Growing up in Scotland: the circumstances of persistently poor children

chapter 3 WHICH CHILDREN ARE MOST LIKELY TO BE PERSISTENTLY POOR?

The aim of this chapter is to identify the children most likely to be persistently poor and compare these with children in short-term poverty and those who avoid poverty. Various background characteristics of children are explored, including family size and composition, parents' work status, education, health, tenure and characteristics of the local area. We first provide a descriptive picture of the types of children in each poverty category. We then use multivariate regression analysis to unravel which characteristics are related to an increased risk of persistent poverty when holding other, potentially confounding, characteristics constant.

The key findings from this chapter are:

• Certain children were more likely than others to experience persistent poverty. These included those in lone-parent families, larger families, families with a young mother, families with parents with low education, and families who live in rented housing, particularly social-rented housing (Section 3.1).
• Some of these factors may not be driving persistent poverty, they may be consequences of being poor, and for others the relationship with poverty is inherently complex. Multivariate analysis - designed to identify the risk of persistent poverty, while controlling for the impact of possibly confounding influences - shows that family work status is the factor that bears most on the risk of persistent poverty. Being continuously out of work is the key driver of persistent poverty (Section 3.2).

3.1 The types of children most at risk of persistent poverty

This section looks at the risk of poverty duration according to a range of socio-demographic and socio-economic characteristics. Particular attention is paid to identifying the types of children most at risk of persistent poverty. In this section we look descriptively at the association between each characteristic and poverty duration. This provides an early indication of some of the underlying factors that may be linked to persistent poverty.

We identified a number of factors that are likely to be associated with poverty duration, covering socio-demographic background, socio-economic characteristics and features of the local area. Some of these were measured just once during the period under investigation, such as ethnicity. Other factors are more dynamic by nature and we take advantage of the fact that GUS is a longitudinal study to construct measures of change (so called time-varying factors), such as changes in the number of children in the family (perhaps due to a new born or an older child leaving home). Table 3.1, along with the relevant section in Appendix 1, presents a detailed description of all factors and explains how they were measured using the GUS data.

Table 3.1 Factors included in analysis of risk of poverty duration

Variable	Year measured	Categories
Socio-demographic
Sex of child	sweep 1	Boy, Girl
Ethnic group of mother	sweep 1	White, Ethnic minority communities
Age of mother at birth of GUS child	sweep 1	Under 25, 25-29, 30-34, 35 and over
Family type transitions	sweep 1 to sweep 4	Stable couple, Couple who separated, Stable lone parent, Lone parent who re-partnered
Number of children at Sweep 1	sweep 1	1, 2, 3 or more
Change in the number of children	sweep 1 to sweep 4	No change, Increase, Decrease
GUS child is mother's firstborn	sweep 1	Yes, No
Mother's health status	sweep 1 and sweep 4	No health problems (at sweeps 1 or 4), Reduced health problems (at sweep 1 but not at sweep 4), Developed health problems at (not at sweep 1 but at sweep 4), Persistent health problems (at both sweep 1 & 4)
Socio-economic
Average Work Intensity	sweep 1 to sweep 4	A measure of household employment. See Appendix 1 for detailed description
Mother's education	sweep 1	Higher grade or above, Standard grade or lower
Father's education	sweep 1	Higher grade or above, Standard grade or lower
Social class at Sweep 1	sweep 1	Managerial/professional, occupations, Intermediate, Small employer/own account, Lower supervisory/technical occupations, Semi-routine and routine occupations, No-one in employment
Family has a car	sweep 1 to sweep 4	At none of the sweeps, At 1-3 sweeps, At all four sweeps
Whether family uses childcare	sweep 1 to sweep 4	Not using at Sweeps 1 & 4, Started using, Stopped using, Using at Sweeps 1 & 4
Tenure	sweep 1	Owner occupier, Social renter, Private renter, Other
Local area
Urbanization	sweep 1	Large urban, Other urban, Towns, Rural
Area deprivation level ( SIMD quintiles)	sweep 1	Least deprived quintile, 2nd quintile, 3rd quintile, 4th quintile, Most deprived quintile

Tables 3.2 - 3.4 present the risk of poverty duration for each group of factors. The results are presented separately for the birth and child cohorts. However, since the patterns of associations are generally very similar for both cohorts, we do not refer to specific cohorts when describing the results (unless the cross-cohort differences are significant).

Table 3.2 presents the risk of poverty duration by socio-demographic background. There are no differences with respect to the sex of the child but children from ethnic minority communities are more at risk of persistent poverty than White children. ¹⁷ Children with young mothers (under 25) faced a higher risk of persistent poverty than those with older mothers; as did those that had lived in a lone-parent family at any time during the observation period (compared to those permanently living in a couple family).

Table 3.2 Risk of poverty duration by socio-demographic background Row % (per cohort)

	Birth cohort				Child Cohort
	Not poor	Temporarily poor	Persistently poor	Unweighted count	Not poor	Temporarily poor	Persistently poor	Unweighted count
Sex of child
Boy	59	17	24	1,817	60	20	20	959
Girl	58	19	23	1,715	60	20	20	921
Ethnic group of mother
White	59	18	23	3,452	60	20	20	1,838
Ethnic minority	34	17	49	80	44	13	44	42
Age of mother at birth of GUS child
Under 25	22	27	50	625	28	34	39	345
25-29	61	19	20	816	59	21	19	423
30-34	74	13	12	1,243	75	13	12	671
35 and over	72	14	14	848	75	12	13	441
Family type transitions
Stable couple	73	15	12	2,883	76	15	9	1,471
Couple who split up	30	30	39	159	34	36	30	68
Lone parent who partnered	13	38	49	158	16	44	40	98
Stable lone parent	7	20	73	332	14	27	59	243
Number of children at sweep 1
1	62	18	21	1,650	61	23	16	612
2	62	18	20	1,253	65	19	16	870
3+	44	19	37	629	46	18	35	398
Change in the number of children
No change	58	19	23	2,251	62	20	18	1,340
Increase	62	16	21	1,143	59	19	22	444
Decrease	27	24	49	138	36	21	43	96
GUS child is mother's firstborn
No	56	18	25	1,831	58	19	23	1,006
Yes	60	18	22	1,701	62	21	17	874
Mother's health status
No health problems	62	18	21	2,624	63	19	17	1,379
Reduced health problems	50	21	29	233	50	20	31	130
Developed health problems	50	19	31	369	57	19	24	185
Persistent health problems	45	19	36	306	46	24	29	186
All	58	18	24	3532	60	20	20	1880

Children in stable lone-parent families were in the family type most at risk of persistent poverty. For example, 73 per cent of birth-cohort children who were in lone-parent families throughout the period were persistently poor, compared with only 12 per cent of children from stable couple-families. Children from larger families also faced a higher risk of persistent poverty as did children whose mother reported health problems or disability, particularly if these were longer-term.

Table 3.3 looks at socio-economic factors. As expected, socio-economic status of the main earner is a very strong predictor of persistent poverty. Virtually all families where no-one was in employment were poor at some point, and about 8 out of 10 of such families experienced persistent poverty. Among the families where the main earner was employed, the risk of persistent poverty decreased in line with increases in socio-economic status of the job. For example, only about 3 per cent of the families where the main earner was employed in a professional/managerial job experienced persistent poverty, compared with about a quarter of the families where the main earner had a semi-routine or routine occupation.

Similarly, the average work intensity ( AWI) ¹⁸ in the household strongly shapes the risk of persistent poverty and poverty in general. Only about 10 per cent of families where all adults worked full-time for virtually the whole period under investigation ( AWI>75%) were affected by any form of poverty, and only 1 per cent of such families experienced persistent poverty. The results were almost as low in the case of the families with AWI in the range 51-75%. Families with AWI of 26-50% had markedly higher risks of persistent poverty - about half of such families experienced poverty at some point and one in five were persistently poor. As expected, the families who only used up to a quarter of their workforce potential faced highest risk of poverty: almost 9 out of 10 such families lived in persistent poverty and virtually all of the remaining 10 per cent experienced temporary poverty at some point between 2005/06 and 2008/09.

It is evident that the risk of persistent poverty is related to parent's education: Higher grades or above offers a good protection against persistent poverty both in the case of mothers' and fathers' education. Ownership of a car was also linked to the risk of persistent poverty (although this, like other factors, could also be an outcome of poverty). Also, families who did not use childcare faced persistent poverty. This could be for a variety of reasons; including having one parent at home caring for a very young child who looks after the family through choice or being constrained by uneconomical childcare costs. Finally, social renters faced a higher risk of persistent poverty than private renters and owner-occupiers.

Table 3.3 Risk of poverty duration by socio-economic background Row % (per cohort)

	Birth cohort				Child Cohort
	Not poor	Temporarily poor	Persistently poor	Unweighted bases	Non poor	Temporarily poor	Persistently poor	Unweighted bases
Mother's education
Higher grade or above	70	16	14	2,755	69	18	13	1,466
Standard grade or lower	28	24	49	777	31	26	42	414
Father's education
Higher grade or above	69	17	14	2,590	69	17	14	1,393
Standard grade or lower	34	21	45	942	37	27	37	487
Social class at sweep 1
Managerial/professional	87	10	3	1,434	88	9	3	753
Intermediate occupations	65	22	13	263	66	29	5	139
Small employer/own account	60	23	18	334	65	23	12	176
Lower supervisory/ technical occupations	68	21	11	473	69	25	7	246
Semi-routine and routine occupations	45	28	27	621	48	29	23	330
No-one in employment	0	18	81	407	2	24	75	236
Average work intensity 19
76-100%	89	10	1	623	89	9	1	348
51-75%	80	16	4	1,316	81	15	4	659
26-50%	47	30	22	685	49	32	19	378
0-25%	1	10	89	327	1	15	85	173
Family has a car
At all four sweeps	73	16	10	2,877	75	16	9	1,540
At 1-3 sweeps	17	32	51	343	17	40	43	167
At none of the sweeps	9	15	76	312	9	24	67	173
Whether family uses childcare
Both at sweep 1 & 4	68	17	15	2,022	68	18	14	1,250
At sweep 1 but not at sweep 4	46	18	36	208	60	20	19	262
At sweep 4 but not at sweep 1	52	22	26	768	37	27	36	219
Neither at sweep 1 nor at sweep 4	37	18	45	534	32	24	43	149
Tenure
Owner occupier	80	13	6	2,526	82	14	4	1,353
Social renter	14	27	59	732	15	34	51	393
Private renter	33	26	41	183	32	24	44	101
Other	31	29	40	91	36	22	42	33
Total	58	18	24	3532	60	20	20	1880

Table 3.4 shows the risk of poverty duration by indicators of the local area. Overall, families living in cities faced higher risk of poverty in general, and persistent poverty in particular, than families living in towns or in rural areas. The risk of persistent poverty was proportionate to the area deprivation level, represented by the value of the Scottish Index of Multiple Deprivation characterising the area: the higher the SIMD value, the higher risk of poverty.

Table 3.4 Risk of poverty duration by area indicators Row % (per cohort)

	Birth cohort				Child Cohort
	Not poor	Temporarily poor	Persistently poor	Unweighted count	Non poor	Temporarily poor	Persistently poor	Unweighted count
Urbanization
Large urban	55	17	28	1,187	57	22	21	593
Other urban	56	19	25	1,180	59	18	23	627
Town	60	20	20	481	63	22	15	279
Rural	67	19	14	684	65	18	16	381
Area deprivation ( SIMD quintiles)
Least deprived	87	9	4	809	87	10	3	451
2	72	17	11	778	74	18	9	444
3	63	18	19	729	61	21	18	393
4	46	22	32	580	46	24	31	284
Most deprived	29	23	48	636	28	28	44	308
Total	58	18	24	3532	60	20	20	1880

3.2 Modelling the key risk factors behind the duration of poverty

Having investigated the separate relationships between the poverty duration and each of the factors, we now turn to multivariate analysis where we include all factors in a single statistical model. The main aim of this analysis is to identify which factors are associated with poverty duration, when accounting for other, potentially confounding, variables. We do so by specifying a statistical model using the poverty categories defined previously - 'no poverty', 'temporary poverty', 'persistent poverty'. Whereas some studies have used multinomial logistic regression for this analysis, we recognize here that the poverty categories are intrinsically ordered and hence we use an ordinal logistic regression model, and compare each poverty category to a 'shorter-duration poor' group ( i.e. comparing short-term poor to those who avoid poverty, and then long-term poor to short-term poor). ²⁰ The factors are represented by the same indicators that were used in the previous section (see Table 3.1 for details). Table A2.1 in Appendix 2 presents the odds ratios from the ordinal logistic models, estimated for each cohort separately. The interpretation of odds ratios is explained in Appendix 1.

It is important to note that the analysis presents significant relationships between the characteristics of families and the risk of persistent poverty - the analysis does not unravel any cause and effect in the relationship. For example, if there is a relationship between tenure and persistent poverty, where families in social rented housing are more likely to experience persistent poverty, the analysis cannot unravel whether living in social rented housing is a cause of persistent poverty. There may also be moderating factors, which may themselves increase the chance of a family experiencing persistent poverty. The main point to note is that the analysis presented here does not provide cause, furthermore respondents were not asked to attribute cause themselves.

Previous research ( e.g. by Adelman et al., (2003), Berthoud et al., (2004), Middleton (2006) and Barnes et al., (2006)) found that factors associated with persistent poverty include work status, ethnicity, health and age. Similar factors were found to play a role in the current research.

Factors significantly associated with persistent poverty in both cohorts were (see Table A2.1 for detail):

• Work intensity, low average work intensity
• Socio-economic status, low socio-economic status of the main earner
• Family type, children that had lived in a lone-parent family at any point during the observation period
• Age of mother, children with young mothers (under 25)
• Education, children whose parents (particularly mothers) had a lower level of education
• Tenure, children who lived in social-rented accommodation
• Local area, children who lived in deprived areas

Low average work intensity (0-25%) was by far the strongest predictor of increased risk of persistent poverty. This is not surprising, as this level of AWI corresponds to persistently workless families or those with working parents in only one of the four years under investigation on average. Also, socio-economic status of the main earner strongly influenced the risk of persistent poverty: the families where the main earner had lower socio-economic status faced higher risk of a longer experience of poverty than other families.

Some of the factors only had a significant effect in one data cohort. Notably, ethnicity appears to be a highly significant factor in the birth cohort, but not in the child cohort, which may be due to a smaller sample size in the latter case. Families with more than one child faced a higher risk of poverty in the child cohort but not the birth cohort.

Finally, it needs to be noted that some of the results of the regression analysis, such as the effect of decrease in the number of children or the child being the first child in the family, despite being significant are rather difficult to interpret and worthy of further investigation. ²¹