WhatDoTheyKnow and Central Government FOI:

2016 Report

Other formats: Plain text, PDF, EPUB.

WhatDoTheyKnow is a public FOI platform (run by mySociety and a group of volunteers) that simplifies the process of sending an FOI request to a public authority and makes the result public so the information can be of use to more people. More generally it is an implementation of mySociety's Alaveteli software — a framework for building FOI sites. There are FOI sites using Alaveteli in 25 jurisdictions around the world.

Every quarter the Cabinet Office releases Freedom of Information statistics for a collection of central government ministries, departments and agencies. This provides a benchmark for understanding how requests made from WhatDoTheyKnow relate to FOI requests made through other routes. We ran several blog posts about this in 2010 but haven't released any comparisons in recent years.

The FOI statistical releases contain quite a lot of information, so to help us explore the data, we've built a website that enables the Cabinet Office statistics to be compared to requests sent through WhatDoTheyKnow - that site can be found here.

Change Over Time

While FOI requests made to central government in total have decreased since their 2013 peak (and have remained relatively flat for several years), FOI requests made through WhatDoTheyKnow have consistently increased each year — but only a small proportion of these go to bodies that are audited by the Cabinet Office.

Chart 1 - WhatDoTheyKnow FOI Requests by Year

Looking just at audited bodies, there is less of a consistent overall increase in the volume of FOI requests. However, the percentage of FOI requests coming from WhatDoTheyKnow has increased every year — from 6.7% in 2010 to 17.14% in 2016. Looking just at ministerial departments, the growth is from 9.1% to 22.28%. So for a rule of thumb: roughly one in ten of our requests go to ministerial departments, and those make up one in five of the requests they receive.

Chart 2 - Requests made to Audited Public Bodies

2016 Statistics

In 2016, out of 45,415 requests to central government, WhatDoTheyKnow accounted for 7,786 (17.14%). On the other hand, requests sent to central government represent a very low percentage of FOI requests made through the site. 88.51% of WhatDoTheyKnow requests went elsewhere (for example, to local authorities).

This percentage varies significantly between government departments. For instance the National Archives received 2,922 requests in 2016 — but only 19 came from WhatDoTheyKnow. While the Health and Safety Executive receives the largest number of FOI requests in central government (4,661), very few (2%) come from WhatDoTheyKnow. On the other hand, the Department for Work and Pensions received the second highest number of requests (4,285) —and a sizeable proportion of these arrived from WhatDoTheyKnow (25%). The case of the Home Office where WhatDoTheyKnow appears to — but probably doesn't — account for 60% of FOI requests is discussed further down.

What this suggests is that different departments are receiving different kinds of requests — and certain kinds of requests are more likely to be sent through WhatDoTheyKnow than others. That very few requests are made to the Health and Safety Executive from WhatDoTheyKnow might reflect that a greater proportion of requests are being made by businesses, journalists or researchers who would prefer to make their requests in private, or who have internal methods of conducting and tracking FOI requests.

There will similarly be different breakdowns of requester types between sectors: estimates of top four requester types to central and local government show that while both receive roughly equal breakdowns of requests from the general public, local government receives more requests from journalists and business and fewer from academics and researchers.1 This presents a problem in using a sample of requests from one area to evaluate the use of FOI in another.

Notes (hide):

1: Worthy, B. and Hazell, R. (2017) ‘Disruptive, dynamic and democratic? Ten years of FOI in the UK', Parliamentary Affairs, 70(1), pp. 22–42. doi: 10.1093/pa/gsv069., p. 27

Comparing requests

One of the key questions when looking at a sample of information like the WhatDoTheyKnow dataset is how representative it is of the whole. We know for instance that certain kinds of requesters who want privacy (such as journalists) will tend not to use the service (although the upcoming WhatDoTheyKnow Pro service addresses this issue by allowing publication to be delayed) — so our dataset is missing certain classes of request that will make it unrepresentative of what is received. Knowing how much this is skewing the composition of our dataset is helpful, so we've explored a basic test to identify whether the success rate of requests is noticeably different between the two datasets.

In the central government statistics each body provides a success rate in terms of how many requests were met in full. This has some problems as an objective measure because of how some requesters might combine questions while others might send each question as an individual request — but it does give us a basis for comparison between over time. The Institute of Government found that from 2010 to 2016 the percentage of FOI requests that were fully withheld grew from 25% to 40%.2

Notes (hide):

2: Institute for Government (2017), Whitehall Monitor - Communicating transparently, avalaible at: https://www.instituteforgovernment.org.uk/publication/whitehall-monitor-2017/communicating-transparently

To compare WhatDoTheyKnow requests and the official statistics, there is an added complication in that the success status of WhatDoTheyKnow requests is self-reported by the requester. Sometimes this isn't updated and so the database has many requests that are either successful or unsuccessful awaiting classification (requesters may also classify inconsistently, or set the wrong classification on their request). To provide an accurate measure of success these reports with uncategorised statuses have to be adjusted. The methodology of that approach is explored here, but in short involves an automated categorization process with aggregate correction of errors for each category. The end result is that in 2016 WhatDoTheyKnow requests to central government had a success rate of 38.5% compared to an official rate of 34.2%. Over the entire dataset the difference is much smaller — with WhatDoTheyKnow being 1% less successful than the official statistics. This doesn't necessarily represent a change as the methodology used to adjust success rates of uncategorised requests is likely to be less accurate for historical data. Similarly it is doubtful that much should be made of the fact that WhatDoTheyKnow has a slightly higher success rate, given the method required to arrive at the result.

What it does show is that WhatDoTheyKnow does not seem to precisely track central government success; it also suggests the discrepancy isn't extreme in the most recent data and there is no obvious platform effect of WhatDoTheyKnow either leading to more or less successful requests. That this comparison method can not be used to compare success rates in individual departments without further manual categorization limits its usefulness for cross-comparison.

Central vs local government

While understanding FOI requests to central government is clearly important in understanding FOI's effect on the running of the country, because the numbers are well documented, there is a risk of looking at these statistics as representative of the picture of FOI in the UK.

There is also a sense that audits of FOI requests hide the complexity of requests. For instance, a single request may contain one or many actual questions — with experienced requesters including substitute questions for speedier replies in the event that one set of data isn't available. Comparisons of counts over time or between areas will not pick up these complexities.

That said, where there are very large differences in number we can certainly draw distinctions. A Constitution Unit Survey found that in 2010, 197,000 requests were sent to local authorities and this is the latest year we have information for.3 In that year there were 43,000 requests made to central government — FOI requests made outside of central government are the majority of FOI requests.

Notes (hide):

3: Worthy, B. and Hazell, R. (2017) ‘Disruptive, dynamic and democratic? Ten years of FOI in the UK', Parliamentary Affairs, 70(1), pp. 22–42. doi: 10.1093/pa/gsv069., p. 26

In 2016 there were only 45,415 requests to central government agencies. Looking at where requests were made in WhatDoTheyKnow, 88.5% of FOI requests sent in 2016 went to public bodies that were not audited as part of the central government.

Examining the type of public bodies WhatDoTheyKnow sends requests to shows what a small (and potentially unrepresentative) proportion of requests central government represents — less than half those sent to local and regional governments. Local Authorities and Central Authorities combined represent 63% of WhatDoTheyKnow requests in 2016 — but this leaves 37% going to other authority types.

Public Body TypeFOI RequestsAdjusted Success Rate %Raw Success Rate %

Local and regional government

31406

57.99999999999999

46

Health

11917

65

49

Audited central government

7802

37

21

Education

4689

48

35

Emergency services

4364

45

35

Unaudited central government

3369

40

26

Other

1513

40

24

Transport and infrastructure

1411

49

37

Law making bodies, the courts and the legal system

564

31

19

Media and culture

335

43

27

Environment and agriculture

223

56.00000000000001

40

Military and security services

146

51

30

The slight discrepancy between audited central government here and on the website results from when the cabinet office has at some point released information for a body (so counts as an ‘audited' body for our purposes), but then in another year has not. In those years the absent statistic is not processed or matched against its WhatDoTheyKnow count for the website — but is counted here. For the 2016 data a small number of FOI requests to the Privy Council Office and the Government Equalities Office cause the problem.

Notably the success rate for local and regional government is much higher than audited central government — over 20% higher in both raw and adjusted success rate. We lack real world information to know how representative this is of all requests to local government, but certainly it is highly suggestive that the picture of FOI looks very different in different organisation types — and conclusions on government success drawn just from FOI information might miss this part of the picture.

The UK Open Government National Action Plan 2016-18 includes a plan for a new Code of Practice that addresses the lack of statistics for most public bodies:

We will also issue guidance in the revised Section 45 Code of Practice to set a standard that public authorities with 100 full time equivalent employees or more should publish statistics on their FOI performance, to better hold public authorities to account.4

As of July 2017, the plan was for a consultation to begin on this revised Code of Practice ‘shortly'.5

Notes (hide):

5: 2016-18 Open Government Action Plan: July 2017 Commitment Progress Updates (2017), Avaliable at: http://www.opengovernment.org.uk/resource/2016-18-open-government-action-plan-july-2017-commitment-progress-updates/

In the meantime, to redress this gap in the analysis, one of our future research goals is to gather comparative material for public bodies outside central government to understand the representativeness of the WhatDoTheyKnow data.

Home Office and immigration

While it has been an outlier since 2014, the percentage of requests WhatDoTheyKnow seems to send to the Home Office is worth further explanation. On a strict ‘requests we sent' out of ‘requests they received' percentage WhatDoTheyKnow accounts for 61.9% of FOI requests to the Home Office. This is a massive increase over the next highest (27.7% to the Department of Education) — so this number needs to be looked at carefully. There are two possible explanations for this number:

  • People are using WhatDoTheyKnow to ask fundamentally different questions to the Home Office than other channels of sending FOI.
  • The Home Office is not recording all messages sent through WhatDoTheyKnow as FOI requests.

These explanations overlap, and in likelihood the number is probably explained by a bit of both. Looking at requests to the Home Office what is immediately apparent is that WhatDoTheyKnow is used as an avenue into government for queries concerning people's personal immigration status or to chase up correspondence and documents. These often aren't answered directly (although sometimes are) but tend to get replied to with a message redirecting someone to the correct channel or procedure. It is plausible that as these are being redirected internally, these may not be being counted as FOI requests (in fact, you could certainly argue it would be inappropriate to class these personal information requests as FOI requests, as they would legally fall under the Data Protection Act, rather than the FOIA) and so our count of FOI requests we send the Home Office is higher than they would consider it to be. This behaviour would explain why the Home Office stands out in our statistics.

It is worth noting we discourage people from using the site in this way — when sending a message to the Home Office users are prompted in large bold writing: "Do NOT use this page to contact the Home Office about your passport, ILR or other documents, or about your application for asylum." This is alongside general warnings not to request information about yourself. When administrators are notified of requests that are not FOI enquiries, these may be removed from public view on the website.

That people continue to do this suggests this is quite a hard behaviour to discourage. Knowledge about immigration status is not just a nice-to-have, and it is understandable that people are ignoring our large warnings if there is a possibility of clarity or progress in stabilising their legal presence in the country. Immigrants through language barriers and lack of experience interacting with the British state will naturally be those least able to navigate state services. WhatDoTheyKnow is accidentally quite an accessible platform for people to ask the state about this kind of information. That people do receive (if not personal) then at least helpful answers may be encouraging others to use the service in the same way, either through word of mouth or by increased googleability.

It is worth noting that this behaviour mirrors the use of FOI laws in India, where 18% of FOIs from rural areas and 15% for urban areas were found to be chasing answers to questions that should have been answered.6 Worthy, John and Volloni (2017)'s finding that when asking a question to parish councils, framing it as an FOI request rather than just a question doubled the success rate (from 5% to 10%).7 An under-appreciated role of FOI is to encourage (and demand) the regular and efficient working of normal state services.

Notes (hide):

6: RTI Assessment and Advocacy Group (RaaG), Samya- Centre for Equity Studies (2014), Peoples' Monitoring of the RTI Regime in India

7: Worthy, B., John, P. and Vannoni, M. (2017) ‘Transparency at the parish pump: A field experiment to measure the effectiveness of Freedom of Information requests in England', Journal of Public Administration Research and Theory, 27(3), pp. 485–500. doi: 10.1093/jopart/muw063.

Future research

This report and website represent early steps in understanding how data from WhatDoTheyKnow can be generalized to provide insight into the functioning of FOI in the UK.

The main initial lesson is that we have no good general dataset for FOI in the UK, and what partial data we have invites us to be cautious in any analysis. WhatDoTheyKnow and previous survey data suggest that the scope of FOI in the UK means that the majority of FOI activity is not covered by the Cabinet Office audit. More than that, the information that we have suggests that measures of success are not generalisable from this audit either. For requests made in WhatDoTheyKnow, the success rate is higher for most other types of public body. Future survey work will be required to determine how representative these results are of the true picture of FOI in the UK.

Related research

Footnotes

1 Worthy, B. and Hazell, R. (2017) ‘Disruptive, dynamic and democratic? Ten years of FOI in the UK', Parliamentary Affairs, 70(1), pp. 22–42. doi: 10.1093/pa/gsv069., p. 27

2 Institute for Government (2017), Whitehall Monitor - Communicating transparently, avalaible at: https://www.instituteforgovernment.org.uk/publication/whitehall-monitor-2017/communicating-transparently

3 Worthy, B. and Hazell, R. (2017) ‘Disruptive, dynamic and democratic? Ten years of FOI in the UK', Parliamentary Affairs, 70(1), pp. 22–42. doi: 10.1093/pa/gsv069., p. 26

4 UK Open Government National Action Plan 2016-18 (2017) , Avaliable at: https://www.gov.uk/government/publications/uk-open-government-national-action-plan-2016-18/uk-open-government-national-action-plan-2016-18

5 2016-18 Open Government Action Plan: July 2017 Commitment Progress Updates (2017), Avaliable at: http://www.opengovernment.org.uk/resource/2016-18-open-government-action-plan-july-2017-commitment-progress-updates/

6 RTI Assessment and Advocacy Group (RaaG), Samya- Centre for Equity Studies (2014), Peoples' Monitoring of the RTI Regime in India

7 Worthy, B., John, P. and Vannoni, M. (2017) ‘Transparency at the parish pump: A field experiment to measure the effectiveness of Freedom of Information requests in England', Journal of Public Administration Research and Theory, 27(3), pp. 485–500. doi: 10.1093/jopart/muw063.