Health and Safety
Executive / Commission
Statistics
HSE use estimates collected from a number of ONS Employer Surveys based on the Inter-Departmental Business Register (IDBR).
The STES takes a sample of 24 000 enterprises from the (IDBR). The STES is run every quarter (June, September, December and March). It takes ONS about three months to produce the latest quarter estimates.
Although HSE receive STES data each month from ONS, HSE make no calculations on the estimates received on a monthly basis: these files would only be used if in-year rates were required. Instead, HSE make calculations on estimates received at the quarterly stage – i.e. estimates for June received in September, estimates for September received in December etc.
Data are for employees only and are broken down by industry, sex and full or part-time employment. ONS send data for Great Britain and also England, Scotland and Wales separately, each quarter. Some SICs are merged in the STES, such as SIC 36/37. To determine employment data for such SICs, HSE compile an estimate using the ratio of, for example, SIC 36 and SIC 37 from the ABI to split the combined figure from the STES.
As the ABI survey was first run in 1997 and replaced the Annual Employment Survey (AES) as ONS’ large-scale survey with a large sample size in 1999, it is useful to set out some background to the AES first.
The AES was first run in 1995 and was carried out each September until it was replaced by the ABI. AES samples were drawn from the IDBR using approximately 120 000 enterprises selected according to certain criteria. The selected criteria meant that the survey covered approximately 400 000 workplaces each year.
For each AES, a sample of enterprises was selected from the IDBR and AES forms sent to the addresses at which employers holds all the pay records for particular parts of their businesses; these are known as 'Reporting Units'. Reporting units are generally equivalent to enterprises. Forms were required for each separate site of a business or for each part of a business within a site in which a distinct business activity was carried out or a separate PAYE scheme is operated. Each part of a business for which a separate return is made is called a 'local unit'. A local unit may, therefore, be a site of a business or part of a site, although in the main, a local unit will represent an entire factory, office or shop and will thus include all employees working there.
Each September, local units were required to indicate how many employees it had. Hence the AES gave a 'snapshot' of the number of employees in employment at one particular point in time each year (this being the case, the data received was different to yearly averages obtained from other surveys, and any seasonal variation would be missed). Also to note is that, although the initial sample was drawn on an enterprise level, the data was actually collected at the more detailed local unit level. The AES only provided data for employees. Working proprietors were not included in the survey; however, company directors on salaries (usually classed as employees) would be included.
The move to ABI – to improve coherence and reduce duplication across ONS’ range of business surveys – took longer than expected due to the level of employee jobs obtained from the ABI being significantly higher than the level obtained from the AES. The main reasons for the differences were found to be under recording by some contributors of their full employee numbers and the estimation systems used by the ABI and the AES. The AES is shown to have been under counting in the past and the new higher ABI estimate of employee jobs found to provide a better estimate. The ABI estimates are more closely aligned with Labour Force Survey (LFS) figures.
The methodology of the ABI is based on the AES. The ABI, like the AES is a point in time estimate of employment; however the ABI is conducted in December rather than September. The ABI became the main employer survey after the last AES in 1998.
The aim of the ABI is to collect employee jobs data from the same sample of businesses that are asked to provide economic data that fee into the National Accounts. Collecting jobs and economic data from the same sample of businesses, and making estimates using standardised estimation techniques enables derived variables such as wages per head and output per head on a much firmer and more consistent basis. The ABI is in two parts, ABI/1 collects information on employment for a December survey date and ABI/2 collects a range of economic data (turnover, purchases etc) for the calendar or financial year.
If more detailed employment data for employees are required than what is given by the STES, then data from the ABI can be broken down by industry to 4-digit SIC level. However, there are a couple of things to note. Firstly, because this is a different survey (an annual ‘snapshot’ of employment rather than the average of 4 quarters), data for aggregated industry codes from the ABI will not be exactly the same as data for the same codes from the STES. Secondly, some 4-digit SIC codes for which there are very few employees are classified as confidential and must not be passed on by HSE to a third party, even in the form of an injury rate – since the data could be derived from the injury numbers and the injury rate.
To obtain ABI data a permissionary note called a “Chancellor of the Exchequer Notice” and a NOMIS account is required. Setting up a NOMIS account is free and can be done through the NOMIS website. This also contains the application forms for the Chancellor of the Exchequer Notices which cost £100 and last until the next ABI dataset is released in December each year.
The LFS is based on a sample of households rather than a sample of employers. It is a quarterly survey of approximately 60 000 households. The survey is carried out in five waves of 12 000 households which are interviewed in five successive quarters. Within each quarter there will thus be five groups of 12 000 households – each group on a different wave of the survey. There is thus an 80% overlap in the sample for each successive quarter. The LFS is based on a simple random sample design, which is representative of the whole of Great Britain.
Since 1992 the LFS has been run on a quarterly basis. Data are collected throughout each quarter in a different way to the quarterly employer data of the STES. Instead of a point-in-time estimate of employment in March, June, September, and December, the LFS yields quarterly averages for March to May, June to August, September to November, and December to February. Hence the LFS data are centred on the months April, July, October and January rather than March, June, September and December.
Employment data for a wide range of breakdowns are available from the LFS (including occupation and age). HSE use the LFS for all of these purposes, but particularly for employment data for the self-employed by industry.
Again, because this is a different survey, estimates of employment for employees from the LFS will not be exactly the same as those from the STES and ABI. However, to improve consistency between LFS and STES/ABI estimates, from 1996/97 estimates for the self-employed include people who are employees in their main job but self-employed in their second job.
LFS estimates of employment do not undergo the same revision process as the STES estimates. This simplifies things a little since all data received from ONS can be treated as final. The LFS works in a different way to the STES. Instead of estimates of employment in June, September, December and March, the ‘June’ LFS estimate actually refers to the period March-May. Similarly, the September, December and March estimates refer to the periods June-August, September-November, and December-February respectively.