The Scottish Government

« Previous | Contents | Next »

Listen

3. simulation results

As mentioned in the previous section, three scenarios were explored by the research team. The three scenarios share the same base macro-economic assumptions but differ in terms of the level of net annual additions to the housing stock. This section is organised as follows. Each scenario begins with a description of the assumed annual level of net additions in each sub-national area over the simulation period. The key outputs are summary charts describing real mean house prices (in 2006 prices), the mean house price to mean household income ratio (affordability measure), the lower quartile house price to lower quartile household income ratio (lower quartile affordability measure) and predicted levels of owner occupation. These charts are supported by tables in Annex A with each chart having a corresponding table. Household migration varies little between the three scenarios, primarily because each scenario is applied uniformly across Scotland, but is discussed briefly in section 3.4. Section 3.5 provides a comparison of the key outputs of the three scenarios. Finally, section 3.6 examines the sensitivity of the simulation outputs to macro-economic variables.

3.1 Scenario 1 (base case)

Scenario 1 assumes the continuation of net additions rates at the levels observed in the 5 years to 2006. These rates are assumed to continue throughout the simulation period (to 2041). The annual level of net additions based on this historic pattern is summarised in table 2.

Table 2 Annual net additions in scenario 1

Figure 3 Scenario 1, simulated mean house prices

Figure 4 Scenario 1, simulated mean affordability measure

Figure 5 Scenario 1, simulated lower quartile affordability measure

Figure 6 Scenario 1, simulated levels of owner occupation

Under scenario 1, mean house prices are projected to increase in real terms and in the long-run in all of the sub-regions. The simulation predicts periodic adjustment with falling real prices between 2012-21 and 2027-31, although the exact timing of these adjustments varies by sub-region. Real price rises are stronger in the Edinburgh city region, rising from approximately £157,000 at the beginning of the simulation period to approximately £280,000 by 2041. The mean real house price in Scotland overall is predicted to rise from around £156,000 to nearly £230,000 in real terms by 2041.

The simulation results suggest that Scotland overall, and the Glasgow, Edinburgh and Dundee city regions, will experience a higher rate of long-run real price growth than other parts of Scotland. In Ayrshire, Aberdeen, Borders/Dumfries & Galloway, Highlands, Moray and Islands and Stirling and Central Scotland, real prices are predicted to rise in the long-run, but at a lower rate overall. The results also suggest that the Edinburgh city region will experience a particularly pronounced long-run rise in real house prices compared with elsewhere in Scotland.

Household incomes are predicted in real terms by the simulation model, and are also predicted to rise during the simulation period, mitigating the effects of rising real house prices to some extent. In most sub-regions (excepting the Edinburgh city region), affordability is predicted to improve early in the simulation period. This is primarily because the housing market element of the simulation predicts periodic adjustment, with such a period predicted to occur broadly between 2010 and 2017. As noted earlier, the exact timing differs between sub-regions. The predicted impact is that affordability should improve between the beginning of the simulation period and around 2014/5, at which time real house price rises are predicted to exceed growth in real household income. In Scotland overall, the mean real house price to real household income ratio oscillates around approximately 7.00 throughout the simulation period. This pattern is reflected in most sub-regions, with the exception of the Aberdeen city region and the Edinburgh city region.

It is worth noting that the structure of the labour market module is a little different in the Aberdeen city region to the other 7 sub-regions. A key difference is the inclusion of net trade in goods and services in the Aberdeen city region equations for full-time earnings. This is an important predictor in the equation, but does not appear in any of the other sub-regional equations. The value of this variable is assumed fixed in the simulation model because predicting its behaviour is outwith the scope of the model. The other exception noted earlier is the Edinburgh city region. Here, the long-run pattern is similar to that predicted for other sub-regions, but rises overall during the simulation period.

The tenure choice module predicts an increase in the proportion of households choosing owner occupation above other tenure choices with the overall Scotland proportion rising from 76% to 88% of households by the end of the simulation period. Stirling and Central Scotland and the Glasgow city region are predicted to witness the highest growth in the rate of owner occupation, rising from 74% and 73% respectively to 88% and 87%.

The effect of rising house prices is particularly pronounced on lower household income groups. Dundee/Tayside, the Glasgow city region and Ayrshire are the least affordable at the beginning of the simulation period. By the end of the simulation period, these conditions are predicted to have exacerbated in the Edinburgh city region, improved slightly in Glasgow and improved noticeably in Ayrshire.

The choice of periods for comparison is clearly important because, as discussed earlier, the simulation predicts periodic house price adjustment. For example, if the period 2007/11 is compared with the second last block of periods in the simulation (2032/36) then different conclusions are reached. In Scotland overall, the lower quartile affordability measures deteriorates, rising from 5.71 to 5.91 and a similar picture emerges in the Edinburgh and Glasgow city regions. Affordability in most other areas is predicted to remain broadly in line with current levels. An exception is the Aberdeen city region, where affordability is predicted to improve, and Dundee and Tayside. The affordability measure in the latter is predicted to fall quickly from a current high of 7.15, before oscillating around 5.00 throughout the remainder of the simulation.

3.2 Scenario 2

Scenario 2 envisages a substantial, and sustained, increase in net additions during the simulation period. An annual level of new-build completions of 35,000 units is assumed to equate to approximately 30,645 annual net additions, based on the historic relationship between new-build completions, demolitions and net additions. The assumed levels of net additions sub-nationally are summarised in table 3.

Table 3 Annual net additions in scenario 2

Figure 7 Scenario 2, simulated mean house prices

Figure 8 Scenario 2, simulated mean affordability measure

Figure 9 Scenario 2, simulated lower quartile affordability measure

Figure 10 Scenario 2, simulated levels of owner occupation

The simulation results for scenario 2 suggest that the higher level of net additions throughout the simulation period would lead to an improvement in affordability in Scotland overall (with the mean ratio falling to 6.2 compared with 6.8 in scenario 1). The model predicts variable impact within Scotland with some sub-regions only modestly affected, such as the Aberdeen City Region, and Stirling and Central Scotland. The improvement in affordability in Ayrshire, Dundee and Tayside and the Glasgow city region is more pronounced with the mean house price to mean household income ratios predicted to fall from 6.4, 7.1 and 7.7 to 5.5, 6.1 and 6.6 respectively. The remaining three sub-regions - Borders, Dumfries and Galloway; Edinburgh City Region and Highlands, Moray and Islands are predicted to benefit from smaller improvements in their affordability ratio. For example, the end of simulation period ratio is 7.0 in the Edinburgh city region compared with 7.5 under scenario 1.

3.3 Scenario 3

Scenario 3 envisages a rise in net additions to 30,645 in the period 2017-2021 (equating to annual new-build supply of approximately 35,000 units). The level of net additions is then assumed to decline over a ten year period until it reaches the annual percentage figures witnessed in the 5 years to 2006. The numbers of annual net additions assumed under scenario 3 are summarised in table 4.

Table 4 Annual net additions in scenario 3

Figure 11 Scenario 3, simulated mean house prices

Figure 12 Scenario 3, simulated mean affordability measure

Figure 13 Scenario 3, simulated lower quartile affordability measure

Figure 14 Scenario 3, simulated levels of owner occupation

The results suggest that the third scenario of net additions has a very limited impact on affordability in comparison with the base case (scenario 1). The mean affordability ratio is predicted to fall from 6.8 to 6.6 in Scotland overall by the end of the simulation period and the lower quartile measure from 5.4 to 5.3. An examination of the results by sub-region reveals a similar pattern to that shown in the scenario 1 and 2 analyses. In the Glasgow city region, for example, the mean affordability ratio is predicted to fall from 7.7 (lower quartile 6.1) to 7.3 (lower quartile 5.8). Meanwhile, in the Aberdeen city region the affordability measures are actually slightly higher in scenario 3 compared with the base case and, in the Edinburgh city region, the ratio changes very little (from 7.5 to 7.4, mean ratio and from 6.0 to 5.9, lower quartile ratio).

3.4 Household migration

As noted above, simulated household migration varies very little between the three scenarios. This is primarily a function of the fact that the simulation scenarios are applied uniformly across Scotland. Clearly, a strong spatial concentration in any rise in net additions would potentially yield a different result.

In the base case (scenario 1), the Aberdeen city region and Stirling and Central Scotland are predicted to see net outward migration throughout the simulation period, although the number of households concerned is modest (5 - 8,000 in each case). Ayrshire and Highlands, Moray and Islands are notable net recipients of households in the simulation. A more mixed pattern emerges in Scotland's urban areas: the Glasgow city region is predicted to see net inward migration early in the simulation period, and outward migration later in the simulation. The Dundee city region is predicted to experience net outward migration, but at a smaller scale compared to Aberdeen city region and Stirling and Central Scotland.

Table 5 Simulated household migration under scenario 1

3.5 Comparison of the simulation results

This section compares the key outputs of the simulation model under the three scenarios. Table 6 summarises mean house price levels (in 2006 prices), tables 7 and 8 report the mean and lower quartile affordability measures and table 9 summarises the predicted rates of owner occupation. All tables show figures for the final period of the model simulation.

Table 6 Mean house prices under three simulation scenarios (2037-41, 2006 prices)

The results clearly suggest that the long-run responsiveness of house prices (and affordability) to levels of net additions is low, but that some improvements to affordability are achievable. In Scotland overall, scenario 2 predicts a mean house price of around £209,000 compared to around £230,000 under the base case, scenario 1. However, table 6 suggests that the impact of increasing the rate of net additions will vary within Scotland. For example, in Ayrshire, Dundee and Tayside and the Glasgow city region, the mean house price level is significantly lower in scenario 2 compared to the base case. Meanwhile, in the Aberdeen city region and Stirling and Central Scotland, the simulation model predicts only very marginally different price outcomes in scenarios 2 and 3 compared to the base case. Borders, Dumfries and Galloway, Highlands, Moray and Islands and the Edinburgh City region fall between these extremes with lower mean house prices under scenarios 2 and 3, but with a smaller proportionate difference compared with Ayrshire, Dundee and Tayside or the Glasgow city region.

Table 7 Mean affordability measures under three simulation scenarios (2037-41)

Table 8 Lower quartile affordability measures under three simulation scenarios (2037-41)

The impact of higher levels of net additions on affordability also varies within Scotland. A more significant improvement is predicted in Ayrshire, Dundee and Tayside and the Glasgow city region. In the Aberdeen city region and Stirling and Central Scotland, the affordability measures are largely unchanged in either scenario 2 or 3 compared with the base case. In Scotland overall, the differences between affordability measures in scenario 3 compared with the base case are trivial. This is suggestive that a sustained increase in the level of net additions is necessary in the long-run to generate any noticeable improvement in affordability.

Table 9 Predicted rates of owner occupation under three scenarios

The levels of owner occupation predicted by the model are almost identical in each of the scenarios. This is mainly because the choice of owner occupation is driven very strongly by demographic factors and by household income. While the cost of owner occupied housing does influence households' choices, the tenure choice modelling results suggest that a very significant change in house price levels would be necessary in order to produce a significantly different outcome.

3.6 Sensitivity of the simulation outputs to macro-economic variables

As noted earlier, the simulation model includes options for the user to alter macro-economic inputs such as RPI / CPI inflation, annual stock market returns and the nominal mortgage rate. In practice, the long-run sensitivity of the simulation model outcomes to most of these variables is very low (although there are, of course, short-run effects to shocks in such variables). The simulation model is designed to provide long-run predictions, hence only a very significant and sustained change to these macro-economic inputs would be likely to create a markedly different long-run outcome.

Each of the following tables summarises the impact of a change in one of the macro-economic, or other additional policy input, variables. Base case figures are shown in brackets for comparison. The sensitivity analysis is carried out on the simulation overall, but the outputs shown simply summarise the difference in outcome at the level of Scotland overall. Tables 10 and 11 examine the nominal mortgage interest rates, this being the most sensitive macro-economic input variable.

Table 10 Assuming 6.5% nominal mortgage rate (rather than 5.5%)

Note: P:Y (Scotland) refers to the ratio of mean house price to mean household income for Scotland as a whole while lq P:Y (Scotland) refers to the ratio of lower quartile house price to lower quartile income

Table 11 Assuming 4.5% nominal mortgage rate (rather than 5.5%)

The sensitivity analysis therefore suggests that a change in the nominal mortgage rate could have an impact of as much as 5% on mean house prices in the long run. A percentage point increase would lead to the affordability measure falling from 6.18 to 6.02 at the end of the simulation period or rising from 6.18 to 6.34 in the case of a percentage point decrease.

The simulation results are not sensitive to the inflation inputs. Changing the rate of CPI inflation makes a very small difference to the simulation results, while changes to the rate of RPI inflation make no discernible difference.

Table 12 Assuming 3.5% CPI inflation (2.5%)

Table 13 Assuming 4.3% RPI inflation (3.3%)

As noted in the earlier discussion, annual stock market returns are important in some of the sub-national housing market equations. The direction of this effect is the opposite of that found in the England model (see Meen et al, 2005). In the latter, stock market returns were positively signed, and this was interpreted as a general wealth effect. In the Scotland model, stock market returns are generally negatively signed, which is suggestive that the housing market is counter cyclical to the stock market. In other words, speculative behaviour in the housing market is more likely when other investment markets are performing poorly. The sensitivity analysis is summarised in table 14. The base case assumes that stock market returns are around 10% per annum initially, but rising to 15% throughout most of the simulation period. The sensitivity analysis assumes 5% per annum stock market returns. The results show house price levels around 2.5% higher than the base case, with a small corresponding change to the affordability measures.

Table 14 Assuming 5% p.a. stock returns (15%)

The sensitivity analysis on GVA growth, shown in table 15, shows that the simulation outcomes are not sensitive to this policy variable. This is not surprising since the labour market module estimation results show that change in the rate of GVA growth, rather than changing levels of GVA growth, is a partial determinant of household incomes.

Table 15 Assuming 4% GVA growth (3%)

Tables 16 and 17 summarise the sensitivity analysis results in respect of the taxation variables (property tax and the basic rate of income tax). A reduction in the rate of property tax, currently council tax, has a small inflationary effect on house prices as might be expected. The impact on the affordability measures is very small. Changes to the basic rate of income tax have a very minor influence on the simulation results.

Table 16 Assuming 0.5% property tax (1.21%)

Table 17 Assuming 25% basic income tax (22%)

Changing the proportion of working population assumed to have university degrees has a noticeable effect on the simulation results. The base case assumed that the 2006 level of 16.8% continues throughout the simulation period. However, over-riding this assumed level is an option in the policy input worksheet. Making a one-off change from 2012 onwards from 16.8% to 25% gives rise to higher mean house prices by around 1.5 to 2.0% throughout the simulation period. This has very minor impact on the affordability measures.

Table 18 Assuming 25% working population with degrees (16.8%)

« Previous | Contents | Next »