disability, or low levels of income to experience hardships such as: not being able to pay rent, bills, have enough food or needing to see a doctor/dentist but not going to one. Similarly, a longitudinal study of UK data from 2001 to 2005 showed that families supporting a child with disabilities were less likely to be income-poor, but they were more likely to experience hardship (activities and items the family cannot access) and financial strain (ability to save, worry about financial well-being) (Shahtahmasebi et al., 2011). However, the same study concluded that, when compared to other families with similar sociodemographic characteristics, the presence of a child with disabilities does not significantly affect poverty trajectories. However, importantly these studies do not take into account the direct effect that unpaid care might have on poverty levels in this context. Thus, this paper will assess the direct impact parental unpaid care has on poverty.

Unpaid care

It has long been observed that care, be it unpaid or not, has been recognised as necessary for both the social and economic development and overall well-being of a society (see, for example, Daly, 2001). For instance, caring for the sick, not as part of one’s job, in the short term can reduce the cost to the state. According to the Office for National Statistics (2018), in 2016 in the UK, informal childcare was estimated to have had a value of £351.7 billion, an equivalent of £5,358 per person, while informal adult care was estimated at £59.5 billion. In the seminal report on the progress of the world’s women, Elson (2000) adopts the feminist economist perspective of the 1990s regarding unpaid care work whereby “work” refers to an activity that takes time and energy, is costly and there is some form of obligation, either social or contractual; “care” refers to an activity which provides nurture to other people; “unpaid” means that no financial remuneration is received (Elson, 2000). This definition has stood the test of time being repeatedly cited in academic and policy papers subsequently (e.g., Xue and McMunn, 2021; Puteh and Kadir, 2022).

It can be argued that, in theory, the three concepts described above—work, care, and payment—along with the composite term unpaid care are generally understood. However, there is no agreed definition of what are the activities which might help identify someone as an unpaid carer, regardless of whom the care is provided. In reality, unpaid care differs across countries and is dependent on the person who receives the care and their needs. For example, in a country that does not facilitate the movement of those with mobility impairments, extra care will need to be provided compared to countries in which these people can, for instance, get access to stores and do their groceries alone.

Additionally, a person with a long-life visual impairment would have different care needs compared to someone who just has a broken arm. Overall, while we might, in theory, be able to conceptualise unpaid care, real-life experiences are diverse and do not easily fit into a box.

Caring for someone, while beneficial to the larger society and one’s personal satisfaction, has its drawbacks. Unpaid care can negatively influence the social, emotional, and financial well-being of the caregiver. For example, one study by Nguyen and Connelly (2014) showed that in Australia in 2008, being a co-resident informal caregiverFootnote 3, reduced employment probability by 13% for women and by 12% for men. The current study will only assess the effect of co-resident unpaid carers (parents/siblings who live in the same house as the child with disabilities). In the EU in 2016, 19% of economically inactive women were inactive as a result of providing unpaid care either to children or incapacitated adults (The European Commission, 2018). Additionally, in a qualitative report by Galandini and Ferrer (2020), unpaid carers expressed feelings of tiredness, social isolation, loneliness, being undervalued by society, etc. As such, further analysis of unpaid care is warranted to find ways to minimise the negative effects.

Research objective

The existing literature has shown a connection exists between poverty and the presence of disability within a family as well as between poverty and provision of unpaid care. A person with disabilities might need extra care provision compared to a person with no disabilities, but the provision of unpaid care is not a necessary condition of disability. With this in mind, we want to observe whether there are any differences, in terms of the effect sociodemographic factors might have on poverty, between families with a child with disabilities who needs extra care and one that does not. To our knowledge, little to no research in the UK has analysed the direct effect of providing extra unpaidFootnote 4 care on poverty in families supporting a child with disabilities. So, the current study aims to explore the tripartite relationship between poverty, unpaid care, and children with disabilities and their families. The analysis delivers this by considering three groups of interest: (1) Families supporting children without disabilities (2) Families supporting children with disabilities (a) who do not provide unpaid care and (b) who provide unpaid care. In this context, unpaid care provision refers to the extra care that is specifically related to a child’s disability. The research questions are:

  1. 1.

    Does the presence of a child with disabilities within a family increase the likelihood of poverty compared to the presence of a child with no disabilities?

  2. 2.

    Given the presence of a child with disabilities, does the provision of unpaid care by co-resident parents increase the likelihood of a range of poverty measures?

  3. 3.

    What are the covariate risk factors for poverty conditioning on membership of group 1 (families supporting children without disability) against group 2 (those supporting children with disabilities)

  4. 4.

    What are the covariate risk factors for poverty conditioning on membership of group 2a (Families supporting children with disabilities who do not provide unpaid care) against group 2b (those who provide unpaid care)?

Methods

Data source

The study used secondary data from the 2018/19 Family Resources Survey (FRS)Footnote 5, a face-to-face, repeated cross-sectional UK survey sponsored by the Department of Work and Pensions (2021a) and carried out by the ONS. The survey’s purpose is to provide a resource for the evaluation, monitoring, and development of social welfare policy. Respondents are adults aged sixteen and over, excluding those who were classed as dependent children. A stratified cluster probability sample design is applied to selected addresses in Great Britain from the Royal Mail’s small users’ Postcode Address File (PAF). For Northern Ireland, a systematic, geographically stratified sample approach was used. A total of 19,169 households and 22,406 benefit unitsFootnote 6 fully cooperated; a response rate of 50% was recorded, which is regarded as reasonable by the FRSFootnote 7. Specific non-response issues for this study are presented in the discussion section. As well as the base FRS datasets, we also used data from the Households Below Average Income (HBAI) (DWP, 2020b). HBAI uses raw data from the FRS and aggregates it to offer analysis targeted at understanding poverty in the UK.

Detailed information covering each household’s demographics, economic situation, and well-being is collected. In the event of a member of the household being absent, most of the questions can be responded to by proxy, but for some of the individual-level questions, answers by proxy are not possible, in which case non-response is recorded.

Sample

We selected an analytical sample of 5451 benefit units (BU) with dependent children (all those aged 0–16 and those 16–19 still in full-time education). The selection of benefit units with children was based on an already available variable in both datasets which indicates the number of children in each benefit unit. The final sample resulted from the merge of the FRS (5466 BU) and HBAI (5451 BU) datasets, with a small number of 15 BU being lost as a result of the merge.

Measures

Response variables

Three response variables were analysed:

  • The Before Housing Costs relative income (BHC) poverty line;

  • The After Housing Costs relative income (AHC) poverty line and

  • Child material deprivation.

This paper uses both measures to compare them across our groups of interest. The HBAI records both BHC and AHC as weekly net incomes from all sources (i.e., employment, benefits, investments, maintenance payments, etc. minus tax, national insurance, contributions to pension schemes, etc.) from all family members. For AHC, housing costs such as rent, mortgage, structural insurance, etc., are deducted from the BHC income. Both AHC and BHC values are adjusted to account for family size and composition using the modified OECD equivalence scales (for further details, see the methodology section in (DWP, 2020b)). For further analysis, a benefit unit is said to be in AHC/BHC poverty when their weekly income is below 60% of the weekly equivalised median AHC/BHC income. Based on this, binary labels (i.e., in poverty/not in poverty) were derived from the continuous AHC and BHC variables. While in line with the observations made in previous sections, the analysis was carried out for both AHC and BHC. However, as no meaningful differences were found between AHC and BHC results, only the AHC results are reported here.

Child material deprivation is measured as a continuous variable with a score from 0 to 100. The score is computed from the answers to a set of 21 questions about goods and services related to children, adults, and household items. Each question is assigned a 0 if the item can be afforded or 1 if it cannot, and then weights, determined through a prevalence weighting methodFootnote 8, are used to obtain the final score for each family (see Supplementary Table S.1, Supplementary Information, for the 21 questions used and their respective weights). Based on this score, a dichotomous variable is created and is available in the HBAI dataset. Children are classified as materially deprived if the score is 25 or above, meaning children live in families who cannot afford (but want) at least five to six of the items used to compute the material deprivation score.

Group classification variables

A variable is available in the FRS dataset, which was used to identify children with disabilities who, according to the FRS’s documentation (Department for Work and Pensions (2020a)), are based on the following criteria:” disabled people have been identified as those who report any physical or mental health condition or illness that lasts or is expected to last 12 months or more, and which limits their ability to carry out day-to-day activities” (p.9). This definition is based on the core definition of disability given in the Equality Act (2010), as described in the child disability and poverty section above. For this study, no distinctions are made between types of disabilities or severity due to low counts. However, the maximum number of difficulties (difficulty with mobility, hearing, vision, etc.) a child is struggling with in a family/benefit unit was also tested, but the results indicated that the variables are not statistically significant (p > 0.05).

In contrast, the identification of unpaid care provision is not straightforward. The FRS is not specifically designed for unpaid care research. Unpaid care can refer to multiple activities and can be provided to anyone and by anyone. However, the FRS only records unpaid care related to” people who receive help or support because they have long-term physical or mental ill-health or disability (or problems relating to old age).” (DWP, 2021a, questions instructions document, p. 112). Moreover, the FRS methodology document (DWP, 2020a) does not provide a fixed definition of what is meant by care and states that informal careFootnote 9 is: “any care where their carer is not doing it as a paid job; it can be for many, or only a few hours a week, and can take several different forms. The survey is intentionally not prescriptive about what counts as care; it could, for example, include going shopping for someone or helping them with paperwork”. (p.9). Overall, the dichotomous unpaid care variable is telling us whether unpaid care, regardless of how many hours, is provided by any co-resident family memberFootnote 10 or not. Moreover, for both disability and unpaid care, the definitions used in the FRS are the standard definitions for UK surveys.

Explanatory variables

Key sociodemographic variables commonly associated with poverty, care, and disability (see: Ghosh and Parish, 2013; Parish et al., 2010; Rothwell et al., 2019) were used as regressors. These included: the highest education level and highest occupational classification in the benefit unit; ethnicity; housing tenure; number of dependent children; number of adults with disabilities; employment status and average general health of the benefit unit measured by an ordinal variable with the lowest value representing better health.

Missing values were found in two variables: education (7.56%) and average health (1.12%). No explanation was found in the documentation as to why education has missing values, whereas health was not missing at random. Questions about health cannot be answered by proxy, so if no person from the benefit unit was present at the time of the interview, no data was recorded. However, due to the relatively modest level of missingness, the missing values were imputed using multiple imputationFootnote 11, with m = 5 and all of the other variables in the dataset used as predictors.

Results

Analytical strategy

We report descriptive statistics for the sample as a whole (i.e., families with dependent children), and separately for each group of interest: families who have at least one child with disabilities and families supporting children with disabilities who do not provide unpaid care, and those who do. Descriptive statistics for each variable by group are shown in Table 2.

Table 2 Descriptive statistics (unweighted data).
Full size table

Model-based logistic regression analysis was employed to examine the effect sociodemographic factors have on the response variables. To address the first two research questions—does the presence of child disability increase the likelihood of poverty and does the provision of unpaid care for a child with disabilities have a similar effect—two logistic regression models were run. In this case, the groups of interest disability vs. no disability; unpaid care vs no unpaid care were treated as binary regressors, and their direct effect on poverty was tested while controlling for other covariates. Results for these models are presented in the next section in Table 3. To address the third aim of the paper—comparing families with dependent children (without disabilities) with families supporting at least one child with disabilities—two logistic regression models were run (see Table 4). The final aim was addressed by comparing logistic regression results for families supporting at least one child with disabilities who do not provide extra unpaid care against those who do provide extra unpaid care (see Table 5). All models controlled for household structure (i.e., number of dependent children, lone parent), number of adults with disability, household average self-reported health, education level, tenure, ethnicity, and occupation classification.

Table 3 Odd-ratios of children’s material deprivation and AHC poverty for two groups: families with dependent children (left) and families with at least one dependent child with disabilities (right).
Full size table
Table 4 Odd-ratios of children’s material deprivation and AHC poverty, comparing families supporting children without disabilities and families supporting children with disabilities.
Full size table
Table 5 Odd-ratios of children’s material deprivation and AHC poverty, comparing families supporting children with disabilities who do not provide unpaid care and those who do.
Full size table

For each logistic regression model, Generalised Variance Inflation Factor (GVIF) tests were used to check for multicollinearity. Overall, GVIF scores for all variables across all models are below the value of two, suggesting no multicollinearity is observed between variables. All descriptive and statistical analysis was conducted using R.

Descriptive findings

Table 2 presents a descriptive statistics overview of the full sample as well as a breakdown by groups of interest. The main point of this table is to illustrate some of the differences between groups. For example, compared to the full sample, those supporting a child with a disability are more likely to be materially deprived, be lone parents, live in socially rented accommodation, and not have a higher education classification. Overall, the results show that families in which child disability is present are more likely to have a lower living standard.

Is belonging to either group—disability vs. no disability; unpaid care vs. no unpaid care—a significant predictor of poverty?

The presence of child disability and provision of unpaid care were treated as dichotomous regressors in logistic regression models. The direct effect of these two variables on material deprivation and AHC poverty was assessed. This section presents the results for the first two research questions—does child disability increase the likelihood of poverty, and given the presence of a child with disabilities, does the provision of unpaid care increase the likelihood of poverty? The first half of Table 3 shows the odds ratio computed using the main subsample—families with dependent children—where the variable of interest is a binary indicator of whether a child with disabilities is present within a family or not. In the case of material deprivation, this variable is not statistically significant (p > 0.05). However, having children with no disabilities seems to be increasing the odds of AHC poverty by 2.173 times. The last model, displayed in the right half of Table 3, only uses a small subsample, families, in which at least one child with disabilities is present. In this case, the variable of interest is a binary indicator of whether unpaid care is provided or not. Similar to the above, this variable is not statistically significant for material deprivation. On the other hand, providing unpaid care has lower odds of AHC poverty by 0.311. This is a slightly unexpected result, the implications of which will be debated further in the discussion section.

Families supporting children without versus with disabilities

The effect sociodemographic factors have on poverty measures for families supporting children without disabilities and those supporting children with disabilities are presented in Table 4. The following results address research question number three. Not owning a house increases the odds of material deprivation by around five times for both groups. Parents of children without disabilities with an elementary job have 4.11 higher odds of material deprivation compared to those with a managerial position. A higher education degree has lower odds (OR = 0.32) associated with material deprivation for those raising children with disabilities compared to the first group (OR = 0.678). However, being of any other ethnicity than white increases the chances of material deprivation by 3.2 times for those supporting children with disabilities.

Families supporting children with disabilities who do not provide unpaid care in comparison to those who do

The final research objective was to compare the effect of sociodemographic factors across two groups: families supporting children with disabilities who do not provide unpaid and those who do (see Table 5). When taking into consideration only the effect of statistically significant variables (p > 0.05), just one difference stands out. Whilst renting increases the odds of material deprivation for both groups, for those providing unpaid care, private renting increases the odds by 7.326 times compared to only 4.621 for those not providing unpaid care. In terms of AHC, no differences are observed between the groups from the statistically significant factors (p > 0.05).

Discussion

The current study’s focus was on families with children with disabilities in the UK. The 2018/19 FRS dataset was divided into two sub-samples: families with dependent children with no disability and families which have at least one child with disabilities. The latter was further subdivided into families with children with disabilities who do not provide unpaid care and those who do. The effect sociodemographic factors have on AHC poverty and children’s material deprivation in the presence of a child with disabilities was analysed. The descriptive results in Table 2 show that child material deprivation levels are higher among families within which child disability is present but that AHC poverty rates are slightly lower (both these differences were statistically significant). Additionally, for families not providing unpaid care, higher levels of AHC poverty were observed. The results in Tables 4 and 5 show a detrimental effect on material deprivation of being non-White or in a lower-paid job (in families where child disability is present). Equally, being a lone parent has a negative influence on those providing unpaid care.

The analysis also shows that having children with no disability increases the risk of AHC poverty in comparison to families raising children with disabilities (see Table 3). This was a somewhat unexpected finding; it was hypothesised that the presence of a disability within a family would increase the risk of poverty. These results could be explained by how income is measured and reported. Disability benefits are included in the overall income of a family, even though they are not means-tested benefits. As a result, the family’s income can rise just slightly above the AHC poverty threshold. Previous research (Schuelke et al., 2022; Morris and Zaidi, 2020) has shown that the direct costs of disability are higher than the benefits received. As a result, if benefits are removed from the total income, the residual income is more closely equivalent to that of a unit without any children with disabilitiesFootnote 12. The effect of doing this with these data is to reverse the direction of the AHC poverty levels with benefit units containing children with disabilities having a higher rate of AHC poverty (35.10%) than those without a child with disabilities (28.12%). This brings the AHC poverty levels approximately into line with those of material deprivation.

On the topic of provision of unpaid care, the sociodemographic factors used in this analysis do not have different effects on AHC poverty classification regardless of unpaid care provision. However, results from Table 3 show that benefit units providing unpaid care have a lower likelihood of AHC poverty in comparison to those who do not. This is a counterintuitive result, and indeed previous research has shown that providing unpaid care leads to a decrease in working hours (Nguyen and Connelly, 2014), which reduces the income of a family. However, according to our data, the average AHC income for those providing unpaid care is higher, £427.69 compared to £395.5 per weekFootnote 13. Additionally, only 13.11% of those providing care are below the poverty threshold compared to 29.73% of those not providing unpaid care. Of those AHC poor and providing unpaid care, 43.75% are lone parents implying income limitations (Joseph Rowntree Foundation JFR (2020); Nieuwenhuis and Maldonado, 2018) (especially if they spend most of their time caring for a child with disabilities).

Given the counterintuitive nature of this finding and its contradiction of research showing that the provision of unpaid care leads to a drop in income, it is reasonable to speculate on the possible explanations that might then be the subject of future research. We might reasonably observe that there are two processes in play: the decision to provide unpaid care and the consequences of that decision. The decision itself is likely to be based on factors that include the degree of disability but also the level of income. If a family’s income level is high, then the possibility of them providing unpaid care without falling into poverty is available to them. In terms of consequences, as previous research has shown (Carmichael and Ercolani, 2016; Brimblecombe et al., 2020), the decision to provide unpaid care will invariably have the effect of reducing income. So there might be two processes operating which have opposite impacts on the outcome variable, and perhaps in these data, there are higher numbers of better-off families making the decision to provide unpaid care to outweigh the negative impact of providing that care when averaged over the whole dataset. These are, of course, speculations, but the result is important enough to warrant further investigation. For example, longitudinal data could be used to offer insight into the processes underlying this finding. Additionally, a full investigation into which factors play a role in the process of deciding to provide unpaid could be achieved through the use of interviews and specific data collection.

The survey had a response rate of 50%, which according to the FRS (DWP, 2020a) documentation, is considered reasonable. Some of those who refused to participate provided their reasons, which included 18% who reported that they were ‘genuinely too busy’Footnote 14 (DWP, 2020a), potentially some of these had caring responsibilities. Moreover, of those who participated in the survey, not all answered for themselves, even though unpaid care data is reported at the individual level. Additionally, when asked the question about care:” Is there anyone in this household who receives any of these kinds of help or looking after?” respondersFootnote 15 are presented with a show card containing a fixed list of activities (see Table S2, supplementary Information), which could be perceived as taking care of someone. The use of a show card comes with both advantages and disadvantages. Giving people an exhaustive list could positively impact the number of people who identify themselves as carers, as some respondents might think the help they are providing is part of their duty/daily routine rather than providing care per se. For example, helping someone with their groceries or paperwork (an activity mentioned in the show card) might have not been perceived as unpaid care by some respondents. Seeing an activity on the list might then prompt them to identify themselves as providing care. This reduces extraneous variability in responses. On the other hand, providing a list can also introduce validity issues into the responses by constraining people’s choices, reducing subjective adaptability. Moreover, based on a prior detailed analysis of variables related to unpaid care, we infer that the survey only captured direct/primary care. In general, provision of care is self-identified, hence the number of unpaid carers might not always be a perfect representation of reality.

The above has addressed some of the potential limitations of this study. A final drawback to be mentioned is the treatment of disability as a binary variable. While the legal definition of disability in the UK is dichotomous in nature—either you have a disability or not—the extent (type) of a disability can have implications on the costs associated with it and the aid required. However, due to low counts in the data available, the type of disability was not used as a predictor in the analysis. Nevertheless, the study offered an exploration into this by looking at the number of difficulties a child was confronted with.

Conclusion

Overall, the study has found that sociodemographic factors have a statistically significant effect on children’s material deprivation when groups are treated independently. On the other hand, groups (disability vs. no disability; unpaid care vs. no unpaid care) have an impact on AHC poverty when treated as direct effects (using dummy variables). The study has offered an initial exploration into the tripartite relation of child disability, poverty, and unpaid care in the UK. Nevertheless, further research is needed to untangle some of the points highlighted, such as the effect of disability benefits on poverty and the distinction between those providing unpaid care and those who do not. Additionally, future research should consider data collection targeted specifically at families supporting children with disabilities to provide better, evidence-led policy.