Tablau cyflenwad a defnydd a thablau mewnbwn-allbwn 2019 ar gyfer Cymru: methodoleg amlinellol

This release outlines the estimation procedures, data sources and transformations for the tables and presents an assessment of where Wales-level data are insufficient and further development will be necessary.

This document outlines:

• the estimation procedures for the 2019 Supply & Use Tables (SUTs) for Wales which are published by Welsh Government as Official statistics in development
• the compilation of a 55-sector product-by-industry tables for Wales, detailing our data sources and transformations, and its transformation into Input-Output Tables (IOTs) and the final publication of 55-sector matrices
• an assessment of where Wales-level data are insufficient to adequately report key sectoral and demand variables, indicating that further data development will be necessary for the publication of robust tables

IOTs have historically been published for Wales by The Welsh Economy Research Unit (WERU) at Cardiff University’s Business School, with the most recent iteration prior to 2019 being IOTs for Wales 2007 (Cardiff Business School). Work undertaken within Welsh Government since 2022 builds upon processes and understandings developed during this history but does not use any Cardiff University data or intellectual property.

The document is intended for readers with some knowledge of national accounting and IO concepts. Those wishing to familiarize themselves are directed to Miller & Blair (2009) ^{[footnote 1]}.

This project has greatly benefitted from assistance from civil service colleagues within Welsh Government, other Devolved Governments, and the Office for National Statistics (ONS), and we are extremely grateful for this guidance and assistance.

Structure of this report

Developing the base year and product-industry structure

Explains the framework for the compilation and publication of the Wales SUTs and IOTs.

Initial estimation of components of GVA

Describes the estimation of the Output and components of Value Added in the SUTs.

Estimating output for sectors in Wales

Explains the estimation of intermediate product purchases by our initial 64 industries (latterly aggregated to 55 sectors for publication).

Estimating final demand

Outlines our approach to the (challenging) estimation of the import propensities, by product for each of our industries (within and outside the UK).

Estimating product purchases in Wales

covers the compilation of our final demand elements – the purchasers of Wales’ final products, including households, public and third sector organisations, goods and services sent to export, and those used in gross fixed capital formation.

Estimating the Initial SUT in basic prices

Describes how we confront supply and demand in common (basic) prices.

Balancing the SUT

Describes the balancing process whereby we ensure that supply equates to demand for each product at Wales level.

Creating the Symmetrical IOT & publishing outputs

Covers the transformation from SUTs to IOTs (including the additional challenges specific to this process), and disclosure and aggregation.

The IOT project was undertaken by staff within Knowledge & Analytical Services (KAS) in Welsh Government, supported by a Project Board drawn from within KAS and more widely across Welsh Government, and including the Chief Statistician and Chief Economist. A peer review process checked and triangulated methodological, conceptual and compilation approaches with colleagues from within ONS and across other devolved governments. Quality assurance was undertaken by KAS staff not directly involved in the compilation process.

Compilation of the Wales SUTs commenced in early 2023.  At this time, SUTs and IOTs were available from ONS and the other devolved governments for 2018 or 2019. Given the usefulness of other tables for comparison, and for the UK IOTs in sourcing some ratios not available for Wales, an initial estimate of SUTs (and later IOTs) for Wales in base-year 2019 was considered appropriate. This also avoided the significant difficulties inherent in compiling tables for 2020, the first year of the COVID pandemic.

At UK level, SUTs and IOTs are published for 104 commodity and industry disaggregations, whilst the Scottish Government publish 98, and Northern Ireland Statistics and Research Agency (NISRA) 62 sectors (but with tables estimated at a finer level).

After internal consultation, it was agreed that Wales’ SUTs should be initially compiled for 64 sectors, generally matching ONS ‘A64[i]’ Standard Industrial Classification (SIC) (ONS) and Classification of Products by Activity (CPA, ONS).

Advantages to this level of disaggregation

• A64 is an accepted classification for UK IOTs and SUTs and used by OECD (although rarely published at this level in the UK).
• ONS publish regional GVA for industries which closely match IO64, providing a statistically robust, credible and transparent ‘anchor’ for our initial industry estimations.
• The extant Cardiff Business School IOTs are published at IO64 level, easing comparisons and triangulation.
• Some work on IOTs undertaken by the Fraser of Allander Institute at University of Strathclyde is available in this classification, enabling some inter-regional comparison and ‘sense-checking’.
• Compilation at a more detailed sector level – 98 or 105 for example – would have significantly increased the level of imputation required.

A64 was therefore chosen as the level of disaggregation which was a priori considered most likely to balance usefulness and robustness (See Annex 1). However, due to issues of disclosure and data quality, it was decided to further aggregate some sectors. This set of tables are therefore published at a 55-industry level of disaggregation. This is discussed further in section 9.

ONS typically separates final demand for products and services into 9 final users (separating out for example exports to the EU and to the Rest of the World (ROW), and exports of services). Scottish Government (unlike the UK) separately reports the purchases of central (Scottish) and local government, whilst reporting exports (aggregate goods and services) to the rest of UK (RUK), ROW, and to non-resident households (the last presumably equating largely to inbound tourism). NISRA also provides detail on central and local government purchases, and details exports to GB, Ireland, rest of EU and ROW.

After considering the above, and considering issues with data scarcity, Wales SUTs and IOTs aggregate the purchases of central and local government, and detail exports to the rest of the UK and ROW.

The ONS publishes balanced estimates of Gross Value Added (GVA), and GDP, annually for countries and International Territorial Levels (ITL) 1 – 3 in the UK. These data are published for 79 industries as both ‘chained volume’ measures (in £m) and indices. The estimates of GVA, aggregated to our 64 IO groups (henceforth IO64) provide the key starting point for our estimated Welsh production functions ^{[footnote 2]}. Our estimates of GVA for IO64 in Wales are thus constrained to ONS published estimates in our initial estimation ^{[footnote 3]}.

For consistency with Systems of National Accounts, our IOT requires the disaggregation of industry GVA into its component parts: initially Compensation of Employees (CoE), Taxes-less-Subsidies (T-S) and Other Value Added (OVA, comprising gross operating surplus and mixed income). ONS publishes annual estimates of balanced UK regional gross value added for UK ITL1 regions by 9 detailed components and for 31 industries (herewith ONSRA31). Sixteen of these industries match with IO64, and our estimates of GVA components (CoE, OVA and T-S) were, for these industries, taken directly from the ONS published 2019 estimates.

The remaining 48 industries required further, indirect estimation of GVA components. Initial focus was on estimating Compensation of Employees by IO64 as firstly, this is usually the largest single proportion of GVA, and secondly the most relevant for regional accounting and future modelling purposes.

Estimation process

1. For each ‘set’ of unknown IO64 (uIO64) sectors with a wider ONSRA31, aggregate CoE was constrained to the ONSRA31 total.
2. Full time equivalent (FTE) ^{[footnote 4]} employment for each uIO64 was estimated from analysis undertaken via Nomis of the ONS Business Register & Employment Survey (BRES), building from 5-digit SIC.
3. Median weekly gross wages for each uIO64 at UK level was estimated from the Annual Survey of Hours and Earnings for 2019 (ONS). UK-level data (rather than Wales-level) were used as the sample sizes and robustness of the Wales-level data were not sufficient at the industry level for our purposes.
4. For each uIO64 within the wider ONSRA31, estimated FTE employment was multiplied by the UK ASHE median weekly wage to approximate weekly compensation of employees for that uIO64 in Wales
5. ONS-published CoE for each relevant ONSRA31 was allocated to each constituent uIO64 based on the wage weighting resultant from 4. above.

We arrived at our initial estimate of CoE for our remaining 48 industries. The allocation process led to some significant differences between Wales and the UK related to the importance of CoE in overall GVA. However, this is also true of a number of wider ONSRA31 sectors where components are directly available. In Mining & Quarrying for example, CoE is 48% of GVA in Wales but only 20% for the UK, likely driven by very different activity mixes at these two spatial scales.

The allocation of CoE to IO64 groups by the wage-weighted average employment led to some anomalies ^{[footnote 5]}. Notably, for some sectors CoE was estimated to be unfeasibly high in comparison to overall GVA. A further reallocation step then moderated these estimates by transferring CoE to other uIO64 groups within the wider ONS regional accounts group.

Step by step process

1. Each uIO64 sector was checked to see if the ratio of CoE/GVA was higher than the maximum reported for any UK sector at 64-sector level (86% in 2019).
2. Those sectors were further checked to see if the CoE/GVA ratio for that target sector was more than +/- 10% the UK ratio.
3. Where 1 and 2 was true, CoE was taken from the target sector such that the ratio of CoE to GVA became the average of (1) the initial Wales estimation and (2) the UK ratio of CoE/GVA for that sector.
4. Now ‘surplus’ Welsh CoE is reallocated to other uIO64 sectors in the same Regional Balanced GVA 31-sector (RBGVA31) group, weighted by on their existing estimate of CoE.
5. Where there is more than one ‘target’ sector within a single RBGVA31 group, the surplus is aggregated and divided amongst remaining non-target sectors as above.

The result is thus an estimate of CoE for each unknown IO64 sector that was consistent with the CoE estimate produced by ONS for the wider group, and which was considered reasonable in relation to overall GVA.

Taxes less subsidies for each of our 48 uIO64 was estimated with reference to the ratio of T-S to GVA as reported in the UK Supply and Use Tables for 2019 (ONS) and taken directly from ONS regional accounts for the remainder.

Finally for our initial estimation of GVA components, other value added (OVA) is estimated for each unknown IO64 group as the remainder of ONS-published figures after the subtraction of our estimated compensation of employees and taxes-less-subsidies.

Background

Estimating economic output for the Welsh economy is difficult, with no publication of System of National Accounts (SNA)-compliant aggregate measures below the level of the UK (where it is reported in UK Supply and Use and Input-Output, IO, Tables). The situation is significantly more challenging for some individual industries due to limitations with the availability and quality of the relevant data. Such estimates, however, form a critically important part of the SUT and Input-Output structure. The ratio between intermediate inputs and GVA (the latter published by ONS for 71 industries at ITL1), which together sum to output, has important economic implications.

Some additional data is available to help derive this estimate of output for, in the Wales’ case, 64 industries.

Additional data

• ONS publishes aggregates of Annual Business Survey (ABS) key variables at ITL1 scale, including turnover (analogous to output for most industries when relevant elements such as changes in inventory, and goods resold with no further processing are accounted for).
• Further aggregates obtained from ABS microdata, including approximate output.
• The annual accounts of Welsh-headquartered companies (taken from the Companies House Data) whose business is wholly or mostly in Wales, and where a small number of businesses comprise the bulk of Wales-territorial economic activity
• Bespoke, sector-specific regional or UK data wherein estimates of output are published or can be inferred.
• The approach to estimation varies dependent on the nature of the data available. This document summarises our approach to estimating output for 2019 across our 64 IO sectors.

Sectors excluded from the ABS

The ABS sampling universe excludes parts of the agricultural sector, finance, insurance and real estate activities, and public administration and defence, for which ONS uses alternative survey and admin data sources in the UK national accounts. These seven sectors (at our level of disaggregation) require a different approach to those which benefit from ABS coverage. The approaches are as follows.

Agriculture (A01)

Agriculture benefits from a longstanding policy focus and public support (at both Wales and UK level), and benefits from a number of bespoke data sources that help estimate economic scale and behaviour, including the annual Farm Business Survey and annual Agriculture Survey (which provide counts of livestock, land and other physical inputs and assets). Our key source for IO estimation is Wales Aggregate Agricultural Output and Income (Welsh Government). This provides details (across a number of years) on output and its components, and value added and its components, including for the latter, CoE, depreciation, subsidies (of various types) and residual income.

The results of the output and income publication were used to compile the 2019 IO vector, providing our estimates of output as well as value added and its components (where they supersede ONS regional balanced estimates of GVA and its components).

Financial services (K64)

We examined the ratio between GVA and output reported in the 2019 UK SUTs for the relevant industry. Adopting this ratio would allocate 1.8% of UK financial output to Wales, on an employee base of 2.4% (implying output per FTE in Wales is approx. 66% of the UK level). We then compared the resultant level of output to Scotland, which showed our estimate for Wales was broadly in line.

In the absence of better data, this estimate seemed reasonable and was adopted.

Insurance (K65)

Estimating insurance activity at sub-national scale is conceptually and technically complex. In summary we have based our estimate in large part on the aggregates for a leading insurance group of companies, which has the large majority of its workers in Wales, and which comprises the majority of all insurance and related employment in Wales. There are however a number of options. 

The ONS has a relatively straightforward approach for measurement of the finance and insurance industries in estimates of regional GVA (ONS) using apportionment: UK estimates are regionalised based largely on the level of employment. Undertaking this exercise for Wales provided an estimate of around £6bn (Option A).

Looking directly at data on FTE employment (BRES) & headquartered companies (Companies House) in Wales suggests this leading insurer employs some 75% of Wales' insurance workers, and accounts for a similar or higher proportion of industry operating revenue. This company however reports a net revenue (gross revenue less premiums passed on to other insurers) of £1,350bn and a gross revenue of £2,850m. This implies (assuming the company is representative in terms of revenue and/or revenue per employee) a Wales insurance gross turnover/revenue of under £4bn.

Using this figure for Output (whilst adhering to ONS balanced GVA), would however significantly reduce the ratio between ONS published GVA and our Output estimate when compared to IO Tables for Scotland and the UK (NI does not publish Insurance separately). Accepting this difference but maintaining congruence with ONS regional accounts is possible (Option B) but has significant implications for the robustness of the result – i.e. it would produce estimates that could be demonstrated as probably false. 

An alternative is to report GVA and its components based on the leading company’s aggregates for wages, profits and taxes (multiplied by the ratio of Wales to company employees) - and to treat regional output similarly (Option C). This would give 'weight' to regionally specific information and imply that production functions were different in Wales from the UK average (e.g. perhaps driven by wage & product differences, and with Wales having no large life insurers). Our estimate of Wales GVA would however become somewhat smaller than ONS published figures for 2019. 

We have here adopted Option C. – but note that due to concerns over data quality, sector K65 is not published separately in our final Tables.

Activities auxiliary to finance & insurance (K66)

ONS balanced regional GVA estimates the Wales share of this sector at £493m (1.0%). With Wales hosting 2.1% of FTEs this implies a considerably lower level of output (and value added) per FTE –under 50% of the UK. Interestingly, a lower level of relative GVA is not reported in the Scotland 2019 IO Tables: 7.2% of the UK, on only 6.8% of FTE employment. 

ONS does not publish components of GVA for K66 at regional scale. However, multiplying ASHE employee gross annual income for the relevant industry ^{[footnote 6]} by Wales FTEs results in an approximate estimate of £230m for CoE. This level of CoE would imply that (unless other value added in Wales is far greater) GVA in the sector is proportionately far lower than for the UK and Scotland. 

We thus adopted the ratio between GVA and Output from the UK SUTs to initially estimate output in Wales at around £900m in 2019.

Real estate activities excluding imputed rents (L68.3 + L68BXL68.3)

In IO64 this sector is the aggregation of real estate on own account, and that undertaken on a fee or contract basis. ONS published GVA was 2.2% of the UK total for 2019. This is significantly lower than Wales’ proportion of UK FTEs (3.1%). Meanwhile, the volume of Wales’ residential sales in 2019 comprised 4.8% of the UK ^{[footnote 7]} – albeit at an average price of £164,000, some 71% of the UK average. This implies the total volume of sales (turnover) of residential properties at 3.4%.

Adopting this ratio for Output would see our estimate significantly higher than simply using the ratio of GVA to Output as found in UK SUTs. Suitable data are not however available to estimate the value of non-domestic/commercial property sales in this sector – let alone by region. These sales are likely to reduce relative sector output in Wales, being a priori more closely linked to GVA or Gross Fixed Capital Formation (GFCF) – both low in Wales (and GFCF especially so).

We here then take the ratio of GVA to Output to estimate Wales;’ output for this sector, but accept the possibility of uncertainty in these estimates. As with all sectors, we must consider issues of data quality at the balancing stage.

Imputed rents of owner-occupied dwellings (L68 (other))

This imputed sector has no employees, wages or actual transactions. As such we simply apply the GVA-Output ratio from UK SUTs to achieve a Welsh Output estimate (3.3% of UK). Note, dividing the number of Wales households in the 2021 Census by the British total, then multiplying by the Wales-UK residential average price difference results in a ratio of 3.45%, so the estimate appears reasonable.

Public administration and defence (and other public sector) (O84 to Q88)

The ABS does not capture output for Public Administration and Defence (PAD); UK national accounts data is instead based on administrative government accounts data. For the purposes of the Wales estimate, the ratio of GVA to Output was extracted from the UK Supply & Use Tables 2019 and applied to Wales’ GVA to report an estimate for output.

 Further examination of ABS estimates for public sector output suggested some potential issues, with Education estimated as having a significantly larger output in Wales than for Health (despite the latter being the larger sector in UK SUTs). Hence, output estimates for Education, Human Health and Social Work activities were derived from UK SUT Output-GVA ratios (as for PAD). 

Other bespoke estimation & triangulation 

Unfortunately, the ABS sample is not stratified at ITL1 spatial scale, and only a limited suite of variables is published at this geography. This means that even where a sector is covered by the ABS, non-return bias can significantly impact the estimation needed for SUT compilation. A small number of sectors can be triangulated against other sources.

Basic metals (SIC 24)

In 2019, Wales had three significant companies in this sector. Basic steelmaking was confined to two sites with a range of rolling, coating and other speciality steel working at other sites.

Wales accounted for 100% of one company’s operations, and another company in Wales for about 85% of UK employment. Another company meanwhile, appeared to employ around 5,000 in the UK, of which c. 200 were in Wales. Applying these proportions would indicate that in aggregate, 79% of these three organisations’ UK operations were located in Wales.

Assuming the ‘per-worker’ equality of all financial metrics across UK operations allows us to apportion output and key value-added metrics to Wales. This results in an estimate of, for example, around £2.4bn of Welsh output/turnover, £330m in gross wages and losses of almost £200m.

The sum of our estimated approximate GVA from company accounts (wages + profits + taxes on production) is, at £310m, under half of the ONS estimate for SIC24 (£628m). We adopt our bottom-up estimate of output/turnover (and GVA) due to:

• the dominance of these three organisations in the relevant SIC for Wales
• the dominance of these companies’ Welsh operations in their overall UK operations
• our ability to present, bottom-up, policy-important components of GVA such as wages, losses and subsidies (ONS does not publish regional components of GVA separately for SIC24)

Using this bottom-up approach will also potentially allow a high-quality estimate of steel exports from Wales at a later stage, based directly on company declarations.

Water supply & sewerage (SIC 36-37)

Water supply and sewerage in Wales is provided almost wholly by two companies. The aggregate of these two companies’ financial metrics for 2019 will comprise a reasonably accurate picture of Wales’ SIC36-37 ^{[footnote 8]}.

This approach estimates regional output/turnover at £810m. Our estimate of approximate GVA is £500m, significantly less than the balanced ONS estimate of £735m.

We note here the dominant role of depreciation in sector other value added and GVA. We estimate this at £300m for Wales in 2019, with wages at £170m and profits £30m. Depreciation will likely vary significantly across UK water companies, in ways unrelated to current employment, and we thus consider our bottom-up, company accounts-based estimate to be likely superior to apportionment by ONS ^{[footnote 9]}, adopting this approach for output and GVA and its components.

Final comment

The adjustments described in this chapter mean that our estimates of sector GVA and its components do not in all cases match those published by the ONS for 2019, instead being chosen to most closely represent territorial economic activity in Wales as best understood for the purposes of these tables. 

Background

Whilst there are a number of data sources available at regional level, or areas where Wales-level estimates for 64-sector commodity demand can be estimated, we did not fully replicate the UK SUT Final Demand structure. There are several reasons for this:

• At the time of estimation, data on international exports of goods from Wales e.g. from the Regional Trade Statistics (RTS, HMRC) are limited, detailing only major commodities and countries and only in 2-digit Standard Industrial Trade Classification (SITC) form, with then significant difficulty in reallocating to 64 sectors.
• Data available on regional international trade in services from HMRC was insufficiently detailed.
• Although Trade Survey Wales provides estimates of Wales to UK and international trade in goods and services, it has limitations, for example being voluntary with a low response rate.
• Exports from Wales to the rest of the UK (i.e. outside our regional reference economy) must be detailed in a way not necessary for the UK Tables.
• The UK division of Government demand by Central and Local government is less useful in our regional context where local, Welsh and UK governments demand products and services, but with limited information on how these are split across commodities.

Elements of final demand in the UK SUTs

Final demand for products in UK supply and use tables comprise the elements detailed below. 

UK SUT final demand

• Households
• Non-profit institutions serving households
• Central government
• Local government
• Gross fixed capital formation
• Valuables
• Changes in inventories
• Exports of goods to EU
• Exports of goods to ROW
• Exports of services

With data and geographical differences in mind, the list below details our subdivision of final demand in Wales, essentially aggregating UK export columns into one international export estimate for Wales, adding a rest-of-UK export element, and aggregating government demands element into one. 

Wales SUT/ IOT final demand

• Households
• Non-profit institutions serving households
• Government (both central and local)
• Gross fixed capital formation
• Valuables
• Changes in inventories
• Exports to the RUK
• Exports to the ROW (that includes the EU)

Despite these aggregations and changes, significant estimation issues remain in achieving an estimate of Wales’ final demand. These are covered in the following sections.

Household Final Consumption Expenditure (HHFCE)

The sole data source used for our estimation of household expenditure is the Living Costs and Food Survey (LCF, ONS) (this source is complemented at UK level by other sources that do not report at ITL1 level). 

Data is collected on the pounds per week spent on products and broken down by ITL1 region. It is produced according to COICOP (classification of individual consumption by purpose) which can be mapped to CPA. 

The sample size for Wales is low (around 220 households per year), so estimates are based on 3-year combinations of data. We have used data from the financial years ending 2017, 2018 and 2019.

Non-Profit Institutions Serving Households (NPISH)

Information on NPISH is limited for Wales (as for other devolved regions). We are however able to directly estimate the largest single element, demand for education services. These are estimated with reference, firstly, to the expenditure of the higher education establishments in Wales as reported by the Higher Education Statistics. The second element relates to the provision of education by the independent school sector in Wales (typically religious or charitable bodies) with this informed by Welsh Government estimates of pupil numbers in the schools' census results (Welsh Government), and fee information from the Independent Schools Council Annual Census.

In the absence of data on the commodity purchase of other NPISH elements in Wales (e.g. charities), cell estimates depended on a straightforward apportionment of UK SUT totals based on the number of Welsh households as a percentage of the UK (drawn from the varied UK Censuses of 2021/2 rather than LFS-based estimates for 2019). 

Government

Welsh Government Budget outturn for 2019/20 was examined and allocated to relevant commodities as appropriate based on major expenditure groups, budget expenditure lines and other relevant information, generally allocating expenditure to commodity groups with non-zero estimate in UK SUTs. Amendments were then made to allow for regional specificity (e.g. an element of government demand in ‘Land Transport’ to account for Transport for Wales) – although this process requires further checking for SNA-compliance. 

ONS estimates of regional managed government expenditure (for 2019, current spending) were used to add product demand elements in reserved areas such as defence, policing/justice and international, resulting in a final estimate of around £24.4bn in regional government product demand. This is about 5.7% of relevant UK Government demand, although it should be remembered that this is government demand we notionally allocate to the region (including for example money spent internationally and that related to administrative activities undertaken elsewhere in the UK) rather than the demand for products supplied from Wales.

Gross Fixed Capital formation

The UK SUTs detail Gross Fixed Capital formation for 26 commodities. These are aggregated to 21 IO64 groups, and then divided by the ratio of Wales-UK Output to obtain an initial Wales estimate of GFCF by commodity. 

ONS report experimental estimates of GFCF for regions for five asset classes: ‘Buildings and structures’; ‘ICT equipment’; ‘Transport equipment’; ^‘Other tangible assets’; and ‘Intangible assets’. Each of our 21 non-zero cells were allocated, best-fit to one of these classes and their value multiplied by the regional intensity of GFCF for that class (essentially a location quotient) in comparison for the UK. 

The final estimate totalled £11.25bn. Each non-zero cell was then inflated by the difference between this total and the total for regional GFCF reported in the ad hoc data output Regional gross fixed capital formation, ITL1 and ITL2, 2000 to 2020 (ONS) (£13.15bn). Note this data source also contains sectoral estimates, but these are GFCF estimates by industries not of the demand for products which we require. 

Valuables and changes in inventories

Whilst no information is published on these items, the ABS includes questions on stocks at the beginning and end of year, and the resultant change (including works in progress). The ABS microdata for Wales in 2019 ^{[footnote 10]} was used to estimate changes in stock/WIP by 64 groups. Given that changes in stock is a volatile and derived (net) variable, the ABS sample will not be representative of missing firms, so we did not gross up to estimated aggregate product supply.

Exports to the ROW

Estimation of exports to the ROW is a complex undertaking. For example, HMRC data is limited in several respects.

Limitations of HMRC data

• Publication was partial, covering only the top products, and with disclosure issues affecting full access, though future estimations will not suffer this restriction with the introduction of the UK trade data report tool (HMRC).
• HMRC is tasked with accounting for physical goods only, rather than goods and services.
• Exports are apportioned on a regional employment basis at firm level, thus leading to potential misallocation – for example where a firm might have a large non-Welsh HQ, but with a Welsh plant perhaps exporting proportionately more of the firm’s products.
• The data are available regionally by 2-digit Standard Industrial Trade Classification (SITC) which is very difficult to reallocation to the product categories used in our IOTs.

The Welsh Government’s Trade Survey for Wales (TSW) fills a number of these gaps, gathering information on the international (and inter-regional) exports and imports of (in 2019) well over 1,000 respondents. Albeit this data has caveats, as detailed in the quality report published in 2021 (Welsh Government). Meanwhile, for some sectors, sector-specific information provides detail not available more generally, or high industry concentration means regional exports can be inferred from company accounts. The headings below provide detail (covering also rest of UK exports where appropriate).

Summary of export estimation

Agriculture, forestry, fishing (RUK, ROW)

Estimated from sector-bespoke sources at Wales and UK level, including those formerly required for EU fund management.

Basic metals (ROW)

Export totals reported in company accounts of Wales’ three steel makers (employment allocation), as well as HMRC data.

Electricity, gas, steam and air conditioning supply (RUK)

Digest of UK Energy Statistics estimates of regional electricity/energy supply & use; OFGEM price estimates and breakdowns. 

Water supply & sewerage (RUK)

Water company accounts; water abstraction license costs. 

Insurance (RUK; ROW)

Company accounts (applied to Wales’ UK population share). 

Other sectors (RUK; ROW)

TSW and HMRC data where available.

For ‘non-bespoke’ sectors we initially estimated using data from TSW, aggregating data on exports of goods and of services, to the UK, EU and ROW, into total exports to the RUK, and to the ROW. TSW sample microdata were analysed to estimate the proportions of export sales in total sales for our 64 IO groups. This proportion was then applied to total estimated ABS turnover ^{[footnote 11]} for that IO group in Wales. 

This methodology ignores the impact of, firstly, non-response in the TSW sampling frame and secondly, the targeting of trade-oriented companies in the original sample (not then representative of the overall Wales firm cohort). We therefore undertook an alternative process whereby we estimated UK export propensities for our 64 commodities based on UK SUTs.

Where TSW coverage is less than 10% of ABS output and where the difference in TSW-ABS export propensity is more than ten percentage points, we default to the SUT proportion, applied to Wales’ industry output. The same steps were followed if the TSW reports no ROW exports by that industry. In all other cases we default to the TSW estimate. A final step constrains total ROW exports to that reported by TSW for 2019 in a later revision, some £24bn.

Exports to the RUK

Trade Survey Wales (TSW) estimates trade from Wales to the UK though other data sources were used where specific sectors benefit from more reliable estimates (e.g. from company accounts or bespoke sector surveys). 

Conclusion and summary

Following the above estimation, we achieved a full accounting of the final demand in 2019 for 64 products, on a ‘combined use’ basis – i.e. including imports, and in purchasers’ prices (but excluding taxes on products).

Background

Very limited data are published on the purchases of intermediate (non-value added) purchases at a sub-UK level. Two ONS surveys however require businesses to provide this information; the ABS and Annual Purchases Survey (APS). We chose APS as our source; it is specifically designed to capture detailed information on company intermediate consumption and is used for that purpose in the UK National Accounts. 

Both surveys suffered pandemic-related challenges with 2019 data (that was collected in 2020) and cannot be aggregated across years (due to weighting and survey response issues). While ONS were able to make effective use of the 2019 data at UK level, due to the lower level of response for Wales, we have used 2018 patterns instead.

In other countries (Scotland, NI) we understand that both surveys are combined, with more detailed APS commodity purchase estimates constrained to wider ABS categories. Other devolved nations however benefit from a regionally stratified (Scotland) or fully bespoke (NI) version of the ABS. This is not currently the case for Wales. 

After careful consideration, the research team opted to base the initial estimation of Wales’ SUT intermediate product purchase proportions on the APS microdata, albeit of course summing to intermediate aggregate totals that are for most sectors estimated via the residual – i.e. the difference between ABS output and a multiply-sourced estimate of regional GVA. 

It should be noted that a hybrid approach might have resulted in an improved estimation, but this would have been at the cost of significant investigative and compilation time, and complex narrative explanation.

Drawing on the APS

Following the above decision, the estimation of commodity purchases for our 64 industries is relatively straightforward.

ONS provided estimates of weighted IO64 sector intermediate product purchases for Wales in 2018 by relevant IO64 groups (but not the excluded Public Administration as an industry), in £Thousands. These were converted into propensities (hence total intermediate purchases = 1). These propensities were applied to our estimate of total intermediate purchases for each IO64 industry in 2019 to achieve an estimate of commodity purchases for each IO64 group in £m (2019).

Further analysis & amendment

Following this arithmetic approach, some further modification was required to two sectors.

Manufacture of coke and refined petroleum products (SIC19)

Firstly, it was noted that the ‘Manufacture of coke and refined petroleum products’ sector purchased the bulk of its intermediate inputs from ‘Mining and Quarrying’. This no doubt relates to the purchase of crude oil and related products. However, the large scale of this sector in Wales and dominance of intermediate purchases in that output has a very discernible impact on the importance of mining and quarrying in regional intermediate purchases overall ^{[footnote 12]}.

Further investigation showed that a single plant survey return accounted for this single reported transaction. Additionally, APS under-reports total regional intermediate purchases (summing to some £37bn for 2018 compared to our initial IO estimate of £69bn in 2019). This undercount then significantly over-weights this single plant’s purchase. In order to soften this impact, we depreciated total sector purchases by the approximate under-coverage of the APS (versus IO) to reflect the over-weighting of that sector and plant.

Repair and installation (SIC33 OTHER)

A similar issue arises with the purchases of ‘other transport equipment products’ by the ‘repair and installation industry’. Again, a single repair plant dominates in Wales, making purchases that are likely to be different from smaller, under-sampled repair businesses, and thus with the associated challenges with the commodity purchase. Here we depreciate the ‘individual product purchase’ by the APS whole-economy undercount. This is because, unlike refining of petroleum, the sector does not comprise a single plant (Companies House reports over 1,700 firms headquartered in Wales) and we thus considered there was value in retaining the original weighting for other purchases (although we accept this is slightly arbitrary). 

Alternative approaches – for example simply using UK SUT-reported proportions of product purchase – were considered, but this then would completely ignore the product purchases of an extremely important element of the regional sector.

Public Administration (O84)

The purchases of Public Administration (O84) are not detailed in APS (or the ABS). We therefore defaulted to the estimates produced in the UK SUTs for 2019 for product purchases (at IO64 level) although accept there will be some differences due to the different activities and competencies of governments at UK and Wales scale.

Introduction

The steps described so far provided an estimate of the use of 64 products in Wales by 64 regional industries, for varied elements of final economic demand. These estimates were produced in purchaser prices: that is with taxes on products subsumed (but invisible) within each estimate ^{[footnote 13]}, and with distribution and trade margins (DTM) also included (but invisible) within the price paid for each product, rather than separately identified and re-allocated to the appropriate distribution activity (motor sales, wholesale, and retail), which are thus reported net of margins.

Additionally, the (product) use table details the supply and use of all products used in Wales (or exported from Wales) with no further detail on whether these products are sourced from within Wales, or imported (from the RUK or the ROW).

Whilst there is a wealth of interesting information in this table, an SUT requires an appreciation of the supply of products from within Wales, differentiated from intermediate or final imports. This supply estimate must also at some point be transformed into basic prices – with taxes on products identified and removed, and DTMs re-allocated to relevant distribution activities Tables – i.e. present a full balance of regional supply and demand of products/industry supply to SNA standards, in common prices (and to enable the estimation and publication of IOTs).

Steps taken to estimating in basic prices: 

1. Estimate the level of remaining taxes on products and deduct this from all product rows and report separately.
2. Estimate the distribution and trading margins for all physical products (plus publishing & broadcasting activities) and reallocate to relevant distribution sectors.
3. Estimate, for each product and industry, the level of non-regional supply and remove from our supply estimate, to be separately reported.

Following the above steps, we achieve an estimate of regional supply in basic prices that can be reconciled with regional demand (albeit through a further balancing process).

Estimating taxes on products

Given the limited variation between Wales and the UK in tax policy and regulation in 2019, tax rates for products in Wales should a priori be expected to be similar to the UK ^{[footnote 14]}. We thus estimated product taxes for IO64 groups from the output-weighted average of 105 rates reported by ONS for 2019 in the UK SUTs. These rates are then applied to the supply of each product in Wales in a manner analogous to that in the UK SUTs to provide an initial estimate of product taxes in Wales.

ONS produces a regional allocation of taxes and subsidies for the UK that was supplied to the research team. A subsequent process then apportioned, where possible, relevant product taxes (and subsidies) to our IO64 groups. The resultant estimate was then used to check the veracity of our proportional allocation, resulting in the supplanting of one estimate with the ONS-supplied direct regional estimate ^{[footnote 15]}. 

It is worth noting that our approach is relatively crude. For example, errors will occur if product taxes are different for different users (e.g. final demand versus intermediate industries) and product use varies between Wales and the UK in respect of the proportions of product use. Significantly more information on product tax rates in the UK is available from ONS, but these data proved too complex to integrate into our current estimation process given resources and timeframe.

Estimating margins

The process of estimating margins was somewhat more complex than that for taxes, not least because margins must be allocated to one of three DTM (distributor trading margins) groups (motor sales, wholesale, retail). 

ONS National Accounts team again provided a wealth of relevant information, notably the level of margins estimated for detailed product classes ^{[footnote 16]}, and with these differing by detailed use – whole economy, households, and for export for example. Additionally, the team provided estimates of the ‘channels’ via which products progressed to market – related to the three DTM groups, with different DTM rates.

Thus, for each IO64 group, UK margin rates were weighted by relevant channel to achieve an estimate of weighted average total margin for that product, for intermediate sales, for exports, and for households ^{[footnote 17]}. Two separate estimates of DTM rates were then applied to relevant sales across intermediate industries, and domestic final demand (the latter using the household rate) ^{[footnote 18]}.

There is however a limitation here, in that the UK National Account does not (and cannot) distinguish between margins accruing to Wales-based distributors, as opposed to those accruing to distribution across the rest of the UK. We dealt with this by turning to the ABS for 2019, and specifically the returns of Wales-based businesses.

Here we then estimated the total DTM accruing in Wales by subtraction of purchases of goods with no further processing from turnover (apart from additional adjustments for example changes in stocks), aggregating the firm returns by three SIC classifications that best represent our three required DTM activity (product) classifications. In aggregate across all relevant industries this then led to an approximation of total margins accruing in Wales. 

Our ABS-based estimate of total margins accruing in Wales is some £10bn. This is allocated to specific cells in the supply table based on our earlier estimate of total DTM for that industry (i.e. the DTM rate for that industry derived from the weighted UK DTM rate applied differentially to each product it supplies). For each industry (and indeed for final demands), the resultant margin is added to the existing output ^{[footnote 19]} of the three DTM activities based on the DTM split reported for all industries in the 2019 ABS. The uncertainties involved in this process led to the aggregation of wholesale, retail and motor repair prior to publication. Nonetheless, following this step we achieved an estimate of regional supply in basic prices.

Estimating regional & non-regional supply

The final step is to estimate the proportion of each product that is sourced regionally, from the rest of the UK, or from overseas. These data are not collected in the ABS or the APS and are not published by HMRC. Unlike for exports, it is usually impossible to infer such proportions from company press releases or annual accounts.

A wide variety of algorithmic methods exist for estimating regional supply for IOTs where no direct data exist. These are all however, relatively crude – typically based on, for example, concentration of employment or output (‘location quotients’), economy size, or latterly aggregate movements of freight vehicles. None of these methods reflect the reality of global supply chains, where movements of good and purchases of services are based on the international structure of multinational corporations, relative prices and/or quality, or long-established commercial linkages, rather than crude levels of activity or production. For example, most of these approaches would have Welsh lamb consumption based squarely on our significant levels of production but this is not the case in reality.

Until relatively recently, this was a very significant – arguably insurmountable – estimation gap, but the Trade Survey for Wales fills a proportion of that gap – albeit with caveats. The TSW includes estimates of the location of purchases (Wales, other UK countries and overseas) for over 1,400 businesses in 2019, and covers both goods and services. Using the TSW, it was possible to estimate an aggregate regional use and import propensity, across all products/services for each (4-Digit SIC) industry in Wales. 

This is, however, not quite what is required. An ideal survey would detail individual product purchases by origin for each industry, and these estimates would (after re-pricing) slot directly into our emerging supply table. However, whilst TSW does ask which ‘top 5’ products are purchased, it does not record their geographic origin.

We are therefore restricted to knowing only the geographic pattern of purchases by industries, not of products. We examined two options.

Option A

To translate industry purchases to the product information, to undertake the following steps.

1. Examine the product purchases of the relevant (reclassified to IO64) industry in the UK IOTs.
2. For each industry, we split total purchases by Wales, RUK and ROW based on the results of the TSW for that industry.
3. Each of the three geographic purchase elements is then classified to products according to UK IO ratios, resulting in an (indirect) estimate of product purchases by each origin.
4. Summing across all industries results in a single estimate for each product of total purchases by each origin.
5. These totals are summed and ratioed to provide a regional use, and RUK and ROW import propensity for each product.
6. This ratio is applied to basic price (post tax and margins-reallocated)  purchases of that product for each industry to achieve an estimate of regional supply in basic prices for 64 industries.

It should be noted that this approach makes two significant assumptions. Firstly, that the production function of each industry in Wales matches it’s SIC-match for the UK in terms of product purchases. Secondly, that our estimate of the aggregate ratio of regional- to import-supply for all industries and final demand, based on the TSW all industry average is appropriate for each industry: i.e. we assume that each industry in Wales uses local or imported products in the same ratio for each product, but with each product having a different regional/import supply propensity.

Option B

We straightforwardly assume that industries’ imports comprise only their own principal (output) product and hence apply industry proportions of regional and non-regional input purchases to that principal product row in the SUT.

These are significant (and incorrect) assumptions in both options. Option A is more ‘technically correct’ (as it allows for UK-level industry input mix) but results in a common regional supply proportion for each industry across all products. Option 1 would, for example uniformly allocate 90% of all water industry intermediate purchases to Wales, but less than 5% of mining and quarrying industry purchases. This would result in significant variation in any industry multipliers derived from later-estimated IOTs, with these differences often based on unrepresentative TSW samples.

Option B conversely varies the estimated regional purchase proportion of each product, but with each industry (and final demand vector) assumed to purchase each product in the same regional proportion. This approach, by varying regional intermediate purchase propensities within each industry will soften the impact of our missing information in its impact on any estimated Leontief multipliers; i.e. they will have a far less extreme distribution (but will rely in part on our industry-product purchase identity assumption).

Option B was considered the most reasonable given that in current data landscape we have no information on the geographic origin of products used by Wales’ industries – and due to the likely future use of multipliers from the (albeit experimental) IO Tables. It should however be noted that making small changes to future Trade Survey Wales coverage – i.e. product supply by origin – would result in significant improvements to the regional supply estimation.

Summary

The above steps took us from a regional use table to a fully featured regional supply table for Wales for 64 industries and products in 2019. With the table available in basic prices, supply can be reconciled with use to assess the level of regional supply and demand for each product. 

Given the nature of the estimation process – using a wide variety of demand- and supply side surveys of varying quality, regional coverage and statistical reliability, together with company accounts and imputed data, it is no surprise that demand does not equal supply for any product. The next stage of the process was to re-examine our estimates and achieve this balance.

The problem

Supply and use (and symmetrical Input-Output) tables bring together a wide range of economic data, drawn from both administrative sources (such as aggregated tax and customs returns) and surveys – including of individuals, households, businesses and other organisations. These data are collected for different reasons, using different concepts and scopes (sometimes for the same notional measure). They may measure quarterly or annually and can represent calendar or financial years.

It is no surprise then, that when these data are compiled and brought into a common SUT framework they do not match, and do not describe industry production functions, the territorial spending of government and households, or the prices in which things are bought and sold, in one coherent price framework.

Sections 1 to 7 of this report describe how many of these issues were addressed, and the amendments made to make data more consistent. However, a final stage is required when, in the SUT, the estimated territorial supply of each identified product is compared with estimated demand and found – very typically across SNA processes – not to match. A final balance is then required to equate the supply and demand of each product for the reference economy.

When the reference economy is sub-national, the above problems are increased significantly, both in the UK and many other countries for several reasons.

Reasons for the problem

• Core economic surveys (ABS; Living Costs & Food) have limited sample sizes at devolved scale, and the structure of these surveys can make this difficult or expensive to address.
• Some other surveys are not designed to report at devolved/regional scale, for example lacking regional stratification or any formal mechanism whereby sample returns can be scaled (at firm or variable level) to properly represent the region.
• Important administrative data – for example on product taxes and trade in goods (both inter-region and international) – are unavailable/irrelevant sub-nationally.
• Some issues – for example the treatment of cross-regional distribution margins, or the status of inter-regional goods or financial movements within a firm have been less discussed and estimations are thus open to different interpretations.

The above are particular challenges in Wales, which for historical reasons has an economic-statistical landscape far more integrated with England than Scotland or Northern Ireland are with the UK, and with commensurately fewer bespoke data and estimates available (or currently even possible). The final balancing of the SUT is therefore more difficult.

The approach

Our balancing of supply and demand was undertaken manually, commodity by commodity. It is important to note that a full supply (or ‘make’) matrix, industry by product, for Wales does not exist and thus balancing must be undertaken by equating the (estimated) territorial supply by industries in Wales (in columns) with total demand (in each relevant row). This required an estimate of the product mix supplied by each of our 64 Wales industries, and for this we relied upon the UK Supply matrix provided by ONS – accepting however it will misrepresent what happens in the region.

Beginning with ‘A01 Crop and animal production, hunting and related service activities’, estimated product supply was equated – approximately – with estimated product demand. Every element of estimation was assessed for likely reliability and accuracy, including product intermediate and final demand elements, and our estimate of industry output (which directly feeds our estimate of regional supply). This process also reflects other contextual data, such as regional employment location quotients, UK ratios of product exports, and the approximate estimate of gross wages that arose from our estimation decisions ^{[footnote 20]}.

We then amended the elements of our estimation considered weakest or giving rise to unreasonable outcomes to balance product supply and demand. Whilst it is desirable to maintain links to ONS estimates of regional GVA, in some cases our estimate of output/regional supply was changed (in the face of better product demand information, or where implied wages appear implausible) thus giving rise to SUT estimates of GVA that differ from regional accounts (balanced GVA). In some cases, output is amended by changing single elements within value-adding elements, or in intermediate purchases, in others, output is scaled linearly. Other table elements were amended as required, including the propensity to source inputs locally (an estimate wholly dependent on Trade Survey Wales and UK ratios), and distribution margins (these are wholly UK estimates at product level). 

As we progressed through the table, we undertook amendments as far back along the ‘supply chain’ of spreadsheets as possible, rather than manually amending values in our final SUT, thus seeking to retain the interlinked nature of our set of workbooks. All amendments were explicitly noted and agreed within the research team. 

The ‘live’ nature of this approach means that linking is protected, and as we progress down our product rows, we can see how changes to industry scale or local sourcing work back through supply chains to ‘unbalance’ other product rows, including those already balanced. The process must be repeated several times therefore before a full, approximate balance is obtained for all products.

Our outputs

SUTs are intended to accurately describe the creation and use of products and services within an economy, identifying which industries produce these commodities. In many cases a single product can be created as the output of industries other than that primarily associated with it. For example, whilst most visitor accommodation is provided within the primary hospitality sector (by hotels, B&Bs and self-catering for example), rooms are also provided by other industries – for example farmers partially diversifying into tourism. Conversely, industry output is not homogenous. Industries can produce a range of products, and for some they will not be the primary producer. For some products, ‘industry spread’ is significant (think of the research and development (R&D) activities spread across many different types of firms), and conversely some industries produce many products and services.

This means that in any SUT, the supply and use of every product is balanced across a single matrix row, but this total supply/use does not then equal the output of its ‘home’ or primary industry (because that product may be additionally produced elsewhere, or the industry may create multiple products. Thus, whilst a SUT is conceptually robust it cannot, due to the demands of matrix algebra inform impact and policy modelling that rely on IOTs (including, for example Computable General Equilibrium, or carbon foot-printing). This is because economic and other multipliers can only be derived via a Leontief ‘square’ matrix where the total reading across the matrix row (until now recording the supply of each product) equals the related column total (the output of the primary industry). 

Most statistical agencies therefore publish both SUT and ‘symmetrical’ IOTs, albeit the latter not always in as regular or as timely a fashion. Symmetrical IOTs can either detail economic transactions between industries (hence IxI) or between products (PxP). The process for PxP effectively moves products into their primary industry, adopting the industry production function, then “transferring inputs associated with secondary outputs from the industry in which that secondary output has been produced to the industry to which they mainly belong” as outlined in Eurostat’s supply, use and input-output tables methodology. Alternatively, for IxI tables we can assume the use (or sales) structure of a product is independent of the industry in which it is produced. 

Both of these transformations bring assumptions that are unlikely to hold in practice (for example, there is no fertiliser involved in the provision of on-farm visitor accommodation, but an IxI table might suggest so), but are made by statistical offices around the world and produce matrices that are extremely useful. For a number of reasons relating to data availability, statistical anomalies and policy relevance, IxI matrices are more commonly derived and used. We thus follow Scottish Government in identifying IxI as the primary (and in our case only) matrix of interest that we publish.

The difficulties at Wales scale

Moving products between industries or adopting different production/sales assumptions requires an understanding of the product-industry supply mix (a fully supply or ‘make’ matrix). Such a supply matrix, if consistent with the full SUT, will mean that some straightforward matrix algebra can produce a symmetrical IxI matrix. This is the approach taken by UK and Scottish Governments and NISRA, albeit the NISRA’s IOTs are designated as official statistics in development. 

Wales does not benefit from a supply/make matrix, which details the product supply of each industry (and vice versa). We cannot therefore undertake the matrix algebra to obtain an IxI which emerges already balanced. One option might be to artificially manage the SUT compilation process such that product mix matches that of the UK (or Scotland/NI) and multiplying use by the relevant make matrix ‘just works’.

This however is problematic: whilst we do not know the product mix that applies in Wales, we know it is not like the UK. To take two earlier examples, the Farm Business Survey (Aberystwyth University) tells us farms here are less diversified; and our industrial structure means far fewer industries will undertake R&D, as demonstrated in relevant publications such as the research and development gross expenditure (Welsh Government). Seeking to fully replicate UK (or other UK nations’) product-use relationships is probably not fitting.

In the absence of bespoke data for Wales, the following approach was used.

Approach

• Our SUT was derived, in PxI (product by industry) form using the best possible information but in the absence of a detailed product-industry supply mix, and then transformed to basic prices using the approaches detailed in Sections 1 - 8 above.
• Three distribution & trade margin sectors were aggregated into one ‘wholesale, retail & motor repair’ sector following deep data concerns about how regional margins could be individually allocated – thus creating a 62 x 62 product-by-industry SUT.
• The UK ‘make’ matrix for 2019 was aggregated to 62 sectors, transformed to PxI and multiplied by our SUT, moving the matrix some way toward symmetry – but notably keeping industry output (and value-added elements) consistent with our original SUT.
• Further balancing was undertaken by adjusting the Inventories/works in progress/valuables column ^{[footnote 21]} where a reduction in estimated inventories, improved the row-column balance.
• A final step calculates the remaining difference across each of our 62 industries then inflates/deflates final demand elements (excluding inventories)  in their existing proportions to eliminate this difference and match industry row and column totals for each ^{[footnote 22]}. 

Two final steps remain. We first ensure that no responses from individual firms to any ONS or Government survey will be discernible in any numbers published in the final publication of IOTs and SUTs. Additionally, a judgement is made on whether estimation issues (data or conceptual) are so severe that their identification in the final publication is not appropriate. Following these checks, aggregation is undertaken to address identified issues, and we publish SUTs and IOTs for 55 products and industries in Wales in 2019 (Annex 2).

Annex 1: the IO64 sectors and corresponding CPA

A01

Crop and animal production, hunting and related service activities

A02

Forestry and logging

A03

Fishing and aquaculture

B

Mining and quarrying

C10T12

Manufacture of food products, beverages and tobacco products

C13T15

Manufacture of textiles, wearing apparel and leather products

C16

Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials

C17

Manufacture of paper and paper products

C18

Printing and reproduction of recorded media

C19_C20

Manufacture of chemical, coke and refined petroleum products

C21

Manufacture of basic pharmaceutical products and pharmaceutical preparations

C22

Manufacture of rubber and plastic products

C23

Manufacture of other non-metallic mineral products

C24

Manufacture of basic metals

C25

Manufacture of fabricated metal products, except machinery and equipment

C26

Manufacture of computer, electronic and optical products

C27

Manufacture of electrical equipment

C28

Manufacture of machinery and equipment n.e.c.

C29_C30

Manufacture of motor vehicles and other transport equipment

C31T33

Manufacture of furniture and other manufacturing; repair and installation of machinery and equipment

D35

Electricity, gas, steam and air conditioning supply

E36

Water collection, treatment and supply, Sewerage

E37T39

Waste collection, treatment and disposal activities; materials recovery; remediation activities and other waste management services 

F

Construction

G

Wholesale and retail trade; repair of motor vehicles and motorcycles

H49

Land transport and transport via pipelines

H50_H52

Other transport; warehouse and supporting activities

H53

Postal and courier activities

I

Accommodation and food service activities

J58

Publishing activities

J59_60

Motion picture, video and television programme production, sound recording and music publishing activities; programming and broadcasting activities

J61

Telecommunications

J62_63

Computer programming, consultancy and related activities; information service activities

K

Financial & insurance activities

L68B

Real estate activities excluding imputed rents

L68A

Imputed rents of owner-occupied dwellings

M69_70

Legal and accounting activities; activities of head offices; management consultancy activities

M71

Architectural and engineering activities; technical testing and analysis

M72

Scientific research and development

M73

Advertising and market research

M74_75

Other professional, scientific and technical activities; veterinary activities

N77

Rental and leasing activities

N78

Employment activities

N79

Travel agency, tour operator reservation service and related activities

N80T82

Security and investigation activities; services to buildings and landscape activities; office administrative, office support and other business support activities

O84

Public administration and defence; compulsory social security

P85

Education

Q86

Human health activities

Q87_88

Social work activities

R90T92

Creative, arts and entertainment activities; libraries, archives, museums and other cultural activities; gambling and betting activities

R93

Sports activities and amusement and recreation activities

S94

Activities of membership organisations

S95

Repair of computers and personal and household goods

S96

Other personal service activities

T

Activities of households as employers; undifferentiated goods- and services-producing activities of households for own use.

Annex 2: the published Wales IO55 sectors and corresponding CPA

A01

Crop and animal production, hunting and related service activities

A02

Forestry and logging

A03

Fishing and aquaculture

B

Mining and quarrying

C10T12

Manufacture of food products, beverages and tobacco products

C13T15

Manufacture of textiles, wearing apparel and leather products

C16

Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials

C17

Manufacture of paper and paper products

C18

Printing and reproduction of recorded media

C19_C20

Manufacture of chemical, coke and refined petroleum products

C21

Manufacture of basic pharmaceutical products and pharmaceutical preparations

C22

Manufacture of rubber and plastic products

C23

Manufacture of other non-metallic mineral products

C24

Manufacture of basic metals

C25

Manufacture of fabricated metal products, except machinery and equipment

C26

Manufacture of computer, electronic and optical products

C27

Manufacture of electrical equipment

C28

Manufacture of machinery and equipment n.e.c.

C29_C30

Manufacture of motor vehicles and other transport equipment

C31T33

Manufacture of furniture and other manufacturing; repair and installation of machinery and equipment

D35

Electricity, gas, steam and air conditioning supply

E36

Water collection, treatment and supply, Sewerage

E37T39

Waste collection, treatment and disposal activities; materials recovery; remediation activities and other waste management services 

F

Construction

G

Wholesale and retail trade; repair of motor vehicles and motorcycles

H49

Land transport and transport via pipelines

H50_H52

Other transport; warehouse and supporting activities

H53

Postal and courier activities

I

Accommodation and food service activities

J58

Publishing activities

J59_60

Motion picture, video and television programme production, sound recording and music publishing activities; programming and broadcasting activities

J61

Telecommunications

J62_63

Computer programming, consultancy and related activities; information service activities

K

Financial & insurance activities

L68B

Real estate activities excluding imputed rents

L68A

Imputed rents of owner-occupied dwellings

M69_70

Legal and accounting activities; activities of head offices; management consultancy activities

M71

Architectural and engineering activities; technical testing and analysis

M72

Scientific research and development

M73

Advertising and market research

M74_75

Other professional, scientific and technical activities; veterinary activities

N77

Rental and leasing activities

N78

Employment activities

N79

Travel agency, tour operator reservation service and related activities

N80T82

Security and investigation activities; services to buildings and landscape activities; office administrative, office support and other business support activities

O84

Public administration and defence; compulsory social security

P85

Education

Q86

Human health activities

Q87_88

Social work activities

R90T92

Creative, arts and entertainment activities; libraries, archives, museums and other cultural activities; gambling and betting activities

R93

Sports activities and amusement and recreation activities

S94

Activities of membership organisations

S95

Repair of computers and personal and household goods

S96

Other personal service activities

T

Activities of households as employers; undifferentiated goods- and services-producing activities of households for own use

These are ‘official statistics in development’ as the method employed is still in development, and there are some known data quality issues. Our statement of compliance with the Code of Practice for Statistics (UK Statistics Authority) produced by the Office for Statistics Regulation (OSR) is included below, providing details of how we comply with standards expected around trustworthiness, quality and public value.

Statement of compliance with the Code of Practice for Statistics

Our statistical practice is regulated by the Office for Statistics Regulation (OSR). OSR sets the standards of trustworthiness, quality and value in the Code of Practice for Statistics that all producers of official statistics should adhere to.

All of our statistics are produced and published in accordance with a number of statements and protocols to enhance trustworthiness, quality and value. These are set out in the Welsh Government’s Statement of Compliance.

These official statistics in development demonstrate the standards expected around trustworthiness, quality and public value in the following ways.  

Trustworthiness:  These statistics were pre-announced on the Statistics and Research area of the Welsh Government website. The published figures were compiled by professional analysts using the best available data and methods. Pre-release access was restricted to eligible recipients in line with the Code of Practice. 

Quality: These statistics are produced to high professional standards and are produced free from any political interference. Methods were peer reviewed by colleagues across the UK. This release is published in line with statement on confidentiality and data access which is informed by the trustworthiness pillar contained in the Code of Practice for Statistics. 

Value: These tables give a snapshot of the Welsh Economy and detail the buying and selling relationships of goods and services between all parts of it. As part of our work over the next 12 months, we will work with users to understand how these tables can improve our understanding of the economic effects of interventions.   

You are welcome to contact us directly with any comments about how we meet these standards. Alternatively, you can contact OSR by emailing regulation@statistics.gov.uk or via the OSR website.

The Well-being of Future Generations Act 2015 is about improving the social, economic, environmental and cultural wellbeing of Wales. The Act puts in place seven wellbeing goals for Wales. These are for a more equal, prosperous, resilient, healthier and globally responsible Wales, with cohesive communities and a vibrant culture and thriving Welsh language. Under section (10)(1) of the Act, the Welsh Ministers must (a) publish indicators (“national indicators”) that must be applied for the purpose of measuring progress towards the achievement of the wellbeing goals, and (b) lay a copy of the national indicators before Senedd Cymru. Under section 10(8) of the Well-being of Future Generations Act, where the Welsh Ministers revise the national indicators, they must as soon as reasonably practicable (a) publish the indicators as revised and (b) lay a copy of them before the Senedd. These national indicators were laid before the Senedd in 2021. The indicators laid on 14 December 2021 replace the set laid on 16 March 2016.

Information on the indicators, along with narratives for each of the wellbeing goals and associated technical information is available in the Wellbeing of Wales report.

Input-Output tables are used within the calculation of a couple of the national indicators:

• The global footprint of Wales
• Emissions of greenhouse gases attributed to the consumption of global goods and services in Wales. 

Further information on the Well-being of Future Generations (Wales) Act 2015.

The statistics included in this release could also provide supporting narrative to the national indicators and be used by public services boards in relation to their local wellbeing assessments and local wellbeing plans.

[1] Miller, R.E. and Blair, P.D. (2009) Input-Output Analysis: Foundations and Extensions. 2nd Edition, Cambridge University Press, Cambridge. 

[2] The 79 GVA classifications generally match or aggregate to IO64, although we deviate from IO64 (and follow Regional Accounts) in our compilation by classifying sewerage with water collection, treatment and supply, rather than with Waste collection, treatment and disposal activities.

[3] As we will see, the subdivision of output and GVA into its components, and the final balancing required in IOT compilation means our final published estimates of GVA diverge in some cases from ONS estimates.

[4] FTE varies across industries in terms of assumed hours and one part time is set at 0.5 full time[1] This reallocation was attempted using only FTEs rather than a UK wage weighting, but created a higher number of anomalies, as well as resulting in the equalisation of implied wage across all uIO64 sectors within a ONSRA31.

[5] This reallocation was attempted using only FTEs rather than a UK wage weighting, but created a higher number of anomalies, as well as resulting in the equalisation of implied wage across all uIO64 sectors within a ONSRA31.

[6] Despite the issues with ASHE sample sizes for industries at regional scale, this is published.

[7] See Registers of Scotland House Price Statistics and ONS Residential property sales for administrative geographies: HPSSA dataset 6

[8] Note there are some minor Wales-England cross border transfers/customers, and some complex transactions between the companies in respect of infrastructure maintenance.

[9] Which does not report components of GVA for SIC36-37 separate from SIC38 waste management.

[10] Note whilst some of our IO compilation relied on the more robust 2018 ABS sample, stock changes will be volatile between years so 2019 is the better choice.

[11] Estimates of turnover are also available from IDBR but are problematic for our purpose. e.g., some include VAT, some do not. 

[12] Note however Wales produces effectively no petroleum crudes so this purchase will ‘leak’ from Wales as an import at a later stage of analysis.

[13] Some product taxes, e.g. VAT are already excluded from some datasets.

[14] There are notional exceptions, e.g. Landfills Disposal Tax, but in reality the impact will be negligible How Landfill Disposals Tax works | GOV.WALES

[15] That for real estate Real estate activities excluding imputed rents, where the ONS estimate for regional stamp duty on property was significantly higher than our proportional estimate.

[16] Aggregated to 24 IO64 groups, albeit here based on a straight average.

[17] This process was slightly more complex than reported here as the ONS system runs algorithmically on a model ‘in-house’ which is not replicable outside. Thus we have to, for example, constrain ONS provided weights to 1.0. There may therefore be errors arising in the estimate of which we are unaware

[18] The use of the Trade Survey Wales to estimate exports to RUK and ROW means that activities are already recorded for DTM sectors for these elements, and are copied across with margins associated with other demand elements reduced appropriately.

[19] Modest or zero, apart from motor trades which also includes repair

[20] Essentially the estimate of industry compensation of employees was divided by BRES estimates of regional FTE employment to return an approximate implied gross wage, and this is compared with UK ASHE median gross annual wage. Some small differences (e.g. in the treatment of pensions or overtime) will be second-order concerns in this context.

[21] These numbers are unknown for Wales in 2019 because the relevant ABS sample variable cannot be summed to represent the industry at Wales level. They are also volatile year-to-year and not policy relevant.

[22] This step obviates the need for a full rebalancing of the matrix, e.g. via the RAS technique (Open Risk Manual).

Statistician: Jonathan Bonville-Ginn
Email: InputOutputTables@gov.wales

Media: 0300 025 8099