Quality

Sample size

OSR’s recent review of Business Demography statistics discussed the sizeable growth in the number of UK businesses since 2009. This increase is confirmed by the latest ONS Business Demography Statistics (for three of the last four reported years) and by BEIS’s estimates of the UK Business Population, indicating that the total number of private sector businesses (including unregistered businesses) stands at approximately 6 million. With an increasing business population and an increased demand for detailed data, ONS will need to consider, in the short-term, increasing the size of its survey sample to meet this need.

Since 1998, the Scottish Government has funded an enhanced ABS sample in Scotland, to improve the quality and accuracy of Scottish figures. In 2018, around 2,600 extra firms in Scotland were sampled, giving a total sample size in Scotland of around 8,300 firms.

By comparison the sample size for Wales in 2018 was 2,143. Given that Scotland has 77% more private sector businesses than Wales (369,945 versus 208,830) this would suggest that Wales was under-represented.

Sample design

The sampling frame for the ABS is the list of UK businesses on the IDBR. Businesses are added to the IDBR if they are: registered for VAT purposes with HMRC; registered for a Pay As You Earn (PAYE) scheme with HMRC or an incorporated business registered at Companies House. The IDBR covers businesses in all parts of the economy, except some very small businesses; the self-employed and those without employees, neither of which would be registered for PAYE, and those with low turnover, which are not registered for VAT; and some non-profit making organisations

There are 2.6 million businesses on the IDBR; covering nearly 99% of UK economic activity. It is used by government departments, including ONS, as the sampling frame for most business surveys.

Administrative data from these sources are supplemented by data from surveys such as BRES to keep information on the IDBR up to date. In 2015, the coverage of the ONS Standard Business Survey Population was expanded to include a population of solely PAYE-based businesses. This increased the population by approximately 92,000 businesses.

IDBR Data uncertainty

OSR’s Assessment Report of UK Business Demography Statistics noted several areas of uncertainty around the quality of business data held on the IDBR. The IDBR is used as the sampling frame for many ONS business surveys and the IDBR’s quality issues therefore directly affect the quality of surveys such as the ABS. It is important that the ABS team understands these uncertainties and how they affect the quality of ABS data. The ABS team must work with the IDBR team to resolve these uncertainties.

The accuracy of the Standard Industrial Classification (SIC) codes on the IDBR relies on a full and accurate business description provided by the business and a consistent and accurate conversion of this to an appropriate SIC code. The IDBR is a live database, and in the case where a business’s activity changes, or its classification is updated following the selection of the ABS sample, there is a risk that such business activity is mis-represented in the sample. This risk is to a certain extent reduced by the ABS reclassifying businesses in its own database live environment.

ONS needs to quantify the impact of incorrectly classified businesses on the quality of ABS data. The ABS team must also work with the IDBR team to improve the accuracy of business data recorded on the IDBR.

Similarly, ONS should consider risks to the quality of ABS data arising from the uncertainty regarding the recording of business structures on the IDBR. BRES is used as the main source for updating multi-site business structures on the IDBR. BRES sample rotation and non-response, however, mean that the structures of large and medium-sized businesses are not always updated each year, so could be inaccurate.

The IDBR does not include unregistered businesses. Users of ABS data, including the ONS’s National Accounts team, must make coverage adjustments to ABS data used in the GDP Supply and Use tables. Working in partnership with BEIS, and using alternative data sources, the ABS team should seek to gain a better understanding of the contribution of unregistered businesses to estimates of total business activity. The ABS team must work with the IDBR team to understand the impact of IDBR data quality issues on the quality of ABS data.

Apportionment

Regional ABS aGVA estimates are produced at the level of the UK’s Nations and the Regions of England, and ONS provides analysis at lower levels by request. Data are collected in the ABS at the reporting unit level (which corresponds to a whole business, or a homogeneous subset of a business). These data are apportioned to produce the regional results. As a result, the regional results are less accurate than if the data were collected directly at local level

Employment information from BRES, which is collected at the local unit (site) level, is used to apportion national ABS data to use as the basis for regional estimates. The quality of sub-national data are therefore reliant not only on the quality of responses to ABS, but information about business size, industrial activity and structure on the IDBR being good quality. As described in para 1.72, business information is not updated comprehensively each year. There is no coverage of these quality concerns in ONS’s published documentation, which undermines the quality of ABS estimates, as users are not well placed to understand their capability and useability across different uses. Below the “minimum domain level” (the threshold below which the number of returns in any given year makes the data too volatile to use for regional apportionment) synthetic estimation replaces actual data.

Given the quality concerns discussed at 1.73, and the increasing demand for more-detailed regional economic data, ONS must explain to users the short-term limitations on the use of apportioned regional data, until such time that it can quantify and improve on the estimates of business structure information. This work is crucial for meeting the short-term demand for more detailed data, whilst alternative data sources are discovered in the medium term to enhance measures of regional economic activity.