West London Alliance: Housing Stock Projections. Prepared for: Mr D McCulloch London Borough of Hillingdon for the West London Alliance.

Transcription

1 West London Alliance: Housing Stock Projections Prepared for: Mr D McCulloch London Borough of Hillingdon for the West London Alliance 15 March 2006 Client report number

2 1 Housing Stock Projections Prepared by Name Robert Flynn Position Principal Consultant Signature Approved on behalf of BRE Name Simon Nicol Position Director, Housing Centre Date Signature BRE Garston WD25 9XX T + 44 (0) F + 44 (0) E enquiries@bre.co.uk This report is made on behalf of BRE. By receiving the report and acting on it, the client - or any third party relying on it - accepts that no individual is personally liable in contract, tort or breach of statutory duty (including negligence). BRE Client report number

3 2 Housing Stock Projections Executive Summary This report provides estimates of local housing conditions within the authorities who are members of the West London Alliance. The estimates are provided at the level of the authority and ward using models developed by BRE which combine national data from the English House Condition Survey 2001 with local census data. The information is provided in tabular form at authority level, and the ward data are mapped and provided in detail in a separate spreadsheet. A further spreadsheet provides some of the census data used to develop each model and other background census information requested by the alliance. The models were originally developed on an all tenure basis. Information is supplied in this format but supplementary information has also been provided on the private sector stock. Information has also been provided on comparisons between the model outputs and survey data provided by individual authorities including Kensington and Chelsea from within the sub-region. The results have been generally positive indicating the models provide a useful guide to local housing conditions. BRE Client report number

4 3 Housing Stock Projections Contents Introduction 4 Description of the project 5 Project brief 5 The development of BRE Local Housing Stock Projections 5 Findings 7 Presentation and interpretation of results 7 Results from the BRE Housing Stock Models 7 Definition of vulnerable 9 The Decent Homes model and its components 9 Comparisons with local authority data 13 Key statistics by census wards 17 Summary and conclusions 38 Appendix A Modelled data: private sector wards Appendix B Predicting Housing Conditions Originally published Jan 2004 Environmental Health Journal BRE Client report number

5 4 Housing Stock Projections Introduction The main objective of this report is to provide estimates of local housing conditions at the level of the subregion, authority and statistical ward. The projections are based on models developed by BRE, which combine local and national data to produce estimates of conditions that would not otherwise be available at the local level. The projections have been made for the seven local authorities comprising the West London Alliance (WLA), i.e. Brent Ealing Harrow Hammersmith and Fulham Hillingdon Hounslow Kensington and Chelsea BRE Client report number

6 5 Housing Stock Projections Description of the project Project brief Prior to the development of the BRE Housing Stock Models, the main method of estimating the numbers and determining the location of private sector dwellings in poor condition has been to undertake a private sector house condition survey. Private sector house condition surveys are usually carried out on a sample basis and information is typically gathered on around 1000 dwellings. It is normal practice to report on subareas with a minimum sample of dwellings, as below this sampling error makes comparisons less reliable. This will mean that such surveys will only be able to report on three or four sub-areas, which is of little value for targeting local areas. The members of the WLA were aware of this, and were interested in an alternative approach that could provide information on key housing indicators at borough, and statistical ward level. The alternative approach was to use the BRE Housing Stock Models to provide projections of key indicators at these levels. A particular advantage of this approach is that it allows comparisons to be made at the sub-regional level, whereas the surveys undertaken by the individual authorities have taken place at different times using different companies, surveyors and methodologies and therefore provide a poor basis for comparison. While the above gives an outline of what the stock models provide, it does not explain how they work and where they can fit in with the gathering of local authority data on housing conditions. A brief explanation of how the stock models were developed, and the projections they provide, is outlined below. The development of BRE Local Housing Stock Projections The BRE local housing stock projections are based upon a series of models developed at the Housing Centre of BRE, initially using funding from the BRE Trust (formerly the Foundation for the Built Environment) and local authority partners. The models make use of data from the 2001 EHCS. They relate the condition of a dwelling (or some other housing measure included in the 2001 EHCS) to the characteristics of the local area in which it is situated. So if the characteristics of a particular small area (for example, a ward in a borough) are known, the models can be applied to provide estimates of the proportion of dwellings in different conditions. Similar, more limited models used in the past have been those for estimating unfitness rates from building age and type, together with the Shaw and CACI housing market classifications. The local area characteristics can come from any available source. This can be a local data base e.g. a housing needs survey, or a national one such as the census The latter is generally more useful and convenient, and the models have been developed using specific national datasets; particularly the census Application of the models requires specific local area data on a variety of demographic and socioeconomic factors. The primary sources used currently are Small Area Statistics from the census 2001 and from other commercially available databases. The smallest geographical unit which can be modelled is the 2001 Census Output area. BRE Client report number

7 6 Housing Stock Projections The dwelling condition factors modelled so far are: non-decent housing with its individual components inadequate thermal comfort, unfitness, modern facilities, disrepair; Non decent homes with vulnerable occupiers (which is the same definition as used for the new PSA target), dwellings with a SAP Rating less than 30 and Households in Fuel Poverty (although the latter two model outputs are not being supplied as part of this project). Ideally it would be desirable to use EHCS data directly at the local level. Unfortunately this is not possible, because the relatively small sample used for the survey (around 17,000 dwellings from the country as a whole in 2001) does not give sufficient coverage for each of the 400 or so districts, and certainly not for the individual wards and census output areas within districts. We are therefore left with no option but to use nationally based models. The penalty this brings is the possibility that small areas which are atypical in condition will not be well modelled. Unfortunately there is no way of predicting just where these areas will be. It is therefore important, wherever possible, to compare them to local data. The models are based on two complex statistical techniques; CHAID which is short for chi square automated interaction detection, and logistic regression. A brief summary of how CHAID is used is included in the article Predicting Housing Conditions reproduced in Appendix B. This article first appeared in The Environmental Health Journal in January 2004 and describes how the models were used to predict conditions in the London Boroughs of Redbridge and Richmond. Logistic Regression is a statistically more robust technique than CHAID and we have used the main independent variables suggested by the CHAID analysis to develop models which respond more smoothly to changes in independent variables. While the final outcomes are broadly similar models, those produced by logistic regression are generally considered to be more robust. BRE Client report number

8 7 Housing Stock Projections Findings Presentation and interpretation of results The data produced by the models (e.g. decent homes etc.) are presented in tabular form by authority and by ward. We currently report on statistical wards based on the 2001 census. These usually coincide with political ward boundaries but cannot be guaranteed to do so. Results are also produced in map form at ward level. Results from the BRE Housing Stock Models The tables describe the stock in terms of the following characteristics at ward and local authority level: Dwellings which would fail the Decent Homes Standard Dwellings which would fail the Decent Homes Standard due to: o o o o Unfitness Inadequate Thermal Comfort Disrepair Non-modern facilities and services Non decent homes with vulnerable occupiers Table 1 summarises these results for all dwellings in the housing stock in the sub-region. While the main reason this report was commissioned was for the purpose of improving the understanding of the private stock, the models that produce the information are all tenure models. Table 1 therefore summarises the results as they were originally output but expresses them as percentages. For all but one of the projections the percentages are reported as a percentage of all dwellings. The one projection where we do not report in this way is for vulnerable households in decent dwellings which we report as a percentage of all vulnerable households. This is not in itself a particularly useful measure; however, if the social rented stock is removed to leave only the private sector stock, then this becomes particularly useful as it relates directly to a government target. The strong demand from all of our clients to find a way of converting the all tenure to private sector projections has led us to commence with the development of a private sector set of models. However these will not be developed for some months and we have therefore arrived at an interim solution. This interim solution is to take the all dwelling projections for a variable e.g. non decent homes, at the level of the census output area and multiply these by the number of private sector dwellings present to arrive at an estimate of the number of dwellings which fail the standard for that variable i.e. for non decent homes we would end up with the number of private sector dwellings that are non decent. These totals can then be BRE Client report number

9 8 Housing Stock Projections summed to ward and borough level. In Table 2 we provide the totals but also express them as a percentage of all private sector dwellings (with one important exception). This information is therefore in a format which can easily be compared to private sector house condition survey results. The only exception is decent homes with vulnerable occupiers which for targeting reasons is expressed as a percentage of all private sector dwellings with vulnerable occupiers. In Table 3 we again provide the private sector totals for each projection, but we now express them as percentages of all dwellings. This provides an immediate indicator of where the main concentrations of housing stress in the private sector are to be found. It is important, however, to remember when interpreting this table and the selection of maps of the private sector which use the same method (i.e. all those with the suffix (b) ) that the percentage will reflect not only the severity of the housing problems in the area but also the extent of the private sector within it. In Hammersmith and Fulham, for example, 43% of dwellings fail the decent homes standard. However, as only 68% of the stock is private sector there are relatively few private sector non decent homes, accounting for only 29% of the stock in the borough as a whole. The information supplied on non decent homes with vulnerable occupiers is of particular importance to the private sector, as the 2002 Government Spending Review expanded the decent homes target to the private sector, with the aim of increasing the proportion of private housing in decent condition occupied by vulnerable groups. The Government's Decent Home Target Implementation Plan also sets out a trajectory for delivery that includes targets for specific years up to 2020, expressed as the proportion of vulnerable households in the private sector living in decent homes. The relevant target percentages are 65% by 2006, 70% by 2010 and 75% by There is also a target to increase this proportion year by year. The way in which the target is to be calculated is contrary to the format used to the develop the models (which are generally expressed as percentages of all dwellings). Table 1 reports in the original format i.e. non decent homes with vulnerable occupiers as a proportion of all dwellings, as this allows the results to be readily compared with the other statistics. We do also, however, report this result as a percentage of vulnerable households living in decent homes, and while this is of little importance to the stock as a whole, it is important in the private sector and this information is provided in Table 2. We have also included in the tables information on the proportion of social rented housing within each authority and have further broken this down into the proportions which are owned by Registered Social Landlords (RSLs) and local authorities. BRE Client report number

10 9 Housing Stock Projections Definition of vulnerable The term vulnerable can take on a number of meanings but in this report it has a very precise definition which has been provided by the Office of the Deputy Prime Minister (ODPM). The ODPM defines vulnerable households as those in receipt of at least one of the principal means tested or disability related benefits. For the purpose of establishing the national 2001 baseline from the English House Condition Survey, the benefits taken into account were: income support, housing benefit, council tax benefit, disabled persons tax credit, income based job seekers allowance, working family s tax credit, attendance allowance, disability living allowance, industrial injuries disablement benefit, war disablement pension. However the ODPM qualify this definition with the following: - The detailed definition of qualifying benefits used to define vulnerable will be subject to change and since 2001 a new range of tax credits has indeed been introduced with different qualifying thresholds. These are child tax credit, working tax credit and pension credit. The definition of vulnerable households used to monitor progress towards the target has therefore been amended as follows. In addition to the benefits described in the previous paragraph, pension credit will be included as a qualifying benefit. Also households in receipt of either working tax credit which includes a disability element or child tax credit will qualify as a vulnerable household providing the person entitled to the tax credit has a relevant income of less than 14,200, as defined for the purpose of determining eligibility for the tax credit. Working Families Tax Credit and Disabled Persons Tax credit have been abolished. We would wish to make it clear that for the purposes of this analysis we can only currently use data based on the original definition. As yet no dataset of amended data has been published, so we cannot update the model using the new definition. The Decent Homes model and its components The decent homes model and each of its components are separately derived models. This means it is quite possible in an individual authority or ward for the sum of the results of the components models to be more, or less, than the overall percentage produced by the decent homes model. This again emphasises the importance of using the models for their intended purpose; i.e. to look at the spatial distribution of housing conditions, rather than for precise predictions of the actual level of unfitness, decent homes etc. in a particular area. For this reason one of the best ways of viewing the data is to use maps which are more effective in displaying patterns rather than precise predictions. BRE Client report number

11 BRE Client report number Table 1 Modelled data, all dwellings (proportions and totals) Brent Ealing Ward Dwellings Non decent Hammersmith and Fulham Harrow Hillingdon Hounslow Kensington and Chelsea TOTAL All Non decent as a percentage of all dwellings Thermal comfort Unfit Disrepair Non modern Inadequate thermal comfort as a percentage of all dwellings Unfitness as a percentage of all dwellings Vulnerable occupants Vulnerable households as a Non modern Disrepair as a as a percentage of percentage of percentage of all dwellings all dwellings all dwellings Vulnerable non decent Vulnerable households in non decent dwellings as a percentage of all dwellings Vulnerable decent Social - LA Social - RSL Social - all Vulnerable households in decent dwellings as a percentage of all vulnerable households Social LA as a percentage of all dwellings Social RSL as a percentage of all dwellings Social all as a percentage of all dwellings ( 38% ) ( 25% ) ( 5% ) ( 12% ) ( 7% ) ( 28% ) ( 11% ) ( 60% ) ( 11% ) ( 13% ) ( 24% ) ( 40% ) ( 27% ) ( 5% ) ( 12% ) ( 7% ) ( 23% ) ( 9% ) ( 60% ) ( 12% ) ( 7% ) ( 19% ) ( 43% ) ( 31% ) ( 7% ) ( 15% ) ( 9% ) ( 26% ) ( 11% ) ( 57% ) ( 19% ) ( 13% ) ( 32% ) ( 34% ) ( 24% ) ( 4% ) ( 10% ) ( 5% ) ( 19% ) ( 7% ) ( 64% ) ( 7% ) ( 4% ) ( 11% ) ( 33% ) ( 24% ) ( 3% ) ( 9% ) ( 4% ) ( 21% ) ( 7% ) ( 66% ) ( 12% ) ( 5% ) ( 17% ) ( 36% ) ( 26% ) ( 4% ) ( 11% ) ( 5% ) ( 25% ) ( 9% ) ( 63% ) ( 17% ) ( 6% ) ( 23% ) ( 45% ) ( 33% ) ( 8% ) ( 15% ) ( 9% ) ( 24% ) ( 11% ) ( 55% ) ( 8% ) ( 17% ) ( 25% ) ( 38% ) ( 27% ) ( 5% ) ( 12% ) ( 7% ) ( 24% ) ( 9% ) ( 61% ) ( 12% ) ( 9% ) ( 21% ) 10 Housing Stock Projections N.B. This table gives the totals as output by the models and then converts them to percentages (excluding the final row which is sourced direct from the EHCS). The percentages are percentages of all the dwellings stock except for the vulnerable decent column which is vulnerable households in decent dwellings as a percentage of all vulnerable households.

12 BRE Client report number Table 2 Modelled data, private sector (as a percentage of all private sector dwellings) Brent Ealing Ward Dwellings Non decent Hammersmith and Fulham Harrow Hillingdon Hounslow Kensington and Chelsea TOTAL Private Private sector non decent as a percentage of all private sector dwellings Thermal comfort Unfit Disrepair Non modern Private sector inadequate thermal comfort as a percentage of all private sector dwellings Private sector unfitness as a percentage of all private sector dwellings Private sector disrepair as a percentage of all private sector dwellings Private sector non modern as a percentage of all private sector dwellings Vulnerable occupants Private sector vulnerable households as a percentage of all private sector dwellings Vulnerable non decent Private sector vulnerable households in non decent dwellings as a percentage of all private sector dwellings Vulnerable decent Private sector vulnerable households in decent dwellings as a percentage of all private sector vulnerable households ( 38% ) ( 25% ) ( 5% ) ( 13% ) ( 7% ) ( 24% ) ( 10% ) ( 61% ) ( 40% ) ( 26% ) ( 5% ) ( 13% ) ( 7% ) ( 20% ) ( 8% ) ( 60% ) ( 43% ) ( 30% ) ( 7% ) ( 16% ) ( 9% ) ( 21% ) ( 9% ) ( 57% ) ( 34% ) ( 23% ) ( 4% ) ( 10% ) ( 5% ) ( 17% ) ( 6% ) ( 64% ) ( 32% ) ( 24% ) ( 3% ) ( 9% ) ( 4% ) ( 19% ) ( 6% ) ( 66% ) ( 36% ) ( 25% ) ( 4% ) ( 11% ) ( 5% ) ( 20% ) ( 7% ) ( 63% ) ( 46% ) ( 33% ) ( 8% ) ( 16% ) ( 9% ) ( 18% ) ( 8% ) ( 54% ) ( 38% ) ( 26% ) ( 5% ) ( 12% ) ( 6% ) ( 20% ) ( 8% ) ( 61% ) 11 Housing Stock Projections N.B In this table the percentages have been calculated by taking the percentage rate of any model output at the level of the census output area and then multiplying it by the private sector stock total. This can then be summed to ward and borough level. This gives a total for the private sector for the particular model. They are then expressed as percentages of all private sector dwellings. These results should be readily comparable to any private sector house condition survey results which may be available. The one output we express differently is private sector vulnerable households in decent dwellings. This is expressed as a percentage of vulnerable households to facilitate comparison with ODPM targets.

13 BRE Client report number Table 3 Modelled data, private sector (as a percentage of all dwellings) Brent Ealing Ward Dwellings Non decent Hammersmith and Fulham Harrow Hillingdon Hounslow Kensington and Chelsea All Private sector non decent as a percentage of all dwellings Thermal comfort Unfit Disrepair Non modern Private sector inadequate thermal comfort as a percentage of all dwellings Private sector unfitness as a percentage of all dwellings Private sector disrepair as a percentage of all dwellings Private sector non modern as a percentage of all dwellings Vulnerable occupants Private sector vulnerable households as a percentage of all dwellings Vulnerable non decent Private sector vulnerable households in non decent dwellings as a percentage of all dwellings ( 29% ) ( 19% ) ( 4% ) ( 10% ) ( 5% ) ( 19% ) ( 7% ) ( 32% ) ( 21% ) ( 4% ) ( 10% ) ( 6% ) ( 16% ) ( 6% ) ( 29% ) ( 20% ) ( 5% ) ( 11% ) ( 6% ) ( 14% ) ( 6% ) ( 30% ) ( 21% ) ( 3% ) ( 9% ) ( 4% ) ( 15% ) ( 5% ) ( 27% ) ( 20% ) ( 3% ) ( 7% ) ( 3% ) ( 16% ) ( 5% ) ( 28% ) ( 19% ) ( 3% ) ( 8% ) ( 4% ) ( 15% ) ( 6% ) ( 34% ) ( 25% ) ( 6% ) ( 12% ) ( 7% ) ( 14% ) ( 6% ) N.B In this table the private sector dwelling totals are expressed as percentages of all dwellings. We express them as percentages of all dwellings as this gives an immediate indicator of where the main concentrations of housing stress in the private sector are to be found. 12 Housing Stock Projections

14 13 Housing Stock Projections The alliance will need to decide how much credence to give to the individual model results. In this context the results of comparisons with local authority house condition survey data are very informative. Comparisons with local authority data The WLA will need to decide how much credence to give to the individual model results. In this context the results of comparisons with local authority house condition survey data are very informative. A total of 16 local authorities throughout the country have now provided house condition survey for comparison with the model outputs including Kensington and Chelsea from within the sub-region. The databases supplied have only occasionally contained all eight variables that could be compared against model outputs. The comparison process itself presents a number of challenges due to unavoidable factors in particular the small sample size in each census output area requires aggregation of census output areas to allow sample sizes to become large enough to make viable comparisons. The process itself begins with the output areas ranked in order, starting with the highest proportion of dwellings likely to fail the variable being compared e.g. non decent homes, and ending with the COA with the lowest. The survey cases provided by the Authority are then matched via postcodes to our model. The cases are then weighted using the variable any survey weighting factor used in the survey. To compare the model and survey data the output areas are then grouped into clusters. Left ungrouped, the small number of survey cases per output area would render any comparisons meaningless. The number of clusters is set such that each cluster includes a minimum of 30 survey addresses per cluster. This is sufficient to allow calculation of a percentage score for each cluster that iss comparable with the model percentage. In order to determine which COA are placed in which cluster, the COAs are sorted into order of predicted percentage of dwellings failing the standard. The first groups COAs were placed in cluster one, the next group in cluster two, etc. The results can then be illustrated in a graph such as that for Kensington and Chelsea for unfitness. The survey data follows a similar trend as the modelled data, indicating that there is a relationship between the modelled and the surveyed data. This can be quantified by calculating the correlation coefficient for the clustered modelled and surveyed scores. In this case, the correlation coefficient is 0.4, indicating a reasonable degree of correlation between the survey data and the model for non-decent homes. To place this results into context it is useful to consider what it is a good correlation coefficient and secondly compare these to others achieved elsewhere. Correlation coefficients are on a scale from -1 to 1. A score of 0 would mean that there is no relationship between the model predictions and the survey data. A score of 1 would mean there is a perfect correlation between the two sets of data. A score of -1 would mean there was an inverse correlation i.e. as the model outputs increase the local authority survey decreases. BRE Client report number

15 14 Housing Stock Projections When we commenced the development of the models our target was to provide useful models. Our research led us to conclude of 0.6 or above would provide useful models while those with results in the might be better regarded as potentially useful providing some local explanation could be found for some of the differences between the model and survey results. The Kensington and Chelsea result falls into this potentially useful range. unfitness Percentage cluster 1 cluster 2 cluster 3 cluster 4 cluster 5 cluster 6 cluster 7 cluster 8 cluster 9 cluster 10 cluster 11 cluster 12 cluster 13 cluster 14 cluster 15 cluster 16 cluster 17 cluster 18 cluster 19 cluster 20 cluster 21 COA cluster unfitness - model unfitness - survey Table 4 summarises the results for all comparison undertaken to date. Not all of the variables were available for comparison e.g. in Kensington and Chelsea only unfitness and disrepair were available, but those that were are reported on and their significance evaluated. The other local authorities are not identified, but we do provide an indication of their geographical region and type. BRE Client report number

16 15 Housing Stock Projections Table 4 Correlation coefficients of comparisons between modelled and local authority survey data Inadequate thermal comfort Unfitness Disrepair Vulnerable occupants of non-decent homes Nondecent homes Nonmodern SAP less Fuel Model than 30 poverty Outer London Outer London Outer London Outer London Outer London Outer London Outer London Outer London 0.5 Outer London South East Urban Borough Inner London Inner London Kensington & Chelsea Inner London Inner London North west metropolitan 0.8 North west metropolitan Yorks and Humberside Metropolitan Yorks and Humberside Metropolitan Unfitness has the best record, with seven correlation coefficients of 0.7, one of 0.8 and three of 0.9. These are exceptionally good results, and indicate that the model is a powerful predictor of concentrations of unfit dwellings. Disrepair also emerges overall as a successful model with nine results in what we generally regard as the useful 0.5 to 1.0 range, and four which are potentially useful in the range. There is one particularly poor result in a Yorkshire and Humberside authority where the database provided had been subject to a exercise to project forward likely rates of non decency and this may account for their -0.4 result as we are comparing these with our 2001 based data. It is more difficult to explain the -0.1 result in the other Yorkshire and Humberside authority although the data for disrepair in did return far higher levels of disrepair than predicted by the model. Further investigation of the survey data would be needed to determine why this should be so. The non decent homes result is more mixed, with excellent result of 0.9 and 0.7 in Yorkshire and Humberside three more of 0.7 but two very unusual ones of -0.1 and -0.4 (the -0.4 result was associated with a large proportion of missing survey data but is still reported for completeness). The inadequate thermal comfort model follows a similar pattern but with lower correlations, five in the useful range of but also two unusually disappointing results of -0.2 and -0.7 (the latter again associated with missing survey data). What appears to be occurring is that the more moderate performance of the inadequate thermal comfort model is counterbalanced by disrepair and unfitness, resulting in a good performance for non decent homes. The main exception to this is where the inadequate thermal comfort measure has performed particularly poorly in two inner London boroughs, and good results from unfitness and disrepair have failed to fully offset this. BRE Client report number

17 16 Housing Stock Projections The non modern variable which has produced only low correlation coefficients seems to have little influence on the overall decent homes performance. Because lack of modern amenities is such a rare occurrence it has always proved difficult to model, but has little influence on decent homes. In Yorkshire and Humberside the poor result is again associated with the forward projections. So far we have six results for the vulnerable occupiers/non decent homes variable of which four are in the useful range and two are from London. Of the remaining two models, SAP less than 30 and fuel poverty, it gives confidence to see a good performance from the SAP less than 30 model with consistent correlations in the potentially useful range When the comparisons are looked at geographically it is fairly obvious that some of the most of the problematic results occur in inner London boroughs. There are two main explanations have been suggested for this and they are worth briefly considering. It has been argued that with their unusual social and housing mix, the stock in the most central of the London authorities may not be susceptible to modelling. The evidence for this is mainly anecdotal and particularly relates to experience of areas where there are very high rates of second home ownership. It is thought that second homes have often been wrongly enumerated by the census as vacant dwellings. High rates of vacancy in an area are usually indicative of housing stress and for some of our models this is an important weighting factor. This is thought to be one of the main reasons why part of Kensington and Chelsea return such high results. A second important factor relates mainly to the thermal efficiency models. London is unusual in that a large proportion of the stock was built with solid walls during the inter war period. In these areas the models will tend to underestimate poor levels of thermal efficiency and the position appear better than it actually is. This will to a limited extent be catered for by the models by an overall geographical factor for London but it will not be sensitive within London. It is also important to remember that the models are not being compared using an exact yardstick. The combination of small samples and surveyor variability in house condition surveys mean that expectations of consistently high correlations with the models are unrealistic. Despite these issues we regard the results overall as very encouraging and with the exception of non modern amenities and fuel poverty the indications are that the models provide a reasonable guide to local housing conditions. The relatively few comparisons undertaken with the fuel poverty data mean that it is probably too early to form a judgement of the utility of this model so this data should be approached with caution. With it s very low incidence it may be that non modern amenities cannot be modelled successfully. We do however caution against completely discarding it as surveys face equally difficult problems in determining the incidence of such a variable. Fortunately it makes a very small contribution to non decency and is therefore not a major deficiency of the housing stock modelling system. BRE Client report number

18 17 Housing Stock Projections Key statistics by census wards The following maps of key statistics are based on data which is provided in a separate spreadsheet (Key Statistics WLA.xls). For the first 11 maps (Maps 1 to 5b) the data has been mapped using the standard deviation of the Census Output Area statistics to define the ranges. This results in the more vivid colours being assigned to the extreme ends of the range, which therefore highlights the best and worst areas. For the maps of vulnerable occupiers of non decent homes (Maps 9a and 9b) the ranges include government targets of 65%, 70% and 75%. It should be stressed that there is no requirement to meet these targets at local level, but other authorities have been interested in these distributions and we therefore continue to include them for interest. The first map is of the distribution of the social rented stock which helps to see the subsequent private sector maps in context. Thereafter the maps include an all dwelling and a private sector stock map for each variable, except for the vulnerable occupiers of non decent homes maps, which are provided solely for the private sector as they have no real relevance to the social rented stock. BRE Client report number

19 BRE Client report number Map 1. Social rented stock as a percentage of all dwellings 18 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

20 BRE Client report number Map 2a. Private sector non decent dwellings as a percentage of all private sector dwellings 19 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

21 BRE Client report number Map 2b. Non decent private sector dwellings as a percentage of all dwellings 20 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

22 21 Housing Stock Projections Across the region, the levels of private sector non decency are generally highest in the south east, and progressively lower away from the centre of London. Kensington and Chelsea shows the highest borough wide level of non decency at 46% (of the private sector stock), and Hillingdon the lowest at 32%. When the private sector stock totals for non decent homes are calculated as a percentage of all dwellings, the percentages are of course lower; but the overall pattern is similar; except that Hammersmith and Fulham, the authority with the lowest proportion of private housing, no longer has the second highest level of non decent homes. At 29% it is now overtaken by Harrow (30%) and Ealing (32%) and equalled by Brent. When the non decent private stock is calculated as a percentage of all the dwelling stock this is arguably a better measure of the absolute demand for private sector housing services. Kensington and Chelsea, and Hammersmith and Fulham, have the highest levels, and Harrow and Hillingdon have the lowest. This is shown up in map 2a where intensity of the highest areas of non decency are shown up in Kensington and Chelsea, Hammersmith and Fulham and the southeastern part of Brent. BRE Client report number

23 BRE Client report number Map 3a. Private sector dwellings failing the decent homes standard due to inadequate thermal comfort as a percentage of all private sector dwellings 22 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

24 BRE Client report number Map 3b. All private sector dwellings failing the decent homes standard due to inadequate thermal comfort as a percentage of all dwellings 23 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

25 24 Housing Stock Projections At authority level, the proportion of private sector dwellings failing on inadequate thermal comfort varies from 23% in Harrow to 46% in Kensington and Chelsea. The inner/outer pattern of inadequate thermal comfort is slightly less well defined than that of non decency, but the distributions for both maps are broadly similar. This is much as would be expected with the major contribution made by inadequate thermal comfort failures to non decency at a national level. BRE Client report number

26 BRE Client report number Map 4a. Private sector dwellings failing the decent homes standard due to unfitness as a percentage of all private sector dwellings 25 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

27 BRE Client report number Map 4b. All private sector dwellings failing the decent homes standard due to unfitness as a percentage of all dwellings 26 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

28 27 Housing Stock Projections At the Borough level Hillingdon (3%) has the lowest private sector unfitness level, and Hammersmith and Fulham (8%) has the highest. At the ward level the most notable concentrations of unfitness are in Kensington and Chelsea, crossing over the border into the eastern side of Hammersmith and Fulham. The lowest levels are as with previous distributions found at the periphery but for unfitness this is more noticeable in the north of the sub-region. BRE Client report number

29 BRE Client report number Map 5a. Private sector dwellings failing the decent homes standard due to disrepair as a percentage of all private sector dwellings 28 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

30 BRE Client report number Map 5b. All private sector dwellings failing the decent homes standard due to disrepair as a percentage of all dwellings 29 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

31 30 Housing Stock Projections Hammersmith and Fulham, and Kensington and Chelsea (16%) return the highest levels of private sector disrepair closely followed by Brent and Ealing (13%) with Hillingdon returning the lowest at 9%. At ward level the percentages follow a very similar pattern to unfitness with the familiar inner outer trend. BRE Client report number

32 BRE Client report number Map 6a. Private sector dwellings failing decent homes standard due to non modern amenities as a percentage of all private sector dwellings 31 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

33 BRE Client report number Map 6b. All private sector dwellings failing decent homes standard due to non modern amenities as percentage of all dwellings 32 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO

34 33 Housing Stock Projections The overall pattern of private homes with non modern amenities is very similar to those of unfitness and disrepair, being highest in the south east of the region. The highest percentage occurs in Hammersmith and Fulham, and Kensington and Chelsea (both 9%), with Hillingdon lowest at 4%. BRE Client report number

35 BRE Client report number Map 7a. Private sector vulnerable occupiers of decent homes as a percentage of all private sector vulnerable occupiers 34 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

36 35 Housing Stock Projections In map 7a the green areas meet 2006 and 2010 targets of 65% and 70% respectively. No areas meet the 2020 target. Only in Hillingdon and Harrow do a substantial number of wards meet the 2006 target. In other authorities, very few areas meet this target, and in Kensington and Chelsea, none at all. It is however important to note that there is no requirement for any one area to meet these targets. They do however serve to show where resources may usefully need to be targeted. The definition of vulnerable groups for this model is the same as used by the ODPM to determine the national average from EHCS data and has been described earlier in some detail. BRE Client report number

37 BRE Client report number Map 7b. Private sector vulnerable occupiers of decent homes as a percentage of all private sector vulnerable occupiers (additional ranges) 36 Housing Stock Projections Boundary data for all maps: Source: 2001 Census, Output Area Boundaries, Crown Copyright Crown copyright material reproduced with the permission of the Controller of HMSO.

38 37 Housing Stock Projections This map shows that of the areas failing to meet the 2006 target of 65% vulnerable occupants having decent housing, the areas in greatest need of private housing assistance are in Kensington and Chelsea, the adjoining eastern end of Hammersmith and Fulham, and the western end of Ealing. The southern areas of Hillingdon and Hounslow also have elevated levels of need in comparison with their overall low level of non decent housing. BRE Client report number

39 38 Housing Stock Projections Summary and conclusions The BRE Housing Stock Models have provided information on: non-decent homes and its four components, non-decent homes occupied by vulnerable groups The information has been provided in data (the spreadsheet Key Statistics WLA.xls) and map format for the whole of the dwelling stock and separately for the private sector stock. Basic descriptions of the authority data and the ward maps of the region have been provided. We have also included a summary of comparisons between the models and local authority survey data to help evaluate the models and this comparison includes information provided by Kensington and Chelsea. We have now developed tools to generate a spreadsheet which provides pertinent census data to complement the model data. The spreadsheet (WLA LA Census Data.xls) is designed for easy import into GIS software which is the best way to interrogate the data. A simple user interface has, however, been added to allow the user to examine individual ward data. The spreadsheet also includes census data not regarded as significant by the models but which may well be of interest to users e.g. tenure, ethnicity. We have refrained from offering any detailed analysis of the data or the distributions as we feel the interpretation of the data is best left to those familiar with their own areas while we concentrate on providing the tools and the data to enable this. We would recommend that the data be carefully evaluated and compared to other sources and ideally imported into a GIS system where it can be overlaid onto street plans and other data held by the authorities. Added to other information sources this should now greatly assist in the further development of the sub-regional housing strategy. BRE Client report number

40 BRE Client report number Appendix A Modelled data: private sector wards Ward Dwellings Non decent Thermal comfort Unfit Disrepair Non modern Vulnerable occupants Vulnerable non decent Vulnerable decent Private Private sector non decent as a percentage of all private sector dwellings Private sector inadequate thermal comfort as a percentage of all private sector dwellings Private sector unfitness as Private sector disrepair as a percentage of all private a percentage of all private sector dwellings sector dwellings Private sector non modern as a percentage of all private sector dwellings Private sector vulnerable households as a percentage of all private sector dwellings Private sector vulnerable households in non decent dwellings as a percentage of all private sector dwellings Private sector vulnerable households in decent dwellings as a percentage of all private sector vulnerable households Brent : Alperton ( 40% ) ( 25% ) ( 5% ) ( 13% ) ( 6% ) ( 24% ) ( 9% ) ( 61% ) Brent : Barnhill ( 36% ) ( 25% ) ( 4% ) ( 10% ) ( 7% ) ( 25% ) ( 10% ) ( 59% ) Brent : Brondesbury Park ( 39% ) ( 26% ) ( 5% ) ( 13% ) ( 8% ) ( 23% ) ( 9% ) ( 61% ) Brent : Dollis Hill ( 34% ) ( 22% ) ( 4% ) ( 11% ) ( 5% ) ( 25% ) ( 9% ) ( 65% ) Brent : Dudden Hill ( 37% ) ( 23% ) ( 6% ) ( 13% ) ( 7% ) ( 23% ) ( 9% ) ( 62% ) Brent : Fryent ( 37% ) ( 26% ) ( 4% ) ( 12% ) ( 6% ) ( 23% ) ( 9% ) ( 61% ) Brent : Harlesden ( 43% ) ( 26% ) ( 7% ) ( 14% ) ( 7% ) ( 39% ) ( 17% ) ( 57% ) Brent : Kensal Green ( 42% ) ( 26% ) ( 7% ) ( 15% ) ( 7% ) ( 26% ) ( 11% ) ( 58% ) Brent : Kenton ( 30% ) ( 20% ) ( 3% ) ( 10% ) ( 5% ) ( 15% ) ( 5% ) ( 68% ) Brent : Kilburn ( 44% ) ( 29% ) ( 7% ) ( 15% ) ( 10% ) ( 33% ) ( 15% ) ( 55% ) Brent : Mapesbury ( 41% ) ( 26% ) ( 7% ) ( 16% ) ( 11% ) ( 24% ) ( 10% ) ( 59% ) Brent : Northwick Park ( 30% ) ( 21% ) ( 3% ) ( 10% ) ( 4% ) ( 16% ) ( 5% ) ( 68% ) Brent : Preston ( 33% ) ( 22% ) ( 4% ) ( 11% ) ( 4% ) ( 20% ) ( 7% ) ( 67% ) Brent : Queens Park ( 42% ) ( 27% ) ( 7% ) ( 15% ) ( 8% ) ( 20% ) ( 9% ) ( 57% ) Brent : Queensbury ( 35% ) ( 24% ) ( 4% ) ( 12% ) ( 4% ) ( 20% ) ( 7% ) ( 64% ) Brent : Stonebridge ( 36% ) ( 24% ) ( 5% ) ( 10% ) ( 7% ) ( 43% ) ( 16% ) ( 62% ) Brent : Sudbury ( 37% ) ( 26% ) ( 5% ) ( 11% ) ( 5% ) ( 24% ) ( 9% ) ( 63% ) Brent : Tokyngton ( 38% ) ( 24% ) ( 4% ) ( 13% ) ( 6% ) ( 22% ) ( 8% ) ( 61% ) Brent : Welsh Harp ( 35% ) ( 23% ) ( 4% ) ( 12% ) ( 5% ) ( 27% ) ( 9% ) ( 65% ) Brent : Wembley Central ( 39% ) ( 24% ) ( 5% ) ( 13% ) ( 6% ) ( 27% ) ( 11% ) ( 61% ) Brent : Willesden Green ( 43% ) ( 26% ) ( 8% ) ( 17% ) ( 9% ) ( 30% ) ( 13% ) ( 57% ) 39 Housing Stock Projections

41 BRE Client report number Ward Dwellings Non decent Thermal comfort Unfit Disrepair Non modern Vulnerable occupants Vulnerable non decent Vulnerable decent Private Private sector non decent as a percentage of all private sector dwellings Private sector inadequate thermal comfort as a percentage of all private sector dwellings Private sector unfitness as Private sector disrepair as a percentage of all private a percentage of all private sector dwellings sector dwellings Private sector non modern as a percentage of all private sector dwellings Private sector vulnerable households as a percentage of all private sector dwellings Private sector vulnerable households in non decent dwellings as a percentage of all private sector dwellings Private sector vulnerable households in decent dwellings as a percentage of all private sector vulnerable households Ealing : Acton Central ( 41% ) ( 27% ) ( 7% ) ( 15% ) ( 8% ) ( 22% ) ( 9% ) ( 58% ) Ealing : Cleveland ( 38% ) ( 26% ) ( 5% ) ( 11% ) ( 7% ) ( 17% ) ( 6% ) ( 62% ) Ealing : Dormers Wells ( 41% ) ( 25% ) ( 5% ) ( 13% ) ( 7% ) ( 29% ) ( 12% ) ( 59% ) Ealing : Ealing Broadway ( 42% ) ( 28% ) ( 6% ) ( 14% ) ( 9% ) ( 13% ) ( 5% ) ( 58% ) Ealing : Ealing Common ( 39% ) ( 26% ) ( 5% ) ( 13% ) ( 8% ) ( 15% ) ( 6% ) ( 61% ) Ealing : East Acton ( 36% ) ( 24% ) ( 5% ) ( 12% ) ( 7% ) ( 26% ) ( 9% ) ( 64% ) Ealing : Elthorne ( 41% ) ( 27% ) ( 6% ) ( 14% ) ( 7% ) ( 19% ) ( 8% ) ( 60% ) Ealing : Greenford Broadway ( 41% ) ( 30% ) ( 5% ) ( 11% ) ( 7% ) ( 27% ) ( 11% ) ( 58% ) Ealing : Greenford Green ( 38% ) ( 25% ) ( 4% ) ( 13% ) ( 6% ) ( 18% ) ( 7% ) ( 61% ) Ealing : Hanger Hill ( 38% ) ( 26% ) ( 5% ) ( 12% ) ( 7% ) ( 14% ) ( 6% ) ( 62% ) Ealing : Hobbayne ( 38% ) ( 26% ) ( 4% ) ( 12% ) ( 6% ) ( 21% ) ( 8% ) ( 62% ) Ealing : Lady Margaret ( 43% ) ( 26% ) ( 5% ) ( 14% ) ( 6% ) ( 22% ) ( 9% ) ( 57% ) Ealing : Northfield ( 40% ) ( 25% ) ( 5% ) ( 14% ) ( 7% ) ( 11% ) ( 4% ) ( 62% ) Ealing : North Greenford ( 38% ) ( 25% ) ( 4% ) ( 12% ) ( 6% ) ( 18% ) ( 7% ) ( 62% ) Ealing : Northolt Mandeville ( 41% ) ( 31% ) ( 4% ) ( 10% ) ( 6% ) ( 25% ) ( 11% ) ( 57% ) Ealing : Northolt West End ( 40% ) ( 33% ) ( 4% ) ( 8% ) ( 5% ) ( 31% ) ( 13% ) ( 57% ) Ealing : Norwood Green ( 35% ) ( 21% ) ( 4% ) ( 11% ) ( 6% ) ( 27% ) ( 10% ) ( 65% ) Ealing : Perivale ( 38% ) ( 25% ) ( 4% ) ( 12% ) ( 5% ) ( 19% ) ( 7% ) ( 63% ) Ealing : South Acton ( 43% ) ( 29% ) ( 7% ) ( 14% ) ( 9% ) ( 25% ) ( 11% ) ( 57% ) Ealing : Southall Broadway ( 41% ) ( 19% ) ( 6% ) ( 15% ) ( 7% ) ( 27% ) ( 11% ) ( 59% ) Ealing : Southall Green ( 43% ) ( 22% ) ( 6% ) ( 15% ) ( 7% ) ( 28% ) ( 12% ) ( 57% ) Ealing : Southfield ( 40% ) ( 27% ) ( 7% ) ( 14% ) ( 7% ) ( 12% ) ( 5% ) ( 60% ) Ealing : Walpole ( 39% ) ( 25% ) ( 6% ) ( 14% ) ( 7% ) ( 12% ) ( 5% ) ( 62% ) 40 Housing Stock Projections