1 Measures of income poverty and the definition of poverty used in this project are discussed in Chapter 2.
2 For further information about weighting in GUS see the user guides on the GUS website www.growingupinscotland.org.uk
3 Prior to the HBAI series the Government produced the Low Income Families ( LIF) statistics, which concentrated on showing the numbers of people living on, or below 140 per cent of supplementary benefit/income support.
4 An alternative way of looking at poverty is through expenditure rather than income and deprivation of essential items. Income and expenditure reveal different aspects of poverty and each has its own strengths and weaknesses. Atkinson (1989) argues that an income measure is about a right to a minimum level of resources, while expenditure is about a standard of living that can be achieved. Income does not completely reflect actual or potential living standards and recently the Government has incorporated material deprivation in its measure of child poverty ( DWP, 2003). On the other hand, patterns of expenditure may be highly dependent on the spending preferences of households.
5 For instance, Callan and Nolan (1994) demonstrate that the method cannot take into account improvements in living standards of low-income groups that are shared by the rest of the population or differences in average living conditions across countries. Furthermore, Veit-Wilson (1998) argues that relative income poverty lines represent nothing more than an abstract statistical construct without independent validity as an empirical indicator of poverty.
6 For instance, Callan and Nolan (1994) demonstrate that the method cannot take into account improvements in living standards of low-income groups that are shared by the rest of the population or differences in average living conditions across countries. Furthermore, Veit-Wilson (1998) argues that relative income poverty lines represent nothing more than an abstract statistical construct without independent validity as an empirical indicator of poverty.
7 An underlying assumption of income equivalisation that has been questioned by much research is that household income is shared equally amongst household members. Research indicates that women often prioritise the needs of other family members over their own and many poor parents tend to protect their children from the effects of poverty (for example Goode, Callender and Lister, 1998; Millar and Glendinning, 1989; and Middleton et al., 1997) although, as Marsh and McKay (1994) showed, parents do not always succeed in this. While the assumption of equal sharing does not always hold and families differ in the extent to which they pool and share their resources equally, larger households do benefit from economies of scale and this report equivalises income to account for this.
8 The official definition uses net income from all sources while GUS collects total gross income information. However, the difference between gross and net income is smallest towards the bottom of the income distribution (as a higher proportion of low income households' income fall below the personal allowance thresholds for income tax and national insurance and/or come from means-tested non-taxable benefits). As this study uses a low-income indicator rather than the whole income distribution the effect of GUS only collecting gross income should not be substantial in this analysis.
9 So, for GUS 2005/06 we obtain income estimates from SHBAI 2005/06, for GUS 2006/06 we obtain income estimates from SHBAI 2006/07, and for GUS 2007/08 we obtain income estimates from SHBAI 2007/08. The SHBAI for 2008/09 is not yet in the public domain, and hence for GUS 2008/09 we obtain income estimates from SHBAI 2007/08.
10 This report includes all families, including those where one or both parents were self-employed. While HBAI has noted that the reported incomes among the self-employed group can be anomalous in relation to their living standards, HBAI analyses also include the self-employed ( DWP, 2009).
11 Analysis carried out by Scottish Government.
12 As the observations are annual it is possible that a child could have been poor in between interviews and this would not be captured in our analysis.
13 None of the analysis takes into account how poor families were when they are poor (the shortfall of income below the poverty line) or the extent to which income was above the poverty line during periods that families were not poor.
14 The temporary poor group of families is not homogenous and contains, amongst other categorisations, families that have escaped or entered poverty over the period. These two groups of families in particular are likely to have quite distinct outcomes related to their poverty transitions and further investigation of these families is beyond the scope of this report.
15 Including imputed households in the final classification of longitudinal poverty status changed the incidence estimates only very slightly. For example, 58 per cent of the birth cohort and 59 per cent of the child cohort were not poor prior to imputation and 58 per cent of each cohort were not poor after imputation.
16 Although other surveys capture income more precisely, it can lead to researchers highlighting small changes in income that push a household over the poverty line, even though it is unlikely to result in a marked change in household living standards. In fact very small income fluctuations are often not a useful way to re-categorise a household's poverty status, and some analysts use a move across the income threshold accompanied by a 'substantial' change in income (say 5 per cent) to identify a transition in to or out of poverty.
17 The results related to ethnicity should be interpreted with a degree of caution due to a small number of children from ethnic minority communities in the GUS sample.
18 AWI measure is based on the average use of household workforce, i.e. the ratio of people in employment to the total number of adults available to work. See Appendix 1 for the full definition and examples of calculating AWI.
19 In the case of AWI, the unweighted bases are lower than for other characteristics. This is because the AWI indicator is a complex variable, derived using several different questions (see Appendix 1 for details), some of which were particularly affected by non-response. For this reason, the poverty estimates for AWI have been calculated separately, using smaller bases (2951 cases for Birth Cohort and 1558 cases for Child cohort).
20 The statistical analysis and approach used in this report represents one of many available techniques capable of exploring this data. Other analytical approaches may produce different results from those reported here.
21 For example, the positive effect of decrease in the number of children on the risk of poverty may be related to family separation, which may lead to drop in available income, or the death of a child, which could lead to a parent ceasing employment either permanently or temporarily thus affecting income.
22 Using the latest sweep of GUS to identify child outcomes means that there may be some blurring of the relationship with the longitudinal poverty groups. This is because some of the persistently poor children, as defined in this research, may not be living in a poor family in 2008/09 (the definition states that to be persistently poor a child has to be living in a poor family for three or more out of four years). Likewise, a temporary poor child may be living in a poor family in 2008/09. However, these potential inconsistencies are likely to average out and not have a major effect on the analysis.
23 We use the chi-square test to test for statistical siginficance at the 95% confidence level. It should also be noted, given the relationship between statistical significance and sample size, that the birth cohort is almost twice as big as the child cohort.
24 Specifically, it would run us into what is known in economic literature as the problem of endogenity of dependent variables.
25 We also ran models to test the relationship between poverty per se, that is either temporary or persistent poverty against no poverty, and found that in all models there was no relationship between any experience of poverty over the period and child outcomes.
26 However, it needs to be noted that this latter study used a different measure of persistent low income and a much more detailed measure of cognitive ability, capable of detecting quite small nuances in ability. Accordingly, a different statistical method (linear regression) was used in this study.
27 See Hills (2007) for an example of how labour market and housing policies can work together to enhance employment opportunities.
There is a problem
Thanks for your feedback