International analysis of child poverty – ukmod/euromod modelling

Footnotes

1 Scottish Government’s Programme for Government 2025 to 2026

2 Best Start, Bright Futures: tackling child poverty delivery plan 2022 to 2026

3 Child poverty summary statistics, accessed 06 May 2025.

4 Tackling child poverty delivery plan annex 2

5 Tackling child poverty priority families overview

6 Summary Report, Danish, Finnish, Slovenian, and Croatian case studies.

7 UKMOD - Centre for Microsimulation and Policy Analysis

8 HOME | EUROMOD - Tax-benefit microsimulation model for the European Union

9 Data are also available for 2003, 2008 and 2009.

10 The 2020 wave of EU-SILC contains information on pre-pandemic (2019) incomes, but response rates to the survey questionnaire are affected by the pandemic, possibly leading to selection issues.

11 See appendix A.2 for definitions of original and disposable income.

12 The figures reported are unconditional differences in wages – not controlling for individual characteristics (such as education) and behaviour (such as work hours).

13 ONS income and earnings statistics guide

14 An example is the council tax band information for the residential home.

15 An example is income, which as we have seen is referred to the previous calendar year in EU-SILC, while it refers to the current calendar year in the FRS. As such, income needs to be uprated for an extra year in SILC-based models.

16 See DWP (2024), Households below average income series: quality and methodology information report FYE 2023.

17 The modified OECD scale before housing costs is also used.

18 On matching methods see, for example, Stuart, E.A. (2010), “Matching methods of causal inference: a review and a look forward”, Statistical Science, 25, pp. 1-21.

19 Benefit unit defined as a single adult or married couple and their dependent children. See Appendix A.2.2 for the definition of a dependent child. Married couples identified as people married or in a civil partnership or living with a partner.

20 This was done by identifying the “closest match” in the recipient database to a donor and adjusting their characteristics to align with the donor.

21 See Eurostat variable ED0-2. Averages between 2011 and 2019 are: Denmark: 28%; Croatia: 21%; Slovenia: 18%; Finland: 20%; and UK: 21%.

22 The DWP Household below average income series: quality and methodology information report FYE 2023 notes “Comparisons between the numbers with no qualifications in the FRS, LFS and the Census indicate that the FRS figures have historically overstated the numbers of working-age adults with no qualifications. As a result of the FRS mode change in FYE 2021 and FYE 2022, the raw FRS sample contained a much higher proportion of working age adults than in the years prior to the COVID-19 pandemic, and much lower numbers with no qualifications.”

23 Pre-adjustment differences in 2024 for Croatia, Slovenia, Denmark, and Finland averaged respectively 8.3, 7.5, 7.6, and 7.5%, which fall in the matched data to 0.6, 1.4, 0.5, and 1.0%.

24 A “log” is a standard non-linear scale adjustment. It adjusts for the fact that the distribution of earnings are compressed at low values because they are (generally) non-negative.

25 A naïve model assigns the same prediction to all observations. In Scotland, that would mean identifying no observation as unemployed, and in the comparator countries it would identify all observations as unemployed. For Scotland in 2011 such a model would be correct for 63% of the sample, in contrast to the 74% correct predictions reported for the regression model.