Around the world today, organisations and communities are recognising the 26th International Day for the Elimination of Violence Against Women. This is a moment to reflect on one of the most prevalent and pervasive human rights violations in the world – but it’s also a call to action.

As one small action in that continued effort, we’ve been working with the End Violence Against Women Coalition (EVAW) this year, to explore how something like our Local Intelligence Hub could help their members organise for change.

Little did we expect that, in building a prototype Data Hub for them to explore their needs, we’d discover a gaping hole in data collection about the safety of girls at schools in 16% of local authorities.

But first, what were we aiming to achieve?

In a post last month, I shared some of the goals of this work – such as using data to galvanise support from MPs, to monitor patterns that official bodies might miss, and to help EVAW’s members make the case for increased local funding to address violence against women and girls (VAWG).

I also shared how we were using the batch request tools in WhatDoTheyKnow Pro (our advanced Freedom Of Information service) to generate new public data on VAWG prevalence in schools.

And, of course, all of this work builds on top of the Local Intelligence Hub we designed and built with The Climate Coalition and Green Alliance – which has already proved its worth as a tool for community organising and public affairs, including through events like this Summer’s #ActNowChangeForever Mass Lobby, and The Climate Coalition’s Great Big Green Week.

Now it’s time for an update – how did we get on?

A replicable pipeline of brand new VAWG data

When we built the Local Intelligence Hub with The Climate Coalition (TCC), much of the data we included was already publicly available: MP information from Parliament, demographic data from the ONS, public opinion data shared by polling companies. Combined with TCC member organisations’ own data on their local support and activities, the Hub was able to present a nuanced picture of climate and nature are being protected across the whole country.

We knew we faced a different challenge with the VAWG data hub. As I explained last month, public data in this space is often incomplete, or missing entirely. We wanted to use this as an opportunity to test how WhatDoTheyKnow and the Local Intelligence Hub could work together to generate and then publish brand new datasets on VAWG prevalence or activity, made public through FOI requests to local authorities and policing bodies.

We chose school safeguarding referral figures as a suitably challenging example that was also indicative of levels of risk to children. When school staff fear a child may be in danger in any way, they are meant to refer it to the safeguarding team at their local authority. The UK government collects some information about these referrals as part of its Children In Need census, but the definition of a “child in need” is somewhat open to interpretation, and we and EVAW both suspected that, as a result, the official data was only telling part of the story. The census also only covers local authorities in England, leaving Scotland and Wales to collect their own, incompatible data (the CRCS census in Wales, and Children’s Services Plans in Scotland).

With the help of the WhatDoTheyKnow volunteers, we drafted an FOI request to be sent to every UK local authority with a responsibility for education, asking for three things:

1. The total number of safeguarding referrals made to them, by schools in their area – this is data that technically should be collected by the CIN census for English authorities, but we suspect is not
2. Any sort of categorical breakdown they held about those referrals, such as a breakdown of the genders of the children involved – this doesn’t currently appear in any public dataset that we know of
3. The total number of schoolchildren in their area

You can browse the requests and responses on WhatDoTheyKnow. Here are some key things we learned through the process:

No matter how much you research your request, something will slip through

Our background research and even our first pilot requests failed to reveal that the total number of schoolchildren is something that’s already published for England, Scotland, and Wales. Thankfully, many authorities simply pointed us to this data (with a “Section 21” refusal – “information already accessible”), but others continued to provide the data for each year we requested. Had we known in advance that the data was already available to us, we could have left it out of our requests to English, Scottish and Welsh authorities. We can only hope, since this is such basic information, the authorities who did go on to provide the data to us didn’t spend too long gathering it.

You will receive information in every format imaginable, and your data extraction process needs to handle that

We asked for responses to be provided in a “re-usable, machine-readable format” if the authority deemed the information to meet the FOI Act definition of a ‘dataset’. We think, in reality, very few of the authorities held this data in a format structured enough to count as a ‘dataset’, but a few did send over their data in spreadsheet format, which was nice to see! Others, however, sent us tables in Word documents, in PDFs, SharePoint links, even ASCII-art tables in raw email text.

We also knew authorities might hold the information by calendar, academic, or financial reporting year, so we gave them the freedom to provide it to us in whichever scheme they had. Unsurprisingly, we received responses across all three (57% calendar year, 35% financial year, 8% academic year).

Happily, the crowdsourcing interface in WhatDoTheyKnow Projects enabled us to make relatively quick work of extracting the data we needed, but we were ultimately only able to extract a fraction of the information some authorities provided and we found that some of the interpretation of the responses (ie: “is this a financial year, or an academic year?”) heavily relied on human intuition, which means we’ll need to think carefully about the way we structure future requests, if we want to process the data through any sort of automated pipeline.

Complex requests are a risk

The more information you request, the more useful it might be to you, but the more you risk the public authority refusing to answer it on “Section 12” cost grounds. WhatDoTheyKnow’s advice is to keep your request as short and focused as possible. But we knew that historical data, across a few metrics and a few years, would be most useful to the VAWG Data Hub’s users, so we asked for as much as we felt we could justify – and it mostly paid off.

70% of responses to our batch request contained both key pieces of data we wanted (the total referrals for multiple years, and the gender breakdown). Another 7% contained just the yearly totals, without any gender breakdown.

7% are still awaiting a response, even now, over a month after the statutory deadline. And 6% of authorities said they didn’t hold the information at all (because they, surprisingly, don’t record the referrals they receive). Which leaves 10% who refused our request on cost grounds. If our request had been simpler, this number of refusals would likely have been smaller.

However, this result is in itself interesting: at least 16% of local authorities responsible for handling safeguarding referrals either don’t record them, or record them in such a way that it would take more than 18 hours of officer time to report how many they received in a given year, or how many relate to girls.

If the government is serious about halving violence against women and girls within a decade, this is precisely the sort of data local authorities will need at their fingertips, in order to monitor progress and allocate resources. The fact that it’s effectively inaccessible to 16% of them right now is a worry.

Combining data for new patterns and new questions

Remember how I mentioned we were adapting the Local Intelligence Hub for EVAW’s needs?

With our FOI data extracted through WhatDoTheyKnow, we were able to very quickly load it into a prototype VAWG Data Hub. Alongside it, we loaded in a whole new area type to filter by—“Policing areas” or Police & Crime Commissioners—as well as some examples of crime prevalence data (the number of reported VAWG-related incidents, by policing area) and public policy guidelines data (the Council Of Europe’s recommended minimums for VAWG service provision).

Thanks to my colleague Alex’s improvements to our TheyWorkForYou Votes infrastructure, we were also able to make quick work of importing VAWG-related MP data into this new hub – including VAWG-related parliamentary groups that MPs might be sitting on, or relevant votes and motions they’d supported.

Plus, of course, there was all the usual MP and area data that campaigners and public affairs teams have already found so useful on the Local Intelligence Hub – things like election results, public attitudes polling, and income and poverty indicators.

Data in action

With the data in place, it was possible for us to give EVAW’s member organisations a demonstration of how they could use a data hub like this as part of their campaigning, fundraising, and policy influencing work. For example, to find the council areas with the most school safeguarding referrals for girls and also the highest overall deprivation:

Or to find the MPs with the strongest support for VAWG prevention, but in constituencies with high VAWG prevalence:

All of the data we demonstrated this Autumn is still a work in progress, but it was reassuring to see almost 70% of members on a recent demo call saying that a VAWG data hub like this would “definitely” be useful to them in their day-to-day work.

We look forward to honing the VAWG Data Hub further with EVAW and their members, to make sure we’re asking the right questions, and presenting an accurate picture of the VAWG landscape.

Header image: Khyati Trehan, for Google Deepmind – Pexels Free License.