Defining local markets in merger cases can be tricky, whether in supermarkets, construction materials or hospitals. Some innovative approaches have been adopted in recent merger cases in France, based around customer data and footprints. Are there any holes in these methods?
From a competition law perspective, all relevant markets have a geographical dimension: this can be local (i.e. subnational), national, regional (i.e. covering several countries), pan-European or worldwide. This article focuses on market definition in the context of transactions with a local dimension in France.
There may be many instances of local competition, from production plants with high transport costs, to distributors and retailers whose reach is restricted to a town or city. In such cases, markets are limited by the willingness of customers to either transport the good, or go to buy it themselves.
The 2013 French Merger Guidelines deal with geographical market definition and introduce concepts that were not present in the previous set of guidelines. One of these is the distinction between food retailing and other types of business with a local dimension. Another is the focus on customers’ location of origin when defining catchment areas.
From the SSNIP test to catchment areas
When defining the relevant product or geographical market in the context of merger control, the applicable test is the SSNIP (small but significant non-transitory increase in price), or ‘hypothetical monopolist’, test. This test starts from the most narrowly defined market and asks whether a hypothetical monopolist of the same product and geography would find it profitable to increase prices by 5–10%. Products and geographies are added to the candidate relevant market until raising prices would indeed be profitable and the test is passed.
The main focus of the SSNIP test is customer behaviour—i.e. how demand reacts to relative price changes. In the context of local markets, this concept can be difficult to implement. In fact, determining the right starting point for the geographical market can be tricky, since it could mean any customer location currently or potentially supplied by both parties to the transaction.
In practice, and in order to circumvent this issue, the focus is on the location of the stores relative to one another, as opposed to the location of customers. A key element of this change in approach is the concept of the catchment area; namely, the area around the store where customers come from (or where ‘bulky’ goods are transported to). In other words, it is the area surrounding the store within which that store exerts a credible competitive constraint on any other stores.
When the catchment area of one of the parties’ stores overlaps that of one of the other parties, some customers who currently have a choice of location for their purchases may face reduced competition after the transaction. Going from demand to supply is therefore a step towards narrowing down the geographical locations where competition may be hindered by the transaction.
From a merger assessment perspective, it has now become standard to consider catchment areas as the equivalent of relevant geographical markets. However, although catchment areas are a useful starting point, they do not represent a set of customers over which the SSNIP test is met: starting with the catchment area is neither necessary nor sufficient for the SSNIP test to be fulfilled. Customers located within a catchment area may face reduced choice, while it is also possible that, overall, customers within the catchment area could buy the goods at stores located outside the catchment area. The market defined by the catchment area may thus be too narrow or too wide to meet the standards of the test. This is an important consideration when defining catchment areas, and when undertaking the competitive assessment having defined the market(s).
Market definition for mergers in food retailing in France
When assessing mergers in food retailing, the French Competition Authority relies on extensive case law from after 2000, and makes a distinction between hypermarkets and supermarkets.
- If the store under scrutiny is a supermarket, the Authority will look at all supermarkets, hypermarkets and discount stores within a 15-minute drive of the store.
- If the store under scrutiny is a hypermarket, the Authority will look at the same local market (supermarkets, hypermarkets and discount stores within 15 minutes of the store), but will also run an additional analysis that will include only hypermarkets located within a 30-minute drive of the store.
This approach can be amended on the basis of specific local conditions. Examples given are supermarkets within Paris (where catchment areas will typically extend to only 300–500 metres) and in overseas departments (which are typically islands, such as Martinique, with specific road patterns).
Until recently, this approach to transactions with a local dimension based on drive-time around the stores was also used for mergers that did not involve food retailing. For these other types of merger, the French Competition Authority’s methodology has evolved from drive-time zones to ‘footprints’.
From drive-time zones to footprints
As explained above, for other mergers (e.g. non-food retailing, or distribution of construction materials), the ‘traditional’ approach consisted of relying on drive-time areas around stores (‘isochrones’) in order to draw ‘circles’ around them. These areas were defined using the same radius (or drive-time) for all stores and therefore did not take into account local geographical conditions (besides elements that might affect drive-time).
However, over the last few years, the Authority has adopted an approach based on the actual location of origin of customers buying from each of the stores. It considers that the local catchment area can be defined as the area around the store where customers representing 80% of sales are located (or where 80% of customers are located). On a map, this area represents the ‘real footprint’ (‘empreinte réelle’) of the store, and it can be very different for different stores.
Although the Guidelines endorse this new methodology, they remain relatively vague about how it is to be implemented in practice. In particular, they do not specify whether the area thus defined needs to correspond to the area closest to the store, which would appear to be a basic logical requirement. Nonetheless, the case law seems to interpret this concept as the area closest to the store.
One of the first Decisions to apply such a methodology, in combination with the more traditional approach based on a standard 30–45-minute drive-time around the store, was Titouan. In this merger between two distributors of home appliances and furniture, customers were asked for their postcode when buying goods in order to determine where the customer was coming from. The Decision did not explicitly explain how the methodology used to draw the footprints is implemented in practice. Nevertheless, it established an (arbitrary) 80% threshold of sales as a value of reference, arguing that the remaining 20% of sales represented marginal purchases that are not relevant for market definition (without any further justification for this choice).
The Saturn Decision a month later related to retail sales of home appliances—i.e. large consumer goods such as refrigerators, stoves or washing machines (‘white goods’), light electronic consumer goods such as TVs, DVD players or hi-fi equipment (‘brown goods’), and computer-related products (‘grey products’). This Decision specified that the 80% threshold could be interpreted in terms of sales value (turnover), as in Titouan, or in terms of the number of customers. It also explained that, where there is a discrepancy in market definition resulting from drive-time analysis and the actual sales area based on customer origin, the latter methodology was to be retained.
The Brossette Decision goes a step further in the refinement of the methodology used by the Authority. In this case, which concerned bathroom and heating products sold to business customers such as plumbers, theoretical distances of 50–75km around the outlet were mentioned, in accordance with existing case law. However, the entire analysis relied on a footprint approach. Because the outlets involved were selling specific products to business customers, invoice addresses were available for the majority of sales. With specific addresses, the local footprint could be defined more precisely than in cases where customers are asked their postcode on the way out of the store. In fact, this was the first case where French local geographical district codes (‘codes INSEE’) were used, which are narrower than postcodes. For each of these codes INSEE, the total turnover of customers with an address in that code was calculated. The footprint brought together the codes INSEE closest to the stores under investigation that represented 80% of the stores’ turnover. The resulting footprints were therefore of very different sizes depending on the outlet.
The Brossette case was also different from the Titouan or Saturn cases, in that the analysis focused on the stores owned by the acquirer (Point.P), as opposed to the target of the merger (Brossette). This was due to better data availability at Point.P, and a roughly equivalent number of stores owned by both parties prior to the merger (i.e. choosing Point.P instead of Brossette did not significantly increase the number of overlap areas).
From footprints with holes, to circles?
Coverpro is the most recent case involving detailed local analysis, and involved stores with a wide range of sales of construction materials as well as stores specialising in products used to make roofs. Although it deals with only five stores, it adds an interesting twist to how ‘sales footprints’ need to be determined. In fact, this technique, when it relies on small geographical areas in order to build up the footprint of 80% of sales made by the store, can lead to ‘patchy’ footprints—i.e. footprints that are not uniform but which have ‘holes’ or ‘quasi-holes’ (the Authority calls them ‘discontinuities’ (‘discontinuités’)).
Figure 1 illustrates such a situation around store A. Customers representing 80% of sales are located within the shaded area. The ‘hole’ is an area where the store currently has no customers. It does not mean that local customers would not find store A a credible supplier. Stores owned by the parties or by competitors may be located in those holes and it would appear unreasonable not to consider them credible alternative stores. It is the case of store B in Figure 1.
Figure 1 Footprint with a hole (or ‘discontinuity’) with a store in the hole
Note: The footprint is highly stylised—in practice, it would be a roughly circular area with irregular edges (even in cases where there are no holes), due to the shape of local district boundaries.
Before Coverpro, the Authority did not specify how to deal with such cases, either because it did not need to (i.e. there were no patchy footprints) or because it did not think it was necessary to specify explicitly how to treat them.
In Coverpro, the Authority explicitly specified that, in such cases, the holes in the footprint needed to be filled. The reasoning is nonetheless that such holes (discontinuities) are not justified from the perspective of competition analysis—which is, indeed, correct in most circumstances in terms of competitive interactions and substitution between stores.
Nevertheless, this approach raises questions about how far ‘filling the blanks’ should go within the footprints. There are some clear-cut cases where the footprint exhibits a true hole—i.e. an area entirely surrounded by the footprint, as in Figure 1. Those cases are uncontroversial.
Other cases may be less clear-cut—like the two illustrated in Figure 2, which exhibit ‘quasi-holes’. On both of these stylised footprints, the area of actual sales is shaded. It is always a sub-part of a circle (shown by the dotted line) as a direct consequence of the way the footprint is constructed. In both cases, store B is within the area of actual sales and is therefore automatically considered as part of the market. However, under a strict definition of the footprint, stores C and D would not be part of the relevant market (at least at the initial screening stage), even though one or both of them may be closer to store A than B is. In such cases, how do we decide whether C and/or D are also part of the market?
Figure 2 Footprints with quasi-holes
If ‘filling the blanks’ is done systematically—i.e. if the area includes all stores that are at a distance shorter than the maximum distance defined by the footprint—it becomes equivalent to drawing a circle around the store, with a radius based on the minimum distance from the store required to cover customers representing 80% of sales, as illustrated in Figure 2. Any other methodology (e.g. manually filling the blanks and deciding whether a quasi-hole qualifies to be filled) could be perceived as arbitrary (and therefore unpredictable), as well as time-consuming.
Carrying out the Brossette method with manual amendments to the footprints in order to process the large number of stores owned by Point.P (more than 200) would have taken a significant amount of time. This analysis cannot even be narrowed down to overlap areas (196 areas in the case of Brossette), since filling in the blanks may lead to the creation of new overlaps.
In Coverpro, there were only five stores, so manually amending the footprints was feasible and was done for several of these. In the case of the store located in Le Havre, France, the footprint looked similar to the diagram on the left of Figure 2. It was explicitly decided to leave out stores outside the actual area of sales but within the local circle, due to local geographical considerations.
Generally speaking, in order to ensure consistency and feasibility (in addition to legal certainty) in cases involving a large number of stores, the approach taken to defining the footprint needs to be systematic.
- The traditional approach of using a single drive-time metric for all stores remains an option, although such an approach does not take into account differences in geographical reach among stores.
- One option would be to define footprints strictly on the basis of location of customers, as in Brossette. Even though there would be a risk of missing out on overlaps between the parties due to holes in footprints, it would be possible to adjust the list of competing stores when undertaking the competitive assessment and calculating market shares.
- An alternative would be to define footprints on the basis of customer location with true holes being filled up—but not quasi-holes. This would reduce the risk of missing areas of true overlap, while preserving the local pattern of sales made by the store.
- Finally, footprints can be defined as the entire area around a store that is within reach of the minimum distance required to cover 80% of sales—i.e. within a circle drawn around the store with a radius based on the 80% threshold. Such an approach includes all own and competing stores, with the risk of underestimating the parties’ local market share by overestimating the number of local competitors. However, it also takes into account the local reach of the store, which can differ from the reach of another store.
Existing French case law in defining local markets has evolved from an isochrone-based approach to a technique relying on customer location. Issues resulting from using isochrone-based approaches have also been raised in other jurisdictions, but methodologies relying on customer location are not exempt from theoretical and technical issues either.
In France, as a result of the specific treatment of food retailing mergers, two sets of standards have emerged depending on the sector. In food retailing mergers, the approach remains focused on the drive-time around supermarkets or hypermarkets, as in (for example) the UK. In mergers that do not involve food retailing, recent Decisions have used local footprints based on the actual location of customers representing 80% of sales. This approach allows local geographical constraints to be accounted for, and acknowledges differences in catchment areas between stores. There are, nonetheless, technical considerations to take into account when implementing this methodology. In order to promote sound economic reasoning as well as predictability, more clarity would be welcome on how to address them.
Contact: Pascale Déchamps
 Autorité de la concurrence (2013), ‘Lignes directrices de l’Autorité de la concurrence relatives au contrôle des concentrations’, July.
 One of the first key cases was the Carrefour/Promodès transaction, cleared on 5 July 2000 (see Arrêté du 5 juillet 2000 relatif à l’acquisition par la société Carrefour de la société Promodès, BOCCRF 2000-11). In France, hypermarkets are retail stores that have more than a third of their turnover from food-related items and are larger than 2,500m2. Supermarkets are retail stores that have more than two-thirds of their turnover from food-related items and are between 400m2 and 2,500m2. (Source: INSEE, Institut national de la statistique et des études économiques.)
 Strictly speaking, circles are based on distances rather than drive-time. In practice, a drive-time zone is typically estimated with a circle using the distance that can be covered within the drive-time as the radius.
 Autorité de la concurrence (2013), ‘Lignes directrices de l’Autorité de la concurrence relatives au contrôle des concentrations’, July, para. 366. Although this is a development in the French Competition Authority’s approach to local markets, it is not new in economics. In 1973, Elzinga and Hogarty proposed a methodology based on the location of customers in their seminal paper, Elzinga, K.G. and Hogarty, T.F. (1973), ‘The Problem of Geographic Market Definition in Antimerger Suits’, Antitrust Bulletin, 18, pp. 45–81.
 Autorité de la concurrence (2013), ‘Lignes directrices de l’Autorité de la concurrence relatives au contrôle des concentrations’, July, para. 366.
 The Guidelines also do not specify how deliveries and sales within the shop should be treated when a store makes a significant number of deliveries.
 Décision n° 11-DCC-78 du 18 mai 2011 relative à l’acquisition du groupe Titouan par le groupe Conforama, paras 24–5.
 Postcodes in France represent fairly large geographical areas, such as an entire city or even several small towns. Only three cities in France (Paris, Lyon and Marseille) have separate postcodes for addresses within the city.
 Later Decisions add that these remaining sales could be very far away, which justifies the choice of the threshold (see, for example, the Brossette case referred to further below, para. 50). The Elzinga and Hogarty 1973 paper referred to a 75% threshold, which was heavily criticised as being too low. In 1978, the authors proposed to use a 90% threshold for a ‘strong market’ (as opposed to a ‘weak market’ when using 75%). See Elzinga, K.G. and Hogarty, T.F. (1978), ‘The problem of geographic market delineation revisited: the case of coal’, Antitrust Bulletin, 23:1.
 Décision n° 11-DCC-87 du 10 juin 2011 relative à la prise de contrôle exclusif de la société Media Concorde SNC par la société High Tech Multicanal Group, paras 41–2. Media Concorde SNC owned the Saturn network of stores. A Decision taken in April 2012 follows the same sort of reasoning (Décision n° 12-DCC-46 du 3 avril 2012 relative à la prise de contrôle des fonds de commerce de la société SCT Toutelectric par le groupe Rexel).
 The Decision does not specify why using the number of customers can also be appropriate. Although it does not say so, relying on the number of customers may be the result of a lack of sales data.
 Décision n°12-DCC-41 du 23 mars 2012 relative à la prise de contrôle exclusif de la société Brossette par la société Point P. Oxera Senior Adviser, Pascale Déchamps, advised the parties in the Brossette case.
 Despite the existence of a large proportion of deliveries, for which the addresses were also known, the Authority considered invoice addresses to be the relevant basis for customer location when customers choose an outlet. In addition, the Authority did not question the fact that invoice addresses might not always represent the actual customer’s point of origin (e.g. a plumber may have an invoice address in one place, live somewhere else, and work in a third location).
 Décision n° 14-DCC-10 du 28 janvier 2014 relative à la prise de contrôle exclusif par Point P de cinq points de vente détenus par Wolseley France Bois et Matériaux. Oxera advised Point.P in this case.
 Décision n° 14-DCC-10 du 28 janvier 2014 relative à la prise de contrôle exclusif par Point P de cinq points de vente détenus par Wolseley France Bois et Matériaux, para. 20.
 There may be several reasons for the absence of customers in a given location: because there are no businesses in the area (e.g. it is too expensive for small businesses, or it is mostly countryside), or because there are specific local geographical circumstances that make this area less accessible from the store than the areas around it.
 ‘La partie notifiante a donc identifié, autour de chaque point de vente cible, une zone de chalandise comprenant les communes les plus proches du point de vente permettant de capturer 80 % des ventes du point de vente. Cette zone de chalandise englobe la totalité des communes qui sont situées dans l’empreinte à 80 %, quand bien même aucun client ne serait situé dans une ou plusieurs de ces communes, l’existence de discontinuités n’étant pas justifiées du point de vue de l’analyse concurrentielle.’ Décision n° 14-DCC-10 du 28 janvier 2014 relative à la prise de contrôle exclusif par Point P de cinq points de vente détenus par Wolseley France Bois et Matériaux, para. 20.
 The footprint is constructed by gradually adding local areas based on their distance to the store until 80% of sales are reached. The footprint is therefore necessarily within the circle defined with a radius equal to the distance of the last local district added, all other districts being at a shorter distance.
 Even quasi-holes that may be justified by local geographical circumstances (e.g. mountains, or a river with no bridge) would still raise issues about how difficult it really is to access those areas relative to other areas from which other customers come.
 See, for example, Geer, T. (2010), ‘Taking stock of the OFT’s approach in supermarket mergers’, Agenda, August; and Oxera (2009), ‘Supermarket mergers: Holding on to market share under competition scrutiny’, Agenda, April.