CHAPTER 4 DESIGN OF RESEARCH INSTRUMENTS, DATA
4.2 Selection of study sites
4.2.1 Aims and sampling strategy of the I’DGO TOO project
This research uses data generated in conjunction with the I’DGO TOO project. The aims of the WISE part of I’DGO TOO were:
- To determine the pluses and minuses of ‘urban renaissance’ developments,
in terms of residential outdoor space and quality of life of older residents.
- To determine how, and to what extent, different types of residential
outdoor spaces (private gardens, shared gardens, balconies, courtyards, etc.) contribute to the quality of life of older people.
- To identify how best to design different types of residential, outdoor spaces
89
The focus was therefore on comparing urban renaissance (post 2000)
developments with earlier ones. The I’DGO TOO team decided to sample sites across Great Britain and to target a mix of different types of developments in the following categories:
- Tenure – private, social and mixed tenure housing
- Region – seven regions: Scotland, Wales and 5 in England
- Settlement types – city, town, village, rural
- Era – from pre-Victorian to Urban Renaissance (post 2000)
- Housing type – terrace, semi-detached and detached houses or
bungalows
and apartment blocks
- Density of the development
Within these categories a range of layouts was also selected as shown in Fig 3.1. The I’DGO TOO project also sampled age-specific housing, both private and social (that is, housing with a minimum age limit). The data from these sites is not included in this study.
The aim of the I’DGO TOO team was to obtain 2,800 responses from age-specific and non-age-specific developments. To this was added 1,190 individual houses from areas surrounding the developments, to give a target sample of 3,990. A response rate of 20 to 25% was assumed, so the aim was to send out 16,000 questionnaires against the sampling frame in Table 4.1. It was assumed that response rates from private housing would be better than from social housing and better from age-specific than from non-age-specific housing (Whitfield, 2003). The grey shaded columns in Table 4.1 represent the sample used in this research.
90 Table 4.1 Sampling frame for I’DGO TOO
Region Private age specific Private non-age specific Social age specific Social non-age specific Mixed non-age specific Total London 60 60 140 140 170 570 Midlands 60 60 140 140 170 570 North 60 60 140 140 170 570 Scotland 60 60 140 140 170 570 South-East 60 60 140 140 170 570 South-West 60 60 140 140 170 570 Wales 60 60 140 140 170 570 420 420 980 980 1190 3990
A clustered sampling strategy was used. Residential developments of interest were identified in each region and a range of housing around these
developments was selected to be included in the sample.
The individual housing was added for two reasons: firstly, it enabled I’DGO TOO to access older people living in their own homes, without having to discover where they were; secondly, it gave this study a sample of residents with their own individual, private residential outdoor space for comparison with those sharing residential outdoor space.
Several strategies were used to identify possible developments. Housing
Associations were approached to join the I’DGO TOO advisory group. Peabody Trust and Places for People joined the group and provided information about their residential developments. Recent (post 2000) developments were
91
identified from the CABE (Commission for Architecture and the Built
Environment) website www.cabe.org.uk (archived on 18th Jan 2011).
For each development considered for inclusion, the location, number of residents, approximate area and population density of the development, the layout and any special features were recorded. Developments were selected from this database to give the range of variables required. Non-age-specific social housing was identified in the surrounding area using information from District, Borough, City and Town Council and urban regeneration web-sites. Appendix A4.1 lists all the web-sites consulted. Sampling within the sites is discussed in Section 4.1.3. Mixed tenure housing was selected in the same areas visually, using aerial and birds-eye views from www.bing.com to give a range of housing age and types.
4.2.2 Summary of relationship between this research and I’DGO TOO Modifications made to the original I’DGO TOO sampling strategy to
accommodate this research are indicated in Table 4.2.
Table 4.2 Integration of sampling strategies
I’DGO TOO approach Modification Explanation
Focus on post 2000 developments Addition of older developments To provide a more balanced sample Comparison of age-
specific and non-age specific developments
Data from age-specific developments excluded
Interest is in all adults, so avoids sample being structurally skewed to older age groups Difficulty in accessing
older adults in their own home Sampling individual houses close to developments Provides a comparison with SROS
(Enables I’DGO TOO to identify incidental older people in the sample)
92
Use and modification of the large database generated for the I’DGO TOO project enabled this researcher to have access to much more data than could have been generated alone. As research assistant on the project this
researcher was fully involved in the design of the I’DGO TOO study and its instruments. All the map-based data collection for I’DGO TOO and for this thesis was done by this researcher alone. 332 developments and streets are included in this research.
4.2.3 Sampling strategy within sites
Ideally, the questionnaire should be sent either to the whole population or to a randomly selected sample. A random sample is one in which every member of the population has an equal and independent probability of being selected (McPherson, 2001). Theoretically, a true random sample (assuming it is not too small) will have a mix of characteristics close to that of the whole population, that is, it will be representative of the population. The usual and simplest method of generating a sample of households is systematic sampling in which
every ith house is selected. In this research all households in developments of
up to 60 dwellings were included, but for larger estates, systematic sampling
was used of every 2nd or 3rd house, to generate a sample of about 60 dwellings.
Response rates are usually considerably lower from those living in social housing than from those who rent privately or own their own homes (Whitfield, 2003). In a few large social housing estates, such as Peabody’s Old Pye Street, all the households were surveyed in an attempt to raise the number of responses from people in social housing in the complete sample.
93
Such systematic sampling is not random because the selection of each member of the population is not independent. This method may, however, give a more
representative sample than a truly random one, as members of thepopulation
may be clustered and sampling systematically ensures an even spread of
representatives from different clusters. It does cause difficulties in the application of statistical methods (McPherson, 2001), but these are avoided by using non- statistical methods of analysis (see Section 3.4.4). The lack of randomness caused by systematic sampling is, in any case, smaller than that generated by the self-selection of the respondents (i.e. those who complete and return the questionnaire). This self-selection ensures that the final sample is not random and is a major drawback of the self-completed survey. The sample is limited to those people who will fill in and return a questionnaire: a relatively small segment of a given population. The views of those who cannot read English, are too busy or are not interested, are, therefore, not represented. This means that the results cannot be generalised to the wider population (Dunn, 2010).