By Paul E. Atkinson
Directorate for Science, Technology and Industry
I very much welcome the opportunity to speak about the way the current availability of data and basic statistical information looks to a would-be user of these data. My perspective is not that of an expert in either SME issues or the existing statistical base -- I am relatively new to this field -- but of an economist who has had to deal with a wide range of issues at various times and now confronts a new set in the SME domain. From this perspective, what I would like to have as a user at some general level is pretty much the same across domains. The scope this provides for comparison offers some perspective.
An important point of departure is the question why do we want data and statistics? Here I come to two basic answers which are really the same ones that I reach when I pose this question about other areas of economic analysis or policy.
First, we want an empirical base to inform policy makers who formulate, design and implement policies. At the most basic level we want to know what is happening, or just an overview of the factual situation. This is at least a starting point for examining issues such as what is important? What is not? Which variables are large in the overall context? Which ones are small? Which are growing rapidly? Which look satisfactory and which unsatisfactory? This gives policy makers empirical guidance that goes beyond theoretical considerations and general principles as to where to direct their attention and priorities.
At another level we want to know why it is happening, i.e. we want some assessment of the forces shaping events. Pure data and statistics, of course, can never answer this; one always needs an analytical framework built on theory. But data and statistical information provide the basis for calibrating models or theoretical frameworks. This in turn allows us to make judgements about which forces are important and which can safely be neglected in a world where, theoretically, everything pretty much depends on everything. Good analysis based on economic principles is extremely important, but if it is not empirically based it risks becoming abstract and focused on things that are not important.
Concretely, in the domain of formulation, design and review of SME policies and programmes, what do we want to do with our empirical base? We can list the following, and the list is probably not complete:
The second reason for wanting an empirical base is to be able to tell coherent and credible stories about what is happening and why. Governments must continuously communicate with the business community, the financial community and the general public about what is happening in all domains under their responsibility and be able to explain why. On this basis they must explain, defend and sell their policies, even if this really means to have no active policy at all. The communications element of public policymaking never ends and SME policies are as subject to this rule as any other policy domain.
Now in this context, what do we want in terms of data and statistics relating to SMEs?
Obviously the answer is “just about everything” but it is perhaps more helpful to list a few basic categories and characteristics that stand out.
First, we know that there is no special economics of SMEs, they are just the bottom end of a size classification of the more general class of units organized to carry out business activity. So we need a good general data set and statistical base providing a wide range of economic and financial information about the business sector. We also need a sensible breakdown by size so that we can distinguish the characteristics and analyze the behaviour of small ones as distinct from large ones. The existing Structural Business Statistics data in many countries should serve this purpose. But perhaps the single most striking thought that I took out of the Workshop the OECD hosted in Paris last September as preparation for today’s workshop was that this is generally one of the weakest data sets that I have ever tried to use. Once we ask for breakdown by size class the useful information base shrivels. When we last produced the SME Outlook in 2002 we prepared a statistical annex to provide readers with a reference source. All we were able to put together were nine lonely tables, one of which came from OECD Labour Force statistics, with most of the data at least three years old and no time series. I know there are intractable problems of definition, the statistical unit, cost, confidentiality and so on. But users would respond positively to almost any improvement, without being too insistent about what exactly it was.
Second, we are interested in human behaviour as well as what happens to business units. In particular, entrepreneurship is a topic of great interest since it is widely seen as crucial to ensuring the dynamism of modern economies in a rapidly changing and globalizing world. We want to know about the people who start and manage new businesses, why they do it, what their goals are, what obstacles they face and how they respond to both success and failure. As one economist put it to me, “the whole entrepreneurship package”. I would guess we have to survey people, both households and business people, rather than try to exploit data on business units, if we are to get very far here.
Third, since much of the interest in SMEs and entrepreneurship stems from a desire to strengthen the performance of economies over time, especially in stimulating productivity and output growth and realizing potential for job creation, we want information about enterprise demography and the dynamics of enterprise behaviour. What facilitates fast growth of SMEs? In what circumstances do successful SMEs add to their employment levels? At minimum we need to be able to capture the impact of births and deaths on the data. The more information we can have reflecting changes over time, rather than just static pictures at specific points in time, the better.
Fourth, very much along these lines, we would like data sets that allow longitudinal studies, since these allow analysis to be undertaken which assesses the impact of various forces on SMEs and entrepreneurial behaviour over time. Such firm-level studies can facilitate the assessment of how various economic and social forces affect developments and thereby provide guidance as to what types of policies are needed, and perhaps how they should be designed or implemented, while they are still in the formulation stage. Or they can be used to identify the impact of policies and programmes after they have been implemented, thus contributing to evaluation exercises. Since such studies can often be designed to shed light on particular issues motivating the study, data which make them possible are particularly valuable.
Fifth, since the experience of other countries offers real world, concrete evidence of how specific policies or programmes operate in practice, good cross-country comparative studies are also very valuable. The studies themselves must always be controlled for various differences in the sample of countries considered that are not the subject of the study. In this regard, the most important aspect is almost certainly the data themselves. Thus common definitions and classification systems are essential if cross-country comparative work is to be fruitful. This is an area where SME statistics seem to be especially weak at the moment.
Sixth, statistics that show information about specific variables broken down by size class are the building blocks of a useful empirical base but by themselves only shed light on some very basic issues. In practice political interest is often multi-dimensional in the sense that responding to it requires information that is broken down by more than one variable. The highest profile area at the moment is probably gender, where there is great interest in many of the questions we might pose about SMEs and entrepreneurial behaviour. Responding to interest here requires most of the building blocks I spoke of further broken down by a sub-category, gender. Such issues multiply the demands of users for data and statistics, but analysis of such issues as “do women entrepreneurs face different financing obstacles from male entrepreneurs?” requires it. It is very easy of users to make demands on statisticians that are excessive, but so long as the political interest is there, such demands are likely to be made. Statistical priorities must be set by policy priorities, and these should guide data collection efforts.
Now let me turn to some broader considerations. We have to recognize that there are constraints which affect the collection of data and statistics of any kind. Statistics relating to SMEs and entrepreneurship are not exceptions. Notably we can point to costs to statistical agencies, although the need to minimize reporting burdens on enterprises, especially the small ones which are of interest here, and confidentiality must also be taken into account. We should also recognize that there are other areas in economic, social, financial and other types of data and statistics where improvements would be welcome. In the current budgetary climate in most countries, resource increases for the purpose of improving statistics of any kind will be rare. So if more resources are to be devoted to improving SME and entrepreneurship statistics they are likely to have to be drawn from other types of data collection. Scope for this must obviously be evaluated in the context of a broader prioritization and cost assessment of all statistical production activities.
At our Workshop last September, participants were very conscious of the problem of cost, as well as those of confidentiality and reporting burdens. Indeed, the proposal for a single identification number as a device to develop an integrated business statistical register was motivated by these concerns. Overall, we at the OECD feel that the recommendations that constitute the Action Plan summarized in the background paper to this session is practical and cost-effective. At the same time, it offers a good prospect of making much-needed progress in improving SME statistics. I recommend it to you.