Statistical methods

This law firm diversity tool gives our current best estimate of diversity across different groups of firms. The data behind the tool has been statistically modelled, so we can give law firms a more accurate picture of diversity in the legal sector.

Law firms collected diversity data in 2014. Within this data, there were people who did not respond to one or all of the questions, or gave the response 'prefer not to say'.

This meant that there were gaps in the data, which we call an 'incomplete data set'. In order to improve the quality of this data and 'fill in the gaps', we used a proven statistical method called 'bootstrap re-sampling' to estimate what the diversity data would look like if nobody had responded 'prefer not to say'.

Bootstrap re-sampling is an established statistical method, used by organisations like the World Health Organisation because it can give better estimates of the health of populations than they could get if they used an incomplete data set.

We analysed 1000 samples of responses where all questions were answered, and then took an average of all these samples to get the figures used in the tool. This analysis of 1000 statistical samples provides a more accurate figure than if we used the data as it was, which would be distorted by the 'prefer not to say' responses.

Comparing the diversity figures in the original response data and those in the tool, we can see the effects that the statistical modelling has had. For example, the tool shows a slightly larger proportion of BME solicitors working in law firms than the original survey response data, because the pattern of 'prefer not to say' responses was not equally spread across the population. Confidence intervals, show how confident we are in the accuracy of each figure.

If you are interested in viewing how everyone responded to the survey, rather than using the toolkit, which relies on statistically modelled data, we report this information annually in our reports on workforce data for solicitors firms.¹