The UK Government's Department for Digital, Culture, Media & Sport (DCMS) is currently running an open call for evidence titled 'The marginal cost restriction on public sector data re-use'. The call for evidence closes at 11:59pm on 8 September 2026 (that's tomorrow).

This post contains my notes on DCMS's call for evidence, much of which I think is misleading. This is not a response to the survey questions or a systematic argument against the DCMS plan.

Short version: the marginal cost limit in the UK's little-used RPSI regulations does not mandate open data – but does provide a convenient excuse for underfunding of public data infrastructure.

Other blog posts you could be reading

I recommend the following posts, which raise concerns about the DCMS plan from various perspectives and are likely to be more digestible than my own post below:

Treasury brain, or the paranoid style in UK data policy

At first glance DCMS's call for evidence may appear to be narrow or technical, but it is best understood as part of an ongoing campaign led by HM Treasury to roll back UK Government policy on open re-use of public sector information (PSI).

The call for evidence should be read in the context of Treasury's Data Valuation Framework, which supplies a permission structure for public bodies to raise commercial income from publicly-owned data assets, as well as recent scoping work by National Archives that seems to be laying the ground for deprecation of the Open Government Licence.

In my view the plan described in DCMS's call for evidence is also likely to reduce access to some information assets – datasets in particular – and in general make it easier for the Government to justify underfunding of the production of public sector information.

You should read what the regulations actually say

The focus of DCMS's call for evidence is a plan to remove or amend a provision in the Re-use of Public Sector Information Regulations 2015 (RPSI).

The 2015 RPSI regulations were originally the UK's transposition of the European Union's 2013 PSI Directive, but have been retained in domestic law following Brexit.

The EU itself has moved on – in 2019 the PSI Directive was updated and renamed the Open Data Directive, with an emphasis on unlocking "high-value datasets" and greater convergence between re-use of PSI and the open data agenda.

Guidance on the UK's RPSI regulations is available from the ICO and from the National Archives.

Strangely, in its call for evidence DCMS does not link to the RPSI regulations or to either set of guidance.

What is marginal cost?

Regulation 15 in RPSI contains numerous caveats but its base statement is that "any charge for re-use must be limited to the marginal costs incurred in respect of the reproduction, provision and dissemination of documents."

Marginal cost is the additional cost incurred when supplying a single copy of a resource, such as a dataset. Where the resource is in a digital format and can be delivered over the internet, the marginal cost for re-use will for practical purposes usually be nil.

The principle of marginal cost is part of the Open Definition i.e. the consensus definition of open data (or open content). In the Open Definition the marginal cost is expressed as "no more than a reasonable one-time reproduction cost".

Marginal cost, then, is a limit on charging for re-use of public sector information.

It is not a "restriction on public sector data re-use" – even the title of DCMS's call for evidence is misleading. Given that data and other digital content is non-rival it should be obvious that, in most circumstances, minimising the cost will encourage rather than restrict re-use of public sector information.

The RPSI regulations do not mandate open data

The central premise of DCMS's call for evidence is flatly false:

"The provision therefore requires most public bodies to make a significant proportion of their valuable data available to businesses for free as Open Data, thereby preventing them from generating revenue through non-Open Data models."

The marginal cost provision is only one of many criteria in the Open Definition that underpins open data.

The RPSI regulations do not mandate or "embed" open data. In fact they do not mention open data or open licensing at all.

RPSI does not require public sector bodies to publish any data proactively, openly or otherwise. If a public sector body receives a request for data or other information, it may impose conditions on re-use, where appropriate through a licence, based on the purpose of re-use stated by the requester.

The conditions must not "unnecessarily restrict" re-use or competition, and the ICO advises in guidance that the easiest way to do that is to use the Open Government Licence. But there is no statutory requirement that the conditions or licence should be open.

Most high-value datasets are already exempt from the marginal cost limit

DCMS's call for evidence acknowledges that some data assets are exempt from the marginal cost provision, but understates the extent of the carve-outs that the UK secured in negotiations prior to the 2013 PSI Directive.

Regulation 15 excuses from the marginal cost limit:

  • any public sector body that is "required to generate revenue to cover a substantial part of its costs relating to the performance of its public task"
  • "documents for which the public sector body making the charge is required to generate sufficient revenue to cover a substantial part of the costs relating to their collection, production, reproduction or dissemination"; and
  • libraries (including university libraries), museums and archives

Other data-rich sectors are excluded from the scope of the RPSI regulations entirely, under regulation 5:

  • public service broadcasters (for most purposes)
  • educational and research establishments including schools and universities; and
  • cultural establishments, other than libraries, museums and archives

In practice that means that nearly all datasets and documents produced by Ordnance Survey, the Met Office, British Geological Survey, HM Land Registry, Registers of Scotland, the UK Hydrographic Office, HMRC's Valuation Office, the BBC, public records held by the National Archives itself, and many other public bodies including a swath of research, education, and cultural institutions, are exempt from the marginal cost limit.

Datasets and other documents are also exempt from the RPSI regulations if a third party owns relevant intellectual property rights in them – even if that third party is another public sector body and even if their intellectual property is already available under an open licence. Given the increased sharing of data between public bodies, and the extensive outsourcing of data work to the private sector, this exemption routinely prevents the effective use of RPSI to unlock public data.

DCMS's call for evidence also fails to mention the Delegation of Authority scheme under which National Archives can permit public sector bodies that are covered by Crown copyright to license specific material on different terms, where they might otherwise be expected to apply the Open Government Licence.

Use of the RPSI regulations is not well understood

My strong impression is that the RPSI regulations are little known and little used, even within the UK's community of open data partisans.

Cabinet Office maintains cross-government statistics on the volume of requests under the FOI Act and the EIRs, but does not collect figures on re-use requests. That itself is indicative. There is nothing in the call for evidence to suggest DCMS has done any research in this area.

In the more than ten years since RPSI was enacted, the Information Commissioner's Office (ICO) has published only 12 decision notices related to enforcement of the regulations. There will be other cases that were resolved informally, but only one decision notice has been published in the past two years.

Only one of those 12 cases involved a dispute over charging for re-use of public sector information, and in that case the ICO found in favour of the public sector body.

As of 7 August 2026, the ICO's active caseload included 4,734 FOI/EIR complaints and only one RPSI complaint. (Some of the "hybrid" cases may also include a re-use element.)

Frankly, if the marginal cost provision in the RPSI regulations was really a significant barrier preventing public bodies from generating much-needed revenue, I would expect considerably more disputes and more involvement from the regulator.

"Open by Default" is an aspiration, not the reality

DCMS's call for evidence overstates the current availability of open public data:

"The Government currently operates an 'Open by Default' policy approach to public sector data. This means that accessible public sector data should be made available for anyone to re-use as Open Data, i.e., with terms that provide predominantly free and unrestricted re-use.

"Information is considered to be accessible if it is required to be communicated to a requester under an information access regime – such as the Freedom of Information Act 2000 – or if it already published. Open Data does not cover data exempt from information access legislation because there are overriding reasons to withhold it, such as personal data that cannot be shared compatibly with data protection legislation, as well as data that is commercially sensitive or linked to national security. For example, identifiable health and social care data cannot be made available as Open Data. Such data must be afforded higher levels of protection, including strict restrictions on access and, in the case of health and social care data, controls to ensure it is never used for marketing or insurance purposes."

There is no way to quantify how much public sector information could be accessed in theory under the FOI Act and related regimes. But, even if rights of access were perfectly observed across the public sector, it is absurd to suggest that information is "accessible" just because it might potentially be disclosed within the standard 20 working days. The FOI Act itself recognises that information is only accessible once the requester has a reasonably certain route to obtain it.

DCMS also seems to have muddled the definition of open data with reasons why a public body might decide not to release data on open terms. Public bodies can in fact proactively release open datasets containing information that they would not be obliged to disclose under access regimes. And definitionally there is no reason why an open dataset cannot contain personal data, including health or social care data (though the OGL would not be the appropriate licence). Data may be re-usable under an open licence even if processing is restricted under data protection law or other statutory regimes.

As a matter of good practice, many public bodies apply an open licence to datasets and content that they publish proactively on their websites. The first incarnation of the UK's Government Digital Service led by example when it added the Open Government Licence to the footer of every page on GOV·UK ("except where otherwise stated"). But public data is not "open by default" in any statutory sense – when seeking unpublished data, and in many other cases, re-users still need to ask for specific permission.

Most open data releases are policy decisions

A significant number of open datasets are published to comply with statutory requirements that are unrelated to the RPSI regulations. Companies House data products and MHCLG's bulk Energy Performance Certificates (EPC) data are made available as outputs from public registers, as are many Environment Agency datasets. But in those and similar cases the underpinning legislation does not mandate open data as such – publication for re-use under an open licence is the result of a policy judgement that this is the simplest way to comply with looser statutory requirements.

There are other fine examples such as DfT's Bus Open Data Service and DESNZ's Fuel Finder, both underpinned by statutory requirements. In those cases the data is collected from the private sector and the public purposes of the schemes could not plausibly be delivered on a non-open model.

But the vast majority of open public data has been released either in support of a domain-specific policy agenda or in general recognition that publication of open data is good practice as a driver for social or economic benefits.

We can see this particularly in local government, where functions are broadly the same from council to council but there is a wide variety in the range of available open data. If compliance with the RPSI regulations was a mandate for publication of open data, we would see much greater consistency in the available datasets.

Charging for re-use will also reduce access to public sector information

In its call for evidence, DCMS is keen to emphasise that any potential reform should not alter freedom of information law or change public bodies' existing compliance obligations under that legislation.

But in practice, making it easier for public bodies to charge above marginal cost for re-use is likely to also undermine existing rights of access to any affected information assets.

In principle, public bodies can disclose chargeable information under the FOI Act or EIRs without granting permission for re-use. There are a few examples of datasets like that, notably the postcode list for Northern Ireland which is publicly accessable in ONS data products but only available for commercial re-use following payment of a fee.

But normally, if a public body is charging for re-use of an information asset, it will try to withhold that information as exempt – on the basis that disclosure would be likely to prejudice its commercial interests due to the risk of unlicensed sharing or other re-use.

If the Government decides to repeal the RPSI regulations or remove the marginal cost limit entirely, that may mean that we fall back on the re-use provisions in section 11A of the FOI Act and the related Freedom of Information (Release of Datasets for Re-use) (Fees) Regulations 2013. Currently section 11A has limited effect because it excludes copyrighted works to which RPSI applies.

Even if there is technically no alteration to FOI law, it will be difficult to remove or amend the marginal cost provision in RPSI without a practical effect on access rights.

Other funding models are available

We are naturally sympathetic to the plight of budget holders in public bodies who are forced to bid, often unsucessfully, for the central funding that they need to collect and maintain data for the purpose of delivering public services. DCMS is counting on that in its call for evidence.

But at the policy level, adequate funding (or not) of public data assets is a political choice. There is a perennial problem in data policy that we speak about data only in generalities, but the business case for funding any particular dataset or data programme is always fact-specific and often requires subject matter expertise that central planners lack.

The assumption in DCMS's call for evidence is that re-users of public sector data have an interest or stake in the ongoing maintainence of that data for its original purpose. But that will not necessarily be the case – sometimes re-users have no other alternative source for data they need, because the relevant public bodies have established an effective monopoly. If we cross the streams, and make the funding of data for public purposes dependent on revenue from re-use, the tendency will be to optimise those datasets or data programmes for the purposes of re-users and away from the public task.

He who pays the piper calls the tune. Would Elon Musk like to sponsor the production of Home Office's migration data? Let's ask.

The principled argument for open public data is only that data that is already funded for public purposes should additionally be available for re-use. Advocates for open data often have interests in particular domains and may have views on whether datasets should be funded in the first place – but it does not necessarily follow that they must. It is reasonable to take the view that some existing public datasets have reached the end of their usefulness and should no longer be maintained or funded. Continuing to maintain public data only for purposes of re-use should fall outside the public task.

That does not mean public bodies should not produce additional data assets to meet a demand from re-users. If a public body has spare capacity, or is the only source capable of producing a particular data asset, it is perfectly reasonable for that body to offer bespoke data outputs and charge the first re-user for at least the costs of production. If the data is suitable, it should be licensed openly – but provided there is no royalty in addition, and that subsequent re-users are charged no more than marginal cost, that model enables a public body to raise revenue from re-use without any conflict with open principles. ONS operates just such a revenue model with its ad hoc statistics.

In general though there is no obvious reason why funding of public data infrastructure should be ring-fenced or linked to re-use. As data is a non-rival good, charging cannot be justified by the need to manage consumption in the way that it might be for other public resources. Charging for public data only creates an artificial scarcity, reducing (and often eliminating) potential for re-use to the detriment of the economy and society.

We have been here before

In 1980 a review led by Sir Derek Rayner, Managing Director of Marks & Spencer, advised the Thatcher government that official statistics should be collected to meet primarily the needs of government, and that non-government users should be charged for statistical data. The review led to deep cuts in funding for official statistics.

The damage to public data from implementation of the Rayner Review took decades to unwind, but by 1990 we can see the emerging shoots of a more open approach. This is from a report produced by the Royal Statistical Society's Working Party on Official Statistics in the UK:

"Many statistical inquiries are financed through public money and depend on public co-operation for the information provided. It follows that, subject to constraints to protect the confidentiality of individual respondents, the public are rightful consumers of the statistical product. Analyses should be in the public domain and basic data made available for further analyses."

But the current Government is in danger of forgetting the lessons of that earlier debacle.

Funding for ONS statistics has been reduced on real terms. Earlier this year ONS announced wide-ranging cuts to statistical outputs, reducing the national evidence base in areas such as health, wellbeing, and climate change. The UK has been without a permanent National Statistician since May 2025.

This where the DCMS plan comes in. One for the advantages of encouraging public bodies to charge above marginal cost for re-use of data is that it shifts the blame for underfunding, away from policymakers in the centre of government. The onus is on delivery departments and other public bodies to make up any shortfall by generating revenue from the market. If they fail to do so they're just not being entrepreneurial enough, and maybe the data wasn't that important in the first place.

Is the current Labour government prepared to roll back progress achieved by the previous Labour government?

You might get the impression from DCMS's call for evidence that UK open data policy began in 2012, with the Cameron government's publication of its Open Data White Paper.

But open data was an agenda in civil society and academia, and internationally, well before that. It was Gordon Brown's Labour ministry that first embraced open data (in its modern form) as government policy.

Transport for London's open data offer – still the go-to case study on realisation of benefits from open public data – was launched in 2007 with the main aim of stimulating the app economy. The first significant suite of open geographic products was released by Ordnance Survey in April 2010, shortly before the general election. While the first version of the UK's Open Government Licence was published later in 2010, that was the result of development work under Labour to replace the pre-existing PSI Click-Use licence.

Although open data was a focus of policy attention mainly in the years from 2010 to 2015, the Tory-led governments of those years arguably failed to build on the promise of the early Labour initiatives, particularly with regard to high-value datasets and foundational data.

Development of open data initiatives has continued in many parts of the UK's public sector, but cross-government data policy has languished over the past ten years. It would be disappointing if the current Labour government decided to roll back the agenda of previous governments, rather than reclaiming and extending open data policy as other countries continue to do.