More than one way to open some data: government owned and government influenced

More than one way to open some data: government owned and government influenced

Guest post by Tim Davies & Jeni Tennison (Open Data Institute)

As work is kicking off on the development of the second Open Data Charter Sector Package – focussed on Agriculture – we’ve been exploring the boundaries of the data that each package should cover.

In the draft template for sector packages we describe them as so:
“…a curated resource, providing guidance on how to make open data available, and to support the use of open data, in a particular sector.”

This is fairly broad. However, an implicit assumption in most conversations about the packages so far has been that focus is on government owned data, understood as the datasets which governments collect and hold directly. But often the dataset that most matter, and that could have the most impact if they were open, do not belong to governments.

Should these datasets be within a scope of a sector package?
The arguments against including non-government datasets might focus on the need to make sector packages actionable: only including those datasets which government can directly decide to publish as open data. However, the counter arguments might include that:

  • Datasets which are government owned in one country, are held in the private sector in others. For example, there is varied picture worldwide on ownership of national address registers. Some are held by central government, some by local agencies, and others by private providers. However, regardless of ownership, this is still important data for policy makers to be aware of and they should be thinking about how to increase accessibility. Government may have powers to mandate open data publication as part of giving licenses to run a register, or may be able to negotiate with private providers to secure access to data which can then be shared as open data.
  • Government can regulate to require that other people publish open data. For example, through targeted transparency policies governments may require private actors to make particular data available for the public good. To date, targeted transparency has tended to focus on localised analogue disclosures – for example through food labelling, or food safety scores-on-the-doors, but there are also cases of digital disclosure requirements, including the UK Competition & Markets Authority requiring banks to publish open data about the location of bank branches and ATMs, and details about the products that they offer.
  • Government has a role in creating an enabling environment for communities to create and share data. Whether through setting standards, or providing core technical infrastructure for validating, publishing, harvesting, aggregating, visualising or otherwise processing data, government can incentivise and facilitate open data publication.

Taken seriously, these arguments would suggest that we include government influenced data as well as government owned data within the scope of sector packages.
This doesn’t mean every dataset should be included in a package: they still need to pass the test of having demonstrated use-cases and demand. However, it would mean that we’re not just looking at evidence of publication on government data portals to show that data can be provided, but would involve looking for cases where governments have adopted a wider range of strategies to make data available.
What do you think? Should sector packages include government influenced, as well as government owned, datasets?