There has been speculation in the press that the price of houses in Birmingham varies according to the proximity to certain local schools. This project will devise methods to represent the sale price of houses as given the land registry with other data such as asking prices quoted by estate agents and data relating to property sizes. The methods will attempt to quantify the different factors responsible for arriving at the price of a house, such as floor area, plot area, vintage, proximity to certain features such as local schools, shops, railway stations etc. The project will require skills in spatial statistics and automatic interrogation of web pages on the internet.
In recent years house prices have risen dramatically. I will look at the root cause of the abrupt change in the housing market and to see if there are any links to the location i.e. schools and town centres and house prices. I will carry out spatial interpolations on the house pricing data I obtain to try and justify my findings.
The following surface was created in a demonstration version of surfer 8 to show the variability in prices for different location in Birmingham for detached houses. The object of my project is to find what caused the peaks in the image.
House prices have risen at a higher rate than general prices as measured by the retail price index (RPI). Construction prices have not risen in line with the external factor of purchasing land. The cost of building a property represents a smaller figure to the actual cost of the property. House prices have risen at a faster rate than construction prices.
Figure 1 National average new house prices and construction
tender prices, and the retail price index 1950–1997.
DETR (1999), Davis Bel. eld & Everest
(1975), BCIS (1999) and the Office of National Statistics.
The graph shows all three underlying factors of house prices being in line with each other until 1970. From 1970 we see that the dwelling prices begin to climb on an upward spiral. Between the years of 1990 to 1995 construction prices in fact are lower than the RPI, showing the reduced cost of materials and increased productivity of the work force [Meikle].
The different types of properties has not been considered, which may undermine the correctness of the trend shown in the graph.
Figure 2 Average earnings and new house prices for
1970–1997. The sources were DETR (1999) and the Office
of National Statistics
Average earnings were in line with the cost of new dwellings until 1985. After this point there is a significant change between the two. The cost of new dwellings could have increased due to the change in the size of the properties and the architectural design changes implemented during this period. For people on less than average earnings will find increasing unaffordable [Meikle].
The British housing market is subject to risk through speculative behaviour, causing a high level of uncertainty and price inflation. Under the French system house prices are more controlled reducing speculative behaviour. The Swedish system has taken the route of local authorities buying land reserves to reduce price inflation but this is difficult to sustain because of the pressure applied by the open market [Barlow].
The British planning system has been criticised for its uncertainty and has been increasing politicised. People do not have an automatic right to building. Under a zoning system we would not have this problem. Local authorities use their discretion to operate control within their constituencies. The lack of fiscal control has shown an inflationary land market. Land owners are reluctant to reduce prices because of their high expectations
The house building industry in France is stimulated through the finance system allowing a range of different loans available on low interest rates. Conditions imposed on the loans limit dwelling size and house prices. From the mid 1970s to 1984 people were given low interest loans to purchase land for future developments. Land that is not owned by the communes is licensed to control the quality and price issues [Barlow]. Taxation of land holding and lower rates of inflation give justified cause for lower expectations to land owners and developer compared to the British system.
The Swedish system is subject to a high degree of state control, providing state house building loans (SHL) with low interest rates. Almost all dwellings built by private developer’s use SHL’s to build. Under the system land must be used from land banks and construction costs are controlled. Over 70% of new housing is built on commune land [Barlow]. Supply and demand is more correlated, local authorities can release land from their land banks to cope with the influxes in higher demand if necessary. However it is not immune from inflation pressure. Communes have found it progressively difficult to replenish their land bank reserves
Real world relationships can be overly simplified, causing incorrect theories to be supplied. In France details of the square footage of properties is available. Unfortunately I do not have data concerning square footage for the area of west midlands. It would have been interesting to see the different comparisons. It may have shown us a better in sight into the relationship between the cost of a property and its size.
Research has shown UK house prices ripple from the south east of England to other regions. The effect is felt gradually in all regions causing prices to move together in the long term after a quartile lag.
The Dickey Fuller test considers whether convergence exists between different regions on house price ratios
Table 1. Augmented Dickey–Fuller unit root tests of
convergence in regional: national house price ratios
Region ADF test statistic
Yorkshire and Humberside _0.48
North West _2.42
East Midlands _1.90
West Midlands _1.65
East Anglia _2.30
Outer South East _2.50
Outer Metropolitan _2.13
South West _1.50
5 per cent critical value _2.89
10 per cent critical value _2.58
Looking at the values obtained from the augmented Dickey Fuller unit root test we cannot reject the hypothesis for even 10% critical value. The convergence determined by the ripple effect does not seem to exist using the test. However the unit roots test can suffer from low power to reject the unit root hypothesis [Cook].
External factors like fuel prices could affect the house prices. The effect was apparent in the early 70’s, where people were using oil heated and gas heated houses but no real correlation was made between two [Halvorsen & Pollakowski].
Factors determining house prices
The supply of houses has a direct impact on prices. If there was a plentiful supply of existing and new houses being built we would possibly see a reduction in prices. Having a reduced supply of properties with high demand provides a catalyst to the housing market to raise its prices.
Confidence is also a crucial factor, along with finance. If consumers are willing to take the risk of purchasing a house they will most likely require a mortgage. Some people are not comfortable taking out large mortgages because of the variability in interest rates set by the bank of England. If the bubble surrounding the housing market was to burst, the values of houses will follow.
Borrower’s income has increased proportionally allowing people to purchase properties. However it has made it difficult for first time buyers to enter the market currently because they do not have any capital behind them.
Inflation is up to 0.7% currently. The more it rises the less value for money we get. It is important to maintain, else it will affect the cost of house prices.
Other Factors effecting house prices
You would probably think high crime rates will cause a reduction in property values over an area. Some theorists have contradicted this notion by saying the cost of crime has virtually no impact on house prices overall, but homes are highly discounted in high crime areas. However I have not found a great deal of information to support this notion.
Unemployment can affect house prices. Over the last 15 years we have seen the decline of the manufacturing industry in Britain. A local issue of mass job losses at Rover MG and its suppliers may adversely affect house prices within the region of Birmingham but only time will tell how a bigger impact it has on the area.
Houses placed in flood zones have always been connected with low house prices and very high insurance premiums. All the insurance companies will be using a GIS package of some sort to determine, which properties are affected based on their postcode. The can be issues with the accuracy of using postcodes because some houses in a flood zone may not actually be at the same high risk compared to others within thee postcode.
Government policy can affect the way council tax banding is carried out. For example the introduction of a local income based taxed would replace the council tax if the Liberal Democrats were to come into government. The changes in the stamp duty threshold have increased from £60.000 to £120.000. The benefit will be very little because first time buyers will most likely exceed the limit. The conservatives want to increase the figure to £250.000
Lower interest rates have played a big part in making purchasing properties easier. A lot of borrowers will find it difficult if interest rates were to increase by a couple of percent. At present the interest rate is at 4.75%. Ten years ago it was over 7%.
Council Tax Banding
Council Tax replaced the Community Charge (Poll Tax) on 1 April 1993 as the way households contribute to the cost of local authority services. The amount of the tax is set by local councils and charged according to valuation bands
The UK Government has a web site allowing UK residents to view a list of all council tax bandings in England and Wales. The site – located at http://www.voa.gov.uk – has been introduced by the Valuation Office Agency.
VOA is currently preparing for a revaluation of all 22 million domestic properties in England. Property values will be reviewed to bring them in line with changes to the property market. A revaluation of all domestic properties in Wales has been taken place, which took effect on the 1 April 2005. The Revaluation for England will take effect on 1 April 2007.
Council tax banding has a direct connection on the value of the property. The banding is considering the physical state of the property and its locality. Every dwelling has to be placed in one of eight bands for council tax purposes.
The bands for England and Wales are as follows:
Band A … up to £40,000
Band B … £40,001 to £52,000
Band C … £52,001 to £68,000
Band D … £68,001 to £88,000
Band E … £88,001 to £120,000
Band F … £120,001 to £160,000
Band G … £160,001 to £320,000
Band H … £320,001 and above
Note: The bands are set to the price they could have been sold for on the open market on 1 April 1991
New bands with effect from 1 April 2005
Band A … up to £30,000
Band B … £30,001 to £39,000
Band C … £39,001 to £51,000
Band D … £51,001 to £66,000
Band E … £66,001 to £90,000
Band F … £90,001 to £120,000
Band G … £120,001 to £240,000
Band H … £240,001 and above
Band A… up to £44,000
Band B… £44,001 up to £65,000
Band C… £65,001 up to £91,000
Band D… £91,001 up to £123,000
Band E… £123,001 up to £162,000
Band F… £162,001 up to £223,000
Band G… £223,001 up to £324,000
Band H… £324,001 up to £424,000
Band I… £424,001 and above
Source: Valuation Office Agency (VOA)
Assumptions made to Assess Council Tax Banding
The assessor has made a number of assumptions to determine the banding applied to a property:
The property was repairable, not considering its actual state.
The size and layout of the property was the same when the valuation had taken place and any mortgages were paid off.
Any damages to the property were repairable and the purchaser would have to pay to keep it in condition.
The property was constrained to be a private household.
The value of the property was only based on the development considered at that time.
Sources of Information
The Nationwide website provides details of historical house process since 1973 in quarterly periods for different regions in the UK, including the West Midlands. The data provided would have been more useful to me if it had been broken into postcode sectors, similar to what I have found from the land registry. Similar data sets can be found using the Halifax website.
Census data was consulted using the 2001 Aggregate Statistics Datasets found on Casweb. The data set combined two tables, KS016 Household spaces and accommodation type, which had the different types of properties with KS019 Rooms, amenities, central heating and lowest floor level for the average number of rooms per household in an OA format. A similar table called the UV057 number of rooms can be located from the univariate tables. To find the average number of rooms you have to calculate it yourself by simply multiplying the number of rooms by itself for each one and then you sum the figures together and divide by the total. The only problem is when you try to calculate the rooms, which are classified as 8 and over, you have to make some kind of assumption on the number of rooms. I will be using the first data set because it has a higher accuracy than the second.
I have used the Land Registry to retrieve information house prices. The average price of different properties was provided in terms of detached, semi-detached, terraced, flats and overall, as well as sales in quarterly periods from October – December 1998 to October – December 2004, restricting the data study to a period of 6 years. The sample size used is of decent range, so I felt there was no need to extend it any further.
Code point data concerning postcode unit point and boundary data was downloaded from Digimap in csv format. This provided details of delivery points and eastings and northings for each post code in the selected areas. There was a clash between the house price data I had obtained on postcode sectors and the code point data.
I found census boundaries of English output areas of the west midlands from the UK Borders website. The data came in two different file sizes of 414mb and 49mb.I downloaded the second more generalised version, which was perfectly adequate for the accuracy I was looking for in my project.
When trying to locate the easting and northings for the postcodes of schools. I had difficulties in finding some of the postcodes x and y position, so I used Multimap when no data was available through a postcode query in Digimap.
While browsing through Multimap I came to a link directing me to yahoo finance. Here you can find out the house price variations over the last five years. The interesting part was that it also predicts house prices five years from now. No assumptions about how it predicts future house prices were given.
Source: Yahoo finance (five year prediction and history)
I had to cut and paste all the house price data I could find from the land registry into several excel work sheets for Birmingham, Coventry, Dudley, Walsall, Wolverhampton and Worcester. Only postcode sectors were considered. The process was very repetitive because the data spanned over a lot of pages, which had to be selected in turn. Once I had all the house prices for the area of west midlands I had to import all the excel worksheets into a database in access as tables. I used the same column headings as before and did not assign a primary key.
I now had to find the weighted centroids of the postcode sectors. I multiplied the easting and northing in the code point worksheet. I had to tally the number of delivery points and then simply divided it by the eastings and northings to find the weighted average of the centroid. A complex query had to be performed to link the two tables together via query. The problem I had was the number of spaces in the postcode structure varied between areas. This would have to be taken care of to get the two tables to join effectively.
The following were developed in the expression builder in access:
Expr2: IIf(Mid([Expr1],3,1)=” “,Left([Expr1],2)+” “+Right([Expr1],1),IIf(Mid([Expr1],4,1)=” “,Left([Expr1],3)+” “+Right([Expr1],1),[Expr1]))
Once the join was completed I had to link all the house price data I had for the areas in quarterly period. Expression 2 above was linked in multiple queries with the region/area field. I now had the house price data for different types of houses linked with their weighted centroids.
All the queries I performed were exported to a folder in dbase 3 format. I created a feature class from an x and y table in arc catalogue. For each dbf file I had created I had to keep repeating the process of selecting the weighted eastings and northings and setting the z value to the average price for detached, semi’s etc. By doing the following operation I created shape files giving me the location and average prices for houses.
Note: At this point the data for different areas is split up into individual areas by default. (Date: Birmingham April – June 2000).
Here you can see all the weighted centroids within the area of Birmingham. The cursor inquiry shows the values of average house prices and sales within the area. You have to appreciate the time and effort required to create almost 700 shape files using the create feature class function.
After this stage I made a mistake of thinking that I need to create rasters for all the shape files I created. I got a bit ahead of myself and created over 600 raster files. I new I would need to convert data set to raster at some point to use in the raster calculator. I never thought about the fact that when I carry out my spatial interpolations I will automatically create rasters using the shape files I created early.
I had used a cell size of 50 meters for all the rasters and I used a British National grid projection with its default parameters:
(Date: Birmingham April – June 2000 for semi-detached properties)
When creating the rasters I forgot set the projection. I looked in arc toolbox for a function to perform a batch conversion but it would only allow me to carry out projection setting for one at a time. I was left a bit confused to how to complete the task, so I posted a question in one of the esri forums. I was told that I could either do this in the raster calculator, where I would have to define all the output files but this would have been just as slow. The other option was to download a visual basic script and add it as a customised button to arc map. The script allowed me to batch convert the rasters to the projection I required.
I had to merge all the six areas I had within the west midlands for quarterly periods. After a lot of time thinking how to do this I missed the obvious solution of simply perform a merge using the geo processing wizard. I had some trouble doing this in arc view, so I reverted back to arc map to get it to work.
(Date: West Midlands, April – June 2001 for semi-detached properties)
Once I had merged the area together I went on to do my spatial interpolations of the average house prices. Before I could carry out any of my surface interpolations I had to remove all the zero values present in my data sets. The zero values would have severely hindered the results I would get if this problem was not emitted. I should have taken care of this problem when I was working in access, it would have been a lot simpler and there would have been little room for error. The solution was not straight forward until I used the select by attributes function in arc map e.g. “S_AV_PRICE” > 0. This removed the issue of zero values with ease but I had to do this several times for the different types of property averages I needed.
There are three different types of spatial interpolations I can carry out in arc map
Inverse distance weighting.(1/d) Works on the principles of data close to one another is similar. Data further away is said to be dissimilar. The search radius considers the number of points to consider around particular points.
First of all I used the inverse distance weighting function (IDW), with a search radius of 12 and a constant cell size of 200 metres.
Potential Hot Spot
(West Midlands average house price, April – June 2001 for semi-detached properties)
As you can see the surface is fairly smooth and not disjointed as you may have expected. For this particular period the average house price for semi-detached properties is quiet low through out. Near the bottom I was a little surprised to see higher values in the area Coventry. I would have expected the majority of the area to be in yellow for this period (£118.000 – £163.000) but the majority of the surface has revealed lower figures, which look acceptable.
There is a particular area of yielding a high average price £375650, within Birmingham’s B15 postcode (Edgbaston). Using my local knowledge of the area this is not surprising and seems correct; however there were 5 sales in the postcode sector.
The spline interpolation performed on the same area is not as smooth as the IDW. We want to find the smoothest possible surface. The spline has produced results I was expecting from my point of view as I mentioned above.
Fluctuation not present in IDW.
(Spline produced using default parameters)
The kriging interpolation has produced a more jagged edge around the variations within the image. The image has not got any smoother. The higher average price in Coventry has multiplied by three fold compared to the IDW interpolation.
Fluctuation in Coventry exacerbated compared to IDW
(Kriging produced using default parameters)
I have looked through the different types of interpolations I can do. Without a doubt the IDW provides the smoothest surface. I will use the IDW to produce the rest of my spatial interpolations for the different types of properties. Above I have looked at the average property price for semi-detached houses. There was no particular reason why I chose the particular quarterly period or property type. It served merely as a visual aid to show the variations. It may be a good idea to look at the overall average prices rather than for any particular type because it should go some way to removing any spikes created by an individual high priced property.
The spatial interpolation showing the average house price for detached properties is expected to have hot spots in varying areas and that’s what I have found.
West Midlands average house prices, October – December 1998 for detached properties)
In the following image 6 years on it shows signs of great similarity. There has only been a subtle change in some areas toward the east.
West Midlands average house price, October – December 2004 for detached properties)
The average house price for flats is low in some areas and high in a hotspot area of Birmingham and Coventry.
West Midlands average house prices, January – March 1999 for flats)
The hot spot has moved away from Coventry and only exists in Birmingham. The highest average price for a flat has more than doubled
West Midlands average house price, October – December 2004 for flats)
Again we have hot spots in the area of Coventry, probably because there are a lot of terraced properties in the area. Prices towards Wolverhampton are also less expensive
West Midlands average house prices, January– March 1999 for terraced houses)
The hot spots have increase proportionally in Birmingham and Coventry but the average price has not been affected greatly.
West Midlands average house prices, October – December 2004 for terraced houses)
The next step was to find the average number of habital rooms. Using the data obtained from the census on household size I can compute an average. I saved the census data set as a dbase 3 file and did the same thing as before by creating a feature class from an x and y table. I used the easting and nothings as the x and y parameters. The z parameter was given the average number of rooms, thus a shape file was created. I tried to produce a surface interpolation on these points but it kept freezing the computer up. This was caused by a single point, which was out in the middle of no where. The point had several thousand points overlaid itself all with a value of zero. I removed the obstacle by selecting the attributes with values based the fid of below 17457.
An IDW surface interpolation was later successfully performed to compute the average number of habital rooms with a cell size of 200, but a reduced number of 6 were used for the search radius because there were a lot of records in the file. The same as the average house price interpolations. The variations in the image are fairly evenly spread, except in the middle where the average number of rooms is between 3 and 4. This is what I would expect for this area, which includes Birmingham, Dudley, Walsall, Wolverhampton and Worcester.
Reduction in average number of rooms.
Average number of habital rooms in the west midlands.
I found the average number of habital rooms image was not at the same extent because I never considered the whole area of west midlands in its entirety. To resolve this problem I tied to just change the extent in the image properties but this was not correct because it was still considering a larger area, causing lower values to be attained. I recreated the average number of rooms by re-running the surface interpolation, where I changed the options in spatial analyst. I matched to the same extent I created for average house prices interpolations.
My selected areas of the west midlands for average house prices.
Census areas of average rooms for west midlands
The area of Birmingham and parts of Coventry seem to have the highest price per habital room between £29.900 and £46.000, although there is an area in Dudley.
Overall price per habital room for October – December 1998
The location of high priced habital rooms varies through different locations between the years but remains fairly constant except for a couple of quarterly periods in Birmingham and Coventry.
It would have been a good idea to create an animation in arc map to be able to visualise the change over the six year period.
The highest price per habital room in this quarter is between £90,000 and £123,000. This is a considerable increase over the years. Compared to image created for the fourth period in 1998, there is £77.000 difference.
Overall price per habital room for October – December 2004.
This particular quarter has the most significant change of all the interpolations. Again Birmingham and Coventry are at the centre of high prices. Other quarterly periods are not as dramatic, this seems to be an exception and possibly at the height of the house price boom, which has slowed down recent months.
Overall price per habital room for July – September 2002.
The next step was to try and find all schools within the west midlands. I found details of primary, secondary and post 16 education from the council website for each area within the west midlands. I created a database with names of the schools and I ran a postcode query search to find their positions. Eventually I gathered a good data set of over 1200 schools in total but I forgot to find any performance tables at this stage. Using department for educations website I located the performance tables for year 2004. I had to create another table to extend the database. The only common field between the schools location and their performance was there name. I understood this was going to cause problems because of the way they have been named. I never took great care with the naming. A query with a relational join between the two tables would only link the fields, which are identical. Any confusion will result in fields not correctly being linked. The problem could have been sorted out fairly easily by simply locating the postcode for the performance tables and then linking the two.
The linking problem caused by linking the two name fields of the schools caused a vast reduction in my data, perhaps by a third, which was very disappointing. Any way I created three different feature classes for each type of school using their easting and northing positions.
In the image there are little no primary schools in high priced areas. The school visible here is in Hockley Heath, Birmingham and has an average point score of 30.2. The England average for primary schools is 27.5. I was surprised to find a high average score for the school. There seems to be some evidence showing exceptions to the idea of good performing schools are not necessarily located in areas with high house prices. I would have liked to have seen the school placed directly at the centre of the largest variation.
Overall average house price for all types of properties for October – December 2004.
Performance tables for different LEA’s were obtained from the department of education and skills (Dfes) for 2004. The average point score was taken for each primary and secondary school (Key Stage 3). The average was calculated by the percentage of pupils obtaining different levels in maths, science and English. However the performance for secondary schools was split into two methods, the first stated above and the second took the students achieving 5 or more grades A-C.
The Post-16 education, which includes colleges and schools, has the value of the average point score per examination entry.
Average Point Score
Secondary Schools (Key Stage 3)
Average Point Score
Secondary Schools (GCSE)
(5 or more grades A*-C)
Average point score per examination entry
Larger data set originally collected for all the schools
The images blow for the remaining secondary and post 16 education show there is no link between schools and high property houses. The majority of the schools are located in cheaper house priced areas. The data set has limited my analysis with the schools but I think it would not have made a significant difference to my findings. We would have still seen the same trend through out the areas.
The majority of schools are located in less expensive areas.
Overall average house price for all types of properties for October – December 2004 with secondary school overlaid.
Overall average house price for all types of properties for October – December 2004 with secondary school overlaid.
Next I took a look at whether proximity to centres affected house prices. I gathered my data set for 8 different supermarket chains using the store finder located on their websites:
I was a little concerned with the quality of the store locator search engines the supermarkets had to offer. I’m confident I managed to get all the above supermarket locations, apart from one or two. I again located there position using a postcode query and inputted them into a database, from which I created a feature class for all the supermarkets I found. I had over 200 records to form the data set.
I had a little teething problem in drawing buffers around the locations, so I carried out the operation in Arc view 3.3 using the buffer wizard. The problem was caused by mapping units, which were easily solved and then imported back into arc map. I have drawn 1 kilometre buffers around all the supermarkets because people would like to live close to centres.
Supermarket locations with 1km buffers
Nearly all of the supermarkets are located in lower priced areas (green). You would probably expect to find more supermarkets in areas with high property price. The image does not show support the argument. There are many restrictions applied to the data set. People do not only shop at big retail chains. No independent privately owned supermarkets are accounted for. The latter may show a different trend.
Overall average house price for all types of properties for October – December 2004 with supermarket location buffers overlaid.
I had a look at some strategy data from the ordinance survey for the south of England to a 1:250.000 scale but I could not find anything useful in the way of features i.e. railway proximity etc. The labelling was very poor and sparse. The only thing it did show clearly was rivers.
Strategy data for the south of Britain 2004
I found some interesting data from Yahoo property finance predictor. It would have been good if gave values for postcode sectors and not just individual postcodes. The particular postcode I have run a search on was for where I live. Using the local knowledge I have of my area this seems about right.
Price history for All (Average) property in B11 3BN
Period (each quarter for the last 5 years)
I also ran a search to predict the prices for the next five years. The only draw back is that has not told me about any assumptions it’s made or how the model was created.
Price prediction for All (Average) property in B11 3BN (could increase by 4.90 % in the next 5 years)
The percentage change in average prices in each quarter
The compound or cumulative change in average prices over the whole 5-year period
My findings have shown there is little no link between house prices and good performing schools. I have seen some exceptions in a few areas but they were really scarce. I managed to get some descent surfaces with very little noise with the Inverse distance weighting interpolations. Noise in the images would have caused jagged edge and jumps but my surfaces have fairly smooth variations.
The average house prices have shown a higher price for houses in the south east and towards the west. You would expect to see this trend in Britain, probably because during the Victorian period pollution from the factories spread towards the north eastern areas where poorer people lived.
The accuracy of data could have affected my overall results. The data obtained on for the average house prices from the land registry is of a good reliable accuracy. The performance tables for schools were retrieved from the department of education, which is considered to be a reliable source. The average point score used for each school can come under some criticism because it is not easy to access performance. For secondary schools I have used the percentage of students who pass with 5 or more GCSE’s graded A to C. I found a few postcodes for schools and supermarkets to not have valid royal mail postcodes, which I found a bit strange.
I was expecting to find some sort of connection with the location of supermarkets and high property prices but nothing concrete emerged. The data sets I compiled for each analysis could have contributed to the variations I found. If I had better strategy data I could have tried to look for trends and other factors affecting house prices.
As an extension to my project it would have been a good to produce a model to predict house prices in five years time. I have looked at the predictive model in yahoo finance to give me a general idea. I could have possibly compared the average prices I got to theirs to find any differences.
The supply and demand for houses has prompted concerns in the housing market. At the moment about twenty thousand new are being built per year. The number is very small compared to demand for houses. The continuing migration of people into Britain has caused the problem to become exacerbated.