Chapter 8 Subnational Statistics 

Richard Prothero and Paul Swinney

How much does the economic performance of the UK vary at the subnational level?

How do we measure subnational statistics and does this differ from measuring national statistics?

What are the measurement challenges of analysing figures at a subnational level and how this can be crucial to interpreting subnational data?

How can we compare subnational estimates across countries?

How does using subnational data help inform policymaking?

8.1 Introduction

If you board a Caledonian Sleeper at London Euston at 21.15 on a weekday evening, then by 08.42 the next morning you can alight in Scotland’s “highland capital”, Inverness, 914 kilometres away. The journey reveals the extremes of the British landscape, from the great city of London to the spectacular mountains of the Scottish highlands.

If you peer out of the window of your sleeper cabin you will see natural and physical beauty, but much too about the nation’s economy. Crucially, it shows that the UK economy is not flat, but clustered in very specific places.

The start and end points of the train journey illustrate this. In 2016, there were more than three times as many jobs in Camden (358,000), the London borough in which you boarded, than in the whole of the Highlands (111,000), despite Camden being just 0.08% of the physical size of the highlands.

Meanwhile, the towns you pass through also illustrate the varying fortunes of different places. Between 2009 and 2017, employment in England and Wales grew by over 10%, but this employment growth was not shared evenly. For example, the train stops at Crewe, which had a 12% decline in employment during this period, and then again at Preston, where employment declined by 5%.

ONS Resource

The ONS release, Employees in the UK by region: 2021, taken from the Business Register and Employment Survey (BRES), is the official source of employee and employment estimates by detailed geography and industry for the UK.

For data specifically on towns, the article ‘Understanding towns in England and Wales: an introduction’, published in 2019, provided both employment and population change data for almost 1,200 urban areas.

The article ‘Understanding towns in England and Wales: spatial analysis’ is the third in a series of towns articles being produced by the Centre for Subnational Analysis at the Office for National Statistics (ONS). It is an expanded update of the first article from 2019. A coastal towns output, Coastal towns in England and Wales: October 2020, was published in October 2020.

This clustering of people and economic activity is what makes subnational statistics so important. Economic statistics tend to focus on the performance of countries, while very little is said about where individuals and businesses locate within countries. Yet, the location has a large bearing on the wages of those individuals, the profits of those businesses and the performance of the national economy.

Subnational economic statistics provide insight on the places where we live and the places where we work. They describe the disparities that exist across different parts of the country in jobs, incomes and productivity. They are used to examine regions and cities, towns, rural areas and neighbourhoods. They link data to location, and in doing so allow us to investigate the factors that influence economic outcomes across different areas, and understand how location itself can influence economic outcomes. And when viewed as a whole, this patchwork of subnational statistics can help us understand how the UK is put together and the performance of the national economy. This helps to explain why the role of local regions has been more prominent in contributing to a better understanding of the UK economy.

This chapter shows how subnational statistics are collected and produced and the unique issues involved. It details some controversies and the anticipation that administrative data will help expand the amount of subnational data available. Regional statistics are a fascinating topic for study and application, and are useful as a basis for policymaking. The demand for local economic data in the United Kingdom has increased in recent years as more powers have been devolved to Scotland, Wales and Northern Ireland. Devolution of powers within England and the creation of combined authorities has also created renewed interest for data for the English regions and local geographies.

8.2 Measuring regional statistics

Many economic statistics are quoted at the level of a nation state. We look at GDP growth in the United States, inflation in France and unemployment in the UK. While these are important metrics for understanding economic performance, as aggregate measures they do not tell us how different geographic areas within a country are performing.

There can often be large regional variations in a country. For example, Figure 8.1 shows the variation in labour productivity in the UK.

Figure 8.1 Labour productivity (GVA per hour worked, NUTS 3) in the UK, 2017

Labour productivity (GVA per hour worked, NUTS 3) in the UK, 2017

Office for National Statistics – Understanding spatial labour productivity in the UK, Figure 1

Figure 8.1 gives us a clue that looking at the sub-national level will provide two main benefits:

  • Economic performance will vary. Statistics representative of the national picture can mask significant variation between different local areas.
  • We understand national issues by looking at what is going on across the country. For example, understanding the UK’s aggregate productivity performance is aided by looking at how productivity varies between regions.

But how do we collate this data? While the demand for economic statistics at a local level is growing, the way statistics are compiled has put limitations on what data can be produced.

8.2.1 Survey or administrative data?

Many economic statistics have traditionally been produced using surveys on a representative sample of the population. If the sample is drawn from a representative group of people, then the overall results should be a fair estimate of the whole population. In the UK the largest regular surveys will have a sample size of tens of thousands of individuals or businesses, but suffer two problems when it comes to looking at sub-national data:

  • The sample might not be representative of each of the geographic areas within a country. We can collect data in a way that provides a nationally representative picture, but it is not representative for subnational areas. This can be overcome by “stratifying” the survey so that it is representative of these different subnational areas. But this may cause a second problem.
  • The sample size might be too small to provide reliable estimates. Sometimes a survey may only give reliable estimates for regions but not for smaller geographies such as local authorities. At other times it might be difficult to combine geographic disaggregations with other disaggregations such as occupation or ethnicity. Individually, each is possible, but combined the sample sizes become too small.

The drawbacks of survey data can be overcome by supplementing or replacing it with administrative data such as tax information. For example, the use of Value-Added Tax (VAT) data collected by HM Revenue and Customs (HMRC) has allowed the Office for National Statistics to produce much more detailed information on local economic output. In addition, Claimant Count information collected by the Department for Work and Pensions (DWP) provides detailed geographical data on unemployment, while detailed house price data is available from administrative data published by the Valuation Office Agency (VOA).

8.2.2 Allocating a location to data

Whether using survey or administrative data, a key to the production of subnational statistics is that there must be some location data included in the data collection phase. At its simplest, this location data allows for a relatively straightforward aggregation of all individuals or firms in an area to enable publication of subnational statistics.

gross value added (GVA)
Gross Value Added (GVA) is the value generated by any unit engaged in the production of goods and services. It measures the contribution to the economy of each individual producer, industry or sector. Simplistically it is the value of the amount of goods and services that have been produced, less the cost of all inputs and raw materials that are directly attributable to that production.

However, particularly with business statistics, it can get more complicated. Consider a large firm that operates across many different sites across the country. To produce regional gross value added (GVA) data, we would want to know how much of its value added the firm adds in each of its locations, so that we can assign this value to the appropriate region. But most firms do not have data on value added for each individual local unit, and this means the data on variables such as turnover and intermediate consumption is only collected at a level which aggregates different local units together (known as reporting unit level, which may include the head office of the enterprise). Figure 8.2 shows this complex relationship.

Figure 8.2 The relationship between local units, enterprises, enterprise groups, and reporting and administrative units

The relationship between local units, enterprises, enterprise groups, and reporting and administrative units

When this occurs, statisticians have a choice. They could allocate all the GVA to the location of the reporting unit, but this could become very misleading (imagine if all of the value added of a major supermarket chain in the UK were allocated to a single local authority rather than spread across all its sites). Instead, in these cases, there is usually some further modelling of the data undertaken to allocate it across the local units of the firm using data on other regional variables such as employment.

gross value added income GVA(I)
Gross value added income is GVA measured at current basic prices, which include the effect of inflation, excluding taxes (less subsidies) on products (for example, Value Added Tax). This involves adding up the income generated by UK resident individuals or corporations in the production of goods and services. It is calculated gross of deductions for consumption of fixed capital, which is the amount of fixed assets used up in the process of production in any period.
gross value added production GVA(P)
Gross value added production is GVA measured at both current prices and in chained volume measures (CVM). It is calculated for a given reference period as the total value of all goods and services produced (output), less goods and services used up or transformed in the production process, such as raw materials and other inputs (intermediate consumption). The production approach to compile GVA is conceptually equivalent to the income approach, but allows deflation of current prices to produce constant price measures, since the production components relate to goods and services that can be broken down into price and volume indices.

How it’s done Allocating business data to a location

The case of Regional GVA can provide a useful example. In 2018, ONS examined the sources of firm data by splitting the observed data into three different categories, depending on the approach taken to using it in the GVA calculations.

The first type of data were those that were directly observed at a regional level, collected in a way that could be immediately and wholly assigned to a single region; at the aggregate UK level, there was around 32% observed data in gross value added income GVA(I) and around 23% in gross value added production GVA(P).

The second type of data are not directly observed, but are estimated using sampling and weighting techniques common to all sample surveys; at the aggregate UK level, this accounted for around 51% estimated data in GVA(I) and around 49% in GVA(P).

The final type are the data that are modelled to provide regional estimates, most often by the apportionment of data collected for a larger area using some regional indicator; at the aggregate UK level, there was around 17% modelled data in GVA(I) and around 29% in GVA(P).1

The key point here is that there are some added difficulties to collecting and producing statistics for regional data compared with national data, principally reflecting the challenges of identifying where in the UK that economic activity took place, and this is particularly the case for business data. However, there are methods and approaches that have been developed over many years to address these issues and ensure that Regional GVA data provides a coherent indicator of local economic activity.

Furthermore, improvements continue to be made to both the methods and outputs. For example, in December 2019, for the first time ONS began to produce Regional Gross Domestic Product (GDP) data as well as the long-standing series of Regional GVA.

ONS Resource

The current national statistic for regional economic output from ONS combines estimates from gross value added income (GVA(I)) and gross value-added production (GVA(P)) to produce a balanced measure of regional GVA, known as GVA(B). The GVA(B) data is published alongside data on regional GDP (see Measuring the data).

8.2.3 What data can be measured?

This book has covered a wide range of economic statistics, and many of these have an important subnational component. But not all economic data are always available at a local level.

Figure 8.3 summarises the main economic statistics and what is and is not available.

Statistic Conceptually relevant UK Data Available Links (April 2023)
GDP and economic output Yes Yes. Data down to local authority level. Annual data

Quarterly Data
Labour Market Yes Yes. Administrative data available at low levels of geography. Most survey data available at NUTS1/regional levels with some estimates available down to local authority level. Monthly Regional Labout Market Update

Employees in the UK (BRES)
Inflation Yes No. Headline measures of consumer price inflation are not currently available below national level.
Trade External trade possible, but intra-country trade not well defined. Import and export data of goods and exports of services available at NUTS1 regional level with some estimates at smaller geographies such as city regions. Goods
Regional trade in goods – monthly

Regional trade in goods – quarterly

Services
International trade in services by subnational areas of the UK
Productivity Yes Yes. Labour productivity estimates available down to local authority level. Regional Productivity

Subregional Productivity
Sectoral Accounts Most concepts make sense at national level only No
Business Statistics Yes Yes. Available down to local authority level. Business Location

Business Demography

Business population
Income and Earnings Statistics Yes Yes. Available down to local authority level. Annual Survey of Hours and Earnings (ASHE)

Gross Disposable Household Income

Figure 8.3 The regional data that can be measured for the UK

The regional data that can be measured for the UK

8.2.4 Residence or workplace?

Is the data showing the outcomes of the residents who live in a geographical area or the outcomes of workers who are employed in the area? For London, for example, a high level of commuting can mean some very large differences between data measured in these two ways. As we will see later, this can also be a source of confusion in interpreting GVA per head data.

When using subnational economic data, you should seek to understand whether any specific data is referring to the location of individuals’ residential addresses or whether it is referring to the location of a workplace. Some datasets only provide information on one of these location types. For example, ONS Regional Gross Value Added and Productivity data are available on a workplace basis only, while the income data in the Gross Disposable Household Income dataset is on a residential basis only. Some datasets, such as the Annual Survey of Hours and Earnings (ASHE), can provide both.

How it’s done Understanding workplace versus residence data

The first buildings of the Canary Wharf business district of London were completed in 1991. It had previously been part of London’s docks, and the final working dock in the area had closed in 1980. It is in the London borough of Tower Hamlets, traditionally one of the poorest in the country.

Canary Wharf is now the location for bank headquarters, international law firms and media companies. But many of those highly-paid service industry jobs are filled by workers who commute into the area. As such, Tower Hamlets provides a good example of how there can be large differences between regional data when considered on a workplace basis compared with on a residential basis.

The metrics of Average earnings (workplace basis) and GVA per hour worked show data that relate to people who work in Tower Hamlets and businesses located there.

Metric Workplace Residential
Income metric (UK=100, 2020) 179.2 (GVA per hour worked) 122.1 (Gross Disposable Household Income)
Average earnings (Median Weekly Pay – Full time, 2022) £1,016 £807

Figure 8.4 Metrics for Tower Hamlets Local Authority on a workplace and residence basis, 2020 and 2022

Metrics for Tower Hamlets Local Authority on a workplace and residence basis, 2020 and 2022

Figure 8.4 illustrates that the borough is one of the most successful business locations in the country, with labour productivity (GVA per hour worked) 79% above the UK average and average median earnings for its workers of £1016 per week, which is 59% above the UK average.

Gross disposable household income and average earnings (residential basis) are, by contrast, measures related only to the individuals and households resident within the borough. The earnings of commuters into the borough do not make it into these metrics.

As a result, the median earnings of a Tower Hamlets resident is £807, 26% above the UK average, but over £200 per week lower than the median of workers in the borough. Meanwhile, Gross Disposable Household Income per capita of residents in the borough is 22% above the UK average, so not fully reflective of the high business productivity in the borough.

So Tower Hamlets is a case in which many high earners who work in the borough commute in from homes elsewhere, leading to higher workplace than residence earnings in the borough. In the local authorities where these commuters live, the opposite will typically be the case, with average resident incomes higher than average workplace incomes.

8.3 Understanding geography

How should subnational data be presented? Is the interest in cities, or regions, or rural areas? What is the correct definition of each of these? Alternatively, maybe there are some researchers interested in analysing data for travel to work areas, or functional economic areas, or users with an interest in using geodemographic classifications in place of contiguous geographies. It can get quite complicated. We explain some options below.

8.3.1 Administrative boundaries

local authority
Local government structure varies across the UK. In England, some areas have two-tier authorities with service provision shared between country councils and district, borough or city councils. Other areas of England (particularly metropolitan areas) have a unitary tier. In Scotland, Wales and Northern Ireland, local government is based on unitary tier authorities.
Local Enterprise Partnership (LEP)
LEPs are voluntary partnerships between local authorities and businesses in England set up in 2011 by the Department for Business, Innovation and Skills to help determine local economic priorities and lead economic growth and job creation within the local area. Enterprise partnerships have also been set up elsewhere in the UK.

The most common request for subnational data is to match administrative boundaries. For the UK, the most commonly used boundaries are those of local authorities. However, a range of other administrative geographies are also in regular demand for economic data, including Combined Authority boundaries covering city regions, and Local Enterprise Partnership boundaries (which in England up to 2020 have been uniquely problematic for statistics because they overlapped).

ONS Resource

This source on Administrative geography explains the often confusing hierarchy of areas in the UK.

modifiableareal unit problem (MAUP)
The modifiable areal unit problem (MAUP) is a source of statistical bias that can have a significant impact on the results of statistical hypothesis tests into geographic areas. The resulting summary values (e.g. totals, rates, proportions, densities) are influenced by both the shape and scale of the aggregation unit.

The advantage of administrative boundaries is fairly clear in that they match the area for which local policymakers have policymaking responsibilities. However, for research purposes such geographies can cause problems, as the shape and scale of boundaries can have an impact upon the statistical results. For example, if the local authority boundary of one city is tight to its urban form, but the local authority boundary of the neighbouring city includes its surrounding suburbs, villages and countryside, then any comparative data analysis will be affected by these different boundary shapes. This is known as the modifiable areal unit problem.

8.3.2 Statistical geographies and classifications

There are a number of ways to minimise this problem. One is to use a statistical geography that has been designed to capture a certain concept or type of place. UK Travel to Work Areas, for example, are designed to capture self-contained labour market areas, while Built-Up-Areas capture only the urban extents of cities, towns and villages, using satellite imagery in their creation.

Another possibility is to use a classification in which either neighbourhoods or local authorities have been classified into different types of areas. The UK Rural Urban Classification is an example of this approach and is used extensively for analysis of rural areas by the Department for Environment, Food and Rural Affairs (Defra) and researchers interested in rural issues. Geodemographic classifications, meanwhile, are particularly favoured by geographers, and include the Output Areas Classification and the Classification of Workplace Zones, alongside a number of private sector versions.

ONS resource

The 2011 rural-urban classification provides a rural/urban view of datasets.

The 2011 area classifications provide information on the classifications used to support the production of official statistics.

The examples given above start to give an indication of how many different geographies might be of interest to different sets of users. And there are many more, as this summary schematic of the different geographical breakdowns of the UK shows.

Given all the possibilities, the aim of statistics producers is to try to be as flexible as possible when designing surveys and data architecture, and in the dissemination tools that they provide, so that users have some flexibility to produce analysis for a wide range of geographies depending on specific needs.

Organisations such as the European Union, OECD and United Nations have worked with nation states to produce a range of definitions for international comparisons, and some of these are used quite regularly in the UK.

Nomenclature of Territorial Units for Statistics (NUTS)
A geographical system, according to which the territory of the European Union is divided into hierarchical levels. The three hierarchical levels are known as NUTS 1, NUTS 2 and NUTS 3.

The most commonly used of these is the NUTS (Nomenclature of Territorial Units for Statistics) geography (see NUTS – Nomenclature of territorial units for statistics: Background).

This is available at three levels, with NUTS 1 including Scotland, Wales, Northern Ireland and nine English Regions (including London, South West, North West, etc.). A second level, NUTS 2, is also commonly used for international comparisons, and this splits the UK into 41 subregions, while NUTS 3 adds a further disaggregation to 179 subregions. There are a series of rules to try to ensure similar-sized areas are defined across Europe.

Under a different name, territorial levels (TL), these same geographies are also used by the OECD alongside boundaries for non-European OECD members, as described in the OECD Territorial Grids.

Meanwhile, Eurostat and the OECD have also combined to create a Functional Urban Area geography that captures major cities and their commuting areas (see Functional urban areas by country).

At the time of writing, this definition was also being considered by the United Nations as a globalised city definition for international statistical comparisons.

A fuller discussion of these international geographies are available at International, Regional and City Statistics on the ONS website.

8.4 Improving regional statistics through new data sources and flexible geographies

In 2015, the then Chancellor of the Exchequer commissioned Professor Sir Charles Bean to carry out a review of the economic statistics produced by ONS. One of the main strands of this review was to examine the need for and provision of regional economic statistics in the face of growing devolution.

The Bean Review made various observations and recommendations, of which the main messages relating to regional statistics were:

  • there was a need for more timely regional statistics
  • there was a need for greater flexibility to provide statistics for different geographic areas
  • that we should look to make more use of administrative data sources in the compilation of regional statistics.

To meet this challenge, ONS began a four year Devolution Programme in 2016 to improve UK regional statistics. This has led to a number of new outputs as well as improvements to the methods, timeliness and geographical disaggregation in many existing products.

8.4.1 Devolution

In November 2014, the UK Government and the Greater Manchester Combined Authority signed a devolution agreement devolving policy powers and responsibilities to Greater Manchester. This agreement, which included the adoption of a directly elected mayor, marked a new step in the process of devolution across the UK. Prior to this there were devolution agreements for the nations of Scotland, Wales and Northern Ireland and also for Greater London, but none for any other city or region in England. However, the agreement with Manchester marked a step change in policy such that, by May 2016, a further 10 devolution agreements had been agreed, mostly with city regions.

These changes, taken together with the recommendations of the Bean Review, led to the creation of the ONS Devolution Programme to improve a range of UK regional statistics. As an example of the work, a new annual publication output was developed that provides data on Country and Regional Public Sector Finances. This provides users with information on what public sector expenditure has occurred, for the benefit of residents or enterprises, in each country or region of the UK, and information on what public sector revenues have been raised in each country or region, as well as the balance between them. This balance, known as the net fiscal balance, is the gap between total spending (current expenditure plus net capital expenditure) and revenue raised (current receipts), which at the UK level is equivalent to public sector net borrowing.

The publication is therefore providing new evidence on the geographical breakdown of the UK’s fiscal deficit (or surplus). This information is important for understanding devolution policy and prospects, and the publication highlights one way in which the Devolution Programme has been able to respond to user needs and provide new data and insight.

ONS resource

Data on country and regional public sector finances, giving public sector revenue, expenditure and net fiscal balance on a country and regional basis, can be found at Country and regional public sector finances; financial year 2021.

These improvements continue, as does the need to continuously examine ways in which administrative or other data sources could help improve the timeliness or granularity of subnational economic data. Regional statistics can take advantage of new data sources to improve the geographic match of statistics to real world requirements. If they lead to more disaggregated data being available, then this will allow more flexible geographic outputs and analysis.

Many of the regional economic statistics published by ONS have traditionally been based only on surveys. However, the sample size of such surveys acts as a limit to the amount of disaggregation that can be provided in the statistics. There is, therefore, a lot of interest, and potential, in using alternative data sources.

8.4.2 Administrative data

In 2018, the ONS Regional GVA output included some of the first statistics at ONS to be published making use of administrative VAT records as part of the production process. In that case they were supplementing existing data sources.

ONS resource

Statistics using VAT data are provided at Regional economic activity by gross domestic product, UK: 1998 to 2020.

New quarterly measures of regional GDP are also being published with VAT as the principal data source; see GDP, UK regions and countries Statistical bulletins.

These outputs are a clear illustration of the potential of administrative data both to expand the range of regional statistics available (the VAT data allowed the publication of local authority GVA data for the first time in 2018) and also to improve on their timeliness (for example the creation of the quarterly regional GDP publication).

8.4.3 Big data

Data from commercial sources has been used by ONS in a pilot project to build small area housing rental prices, while the commercial data of banks is being investigated to potentially provide early indicators of regional economic change from their daily transactions data. Mobility data from telecommunication and internet companies are also potential sources of new subnational data.

Because these data are collected and owned by private sector companies, there are barriers to be overcome in terms of using the data in official statistics. However, the possibilities are being investigated and there is significant potential amongst such data sources for the provision of more timely and granular subnational data.

8.4.4 Geospatial data

Geospatial data encompasses many different elements of locational data. This includes the geographic boundaries discussed in sections 8.3.1 and 8.3.2, but in addition it also captures, amongst other things, data on the natural environment such as waterways or flood plains; data on transport networks such as road and rail; data on land use types and habitats; and data on the built environment such as the location and type of buildings or of public infrastructure.

Geospatial data has a great deal of potential in expanding the range of regional data available and in adding further insight to economic data. For example, by combining the land use data collected by Ordnance Survey with economic and social data collected by ONS, new data on high streets were published in a joint project in 2019, with the approach potentially very useful for other similar projects to define subnational data more sharply to policy topics of interest. Geospatial approaches, for example using satellite data, can also help to define new approaches to producing regional data.

ONS resource

High streets in Great Britain provides information on the mapping of the location and characteristics of the high streets.

It is unlikely that greater use of these alternative data sources will completely negate the need for survey based statistics. But they do provide opportunities to produce new datasets that are more disaggregated and more timely, and that can be targeted more directly for understanding local areas or local policy topics. It will then be possible to reorganise the survey based data to compliment these new data sources and fill in the gaps where administrative data, big data or geospatial data are not available.

8.4.5 Flexible geographies

Traditionally, many published statistics have only been available for UK NUTS 1 regions, or sometimes for local authority. However, these can be a blunt instrument for assessing policy questions related to location. If the subject of interest is the health of coastal towns, or the future of suburbs, then data provided at local authority level will not be particularly helpful.

As larger data sources become available allowing more disaggregated geographic analysis to take place, attention is turning towards how best to identify different geographies of interest.

Have a look around the Centre for Cities data tool.

One example can be found in the work of Centre for Cities, who have moved beyond just an examination of overall city data to examining the different aspects of a city. In particular, they have run analysis that analyses a city based on data for the city centre, its suburbs and its hinterland. In order to create as consistent a definition as possible, the city centres were defined as a circle around a central point in the heart of the city centre of each city. The radius of this circle varies according to the size of the city. The suburbs were defined as the rest of the Primary Urban Area. Hinterlands were defined as the average distance that those workers who live outside of the city commute into the city. And rural areas were the remaining land not covered in the above definitions.

Figure 8.5 The varying roles of different parts of the UK in the national economy, 2015

The varying roles of different parts of the UK in the national economy, 2015

Business Structure Database

For producers of statistics, the wide potential uses of the regional data mean that the aim should be to be as flexible as possible in their production. Statistical agencies will continue to publish statistics for the most popular and widely used geographies and seek to develop geographies with wide analytical interest to users. But there will be many times where users of the data want to build statistics to their own geographical requirements, and publishing data at a small area level gives the opportunity for these users to build the data back up to their geography of choice.

8.5 How can we measure inequality between regions?

The chapter has shown that in compiling and disseminating subnational data and analysis, there are a number of choices that have to be made either in the process of compiling the statistics or in the dissemination of the statistics. In most cases, researchers tend to agree broadly on the appropriate methods and metrics to use, but there can be cases where disagreements arise and controversies break out.

One question that tends to lead to much such discussion is how large is regional inequality in the UK compared with other countries? It is a reasonable question, but one that can be surprisingly difficult to answer, with a number of booby traps lying in wait for the unwary.

The two main issues that need to be considered to avoid the traps are both issues that have been mentioned earlier in the chapter. In the case of regional inequality analysis they become important. The issues to consider are:

  • The difference between where people live and where they work caused by commuting.
  • The size and types of the geographic areas being compared.

8.5.1 Distortions from commuting

Measures of economic output tend to be measured on where people work, but commuting means people do not necessarily live where they work. This can cause problems if a metric that mixes residence and workplace data is used. GVA per head (or GDP per head) is an example. The typical measure of subnational economic output is GVA, which is a workplace based measure. But if you divide it by population, then you are dividing a workplace measure by a residential measure, and the result can be distorted by commuting.

Take the extreme example of the City of London with its resident population of less than 10,000, but where commuting means hundreds of thousands come to work every day. The resulting output per resident total for the City of London is huge, but essentially meaningless. It provides neither an accurate measure for the economic productivity of the area or for the household incomes of its residents.

In many parts of the country, commuting takes place within the locality, so the effects are small. But in places like London, the effects can be large. For this reason, in most cases we do not recommend using the GVA (or GDP) per head measure for economic performance or inequality comparisons. Rather, measures of labour productivity – either GVA or GDP divided by the number of workers – are likely to give a better indicator of economic performance.

8.5.2 Distortions from area size and geographies

The more areas you divide the country into, the more likely you are to see high inequality. This is particularly important when making international comparisons.

To illustrate this point, let us consider Bishops Avenue in London. This road contains some of the most expensive houses in London, with prices measured in tens of millions. The average price of a house here would be astronomical – perhaps a hundred times the UK average. But if we took the wider area – say the whole borough of Barnet – the average would be much lower. The smaller the size of the geographical area, the more one is likely to see extreme results. This applies to measures of economic output and income as well as house prices. To illustrate this, the table below shows data comparing productivity levels in terms of ratio of the highest to lowest average GVA per hour worked in three tiers of geography from the highest NUTS 1 to the lowest NUTS 3 region.

NUTS 1 NUTS 2 NUTS 3
Number of areas 12 41 179
Inter-quartile range 10.7 17.7 18.7
Variance 249.3 352.1 418.4

Figure 8.6 Ratio of highest to lowest UK regional gross value added per hour worked, 2004–2018. Note: Each region and sub-region is measured compared to UK overall productivity set at 100, 2017.

Ratio of highest to lowest UK regional gross value added per hour worked, 2004–2018. Note: Each region and sub-region is measured compared to UK overall productivity set at 100, 2017.

ONS Regional and Sub-regional Productivity, February 2019.

How it’s done International regional comparisons

When comparing between countries, you need to ask whether the data are truly comparable. When we compare UK data with that of, say, France, we may find we are not comparing the same-sized areas.

For example, if you examine the NUTS 2 data available on the Eurostat website, you will find that in France, Paris is not separately identified, and the data provided is for the wider Île-de-France, which has a population of over 12 million, covering the Greater Paris area.

Now look at the UK data. Using the same NUTS 2 geography, London is broken up into five areas each with a population between 1.1 million and 2.2 million.

As we are measuring smaller units in the UK, this is likely to give you more extreme results and suggest higher regional inequality. In other words, this higher apparent regional inequality might just be because of different-sized areas. We would therefore want to check the results by looking at alternative geographies that provide more similarly-sized areas to get a better comparison, or provide a statistical geography on a consistent conceptual basis (for example the Metropolitan Areas geography).

8.6 Regional policy

The amount of policy questions that can make use of regional data is seemingly endless. Economic data can help a local area understand the performance of its economy and how this is influenced by the types of businesses in the area, the qualifications of its residents and the impact of transport links. It will also provide information on the incomes of its residents and the numbers of residents unemployed or underemployed.

Given the huge interest, there are many people and organisations using regional data to provide analysis, ranging from civil servants working within local and national government to think tanks specialising in regional issues to academics focusing on testing hypotheses or providing policy advice.

Figure 8.7 is an example of research results from the Centre for Cities on the productivity performance of firms, comparing cities in the Greater South East of England with those in the rest of the UK.

Figure 8.7 Productivity and employment shares of different industries in cities in the Greater South East and the rest of Britain, 2015

Productivity and employment shares of different industries in cities in the Greater South East and the rest of Britain, 2015

ONS Regional Gross Value Added (Income Approach) by Local Authority in the UK; ONS Business Register of Employment Survey

There are two things to note from the figure:

  • Sectors that tend to be more export-focused in their activities (left) are more productive than local services sectors (right). This is true for both types of city.
  • The variation in the performance of exporters drives overall differences in productivity. While there is little difference in the productivity of local services between the two cities, there is a clearer difference in the productivity of exporting industries. The biggest divergence is for finance. Beware, though, for the reasons given in Section 6.4.3 earlier, of too easily accepting the correlation as being causal. Exporting might assist productivity. But it might also be that firms who have high productivity for other reasons also tend to be exporters.

Understanding the drivers of regional productivity differences is a major topic of interest to policymakers, with both ONS and Centre for Cities among those who have been exploring the subject.

8.6.2 Using data to inform local industrial strategies

Policymakers will typically use information from all these sources in helping reach policy decisions. For example, Manchester has developed the evidence base on its local economy to help inform its local policymaking and more recently to develop a local industrial plan.

In 2009, the Manchester Independent Economic Review was published, which was an extensive report into the state of the Greater Manchester economy that informed policy in the city region in the years that followed.

You can read the many reports of the Manchester Independent Economic Review.

Read the Greater Manchester Independent Prosperity Review.

In 2019, the work was updated in the Greater Manchester Independent Prosperity Review and its associated background reports with their focus on productivity, innovation and global competitiveness, skills and employment, and infrastructure.

The evidence from this review then informed the Greater Manchester Local Industrial Strategy, published in June 2019, which is a long-term plan to increase productivity in the city region, developed as part of Government’s commitment to agreeing local industrial strategies across the UK.

Similar evidence and research gathering goes on regularly across the country, whether to inform local strategic economic plans, local industrial strategies, or just to provide local policymakers with some evidence on which to make decisions.

8.6.3 Using data for regional policy evaluation

Using data for regional policy evaluation can be a tricky job, often requiring access to microdata sets or the running of a small survey alongside the building of a framework that allows recipients of a policy intervention to be compared to a comparator group who were not involved. However, when successfully carried out, a project evaluation can provide extremely useful information and feedback to policymakers as they consider future policy interventions.

For example, since 2013, the What Works Centre for Local Economic Growth (WWCLEG) has conducted a review of thousands of existing project evaluations to unearth the most useful on ten policy areas affecting local economic growth, from business support to employment training and innovation. They have also been supporting local practitioners in building robust evaluation procedures to build local evidence bases further and to increase understanding on the impacts of local policy interventions.

The WWCLEG has also produced a very informative guide to using data for local economic policy, with sections on finding data and applying data to local economic strategies, along with ten different case studies. The WWCLEG guide is a useful additional reading for anyone interested in learning more about using regional and local economic data in a policy setting. The link can be found in the further reading section below.

8.7 Summary

Subnational statistics are an important topic of study in their own right. The demand for good subnational data is substantial with policymakers, both local and national, keen for good evidence detailing the economic differences across the country and also evidence on what drives these differences and how policymakers can respond.

To work on, or with, subnational statistics is to work with a range of issues that are rarely a problem at the national level. The sample size of a survey may determine whether there is suitable data for your chosen geography; the way a survey was stratified may affect its reliability; the way data is allocated to location can affect the interpretation of a statistic, as can whether it measures data for residents or workplaces; and the geographic boundaries of comparator areas could mislead the unwary. So, don’t assume working with subnational data is easy. What is true is that working your way through these challenges has a huge potential payoff in terms of being able to investigate and understand the economies of the local areas where we live and work.

8.8 Further reading

Notes