Comments from ACCA to the Department for Business, Enterprise and Regulatory Reform (BERR), March 2009.
ACCA welcomes this opportunity to offer our twin perspectives on the BERR SME Statistics: our members' perspective as providers of business statistics and that of our technical experts as users of such.
SME Statistics are a useful, high-quality resource, and we are keen to see them achieve wider and better use. We believe that the most important function of SME Statistics is to help determine the stakes involved in policy interventions and inform impact assessment. This has very specific implications for the compilation and presentation of the data.
We believe that accuracy is more important for the function of SME Statistics than inter-temporal consistency. It is also very important that the presentation of statistics is consistent with other widely-used publications, such as ONS' UK Business: Activity, Size and Location series. We support the publication of non-disclosive data at low levels of confidence, but would like to see users made more aware of the uncertainty involved in such estimates, and encouraged to qualify their own assessments accordingly.
From a technical point of view, we believe that more detailed use of LFS data and validation of assumptions through existing, large-scale SME surveys can improve the quality and relevance of estimates without adding to the substantial burden of preparing business statistics. We would also like to see a robust definition of business activity that will be relevant to policymaking by focusing on capturing real enterprises operating in real markets, whatever the stage of their development.
Finally, we believe that users of SME Statistics are now far more likely to need the capability to tailor their data queries, and that the presentation of data can be used constructively to help policymakers and other users control for the influence of policy and regulation of SME trends.
ACCA welcomes this opportunity to offer our views on the SME statistics prepared by BERR's Analytical Unit. Ours is a twin perspective, as our members are the primary providers of statistics on registered businesses, and our technical experts are frequent and engaged users. BERR SME statistics are an especially useful resource for our SME Unit, used regularly to demonstrate the importance of the SME agenda, but also frequently in assessing the impact of policy and sizing the market for different policy interventions. On the whole, we extremely satisfied with the level of detail and quality of the information provided.
We believe that market sizing and impact assessment are easily the most important uses for SME statistics, influencing the decisions and effectiveness of businesses, regulators and local government. This suggests to us that some long-term intertemporal consistency can be sacrificed in favour of accuracy in the static sense. Consistency with other often-used data and publications should, however, remain a high priority.
We believe that basing the production of statistics on unregistered businesses largely on LFS data is the correct approach. Using official information from sources such as DWP or HRMC (especially data on NI contributions) involves implicit assumptions about rates of compliance or use of services, which can be difficult to validate. If such information is to be used, assumptions on compliance rates will have to be made explicit and verified regularly.
Finally, we would strongly oppose the creation of new statistical reporting requirements for the purpose of compiling SME statistics, as data of reasonably good quality can be sourced from ONS without further burdening small businesses.
As with all survey data, LFS information is subject to error, and this should be made explicit to users, who may otherwise not fully appreciate the degree of uncertainty attached to the estimates they produce. We fully support the policy of not suppressing data due to low confidence levels but believe that there are user-friendly ways of accounting for these, as for instance with the reporting of data from the Annual Survey of Hours and Earning (ASHE), where findings are shaded to provide a visual representation of confidence levels. Where possible, users should be encouraged to consider confidence intervals alongside the precise estimates – it is unlikely that many will do so unless some explicit reference is made.
It can be difficult to distinguish between genuine working proprietors and special labour market arrangements within LFS data. We are aware of a similar difficulty emerging in a recent attempt to estimate the size of the UK freelance workforce. That work, carried out by Kingston University researchers, relied on LFS data on payment methods in order to separate those genuinely working for themselves from other types of self-employed individuals. We too believe this to be the most appropriate approach.
The recent BERR Survey of SME Finances (2007) found that the average number of proprietors/directors involved in the day to day running of both partnerships and limited liability companies with no employees was smaller than what is assumed by BERR statistics (1.3 per business as opposed to 2). Consequently the BERR statistics may be overstating the number of self-employed business partners by about 140,000, and significant variation between sectors may have been disguised.
The assumption that unregistered businesses have half the average turnover as registered businesses with no staff appears plausible. A naïve examination of the 2007 BERR SME finances survey data suggests that the turnover for unregistered businesses was at least ca. £27,700 (37% of registered), and an estimate based on range mid points was ca. £48,200 (64% of registered). Where the derived average turnover figures breach VAT thresholds, it may make more sense to attribute this to the joint effects of tax deductible costs, exemptions and non-compliance than to cap average turnover.
Most important, however, is the need to corroborate the 50% ratio and other assumptions regularly using large surveys of SME finances – BERR could look into using the existing Survey of SME Finances for this purpose, provided it is adequately expanded and resourced.
Users are often confused by the distinction between IDBR, VAT and BERR data on business figures, and will only have become more so following changes to the UK Business: Activity, Size and Location publication in 2008. We must note that this does not substantially affect the relevance of headline data used to establish the important of the SME sector: smaller businesses' share of employment, turnover and growth may vary under different definitions but the importance of the sector is easy to demonstrate regardless of the definitions employed. Inconsistencies are, however, much more likely to compromise policymaking through impact assessments. It would therefore be appropriate to streamline the approaches of these equally respected publications in order to avoid confusion and errors.
That said, it will always be useful for policymakers to know how many businesses are subject to VAT (regardless of PAYE registration). VAT was central, for instance, to the fiscal stimulus package unveiled in last year's PBR. Since 2008, this information has not been separately available in UK Business: Activity, Size and Location.
In line with our argument that impact assessment is the more important function of SME statistics, we would favour a definition that includes only those businesses active at the start of the calendar year. The major consequence of counting the number of businesses operating at any time during the year would be to include failed start-ups and non-recurring, short-term self-employment in estimates. This would hopelessly conflate labour market trends with the business start-up cycle, and furthermore introduce a large number of ‘businesses' that are unlikely to be affected by regulatory changes or respond to policy through changes of behaviour.
Ad-hoc contract work can, of course, eventually crystallise into genuine freelance work or business start-ups. But this effect can be captured equally well through first- or second-job self-employment at the start of the year. We advise accordingly below.
We do not agree with excluding second job self-employment (regardless of SIC) from estimates of the total number of self-employed individuals. As a waiting room for full-time self-employment, this can be a useful indicator of early enterprise activity. The assumption that the economic contribution of second-job self-employment is very low is irrelevant as these figures can be extremely useful for designing and assessing enterprise policy.
Our argument, it must be noted, is only valid if applied to the stock of businesses active at a particular point in time. If the full stock of businesses operating at any point in the calendar year is used, then it may be better to exclude second-job self-employment as a proxy for ad-hoc work.
We welcome the suggestion of pooling Q4 and Q1 LFS data, as self-employment data from LFS will always be subject to the survey's sample limitations. Our concern is that, once selecting for region, sector and self-employed status, LFS data can become very unreliable, excessively volatile and often disclosive. Worse, statistically insignificant differences in regional or local business figures can, especially in the current economic environment, become highly publicised and prompt actual policy (over)reactions.
We appreciate that ONS and the Analytical Unit work under limitations which make the frequent production of detailed data very difficult. That said, accurate and up to date information is extremely important for users of SME statistics. For providers of statistical information, the lack of timeliness in statistical releases in general can be extremely frustrating; statistical reporting consumes substantial resources and providers need to see a return on their time and effort. The data available within the current scope of the publication have in the past been, and should continue to be, available annually. Additional data would probably need to be published biannually, and we advise accordingly below.
Policymakers studying the start-up or business cycles will find that the inclusion of nonprofits biases assessments of the private sector much more than it would assessments of the public sector. Thus the not-for-profit sector should not be considered part of the private sector for the purposes of BERR SME statistics. The behaviour of non-profits is less reducible to economic fundamentals than that of profit-seeking companies, their regulation is in many ways different and many types of non-profits are strongly countercyclical in the volume of their employment and activity.
For assessments of the relative importance or productivity of different sectors, value-added is in many ways a much more relevant variable than turnover and such data is not impossible to report on the basis of IDBR and Annual Business Inquiry data (plus appropriate assumptions). But impact assessment rightly makes much more extensive use of turnover. If reporting on both is impossible, then we agree that turnover should take precedence as it currently does.
ACCA came across the limitations of group-level (3-digit) SIC in a recent attempt to determine a naïve measure of the SME sector's ‘ideal' share of public procurement. Many (though not most) of the SIC group-level sectors yield only disclosive data – and many more will do so at the class and subclass level. Information at this level is no doubt very useful, not least to many industry and representative bodies whose remits cannot be defined at 3-digit SIC. For instance, accounting practices are spread across SIC 74.12 and 74.14/2, and many intuitive sector groupings in financial services are impossible above the subclass level. We therefore see a use for SIC class and even sub-class data, but would not expect these to be published on an annual basis.
We believe that sub-regional information is, strictly speaking, less significant than sub-sector information, as it is less relevant to impact assessments. We appreciate that many other stakeholders, especially at the regional and local level, will take a different view. More importantly, we appreciate that as government functions (including regulatory services) continue to decentralise, more impact assessments might be required at the local level. This would justify sub-regional publications, although we expect that these would still be highly aggregated, as most of the users of regional SME statistics rely mostly on headline data. Until such changes take place, releasing sub-regional information on a biannual basis (and possibly using pooled data from a biannual sample) would be the best course of action. Detailed reports at regional and sector level (see above) should ideally alternate, allowing for reasonably up-to-date extrapolations.
The size-bands employed by SME statistics are broadly suitable to the needs of most users. For the purposes of post-hoc impact assessment, however, we would argue that a ‘1 employee' size band be included in the published figures, wherever a zero employee size band is also provided. BERR research confirms that the impact of all types of regulation rises dramatically as soon as employers take on staff and become subject to employment regulation. But employers and non-employers are too different in many respects for non-employers to be a suitable control group. Even the 1-4 size band incorporates a relatively broad range of businesses in terms of size and turnover. A ‘1 employee' size band, on the other hand, would incorporate businesses virtually identical to zero employee businesses in every respect bar employment.
Realistically, most users of SME statistics will either be interested in headline information on business numbers, employment, and turnover by size band, or a much more detailed analysis than any release can incorporate. Statistical releases would ideally contain headline information, a schedule of future releases, and links to the detailed data. BERR could additionally consider offering an automated alert facility which would notify users of new SME statistics releases.
Specific areas of interest such as social deprivation, rural, enterprise or sector-specific policy, could benefit from an expansion to the scope of BERR SME statistics. While in principle we would welcome the release of any information on additional aggregations (especially business age), we appreciate that this would be a costly and time-consuming exercise. Rather than commit substantial resources for the benefit of very few users at a time, we would encourage BERR and the ONS to reflect on such issues through ad-hoc papers and data releases, which can then be much better suited to the policy issues at hand and tailored to the needs of potential users.
We appreciate that the existing format of SME statistics is optimised for printing. For the purposes of analysis across regions, size-bands, industry sectors and years however, we believe a pivot table format would be more suitable.
TECH-CDR-851 – Consultation on SME Statistics for the UK and the Regions
The LFS may also be affected by a reluctance to report unregistered activity, but this is much less likely for an ONS survey than for data solicited by government agencies capable of taking enforcement action.
J. Kitching and D. Smallbone, ‘Defining and estimating the size of the UK freelance workforce.' PCG, October 2008.
The variable in question is SELF1-4 ‘Other methods of payment aside from receiving a salary or wage direct from an employer' See ONS, ‘ LFS User Guide Volume 3: Details of LFS Variables' 2005.
BERR Survey of SME Finances, 2007. The survey found that SMEs' mean ratings of regulation and tax issues as a burden (1= no problem, 10 = critical problem) remained constant at 4.3 across size bands for employers. For non-employers, the mean was 3.3. This pattern is not replicated in among turnover size bands.
Last updated: 11 Apr 2012