•   National Survey of Charities and Social Enterprises

Methodology and interpreting findings

Important note on the methodology and findings

  • The National Survey of Third Sector Organisations was conducted amongst third sector organisations (TSOs) across 149 upper tier local authorities in England from September to December 2008.

  • In total 104,391 TSOs across all 149 single and two-tier authorities in England were invited to participate in the survey. Initially pre-notice letters were mailed to all selected organisations (along with a link enabling them to access the survey online). Paper questionnaires were then mailed to all selected organisations shortly afterwards. Thank you/reminder postcards and reminder questionnaires were also sent out to organisations, with telephone reminders conducted with a sample of non-responding organisations. TSOs were able to complete the survey either online or on paper throughout the fieldwork period.

  • A database of TSOs drawing on the list of registered charities and registers of Community Interest Companies, Companies Limited by Guarantee and Industrial and Provident Societies in England, was supplied by Guidestar UK. In some areas a census survey was conducted (i.e. all TSOs were asked to take part), and in others a random stratified sample of TSOs was constructed, using organisation type (registered charity, CIC, CLG or IPS) as the main stratifier. Registered charities were further stratified according to income.

  • A total of 48,939 third sector organisations responded to the survey – this was a 47% response rate.

  • Data from the survey have been weighted within each local authority and at a national level to ensure results are representative of the make up of TSOs according to organisational type.

  • Because a sample, rather than the entire population, was interviewed the percentage results are subject to sampling tolerances – which vary with the size of the sample and the percentage figure concerned. For example, for a question where 16% of the respondents in an unweighted sample of 200 respond with a particular answer, the chances would be 95 in 100 that this result would not vary more than five percentage points, plus or minus, from the result that would have been obtained from a census of the entire (infinite) population (using the same procedures). Strictly speaking these tolerances apply only to random non self selecting samples. However, they are a good guide here.

  • Tolerances are also involved in the comparison of results between different elements of the sample. A difference must be of at least a certain size to be statistically significant. Each comparison made involves a number of elements to calculate whether a difference is statistically significant. This includes the proportion answering a question, the number of respondents and the population size. Caution must be taken when comparing any results of this survey. This is particularly the case where there are small sample sizes in a number of local authorities, due to small population sizes (where a census was conducted).