Bringing Hope, Building Futures: Tackling child poverty delivery plan 2026-2031 – annex 5: Cumulative Impact Assessment

4. Projection uncertainty

The figures presented in Section 3.1 are calibrated to align with the Poverty and Income Inequality in Scotland Statistics. This is achieved by modelling child poverty rates in the latest years where official statistics are available to identify the difference, before calibrating the model in future years to take account of this difference. This difference has grown in this publication, indicating a divergence between the modelling and the published statistics.

The latest model estimated relative child poverty at 5 percentage points below the observed statistics, and absolute child poverty 7 percentage points below the observed statistics. This divergence is larger than seen in the 2022 modelling, where the difference was 2 percentage points, and the 2024 modelling, where there was no difference between the modelled and observed relative and absolute child poverty rates.

This divergence is generally caused by different approaches in identifying social security take-up, with the statistics identifying observed take-up rates from the survey data. In contrast, the model first identifies eligibility for social security benefits based on the survey data, before applying an assumed take-up rate based on historical data. As such, this divergence could be caused by true falls in take-up rates of benefits for eligible persons, or due to increased under-reporting in the survey data compared with previous years. As such, we cannot determine whether the calibrated or pre-calibration figures are the most accurate. The pre-calibration figures have been presented below, but they do not replace the projections detailed in Section 3.1. Until we find robust evidence that our modelled figures are more accurate, it is prudent to use the calibrated figures.

Table 5 presents the pre- and post-calibration analysis for the baseline relative child poverty rate. This shows the modelled relative child poverty rate is fourteen percent pre-calibration in 2026-27, falling to thirteen percent pre-calibration in 2030-31. Table 4 presents the pre- and post-calibration analysis for the baseline absolute child poverty rate. This shows the modelled absolute child poverty rate is eight percent pre-calibration in 2026-27 through to 2030-31.

A further challenge caused by this divergence is that it is not currently possible to model an absolute poverty rate below 7 percent owing to the calibration always adding 7 percentage points to any modelled rates. This presents a challenge in modelling a credible path towards the child poverty targets, given the size of the calibration relative to the child poverty targets. However, this only manifests as an issue at lower rates of child poverty than the model or statistics currently suggest.

Of note, uncertainty surrounding calibration primarily affects the projections of outturn child poverty rates rather than the estimated impacts of the policy package or particular policies. This is because the policy scenario and the counterfactual scenario are likely to be similarly affected.

Further detail on calibration is available in Section 8 and in previous publications.

Source: SG analysis using UKMOD

Source: SG analysis using UKMOD