Taxbase Projections for Three-Year Settlements
Andrew Presland, DCLG
Definition of taxbase 1. A local authority’s taxbase for Formula Grant purposes1 is a measure of each authority’s potential capacity to raise council tax. It is defined as the number of Band Dequivalent properties in a local authority’s area. To calculate this, each domestic property expressed in terms of its Band D equivalence, and they are then added up. For example: a Band A property with no discount is 2/3 of a Band D-equivalent because 2/3 of the Band D amount of council tax is payable for it; and a Band D property subject to a 25% single adult discount is 75% of a Band D equivalent because 75% of the full Band D amount of council tax is payable for it.
2. The total taxbase across England in October 2005 was 17.9 million. The taxbase is used when calculating the ‘Relative Resource Amount’, as part of the four-block model for calculating Formula Grant allocations. Background 3. Multi-year local government finance Settlements are being introduced to bring increased stability and predictability to local government funding. The first announcement, in Autumn 2005, covered only two financial years: 2006/07 and 2007/08, with the first full three-year one being expected to cover 2008/09 to 2010/11, the years of the current Comprehensive Spending Review. 4. As well as moving from a one-year Settlement to one covering two years at a time, there has been a move from a historic snapshot of the taxbase taken before the start of the financial year to a figure projected to a point approximately in the middle of the financial year. This is consistent with the move from population estimates to population projections within the rest of the Formula Grant calculations. Possible methodologies 5. Several ways of rolling the taxbase figures forward - both in terms of the data used, and how they would be rolled forward - were considered, in consultation with local government and housing colleagues. 6.
1
Potential data sources considered included:
This is not the same as the taxbase for tax-setting purposes, which is set by each local authority in advance of setting its council tax, takes account of the assumed collection rate, and treats second home discounts differently. The taxbase for Formula Grant purposes is subsequently referred to in this paper simply as the ‘taxbase’.
Published household projections (based on ONS population projections) Recent trends in house completions Recent trends in relevant planning applications Housing growth predictions from annual monitoring by regional assemblies Recent trends in taxbases.
Projections vs forecasts 7. An issue underlying the choice of data is whether projecting past trends into the future is preferable in principle to making forecasts based on local intelligence on future housing developments. 8. A benefit of the latter approach is that it is potentially better at capturing information at a local level in cases when it is clear that historic trends are not continuing. Growth in Milton Keynes is, for example, levelling off, and so rolling forward past trends would be over-stating the future position. In contrast, some others areas, such as Corby, are now experiencing significant growth following recent decline – and using projections would project their decline into the future, even if that’s the opposite of what is happening on the ground. Looked at methodologically, an inevitable disadvantage of projecting past trends is that such an approach cannot, by definition, identify when a turning point in a trend is likely to be reached. 9. Practical disadvantages of using forecasts are that data are often patchy in practice, being not available on a consistent basis across all local authorities. In addition, they reflect only promises rather than actual buildings, and the timescale of future developments is uncertain, making it impossible to be sure of the impact of new developments upon the taxbase for any particular financial year. 10. On balance, the practical disadvantages of using forecasts were thought to outweigh their theoretical benefits, and so it was agreed that projections based on past trends should be produced. And it was agreed that it was the taxbase figures themselves that should be projected forwards, because it was assumed that the alternative data sources (housing completions or planning applications) would be worse than the taxbase itself for producing projections of the taxbase. 11. As stated above, assuming that past trends will continue into the future won’t reflect reality for some local authorities. But the resulting differences between projections and actuals are usually cancelled out to some extent in the Formula Grant calculations. For example, all else being equal, an authority whose future taxbase is being overstated is likely to be given an understated amount of Formula Grant. But all else isn’t equal in practice. This is partly because the population projections used in the relative needs block are also likely to be overstated, leading to a higher Formula Grant allocation than otherwise, and partly because any adverse effects on Formula Grant allocations are likely to be minimised or eliminated altogether as a result of the operation of the floor damping block.
Choosing the methodology to use 12. Having decided to roll forward each local authority’s taxbase in line with recent taxbase trends, a range of possible methodologies was considered, some more sophisticated than others: Fitting linear regression lines for each local authority; Use of exponential smoothing; Use of the Henderson curve moving average, with weights following bell-shaped curves; and Simply applying the annual average increases for each local authority (with different options covering varying numbers of years, and varying weights for different years).
13. Whilst a case could be made for any of these approaches, at least some of them had drawbacks, such as the need for the linear regression lines to be based on figures based on many years of data, going a long way back in time, if the results were to be robust. 14. In addition, it was clear from discussions with local government representatives in the Settlement Working Group that a relatively simple method was likely to be preferable for it to be able to be explained to and understood by a wide range of people, including elected councillors and members of the public. 15. To compare the performance of the various variants, each one was applied to taxbase figures reported on CTB1 returns as at Autumn in various previous years to see how good the various methodologies were at predicting Autumn 2003 and Autumn 2004 figures (referred to in the chart as 2004/05 and 2005/06 respectively, being the financial years’ Settlements for which the figures were used). 16. A straightforward method of comparing the results from each variant was adopted: for each local authority, the absolute difference between its projected figure and its reported figure was calculated, and they were summed to give an England total.
National sum of absolute differences between reported and projected 2004/05 and 2005/06 figures for all options considered
Sum of absolute differences (thousand)
300 250 200 150 100 50 0 One Year Average Two Year Three Year Average Two Year Weighted Average Three Year Weighted Average Three Year Weighted Percentage Four Year Weighted Percentage Linear Regression
Projection Methodologies
2004/05
2005/06
17. The graph above compares the results for each variant, where the weighted averages were simply judgemental and rather crude weights (e.g. giving the later year twice the weight of the earlier year in a two-year average). 18. The preferred option, taking account of the relative sums of absolute differences and the need for simplicity, was the two-year average: i.e. applying for each authority the average annual percentage increase in its taxbase, averaged over two years. 19. The geometric, rather than the arithmetic, mean was used to calculate each authority’s average increase over the two years (i.e. the annual % increases were multiplied together, and the square root of their product then taken). The differences between using geometric and arithmetic means were compared, and found to be small, but the former was regarded as more suitable for this application. 20. In addition, the taxbase figures were adjusted to take account of one-off increases in the taxbase due to reductions in the council tax discount for long-term empty homes made using power provided by the Local Government Act 2003. This was done by identifying the extent to which each authority’s Autumn 2004 taxbase had been inflated as a result, and then deducting that amount, to put each year’s figures on a like-for-like basis. 21. The graph below shows how the pattern of differences between projections and reported figures varies across the Government Office Regions for the period studied. The biggest differences are in London (in particular) and the South.
Sum of absolute differences between reported and projected figures for two year geometric average method
35
Absolute difference between reported and projected taxbase (thousands)
30
25
20
15
10
5
0 Yorkshire & Humber West Midlands South West South East North West North East London East Midlands East
Region
2004/05
2005/06
22. There is no clear pattern to this, however. To fully understand what is happening, it is necessary to get beneath the absolute differences, and look at the extent to which they were due to the projections under-predicting taxbases, or over-predicting them. In London, for example, almost 25,000 of the total absolute difference of 31,500 was for boroughs where the projections had over-predicted the reported figures, with the other 6,500 therefore being from boroughs where the taxbase had been under-predicted.
23. The map below compares actual (as reported) and projected taxbase figures for 2005/06 (i.e. Autumn 2004). The fact that changes in each size category tend to be scattered across England, rather than clustered in particular areas, provides some reassurance that the methodology used did not involve a systematic bias – although it is still possible that mapping at a different geographical level would identify such a bias.
24. There were nevertheless large differences between projected and reported figures for some individual authorities, both when projecting forward one year to 2004/05…
Percentage difference between reported and projected 2004/05 figures for all local authorities for two year geometric average method
3
2
Percentage Difference
1
0
-1
-2
-3
-4
-5
-6
Local Authorities
…and in projecting forward two years to 2005/06…
Percentage difference between reported and projected 2005/06 figures for all local authorities for two year geometric average method
4 3 2
Percentage Difference
1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10
Local Authorities
As might be expected, wider variation was seen over the two-year period to 2005/06, rather than the single year to 2004/05, because there was twice as much time for the projections to stray from reality.
25. As stated above, it is clear that the methodology used fails to be accurate at predicting figures for a few local authorities. But the floor damping block means that such authorities are protected from any adverse changes in the amount of Formula Grant received that might otherwise have occurred as a result. 26. To summarise the results for individual local authorities:
For 2004/05 (i.e. projecting over one year): 44% of authorities had projections within 0.25% of their reported taxbase figure; 77% of authorities had projections within 0.5% of their reported taxbase figure; and 95% of authorities had projections within 1% of their reported taxbase figure. For 2005/06 (i.e. projecting over two years): 23% of authorities had projections within 0.25% of their reported taxbase figure; 45% of authorities had projections within 0.5% of their reported taxbase figure; and 75% of authorities had projections within 1% of their reported taxbase figure. Changes in reported taxbases
Average annual percentage increase between 2003/04 and 2005/06 (geometric mean)
Percentage increase
1 0.8 0.6 0.4 0.2 0 Yorkshire & West Humber Midlands South West South East North West North East London East Midlands East England
Region
27. The above chart moves away from comparing projections and reported taxbase figures and looks at how taxbases changed in practice, for each GOR, during the two-year period studied. The national annual average increase was 0.6%, with figures for individual GORs ranging from 0.5% in the North East to 0.9% in the East Midlands. 28. As might be expected, wider variations can be seen at a local authority level, than at a regional level. Annual average changes in reported figures varied from a decrease of just over 0.9% for Barking and Dagenham to increases of over 6% for the City of London and over 2% for Tower Hamlets and Fenland.
•
Greatest average annual taxbase increases – City of London 6.5% – Tower Hamlets 2.4% – Fenland 2.1% Greatest average annual taxbase decreases – Barking & Dagenham 0.9% – Hammersmith & Fulham 0.8% – Ealing 0.5%
•
Comparison with Household Projections 29. Some work has been done recently to compare the taxbase projections with the recently published DCLG (formerly ODPM) household projections (which, in turn, are based on ONS population projections). Comparing with the DCLG household projections: Nationally, the taxbase and number of households are both projected to increase by an average of 0.9% per year between 2005 and 2007. Similar patterns of increase can also be seen at Government Office Region level. But the projected increases in numbers of households tend to be slightly higher than increases in taxbases in London, the South and the East, and slightly lower in parts of the north
30. Overall, the similarities are reassuring. The small differences that do exist for some regions might simply reflect the fact that Band D-equivalent properties and households are different things, and so shouldn’t be expected to change in exactly the same way. This will need to be considered as further data become available. Looking Ahead 31. The first test of the accuracy of the projections will take place later this year, when it should be possible to compare the projections with figures for October 2006, as reported on CTB1 returns. 32. It will then be necessary to decide whether it is right to carry on using the existing methodology, or whether it needs refining in any way. In particular, it will be interesting to see whether the designation of DCLG growth areas in parts of the country is weakening the link between past trends - on which the projections are based - and the future changes that they are trying to predict. 33. On a related point, it may be helpful to monitor any developments in the availability of consistent information for each local authority on forecasts of future housing developments based on local developments, in case they are thought more likely to produce accurate figures for future years than the current approach of rolling forward past trends. 34. Depending on priorities, other work might include using reported CTB1 figures to examine the extent to which taxbase growth has varied across bands in recent years (e.g. are numbers of properties in the higher bands tending to increase by a greater percentage than those in the lower bands?). Any such findings could then be taken into account in the methodology. 35. Unless it becomes necessary to change the methodology, it is expected that the same approach will be used in due course to produce projections for October 2008,
October 2009 and October 2010, to cover the three years of the Comprehensive Spending Review (2008/09, 2009/10 and 2010/11). Andrew Presland Local Government Finance Analysis Team, DCLG June 2006