Chapter 1 1

A Road Map to Assessing Local Market Mergers David Wirth

Ashurst LLP Tom Punton

Introduction It is also important to consider whether the aggregation of local overlaps may affect national aspects of competition. Many mergers that are reviewed by competition authori- For example, where the parties are close competitors across a ties involve an assessment of competition in markets that are number of different local markets, this is more likely to impact narrower in scope than national, such as mergers between on the dimensions of competition that they set nationally. In retailers, wholesalers and local service providers. Such sub-na- Sainsbury’s/Asda (2019), the CMA explained that elements of the tional markets are often termed “local markets”, even in cases competitive offering that are set centrally and apply uniformly where competition takes place at a broader regional level. across all stores “affects their effectiveness as a competitor in every local Whilst the substantive threshold for assessing mergers area where they operate”, and therefore it considered the impact of involving local markets is the same as for all other mergers, local the transaction both at a national level as well as in each of the market cases raise their own unique issues, most notably since areas of local overlap. they often result in a large number of individual markets having Whilst the extent of the local overlap is a relevant factor to to be scrutinised as part of the same transaction. For example, assessing the impact of the transaction at a national level, it is not in Greene King/Spirit (2015), the transaction involved the acqui- the only factor. In its Phase 1 decision in JD Sports/Footasylum sition of over 1,200 pubs, Ladbrokes/Coral (2016) involved over (2019), the CMA explained that is also relevant to consider other 1,800 betting shops, and in Sainsbury’s/Asda (2019) the merging evidence that more accurately reflects the overall aggregate parties operated over 1,000 supermarkets and 600 petrol filling constraints on the parties, such as internal documents, third- stations. In such cases, it would be impractical to conduct a party views, evidence on the Parties’ service propositions and detailed competition assessment into each local area, which can third-party reports and commentary. In Sainsbury’s/Asda (2019), raise practical and timing issues both for the parties and the as well as undertaking a GUPPI assessment at the local level, competition authorities reviewing the transaction. the CMA also carried out a nationally weighted GUPPI assess- The aim of this article is to provide an overview of the issues ment in order to assess the closeness of competition between the that arise in the assessment of local market mergers. We also parties at a national level.2 consider a number of the techniques that have been used by A merger between retailers may also give rise to other effects the UK’s Competition and Markets Authority (CMA) and other competition authorities in assessing whether local market that are not merely a reflection of the aggregated loss of compe- mergers are likely to give rise to competition concerns. tition at a local level. In particular, harm may arise at a national level where retail mergers result in the loss of potential compe- tition that goes beyond the parties’ respective current estates. Local Versus National Competition Similarly, the loss of competition in relation to innovation in a While this article focuses on assessing competition in local given retail sector or certain nationally set pricing parameters markets, it is important also to consider whether a transaction may not be limited to an aggregation of the loss of local compet- involving local markets has an impact on competition at the itive constraints. national level. For example, in Ladbrokes/Coral (2016) the CMA had histor- In this regard, even if consumers’ choices are limited to local ical information on shop openings and closures in the industry. retailers, some parameters of competition, such as price, quality, The CMA noted that the parties had frequently opened stores range or service (collectively referred to as PQRS), may be set in competition with each other over the five years preceding nationally. For example, for firms with many outlets, it may be the merger. Similarly, in Pure Gym/The Gym (2014), the CMA more cost effective to set national prices, rather than giving each at Phase 1 found that the parties were the two leading budget store the discretion to flex prices locally. gym players and were each significantly larger than the third and It is therefore important first to consider what elements of fourth largest players. Their expansion plans were also materi- competition are set locally (rather than nationally), and whether ally more significant than those of other budget gym players, the parties would have an incentive to adjust their retail offer and meant that they would open a large number of new gyms in in response to local competitive conditions post-merger. The the near future. CMA’s commentary on retail mergers emphasises that the CMA’s “strong starting assumption at phase 1 has been that there will The Use of Catchment Areas be material local competition on one or more aspects of the PQRS retail offer”.1 It is important, however, that there is a nexus between The aim of geographic market definition is to focus on the area the theory of harm being considered at the local level, and the over which competition takes place. When defining the relevant incentives and ability for local retailers to behave in the way geographical market, competition authorities typically refer to advanced in the theory of harm. the SSNIP test (Small but Significant Non-transitory Increase

Merger Control 2020 2 A Road Map to Assessing Local Market Mergers

in Price). This test starts from the most narrowly defined catchment area using the Parties’ data, and two narrower geographic market and considers how customers and suppliers catchment areas (20 miles and 50 miles) based on the views would behave towards a relative increase in prices by 5–10%. of third parties; and In local market cases, however, competition authorities tend ■ are catchment areas based on the locations of customers to focus on the catchment areas of customers (described in the consistent with the other evidence available (e.g. the CMA’s Retail Mergers commentary as a pragmatic approach parties’ internal documents and monitoring of compet- to identify the geographic market), which is unrelated to the itors, the parties’ decision-making and how they react SSNIP test. More specifically, the CMA’s guidelines refer to to competition and new entry, how customers view the catchment areas accounting for the closest 80% of customers market and so on). being used as a proxy for geographic market definition, which Catchment areas can also be derived from customer surveys involves ranking customers by distance from each of the sites which can ask about the customer’s journey and where they in question and using the 80th percentile as the catchment area.3 might shop if the store was not available (e.g. Celesio/Sainsbury’s The following chart provides an example of the distance (2016)) or from the parties’ internal documents, which monitor profile of customers (e.g. around a particular site) that is often the competitive offering of local competitors. It may also be found in local market cases. Of note, the distances of customers possible to use other quantitative techniques to estimate the size from a site will usually increase significantly beyond a certain of the catchment area, such as by looking at the impact on sales threshold, often reflecting some inconsistencies in the data of new store openings within a local area or the impact on prices (e.g. people on holiday or visiting/working in a different area of varying levels of concentration at different distances. to where they live). This can have a significant bearing on the overall catchment area used for the competitive assessment Methodological Issues With Using th unless a lower threshold, such as the 80 percentile, is applied. Catchment Areas In this regard, it is often helpful to plot the distance profile of customers in order to identify whether there are any obvious There are also a number of factors to take into account when kinks in the data. considering catchment areas. First, the size of the catchment area depends on the specific measure used (e.g. straight-line distances; road distances; drive- time distances; public transport journey times, etc.). These different approaches have all been used in previous cases, and can result in some material differences in the catchment areas defined. Different software and modelling assumptions may also generate different results as to the distances which can be travelled within the relevant journey times (with journey times and distances varying depending on the time of day of the journey and the road layout). There appears to be little discus- sion in merger decisions as to why one particular approach is The use of catchment areas typically requires details of favoured over another. customers’ addresses to be available. The CMA has used a Second, it is important to consider where to centre the catch- number of different types of evidence to identify the location ment area. The approach adopted by the competition authori- of a store’s customers. Sources include: (a) data from loyalty ties has typically been to centre the isochrones on (i) each of the cards or mailing lists (e.g. Greene King/Spirit (2015)); (b) home/ target’s sites, and (ii) each of the acquirer’s sites, as the compet- site delivery records (e.g. Edmundson Electrical/Western Electrical itive constraints facing the parties could vary depending on (2014)); location of a healthcare funder as a proxy for patient where the isochrone is centred. Clearly, in mergers involving addresses (e.g. Cygnet/CAS (2017)); and customer contact details a large number of local overlaps, this can result in a significant (e.g. Nationwide Building Society/Derbyshire Building Society (2008)). amount of work. The use of catchment areas raises three significant issues: Whilst centring the isochrones on the parties’ stores focuses ■ a catchment area approach is based on static analysis (i.e. it on the areas where they typically win customers, it does not is based on the current purchasing patterns of customers) necessarily reflect the choices facing customers, which will and fails to consider how those customers would behave vary depending on where customers are located. Accordingly, if prices increased by 5–10 per cent (i.e. in response to the in some cases, it may be practical also to consider isochrones SSNIP test question). The CMA’s guidelines note that from the perspective of the customer (known as demand re-cen- geographic markets tend to be wider than catchment areas, tring). Demand re-centring has been applied by the CMA in although, in practice this is often not taken into account in assessing mergers between grocery retailers, where it has re-cen- local market assessments; tred the isochrones on census output areas in order to better ■ is the 80% threshold meaningful (i.e. does it make sense reflect the choices for consumers in the main areas of popula- to exclude 20 per cent of customers)? There appears to be tion. For example, if customers happen to be located in a small no economic basis for using an 80% threshold, other than town (Bury St Edmunds in this example), centring the isoch- to exclude some potential outliers. In some instances, rones on each store (on opposite sides of the town) may fail to authorities will also look at other catchment areas, e.g. 75% recognise the competition that exists between the parties (i.e. or 85% but again without a clear justification. This can isochrones centred on each store suggests that they are not in lead to a further departure from the SSNIP test, and add the same market, when in reality they are the two best alterna- uncertainty as to the scope of the geographic market. For tives and similar distances apart for many customers living in example, in CareTech/Cambian (2019), the CMA considered the town). three different catchment areas, one based on the 80%

Merger Control 2020 Ashurst LLP 3

(2016) the CMA found that Sainsbury’s pharmacies drew customers from a wider area than Lloyds, partly because they were located in a supermarket. As a result, the CMA used different catchment areas for supermarket and non-supermarket pharmacies.

Filtering The next step in most local market mergers is to identify those areas that are most likely to give rise to competition concerns. In cases where the number of local overlaps is low (e.g. fewer than 20), it may be possible to consider all overlapping areas in detail avoiding the need for filtering. For example, in David Lloyd/Virgin Active (2017) the CMA identified seven overlap areas and assessed competition in each area individually.7 However, in many cases the number of local overlaps can be in the tens or hundreds and it is often impractical (and expen- sive for the parties) to conduct an in-depth investigation into each local area, particularly within the merger control time- table. Accordingly, it is usually important to identify filters or a decision rule in order to focus on the key areas of concern. In cases that involve hundreds of individual local markets, the filter may be the same as the decision rule due to the impracticality of assessing each local area in detail. The first step in choosing a filter is to identify a relevant measure of concentration and/or competition between the parties. Ideally the filter should be based on a measure that best captures the closeness of competition between the merging parties in each local area so that only areas that fail the filter are likely to give rise to competition concerns (i.e. minimising the risk of Type I errors). However, in reality the design of the filter needs to be pragmatic and often depends on what data is readily available. The CMA’s Retail Mergers Commentary iden- tifies three measures of concentration that have been used as a Third, whether to use an average catchment area across all filter in the past: fascia count; store count; and market shares.8 sites within the relevant product market definition, or to rely on ■ A fascia count filter is likely to be appropriate if branding site-specific catchment areas (which may vary by each site). The is an important element of competition and there is limited CMA answered this question in Cygnet/CAS (2017) by saying variation between stores of the same fascia (i.e. customers that “our usual starting point in mergers in local markets is to calculate an care about the number of fascia in an area rather than the average catchment area and apply this… across all of the merging parties’ number of stores). Fascia count filters have commonly overlaps”.4 The CMA also explained that the reason for using been applied in grocery mergers, e.g. in Martin McColl/ an average catchment area is that “it should capture a consistent rela- Co-operative the CMA applied a four to three fascia filter tionship between customer behaviour” across all sites, and therefore (areas failed the filter if post-merger there would be fewer concluded that “rarely are site specific catchment areas used”.5 than four fascia in an area).9 These statements contrast with the use of site-specific catch- ■ A store count filter is more appropriate where brand is not ment areas by the CMA in a previous Phase 1 case in the same very important or visible to customers. For example, the markets (i.e. Acadia/Priory (2016)). In that case, the CMA said CMA used a filter based on the number of pubs in a local that “where there were sufficient observations to form a hospital specific area in Greene King/Spirit (2015) (areas failed the filter where catchment area, this was used since hospitals are likely to have different catch- the merging parties had more than 35% of the pub count ment areas reflecting local competitive conditions”.6 However, the CMA and the merger resulted in a 5% increment).10 considered that just 10 data points was sufficient (which means ■ In many cases a market share filter may provide more that just eight data points are used at the 80 per cent level). There information about the effect of the merger on competi- is a danger that this approach, based on such a limited number tion than either a fascia or store count filter. However, it of data points, creates a false sense of precision, which becomes is only possible in cases where information on the market extremely sensitive to the locations of just one or two people. size is available at a local level, which is not the case in Fourth, there may be different catchment areas depending on many local market mergers. In Cygnet/CAS (2017) and the willingness of different types of customer to travel (e.g. in CareTech/Cambian (2019), the CMA adopted a filter based urban and rural areas, reflecting the fact that customers may have on the number of beds in a local area based on publicly to travel further in rural areas). For example, in assessing the available information on bed numbers. merger between Greene King and Spirit, the CMA considered A further important step is to determine which competitors that it should adopt a two-minute drive time isochrone (centred should be included in the competitor set that is used to calcu- on each of the parties’ pubs) in city centres, a 10-minute drive time late concentration in each area. The CMA’s Retail Mergers in City Urban, Other Urban and Suburban areas, and a 15-minute Commentary notes that the CMA will consider a range of drive time in Rural and Rural Town and Village areas. In such factors in assessing the effective competitor set, including: cases, whilst it is still relevant to consider average catchment areas, internal documents; evidence on store characteristics; evidence it is important to distinguish between different cohorts of users. from competitors and customers; and surveys. For example, in Fifth, there may be asymmetric catchment areas between the the casino merger Rank/Gala (2013), the CMA concluded that parties and different competitors. For example, in Celesio/Sainsbury’s other types of leisure venues such as cinemas and bingo halls

Merger Control 2020 4 A Road Map to Assessing Local Market Mergers

should not be included in the effective competitor set based on a an average catchment of 60 miles for Long Term Mental Health review of the merging parties’ internal documents. (LTMH) patients. This was slightly narrower than the parties In some cases the CMA has applied a weighting to the concen- estimate of between 70 and 80 miles.14 The CMA used this catch- tration measure if a simple fascia or store count is not reflective ment area for identifying the overlaps between the parties and of competition in an area (e.g. due to the differentiated nature of for applying a stage 1 filter. For those sites that remained after the offering of different competitors). This approach introduces the application of the filter, the CMA then subjected them to additional complexity and therefore is more commonly applied a more detailed local competition assessment (stage 2 analysis). in Phase 2 cases. Weighting adjustments fall into two categories: According to the CMA’s Provisional Findings report, one ■ adjusting the weight applied to specific competitors where area that was identified as raising concerns was the overlap they exert a more limited competitive constraint. For between Cygnet Hospital Coventry and Cambian Raglan House. example, in Ladbrokes/Coral (2016) the CMA found that inde- In responding to the CMA’s Provisional Findings, not only did pendent licensed betting offices (LBOs) exerted less of a the parties comment on the number of competitors within the competitive constraint on the merging parties than national catchment area, but they also focused on the large number of LBOs. As a result, the CMA applied a 0.9 weighting to inde- competitors located just outside the catchment area, all of which pendent LBOs in its store count filter;11 and had a catchment area that overlapped in part with that of Cygnet ■ weighting all stores based on distance. This approach has Coventry, but which had been excluded entirely from the CMA’s been used where distance is particularly important in deter- analysis. mining customer choice. For example, in Celesio/Sainsbury’s To illustrate this point the Parties provided a map showing (2016) the CMA used a weight based on the straight line the catchment areas of the ten sites located between 60 and 80 distance from the focal store to calculate a weighted store miles of Cygnet Coventry (i.e. excluding the other competing count in each area (i.e. a store near the edge of the catch- sites within the catchment area), which is shown below.15 These ment area counts for less than a store near the focal store).12 were competing sites that would have fallen within the catch- In Sainsbury’s/Asda (2019), the CMA sought to calculate the ment area based on the Parties’ calculations, but which were GUPPI in all the local areas of overlap between the parties. excluded on the basis of the CMA’s refined approach. In order to calculate diversion ratios, the CMA carried out a customer survey at 150 of the parties’ stores. The survey results were used to calculate the weights that should be applied to different competing stores within a 15-minute catchment of each of the Parties’ stores (with the weights varying depending on distance, brand, and size of store). The weights were then used to calculate the diversion ratios between the parties in each of the local areas of overlap. One of the most challenging aspects of choosing an appro- priate filter is setting the threshold at which areas will fail the filter. In some cases, the CMA has applied a four to three fascia threshold, but in other sectors it has taken a more conserva- tive approach and used a five to four fascia threshold. In rela- tion to store count and market share filters, the CMA has typi- cally based its filter on either a 35% or 40% threshold and the CMA’s Retail Mergers Commentary refers to 40% as “a starting point when assessing the appropriate share of store threshold”.13 However, in CareTech/Cambian (2019), the CMA adopted a lower 30% filter on the basis that there was some uncertainty as to the services and bed numbers provided by competing providers of chil- dren’s care homes and specialist schools which may result in the Parties’ market share being understated.

Constraints From Outside the Catchment The map above shows that nearly all of the Cygnet Hospital Area Coventry catchment area is overlapped by the catchment areas of competitors located more than 60 miles away, and in many parts In local market cases, the use of catchment areas is designed to of the catchment area there are multiple overlaps. This means create a frame of reference for identifying the areas of overlap that, for the vast majority of patients that are located within a between the parties. However, there is often a tendency to define 60-mile catchment area of Cygnet Coventry, they have other distance or travel time catchment areas (based on the parties’ data), suitable alternatives from outside the catchment area of Cygnet and then to only focus on competing sites located within that area. Coventry (as well as other alternatives within the catchment area). This can give rise to the so called “binary fallacy” on the basis In the CMA’s final report, the CMA flexed the catchment area that it considers competition to be equally effective between all to 70 miles to test the sensitivity of its analysis to competitors competitors within the catchment area, but assumes that compe- located just outside the 60-mile catchment area, which resulted tition from competitors outside the catchment area is zero. in the parties’ combined market share falling well below 40 per The binary fallacy is relevant in all merger assessments, but is cent. This analysis (in combination with a number of additional particularly important in local market assessments. This reflects pieces of evidence submitted by the parties) appears to have the fact that the catchment area is not a defined geographic persuaded the CMA to reverse its provisional finding that the market and there may be a number of competing sites located overlap between Cygnet Hospital Coventry and Cambian Raglan just outside the catchment area which may represent better alter- House in the West Midlands raised competition concerns.16 natives to consumers than the parties’ sites. In Sainsbury’s/Asda (2019), the CMA acknowledged that there For example, in Cygnet/CAS (2017) (which concerned overlaps may be an additional competitive constraint from stores located in the provision of mental health hospitals), the CMA identified outside the 15-minute catchment (which was informed by its

Merger Control 2020 Ashurst LLP 5

customer survey). In doing so, the CMA applied a separate out-of- in response to local competition. The CC concluded that market weight (of 25%) to the GUPPI analysis in order to account Sports Direct varies a number of non-price factors (QRS) for those customers that would switch to stores located further at a local level, but there was no evidence linking this varia- afield as well as customers that would switch to shopping online. tion to differences in the strength of local competition; and ■ bidding market analysis. For example, an assessment of The Competition Assessment tenders that had taken place in local markets was consid- ered by the CMA in Menzies/Airline Services Ltd (2019). In For sites that remain after the filters have been applied, the particular, the CMA considered the frequency with which competition authorities will typically conduct a more detailed both parties submitted bids for tenders (for de-icing and competition investigation into each of these local areas. There ground handling services) at individual airports, and the are two main issues considered in this respect: (i) how important outcome of those tenders, including how frequently the is the rivalry between the merging parties in each of the areas parties were the only two bidders.17 of overlap; and (ii) the extent of competition that remains from The above list is by no means exhaustive. For sites that fail other competitors in that local area. the initial filter, the full economic toolkit of quantitative tech- The techniques that are applied for assessing competition niques is available and should be considered. However, one of tend to be case-specific, but are similar to the techniques used the principal issues that arises in local market cases is the level for assessing unilateral effects in all other types of mergers. The of detail required in order to assess competition in the overlap techniques applied in previous local market cases include: areas that fail the filters. This is often a function of both the ■ refined market share analysis (e.g. with isochrones severity of the concerns raised, and the number of overlaps that flexed to consider the sensitivity of the market shares to fail the initial filter. constraints from outside the catchment area, with addi- Clearly, if the stage 1 filters are set at too cautious a level such tional competitors included within the competitor, or with that they result in a large number of local overlaps having to be different weights being applied to different competitors). considered in detail, then it will clearly impact on the types of For example, as set out above, the CMA flexed the catch- analysis than can be considered in the competitive assessment, ment area analysis in Cygnet/CAS (2017); particularly within the Phase 1 timetable. In such cases, the addi- ■ product and geographic differentiation. For example, in tional time available at Phase 2, and the higher decisional threshold Greene King/Spirit (2015) the CMA considered the extent for reaching an adverse finding, could mean that a Phase 2 inquiry of the competitive constraint exerted by wet-led pubs (i.e. is more likely to lead to a better outcome for the parties. pubs that focussed on serving drinks) on dry-led pubs (i.e. pubs which focus on serving food as well as drinks), and Is Phase 2 the Only Option? the geographic proximity of the parties’ pubs in a given area; As noted above, a significant proportion of mergers referred to ■ impact studies. For example, impact studies were used in Phase 2 by the CMA in the last three years have involved an assess- Cygnet/CAS (2017) to assess the impact on patient volumes ment of local markets. It is also of note that in a number of those and the level of fees paid at a Cygnet hospital (that failed cases, the outcome achieved by the parties at Phase 2 was mate- the stage 1 filters) following the opening of a new mental rially better than would have been the case had remedies been health hospital by CAS, which the CMA considered accepted at Phase 1 (i.e. in terms of the number of divestments showed a lack of competition between the parties’ sites. required). In this regard, the following table compares the number Similarly, in Poundland/99p Stores (2015), the CMA sought to of SLC findings at both Phase 1 and Phase 2 in the local markets estimate the effect of the opening and closing of different cases considered by the CMA at Phase 2 in the last four years. retailers’ stores on the store-level revenue at Poundland and 99p Stores, in order to assess which competitors Merger18 Number of local impose the most significant competitive constraints on each of the parties; SLCs at: ■ customers surveys and diversion ratios. For example, Phase 1 Phase 2 in David Lloyd/Virgin Active (2017) the CMA carried out surveys in three areas (Northwood, Clearview and Ashford St Peter’s NHS Foundation 13 0 Brighton) to calculate the diversion ratio of Virgin Active Trust/Royal Surrey County NHS members to David Lloyd gyms, and vice versa (i.e. asking Foundation Trust (2015) which alternative gym customers would use in the event Poundland/99p Stores (2015) 92 0 that their current gym was closed). The high diversion ratios in Clearview and Brighton led to the CMA reaching Ladbrokes/Coral (2016) 798 642 a Significant Lessening of Competition (SLC) finding in those two areas; Celesio/Sainsbury’s Pharmacy Busi- 78 12 ■ price pressure tests. For example, in Cineworld/City Screen ness (2016) Limited (2013) the CC carried out a survey of Cineworld Arriva Rail North/Northern rail 82 3 and Picturehouse customers to calculate diversion ratios franchise (2016) between the parties’ sites. The CC used these diversion ratios, along with a measure of the parties’ gross margin, Central Manchester University 25 0 to calculate the GUPPI in each local area. As mentioned Hospitals/University Hospital of above, a GUPPI analysis was also carried out in Sainsbury’s/ South Manchester (2017) Asda (2019); ■ price/margin concentration analysis. For example, in Sports Cardtronics/DirectCash Payments 848 0 Direct/JJB Sports (2010), the CC conducted a margin-con- (2017) centration analysis to see whether there was any evidence Cygnet Health Care/Cambian adult 12 1 of Sports Direct flexing the quality, range and service services division (2017) (QRS) variables at the local level to earn higher margins

Merger Control 2020 6 A Road Map to Assessing Local Market Mergers

Tesco/Booker (2017) 435 0 Note Euro Car Parts/Andrew Page 92 9 * The authors advised the merging parties in relation to Cygnet/ (2017) Cambian Adult Services (2017), Cardtronics/DirectCash Payments (2017) and CareTech/Cambian Group (2019). The Ausurus Group/MWR (2018) 5 3 authors also advised Morrisons in relation to Sainsbury’s/Asda Menzies/Airline Services Ltd (2019) 5 0 (2019).

The above table shows that, of the local market cases consid- Endnotes ered by the CMA at Phase 2 since 2015 (with the exception of 1. Paragraph 1.6, Retail mergers commentary, CMA, 10 April Sainsbury’s/Asda), there have been some notable narrowing of 2017. concerns compared to the Phase 1 decisions, with five cases 2. GUPPI is the acronym for the gross upward pricing pres- being cleared unconditionally, and a number of other cases sure index. It provides an indication of the upward pricing resulting in far fewer divestments. In Cardtronics/DirectCash pressure that would be expected to arise from a merger as Payments (2017), the CMA identified 848 local areas of concern at a result of a combination of (i) the diversion ratio between Phase 1. At Phase 2, the CMA concluded that the structure and the parties; (ii) the parties’ gross margins, and (iii) relative dynamics of the market meant that the parties did not have the prices of the parties. incentive to increase prices at the local level, and therefore there 3. Catchment areas based on the location of customers have were no areas of concern. The difference in outcomes between Phase 1 and Phase 2 been used in numerous cases by the UK, the European reflects, in part, the difference in the CMA’s decisional threshold Commission and other European Competition Authorities. between Phase 1 and Phase 2 investigations (i.e. with Phase 2 The US horizontal merger guidelines also refer to defining investigations being based on a balance of probabilities test, local markets based on the location of customers. whilst Phase 1 investigations are based on a lower realistic pros- 4. Paragraph 5.82, Cygnet Health Care and Cambian Adult pect test). However, the difference in outcomes may also reflect Services Phase 2 report, CMA, 16 October 2017. the difficulty facing regulators in assessing transactions that 5. Ibid. involve a large number of overlaps within the Phase 1 timeframe, 6. Paragraph 366, Completed Acquisition by Acadia which, when combined with the relatively low threshold for a Healthcare Company, Inc. of Priory Group No. 1 Limited, reference to Phase 2 (a realistic prospect of an SLC), inevitably CMA, 15 August 2016. leads to many more problems being identified, and therefore 7. CMA, David Lloyd/Virgin Active, Phase 1 Decision, para- more divestments being required to avoid a Phase 2 reference. graph 35. Accordingly, in some cases it may make sense to fast-track 8. CMA, Retail mergers commentary, 10 April 2017, paragraphs the merger assessment to Phase 2, although this will depend on 3.20–3.25. the parties being sufficiently advanced in their own analysis to 9. CMA, Martin McColl/Co-operative, Phase 1 Decision, para- make a fast track reference worthwhile. A fast track reference graph 47. was made in relation to Ladbrokes/Coral (2016), and in Sainsbury’s/ 10. CMA, Green King/Spirit, Phase 1 Decision, paragraph 6. Asda (2019). In other cases, depending on the number of prob- 11. CMA, Ladbrokes/Coral, Phase 2 Final Report, paragraph lematic areas identified at Phase 1, it is worth the parties consid- 7.104. ering whether the additional cost and time delay of going 12. CMA, Celesio/Sainsbury’s (2016), Phase 2 Final Report, para- through a Phase 2 investigation might be justified, as there is a graph 7.224. reasonable probability of achieving a material better outcome. 13. CMA, Retail mergers commentary, 10 April 2017, paragraphs 3.36. Conclusion 14. The exact numbers are redacted from the final report. 15. See the parties’ response to the CMA’s Provisional Local market cases can involve a significant amount of Findings report, dated 14 September 2017. complexity, particularly in transactions where there are 16. This was on the basis of a split decision. Two Panel numerous local overlaps between the parties, with the recent Members were of the view that the Merger would result Sainsbury’s/Asda (2019) case being a clear example of this. Whilst in an SLC in the West Midlands, and two Panel Members the competition test for assessing mergers involving local were of the opposite view. As a two-thirds majority is markets is the same as for all other mergers, they often result required for a CMA panel to find a SLC, it was concluded in a large number of individual local markets having to be scru- tinised as part of the same transaction, which can have both that there was no SLC in the West Midlands. timing and resourcing implications for the parties and the regu- 17. For a more detailed discussion of bidding markets, see lators assessing the transaction. Wirth, D. “To bid or not to bid: the assessment of bidding markets The level of detail required to scrutinise local market cases in merger control”, International Comparative Legal Guide to: also appears to be having an impact on the ability of regulators Merger Control 2017. to adequately address all of the issues within the Phase 1 time- 18. Excludes Iron Mountain/Recall (2016) in which at Phase table, and it is notable that a number of Phase 2 cases involving 1 the CMA identified national concerns and therefore did local markets have resulted in a materially better outcome for the not find it necessary to consider the prospect of an SLC in parties (i.e. in terms of far few site divestments being required, the supply of RMS and OSDP on a local/regional basis. or the transaction being clearly unconditionally). Accordingly, there is a clear balance to be struck between the temptation to dig ever deeper into local market cases in search of the holy grail of reaching the perfect answer, and the prospect that, even if certain local areas did slip through the net, whether competition would be “substantially” lessened as a result.

Merger Control 2020 Ashurst LLP 7

David Wirth is a Director, Economist in the Competition and EU department at Ashurst, London. David is an industrial economist with signif- icant experience of advising clients on the economic analysis of mergers and acquisitions under UK and EU competition law. In addition, he also represents clients in relation to market investigation references, advises on abuse of dominance and cartel cases, and specialises in the calculation of damages in cartel litigation cases. Prior to joining Ashurst, he spent seven years at the UK Department of Trade and Industry as a government economist, working primarily in the Consumer and Competition Policy division. He worked on the development, drafting and enactment of the Enterprise Act 2002, which overhauled UK mergers and market investigations legislation and introduced the criminal cartel offence into UK law. He also worked closely on the UK’s input into the reform of the EU Merger Regulation.

Ashurst Tel: +44 20 7638 1111 London Fruit & Wool Exchange Email: [email protected] 1 Duval Square URL: www.ashurst.com London E1 6PW United Kingdom

Tom Punton is a Senior Competition Economist in the Competition and EU department at Ashurst, London. Tom has significant experience advising clients in relation to a range of competition matters including mergers and acquisitions, investigations of alleged market power and anti-competitive practices and agreements, and competition litigation. Prior to joining Ashurst Tom worked for the UK communications regulator.

Ashurst Tel: +44 20 7859 3854 London Fruit & Wool Exchange Email: [email protected] 1 Duval Square URL: www.ashurst.com London E1 6PW United Kingdom

Ashurst is recognised as having a market-leading, international compe- tition practice in Europe and Asia-Pacific, with top-ranked advisers throughout our global teams. Our multilingual network has a strong track record of providing high-quality advice and achieving successful outcomes for clients engaged in a full range of matters, including mergers, anti-trust investigations, appeals and litigation. The experience of our team spans many industries, including building mate- rials and construction, communications, resources and energy (including gas and electricity), financial services, food, gaming, healthcare, insurance, IT, manufacturing and distribution, entertainment, media, mining, petro-chemi- cals, primary industries, professional services, water, ports and rail. www.ashurst.com

Merger Control 2020