The transition to universal credit (UC) from legacy systems is a key component of our welfare spending forecast, as UC spending will represent 27% of total welfare spending by 2024-25. As detailed in our 2018 Welfare Trends report (WTR), we construct UC spending by forecasting the legacy system as though UC did not exist and then incorporate an estimate of the marginal cost of UC. This allows us to base as much of the forecast on as much administrative data as possible, but it does not directly reflect the real-world change in spending on legacy benefits as spending on UC rises. This approach is unavoidable at present but generates inevitable difficulties for our forecast. This box explored the challenges associated with this approach.

When preparing our fiscal forecasts, we are inevitably faced with surprises in the latest data relative to what would have been consistent with our most recent published forecast. How we interpret these surprises helps shape the revisions to our medium-term forecasts. That places a premium on being able to scrutinise the flow of administrative data against the assumptions underpinning our forecasts so that we can identify the source of upside or downside surprises.

For the £64.0 billion of spending on UC, and its legacy equivalents in 2019-20, that is extremely difficult. Analysis is hampered by the less than ideal, but unavoidable, way in which the forecast is constructed, and by uncertainties around the UC rollout. As detailed in our 2018 Welfare trends report, to understand the likely future trends in the UC claimant base it is essential to analyse and interpret past and present trends in the same base receiving legacy benefits, which requires us to be able to group claimants in a comparable way across both systems.

By way of illustration, if we were faced with a 1 per cent upside surprise in spending in the year in progress – i.e. around £0.6 billion, close to the upward revision we have made in this forecast – we would broadly have four alternative ways to interpret it, namely as:

  • A one-off event that will reverse next year. For example, if a build-up of arrears was identified and paid out, we would know that this would not be repeated in future. This would imply a 1 per cent fall in spending next year and not affect spending in 2024-25.
  • A one-off event that will persist. For example, if CPI uprating was 1 percentage point higher than assumed in the previous forecast. This would raise spending in every year by the same proportion, so in 2024-25 we would revise it up by £0.7 billion.
  • News about the cost of UC relative to the legacy system. UC is expected to cost more than the legacy system, with the net additional cost reflecting large and offsetting gross costs and savings associated with different types of UC cases. If the upside spending surprise were interpreted as a higher marginal cost of UC – perhaps because take-up of UC was higher than assumed – then it might be expected to build up at the same pace as the UC rollout progresses. Our latest forecast assumes that around a third of the eventual UC caseload will already have moved to UC in 2019-20 and that three quarters will have moved by 2024-25, so a 1 per cent surprise this year would imply a 2.5 per cent upward revision in 2024-25 – £1.8 billion.
  • News about the rate of spending growth. For example, it might appear that previous assumptions about trends in the prevalence of benefit receipt among the working-age population were too low. If this translated into spending growth being 1 percentage point higher each year, it would be revised up by 6.2 per cent in 2024-25 (£4.4 billion).

In some cases, it is relatively simple to interpret news. DWP knows how much it has spent addressing past underpayments, although there is much more uncertainty over how much future arrears payments will cost. And we know how actual uprating differs from the assumptions in our previous forecast. But in many cases it is much harder to attribute changes in spending to these individual interpretations. DWP analysts try to assess what the actual UC caseload would have looked like under the legacy system and to understand whether surprises in actual UC spending relate to surprises in the pace of rollout, in the take-up of UC relative to the legacy benefits, or in the amounts awarded relative to the amounts predicted by forecast models. This often proves particularly difficult for UC cases that bear most resemblance to HMRC-administered tax credits cases in the legacy system. Unfortunately, it is still difficult to track whether, and how, claimants leaving the tax credits system have moved into the UC system. And we cannot know who would have made a new claim to tax credits but now claims UC instead.

These inevitable difficulties are compounded by problems we have in interrogating the UC forecast. For example, management information is not routinely mapped onto forecast assumptions to test their veracity, while DWP’s modelling of the flows associated with our judgements about the stock of different cases – essential to understanding the UC rollout – involves a particularly cumbersome process. These and other factors mean that it is often almost impossible to judge with confidence whether news this year should affect assumptions about the level or growth of the legacy counterfactual forecast or the UC marginal cost.

We have revised UC and legacy equivalent spending in 2023-24 up by £4.9 billion in this forecast – essentially a ‘spending growth’ interpretation of news about the incapacity benefits caseload. The examples above show how different interpretations of equally-sized surprises could result in very different revisions to future medium-term forecasts.

This box was originally published in Economic and fiscal outlook – March 2020