DWP changes “open data” tool after Help Me Investigate raises concerns

The Department for Work and Pensions has promised a series of improvements to an online tool promoted as an example of “open data” after concerns were raised by Help Me Investigate.

Stat-Xplore was one of the case studies described in a DWP report on open data earlier this year. But on launch the site failed to meet basic open data principles listed by the Government’s own Public Sector Transparency Board (PDF), making it difficult for citizens and developers to use the data.

One of the main issues was the site’s use of cookies, which meant it was not possible for computers to directly access datasets via a specific URL.

The DWP have now announced forthcoming improvements including:

  • “the ability to enter a URL to directly link to a Stat-Xplore database, a system-generated table (i.e. the ‘ready-made’ tables), or a user-generated table (i.e. one created and saved by yourself).
  • “the ability to utilise a RESTful web service to access data and metadata directly in SDMX format, allowing you to develop your own tool to query the data.
  • “we are investigating the possibility of releasing a single data ‘cube’ outside of Stat-Xplore which users could download and analyse using their own tools.”
Andrew Mackenzie, a member of Cabinet Office advisory committee the Open Data User Group said the product was an improvement over the previous ways the department published statistics. However, he added:
“Publishing larger data sets and offering more sophisticated ways of viewing information does not qualify as open data if the data can only be used within one product. To realise the benefits of open data the DWP dataset has to be available as a complete dataset without restriction on re-use. Then other providers can generate new services by combining datasets or offering alternative ways of exploring the data.
“I welcome the news that DWP statisticians are considering publishing the dataset for download so that users can explore and analyse it using their own tools. Subject to confirmation that the dataset is to be published under the Open Government Licence, this would create a significant research resource.
“There are genuine concerns about publishing information which could identify individuals. So, where a full dataset is made available there is a need to develop robust anonymisation methods. The principles are well understood by statisticians, but the discussion about how to implement these as technology needs to include civil society organisations and security specialists.”
Comments on the service and the new developments are being invited on the discussion forum on StatsUserNet.
Meanwhile, the Public Administration Select Committee (PASC) has begun an inquiry into statistics and open data in Government which runs until 3rd September. People can submit evidence here.