Growing up in Scotland: the circumstances of persistently poor children

chapter 2 HOW MANY CHILDREN EXPERIENCE PERSISTENT POVERTY?

This chapter sets out how we define persistently poor children using GUS data. It begins with a brief discussion of using low income as an indicator of poverty. It then describes how household income is collected in GUS and how this compares to the data collection methods of other surveys. There follows a description of the poverty threshold used in this project and how persistent poverty is defined. The section concludes with an analysis of the prevalence of persistently poor children over the first four sweeps of GUS, covering the period from 2005/06 to 2008/09.

Key findings from this chapter are:

• Relative low income is used as an indicator of poverty (Section 2.1).
• GUS collects income in a different way than other social surveys used to estimate income poverty.
• The GUS surveys find approximately three in ten young Scottish children to be living in income poor households at any one point in time (Section 2.3).
• We define persistent poverty as living in a low-income household at three or four of the four annual sweeps of GUS from 2005/06 to 2008/09 (Section 2.4).
• Using this definition we find that one in four 3-4 year-olds and one in five 5-6 year-olds were persistently poor (Section 2.5).
• When considering just those children living in a poor household in 2008/09, we see that the majority, four in five, had been living in persistent poverty over the previous four years (Section 2.5).

2.1 Using low income to conceptualise poverty

Traditionally, the understanding of poverty has focused on the lack of resources at the disposal of an individual or household to ensure a suitable standard of subsistence or living. Despite the abundance of theoretical work in the conceptualisation of poverty, it is only relatively recently that the UK and Scottish Governments have adopted an official low-income threshold (for children) following the announcement of the target to eradicate child poverty by 2020.

This 'official' conceptualisation of poverty is provided in the annual series of statistics called Households Below Average Income ( HBAI), first published in 1988 ³ by the UK Government, and its Scottish equivalent Scottish Households Below Average Income ( SHBAI), first published in 2006 (for the latest versions see DWP, 2009b and SG, 2009). The concept of poverty used in the HBAI and SHBAI series is regarded primarily according to ' potential living standards as determined by disposable income' ( DWP, 2007b). However, it has been pointed out in the literature that it is problematic to determine what is meant by a minimum level of subsistence, or living standards, and to equate this with a sum of money from which this can be achieved ( e.g. Gordon et al., 2000). ⁴

An alternative method of measuring poverty according to income levels is through the construction of relative poverty lines. This approach defines as income poor those who fall a certain distance below an average income level. Similarly to other approaches, relative measures of poverty have attracted some critique. ⁵ However, despite the criticism, the relative poverty lines remain the most commonly used approach to the measurement of poverty.

This project will define poverty according to the Scottish Government's most often used poverty indicator - that is, relative low income or more precisely below 60 per cent of median equivalised household income before housing costs. The construction of this measure using GUS data is described in the following sections while technical details are further explained in Appendix 1.

2.2 Measuring household income in GUS

Before categorising households as income poor or not, we need to be able to establish the amount of income each of them receives. We measure total household income using the single question asked to the mother (or main carer) of the GUS child. This question asks the mother to indicate the total income of their household from all sources before tax - including benefits, interest from savings and so on. Respondents are asked to choose from 17 income bands, ranging from 'Less than £3,999' to '£56,000 or more'. The wording of the income question in GUS is provided below.

Methodology box 2.1

I would now like to ask you some questions about your employment and income. As with all your answers, the information you give will be entirely confidential. This card shows different income levels as weekly, monthly and annual amounts*. Which of the letters on this card represents the total income of your household from all sources before tax - including benefits, interest from savings and so on? Just tell me the letter beside the row that applies to you. Q Less than £3,999 J £23,000 - £25,999 T £4,000 - £5,999 D £26,000 - £28,999 O £6,000 - £7,999 H £29,000 - £31,999 K £8,000 - £9,999 A £32,000 - £37,999 L £10,000 - £11,999 W £38,000 - £43,999 B £12,000 - £14,999 G £44,000 - £49,999 Z £15,000 - £17,999 N £50,000 - £55,999 M £18,000 - £19,999 E £56,000 or more F £20,000 - £22,999

* Only the annual amounts are shown here.

The first thing to note is that the width of the income bands differs, with wider bands towards the top of the income scale. As shown in Table 2.1, this results in a relatively even spread of each cohort sample across the bands; the majority of the bands contain between four and seven per cent of each of the cohorts. ⁶

Table 2.1 Distribution of total annual household income, GUS 2008/09

Band	Birth Cohort		Child Cohort
	Per cent	Unweighted count	Per cent	Unweighted count
Less than £3,999	0	14	0	6
£4,000 - £5,999	3	94	2	38
£6,000 - £7,999	3	100	4	63
£8,000 - £9,999	4	104	5	76
£10,000 - £11,999	4	122	4	69
£12,000 - £14,999	7	205	5	102
£15,000 - £17,999	6	215	7	118
£18,000 - £19,999	4	143	4	79
£20,000 - £22,999	5	199	6	111
£23,000 - £25,999	7	239	6	125
£26,000 - £28,999	6	239	6	120
£29,000 - £31,999	6	246	5	109
£32,000 - £37,999	9	384	8	177
£38,000 - £43,999	8	334	9	192
£44,000 - £49,999	7	293	7	167
£50,000 - £55,999	6	270	6	132
£56,000 or more	13	592	15	365
Total	100	3,793	100	2,049

The way GUS collects income information is different from the more specialised income surveys. For example, the Family Resources Survey ( FRS), used as the basis for HBAI and SHBAI, asks each adult household member about their own income and totals household income from all sources. The FRS also verifies income amounts during the survey interview, for example by asking respondents to show details of pay slips and benefit awards.

Clearly there are likely to be differences in quality when just one question collects information on total income, when this is asked about the household rather than the individual, and when banded income is used. Research by Micklewright and Schnepf (2007) shows that differences in quality are more noticeable at the lower ends of the income distribution. They also found that when using a single question more accurate estimates of household income are generally obtained from men compared with women, and from respondents with income from employment rather than mainly from benefits or pensions. There is also evidence of income being underestimated by women with children.

On the other hand, there are indications that prior questioning on sources of income (as is the case in GUS) might improve the reporting of income. Furthermore, the loss of information in using income bands rather than a continuous measure is minor when looking at the lower end of the income distribution as most of the loss of variation is in the top (uncapped) category. Overall, the loss in accuracy of income estimates obtained from a single question tends not to be 'catastrophic' (Micklewright and Schnepf, 2007, p.20) and have to be weighed against the cost and feasibility of collecting detailed income information in GUS given the competing demands from other topics in the survey. However, using a poverty measure based on income collected in this way may well impact on findings, and hence this should be borne in mind when interpreting the analysis presented in this report.

2.2.1 Equivalising household income

Clearly the standard of living provided by a household's income depends on the size and composition of the household. For example, given two households with £1000 a month income (and everything else equal), we would not expect a lone mother with one child to have the same living standards as a couple with four children. The £1000 has to provide for more people in the couple household and hence we would expect their standard of living to be lower.

To better reflect how a household's financial resources relate to the living standards of its members, we use 'equivalised' income. The equivalisation of income is the process by which total income is adjusted for the number of adults and the number of children of different ages in the household. This enables a comparison of the potential living standards of different types of household. ⁷ There are a number of equivalisation methods and the one used in this report is the so-called 'modified OECD' equivalence scale, which is also used in the SHBAI series. To equivalise income using banded income, we apply the equivalisation calculation to the mid-point of each band. Clearly there is no mid-point of the top unbounded category (£56,000 and above), so here we used a value of £60,000. Appendix 1 to the report explains in detail the process of income equivalisation applied in this project.

2.3 Measuring income poverty using GUS

We define our poverty threshold in the same way as used in official government statistics; that is, we define a household as poor if its equivalised weekly household income before housing costs is below 60 per cent of the population median income. ⁸ Clearly we can not obtain the population median from the GUS dataset, as this covers only two cohorts of young children. Therefore, for each corresponding year of GUS, we obtain estimates of median equivalised income for the Scottish population from the Government's SHBAI series. ⁹ We then calculate 60 per cent of this figure to obtain the low income, or 'poverty', thresholds. ¹⁰

GUS households with income below the poverty threshold are categorised as income poor for each sweep of data. Table 2.2 shows the income below which different family types and sizes would have been considered income poor in 2007/08 (corresponding to Sweeps 3 and 4 of GUS), using the 60 per cent of median income reported in SHBAI for 2007/08 ( SG, 2009). This shows, for example, that the poverty threshold is just under £11,000 a year for a lone parent with one child under 14.

Table 2.2 Low income or 'poverty' thresholds, annual income, 2007/08

	Couple family	Lone parent family
One child under 14	£14,714	£10,668
Two children under 14	£17,167	£13,120
One child under 14, one aged 14 or over	£18,761	£14,714
Three children under 14	£19,619	£15,573
Two children under 14, one aged 14 or over	£21,213	£17,167

The poverty rates for children in GUS for 2005/06 to 2008/09 are presented in Figure 2.1, along with estimates for families with young children in Scotland and the UK based on analysis of FRS data. ¹¹ The apparently higher proportion of income poor families in GUS compared with comparable families with young children in Scotland is likely to be due to the differences in how the income information is collected. As discussed earlier, GUS collects information on household income using a single question generally asked of the mother, who we know tend to report lower incomes.

Figure 2.1 Percentage of children living in income poverty 2005/06-2008/09, according to FRS and GUS

2.4 Measuring persistent poverty in GUS

Research on low income has found that individuals experience different durations of low income (see e.g. DWP, 2007b; Smith and Middleton, 2007, which includes a summary of ways in which persistent poverty is measured). This implies that the low-income population is heterogeneous, comprised of those who experience low income for varying lengths of time. This report uses four years of GUS data to investigate issues of persistent poverty.

The choice of the length of period over which to observe household income is restricted by the availability of GUS data. At the time of the analysis four sweeps of GUS data were available, covering the period from 2005/06 to 2008/09, and we use them all in this research. Having a short observation period means that there is relatively little information from which to categorise patterns of low income. Categorizing patterns of low income over short periods is complicated by the fact that some starts and ends of poverty spells are not observed in the data (the problem of 'censoring'). However, having a short observation period means that attrition is less of an issue and the sample for whom four waves of data are available are more representative (and larger) than samples using longer observation periods.

This research therefore uses a relatively straightforward summary measure of persistent poverty. The methodology used to identify persistently poor families mirrors that developed for the HBAI series ( DWP, 2007b) and used in Opportunity for All ( DWP, 2007a). This methodology counts the number of times a child was observed to be poor at the four consecutive annual GUS interviews. ¹²

Figure 2.2 presents a count of the number of times a family had income below the low-income threshold between 2005/06 to 2008/09 - from a minimum of zero (not below the low-income threshold in any of the four years) to a maximum of four (below the low-income threshold in all of the four years).

Figure 2.2 Number of times families were living in poverty, 2005/06-2008/09

Note: 20 per cent of the birth cohort and 21 per cent of the child cohort did not answer the income question in all four sweeps. These households are excluded from Figure 2.2.

Based on the number of times a family is in income poverty, our longitudinal poverty status classifies GUS children into three categories:

• 'Not poor' - Not poor at any of the four annual interviews
• 'Temporary poor' - Poor at one or two interviews
• 'Persistently poor' - Poor at three or four interviews

Persistent poverty, therefore, is defined as having low income at three or four of the four annual GUS interviews from 2005/06 to 2008/09. ¹³

Before presenting the proportion of children in each category, the next stage of categorisation sought to impute information for those households with missing income information in one wave of GUS. This affected one in five GUS panel households and meant that we were able to improve the sample size for later analysis.

2.4.1 Poverty status imputation

As mentioned above, not all panel households answered the income question in every sweep - 20 per cent of the birth cohort and 21 per cent of the child cohort had missing income information. However, the majority of these households had in fact answered the income question in three of the four sweeps. We therefore decided to impute the longitudinal poverty status for these households and the procedure we used is explained in detail in Appendix 1.

Table 2.3 shows the unweighted sample size in each cohort; those who reported income in all sweeps; the number with missing income in one sweep only and the final sample size after imputation. As the table shows, we were able to impute for the vast majority of households (480 of the 550 in the birth cohort, and, 261 of the 304 in the child cohort).

Table 2.3 Birth and child cohort sample sizes before and after poverty status imputation

	Birth cohort	Child cohort
Complete sweep 1-4 panel sample	3,844	2,100
With income in all sweeps	3,118	1,680
Missing income in 2-3 sweeps	176	116
Missing income in 1 sweep (impute)	550	304
Imputations made	480	261
Final analysis sample	3,598	1,941

2.5 The incidence of persistently poor children in Scotland

Over one fifth of GUS children (24 per cent of the birth cohort and 21 per cent of the child cohort) were in persistent poverty during the period 2005/06 to 2008/09. Nearly six in ten (58 per cent of each cohort) GUS children lived in families which had income above the low-income threshold in all of the four years, while one in five (18 per cent of the birth cohort and 20 per cent of the child cohort) were poor in one or two years - the temporary poor. ¹⁴ The number of GUS children in each of the longitudinal poverty categories is also given in Table 2.4 and demonstrates adequate sample sizes for further analysis. ¹⁵

Table 2.4 Longitudinal poverty status of GUS children, 2005/06-2008/09

	Birth Cohort		Child Cohort
Longitudinal poverty status	Per cent	Unweighted count	Per cent	Unweighted count
Not poor	58	2,333	58	1,261
Temporary poor	18	611	20	356
Persistently poor	24	654	21	324
All	100	3,598	100	1,941

The proportion of GUS children (42 per cent of each cohort) that experienced poverty at least once in a four year period is similar to that found by Barnes et al., (2008) for British families with children using FACS data from 2001 to 2004. However, disproportionately more GUS children experience persistent poverty (24 per cent of the birth cohort and 21 per cent of the child cohort, compared with 12 per cent of the FACS families).

There are a number of reasons for these differences across the two surveys, most notably the different way income is collected and the different samples of children across the two surveys. FACS uses a similar methodology to collect income as the FRS, so collects far more detailed information than GUS. Here it is worth mentioning how using banded income, rather than actual income, may result in fewer observed changes in income from one year to the next - which may also help to explain why more children remain in poverty (and hence are persistently poor). Assuming the family composition remains the same, a larger change in household income is required for a GUS family to move across the poverty threshold as they would have to report a different band from the previous year. When actual income is used, a very small change in household income can push a family across the low income threshold.

Another difference between the two studies is that GUS focuses on families with at least one 'young' child ( i.e. aged 3-4 years or 5-6 years in 2008/09), whereas FACS includes families with dependent children of any age. It is possible that families with young children are more likely to be income poor as they are more likely to have one parent not in paid work, due to childcare responsibilities. Also, the higher poverty rates found in GUS, whether due to the aforementioned reason or simply measurement error, are also more likely to lead to higher persistent poverty rates by definition.

Previous research has shown that cross-sectional survey measures underestimate the number of families who experience poverty over time (Barnes et al., 2008). Again, we see here that although approximately 3 in 10 households were poor at any one sweep of GUS, when looked at over a four-year period we see that over 4 in 10 experienced poverty at least once.

Table 2.5 illustrates the duration of poverty for children who are currently poor (that is, poor in the last sweep of GUS in 2008/09). Here we see that the majority of poor children (82 per cent of the birth cohort and 76 per cent of the child cohort) have been living in persistently poor households over the previous four years. This suggests that poverty can be a lasting experience, although the previously mentioned issues with the way GUS collects income may mean that households are less likely to report a change in income. ¹⁶

Table 2.5 Longitudinal poverty status of GUS children income poor in the latest sweep (2008/09)

	Birth Cohort		Child Cohort
Longitudinal poverty status	Per cent	Unweighted count	Per cent	Unweighted count
Not poor	N/A	N/A	N/A	N/A
Temporary poor	18	156	24	104
Persistently poor	82	541	76	280
All poor in 2008/09	100	697	100	304

Despite various limitations with the way GUS collects income information, this longitudinal measure of poverty can be used to compare GUS children with different durations of living in low-income households. The rest of this report adopts these categories to investigate the circumstances of children living in persistent poverty. This begins by looking at the types of children who are persistently poor (Chapter 3) and then moves on to focus on their health and developmental outcomes (Chapter 4).