This article presents analysis of numbers of workplaces and employees in London and its 33 local authorities, broken down by enterprise size band and SIC 2007 industry sector, for the period 2001 to 2012. The figures are calculated using data from the Inter-Departmental Business Register (IDBR). Although workplaces and employees are counted at local level, the methodology uses enterprise size bands that are based on numbers of employees working for the enterprise in the UK as a whole. Thus it is possible answer questions like ‘How many employees work for large retailers in Tower Hamlets?’ and ‘How has this changed over the past decade?’
Local planners and policy makers need information about businesses. The kinds of questions they ask are: How many businesses are there in my area? What sector of the economy do they belong to? Are they large or small? How many people do they provide employment for? How has the picture changed over time?
The first part of this article (Part 1: Methodology) explains how we have developed a way to address these questions, and the challenges involved. The second part (Part 2: Results) presents key results for London and each of its 33 local authorities over the period 2001-12, accompanied by a series of charts and maps. In the third part (Part 3: Next steps) we outline further analysis that will be published shortly and future work.
In 2008, Greater London Authority (GLA) Economics published a working paper entitled Employment in London by firm size (Prothero, 2008) which proposed a new method for analysing official statistics on the size structure of firms in London and other parts of the UK. The paper pointed out that official sources such as the Small- and Medium-Sized Enterprise (SME) statistics of the Department for Business Innovation and Skills (BIS) and the business statistics publications of the Office for National Statistics (ONS) present different figures on this topic, but none of them meet the policy and planning requirements of organisations like the GLA.
The 2008 GLA Economics paper proposed a new approach based on the ONS’s Inter-Departmental Business Register (IDBR). The IDBR is a statistical register based on administrative sources and surveys which contains information on businesses in all parts of the economy. It covers nearly all of UK economic activity including that of public sector bodies (which are not – strictly speaking – businesses). An organisation will be on the IDBR if it is registered for Value Added Tax (VAT), and/or pays employees through a Pay As You Earn (PAYE) scheme and/or is an incorporated business registered at Companies House1. Some very small businesses, self-employed people and non-profit-making organisations are not on the IDBR as they are not registered in any of these ways.
BIS has developed a methodology for estimating ‘unregistered’ businesses using self-employment data from ONS’s Labour Force Survey2. Such information can be added to the IDBR data if an analysis of total employment is required. However, the approach presented in this article focuses on numbers of employees (rather than total employment) and workplaces of registered businesses.
The information about employees on the IDBR is drawn mainly from the Business Register and Employment Survey (BRES). Estimates from other ONS surveys are also included. For the smallest businesses, either PAYE jobs or figures imputed from VAT turnover are used.
The IDBR database is organised on two levels: one dataset contains enterprise-level information while the other has information about ‘local units’ (the physical locations or workplaces which make up the enterprise). In the case of a small business, the enterprise and the local unit may be the same thing, but large enterprises may have many local units. For instance, a supermarket chain would appear in the enterprise-level dataset as employing several thousand people, and the entire workforce would be recorded at the address where the enterprise is registered. By contrast, in the ‘local unit’ dataset, the supermarket chain’s workforce will be recorded where people actually work, both at the head office and in the local branches. Adding up numbers of employees in all of the locations of the business recorded on the local unit database gives the employee figure on the enterprise-level database.
From the point of view of local planners and policy makers, a combination of enterprise-level and local unit level information from the IDBR is needed. This is because workers need to be counted in the location where they work (their workplaces) but they would normally be thought of as working for a ‘large firm’ if their branch were part of an enterprise that employs many people nationally, even if their own workplace was a small one.
The methodology developed for the 2008 paper by GLA Economics combined enterprise-level and local unit level information from the IDBR. First it created ‘enterprise size bands’ based on numbers of employees working for the enterprise in the UK as a whole. Then it produced analysis of numbers of people employed by local units in London within these enterprise size bands.
This analysis was further split by industry sector (the part of the economy that the business belongs to). The results showed, for example, how many employees work for SMEs in the retail sector in London. Industry sector could be defined either on the basis of the enterprise (reflecting the predominant activity of the whole firm), or at local level (reflecting the activity of the workplace). For the analysis presented in this article, we have chosen the latter approach (see Definitions). If the figures for employees working in local retailing units of all sizes in London are added up, they give the total number of employees in the retail sector in London.
It should be noted, however, that the IDBR does not provide official estimates of numbers of employees in the UK. These are published in ONS’s monthly Labour Market Statistics series (see References). ONS also recommends using the BRES for detailed industry sector and geographical breakdowns of official employee statistics.
In 2013, at the request of the GLA, ONS’s London Region Office undertook to develop the methodology proposed by the 2008 GLA Economics paper with a view to creating a standardised approach that could be used in future, and could be replicated for other parts of the UK. A particular emphasis of this work has been to produce figures for local authorities and smaller areas within London and to provide a time series as far back as possible. The earliest year for which IDBR datasets are available is 2001.
The need for results for small geographical areas has proved challenging because of the requirement to protect the confidentiality of the IDBR data. This means making sure that it is impossible for users to identify a particular business from the published results, in practice by suppressing results where there are too few local units or one ‘dominant’ local unit. This is more likely to happen when analysing results in small geographical areas than when looking at London or the UK as a whole (see below: Protecting confidentiality).
It is possible to present information for local authorities in London broken down by industry sector with relatively low levels of suppression. It is also possible to present results for the small areas known as ‘Middle-layer Super Output Areas’ (MSOAs) which nest within local authorities3, but when the analysis team tried producing breakdowns for MSOAs by industry sector most of the results had to be suppressed. Therefore, MSOA-level information, which will be published shortly (see Part 3: Next steps), will not contain industry sector breakdowns.
The following definitions have been adopted for the purpose of this work:
Enterprise size bands. Local units are placed in size bands according to the number of employees working for the business/organisation in the UK as a whole (the ‘enterprise’). These are referred to as enterprise size bands. The standard definition of a SME (used by the UK government and the EU4) is an enterprise with fewer than 250 employees. For the purpose of this project, the full set of enterprise size bands are as follows, with enterprises in the first four bands being SMEs:
zero employees (the business has one or more owners/proprietors but no employees)
500 to 2,499 employees
The full set of enterprise size bands can be analysed at London level, but for lower geographies it is necessary to combine size bands. The simplest breakdown is two categories: (i) SMEs made up of size bands 1-4 and (ii) large enterprises made up of bands 5-7. SMEs with fewer than 10 employees (size bands 1-2) are known as ‘micro’ enterprises; these include enterprises which do not have any employees.
Industry sectors. The industry sectors used for this analysis are:
Primary and utilities
Wholesale and motor trades
Transportation and storage
Accommodation and food service activities
Information and communication
Financial and insurance activities
Real estate activities
Professional, scientific and technical
Administrative and support service activities
Public administration and defence; compulsory social security
Human health and social work activities
Arts, entertainment and recreation
Other service activities
These sectors are based on ONS’s Standard Industrial Classification 2007 (SIC07) sections (also known as ‘1-digit’ classifications). However, some sections have been combined and one has been split in response to GLA requirements5.
We have chosen to define industry sector based on the economic activity of the local unit (reflecting the activity of the workplace) rather than that of the enterprise (reflecting the predominant activity of the whole firm). This is because the activity of the local unit is of most interest to local planners and policy makers. Users should note, however, that there may be substantial differences between the two approaches if enterprises have workplaces that carry out different economic activities. In addition, under SIC07 head office activities are identified separately at local unit level and appear in the professional, scientific and technical sector. For instance, in this analysis employees working at the head office of a manufacturing company would be appear in the professional, scientific and technical sector rather than in the manufacturing sector.
For years before 2009, it has been necessary to convert the IDBR data from the bases on which it was collected (SIC92 and SIC03) onto a SIC07 basis to allow comparability over the whole time series. Caution should be exercised in interpreting the pre-2009 industry sector breakdowns because the conversion process relies on modelling, which is based on assumptions (see Appendix 1). Also, we believe that the conversion process does not fully account for changes in the treatment of head office activities before the introduction of SIC07. This means that particular caution is advised when looking at pre-2009 results for places where head office activity is important (see Appendix 1). For both of these reasons, the pre-2009 industry sector breakdowns on a SIC07 basis that are presented in this article and the accompanying data tables should be treated as experimental statistics, even though the underlying data (collected on SIC92 and SIC03 bases) is a National Statistic.
Geography. Geographical location is defined at local unit (workplace) level. For instance, when reporting on the number of employees working in Havering, the numbers refer only to people who work at locations in Havering. If a firm has its headquarters in Havering but half of its employees are based at a workplace in Redbridge, the employees working in Redbridge are not included in the Havering results but are included in the Redbridge results.
Empty premises. For the purposes of the IDBR analysis presented in this paper and accompanying tables, local units with no employment have been removed on the grounds that these units are ‘empty premises’, not workplaces. In 2012, this reduced the total number of local units in London by 0.6% compared with the standard IDBR tables available on the ONS website (see References).
As already mentioned, the ONS is required to protect the confidentiality of the IDBR data. In areas where there are few firms or a ‘dominant’ firm, if results were to be published it might be possible for users to infer how many people were employed by a particular business. Although numbers of people employed is not generally seen as commercially sensitive, it is a legal requirement for ONS to protect this information because it has been provided in confidence by the businesses concerned.
For this analysis, four steps have been taken to ensure that confidentiality is protected. Results are suppressed if:
There are fewer than 5 local units in any cell of a table (primary suppression).
One local unit is dominant in any cell and it might therefore be possible for a reader to work out which that local unit was (primary suppression).
Results in other parts of a table or set of tables allow readers to work out the results that have been removed during primary suppression. This involves removing information that does not need to be suppressed in its own right but could be used to deduce suppressed results (secondary suppression).
In addition, as a fourth step, results for numbers of local units are rounded to the nearest 5 and results for numbers of employees are rounded to the nearest 100. This also helps to protect confidentiality.
Comparisons with other data sources
Most of the publications that are currently available on this topic – such as the BIS SME statistics and ONS’s business statistics – do not use the definition of enterprise size adopted here. Therefore our results (see Part 2: Results) are not comparable with them. However, employment data from ONS’s BRES will soon be available on NOMIS on a similar basis as our analysis, with enterprise size band defined at the level of the UK as a whole.
ONS generally recommends BRES as the definitive source of official employee statistics for detailed analysis by industry sector and small geographical areas. However, for this analysis we have chosen to use the IDBR because it is possible construct a time series from 2001 to 2012, whereas BRES data is only available for 2009, 2010 and 2011. Also, for the BRES data published on NOMIS, enterprise size bands are expected to be based on the total number of people employed (including owner/proprietors). By contrast, the standard UK Government and EU definition of an SME is based on employee numbers (see above: Definitions).
Nevertheless, readers should be aware that BRES data will shortly be available on a similar basis, and that there are some differences between the results based on the IDBR and those based on BRES. Apart from the difference between the use of employee-based and employment-based size bands, these differences are mainly attributable to the following factors:
The BRES is a survey, so it provides an up-to-date snapshot at a point in time (September each year). The IDBR is a register of businesses. The IDBR extracts used for this analysis are taken in March, but the employment information on them is drawn from the BRES of the previous September and from other ONS surveys. The IDBR is not updated in full every year because only a sample of enterprises is surveyed each year. Therefore the employment information on the IDBR relates to a variety of time periods.
In the IDBR, much of the information on micro enterprises comes from PAYE records or is imputed from VAT turnover. It is for such enterprises that the differences between the IDBR and BRES estimates are greatest.
The information presented in this section is based on analysis of ‘local units’, or workplaces. Workplaces may belong to private sector businesses or to public sector or not-for-profit organisations. Workplaces are allocated to enterprise size bands according to the number of employees working for the business/organisation in the UK as a whole (see above: Definitions).
The analysis looks at numbers of workplaces and numbers of employees in London and its 33 local authorities. These two variables are classified according to enterprise size bands and which industry sector (or part of the economy) they belong to.
Size of firms in London
Chart 1 shows that the number of workplaces in London rose from 366,290 in 2001 to 414,375 in 2012. There were decreases in the early 2000s and in 2010-11. The number rose in 2012, but the magnitude of the increase is overstated in the figures, as improvements to HM Revenue and Customs computer systems lead to the inclusion in 2012 of businesses that had not previously been recorded on the register. The pattern was similar in the UK as a whole, where the number of workplaces rose from 2.4 million to 2.6 million over this period.
The growth in the number of workplaces was accompanied by growth in numbers of employees between 2001 and 2012. According to the IDBR, the number of employees rose from 3.6 million to 4.2 million in London and from 23.9 million to 26.8 million in the UK, with slower growth or downturns in the early 2000s and in 2010-11. It should be noted that the IDBR figures are not official estimates of numbers of employees in London and the UK. These are published in ONS’s monthly Labour Market Statistics series (see References).
Chart 2 shows that the proportion of SME workplaces in London has remained steady since 2001 at 88-89% of all workplaces. The proportion of SME workplaces is slightly lower in the UK as a whole than in London: it was around 85-86% over this period.
Workplaces belonging to large firms (with 250 or more employees) comprised 11-12% of workplaces in London between 2001 and 2012. Most of these were associated with enterprises with 2,500 or more employees.
Most SME workplaces belong to micro enterprises (those with fewer than 10 employees): in 2012, 87% of SME workplaces in London were associated with micro enterprises. Within this category, the proportion with zero employees fell over this period, while the proportion with 1-9 employees rose. This is understood to be because of the increasing propensity for sole traders to register as limited companies because of the legal and commercial advantages available and because registration has become easier over the past decade. When a sole trader sets up as a limited company, they are recorded as an employee of that company rather than as an owner/proprietor.
Chart 3 shows the proportion of employees in London working for different sizes of enterprise. Although SME workplaces comprise nearly 90% of workplaces in London, only two-fifths of employees work for an SME. Over three-quarters of London’s workplaces belong to micro enterprises, but micro enterprises provided work for only 15% of its employees in 2012. Most employees work for large enterprises, and almost four out of ten employees in London work for a firm employing 2,500 or more employees. The proportions are similar for the UK as a whole.
Chart 4 shows that in 2012, the top five industry sectors in London in terms of where employees worked were the professional, scientific and technical sector, the administrative and support services sector, the human health and social work sector, the retail sector, and financial and insurance activities. Nearly half of all employees in London worked in one of these sectors. The biggest increases in numbers of employees over the decade to 2012 were in real estate activities, education, and health and social work activities. Numbers also appear to have risen in the professional, scientific and technical sector, but much of this is likely to be due to improved classification of head office activities over this period. It has not been possible to capture this classification change fully in our time series (see Chart 6 below and Appendix 1).
Chart 5 shows that most industry sectors saw little change in the proportion of employees working for SMEs over the decade to 2012. The main exception was the education sector, which saw a sharp increase. It should be noted that an increase in the proportion of employees working for SMEs does not necessarily imply that there are more SMEs or that SMEs are expanding; it may be because large businesses have been broken up or closed down. In the case of the education sector, it may also reflect changes in government policy.
Overall, the industry sector with the highest proportion of employees in SMEs in 2012 was ‘other service activities’ (74%), while the sector with the lowest proportion (2%) was public administration and defence.
Size of firms in London’s local authorities
ONS is publishing detailed tables for London’s local authorities alongside this article (see below: Data available with this publication). This section provides examples of the analysis that can be produced using these tables by looking at a selection of industry sectors and local authorities.
Chart 6 shows the ten local authorities in London with the largest number of employees in the professional, scientific and technical sector in 2012. In terms of where employees work, this was the largest industry sector in London in 2012, with 513,700 employees. It should be noted that there are also many self-employed people working in this sector, but they are not covered by this analysis.
The professional, scientific and technical sector includes legal and accounting firms, management consultancies, architectural and engineering firms, scientific research bodies and advertising/market research firms. It also includes head offices of firms whose main activity may be in other industry sectors. In some London local authorities, employees working at head offices accounted for a large proportion of those working in the professional, scientific and technical sector in 2012 (for instance 19% in Westminster and 43% in Hillingdon, see Appendix 1).
Chart 6 does not show figures for earlier years. This is because comparisons over time may be misleading for the professional, scientific and technical sector in some local authorities because the process of converting the IDBR data from SIC92 and SIC03 onto a SIC07 basis (see Definitions) does not fully capture changes in the treatment of head office activities before the introduction of SIC07. Therefore, employees of head offices may be recorded in other sectors before 2009, and numbers of employees in the professional, scientific and technical sector are likely to be underestimated.
The proportion of employees in the professional, scientific and technical sector working for SMEs varies considerably between local authorities. For instance, in 2012 it was 60% in Westminster but only 36% in the City of London.
While jobs in the professional, scientific and technical sector are concentrated in certain local authorities, in retailing they are more evenly spread across London. With the exception of Westminster, where there were 57,500 employees working in the retail sector, the number of employees in each local authority in 2012 ranged from 4,000 in Barking and Dagenham to 20,200 in Kensington and Chelsea.
Chart 7 shows how the numbers of employees have changed in the retail sector in selected local authorities over the past decade. There have been big increases in Hammersmith and Fulham and in Newham (probably linked to the opening of the Westfield shopping centres). In both these boroughs, the increase in numbers of employees has been accompanied by a decrease in the proportion working for SMEs, from 51% to 27% in Hammersmith and Fulham and from 38% to 31% in Newham. Numbers of employees in the retail sector have also risen in Tower Hamlets. Redbridge and Sutton have seen a decline in numbers over the decade, while in Brent and Croydon there has been little change.
Financial and insurance activity takes place mainly in three London local authorities: City of London (with 151,300 employees in 2012), Tower Hamlets (with 75,200) and Westminster (with 37,900). Chart 8 shows that each of these local authorities has a different profile in terms of numbers of employees working for large firms in the financial and insurance sector, and that these patterns have been diverging over the past decade. There has been increasing concentration in the City and Tower Hamlets (where 76% and 94% of employees respectively worked for large enterprises in 2012), while the opposite has been happening in Westminster (where only 39% worked for large enterprises in 2012).
It is also possible to show this kind of information on maps. Maps 9 and 10 show the distribution of employees working for SMEs and large enterprises in London’s accommodation and food services sector in 2012. In Westminster, there were large numbers of employees working for both SMEs and large enterprises. In the City of London, large enterprises were the main employers, with over 10,000 employees in 2012; there were fewer than 5,000 employees working for SMEs. By contrast, an employee working in Hammersmith and Fulham or in Southwark in 2012 would be more likely to have been working for an SME than a large firm.
Map 10: Number of employees in large enterprises in London's accommodation and food services sector, 2012
Data available with this publication
The data tables published with this article present breakdowns of IDBR data at London level and for each local authority within London. UK figures are also shown for comparison. The tables show numbers of workplaces and numbers of employees by enterprise size band and by industry sector. They answer the questions: How many businesses are there in my area? What sector of the economy do they belong to? Are they large or small? How many people do they provide employment for? How has the picture changed over time?
The following tables are available with this publication for 2001 to 2012:
Workbook entitled ‘Size of firms in London and the UK by enterprise size’:
Table 1: Number of workplaces in the UK and London by size of enterprise, 2001-2012 (all seven enterprise size bands)
Table 2: Number of employees in the UK and London by size of enterprise, 2001-2012 (all seven enterprise size bands)
Table 3: Number of workplaces in the UK, by industry sector and by size of enterprise, 2001-2012 (three bands: <10, 10-249 and 250+ employees)
Table 4: Number of employees in the UK, by industry sector and by size of enterprise, 2001-2012 (three bands: <10, 10-249 and 250+ employees)
Table 5: Number of workplaces in London, by industry sector and by size of enterprise, 2001-2012 (three bands: <10, 10-249 and 250+ employees)
Table 6: Number of employees in London, by industry sector and by size of enterprise, 2001-2012 (three bands: <10, 10-249 and 250+ employees)
Workbook entitled ‘Size of firms in London local authorities by enterprise size’:
Number of workplaces and employees by enterprise size band (three bands: <10, 10-249 and 250+ employees) and by industry sector for each of the 33 London local authorities
ONS will publish another article on 8 August 2013 entitled Small and Large Firms in London, 2001 to 2012. This will contain the following analysis of MSOAs in London for 2001 to 2012:
Number of workplaces by enterprise size band (two bands: <250 and 250+ employees) for each MSOA in London
Number of employees by enterprise size band (two bands: <250 and 250+ employees) for each MSOA in London
Now that a standardised approach to producing analysis by enterprise size band from IDBR data exists, similar analyses could be undertaken for other parts of the UK. It would also be possible for local planners and policy makers to add estimates for the self employed to the IDBR figures, following the methodology developed by BIS (see Part 1: Methodology).
Until recently, the only low-level geographies available to analysts were based on political boundaries, such as parliamentary constituencies, or on administrative boundaries, such as local authorities, or on where people live1. None of these approaches is designed for analysing business data such as that contained in the IDBR. However, ONS has recently published a new classification known as ‘Census Workplace Zones’, which is based on information about where people work from the 2011 Census. This provides the foundation for improving geographical analysis of businesses.
The Census Workplace Zones are very small, so they cannot be used for IDBR analysis without compromising confidentiality. Nevertheless, it should also be possible to group Census Workplace Zones together to create larger business zones for which results can be presented. This is a key challenge for the future.
Different researchers may group the Census Workplace Zones together in different ways, reflecting different requirements. However, this could create problems because once results have been published for one grouping, other groupings may no longer be possible. This is because if there are overlaps between different groupings, publication of results might compromise confidentiality since information about particular businesses could be deduced by comparing the results. To prevent this happening, standardised groupings of Census Workplace Zones are needed, like those which are already available for geographies based on where people live. A project is now under way, coordinated by ONS Geography, to create such standardised groupings.In putting together the groupings, this project must take into account the need for the areas to be suitable for analysing other business statistics sources such as the Annual Survey of Hours and Earnings (ASHE) and the BRES as well as the IDBR. It must also take into account the probable needs of different types of user, from policy makers and planners to academics and members of the business community. Furthermore, it must take into account existing residential population-based groupings such as MSOAs. Therefore, this is a complex task which is likely to take time to achieve.
Details of the policy governing the release of new data are available by visiting www.statisticsauthority.gov.uk/assessment/code-of-practice/index.html or from the Media Relations Office email: firstname.lastname@example.org
These National Statistics are produced to high professional standards and released according to the arrangements approved by the UK Statistics Authority.
The IDBR datasets for 2001 to 2012 contain information that was collected each year on the basis of the UK Standard Industry Classification (SIC) that was in place at the time. For the 2001 and 2002 datasets, this was SIC92. For 2003 to 2008 it was SIC03. For 2009 to 2012 it was SIC07. This presents problems for users wishing to look at change over time. Therefore, the results presented here have been converted onto a consistent (SIC07) basis.
However, this means that all breakdowns by industry sector before 2009 contain an element of modelling (required for the conversions) and users should be aware of the assumptions used in this modelling. Also, we believe that the conversion process does not succeed in putting the classification of head office activities before 2009 onto a basis that is fully comparable with SIC07. This appendix explains how the conversion process was done and also discusses head office activities.
Creating SIC07-consistent industry sectors before 2009
Originally, the conversion was done in two stages, both using information at local unit level:
Data collected on a SIC92 basis was converted onto a SIC03 basis using a table showing the relationship between 5-digit SIC categories. This was relatively simple because the changes were between 4-or 5-digit SIC categories and had no impact at 1-digit level, which was the level at which the data was being analysed.
Data on a SIC03 basis (whether originally collected on this basis or converted from SIC92) was converted onto a SIC07 basis. This was complicated because the change from SIC03 to SIC07 affected a large number of lower-digit categories and had an impact at 1-digit level. The method used involved a lookup table and a weighted correlation table. This probability-based allocation process is described in detail below (see Converting from SIC03 to SIC07).
The results of all probability-based conversion processes are imperfect. In order to improve the outcome, we decided that if we had the SIC07 code of a local unit after SIC07 was introduced and this local unit had been in the dataset before then, we could use this information to 'backcast' SIC07 codes for previous years. The process we adopted was: if we knew what the SIC07 code of the local unit was in 2009 to 2011 and it was the same code in all three years, we assumed that this code could be allocated to the local unit (and its employees) in previous years if the local unit existed in those years.
Thus the final results are based on a combination of two methods: a) a probability-based process of conversion from SIC03, and b) 'backcasting' SIC07 codes. Table A1 shows the proportion of employees in London for whom industry sector before 2009 is the result of backcasting.
|% of employees for whom industry sector is backcast||38||42||49||53||58||62||66||70|
|% of employees for whom industry sector is based on conversion from SIC03||62||58||51||47||42||38||34||30|
Converting from SIC03 to SIC07 using the probability-based allocation process
ONS provides two tools for converting data from SIC03 to SIC07. The spreadsheet ‘Correlation between SIC 2003 and 2007’ is a lookup table showing the relationship between each of the 5-digit SIC03 codes and their SIC07 equivalents. However, one SIC03 code may become several SIC07 codes and some SIC07 codes cover more than one SIC03 code. Therefore it is necessary to use the ‘Weighted Tables with percentages SIC03 - SIC07’ workbook, which shows what proportion of each SIC03 code will end up in the SIC07 code shown in the lookup table. This is provided separately for number of local units (count %), employment and turnover. We used count % and employment % at 5-digit level, with the latter being used as a proxy for employees.
To show how this works, take the example of a workplace which was in SIC03 5-digit code 01210 and had 100 employees. According to the correlation table (count %), this SIC03 code was split such that 74% was allocated to SIC07 code 01410 and 26% was allocated to SIC07 code 01420. Therefore the workplace was split such that it was counted as 0.74 units in category 01410, and 0.26 units in category 01420.
For producing analyses of the number of employees by industry sector, the conversion from SIC03 to SIC07 needed to use the employment % in the correlation table (rather than the count %) and this had to be applied to employees individually rather than to local units, which might contain several employees. Not doing so meant that sectors with larger workplaces ended up with a higher proportion of employees than they should, while those with smaller workplaces got a lower proportion. Extending the example, according to the correlation table (employment %), 71% of employment in SIC03 code 01210 was allocated to SIC07 code 01410, and 29% to code 01420. Therefore 71 of the 100 employees in the workplace would be allocated to 01410 and 29 employees would be allocated to 01420.
This complication meant that it was necessary to do the analysis of numbers of employees by industry sector separately from the analysis of number of local units by industry sector. Therefore, when interpreting tables by industry sector that use data converted from SIC03 (before 2009), the employees in a particular cell of the 'number of employees' table do not necessarily work for the local units in the equivalent cell of the 'number of workplaces' table. Conceptually, this is because a small number of records relating to local units with more than one employee will have been split between 5-digit SIC07 categories during the SIC03 to SIC07 conversion process.
Head office activities
During the analysis, we found evidence to suggest that the probability-based conversion process is not able to identify and convert all head office activities onto a SIC07 basis. The issue arises because originally, in SIC92, head offices were recorded against the principal activity of the firm to which they belonged (they had no separate SIC code). In SIC03, this approach was retained at enterprise level, but local units which were head offices were recorded under code 7415 (management activities of holding companies). In SIC07 activities of head offices were split out from those of holding companies and given a unique code (70100) which puts them into the professional, scientific and technical sector.
The process of converting data from SIC03 to SIC07 based on lookup and weighted correlation tables should, in theory, make it possible to convert from SIC03 code 7415 to SIC07 code 70100. However, the results indicate that this was not as successful as other parts of the conversion process. Moreover, it is not possible to convert head office information from data originally collected using SIC92 because head offices were not identified by a separate SIC code under SIC92.
Therefore, before 2009, numbers of employees in the professional, scientific and technical sector are likely to be underestimated and numbers in other industry sectors may be overestimated because employees of head offices may be recorded in other industry sectors rather than in the professional, scientific and technical sector. This means that caution is advised when looking at pre-2009 breakdowns by industry sector, especially in early years when much of the data is based on the probability-based conversion process rather than on 'backcasting' (see Table A1 above). This is particularly so in places where head office activity is important (see Table A2).
|Kensington and Chelsea||18|
|Hammersmith and Fulham||9|
|Kingston upon Thames||8|
|Barking and Dagenham||8|
|Richmond upon Thames||7|
|City of London||2|