Most of the names of CAC 40 companies and, more generally, of businesses that make the headlines on a daily basis are in reality names of "enterprise groups". However, although they account for two thirds of the added value of enterprises (a clear indication of their weight), enterprise groups have no legal existence and are not legal entities as such. This explains why enterprise groups have tended until recently to be ignored in practice in official statistics, with most countries basing their statistics on legal units only (i.e. individual entrepreneurship or companies), despite the fact that an increasing number of laws and regulations now refer to them. The result was that transactions between units of the same group, which in most cases do not comply with market rules and therefore do not have any real economic significance, were taken into account in business statistics. Having concluded that it was no longer relevant to rely solely on legal units to meet French or European regulatory requirements relating to business statistics, INSEE therefore decided to disseminate structural business statistics based on enterprises in the economic sense, the definition of which takes into account the notion of group. Doing so requires the design and implementation of a sophisticated methodology: profiling.

In the wake of general reflections on profiling in the 1990s, the first operational work was carried out in France in 2006 and 2007 through a CNIS working group (Ouvrir dans un nouvel ongletDepoutot, 2008). This is why INSEE, within the framework of the LME, set out to amend the definition of enterprises in the French official statistical system (Box 1).

These concerns are also reflected at a European level under the aegis of Eurostat, which has undertaken a major overhaul of the regulations governing European business statistics in the form of "FRIBS", which has now been adopted: a network-based working group (ESSnet) was launched in 2014 for a period of two years to work on the subject of profiling and the definition of "enterprise".

At the time, Eurostat noted that the official definition of "enterprise" was often imperfectly applied by many countries. The definition drawn from the European regulation (No. 696) of 1993 was subsequently included in the French LME in 2008 to define the various categories of enterprises. In 2015, Eurostat requested countries to comply with this definition when compiling structural statistics and to draw up an action plan to ensure such compliance is achieved by 2020 at the latest.

The need to "profile" certain enterprise groups first emerged at the end of the 1990s during a legal restructuring of the two major French car manufacturers, which had decided at roughly the same time to give legal personality to each of their local units. Thus, new cash flows, with no real economic consistency and corresponding to the physical flows of car components, emerged between the legal units of the enterprise groups. As part of the restructuring, the legal production units leased labour and machinery to certain legal units within the same group and sold their production to other legal units within the group in charge of their commercialisation. None of the new legal units had any real autonomy in terms of production or decision-making, with decisions clearly continuing to be taken at group level. As such, they did not meet the traditional criteria used to define an "enterprise" and were not full-fledged "economic actors". The restructuring referred to here resulted in the turnover, subcontracting, etc. of the French car industry being "fictitiously" multiplied by nearly three when in reality nothing had changed in terms of production capacity as a result of the restructuring, so that it was not possible to leave the new data as they were if the aim was to arrive at an economically significant situation.

This kind of decision, which involves creating several legal units within a new entity, can occur at any time within large groups. Therefore, the assimilation of the legal unit to the enterprise is no longer relevant in this type of case: decision-making autonomy and management of all factors of production take place at group level and no longer at the level of the legal units. Moreover, continuing to reason in terms of legal units not only changes the levels of the aggregates but also their sectoral distribution.

The proposed approach can be illustrated by the following fictitious example (Figure 1) involving an enterprise consisting of a single legal unit. In year N, the enterprise generates a turnover of €300K and creates an added value of €200K. It has several activities in its production process; let A be its NACE code which is assumed to be an industrial activity.

Let us now imagine that for tax reasons the managers of the enterprise decide to create a group structure in N+1 in which there are as many legal units as there are different activities in the initial production process. While the operation remains unchanged from an economic point of view, four legal units are thus created within group G:

The group’s NACE code remains industrial, with such activities creating the most added value. With the simplifying assumption that there is no change in the economic activity of the company between the two years, we arrive at the result shown in Figure 2, although nothing has actually changed in economic terms, with the fact of reasoning in terms of legal units (which have themselves changed) disrupting the statistical landscape.

Figure 1. The simplified description of the flows of a Legal Unit (LeU)

Figure 2. The simplified description of the flows inside a group

Some of the consequences include:

On the contrary, the fact of considering "the group as a whole" as "an enterprise" results in reverting to the previous situation which, by assumption, has not changed. Implicitly, the example given here highlights the changes brought about by group profiling, specifically:

Based on this observation, it was therefore decided to take into account the notion of group in compiling business statistics using the profiling method. The method essentially covers two entirely distinct activities: the determination of the perimeter of such "enterprises" in terms of legal units and the calculation of statistics on the units thus built.

For a given year, the perimeter of these economic enterprises are defined on the basis of the perimeter of the enterprise groups contained in the LIFI (Liaisons Financières, or Financial Links) register (Ouvrir dans un nouvel ongletMariotte, 2017). The profiling of a group begins with the delimitation of one or more enterprises in the economic sense within the group (several in cases where the group is composed of several operational units producing goods and services with a degree of autonomy).

Figure 3 shows the prevailing situation based on a real example involving 10 legal units (LeU1 to LeU10) drawn from the SIRENE register: we find the enterprise groups (G1 and G2) included in the LIFI register based on the financial links between legal units and the enterprises (E1 to E6) created by profilers in the Enterprise Creation Database (Base de création d’entreprise, or BCE).

The national dimension of statistics introduces an additional complexity: compiling statistics requires that a set of legal units in France be considered as an enterprise in France. Therefore, in the case of an enterprise in the economic sense that extends beyond national borders, it is necessary to limit the scope to the latter’s French trace. "French trace" is understood to mean the simple restriction of the activity to the national territory, regardless of the nationality of the owner or owners (e.g. shareholders). Therefore, in France an enterprise in the economic sense will ultimately be:

It should be noted that by restricting the scope to the national territory, part of the "autonomy" that is assumed to characterise an enterprise as an economic actor can be lost. This is, in fact, the challenge of the current globalised economy: its operation cannot be accurately described solely on the basis of "national actors". The (national) sets or wholes that are ultimately considered and used to produce national statistics (including GDP) do not present the desired autonomy that typically characterises an "enterprise". However, only a supranational (European or even global) level would allow for truly economically independent actors only to be taken into account. Such is the aim of European profiling. The latter is led by Eurostat, with France acting as a major contributor.

Figure 3. The links between the different types of units and associated repositories

However, it is currently at the experimental stage and to date no statistics have been produced from the "Global Enterprises" (name of the statistical units defined by European profiling).

In many National Statistical Institutes (NSIs), profiling is limited to this first step of defining and recording perimeters in national registers.

In other NSIs (INSEE, but also the Dutch, Canadian and Australian institutes), profiling is considered meaningful only if it is used to compile structural business statistics. Thus, having identified the perimeters of the enterprise, it is necessary to:

Under the ESANE scheme, the method used to calculate the tax and survey characteristics is a bottom-up method (Figure 4). Thus, the observation unit is the legal unit and the values of the characteristics at the enterprise level are obtained by consolidating the data collected at the level of the legal units. However, two scenarios must be distinguished:

Different methods can be used, ranging from "manual" profiling, resulting in a group monograph produced by a profiler, to fully automatic profiling, which requires no a priori human intervention. Between these two methods, there are alternative methods – i.e. more or less sophisticated semi-automatic methods.

In France, the results of manual profiling have been taken into account in the calculation of structural statistics (ESANE) since the 2013 financial year published in June 2015. The automatic profiling results first became available for the 2017 financial year published in June 2019.

Figure 4. Description of the Bottom-Up Method

Semi-automatic methods are merely exploratory at present, meaning that manual profiling remains the most advanced method. Today, this concerns around fifty of the largest groups (in terms of added value or employment) present in France. The aim is to reach 70 enterprise groups within the next 3 years. Carried out within a dedicated INSEE unit, manual profiling has led to the development of new working methods and even to a new profession: profiler.

The work of profilers consists mainly in making contact with enterprise groups and meeting with them in order to determine:

Once the data have been collected, profilers are required to format, verify and reconcile the information with the data of the legal units they also hold. The validation process is made complex by the number of legal units involved and the difficulty (in some cases) of obtaining reliable data from contact persons from the very first delivery.

These activities – a very new area for INSEE – require specific expertise: the team is therefore made up of a very wide range of profiles, including INSEE "attachés" and administrators, but also contract workers with private sector experience (former accountants, or consolidators, etc.). Experience in accounting and the private sector makes it easier to communicate with enterprise groups by speaking "their language", but also allows for more robust consolidation algorithms to be developed.

The first activity of the profiler is to identify those subgroups within the enterprise groups that meet the definition of enterprise as set out in the LME. The most difficult concept to understand and which only the group is able to provide with certainty is that of autonomy. However, the autonomy they have is not absolute: a subgroup of companies will only constitute an enterprise if it enjoys a "degree of" autonomy.

INSEE has adopted a set of criteria to objectify the matter. This relative restriction on autonomy echoes the parent company’s ownership of company shares: an enterprise will (almost) never hold intellectual property or long-term financing, nor will it have the ability to decide on heavy investments (at group level), all of which are defining features of the power of a holding company. The autonomy referred to here means the possibility of taking all other decisions, most of which relate to day-to-day management, or in the more or less short term.

Thus, for a subgroup of companies to be able to form an enterprise, it must first have the means to make its own decisions: does it have the necessary staff for its production, trading or service activity? Does it have its own general manager, real estate resources, etc.? It must also form an economically coherent whole and, therefore, a whole capable of operating in an economic environment, and its activity must be primarily market-oriented and not group-oriented. Finally, it is essential that the group be able to provide tax and statistical data on these wholes. Although enterprise groups often hold data in their information systems, they are not always available for the French scope of the subgroup or set out in accordance with the French national accounting code. Therefore, enterprise groups are often required to set up an "ad hoc" information system to meet INSEE’s needs.

Once the perimeters of the enterprises have been defined, the profiler collects the necessary information from the group in order to create a tax data and a consolidated ESA questionnaire for each of the "enterprises". The profiler must also obtain a set of intra-group flows from the group to be subtracted from the combined legal units of the group for non-additive variables (Box 2).

For tax data, a flow list has been defined. It concerns:

For ESA data, the aim is to recover the turnover flow broken down by activities as well as the response to the ESA sectoral questions. This type of information is difficult to retrieve because the profilers’ contact person within the group generally holds a financial position (rather than performing a production role) and therefore does not have access to such information.

In some cases, profilers also play an intermediary role in facilitating the collection of thematic surveys where a response at the enterprise level has been agreed (e.g. ICT and CIS surveys).

Unlike manual profiling, automatic profiling is not the exclusive responsibility of profilers. Although the unit responsible for profiling manages the definition of perimeters, the analysis of the data resulting from automatic consolidation is carried out by the department in charge of structural business statistics.

The starting point for forming enterprises in the economic sense is the hard core of the group as defined by the LIFI register. It includes all the legal units in which the head of the group holds a stake of more than 50%.

Since it is impossible to assess the autonomy criterion based on an algorithm, it was initially decided, for automatic profiling, to consider the whole group as a single enterprise: this is a strong hypothesis, but a necessary one in practice (Ouvrir dans un nouvel ongletChanteloup, 2017). It was also decided to propose a second hypothesis: to define the perimeters of enterprises, we may limit ourselves to the legal units of the enterprise groups covered by INSEE structural business statistics (ESANE system). Otherwise, the risk would be to introduce inconsistencies, with legal units within coverage belonging to out-of-coverage enterprises and out-of-coverage legal units belonging to within-coverage enterprises – which would have had unfortunate consequences for the statistical apparatus, including double counts that can be difficult to avoid and, conversely, unintentionally excluded units.

The choices made run counter to the pure definition of enterprise: a group may thus be split into several enterprises based on a criterion not directly linked to the European definition of enterprise and, in particular, to decision-making autonomy. However, it has the merit of simplifying the matter and of undoubtedly facilitating the adherence of the actors concerned to automatic profiling. In addition, the practice is currently being implemented for manual profiling. Where possible, profilers group the financial activities of groups within a full-fledged enterprise.

Ultimately, automatically profiled groups are broken down into:

As with manual profiling, the second phase of automatic profiling involves calculating the characteristics of these enterprises (or rather, in this case, estimating them). Different algorithms adapted according to the variables have been developed:

An illustrative algorithm is shown in Box 2.

The decision to restrict the initiative to the 70 largest enterprise groups in manual profiling was made on the basis of the resources available within the unit dedicated to profiling. This entirely arbitrary threshold does not reflect a major split in the hierarchy of enterprise groups in France. In other words the 71st group on French soil has an economic weight very similar to the 70th group. However, while automatic profiling, which is based on relatively crude assumptions, generates satisfactory results on enterprise groups with fairly simple structures, the method is less effective in dealing with more complex structures, the main criticism being to systematically consider the whole group as a single enterprise. This is why work is currently underway to find ways of improving the treatment of intermediate-sized groups, particularly those listed on the stock exchange, at a lower cost than manual profiling.

Such attempts include:

See (Deroyon, 2015), (Ouvrir dans un nouvel ongletHaag, 2016), (Bacheré, 2017) and (Ouvrir dans un nouvel ongletBalcone, Schweitzer, 2018).

Enterprises in manually profiled groups represent around 12% of FTE employees and 16% of the added value of French enterprises. Enterprises from automatically profiled groups account for around 58% of FTE employees and 56% of the added value of French enterprises. The impact of profiling on the results relating to the accounting and financial characteristics of enterprises is therefore significant. The main implications related to the consideration of enterprise in the economic sense are:

The latter two results reflect the fact that for a given enterprise, the results of all the legal units making up the enterprise will contribute to the sector of the enterprise, whereas in the approach based on legal units they contributed to their own sector. In the example of Figure 2, in the corporate vision, the results of the commercial LeU11 contribute to the sector of the enterprise which is industrial, whereas in the legal unit approach its results are recorded in its commercial sector.

The history of profiling at INSEE goes back more than 25 years. The earliest contribution to the field dates back to the early 1990s, with the work of Emmanuel Raulin (Raulin, 1995), though it is also worth noting that at around the same time the 1993 European regulation provided a definition of the notion of enterprise that remains in force today. An in-depth analysis of statistical units and profiling followed, in particular through the study of the Dutch, Canadian and Australian cases, which were already operational. However, at the time, the French approach remained largely theoretical. The first concrete implementations date back to the early 2000s, with the profiling of car construction groups whose results were taken into account in the Annual Business Surveys (in French, Enquêtes Annuelles d’Entreprises). Subsequently, the results of other manually profiled groups were taken into account in the results of ESANE.

2019 marks an important milestone, with the inclusion of all enterprise groups in structural business statistics. Moreover, some areas for improvement are being considered: in addition to current thinking on semi-automatic profiling, further work is underway in order:

Moreover, while France remains a key driving force in this area, alongside the Netherlands in Europe and Canada or Australia at the global level, national profiling is beginning to develop in other countries following recommendations made by Eurostat.

On the other hand, European profiling (Ouvrir dans un nouvel ongletXirouchakis, Hecquet, 2018) remains more limited in scope and has yet to move beyond the experimental stage. More than 300 enterprise groups have been profiled, although they are not the largest one. In addition, no statistical results taking into account the enterprises obtained by this work have been published. However, with increasing globalisation and the increasing interdependence of economies, this approach no doubt has a bright future.