2011 UK Census

Disclosure Control is a necessary evil. We have an obligation to protect people’s information and if we didn’t they wouldn’t complete their Census forms. Applying disclosure control though means that we are always going to lose something - but there are choices and trade-offs to be made. The survey asked which SDC features users consider to be most important for example, maintaining additivity in tables, consistency across tables, counts that are as near as possible to true counts, a method that is easy to understand, being able to take account of the effects of the method in analysis and the relative importance of these features. Users from different communities may have different priorities and we will be reporting on this and giving feedback to the methodologists who are evaluating the short-listed SDC methods in terms of the protection they give and the effects on the data -a risk/utility continuum.

We are aware that users were very unhappy about the effects of the late addition of Small Cell Adjustment in the 2001 Census, particularly to tables with low counts such as the OA level Origin/destination tables which resulted in users reporting that such outputs were “unfit for purpose”. Whichever SDC method is applied the data quality of sparse tables are going to be disproportionately affected. For these types of tables we are considering the option of applying lower levels of SDC and making the output available to users only under special licence or accessed in data labs with outputs checked prior to removal. For example, some outputs from the Origin/Destination tables could be made available publicly at Super Output Area (SOA)/Data Zone level, made available under licence at Output Area (OA) level, and allow users interested in specific journeys access to detailed micro-data in a safe setting.

What would you feel about access arrangements such as this?

I was wondering why this is an either \ or choice? Why can't the users be allowed to choose the most appropriate for themselves.

If some are prepared for the blurring of data because of SDC at the lowest levels in order to get ease of access fine, whereas others may want the greater accuracy and are prepared to use restricted access, then also fine. I believe that some users will only be using the OA to aggregate to their own higher geographies and so are happy with the first option, whilst others will want to look at the detail and so the latter will satisfy them.

I don’t see that this would undermine SDC and would give users the choice to meet their needs.

If we are not careful there is a possibility with all of these consultations that we will be fixing the problems of the last census. The census agencies hear from those who were dissatisfied, but those who were happy will not be commenting. I would not be at all surprised if this was one of these situations.

We would welcome any ONS proposal to release potentially disclosive data by means of a licence. We have been arguing for such a scheme for some time. Together with clear rules on what may and may not be published, this would enable us to obtain much more accurate data and make the whole SDC process much more transparent.

The idea of access via a 'Data Lab' is a non-starter - inconvenient and slow, and completely absurd in the 21st Century!