Home | Data | Methodology | Blog | Contact
Methodology
Data Collection
This project uses publicly available data on charity registrations (foundations) and de-registrations (dissolutions) from seven charity jurisdictions. We use a programming script, written in the popular Python language, to collect data for each jurisdiction. The exception is Canada, where the data are manually downloaded using the public search facility here: [LINK] .
More information on our data sources can be found here: [LINK]
The data collection script can be found here: [LINK]
Data Cleaning
Each data set is loaded into the statistical software package Stata, where the following general processes are implemented:
- Collapse the data down to monthly observations i.e., each row in the data represents a month.
- For each month, calculate the number of charities:
- registered
- de-registered
- Create additional measures, such as:
- average number of registrations/de-registrations per month
- excess number of registrations/de-registrations per month, etc
- Save and export to CSV for sharing
Data Analysis
The analysis is based on an “excess events” analytical approach, comparing the numbers of registrations and de-registrations in 2020 to what we would expect based on the trends from previous years. For example, let’s say there were 50 new charities registered in Scotland in January 2020 - is that number large or small, expected or unexpected based on previous figures for January?
| Month | Number of new charities |
|---|---|
| January 2015 | 60 |
| January 2016 | 55 |
| January 2017 | 82 |
| January 2018 | 65 |
| January 2019 | 75 |
The average number of new charities in January between 2015-2019 is: (60 + 55 + 82 + 65 + 75) / 5 = 67. So there are 17 fewer new charities in January 2020 compared to the average for 2015-2019. Of course, the figures for January vary each year, and thus we need to know if the figure for January 2020 falls outside the range of expected registrations for that month. Therefore we calculate the standard deviation of the average, which is 10, and use this to construct the range: 57 to 77 (i.e., 67 +- 10). So there are fewer new charities in January 2020 than we would expect.