In our 2018 Forecast evaluation report
we set out the latest conclusions of our review of fiscal forecasting models. This section sets out the conclusions of our review of some of our income tax forecasting models.
Self-assessment income tax
‘Self-assessment income tax is a relatively complex receipts stream because the underlying liabilities relate to a range of different income sources about which there is generally a lack of timely data. This generates uncertainty about the tax base assumptions fed into the forecasting model. The effective tax rate modelling is relatively simple, based largely on historical averages, forecast judgement and off-model adjustments for the effects of policy measures, of which there have been many in recent years.
The average absolute two-year ahead fiscal forecasting difference since our March 2012 forecast has been 7.9 per cent, which on a volatility-adjusted basis is higher than average across our fiscal forecasts. The average difference has been negative at 7.9 per cent, again larger than average, with all the past five March forecasts over-predicting receipts at the two-year horizon.
While much of the relatively poor forecast performance will relate to tax-base assumptions and policy costings, we consider modelling issues to have contributed too and have agreed the following priorities with HMRC for development work over the coming year:
- Deeper analysis of trends in the tax base and (with the ONS) consider how these are reflected in the National Accounts so that the model can appropriately transform elements of our National Accounts-based economy forecast into tax base assumptions.
- Improve the diagnostic outputs of the model to facilitate better model scrutiny.
- Improve the transparency of the assumptions used to adjust the SA liabilities forecast to the published cash receipts forecast.
PAYE and NIC1 integrated payment model
‘Income tax and ‘Class 1’ NICs paid through PAYE are the largest item in our receipts forecast, so while they are not particularly volatile relative to some taxes, even small percentage forecast differences can be material for our overall fiscal forecast. Our PAYE forecast model uses our earnings and employment growth forecasts to project the tax base from the latest base year and applies average effective tax rates (AETRs) derived from the ‘Personal Tax Model’ (the PTM, which we have reviewed separately). The model was extended to incorporate NIC1 payments in 2015, which had a number of advantages relative to the previous modelling approach (see our November 2015 EFO for more detail). Abstracting from uncertainty around the key determinants, the main fiscal forecasting challenge relates to the judgements that are fed into the model, in particular the current year receipts estimate and assumptions about differential earnings growth, which drive the shape of the income distribution that underpins the effective tax rate assumption.
For PAYE income tax, the average absolute two-year ahead fiscal forecasting difference since our March 2012 forecast has been 2.0 per cent, which on a volatility-adjusted basis is lower than the average across our fiscal forecasts. The average difference has been negative at 0.8 per cent, smaller than average, with three of our March forecasts since 2012 over-predicting receipts at the two-year horizon. NICs fiscal forecasting differences have tended to be larger and positive, although these mostly reflect differences from forecasts produced using the old model.
We have agreed the following priorities with HMRC for development work over the coming year:
- Exploit real-time information (RTI) data more to help inform key assumptions (including monitoring of the monthly data, generation of the in-year estimate and to inform the differential earnings growth assumption).
- Improve the plausibility and transparency of smaller receipts streams in the model (such as the tax on occupational pensions and company car tax).
Personal tax model (PTM)
‘The Personal Tax Model (PTM) is a micro-simulation model based on a survey of taxpayers’ liabilities. This model calculates the average marginal tax rate on additional income by taking account of reliefs, allowances and our assumptions about inflation and any differences in earnings growth at different points in the earnings distribution. The model then applies these tax rate forecasts to income growth to generate a receipts forecast. The main forecasting challenge relates to the shape of the income distribution and the differential earnings growth assumptions that are fed into this model to project the distribution.
The PTM model is a component of our overall PAYE income tax and NIC1 forecasts. The fiscal forecasting differences associated with these overall forecasts are discussed in relation to the PAYE and NIC1 integrated payment model.
We have agreed the following priority with HMRC for development work over the coming year:
- Exploit real-time information (RTI) data to help inform the projection of the base-year micro data.