International analysis of child poverty – ukmod/euromod modelling

4. Decomposition

The headline decomposition results show that Scottish demographics, the Scottish labour market, and the interaction between these two terms work to widen the differences in child poverty against the comparator countries. Both labour market and demographics play a broadly similar role in explaining the differences between Scotland and the comparator countries. Of note, if Scotland had the same demographics as the comparator countries, it would have experienced an even faster reduction in original income poverty rates over the period 2011-2024.

The Scottish transfer policies were found to reduce child poverty by greater amounts than all comparator countries, ranging from 6 to 19 percentage points. However, the extent to which Scottish transfer payments depress child poverty appears to have declined between 2011 and 2024, which is the driving factor in the increase in child poverty in Scotland between those periods.

Findings for the general population with children are reflected to an even greater extent among lone parents. Child poverty rates based on original income among lone parents are generally around twice the rates reported for the wider population, both in Scotland and the comparator countries.

While families with young children obverse higher original income revlative poverty rates than the general population with children, there were several instances where relative poverty rates were below the general population when based on disposible incomes. This suggests that they are notable beneficiaries of the tax-benefit system. However, the influence of transfer payments in mitigating child poverty in Scotland decreased considerably between 2011 and 2024 relative to our comparators.

Finally, focusing on families with three or more children showed higher rates of poverty than the general population with children, with a similar decrease in the influence of transfer payments over time relative to our comparators.

Headline results for the decomposition analysis are reported in Table 4.1. Discussion of the counterfactual proxies for demographics and the labour market upon which our analysis of child poverty rates is based (Appendix B). The top two rows of this table report child poverty rates evaluated on original incomes by year and country, as discussed in Section 3. In this regard, child poverty rates in Scotland are reported to exceed those of all comparator countries in all years, which is a design feature of the analysis as discussed in Section 2. The extent to which Scottish child poverty rates evaluated on original incomes exceed those of comparator countries is reported in row (3) of the table. These differences are greatest with Slovenia in 2011, and smallest in Croatia in 2024.

The last two columns of Table 4.1 indicate that child poverty rates evaluated on original income declined between 2011 and 2024 in Scotland, Croatia and Slovenia, with the largest decline observed in Scotland.

Row (4) of Table 4.1 reports the effects on child poverty of replacing the matched Scottish data for the comparator country with the data for Scotland. Positive figures indicate the Scottish demography increases child poverty relative to the demography observed for the comparator country. This permits consideration of the extent to which demographic differences between countries explains observed differences in child poverty. Hence, statistics reported in row (4) of the table permit consideration of the question:

What is the impact of Scottish demographics in explaining differences in outcomes (child poverty rates or reduction in child poverty rates), with respect to a comparator country?

Row (4) of the table indicates that Scottish demographics work to increase child poverty measured in terms of original income for all countries and years considered for analysis. Effects of demography are largest relative to Slovenia, where the statistics reported in the table indicate that replacing Slovenian with Scottish demography would increase child poverty by 3.4 percentage points in 2011, rising to 4.6 percentage points in 2024. Furthermore the last two columns of row (4) reported in Table 4.1 indicate that the influence of Scottish demographics on child poverty rates relative to comparators has increased through time, albeit to a smaller extent when Croatia is considered (0.4 percentage points) than when Slovenia is considered (1.2 percentage points).

If Scotland had the same demographics as the comparator countries, it would have experienced an even faster reduction in original income poverty rates over the period 2011-2024. Discussion in Appendix B.1 highlights the associated role of education underlying these differences, where the population shares with less than higher secondary education are reported to be considerably higher in Scotland than the comparator countries. Nevertheless, care should be exercised when interpreting these findings, given uncertainty concerning the validity of the education comparisons upon which analysis is based.

Source: Authors’ calculations on simulated data derived UKMOD (Scotland) and EUROMOD (Croatia, Slovenia, Denmark and Finland).

Notes: All statistics report poverty rates as percentages of children residing in households with equivalised household income below 60% of the median of the respective distribution (distinguished by country, year). All incomes measured before housing costs and equivalised using the revised OECD scale. Data for Scotland projected from input data for 2012 and 2022 derived from the Family Resources Survey (FRS). Data for all other countries projected from input data for 2012 and 2019 derived from the European Union Survey of Income and Living Conditions (EU-SILC). See Appendix A.1 for further detail.

Comparator (2,8) refers to country in each column heading. Scotland (1,7) refers to the poverty rate seen in Scotland for the year in question. (3) Difference reports poverty rate on original income for Scotland (1) less comparator country (2). Differences due to (4) demography, (5) labour market and (6) interaction identified as described in 0. Difference (9) reports poverty rate on disposable income for Scotland (7) less comparator country (8). Difference due to original income (10) equal to (3), and difference due to transfer payments (11) given as a residual, such that (9) = (10) + (11). Sample sizes in table reported for comparator countries. Sample sizes for Scotland are 6,638 in 2011 and 3,887 in 2024.

Row (5) of Table 4.1 reports the effects on child poverty of replacing Scottish earnings, hours worked, and unemployment with projections reflecting labour market characteristics in comparator countries. This permits consideration of the question:

What is the impact of the Scottish labour market in explaining differences in outcomes (child poverty rates or reduction in child poverty rates), with respect to a comparator country?

Row (5) of the table indicates that the Scottish labour market also tends to widen child poverty, with the exception of Croatia in 2011. Averaging the differences due to demography across all countries and all years (row (4)) obtains a value of 2.9%, contrasting with 2.6% when averaging the differences due to labour markets (row (5)). These statistics consequently suggest that broadly similar roles are played by demographics and labour markets in explaining differences between child poverty rates in Scotland and the comparator countries. Looking across countries in row (5), the effects of replacing Scottish labour market conditions tend to be smaller with respect to Croatia, and bigger for Denmark. The role of the labour market also appears to have increased appreciably through time for explaining child poverty differences between Scotland and both Croatia and Slovenia. Discussion in Section 3 and Appendix B.2 highlights the role played by employment rates and hours of work, particularly among families with higher numbers of dependent children.

The effects reported in row (6) of Table 4.1 indicate that interactions between demographics and labour market outcomes play a significant role in explaining differences in outcomes between Scotland and the comparator countries. The interaction effects are quite large, accounting for over half the differences in child poverty evaluated on original incomes in all considered cases but Denmark. These interactions play against Scotland both in 2011 and in 2024, although their effects diminish over time for Croatia and Slovenia (hence the negative contribution reported in the last two columns).

The large interaction effects reported in row (6) of Table 4.1 are alluded to by the descriptive statistics reported in Section 3. Specifically, hours of employment in Scotland are shown to systematically fall short of those of comparator countries, with these differences widening as the number of dependent children increases, particularly among single parent households. Lower employment in households with many children tends to depress earnings, which supports higher child poverty rates.

Child poverty rates evaluated on disposable incomes are reported in rows (7) for Scotland and (8) for comparator countries. The statistics reported in rows (7) and (8) are for disposable income, as discussed in Section 3. In contrast to the measures of poverty reported for original income, Scottish child poverty evaluated on disposable income is lower than reported for Croatia and Slovenia for 2011, and for Croatia in 2024. Hence, relative to these instances, the Scottish transfer system more than off-sets the higher child poverty evaluated on original incomes. This conclusion is confirmed by rows (10) and (11) reported in the table.

Row (11) of Table 4.1 indicates that the Scottish transfer system acts to reduce child poverty, relative to all comparator countries and all years considered for analysis. In part, this reflects the higher poverty rates evaluated for poverty measured on original income: in context of a progressive tax and benefit system, a higher redistributive effect will be observed in context of more unequally distributed original incomes. Nevertheless, while the relative effects of Scottish transfer policy are insufficient to fully off-set coincident differences for child poverty measured with respect to original incomes, they are substantial. Scottish transfer policy is found to reduce child poverty by a further 6 percentage points relative to Denmark in 2024, rising to 19 percentage points for Croatia in 2011.

However, relative to both Croatia and Slovenia, the extent to which Scottish transfer payments depress child poverty appears to have declined between 2011 and 2024, by 7 percentage points compared with Croatia and 4 percentage points compared with Slovenia (row 11). The increase in child poverty observed in Scotland between the periods (row 9) was driven by the decline in impact of the transfer payments. This was marginally off-set by reductions in child poverty driven by changes in original income (row 3).

Table 4.2 reports decomposition statistics focussing only on child poverty rates calculated for lone parents. Comparing the statistics reported in Tables 4.1 and 4.2 reveals that the above discussion of statistics reported for the general population generally extends in exaggerated form to lone parents. This general observation highlights the role of lone parents underlying the population aggregate statistics reported in Table 4.1.

Child poverty rates based on original income among lone parents are generally approximately twice the rates reported for the wider population, both in Scotland and the comparator countries. Row (1) of Table 4.2 in particular indicates the high incidence of low original incomes among lone parents in Scotland, with child poverty rates equal to 73 percentage points in 2011, falling to 70 percentage points in 2024. As mentioned previously in this section, these low original incomes are alluded to by the limited employment hours among lone parents in Scotland that are further discussed in Appendix B.2.

Notes: See Table 4.1. Sample sizes in table reported for comparator countries. Sample sizes for Scotland are 407 in 2011 and 133 in 2024.

Some interesting disparities can be identified that contrast with the basic similarities between the statistics reported in Tables 4.1 and 4.2. First, row (5) of Table 4.2 indicates that labour market differences between Scotland and comparator countries help to explain a more substantial proportion of differences in child poverty rates among lone parents that the wider population with children. Again, this result is alluded to in Section 3, as displayed in Figure 3.3.

Secondly, row (6) of Table 4.2 indicates that the interaction effect plays a smaller role in explaining child poverty differences between Scotland and the comparator countries among lone parents than it does for the wider population with children. In part, this is a counter-part to the larger role played by labour market differences, as discussed above. Yet this conclusion also depends on a persistent role played by demography; the increase in labour market effect could have been associated with a decline in the effect of demography with a constant interaction effect. Note that it would also be reasonable to anticipate a decline in the demographic effect reported in Table 4.2 as statistics are limited to single parents which limits the scope for demographic variation between countries. That demographics continue to help explain a substantial share of child poverty differences between Scotland and the comparator countries in Table 4.2 suggests that factors other than relationship status and presence of children are important in underlying measures of child poverty. In this regard, discussion in Section 3 highlights the potential role of education differences between countries.

Child poverty rates among families with children up to 1 year old are reported in Table 4.3. Child poverty rates of families with young children evaluated on original incomes (top panel of Table 4.3) remain higher than for the wider population (Table 4.1).

However, differences between child poverty rates evaluated on disposable incomes reported for families of young children (Table 4.3) are in some cases lower than for the wider population (Table 4.1). This is true for Scotland in 2011, Croatia in 2011 and 2024, and Denmark in 2024. In these cases, tax and benefit payments work to more than off-set the higher rates of child poverty evaluated on original incomes. Comparing the bottom rows of Tables 4.1 and 4.3, however, indicates that the influence of transfer payments in mitigating child poverty among families with young children is muted in Scotland relative to the comparator countries, and declined considerably relative to Croatia and Slovenia between 2011 and 2024 (last two columns).

Notes: See Table 4.1. Sample sizes in table reported for comparator countries. Sample sizes for Scotland are 269 in 2011 and 121 in 2024.

Notes: See Table 4.1. Sample sizes in table reported for comparator countries. Sample sizes for Scotland are 323 in 2011 and 192 in 2024.

Table 4.4 reports statistics for child poverty among families with three or more children. Child poverty rates are elevated among families with three or more children relative to the wider population, both when poverty is evaluated based on original income and disposable income. As is generally the case for the population groups discussed above, the bottom two rows of Table 4.4 indicate that differences in distributions of original income tend to exacerbate measures of child poverty among families with three or more children in Scotland relative to comparator countries, at least partly off-set by the respective tax and benefit systems. Also as discussed previously, the weakening of the redistributive effect of the tax and benefit system in Scotland relative to Croatia and Slovenia between 2011 and 2024 is a key driver of divergences in child poverty rates across countries for families with three or more children.

Concerning factors underlying child poverty rates evaluated on original incomes among families with three or more children, row (5) of Table 4.4 highlights the role played by labour market differences between Scotland and the comparator countries. In common with the statistics reported for lone parents (Table 4.2), these are found to contribute positively to differences reported for all comparator countries, accounting for around half of the differences on average. As noted in previous discussion, these effects are alluded to in Section 3 where it is shown that the relatively lower employment engagement in Scotland in comparison to comparator countries tends to widen with the number of dependent children in a benefit unit.