2011 Census

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias

In South Gloucestershire we are also grappling with this problem. 
Using a similar approach to Piers (but using a snapshot of our LLPG taken on census day instead of Address Point) I estimate that the population of some wards is out by up to (-/+9%) .  Clearly this is far from ideal and will have political implications. 

I expect that we will have to use the ONS figures in the profiles we are planning to produce (with a caveat) and, at a later date, produce alternative figures using SASPAC.

regards
andy

Show Original Message -------------------------------------------
Original Message:
Sent: 20-12-2012 09:16
From: Piers Elias
Subject: Changes in Ward Boundaries and Best Fit

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias

The same exists issue in Wolverhampton and other Black Country Councils.

A problem with doing in-house matching to exact boundaries is that other ONS ward-based statistics are (presumably) now issued on a best fit basis.I think that birth and death statistics will be issued  on a best-fit basis from now on.

Greg ball




Show Original Message -------------------------------------------
Original Message:
Sent: 03-01-2013 05:36
From: Andy Cornelius
Subject: Changes in Ward Boundaries and Best Fit

In South Gloucestershire we are also grappling with this problem. 
Using a similar approach to Piers (but using a snapshot of our LLPG taken on census day instead of Address Point) I estimate that the population of some wards is out by up to (-/+9%) .  Clearly this is far from ideal and will have political implications. 

I expect that we will have to use the ONS figures in the profiles we are planning to produce (with a caveat) and, at a later date, produce alternative figures using SASPAC.

regards
andy

-------------------------------------------
Original Message:
Sent: 20-12-2012 09:16
From: Piers Elias
Subject: Changes in Ward Boundaries and Best Fit

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias

Re.   ...other ONS ward-based statistics are (presumably) now issued on a best fit basis.I think that birth and death statistics will be issued  on a best-fit basis from now on...

How can we get confirmation from ONS that this will be the case?  I presume it will also affect other stats on NOMIS (JSA ect)... 




Show Original Message -------------------------------------------
Original Message:
Sent: 04-01-2013 12:03
From: Greg Ball
Subject: Changes in Ward Boundaries and Best Fit

The same exists issue in Wolverhampton and other Black Country Councils.

A problem with doing in-house matching to exact boundaries is that other ONS ward-based statistics are (presumably) now issued on a best fit basis.I think that birth and death statistics will be issued  on a best-fit basis from now on.

Greg ball



-------------------------------------------
Original Message:
Sent: 03-01-2013 05:36
From: Andy Cornelius
Subject: Changes in Ward Boundaries and Best Fit

In South Gloucestershire we are also grappling with this problem. 
Using a similar approach to Piers (but using a snapshot of our LLPG taken on census day instead of Address Point) I estimate that the population of some wards is out by up to (-/+9%) .  Clearly this is far from ideal and will have political implications. 

I expect that we will have to use the ONS figures in the profiles we are planning to produce (with a caveat) and, at a later date, produce alternative figures using SASPAC.

regards
andy

-------------------------------------------
Original Message:
Sent: 20-12-2012 09:16
From: Piers Elias
Subject: Changes in Ward Boundaries and Best Fit

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias

I put this discussion on the Knowledge Hub as well so may be worth looking there as well - pity there isn't a cross-link.

https://knowledgehub.local.gov.uk/group/2011census/forum/-/message_boards/message/9511196

For Census data I would agree that you need to be consistent but do all government agencies now have to use a Postcode to OA lookup on a "best fit" basis and then aggregate to LSOAs/Wards or are postcodes matched directly to LSOAs/Wards by DWP/HMRC etc? I doubt all agencies will be using the same lookups.

 
Beyond Census data, LSOA is the lowest level published by ONS. For population estimates, births and deaths, they will be aggregated from OAs on "Best fit" but I don't know how data from Central Govt Departments will be matched/calculated. If anyone knows the answer, please let us know.

Happy New Year

Piers

Show Original Message -------------------------------------------
Original Message:
Sent: 08-01-2013 04:42
From: Andy Cornelius
Subject: Changes in Ward Boundaries and Best Fit



Re.   ...other ONS ward-based statistics are (presumably) now issued on a best fit basis.I think that birth and death statistics will be issued  on a best-fit basis from now on...

How can we get confirmation from ONS that this will be the case?  I presume it will also affect other stats on NOMIS (JSA ect)... 



-------------------------------------------
Original Message:
Sent: 04-01-2013 12:03
From: Greg Ball
Subject: Changes in Ward Boundaries and Best Fit

The same exists issue in Wolverhampton and other Black Country Councils.

A problem with doing in-house matching to exact boundaries is that other ONS ward-based statistics are (presumably) now issued on a best fit basis.I think that birth and death statistics will be issued  on a best-fit basis from now on.

Greg ball



-------------------------------------------
Original Message:
Sent: 03-01-2013 05:36
From: Andy Cornelius
Subject: Changes in Ward Boundaries and Best Fit

In South Gloucestershire we are also grappling with this problem. 
Using a similar approach to Piers (but using a snapshot of our LLPG taken on census day instead of Address Point) I estimate that the population of some wards is out by up to (-/+9%) .  Clearly this is far from ideal and will have political implications. 

I expect that we will have to use the ONS figures in the profiles we are planning to produce (with a caveat) and, at a later date, produce alternative figures using SASPAC.

regards
andy

-------------------------------------------
Original Message:
Sent: 20-12-2012 09:16
From: Piers Elias
Subject: Changes in Ward Boundaries and Best Fit

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias

Sorry for being late onto this thread but I can confirm that 2011 Census estimates will be in line with the GSS policy that all national statistics estimates for any geography should be best-fitted from output area (OA). Best-fitting is a key part of the Geography Policy for National Statistics and is mandated best practice for producing national and official statistics by geography, so that  outputs are consistent and comparable in their use of geography. Aggregating whole statistical building blocks of OAs for outputs also ensures there are no potentially disclosive slivers when comparing different geographies. The policy can be downloaded at http://www.ons.gov.uk/ons/guide-method/geography/geographic-policy/best-practice-guidance/geography-policy-for-national-statistics.pdf .

 

GSS departments should be using the recommended postcode directory produced by ONS to assign postcoded data to all geographies - wards, parliamentary constituencies etc, best-fitted from the postcode's OA. This is the National Statistics Postcode Lookup (NSPL) that is updated every quarter and can be freely downloaded at http://www.ons.gov.uk/ons/guide-method/geography/products/postcode-directories/-nspp-/index.html. NS also publishes its ONS Postcode Directory (ONSPD) - which assigns postcodes directly via the postcode centroid directly to the same output geographies. Some of the ONSPD postcode allocations will therefore be different to those on the OA based NSPL, and shouldn't be used for official statistics as they wouldn't be comparable with other statistics produced using the NSPL.

 

ONS published lookups from 2011 OA to a number of output geographies, including 2011 wards,  in October, downloadable at http://www.ons.gov.uk/ons/guide-method/geography/products/census/lookup/2011/index.html . At the end of January ONS will publish, for each of the lookups, how "well" the best-fit is, expressed as a percentage of the OA's population that falls exactly within the output geography's boundary. This should give you a better measure of how the population splits across the boundaries, as it uses the grid references of the households plotted into the output geography's boundaries.

 

Estimates of number of households and population by sex for each postcode will be published at the end of February. These postcodes are only ones that contained enumerated households or persons during the census, and a separate dataset will indicate which of these postcodes were split across an output area. These postcode estimates could be used to build estimates for wards and other geographies, using a postcode lookup like the ONSPD which is not best-fit from OA.  

 

So while the best-fit policy is to make sure that national statistics are consistent, comparable and non-disclosive, we are providing tools that allow alternative estimates to be compiled. We do recognise that best-fit - in some instances of small wards and parishes that cross OA boundaries particularly - doesn't meet some customers' requirements for more exact estimates. It's really helpful that Piers intends to make clear that his own ward estimates are not the official ONS best-fit, and would ask that this is always made clear by others thinking of doing the same.

 

A Happy New Year to everyone.

Show Original Message -------------------------------------------
Original Message:
Sent: 20-12-2012 09:16
From: Piers Elias
Subject: Changes in Ward Boundaries and Best Fit

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias

thanks for the alert Piers, and the very nicely laid out note - can I plagiarise, pretty  please?  :-)

We use a similar method to yours  for estimates in bespoke areas, except that we base it on our LLPG addresses rather than Address Points.

we have checked and found a pair of Edivs (out of 73)  in Powys that will definitely give us this problem, Bwlch and Crickhowell ;
   105 people are assigned to Bwlch that should be in Crickhowell
    Bwlch is overestimated by 11.4%
    Crickhowell underestimated by 3.6%.

Frustratingly this is not due to ward boundary change;  the split COA in question is part of a large new rural LSOA which has been created as a result of our own representations to ONS Geography about lack of homogeneity in the old LSOAs <argh>

There are other minor boundary changes, involving very few houses which we wont worry about.


I am not sure how best to publish our 'even better fit' data.  How to label it is key I suppose, to explain why people will find different estimates for Builth and Crickhowell in the ONS tables - will it be simple to override the ONS data, and standard names with bespoke data and labels  in an Instant Atlas system like yours?

kind regards

Di Greaves











Show Original Message -------------------------------------------
Original Message:
Sent: 20-12-2012 09:16
From: Piers Elias
Subject: Changes in Ward Boundaries and Best Fit

Some Local Authorities are wrestling with the impact of the ONS Geography policy of "best fit" that allocates whole Output Areas to a Ward - based on the Population Weighted Centroid of the OA rather than an exact fit method or a "postcode best fit" method (ONS' method for Small Area Population Estimates up to Mid 2010). 
This only affects Local Authorities that have had new Wards since 2003.

In most cases, the impact is minimal but we have examples in Hartlepool where the Ward population is 10% different (and 4 wards, more than 5% different) when compared to a more exact fit method using Residential Addresses from Royal Mail AddressPoint.  We have used this method here for several years for calculating bespoke areas, including changed Wards, from OA population and household estimates that we generate.
 
Clearly a 10% difference has a big impact on actual population and numbers of households and any Ward-based rates (e.g. unemployment, fertility, mortality) and as result we will be re-calculating all Ward data for the two Local Authorities affected here using our own method (via SASPAC). More details are available on this note - https://www.teesvalleyunlimited.gov.uk/media/139547/tvu_census_based_estimates_for_sub-borough_areas_update_dec_12.pdf

The main issue with the above approach is that there will be two sets of Ward results around which will differ slightly.  All Ward data we put on our web-site will be calculated using our method and we will have to carefully label the results to distinguish them from the ONS best fit results.

I would be interested to hear if other Local Authorities are considering a similar exercise.

Seasonal greetings,
Piers Elias