0115 966 7955 Today's Opening Times 10:00 - 20:00 (BST)
Place an Order
Instant price

Struggling with your work?

Get it right the first time & learn smarter today

Place an Order
Banner ad for Viper plagiarism checker

Relationship Between Childhood Well-being and Poverty

Disclaimer: This work has been submitted by a student. This is not an example of the work written by our professional academic writers. You can view samples of our professional work here.

Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of UK Essays.

Published: Tue, 20 Feb 2018

Introduction

This paper explores the relationship between childhood well-being and poverty. Using structural equation modelling a multidimensional picture of child well-being is developed which is linked to previous work on multidimensional poverty indicators at household level (Tomlinson et al. forthcoming). Following a brief literature review of childhood poverty and well-being research, there follows an analysis of several waves of the British Household Panel Study – a valuable source of data collected directly from children as well as adults in the same households. The paper attempts to map the experience of poverty at household level and relate it to the child’s well-being. Rather than seeing poverty as a facet of child well-being, as other researchers often do, this work conceptually distinguishes between the two and shows how they are linked.

Following the literature review various structural equation models are estimated that measure different dimensions of child well-being. These dimensions are then related to other aspects of the child’s life including the experience of poverty, age and gender, household composition, income, parental education and employment status. The effects of poverty are broken down into more detailed dimensions and the relative impact of each dimension is discussed. Finally, the models are used to inform targeting strategies with respect to child welfare policy. Crucially the differential impact of various potential policy instruments is assessed through the models.

Mainstream child poverty research

Since New Labour took office and pledged to eliminate child poverty by 2020 a myriad of policy changes and political statements has been issued to address the problems associated with poverty and deprivation during childhood. Indeed the costs of child poverty and its immediate and future effects are becoming increasingly alarming. For instance, recent research has found that poor children are more likely to get into trouble inside and outside school and more likely to be involved in drug abuse (ONS 2002). The direct costs of this are estimated to be considerable. For example:

  • £6000 for a 6 month non-custodial sentence
  • £21000 for a custodial sentence of 6 months
  • Cost of attending pupil referral unit: £10000/year
  • Drug programmes cost on average £15000/person over a 4 year period

(Source: Godfrey et al. 2004)

Much of the literature relating to child poverty in the UK has focussed around two areas: first the identification of households where risk is greatest and second, the so-called ‘scarring’ of children and the transmission of disadvantage into adulthood. With respect to the former it is now well known that poor children in particular are more likely to come from the following types of household:

  • Workless households
  • Benefit dependent households
  • Lone parent families
  • Low income households
  • Families with younger children are more likely to be poor
  • Large families
  • Ethnic minority households
  • Those in rented accommodation

See, for example, Hirsch (2006a), Lloyd (2006). In addition Bradshaw (2006a) has extensive breakdowns of poverty rates for different social groups with children; Platt (2007) has an analysis of ethnicity, employment and child poverty; large families are extensively discussed in Iavacou and Berthoud (2006) and so on. In other words it is no longer an issue of identifying which types of environment – from a household perspective – are important, but rather moving towards a measurement model that can assess the impacts of the various dimensions associated with poverty on the child and its well-being. This is the approach taken in this paper.

With respect to the second set of literature on scarring and transmission, the impact of poverty on a child’s future life-chances has also been extensively researched. Moreover, these impacts appear to have increased as child poverty increased during the 1980s and 1990s (Fahmy, 2006). Gregg and Wadsworth (2001) have noted the increased polarisation of working versus non-working households and the effects that this has had on poverty rates. That is the growth of dual-earner versus no-earner households. Using cohort studies such as the British Cohort Study (BCS) and National Child Development Study (NCDS), a series of papers has shown that low income in childhood leads to poor educational attainment in later life. For example, see Blanden and Gregg (2004) which also provides a useful review of the US literature on this topic. Gregg and Machin (2000) and Glennester (1995) come to similar conclusions.

Fahmy has also reviewed the literature with respect to youth poverty (youth being defined as being aged 16-25). The consequences of poverty identified for this group, referred to as ‘hazardous transitions’ into adulthood, include:

  • A high probability of becoming a ‘NEET’ (not in employment, education or training – see Istance et al 1994 for an earlier study)
  • A bad career track (Craine 1997)
  • A reduced level of citizenship and civic participation (Dean 1997)
  • A higher risk of homelessness (Smith 1999, see also Flouri and Buchanan, 2004)

Stewart has also documented various consequences of child poverty in later life. Adding low self-esteem, low expectations, reduced educational attainment, benefit dependency and poor labour market outcomes to the list. See Stewart, (2005) and also Hobcraft (1998) and Ermisch et al. (2001).

While all this work is very convincing and commendable there is relatively little literature relating child poverty in the here and now and its immediate impact on the life and environment of the child. It is almost as if this were less important than the future costs. However, there is also a growing interest in the current well-being of children and its measurement. Early literature on this is extensively reviewed in Pollard and Lee (2002). This covers definitions of well-being, the indicators developed and instruments used in the measurement process. Moreover, two recent special issues of Social Indicators Research (SIR, 2007a, 2007b) have already been devoted exclusively to the topic (and a third issue is on the way).

Interestingly, one strand of this work relates to human rights which shows the level of importance now being attached to these issues. Bradshaw et al. (2007) discuss concepts of well-being which are predicated on the UN convention on the rights of the child (UNCRC). Essentially this accepts the multi-dimensional nature of well-being from at least four perspectives: first that it is non-discriminatory, second that it is in the best interests of the child, third that it relates to the child’s survival and development, and fourth that it respects the views of the child (Bradshaw et al 2007: 134).

The link to poverty and deprivation is sometimes made explicit in this literature: for example, ‘child well-being and deprivation represent different sides of the same coin’, Bradshaw et al. (2007). On the other hand, US, and very recent British, research shows well-being to be related to, but not the same as childhood poverty (Land et al., 2006; Bradshaw and Mayhew, 2005) for reasons that are not well-understood, but which probably include protective behaviour by parents (e.g., Flouri, 2004) and individual resilience (e.g., Masten and Coatsworth, 1998, Masten, 2001). Thus there is confusion about the relationship between well-being and poverty. Sometimes poverty is cited as a specific dimension of well-being, and sometimes as a separate concept entirely.

For example, Bradshaw et al. (2007) have developed an eightfold classification of child well-being and generated one composite summary indicator from internationally comparable data. The eight dimensions being:

  • Material well-being
  • Housing
  • Health
  • Subjective well-being
  • Education
  • Relationships
  • Civic participation
  • Risk and safety

These are measured by standardised scores which are added together to form the individual indices and an overall summary index which is then used for international comparison.

There is then no accepted or uncontroversial measure of child well-being. The general thrust of the debate is that child well-being must be measured along several dimensions and poverty (or particular dimensions of poverty such as material deprivation) is sometimes included and sometimes not. The approach taken in this paper is somewhat different in that the two concepts are kept completely distinct as explained in more detail below.

The measurement of poverty and well-being

The approach here uses two sets of measures reflecting two aspects of the situation of children living in British households. First of all we measure poverty at the household level using structural equation models. This is done along several dimensions using data from the British Household Panel Study (BHPS) and is discussed in Tomlinson et al. (forthcoming). The dimensions are: financial strain, material deprivation, the environment, psycho-social strain, civic participation and social isolation. These are combined into an overall weighted index referred to as the Poverty Index (PI).

Second we use structural equation models to measure various dimensions of childhood well-being. We are restricted in the questions that are asked and cannot include all the dimensions listed by Bradshaw et al. (2007). However, we measure four different aspects of child well-being including ‘home life’ which relates to family relationships and parental control (similar to Bradshaw’s ‘relationships’ dimension), ‘educational orientation’ (again similar to Bradshaw et al.), ‘anxiety’ (based in part on Bradshaw’s subjective well-being indicator) and ‘delinquency’ (which also relates to risk and safety).

However, a crucial difference with our approach is that we treat dimensions such as material well-being and housing as aspects of household level poverty rather than childhood well-being. Thus we keep poverty and well-being conceptually distinct and analyse the relations between the two. It is the association between these four measures of child well-being and the numerous measures of poverty already developed that is the ultimate focus of the paper.

In summation we take a multidimensional approach to both well-being and poverty and we examine the correlates of poverty with a child’s current well-being. In this way we can assess the impacts of poverty on the child’s immediate social environment and state of mind rather than what the future might hold. Models which can link together different aspects of poverty with various aspects of children’s livelihoods will assist in developing strategies to alleviate some of these problems. In other words we identify which aspects of poverty have the most serious impacts on the child (and hence will probably affect their future life chances to the greatest extent).

Using structural equation models (SEM)

There are now many academics using more advanced statistical techniques to measure poverty from a multi-dimensional perspective (e.g., Jenkins and Cappellari, 2007, Tomlinson et al., forthcoming, Whelan et al. 2007a, 2007b). These techniques, such as item response theory, structural equation modelling and latent class analysis, can be used not only to analyse which families with children are actually in poverty, but also which particular aspects of this poverty are more intense (such as bad housing, material deprivation, financial strain and so on). This is the approach taken in this paper with respect to the measurement of poverty and the measurement of child well-being – the two being linked together within a coherent methodological framework and then related specifically to policy and policy targeting.

Like the more traditional method of factor analysis, a SEM reduces a large number of observed variables to a smaller number of factors. However, in a SEM the variables are conceptualised as observed manifestations of an underlying or ‘latent’ dimension. Each observed variable in a SEM also has an error term associated with it, allowing measurement error to be isolated and controlled for in a way that is impossible with factor analysis. But, most importantly, a SEM requires a strong theoretical justification before the model is specified. Thus the researcher decides which variables are to be associated with which latent unobserved factors in advance.

There are two fundamental types of SEM used to measure or test the validity of latent concepts – first and second order confirmatory factor analysis models (CFAs). We use first order CFAs below to measure child well-being. A first order CFA simply attempts to measure preordained underlying latent concepts. The left side of figure 1 shows a simple CFA which has two latent unobserved variables: L1, material deprivation; and L2, financial strain. L1 is measured by the observed variables V1 to V4 and L2 is measured by variables V5 to V7. The single headed arrows represent coefficients or loadings in the model and are usually shown in standardised form much like beta coefficients in regression analysis. The covariance between material deprivation (L1) and financial strain (L2) is represented by the double headed arrow. The associated error terms are shown as the circles labelled e1 to e7. Using statistical techniques such as maximum likelihood estimation and making assumptions about the distributions of the variables and error terms in the model, the coefficients and covariances can be estimated. In all SEMs a variety of fit statistics is available to assess the validity of the models constructed (see Klein, 2005, Byrne, 2001). Usually it is assumed that the observed variables in the model are continuous and that the distribution of the variables is multivariate normal. More recently available software is beginning to allow the explicit modelling of categorical, binary and censored variables (such as MPlus which is used in this study).

Models of this kind can be made as complex as necessary to describe real-world situations and employ many latent variables and various interactions between them. Covariates or controls can also be applied to the overall measurement models to assess differences between groups or to assess the impact of a particular variable on the latent concepts under consideration. Furthermore, scores can be generated for the unobserved latent variables. These scores are analogous to the factor scores obtained using factor analysis.

The BHPS and the measurement of childhood well-being

The analysis that follows utilizes data from the British Household Panel Study (BHPS) and follows the methods discussed in Tomlinson et al. (forthcoming). The BHPS commenced in 1991 with an initial sample of around 10,000 individuals resident in some 5,000 households. These individuals have subsequently been re-interviewed each year and the sample has also been extended to include more households from Scotland and Wales and to embrace Northern Ireland (although Northern Ireland is excluded from this analysis). The data can be weighted to provide an accurate picture of life in Great Britain at different points in time.

The analysis here covers the period 1997, 1999 and 2001 (i.e. BHPS waves 7, 9 and 11) and draws on information concerning the following topics for the measurement of poverty: income, finances and benefits; stress; material deprivation; general housing and neighbourhood characteristics and social exclusion and civic participation. The level of poverty at household level is measured by the responses given by the head of household and calculated as detailed in Tomlinson et al. (forthcoming). Each individual dimension of poverty as well as an overall score (the Poverty Index) is computed via a SEM for each household with children. Households with heads under 18 years of age or over 64 years of age are excluded from the sample analysed to calculate poverty scores.

We also use a unique data resource available within the BHPS and consistently applied across the three waves. Children aged between 11 and 15 within these households were also asked to complete a separate questionnaire which forms the basis for the measurement models of child well-being. Questions included relate to home life, schooling, anxiety and psychological aspects of life, social isolation and delinquent behaviour.

Estimating a structural equation model of childhood well-being

As with the measurement of our multi-dimensional poverty index we attempted to create measures of multidimensional childhood well-being using 1st order CFAs based on the responses given by the 11 to 15 year olds in the BHPS panel for the years 1997, 1999 and 2001. The models have been estimated separately for all three waves. Questions change significantly in other available waves and these waves have not been included in the present analysis. The four dimensions of well-being are estimated using the following variables (which are all measured as ordinal scales except the variable relating to suspension from school which is binary):

1. Home life is a measure of the children’s relations to their parents and family and how much control the parents have over them:

How much children talk to their parents

How much control parents exercise over TV

How much the family share meals together

2. Educational orientation is a measure of how well the child is doing at school and their attitudes to teachers and so on:

How much the child likes his/her teachers

Whether the teachers ‘get at me’

General feelings about school

Whether the child is doing well at school

3. Anxiety is a measure of the child’s psychological health and feeling of self-worth

Whether the child feels unhappy

Whether the child has lost sleep

How useless the child feels

How much of a failure the child feels

Whether the child feels no good

The extent to which the child feels lonely

The extent to which the child is left out of activities

4. Delinquency is an attempt to measure aspects of criminal tendencies or anti-social behaviour:

Whether the child has ever been suspended from school

How often the child plays truant

How much experience the child has with smoking cigarettes

Whether the child vandalises property

Whether the child has friends that use illegal drugs (there is no direct question about the respondent’s own drug use)

A first order confirmatory factor analysis model was estimated to measure the four dimensions (see figure 2 for an example from wave 11) and further models developed with controls for gender and age of the child and the overall Poverty Index of the head of household. We attempted this with each of the three waves of the BHPS, but all three models gave similar results and good fit indices. The model estimation was done using MPlus 4 with the observed variables being treated as ordinal rather than continuous where appropriate.

Results and discussion of the basic model

The first order models produce a good fit to the data (see Table 1) and the coefficients on the observed variables are all in the expected direction and all statistically significant at the 1% level. Some error terms were allowed to co-vary as illustrated in the figure based on very high modification indices in the initial modelling attempts. Examining the latent constructs themselves and the correlations between them reveals the relationships between the various dimensions of well-being. That is educational orientation is strongly associated with parental influence and negatively associated with anxiety and delinquency. Delinquency is also positively associated with anxiety etc. (Table 1).

Table 1Fit statistics and correlations for the simple models (wave 11)

Fit statistics:

(N=1201)

Without controlsWith controls

Chi-square 426.959 (79 d.f.)639.104 (130 d.f.)

CFI0.9370.902

TLI0.9550.921

RMSEA0.0570.057

Correlations between latent variables in controlled model (all significant at 1%): 

 

Home life

Educational

Orientation

Anxiety

Educational

Orientation

+.54

   

Anxiety

-.18

-.36

 

Delinquency

–.63

-.54

+.22

The controlling variables are also salient. Girls are more anxious than boys, but have better educational orientation and relations with their parents. There is no significant difference between girls and boys with respect to delinquency. The age controls show that home life diminishes with age, while delinquency increases. Children of 11 and 12 also have stronger educational orientation than their older peers. However, the most striking result is that poverty (measured by our composite multidimensional index) has a highly significant and detrimental effect on all four of the well-being dimensions. That is it contributes to anxiety and delinquency and detracts from educational orientation and home life. Thus we can show that poverty has a serious debilitating effect on child well-being in the here and now. The relative importance of poverty for each dimension of well-being is also evident. The strongest effect appears to be on home life (–0.22) followed by educational orientation (–0.13). The impact on anxiety and delinquency is less strong (both at 0.10), but still highly significant. Thus we can show that the overall impact of the experience of poverty appears to affect home life and education the most while still having an effect on anxiety and anti-social behaviour.

However, one of the issues we wish to deal with (not least from a policy targeting perspective) is to see which sub-dimensions of poverty are the most salient with respect to child well-being. For example, as we have measured poverty in a multidimensional way, which particular dimensions have the biggest impact? In our previous measurement work we developed several indicators of multidimensional poverty. Namely the poverty index is a weighted summation of several sub-indices:

  • financial strain based on bad finances and missed housing payments
  • material deprivation based on the levels of material possessions in the household and whether the household could afford to do certain things
  • the environment which is based on a combination of housing and neighbourhood characteristics
  • social isolation based on lack of social support
  • civic participation based on participation in civic life
  • psycho-social strain based on stress, mental health and anxiety

The most desirable way to test the effects of the various dimensions on well-being would be to include them all as covariates in a measurement model similar to that shown in figure 2. However, because the various dimensions of poverty are highly correlated with each other this presents problems for the estimation (that is there is a multicollinearity issue). Rather than attempt to do this, individual models have been estimated with each sub-dimension of poverty included by itself in place of the overall poverty index in a similar fashion to the model in Figure 2. The relative sizes and significance of the coefficients relating to the individual sub-dimensions of poverty will allow an assessment to be made as to which elements of poverty are the most serious with respect to the child’s welfare. The results are summarised in figure 3 (this is a diagrammatic summary of results from wave 11 (2001) and shows only the significant effects).

The results show that different aspects of poverty have different effects on the various aspects of well-being. For example, the financial dimension affects all the aspects of well-being whereas material deprivation only affects two (being detrimental to home life and increasing delinquency). A poor environment in terms of bad housing or neighbourhood results in reduced quality of home life, increased anxiety and delinquency. By using these results it becomes clear that policy aimed at poverty reduction could in principle be targeted in particular ways that would have different benefits as far as the diverse dimensions of child well-being are concerned. Improving the environment of children – both within and outside the household – may well have a greater overall impact on well-being than improving material deprivation. On the other hand if educational performance is the main criterion then financial strain, and civic participation of the household become the key areas. If home life is seen to be the main issue then finance, material deprivation, the stress of the parents, the environment and civic participation would be the key foci. This policy dimension is returned to below. It is also interesting to note that social isolation (a measure of social exclusion) of the head of household has no bearing on the four well-being indicators.

However, there are also other controlling factors that can be incorporated in the models determining child well-being besides poverty, age and gender. Using the structural equation framework with covariates allows several alternative model specifications to take into account different offsetting factors with respect to child welfare. There is already evidence from the UK that certain situations in childhood can ‘buck the trend’ in reducing the negative outcomes of child poverty. For example, Blanden (2006) has shown that parental interest (mainly the father for boys and the mother for girls) has a positive impact on adult educational outcomes. She also shows that higher educational attainment early in the child’s life has a positive impact later on as does the school’s characteristics and the social mix of the child’s school.

So research has shown that there may be mediating effects (such as parenting or living in a good neighbourhood) that offset the deleterious impact of poverty and deprivation. For example, McCulloch and Joshi (2001) found using the National Child Development Survey that although poverty and living in disadvantaged neighbourhoods does correlate with lower test scores at school, the family environment and family support can offset this effect. In the US the extensive work of Aber and his colleagues has also shown that there are negative effects on child specific outcomes from poverty and material hardship and that cognitive and emotional outcomes are affected by low income and material hardship (e.g., Gershoff et al, n.d.), but that this is mediated by parental characteristics.

With this idea of mediation in mind several alternative models have thus been estimated to take account of the following factors which are included as further controls in the models:

  • Household composition (such as the presence of other children and single versus multiple adult households)
  • Educational attainment of the household head
  • Employment status of the head of household
  • Income rather than multidimensional poverty indices

The household composition model will enable an assessment of family relations and its impact on well-being. The education model will assess the impact of parental human capital irrespective of other considerations. While the employment and income models can be usefully compared with the Poverty Index model (in other words can income or employment status merely substitute for poverty)? These results are summarised in table 3 for wave 11 (2001). The models were essentially the same as shown in figure 2, but without including the Poverty Index as a control which confounded the income and employment status models (again because of multicollinearity).

Household composition was tested by including a variable indicating whether the household was a single adult household (versus other types) and dummy variables representing the number of children in different age categories. The results show clearly the influence of adults is significant when it comes to home life and delinquency (whereas being a single adult household has no effect on anxiety or educational orientation). Single parent households are therefore at a possible disadvantage when it comes to controlling their children. Even when a control for income is included in this model in an attempt to separate out the impact of low income from single parenthood the single adult variable is still significant in the same way. The presence of other children or siblings appears to have no impact on the child respondent’s well-being.

Education of the head of household also has an impact on home life and educational orientation of the child, but only where the household head is educated to a higher educational level (that is degree level). The models for employment status included variables for self-employed status, unemployed and non-employed (i.e. not working and not actively looking for a job). Clearly the household head not having a job has an effect on the child’s well-being (although this is also correlated with the Poverty Index). In the case of being non-employed (which includes housewives, the disabled, and other economically inactive people) this has an impact on all four well-being dimensions to the detriment of the child whereas being unemployed only affects home life and delinquency. Self-employment has no effect. One possible explanation for the difference between unemployed and non-employed effects might be a reflection of the impact of long-term poverty and deprivation on children. That is those household heads that are not economically active for one reason or another and classed as non-employed rather than unemployed may well suffer from longer periods of chronic financial hardship, whereas the unemployed may be intermittently working and thus have experienced periods where they were no longer poor.

Table 3Effects of various controls on the basic well-being model with various controls in addition to age and gender of the child (wave 11). Significance level is 1%. Standardised coefficients shown.

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.