Stamp duty land tax (SDLT) is one of the more volatile sources of receipts. In our 2016 Forecast evaluation report, this box identified a number of reasons why forecasting SDLT receipts is challenging, including the concentration of receipts in a small proportion of expensive properties and the effects of significant policy changes.

Stamp duty land tax (SDLT) is one of the more volatile sources of receipts – the standard deviation of annual growth over the past five years has been 11.9 per cent, compared to just 1 per cent for overall receipts. In recent years, SDLT receipts have been a large source of fiscal forecasting errors. In line with that, we have revised our SDLT forecasts proportionally more than for any other major tax (bar the even-more-volatile CGT and North Sea oil and gas revenues).

There are many reasons why forecasting receipts from a tax like SDLT can be challenging. Three in particular are worth highlighting:

  • it is difficult to map our economic determinants to the true tax base. Property transaction taxes involve a very small minority of all potential taxpayers each year. This differs from most other taxable activities, where most potential taxpayers will be actual taxpayers each year. Someone who earns income this year is likely to earn a similar amount of income next year, so in an income tax micro-simulation model the data points representing the tax base are simulations of actual taxpayers. By contrast a house purchased this year is unlikely to be bought and sold again next year. So SDLT micro-simulations are based on a set of properties that are unlikely to transact again in the early years of the forecast. There are around 28 million dwellings in the UK, but only around 1.2 million residential property transactions a year. Any changes in the composition of transactions relative to the simulated tax base will be a source of fiscal forecasting error;
  • the tax schedule is very progressive – a £200,000 residential transaction pays £1,500 in tax, whilst a transaction for ten times this price (£2,000,000) pays over one hundred times the tax (£153,750). In 2015-16 around a quarter of revenue came from the top 1 per cent of transactions (which is a similar concentration to income tax). In the past, trends in prices and turnover of a small number of highly priced prime London properties have followed a different trajectory to the market as a whole. This compositional effect has led to a volatile effective tax rate and has been responsible for much of our fiscal forecasting error; and
  • the SDLT regime has been subject to large policy changes in recent years. The tax rate for all residential transactions was changed in December 2014 from a ‘slab’, applied on the whole value, to a marginal ‘slice’, applying on the value over a given threshold. Similar changes were made in March 2016 for non-residential property. In November 2015, the Government pre-announced a 3 per cent SDLT surcharge on ‘additional properties’ that would come into effect from 1 April 2016. We have looked into the effects of pre-announcing changes to SDLT in a working paper published alongside this FER. It shows taxpayers react strongly to the chance to bring forward transactions to reduce their tax bill – in the case of the additional properties surcharge, much more so than we expected. These policy changes, especially when they are pre-announced, add uncertainty to our forecasts. They mean that historic relationships are now less useful when producing our new forecasts. The new ‘slice’ regime has further increased the concentration of SDLT receipts among high-priced transactions, meaning the mapping of determinants to the tax base is likely to become even more challenging.a

The errors reported in this FER

Our March 2014 forecast for 2015-16 over-estimated revenues by £3.6 billion. In part that was because between March 2013 and March 2014 we revised up our 2015-16 forecast substantially (£5.1 billion or 55 per cent) in response to strong receipts growth at the time from the London property market. Only a small amount of the error (£0.2 billion) reflects weaker-than-expected UK-wide house prices or transactions. The much larger fiscal forecasting error of £2.7 billion mainly reflects off-model adjustments that we made to reflect an assumption of continued strong growth in the top-end of the London property market that did not materialise.

By the time of our March 2015 forecast, both residential SDLT policy and our forecast model had changed considerably. To model the residential ‘slice’ policy change it was necessary to use a micro-simulation model. At the time we used a 10 per cent sample of historic transactions and grew them in line with our price and transactions forecast, correcting for the distortive effects of the ‘slab’ regime as well as applying behavioural elasticities proportional to the tax change.b In March 2015 this policy costing model became our forecasting model. Our time series model had produced higher forecasts than the micro-simulation model, so even if there had been no policy change, the modelling change would have led to a lower forecast. Whilst the overall March 2015 error is much lower this is only because the economic and fiscal forecasting errors were largely offsetting, and a sizeable fiscal forecast error remains. We included an estimate of the greater-than-expected forestalling from the ‘additional properties’ surcharge – described in the accompanying working paper – within the ‘policy error’ part of the decomposition. It appears that around 60,000 transactions were brought forward to avoid paying the new surcharge, an estimate that is subject to considerable uncertainty.

This box was originally published in Forecast evaluation report – October 2016

a For example, the boroughs of Westminster and Kensington and Chelsea account for approximately 0.01 per cent of the UK’s land area with less than 1 per cent of the population and dwellings. In 2014-15, 10,000 transactions in these boroughs accounted for £0.9 billion (12.5 per cent) of total UK residential SDLT receipts. In 2015-16 the number of transactions decreased to 9,250, but residential yield increased to £1 billion (14.0 per cent of the total).
b OBR supplementary release 22 January 2015, Stamp duty land tax policy costing elasticities – December 2014