The Cabinet Office recently published some data on social investments made by charitable foundations in the UK, along with some interesting visualisations.
It looks really quite interesting, but before getting too excited about the output, I decided to look into the data a little more carefully. I wanted to see how consistent data entry was. Given that social impact bonds are a core area, and all three of the participating foundations had been involved in the Peterborough Social Impact Bond, that seemed a natural choice to look at in more detail.
What you can see in the table below is, unfortunately, a lot of inconsistency in the data.
| Foundation | The LankellyChase Foundation | Panahpur | The Barrow Cadbury Trust |
| Source Of Investment | WC1R 4BH | tn9 1ap | WC2B4AS |
| Investment date (commitment for investment/ facility/ guarantee) | 27 October 2010 | 10 November 2010 | 10 November 2013 |
| Investment date (cash draw down) | 1 November 2010 | 10 November 2010 | Split over 6 dates |
| Original scheduled final full repayment/ redemption date | 1 November 2018 | 20 November 2013 | 20 November 2019 |
| Commitment or facility provided - amount guaranteed or underwritten | £- | £100,000 | £100,000 |
| Cash invested - principal invested/ amount guaranteed or underwritten | £500,000 | £54,193 | £54,193 |
| Total Value incl. co-investment | £5,000,000 | £5,000,000 | £4,200,000 |
| Payment Frequency | Other | Annually | Other |
| Bullet Payment | Yes | No | Yes |
| Balloon Payment | Yes | No | No |
| Product Type | Bonds | Equity | Equity |
| Purpose Of Investment | Working capital and cash-flow finance | Working capital and cash-flow finance | Working capital and cash-flow finance |
| Location Of Investee | UK,W1W5BB | UK,W1W | |
| Geography Of Beneficiaries | PE3 7PD | PE | |
| Sector | Criminal Justice and Public Safety | Criminal Justice and Public Safety | Criminal Justice and Public Safety |
| Type Of Organisation Invested In | Limited Partnership | LLP | Company limited by shares |
| Charity Number | |||
| Company Number | LP013829 | 7240908 |
Now to be fair, this is a complicated investment structure, and investors could enter it via different routes (with a partnership interest or through a corporate feeder structure), so some of the differences are explicable – if not terribly useful when you are trying to draw meaning from this style of data.
What about the analytics? Well there does appear to be some double-counting, but not nearly as much as there might have been: for example, the visualisation (derived from http://data.gov.uk/data/viz/social-investment-and-foundations) appears to show that these three foundations invested £608k and others invested £8.6m in this social impact bond, where I suspect true figures should be more like £379k with others investing around £2.3m. (These latter figures are estimates based on a few assumptions: if you know better, feel free to correct me.)
As a self-confessed data-geek, I fully appreciate that it is hard to get perfectly clean data – particularly on the first go. And I applaud the Cabinet Office and the contributing foundations for the time and effort in getting this far. But if we are to try to draw meaningful information and trends from this data we have to, collectively, do a little bit better.
Update: 4th Feb 2014 - The Cabinet Office wrote to me on 20th Jan 2014 to say that they had removed one of the inconsistencies - updating Barrow Cadbury's reporting of the Total Investment into the Peterborough SIB from £4.2m to £5m. Since the code behind the visualisations on the Data.gov website assume that all deals with the same total investment size are actually the same deal(!) this does address the double count in this instance. It doesn't, however, resolve some of the other issues – for example the revised chart now shows that the three foundations contributed £608k of the £5m in the Peterborough SIB; in reality this is mixing commitments to invest with drawdowns actually made – it is known that these three foundations collectively committed £700k. This could be fixed in the data set too, but the point is not to say that there is one piece of data wrong - this example was picked as one that was easy to validate as there is so much public information on the deal - the point is that we have no idea as to the data quality of the rest of the data set.