How important are postcodes?

Posted by Nick Stripe on January 19th 2007

OAs were specifically designed to have a strong relationship with postcode geography.  Census 2001 enumerated postcodes were the building blocks used in the construction of OAs.  Only where postcodes straddled ward, parish (England), and community (Wales) boundaries, were they split into more than one OA polygon.  This affected approximately 2.5% of all enumerated unit postcodes.

The boundary set that resulted was irregular, but the postcode foundations of the geography were considered important for data linkage purposes.

As with administrative geographies, postal geography moves around over time.  It is a convenient, rather than ideal, geography for statistical purposes.  Ideally data records would be geo-referenced to addresses or grid references, and lookups from these would be easy to produce and distribute.  But further progress in this direction is still needed.  The desire has not yet proved strong enough to facilitate necessary data developments.

In proposing a policy of stability, we are also proposing that the postcode to OA link be allowed to degrade, as it has done with every postcode update since the last Census.  What if this meant that 10% of Census 2011 enumerated postcodes were split between OA polygons?

Can we start to look beyond our reliance on the postcode?

add new comment

Postcodes are important

Posted by Michelle vonAhn (not verified) on 15/02/2007 - 16:47

We use postcodes a lot as a means of linking data to geographies, but more to SOAs than to OAs. I am not particularly exercised about the prospect of OAs being re-drawn within the constituent SOA, but I would not like there to be a loss of linkage to postcode areas wherever possible.

We use the AFPD to link administrative data to SOAs for thematic mapping of summary adminstrative data. In a perfect world, the LLPG would be the ideal way of linking the data to a geography, but it is not a perfect world (yet), so we would prefer that SOAs retain a close link with postcodes. I too would advocate that OS have control over postcodes, given their essential building-block role in the creation of this statistical geography.

reply

There is something

Posted by Simon Dickson (not verified) on 07/02/2007 - 12:31

There is something fundamentally and absolutely good about using a geographic identification system which mere human beings can buy into. Inherent in the concept of government statistics, which are paid for out of my and your taxes, is the fact that ordinary people should be able to get access to them. Websites are falling over themselves to offer local information, because there's a demand for it. ONS must face up not only to the demand, but the obligation to offer it. A better informed population will be a better population.

ONS can't operate in a vacuum here. Postcode-based searching is the number-one established way to find locally relevant information via the number-one information channel, namely the web. Please guys - accept the reality, accept that it's beyond your control, and work with it. It may not be a perfect standard, but it's a standard.

Is there room for an 'out of the box' solution? Can ONS take ownership of postcodes away from Royal Mail, in the national interest?

reply

Don't ditch the postcode!

Posted by Anonymous (not verified) on 02/02/2007 - 15:31

I appreciate that other users may have control over the databases that they are using, and require accuracy that means that the 2.5% of postcodes that straddle OA boundaries is problematic (well, I'd rather they didn't straddle, but for me it's tolerable).

I'm involved in the secondary use of customer databases, joining them up to create a dataset that represents almost an entire sector. The data that is collected is for customer service/marketing requirements, and as such I don't see that they are going to go beyond collecting postal addresses any time soon. As such, postcodes are an invaluable glue for datasets over which we have little control but which can be put to many good uses.

reply

Postcodes

Posted by Tim (not verified) on 26/01/2007 - 13:28

For the research we use the site for, the Postcode location remains the simplest, quickest and easiest way by which to identify a site's location, using your interactive map. Once we've identified the exact location, we use Output and Super Output data and postcodes to us are then irrelevant. The fact is using XY co-ordinates for example would be a pain for our business, simply because everyone knows their postcodes, few know the co-ordinates of their property. We ned to drill down to very small areas for Demographic information and the sizes of Output areas, in most cases, is ideal for us in the work we do. We wouldn't want to see their size increased and especially for comparing past data, changing them could cause us problems. It depends what benefits there are to be had by changing the boundaries. I can't see any change is needed from our perspective, the site, the data, being able to search using postcodes to pull up your map and the boundaries work for us and very well too!

reply

well, just one more comment on postcodes

Posted by Di Greaves, Powys CC Research and Information manager (not verified) on 01/02/2007 - 14:36

I said I'd get on with the consultation now but....

Tim says " The fact is using XY co-ordinates for example would be a pain for our business, simply because everyone knows their postcodes, few know the co-ordinates of their property."
but the beauty of the NLPG system is they dont need to know the grid reference and neither do you, the researcher! You just have to also record their house name or number which is unique within the postcode, (now dont tell me they dont know that either!) Given these basic address elements the NLPG will give you the UPRN,and the grid reference .

Suppose they won't give you the house name/number because they like the faint element of anonymity within the postcode group of addresses, or they dont want to get on your mailing list? then that's another question, a choice to be made in the research design; EITHER
to provide reassurance / incentives sufficient for collection of house names/numbers
OR
straight away attach to the record the grid reference for the current population weighted centroid of the postcode or the best available postcode polygon. (But please dont wait a year before doing this because centroids and polygons will surely have changed by then)

Di

reply

Postcodes

Posted by LSC Planning Manager (not verified) on 26/01/2007 - 12:06

The proposal to use alternative grid references in place of postcodes overlooks the use of the postcode by planners such as the LSC in determining access to and penetration of services.

Postcodes can be linked to individuals in a way that other references can't. As a result, they are hugely useful in supporting the use of multiple, varied datasets that enable monitoring of participation/non-participation by ethnicity, disability, gender etc. They allow comparisons to be made between a range of factors (i.e. compare learning take-up with receipt of Incapacity Benefit.

The prospect of having to identify a grid reference for every person who enrols in learning (or uses any public service) is enough disincentive for anyone dealing with data. Not to mention the havoc it would wreak on all existing MI systems.
If this is desirable, then introduce it alongside postcodes as an additional planning and modelling tool for data and research.

I can see there would be benefits of this approach, but I also know that the likelihood of introducing any reliable method of matching postcodes to the new system would be remote - at a stroke disabling most govt MI systems.

reply

No need for postcodes anymore

Posted by Anonymous (not verified) on 24/01/2007 - 11:19

Postcode data is innacurate when drawn on maps, and has little value other than to aggregate statistics to local geographic areas. Depending upon the size of the postcode area, these statistics are then incomparable. Ultimately, postcodes were designed to aid the delivery of post - not the analysis of small area data therefore it was always going to be unfit for purpose.

As for geo-referencing, I use XY data which is highly accurate and thus eliminates the innacuracies that I would encounter using postcodes. There should be a greater emphasis on using XY references rather than the now highly outdated postcode / postcode centroid system. This emphasis should be directed towards the data custodians who collect data in the first place. For example, all addresses we deal with are verified with the LLPG, if they don't match to a property then work is carried out to physically locate it's reference.

The NLPG should do this - that's if it ever works. A shame then, that other developed (and none developed countries) seem to be better at that than we do.

In answer to your question - we are already independently looking beyond the postcode as we have chosen to use the most accurate referencing available (XY LLPG matching).

The ONS should disregard postcode data in favour of geographically independent boundaries. No effort should be placed on postcodes as it would be a waste of time and resources catering for a system that has long since had it's day in the world of local area spatial-statistics.

reply

Postcodes are relevent & very useful

Posted by Clive (not verified) on 26/01/2007 - 09:22

I use the postcode fields whenever I use the ONS data - they are the best method I have to link population or household data to my customer database - eg I know how many customers I have in BD1 from a search on my database - I can compare this to the population (by segement, household etc) within the census data.

The postcode data is the only way I can do this as it is a constant geographical reference within both the census data and every address for every customer.

WIthout the postcode the greatest level of detail I could achieve would be to Town level, which would severely impair the usefullness of the census data, in relation to my industry sector

reply

If only they were!

Posted by Di Greaves, Powys CC Research and Information manager (not verified) on 01/02/2007 - 13:47

Unfortunately, Clive, it is not true that postcode data "is a constant geographical reference within both the census data and every address for every customer."

Post codes are created, deleted and frequently revised solely for the business purposes of the organisation that owns them. They are not consistent over time, the postcode set used in the 2001 Census does not match the postcode set as it exists today. Postcodes don't nest within any administrative boundaries anywhere in the UK, they can't be used to assign records accurately between Wales and England, let alone between two wards or parishes.

The data and the metadata are not freely accessible and creating an accurate set of statistics by matching postcodes from a data set compiled over a number of years is a logistical nightmare.

They have been useful, but only because they were the first small area, hierarchical, geographic coding with national coverage to become available. Their day will soon be past for research and statistics.

Now we have these excellent LLPGs established in the NLPG framework, created by local government and stamping a UPRN on a property at the very start of its address lifecycle when it acquires a street name /number / house name. Right from the start there is no question about where exactly it is, because the grid reference is recorded and attached to the UPRN at the moment of creation by the local authority who are there on the ground inspecting the building. The LLPG also of course provides a vehicle for recording and date stamping the changes in postcode for a property. (And for adding in the TOID when the building is surveyed and added to the OS mapping)

The UPRN system devised in the NLPG is so much better than the postcode set as a geo-referencing tool; its accurate, stable, and can be used to build small area geographies from single addresses upwards. With the help of the NLPG we can actually make better informed use of postcoded data, but once having access to the UPRN, no-one in their right mind would record the postcode only on a database record, (would they?)

As you can tell I am passionate about this! Now I'v got that off my chest, on with the consultation, sorry to come in so late in the day, have been a bit busy here,

Di

reply

Seems there are two customer groups

Posted by Clive (not verified) on 07/02/2007 - 12:42

Seems there are two sets of views here, relating to the two types of user for this data :- those that relate this data to products and services (commercial users), & those that use the data for "true" statistical/geographic reasons (Scientific & Authority users).

Each group uses the data differently, for my part things are fine as they are - any change would be "a pain" for me also, however I can see how much this change means to Di, and can appreciate the value of having an anchored geographical reference if I were sat on the other side of the fence, boundry or polygonal division.

I'm for the new system, & eager to see what extra data can be delivered, but for my level of use postodes are fine.

reply