Disaggregating annual subnational gross value added (GVA) to lower levels of geography: 1998 to 2019

• We are publishing new gross value added data at lower levels of geography than ever before, including Middle layer Super Output Areas, Intermediate Zones, health board areas and Parliamentary constituencies.

• Lower layer Super Output Areas, Data Zones and Super Output Areas level data form the building blocks that are used to build upwards to derive gross value added and productivity per job filled for new geographies such as towns and cities, and travel to work areas.

• We demonstrate the flexibility of using building blocks data by building maps of user-specified geographies.

• The building blocks data will be accessible to accredited researchers through the Secure Research Service (SRS).

For a number of years, our users have been asking for statistics for small area geographies to target and monitor the impact of local policy making. This interest has been rapidly increasing.

The 2016 review of economic statistics by Sir Charles Bean recommended that the Office for National Statistics (ONS) should aim to provide greater granularity of economic statistics, in terms of both the amount of detailed information provided, and of the levels of geographic areas covered by the statistics.

The Government Statistical Service’s subnational data strategy is also driving the supply of granular data, and this project contributes partly to meeting the growing demand. The publishing of experimental gross value added (GVA) data at sub-regional level is one of the initiatives seeking to disaggregate National Statistics to local level. The experimental statistics are based on modelling and apportionment and are in current prices.

This article describes the aims of the “flexible geography” project to develop methods and data needed to produce statistics for “building blocks”, which can then be used to aggregate and build bespoke user-defined geographies. It covers progress made to date, and a look ahead at what might be achieved in the future.

It is accompanied by a new dataset with experimental estimates of GVA for a set of geographic areas that have never been covered before. These include Middle layer Super Output Areas (MSOAs), Parliamentary constituencies, travel-to-work areas (TTWAs), health boards, towns and cities, and a selection of user-defined bespoke areas.

Our Census geography methodology article defines the main building blocks of MSOAs and TTWAs. MSOAs in England and Wales are broadly equivalent to Intermediate Zones in Scotland. Although there is no MSOA-equivalent geography for Northern Ireland, we have included district electoral areas (DEAs) as an intermediate level between Super Output Areas (SOAs) and local government districts (equivalent to local authorities). You can find out more about the Northern Ireland DEAs in this geography fact sheet (PDF, 506KB).

Lower level geographies:

• Lower layer Super Output Area (LSOA, England and Wales): population of 1,000 to 3,000 people (400 to 1,200 households)
• Data Zone (DZ, Scotland): population of 500 to 1,000 household residents
• SOA (Northern Ireland): population of 300 to 6,000 people

Middle level geographies:

• MSOA (England and Wales): population of 5,000 to 15,000 people (2,000 to 6,000 households)
• Intermediate Zone (Scotland): population of 2,500 to 6,000 household residents
• DEA (Northern Ireland): population of 14,000 to 45,000 people; consists of 5, 6 or 7 wards

We have built higher levels of geography from LSOAs, DZs and SOAs as the lowest level of geography. However, because of economic dominance and the risk of disclosure in some LSOAs, DZs and SOAs, we have not published the dataset, but will make the dataset accessible to users through the ONS Secure Research Service (SRS).

Historically, we have produced regional economic statistics for a rigidly controlled set of geographic areas, specified according to the European Union’s Nomenclature of Units for Territorial Statistics (NUTS) classification.

When the UK exited the European Union, International Territorial Levels (ITL) were introduced as a new UK-managed international statistical geography at the beginning of 2021 (A Beginner’s Guide to UK Geography (2021) v1.0 (PDF, 1.23MB)). The new system mirrors the NUTS classification that it replaced and follows similar standards, for international comparability. The compilation system is complex and designed around the NUTS areas, which has historically limited the flexibility to change geographic classifications.

With the introduction of Local Enterprise Partnerships (LEPs) in England (2010 to 2015), it became clear that we needed ways to meet the needs of emerging new areas that lay outside of the controlled geographical framework. The LEPs cover a hugely variable range of geographic areas, some contiguous with existing ITL areas, others able to be built up from whole ITL areas, and still others that needed lower-level building blocks to be constituted.

The demand for building block level data and statistics that enable “flexible geographies” continued to increase, and the concept evolved. This has culminated in a multi-layered approach that provides the greatest possible amount of information for each area while preserving the confidentiality of data relating to identifiable enterprises.

The Digital Economy Act (2017) has enabled access to an administrative source of data collected through Value Added Tax (VAT) returns, which contains records for almost all businesses operating in the UK (all those registered for VAT) and includes variables for the company’s turnover and expenditure.

By matching these records to the Inter-Departmental Business Register (IDBR), we have created a dataset that contains a separate record for each physical site (local unit) where a business operates. The IDBR holds information on the employment and main activity carried out at each site. Crucially, it also holds the postcode of the site, allowing us to allocate VAT turnover to business sites in any geographic area, even very small ones.

Turnover information provided for the whole company is allocated between the sites of a business according to the relative share of the total workforce located at each site. This approach assumes workers across all sites contribute equally to a company’s gross value added (GVA), meaning they are treated as equally productive. This may not always be the case because of differences in human capital and also because of differences in activities performed at different business sites, but for the purpose set out here it is a reasonable assumption.

The VAT records do not include turnover for households with employees and own-account production, imputed rental of owner-occupied dwellings, and non-market activities of public sector industries. For this information, we used the following alternative data sources to apportion:

• for households with employees and own-account production, we used population estimates
• for imputed rental of owner-occupied dwellings, we used dwelling stock where available, otherwise we used population estimates
• for non-market activities of public sector industries, we used employment

The idea of breaking down economic activity to small geographic areas depends on the availability of data with sufficient coverage of all areas, and the ability to derive data extracts for very small areas. The whole process uses administrative data sources that cover different time periods. Missing years (1998 to 2011) are modelled from the existing data.

We refer to the Government Statistical Service’s disclosure control guidance for tables produced from surveys (PDF, 250KB) to address disclosure risk. Historically, we have guarded against disclosive data by ensuring whatever industry level data we publish at any given geographic level includes a minimum of at least four separate business enterprises. At this low level of geography, we do not provide an industry breakdown, however, even with total industry gross value added (GVA), the risk of perceived disclosure remains because of the potential dominance of some industries in local areas.

We have defined dominance as a particular enterprise (or industry if the industry includes a dominant business) that contributes at least 85% of the area’s GVA. Dominance becomes a potential disclosure issue if users can estimate data points to within 10% to 15% of the true values.

Economic dominance, in this case, does not necessarily mean disclosure of information for a specific business unless there is only one enterprise in a particular industry in a particular area. The risk is further mitigated by the fact that all the GVA figures are estimates compiled using auxiliary variables to break down figures from the UK total (thus ensuring that our estimates sum to all International Territorial Levels (ITLs)), meaning that there is no way to derive precise company values from our published data.

Because of the perceived risk of disclosure, we have decided not to publish Lower layer Super Output Area, data zone and Super Output Area (LSOA/DZ/SOA) data. We will make data available in the Secure Research Service (SRS) for access to accredited researchers in a secure environment while also seeking feedback from users through a consultation process. We may publish the lower-level geography data in future as we continue to work on understanding how best to manage the disclosure risk.

At Middle layer Super Output Area (MSOA, England and Wales) or intermediate zone (IZ, Scotland) level, there are fewer incidences of economic dominance (of enterprise or industry) and therefore lower risk of disclosure. Here, potentially disclosive MSOAs have been paired with neighbouring MSOAs to remove this risk. The merged MSOA and IZ will remain so for the whole data time series.

The first phase of this project focused on breaking down gross value added (GVA) to local authority districts across the UK, including unitary authorities, metropolitan districts, London boroughs and Scottish Councils. From these building blocks we constructed estimates for local enterprise partnerships (LEPs), combined authorities with elected mayors, other city regions and enterprise regions of interest to people across the UK nations. The first estimates of GVA at these levels were included in our Regional economic activity by gross value added (balanced), UK: 1998 to 2017 bulletin, and were then refined and developed in the December 2019 dataset and the May 2021 dataset.

The GVA estimates for local authorities are provided with an industry breakdown into 34 industry sectors. This is less detail than is provided for higher-level geographic areas of the UK, because as we break down the data to smaller areas, the number of business enterprises included in each region diminishes. The level of industry detail provided is the maximum that can be published without risk of disclosure of confidential information about individual businesses.

For the experimental statistics presented here, we have taken the current price estimates at local authority level and apportioned them to lower geographical levels using the aforementioned data sources.

The next step in this development is breaking down the total industry gross value added (GVA) to Lower layer Super Output Areas, Data Zones and Super Output Areas (LSOAs, DZs and SOAs). These lower-level census geographies are designed to divide the UK based on the number of households, and include any businesses operating in the same areas. They are built to be comparable in terms of population size but not GVA, which is a workplace measure of economic output that includes commuters’ contributions. There is need for caution when combining or building areas across country boundaries.

The breaking down of GVA to these lower-level geographies represents a significant improvement in granularity, which allows us to construct much more detailed geographic areas.

The data from these small building blocks allow us to aggregate by adding flexibly to larger areas. These areas include Middle layer Super Output Areas (MSOAs) in England and Wales, Intermediate Zones in Scotland, and district electoral areas (DEAs) in Northern Ireland, Parliamentary constituencies, travel-to-work areas (TTWAs), health boards, and towns and cities. They have also enabled us to build examples of user-defined bespoke areas. We provide time series data for each of these areas for the period 1998 to 2019.

We demonstrate the flexibility argued for in this article by applying the flexible geography concept to four user-defined test cases. The user-defined areas are the West Midlands Metro region, the Old Oak and Park Royal Development Corporation region, the Clyde Gateway, and Clyde River region.

The standard way of presenting lower-level geographies as per the Government Statistical Service (GSS) Geography Policy is to plot their population-weighted centroids into the defined geography’s boundary (the best-fit method). The statistics we are producing with this article are experimental and not official statistics. In line with the Office for National Statistics’ (ONS’) current strategy, we have adopted an ambitious and rather radical approach to illustrate the flexibility that we anticipate our users will require in the future. We therefore use the recommended approach to map Lower layer Super Output Area, data zone and Super Output Area (LSOA, DZ and SOA) centroids for best-fit results, together with an all-inclusive approach that includes LSOAs, DZs and SOAs with contact points with the mapped area. We use the latter approach for some user-specified areas where the centroid approach would not provide adequate coverage of the area of interest. In this way, we are able to show the data for standard geographies and to flexibly use the building blocks to construct non-standard geographies’ analytical areas.

Where we have not used the standard centroid approach, we have provided an explanation and the rationale for it. Such clarity of methods and assumptions is important for transparency and prevents multiple interpretations of the data.

West Midlands Metro region

The West Midlands Metro is a light-rail system operating between Wolverhampton St. George’s and Birmingham Bull Street. The rail line was built in 1999 and further extended in 2015 and 2019, enabling us to see the changes in local gross value added (GVA) using our data time series.

Please note that our results do not allow for conclusions regarding cause and effect. They are for illustrative purposes only, and further investigation into the causes of economic change are beyond the scope of this article.

The West Midlands Metro is an interesting case that illustrates the geographical flexibility we are exploring. It shows how we can fit the analysis in an area with no defined boundary. Since the metro is a line, we include all LSOAs through which it passes, as shown in Figure 1.

Figure 1: West Midlands Metro region

Source: Office for National Statistics

Notes:

1. OS Zoomstack; Ordnance Survey; West Midlands Metro, for list of stations operating as at 4 October 2021.
2. Contains OS data © Crown copyright 2021.
3. Graphic created by ONS Geography.

Download this image Figure 1: West Midlands Metro region

.png (120.7 kB)

The West Midlands Metro has 24 stations and passes through 42 LSOAs. In 2019, the region had a total GVA of £8,496 million, more than doubling from £4,092 million in 1998, all measured in current prices.

The data time series in Figure 2 shows a smooth increase in GVA with a marked upturn from 2015, which may reflect the first major line extension from Snow Hill to Grand Central. The increase includes both volume and price growth.

Old Oak and Park Royal Development Corporation (OPDC) region

The Old Oak and Park Royal Development Corporation (OPDC) is a 650-hectare regeneration project in West London, which was officially launched in 2015. Using the connections with the HS2 high-speed rail network, the project aimed to construct thousands of homes and job opportunities.

The user-defined LSOAs for the OPDC area allow us to illustrate two methods for flexible mapping. Firstly, we produce a standard map based on the recommended centroids approach. Secondly, we use an alternative method, the all-inclusive approach, that includes all LSOAs that intersect with the user-defined boundary. These two approaches are illustrated in Figure 3.

Figure 3: Old Oak and Park Royal Development Corporation region

Source: Office for National Statistics

Notes:

1. London Datastore, Greater London Authority licensed under the Open Government Licence v.3.0.
2. Contains OS data © Crown copyright 2021.
3. Graphic created by ONS Geography.

Download this image Figure 3: Old Oak and Park Royal Development Corporation region

.png (101.1 kB)

Fitting the OPDC map using centroids results in six LSOAs from the user-specified list falling within the boundary. The six cover the bulk of the OPDC map, with total GVA of £2,712 million in 2019 (increasing by 20.6% from 2014). Some of the six LSOAs straddle the OPDC boundary. There are also some LSOAs that stretch into the OPDC boundary but have their centroids outside the map.

Applying the intersection approach means we include all LSOAs that have a contact point with the boundary of the OPDC map. In all, 24 LSOAs have contact points with the OPDC boundary, which stretch the area of interest beyond the OPDC map as illustrated in Figure 3 in lighter colour. Using this method, the inclusion of 18 additional LSOAs results in a higher estimated GVA of £3,936 million for 2019 (increasing by 11.1% from 2014). This method might be preferred if it is important to include the entire area of interest in the estimate. Figure 4 shows the two GVA series based on the two mapping methods.

Clyde Gateway and Clyde River region

In Scotland, we have two user-specified areas: Clyde Gateway and the Clyde River region. Clyde Gateway is an urban regeneration project which exists to drive inward investment and improvement for the people and communities across the east end of Glasgow and South Lanarkshire. The area (840 hectares) sits within the wider Clyde River region, which stretches approximately 110 miles from South Lanarkshire to the Firth of Clyde.

We illustrate another element of the flexibility we seek to introduce using the two areas: that users can analyse smaller bespoke areas that are nested within larger areas. The 26 Data Zones that map the Clyde Gateway have been located using the best-fit method. The 762 Data Zones that map the Clyde River region are located within a mile radius of the River Clyde. The superimposed maps of the two regions are shown in Figure 5.

Figure 5: Clyde Gateway and Clyde River region

Source: Office for National Statistics

Notes:

1. OS Open Rivers, Ordnance Survey; Scottish Government licensed under the Open Government Licence v.3.0.
2. Contains OS data © Crown copyright 2021
3. Graphic created by ONS Geography

Download this image Figure 5: Clyde Gateway and Clyde River region

.png (198.5 kB)

The Clyde Gateway data time series increased steadily from an estimated GVA of £367 million in 1998 to £596 million in 2015 and then steeply to £845 million in 2019 as shown in Figure 6.

Figure 6: Clyde Gateway GVA constituted 2.9% of Clyde River region GVA in 2019

Clyde Gateway and Clyde River region, total GVA (£ million), current prices, 1998 to 2019

Download the data

.xlsx

In 2019, the Clyde River region-estimated GVA amounted to £28,727 million; the data time series shows a steady increase from £14,010 million in 1998. Thus, the Clyde Gateway contributed 2.9% of the Clyde River region’s GVA in 2019.

The main results of this project are three datasets.

UK small area GVA estimates

The dataset consists of four tables:

• Table 1: Middle layer Super Output Areas (MSOAs) (England and Wales) – total GVA (£ millions)
• Table 2: Intermediate Zones (IZs) (Scotland) – total GVA (£ millions)
• Table 3: District electoral areas (Northern Ireland) – total GVA (£ millions)
• Table 4: Geography reference table

This dataset allows users to create their own bespoke areas by filtering MSOA and IZ of interest. We have also provided a lookup sheet detailing how the Lower layer Super Output Area, data zone and Super Output Area (LSOA, DZ and SOA) building blocks are nested in local authorities, ITL1 regions and other geographies. Users will be able to access LSOA, DZ and SOA data through the Secure Research Service (SRS).

GVA is a workplace measure of economic output, which includes the contribution of in-commuters. While GVA per head can be a useful way of comparing regions of different sizes, it is not such a good indicator for those areas with high net in- or out-commuting. We have therefore not provided estimates of GVA per head, which uses a residence-based denominator. Productivity measures such as GVA per job filled are suggested as a better means of comparing economic output between areas and these data are provided for TTWAs and towns and cities.

GVA estimates for bespoke areas

This dataset includes the total estimated GVA, per year, for each bespoke area described previously. Refer to the geography reference table in the small area estimates dataset for each area’s constituent LSOAs, DZs and SOAs, and the dataset will be accessible through the SRS.

UK GVA and productivity estimates for other geographies

This dataset consists of estimated GVA for other user-specified geographies, for the period 1998 to 2019. It also has productivity estimates for travel-to-work areas (TTWAs) and towns and cities for the period 2009 to 2019.

Towns and cities in England and Wales are based on the built-up area subdivision boundaries (BUASD), or built-up area boundaries (BUA) where no subdivisions exist. Towns are the built-up areas or subdivisions with a population between 5,000 and 225,000 at the census in 2011. Cities are built-up areas or subdivisions with a population over 225,000 usual residents. Towns and cities in Scotland are based on localities; and in Northern Ireland, they are based on settlements.

The productivity estimates are based on the GVA per job filled productivity measure, which has two inputs: GVA data in the numerator, and labour market data (number of jobs) in the denominator. Data were first estimated for LSOAs in England and Wales, Data Zones (DZs) in Scotland, and SOAs in Northern Ireland, and then aggregated to towns and TTWAs (that is, the building blocks). The productivity data have been calculated to be consistent with the data for local authorities and ITL geographies within the Office for National Statistics (ONS) Subregional Productivity publication. Further explanations of productivity, and towns and cities are in the information section of the dataset.

These data consist of the following tables:

• Table 1: Travel-to-work areas (TTWAs) – total GVA (£ millions)
• Table 2: TTWAs GVA per job filled (£)
• Table 3: Towns and cities – total GVA (£ million)
• Table 4: Towns and cities – GVA per job filled (£)
• Table 5: Health board areas for devolved administrations – total GVA (£ million) (excluding England)
• Table 6: Parliamentary constituencies (Westminster Government) – total GVA (£ millions)
• Table 7: Devolved administrations’ constituencies – total GVA (£ millions)
• Table 8: Highlands and Islands area offices – total GVA (£ millions)

Gross value added

Balanced UK regional gross value added (GVA(B)) is a measure of output. This measure “balances” the income and production approaches to measuring the economy into a single estimate at a regional level. The balanced measure of regional GVA has National Statistics status following a review by the Office for Statistics Regulation.

Nominal versus real gross value added data

The balanced GVA(B) data used in this article are measured at current basic prices, which includes the effect of inflation. “Real terms" (constant prices) GVA is in chained volume measure (CVM), with the effect of inflation removed.

Gross domestic product

A measure of the economic activity produced by a country or region. Gross domestic product (GDP) growth is the main indicator of economic performance. There are three approaches used to measure GDP:

• the output approach
• the expenditure approach
• the income approach

UK geographies

Subnational GVA data published in this article are a product of disaggregation of local authority level GVA to Lower layer Super Output Areas (England and Wales), Data Zones (Scotland) or Super Output Areas (Northern Ireland). These are then aggregated upwards to higher levels of geography like Middle layer Super Output Areas (England and Wales), Intermediate Zones (Scotland) and district election areas (Northern Ireland).

Our data sources include regional gross value added (GVA) data, Value Added Tax (VAT) turnover data, the Inter-Departmental Business Register (IDBR), the Business Register and Employment Survey (BRES), and census data.

Regional GVA data are published to reflect the economic activity in regions of the UK. Regional GVA is presented in current basic prices, which means that it does not include the effect of market taxes and subsidies (such as VAT) that are within the measure of gross domestic product (GDP), and it does not account for the effect of inflation.

The Value Added Tax (VAT) turnover dataset is HM Revenue and Customs (HMRC) Value Added Tax (VAT) turnover (administrative) data that are used within the compilation of the output approach to gross domestic product (GDP(O)).

The IDBR is a comprehensive list of UK businesses used by government for statistical purposes. It is fully compliant with the European Union regulation on harmonisation of business registers for statistical purposes (EC Number 177/2008). It provides the main sampling frame for surveys of businesses carried out by the Office for National Statistics (ONS) and other government departments. It is also an important data source for analyses of business activities.

The IDBR is updated from four main sources: Value Added Tax (VAT), Pay As You Earn (PAYE), ONS business surveys and Companies House, and Dun and Bradstreet. The updates occur in various frequencies from daily to annually. Because the important PAYE update is quarterly, a shorter time period analysis of business creations and closures would be very volatile. As such, this source is best suited to quarterly publications.

All data on the IDBR are treated as “Official Sensitive” and are protected by the Code of Practice for Statistics and by specific legislation.

The Business Register and Employment Survey (BRES) is the official source of employee and employment estimates by detailed geography and industry. It is also used to update structural information on the IDBR, the main sampling frame for business surveys used to produce UK official statistics.

The UK census results show population estimates of the usually resident population for the UK, constituent countries and all local authority areas (or their equivalent) by various social and economic characteristics. The census is run by the Office for National Statistics (ONS) in England and Wales, National Records of Scotland (NRS) in Scotland and the Northern Ireland Statistics and Research Agency (NISRA) in Northern Ireland.

Details of the policy governing the release of new data are available by viewing the Code of Practice for Statistics, or you can email the Media Relations Office.

We highlight three areas for future development.

Future development of lower-level geography gross value added (GVA)

Alongside this article, we have launched a consultation with the aim of continuing to improve the availability and quality of subnational statistics. Our future plans will be driven by data availability and the outcomes of such stakeholder consultations.

In the future, we may produce constant price estimates of GVA (real GVA). Existing local authority estimates are provided in both current basic prices and in constant terms (as chained volume measures) with the effect of price inflation removed. Because chained volume measures are intrinsically non-additive, they cannot be derived simply by summing component areas, making the methodology more complex to develop and apply. We are also considering estimates of gross domestic product (GDP) and a potential industry breakdown at the Middle layer Super Output Area (MSOA) and intermediate zone (IZs) level.

Exploring other variables and administrative data sources

The next phase of the development work is linked with further exploitation of administrative data to produce more timely and granular indicators across other economic indicators.

We are considering new sources of income data to break down gross disposable household income (GDHI) to lower-level geographies. We will also explore the potential of financial transactions data to inform our analysis of household final consumption expenditure (HHFCE). We will exploit available data sources (for example, administrative or survey data) to improve the data coverage and breakdown using a similar approach to GVA disaggregation.

Creating an interactive online data portal to explore subnational statistics

Future work will also involve developing a new “Explore Subnational Statistics” service. The aim is to be a one-stop shop to find, analyse, visualise, and compare subnational data and statistics, organised by standardised geographies and with the flexibility for user-defined areas.

The GSS subnational data strategy sets out the vision for the “Explore Subnational Statistics” service. The strategy presents a series of commitments towards the ambition of delivering better, timelier, and more granular subnational statistics.

Our results have shown that it is possible to specify bespoke areas for analysis below the local authority level. We encourage our stakeholders to use the building blocks to build different levels of geography for analysis and to inform us of the challenges they encounter through the consultation process and/or by emailing our team: subnational@ons.gov.uk.