In our 2018 Welfare trends report, Chapter 4 described how we model the effects of universal credit on spending. This draws heavily on two models owned and operated by DWP - the Policy simulation model (PSM) and the Integrated forecasting model (INFORM). INFORM is a dynamic micro-simulation model that uses DWP administrative data to forecast the caseloads for different benefits. This box described how INFORM is used in our UC forecast and some of its limitations.
INFORM is a dynamic micro-simulation model that uses a 5 per cent sample of administrative data from DWP and HMRC systems that are merged on the basis of individuals’ National Insurance numbers to identify the combination of benefits received in each household. As this involves confidential individual-level data, the raw inputs to the model are not seen by OBR staff or members of the Budget Responsibility Committee.
INFORM produces a monthly profile of benefits received in each household across the forecast period. These projections are estimated from transition probabilities – i.e. the likelihood of moving onto, off or between any given benefits in any given month – based on historical data. It aligns the caseload projections for each benefit to our forecasts for total caseloads across the legacy benefits and tax credits. Finally, it applies the transition and migration rules to the monthly profiles to estimate the build-up of the UC caseload and the associated rundown of the legacy benefit caseloads. The scenario-based model then converts these INFORM caseloads into the hierarchical legacy benefit breakdown used to calibrate PSM outputs.
But INFORM has some disadvantages. Relying on historical data to model transition rates means that when the past is not a good representation of the future, outturns will deviate from forecasts. INFORM is based on data relating to the period from 2009-10 to November 2015. The divergence in recent years from the transition probabilities during that period mean that INFORM is no longer used in forecasting legacy benefit caseloads. We also cannot use it to model newly eligible cases. That said, it remains the best available source of information on receipt of more than one benefit by individual benefit units, and on the flows in, out and between them. By calibrating the INFORM outputs to our legacy benefit forecasts, we can reduce the forecast risk associated with the transition-probability inputs diverging from actual experience. The marginal savings approach also means that the data problems that led us to stop using INFORM for the legacy forecasts have less impact as they would affect both legacy and UC estimates, but would have little effect on the difference between them.
This box was originally published in Welfare trends report – January 2018