GSS Logo

High quality statistics, analysis and advice to help Britain make better decisions.

2021 Census – a new approach to counting rooms and bedrooms

Posted by


The Office for National Statistics released its first assessment of an alternative method to estimate rooms and bedrooms using Valuation Office Agency (VOA) data on 28 June 2017.

Based on the initial analysis, it’s ONS’ view that with some further development, the VOA data are of sufficient statistical quality to replace one or both of the questions on rooms and bedrooms in the 2021 Census.

We’ve been inviting users to explore the data and take part in a consultation to tell us what impact this could have. The consultation is open until midnight on 20 September 2017.

So what is Valuation Office Agency data?

The VOA data holds information on properties including number of rooms and bedrooms for addresses across England and Wales. The information VOA collects and holds about domestic properties is in support of their statutory functions for valuation and maintenance of Council Tax lists under Local Government Finance Act 1992.The data are checked and updated whenever VOA visits the property, or through its communications with billing authorities, builders, developers and the public.

As with many administrative data sources, there are definitional, coverage, timing and data collection differences between the 2011 Census and VOA data. To understand the extent of these differences, we’ve compared the VOA data from 2011 for rooms and bedrooms against those produced from the 2011 Census.

How do we know it’s of sufficient statistical quality?

For the number of bedrooms, 76% of records were the same in the VOA data as they were in the 2011 Census for England and Wales. This high agreement rate is attributable to the closely aligned definitions of a bedroom between the VOA and census. However, the agreement rate for number of bedrooms differs somewhat between each local authority.

To understand why they differ we’ve looked at local authorities at output area level, an area that covers between 40 and 250 households. We’ve looked at three factors that may be leading to differing agreement levels: over-crowding and under occupancy, accommodation type, and census response rates.

We’ve focused on two local authorities to demonstrate our findings: Harlow and Haringey. Harlow has one of the highest agreement rates between VOA and census data at 85%.  Haringey, meanwhile, has one of the lowest at 67%.

Figure 1: Direct agreement rates of number bedrooms between 2011 Census and VOA by output area, addresses in Haringey

Figure 2: Direct agreement rates of number of bedrooms between 2011 Census and VOA by output area, addresses in Harlow

How does overcrowding or under occupancy create differences in the number of bedrooms recorded?

A possible reason for the difference between levels of direct agreement between the two local authorities could be overcrowding. In 2011 levels of overcrowding in Haringey were estimated at 16%, whilst in Harlow the estimated level was 6%. This could affect agreement rates because of the different definition of a bedroom between VOA and census.

In the 2011 Census a bedroom is defined as any room intended to be used as a bedroom when the property was built, or any room that has been permanently converted for use as a bedroom.

In the VOA data, bedrooms include rooms that are built as bedrooms, even if they aren’t being used as such. It also excludes rooms incapable of comfortably holding a single bed (approximately anything less than 2m x 2m).

This definitional difference means that when levels of overcrowding are high, occupants may use rooms as bedrooms that VOA don’t consider to be bedrooms because they’re too small. In contrast the census would still record it as a bedroom as long as this was how the respondent intended to use it.

Accommodation Type

Some of the differences in bedroom counts between the two datasets can be attributed to accommodation type.  In 2011, 75% of households in Harlow lived in houses or bungalows but in Haringey, it was 41%. The remainder lived in flats, maisonettes or apartments. The distribution of agreement rates for the different output areas follows a similar pattern to that of accommodation type.

There are higher agreement rates in areas with a higher percentage of terraced houses and bungalows than those dominated by converted/shared houses and apartments in commercial buildings. Apartments in converted/shared houses and in commercial buildings are not your conventional property type, they present challenges in accurately defining and accounting for number of bedrooms and collection of address information used to link the datasets.

Census response rates

Differences in agreement rates could also arise from how the VOA is checked, updated and, how the census is answered and processed. Haringey was below the national average in terms of person response, household response, household return and person coverage rates in 2011. As for Harlow, it was at least equal to the national average.

Proportion of census records with missing information on bedrooms (non-response rate) and, records that had bedrooms information imputed due to missing or inconsistent data were much lower in Harlow (below national average), compared to Haringey.

So what happens now?

Whilst we recognise there are challenges of moving towards admin data to estimate number of rooms and bedrooms, it is Office for National Statistics’ (ONS) view that with some further development, the VOA data are of sufficient statistical quality to replace one or both of the questions on rooms and bedrooms in the 2021 Census.

We’re asking users to take part in the consultation so we can understand the impact this could have on services, policy development and evaluation and other uses of the data. Our report includes summary analyses and an output area analysis tool, which you can explore and see what the differences means for your geographical area.

We’ll evaluate responses to the consultation based on the strength of impact on uses of the data. We’ll then weigh this up against the benefits of having the information available more frequently.

This will, in turn, inform our recommendations on the topics for collection on the 2021 Census. These will be set out in a White Paper scheduled to be laid before Parliament in 2018.

The consultation is open until midnight on 20 September 2017.

For more information, please email using ‘Estimating rooms and bedrooms’ as the subject line.