Wed 17 Jun 2015 2:50AM

Consistency over time and between collections

LM Lisa Meehan Public Seen by 292

Hi - I think it's mentioned in the consultation document, but all the discussions on what to add to/change about the census got me thinking about consistency. I've done some analysis over time using HES, and changes in questions makes analysis over time more difficult. It can also be quite tricky/time consuming to track down all the information on which questions changed and how they changed. A couple of examples are similar to past changes to the census: changes in the way ethnicity was recorded and changes in home ownership questions to include trusts.

I understand that sometimes the benefits from changing things will outweigh the costs - in the 2 examples given above it seems clear that the change was for the better. But, I just wanted to flag that careful consideration should be given to any changes that mean that variables are not comparable (or are less comparable) over time.

A related issue is comparability across collections- eg, making sure that where HES and census collect the same information that this is done in a consistent way where appropriate.


Andrea(facilitator) Tue 23 Jun 2015 11:38PM

Hi Lisa, thanks for your comments. Yes Census are very aware of the impact of changes to the collection and what this may mean for the variables and they work hard to strike the right balance. Thanks for the reminder and we'll keep your thoughts in mind.


Daniel John Exeter Wed 24 Jun 2015 8:56PM

The harmonisation of data over time is crucial to being able to measure population changes. Is anybody (@Stats or elsewhere) doing some analysis on the impacts changing questions have on measuring different aspects of society?


Andrea(facilitator) Fri 26 Jun 2015 12:16AM

Yes there is a great deal of work that goes into analysing the impacts changing questions have on the data. Firstly before finalising any changes to questions we carry out extensive testing before so we can see the likely impact on the data quality. There has also been analysis done on the impact new content or changed questions have had on the data. Each census we release information about the census variables and this contains information about the comparability with previous censuses, whether there are any significant issues with that variable and other things to be aware of. A couple of examples where the question has changed between censuses are the ethnicity and telecommunications questions. These examples are both from 2006. Here is a link to some more information about the impact the ethnicity change had: ethnicity question and here’s some more information about the change to the telecommunication question.
I hope this helps.