It’s a well-known fact that there is too much retail floorspace currently available relative to demand. Or is it?
It’s an article of faith that town centres are being adversely affected by competition from out of town retail developments. But what are we basing that faith upon?
We’re not short of reviews and recommendations about the future of high street shops, but we may be shorter than we think of reliable, comparable retail floorspace statistics. Even published government figures have significant holes in them, as we’ll see.
Does this matter? It does to the planners, policy makers, property professionals and retail developers whose task it is to make decisions affecting our retail future. It does to people whose livelihoods depend on sustainable retailing, and the inhabitants of towns and cities relying upon them. We think that’s reason enough to call for change, with your help.
What’s wrong with our floorspace data?
Retail floorspace data is often derived from a sound, comprehensive and reliable dataset: the Valuation Office Agency (VOA) Summary Valuation tables (and their precursor the Valuation Support Application). The figures have gone through various stages of aggregation and modelling to create headline statistics and data series.
There are always issues with data which is collected to answer one set of questions being used to answer another set. The VoA, as its name suggests, collects data to value properties for non-domestic rates. It categorises properties by their business use, and may change these categories between rating lists to suit its own requirements.
This means that analysts have to take account of definitions, changes and assumptions in basing conclusions upon them, and that’s not always easy to do.
What makes it hard to compare and use the data?
We know this from long experience using the VOA Summary Valuation tables. We’ve found a number of dimensions to the data, some expected, some less so.
Definition and categorisation of business types
VOA businesses are categorised by an internal classification of ‘special category’ or SCAT codes. Businesses may be recorded under more than one of these codes.
The generic use of the code defining ‘shops’ is a key example, where more specific categories may exist but the ‘shop’ definition is used. This applies to both the national variation in businesses classified by the VOA, but also to any organisation reaching a definition of ‘retail’ by using these codes. It may be consistent with the definition used by another think tank or government adviser, but more likely it won’t be.
Data is collected and co-ordinated through eight different England Wales VOA regions. We have found a significant regional differentiation in the ‘market tone’, or the base level at which the value of properties is set, which may reflect local political or economic priorities. Similarly, there exists a regional variation in the use of SCAT codes to describe businesses.
This kind of differential may suit a particular region’s economic masterplan, but it doesn’t suit an analyst trying to define a coherent national picture.
We also know that high street retail floorspace is calculated on the basis of Net Internal Area, while retail warehousing and large superstores are calculated using Gross Internal Area, which is usually around 5% bigger. Again it is not clear how the experimental VoA stats handle this difference.
The VOA Summary Valuation is derived from a complete list of all hereditaments (broadly businesses at the property level) in England and Wales from a ‘Central Ratings List’. Not all properties are included: the Summary Valuation includes those records which have floorspace information collected.
Businesses such as food courts, non-livestock markets and public houses are omitted, and these categories may changed with the release of a new Rating List every five years.
The potential is clear for figures to reflect only parts of a total retail economy, and for conclusions based on them to be partial as a result.
The Summary Valuation tables have two time dimensions for consideration. The entire property portfolio is re-assessed and recorded every five years (2000, 2005, 2010 etc). Between these dates continuous updates are based on appeals and redevelopments, which create a dynamic dataset, prone to a lag-time in updating information compared to what’s ‘on the ground’.
The VOA has recently released some ‘experimental floorspace statistics’, including ‘retail’ floorspace. There is a summary explanation* of the methodology used to reach the numbers, but not in enough detail for this kind of analysis.
Imputation methods are used to reach these estimates. A methodology note is paramount to the useful interpretation of data. In their review Commercial and Industrial Floorspace Statistics, 1974 – 1985, what was then the Office of the Deputy Prime Minister explicitly notes that ‘the underlying methodology used to produce them has changed in response to changes in data collection methods.’
An open methodology
Overall, in fact, the issue is about availability of data and more of a problem with the lack of available details on the modelling or methodology used. We can cope with data collected for rating purposes, can re-categorise the many definitions of ‘shops’ and re-weight regional differences if we fully understand how the data have been manipulated so far.
We understand the need for open data. We have floorspace data, although we’d like it a lot less expensively. What’s lacking here is an open methodology which would allow published statistics to be usefully and consistently interpreted.
So if you fall into the categories of planners, policy makers, property professionals or retail developers, or have other reason to be concerned about our urban retail future, this matters. Retail data could be collected, modelled and reported in a more consistent and transparent way, and it could be made available more affordably to everyone.
There’s a basic need for participants in this analysis to define their data needs, harmonising requirements and encouraging the VoA to help meet them. The changes required are not fundamental. The benefits to all of us could be.
Under the skin of retail floorspace data
1. Changing custodians and gaps in the record
- Department of the Environment : 1974 – 1985
- Not available: 1986 – 1993 (Collected in 1986 by Hillier Parker but not released)
- Department for the Environment: 1994
- Not available: 1995 – 1997
- Dept for Transport, Local Govt and the Regions / Office for the Deputy Prime Minister: 1998 – 2005
- Valuation Office Agency : 2000 – 2012 (experimental stats)
- Geofutures / BCSC: 2005 – 2012
Our work with these datasets helped to identity some of these issues. Taking VoA data (which had to be paid for) we were unable to reach the same totals using the published SCAT codes. There may well be excellent reasons for the discrepancy, but without any insight into the methodology used, we’re left with a model which produces considerably larger estimates of available retail floorspace – see Fig 1 below.
2. Anomalous step changes
Examining the published data series, what becomes immediately evident are a number of sudden ‘leaps’ in estimated floorspace. These don’t accord with what we know about the state of the economy or industry, nor do they appear to be attributable to known error margins.
An example: Fig 1 shows the results of the recent VOA experimental statistics representing retail floorspace, alongside our own modelling of retail floorspace statistics using the VOA Summary Valuation tables both for the England and Wales extent from 2000 – 2012. The degree of difference is notable, and it’s down to different modelling methodologies. Fig 3 shows our SCAT methodology. You can read the VoA look-up tables if you like that kind of thing** but there are no codes included, so you can’t cross-reference them.
Fig 2 illustrates the same comparison. It also throws into relief the difference in the VoA’s own floorspace totals between 2004 and 2005. Not a time of retail slowdown by any of our reckonings – but a time when a new rating list was published. Clearly this fall in floorspace reflects a change in categoriation or modelling methods, but who knows what it was? And how can we make decisions based on data including such obvious anomalies?
Fig 3 Our methodology includes the following SCAT codes