Skip to content

Sampling charities

For the purposes of manual classification, we needed to create a number of batches of charities that would be manually classified.

The population used for sampling included all charities on the main registers for the three UK charity regulators (Charity Commission for England and Wales, OSCR (Scottish Charity Regulator) and the Charity Commission for Northern Ireland).

The sampling strategy was as follows:

First, we removed any inactive charities from the files, leaving 200,544.

Keyword searches were performed for a small number of groups of easily identified, homogenous charities, based on their name. These groups could then be excluded in bulk, meaning that the remaining charities would have more variety. The groups that were excluded were:

  • Parent Teacher Association (7,430 charities)
  • Village Hall (6,865)
  • Scout groups (4,266)
  • Girlguiding groups (2,805)
  • Parochial Church Council (2,746)
  • Playgroups (1,740)

The remaining 174,692 charities were then split into two income bands, based on the latest available income of each charity. The threshold used was £100,000. 39,826 (23%) of the remaining charities were above this threshold and 134,866 (77%) below it.

Between the different members of the research team 4,203 charities were ultimately classified using the UK-CAT and the ICNP/TSO. This included 1,328 charities with an income of £100,000 or over and 2,875 charities with a lower income.

Larger charities were, therefore, proportionately over-represented in the manual classification. This is partly due to their financial weight, to try and improve eventual estimations of financial totals for each service area. Secondly, some types of charities are disproportionately found amongst larger charities (for example medical research charities). In contrast, many smaller charities fall into a relatively small number of categories, such as places of worship, grant makers and community associations. Without oversampling larger charities, therefore, some classification types would be likely to be underrepresented or missed out entirely from the manually classified sample.

In addition to the over-representation of larger charities within the random sample, a separate sample of the largest 2,000 charities, as measured by their latest total incoming resources was constructed. All of these charities were classified into ICNP/TSO categories, but UK-CAT tags were not applied to these organisations.