Concepts of “closeness of competition”
Point 1: Market shares may over- or under-state consumers’ “second choices” considerably due to product differentiation.
Point 2: Unilateral effects are more likely, other things being equal, the greater the diversion between the merging parties.
Implications for market definition
Summary of key points
Where would you choose to shop if this store were closed down?
Holland & Barrett/Julian Graves (2009)
StatoilHydro/ConocoPhilips (Jet) (2008)
Looking beyond market shares: the theory, evidence and meaning of closeness of competition in the manufacture, wholesale and retail of fast-moving consumer goods in South Africa and Europe
Adrian Majumdar & Richard Murgatroyd, RBB Economics, August 2009
It is well documented in both economic and legal literature that: (i) market shares may be unreliable measures of market power (the reasons for which are explained below in section 2) and (ii) where markets feature differentiated products, traditional measures of concentration may considerably under- or over-state the extent to which competing firms constrain one another (as we explain in section 3).
Moreover, most products and services are differentiated to some degree – especially retailed products such as fast moving consumer goods (“FMCGs”). As such, detailed assessment of the closeness of competition between firms forms an integral part of both merger inquiries and market investigations in the FMCG sector.1 In this paper we discuss the true meaning of closeness of competition in the context of horizontal merger analysis and its implication for the role of market definition (section 3). We then go on to describe a number of the most prominent empirical techniques used to assess how closely firms compete, drawing upon recent experience from markets for the manufacturing, wholesaling and retailing of fast moving consumer goods in South Africa, Europe and the United States (section 4).
We conclude (in section 5) that to demonstrate that a merger would be harmful, it is not sufficient to show that firms are “close” (or “closest”) competitors, in the sense that a small price rise by one of the parties leads to a high (or relatively high) diversion of demand to the other (and vice versa). It must also be shown that rivals are currently (and likely to remain in the future) sufficiently “distant” from the merging parties – i.e. that rivals are not currently (and will not soon) be effective at targeting the same customer base as that of the merging parties. This is ultimately an empirical question and so we provide case studies to demonstrate several examples of the assessment of closeness of competition in practice. These cases thus provide examples of “practical indicia” (to use the terminology of the South African Competition Tribunal in Massmart/Moresport) which can be used to assess the strength of competitive constraints.
Market shares are not necessarily good indicators of market power
A horizontal merger will harm end customers only if it creates, strengthens or protects “market power”, the latter being the ability profitably to sustain prices above competitive levels.2 Put another way, the concern is that the merger will lead to an increase in market power in the sense that prices will be higher than they otherwise would have been (without there being a compensating increase in quality, range or service3).
There are three “competitive constraints” that may prevent the merged firm exercising higher prices post merger.4 These are:
^ – i.e. the possibility that post merger price rises would not be profitable due to customers switching to (and reactions by) firms already in the relevant market;
Potential competition – i.e. the possibility that post merger price rises would not be profitable due to reactions by firms not currently in the relevant market but able to enter relatively quickly;
^ - i.e. the possibility that post merger price rises would not be profitable due to strategic responses by powerful buyers (such as sponsoring new entrants to the relevant market or, in the case of retailers, diverting substantial volumes to their own private label products).
It is well documented that market shares (and measures of concentration) may not be reliable indicators of market power. While a firm without a high market share is unlikely to have market power, high market shares themselves are not necessarily signs of market power for several reasons. For example, any attempt by the firm with a large market share to attempt profitably to sustain prices above competitive levels could be thwarted in a wide range of different ways, including:
firms with low market shares offering close substitutes to the products sold by the merging firms could expand rapidly without substantial sunk cost;
firms with low market shares (or those not currently in the market) introducing products which are close substitutes to the products sold by the merging firms rapidly without substantial sunk cost; and
buyers with strong bargaining positions based on credible threats to switch to alternative suppliers (e.g. suppliers with low market shares in the market in question or operating in neighbouring markets with sufficient spare capacity to expand considerably in the market in question).
For these reasons, and others, we should be wary of relying on measures of concentration as measures of market power. Put differently, just because a merger increases concentration, this does not mean that it necessarily gives rise to an increase in market power. In our view, therefore, any presumptions based on measures of concentration are therefore better used as safe harbours to rule out concerns as opposed to indicators of harmful effects. Indeed, there are further reasons to be wary of market shares when we analyse mergers in differentiated product markets, which we now go on to explain in detail.
Market shares are not necessarily good indicators of likely unilateral effects
We now turn to the theory of unilateral effects with differentiated products (and services).5 In practice, most products are differentiated to a certain degree – especially at the retail level. This could be due to:
^ For example, two otherwise identical goods are geographically differentiated where a consumer faces different “transport” costs in purchasing each good. For example, it may be more convenient to purchase a new car from a local dealer than from a dealer further away.
^ There are countless examples:
a digital camera is differentiated from a non-digital SLR camera;
coffee shops are differentiated by their internal décor;
cars are differentiated by their features;
Internet retailers are differentiated from “brick and mortar” outlets;
bus services are differentiated from train services, and so on.
^ For example, peak travel is differentiated from off-peak travel.
Switching Costs: For example, two loan products might be identical before a decision to take out a loan is made. However, having taken out the loan there may be penalties for early repayment which mean that ex post the products are different (i.e. switching from one loan to the other would involve a substantial switching cost ex post).
Advertising: Heavily advertised products may be differentiated from less well known products.
^ For example, a branded pain killer may be perceived to be different from a generic pain killer even though the pain killers are made by the same formula and so their physical properties are identical.
In the context of horizontal mergers with differentiated products, unilateral effects arise where the merger creates the incentive for the merged entity to increase prices and where the profitability of that price rise does not depend on a coordinated response by other firms in the market.6 The intuition is straightforward. If two firms that were once competing are brought under common ownership, the loss of competition between the merging firms provides each firm with an incentive to increase price (other things being equal).
The theory of unilateral effects with differentiated products is straightforward. For example, suppose that there are three firms and three goods in the relevant market:
Firm A, that produces good A;
Firm B, that produces good B; and
Firm C, that produces good C.
If firm A increased the price of good A, it would lose some sales as consumers switch to goods B and C, or where consumers leave the market. Similarly, if firm B increased its price, it would lose some sales to A and C, or where consumers leave the market.
Profit maximising firms increase prices up to the point where further increases are not profitable because the lost margins on consumers switching or leaving the market are not recovered by the higher margins earned on those consumers who continue to buy. This is the “pre-merger equilibrium”.
If A and B merge the situation changes. If the price of A goes up, the sales that were lost to B are no longer lost since they stay within the merged entity. This means that (other things being equal) the merged entity may possess an incentive to increase the price of A compared to the pre-merger situation. Similarly, any sales that B lost to A before the merger are now “internalised” (they stay within the merged firm) and so there may be an incentive to increase the price of B as well. 7
A complete analysis of unilateral effects must also take into account other factors that may prevent price rises occurring post merger. First, it is important to assess the scope for supply side responses (e.g. new entry or the introduction of new products by existing competitors) and strategic responses by buyers (e.g. sponsoring new entry). Second, if there are substantial merger synergies that lower marginal cost, the merged entity may wish to lower its prices relative to pre-merger levels.
There are various possible meanings of closeness of competition, and so it is important to explain precisely what we mean when using terms such as “close competitors”, “closest competitors” and “closeness of competition”. We explain this by setting out some points to keep in mind, while referring to our example with firms A, B and C in the preceding sub-section. For convenience, we focus on constraints on product A, although in a merger between A and B we would also be interested in constraints on B.
Suppose that prior to the AB merger firms A, B and C had market shares of respectively 20%, 20% and 60%. If market shares are reliable indicators of “second choices”, i.e. products to which consumers would switch when the price of their first choice goes up beyond a critical level, then we would expect that if the price of A goes up, for every 4 units that leave A and stay in the market, firm B would pick up 1 unit and firm C would pick up 3 units (because C’s share is three times B’s share). More precisely, we expect B’s share of the “market excluding A” (which is 25%8) to equal the share it captures of A’s lost demand (given that this demand remains in the market).
To examine the reliability of market shares as indicators of second choices, we can compare the predictions of market shares with previous actual diversion (where such evidence is available). For example, suppose that we find that B usually captures 20% of A’s lost custom, C captures 30%, while 50% leaves the market. In this case, B captures 40% of the demand that remains in the market – substantially higher than the 25% predicted by market shares. Although C captures 60% of that demand, this is substantially less than the 75% suggested by market shares.
The first important point to take away here is that market shares (in this example) are not good indicators of second choices. This gives rise to our first concept of closeness of competition: “product B is a closer substitute to A than its market share would suggest, while product C is a more distant substitute to A than its market share would suggest”.
The second issue to note is that there is a distinction between which product captures the greatest degree of diversion and which products capture more or less diversion that their market shares would predict. Of the products in the relevant market, C is the strongest individual constraint on A (since it picks up more of A’s diverted demand than B). So it can be argued that: “product C is the closest individual competitor to product A”.
Finally, note that diversion patterns to products outside the chosen relevant market may be important. Since an increase in A’s price leads to 50% of its demand leaving the market, we should not ignore the fact that the collection of “outside products” to which A’s customer base would switch are (collectively) a far stronger constraint than product C.9
There is no clear yardstick for defining when products A and B are particularly “close” substitutes. However, it is fair to say that (other things being equal) when the price of A goes up, the more of A’s demand that is diverted to B, the more likely that after a merger (between A and B) it will be profitable to increase the price of A (and vice versa). Thus, even if firms A and B have relatively low market shares, if market shares substantially under-state diversion between the two parties the AB merger may be anti-competitive. Conversely, if the parties’ combined market share is quite high, this need not indicate a competition problem if market shares may substantially over-state the true constraint that the parties place on each other (because in fact their customer bases do not substantially overlap).
Point 3: A merger of “close” or “closest” competitors may not be problematic due to the presence of other “close” competitors or the ability of rivals to introduce “closely” competing products sufficiently quickly.
An AB merger need not be problematic even where B is a “close” or “closest” competitor to A. First, suppose that when A’s price increases, B captures more of A’s former demand than any other individual product. However, suppose that the absolute amount captured by B is small (e.g. less than 10%) because the large majority of consumers switch to other products. In this case, we might argue that: “while B is the ‘closest’ individual competitor to A, there are also many other products that are ‘close’ competitors to A (being targeted at a similar customer base to that targeted by A)” such that the AB merger is unlikely to give rise to competition concerns.
Second, even if there is historical evidence of high diversion between products A and B, this does not necessarily indicate that an AB merger is problematic where likely future demand and supply side responses are not captured in historic data, for example:
Suppose that historical diversion between A and B has been very high, perhaps because A and B produce the only two products that are compatible with a certain platform (e.g. printer cartridges compatible with a specific printer), A merger between A and B may not lead to a price rise if any increase in prices would make it worthwhile for consumers to invest in a different platform (e.g. switch to a different printer), and having done so they would no longer need to purchase from A and B at all (e.g. they would buy a different brand of printer cartridges).
In some cases, an attempt by firm A to sustain higher prices post merger would lead to other firms (existing competitors or new entrants) introducing a new product to market, which targets a similar customer base to that targeted by firm A. Since this has not yet occurred, switching to the new product will not yet be measurable in the data, although it could be substantial in the event of that product being introduced.
The latter two examples give rise to another concept of closeness of competition: it could be argued that “product B is indeed a close competitor to product A, but other products (including those yet to be introduced to the market) compete with A (almost) as closely” such that the AB merger is unlikely to give rise to competition concerns.
We have explained that market shares may be poor indicators both of market power and of the degree of competition between two merging firms (e.g. where market shares do not accurately reflect second choices for customers of the merging parties). For these reasons it is sometimes argued that there is no need for market definition at all – after all, if market shares are so unreliable why bother defining a market?
In our view, defining the relevant market is important for several reasons. First, the process of market definition – and in particular gathering informative evidence on the scope for demand and supply side substitution – provides an important starting point for understanding the competitive effects of a merger.10 A rigorous attempt to define the relevant market provides an important framework for identifying at the outset the immediate competitive constraints on the merging parties as well as an appropriate reference point for understanding potential competition and buyer power. While this framework can, in theory, also be established via a direct analysis of competitive constraints on the merging parties, in practice, the process of market definition provides greater certainty that competition authorities will fully consider the broad set of potential competitive constraints, and greater transparency as to the approach used by the Authorities to identify them.11
Second, market shares may provide an important filter. In some cases it is possible to identify at an early stage that even on the narrowest plausible relevant market (for example, which takes into account possible closeness of competition by placing all allegedly “close competitors” in their own separate market) the parties’ combined shares are low enough to indicate sufficient post merger competition from rivals.
Unilateral effects are easily understood in principle. If two firms that were once competing are brought under common ownership, the loss of competition between the merging firms may provide each firm with an incentive to increase price (other things being equal). However, to demonstrate that a merger would be harmful, it is not sufficient to show that firms are “close” (or “closest”) competitors, in the sense of there being high (or relatively high) diversion between the two parties. It must also be shown that rivals are currently (and likely to remain in the future) sufficiently “distant” from the merging parties – i.e. that rivals are not currently (and will not soon) be effective at targeting the same customer base as that targeted by the merging parties.
This means that it is important to assess with hard evidence the extent to which firms are close competitors and how close or distant are other competitors. This may include an examination of whether the merging parties can engage in price discrimination and thus increase prices to only a particular group of customers that might be “captive” to the merged firm.12 In the next section, we describe empirical techniques commonly used for this purpose and – where relevant – how they were employed to address the points on closeness of competition identified in section 3.1.
Empirical techniques in assessment of closeness of competition
In this section we provide an intuitive description of many common empirical techniques used in assessment of closeness of competition (we do not seek to provide an exhaustive description of each of these techniques here). However first we set the scene by explaining the danger of relying too heavily on a comparison of characteristics of products as a guide to substitution patterns. We then focus on three of the most popular and accessible ways through which to examine the closeness of competition between firms, namely: survey evidence; impact analysis; and overlap vs. non-overlap analysis.
As a first indication of closeness of competition, consideration might be given to the characteristics of the products of the merging parties. If these products can be said to be “more similar” along certain dimensions than other products in the relevant market, these firms might be considered to be close competitors.
For example, consider a hypothetical example where the relevant market (for the purposes of assessing a merger between two car manufacturers) has been defined to be all new motor cars. Within this wide market, there is a great deal of product differentiation. One way in which the closeness of the merging parties could be assessed is by considering how similar their cars are along a number of dimensions. If the parties both produced family cars, they might reasonably be considered closer competitors than if one of them produced only sports cars and the other only family cars.
However, it must be noted that such an approach does not directly assess the extent to which customers of one product may switch to the other in response to a change in their relative prices. It is not differences in the physical characteristics of the products that determine how closely they compete but the effect those differences have on the substitutability of the products. Comparing characteristics without recognising this point can lead to the wrong conclusions about a firm’s closest competitors. For example, a simplistic product characteristics approach could lead to the conclusion that the price of red sports cars is constrained more by the price of red family cars than by the price of blue sports cars. However, unless colour is the crucial determinant of demand, it is likely that the constraint on red sport cars is stronger from blue sports cars than from red family cars.
A characteristics approach might also naively assume that where firms have different price levels, they cannot be close competitors. However, such a presumption is not appropriate. For example, many retail markets may be characterised as a combination of low cost retailers (e.g. benefiting from procurement and/or scale economies) and specialist retailers that are unable to compete purely on price, but seek to compete through other means such as better range or service levels. A simple characteristics-based approach could lead to the conclusion that the prices charged by a specialist retailer are constrained more by other specialist retailers than by the prices of low cost retailers. However, since consumers are likely to choose retailers on the basis of a combination of price and non-price factors, firms may often be found to differentiate themselves (e.g. by offering better service and quality) precisely in order to compete with rivals that are better able to deliver lower prices. In this sense differentiation can be seen as being driven by competition as opposed to a sign that competition is absent.
This view is expressed in Massmart/Moresport where the Competition Tribunal noted that although sports and outdoor equipment may be divided according to quality and price into three levels, entry, middle and prime, whether these different products categories ought to be considered to be part of a single relevant product market depends upon the willingness of and extent to which the consumer and suppliers would switch between these categories.
Moreover, the Competition Tribunal noted that it is often not possible to identify directly the degree of competitive constraint firms pose upon one another, but rather it is necessary to attempt to infer the strength of competitive constraints via a range of “practical indicia”.13 The remainder of this section provides examples of techniques which could be used to provide such indicia.14
Surveys are commonly used in retail merger investigations to assess (among other issues) the closeness of competition between the two merging firms, and are often conducted outside (or just inside) the stores of the merging parties in order to obtain responses from consumers who have just purchased at one of the merging parties’ stores (although both telephone and online surveys have also been presented in a number of merger investigations). In relation to closeness of competition, surveys are often used to establish which retailers a firm’s customers consider to be the next best alternative(s) to their chosen retailer and thus to help understand which firms are considered to be particularly close competitors to each of the merging parties. For example, typical questions might be:
“Would you switch to another retailer in response to a 10% price increase? If so, which ones?”
Survey analysis may be particularly informative in situations where there is a lack of actual data regarding consumers’ preferences and likely substitution patterns between firms. However the use of survey data must be accompanied by an important caveat, namely that survey results may be subject to a number of different biases. These include those associated with failing to obtain a representative sample, influencing the respondent’s answer by the phrasing of the questions or the timing of the survey, and the fact that consumers may not be able to provide accurate answers to hypothetical questions (such as how they would respond to store closure or price changes).
A recent example of where survey evidence played an important role in merger assessment was in the acquisition of Gamestation Limited (Gamestation) by GAME Group plc (GAME). The transaction was unconditionally cleared by the UK Competition Commission (CC), despite the fact that the two firms were the only high street retailers of both new (“mint”) and previously owned (“preowned”) video games.15
A key question was whether the parties (specialist video games retailers) were sufficiently constrained by other types of retailers (such as generalist entertainment stores, supermarkets, online retailers / auction sites and independent stores). During the investigation, the parties presented considerable survey evidence from internet, telephone and store intercept surveys, while the CC also commissioned a store intercept survey (the CC survey).
One particular area of interest addressed by the surveys was the extent to which the parties were close competitors in the retailing of preowned games and the extent to which the parties would remain competitively constrained post-merger by other retailers of mint and/or preowned games. Survey evidence consistently indicated that customers purchased games from a wide range of retailers and retail channels. This in turn suggested that barriers to switching between retailers were low. Surveys also identified that consumers were price focussed, with a high proportion of consumers comparing prices across retailers prior to making their purchase. This was consistent with the parties’ view that consumers were price sensitive.
The CC also examined which alternative retailers’ consumers would switch to following a price increase. The CC survey asked customers of the parties’ stores where they would go to purchase their preowned games if their current store were unavailable. The results of the survey demonstrated that although customers considered GAME and Gamestation to be close competitors, there were also numerous alternative retailers, such as eBay and Amazon Marketplace, to which consumers would be willing to switch.
It was also found that the large majority of consumers who purchased preowned titles also bought mint titles as well. As such, if one of the parties were to increase the prices of pre-owned games post merger, a substantial share of consumers would be expected to switch to pre-owned alternatives (e.g. available from Amazon Marketplace and eBay) or mint products for sale in other brick and mortar retailers.
Overall, the survey evidence on price-sensitivity, the ease of switching and multi sourcing by consumers across a wide range of retailers, and the propensity of customers to switch to retailers other than the parties of both “mint” and “preowned” games was of considerable importance in allowing the CC to come to the view that effective competition would remain post merger.16.
Impact analysis essentially entails an examination of past industry “shocks” or discrete events in order to gain insights into the customer substitution patterns. Such insights may then be used to draw inferences regarding the extent to which different firms/products are close competitors/substitutes.17 One such example of a shock in a retail sector might be the temporary closure of a retail outlet (e.g. for one month due to refurbishment). Following the closure of the store one may be able to observe where consumers that used to purchase from that store chose to make their purchases in that store’s absence.18
Other examples of impacts to be tested include temporary outages in production, permanent entry or exit, and significant promotional activity. Some examples are discussed below from a number of merger and market enquiries.
An impact analysis based upon promotional activity was undertaken in the UK by the CC in 2006 as part of its analysis of the Heinz/HP Foods Group merger. This gave rise to potential competition concerns as regards the manufacture of tomato ketchup, brown sauce, barbecue sauce, tinned baked beans and tinned pasta products in the UK. In particular, with regard to the supply of tomato ketchup, the CC noted that the transaction would give rise to a significant increment in concentration by bringing Heinz ketchup and HP’s Daddies brand (Daddies) under the same ownership.
The CC sought to analyse the effect of Daddies and Heinz promotional activity on Daddies, Heinz and own label ketchup sales (volume and value share) with a view to being able to draw inferences as to the extent to which consumers considered these products to be substitutes.20 For example, if Heinz ketchup and Daddies ketchup were considered to be close substitutes by consumers, one would expect to see, all else equal, that promotional activity by one of the two brands would cause a substantial increase in volumes for that brand at the expense of the rival brand.21
The CC found that certain promotions by Daddies coincided with significant reductions in the sales volumes of Heinz ketchup, but posited that this reduction in sales volumes was more likely to be a consequence of increases in the price of Heinz rather than any increased competition from Daddies, particularly since the total volume of ketchup sold by the retailer decreased in the relevant period, i.e. Heinz ketchup volume reductions were not compensated for by increases in Daddies ketchup sales volume.22 The CC thus came to the view that despite the increase in concentration in ketchups, the merger would not harm consumers because Heinz and Daddies ketchup products were not important constraints on each other. The transaction was cleared unconditionally. In terms of our analysis of section 3, market shares in this case were found to considerably over-state the constraint the parties placed on each other.
An impact analysis based on the effects of store entry was undertaken by the CC in 2009 as part of its assessment of the acquisition of Julian Graves Limited (JG) by NBTY Europe Limited, in which an overlap existed in the retailing of nuts, seeds and dried fruit (NSF) due to NBTY’s ownership of Holland and Barrett Retail Limited (H&B). At the time of writing the CC had provisionally cleared the acquisition unconditionally, having identified that although the relevant product market was differentiated and that the merging parties were close competitors, they would continue to face effective competition from supermarkets and other retailers such as independent health food stores even though some of these stores possessed a markedly different offering from those of the parties.
The CC sought to examine the effect of entry of competing fascia (supermarkets, JG and drugstores) in local areas on revenues of H&B stores located in the same areas in order to determine whether or not JG was likely to pose a competitive constraint upon H&B, and if so whether this constraint was unique or whether other retailers such as supermarkets were also likely to pose a competitive constraint. The intuition behind this approach was that if JG and/or supermarkets competed with H&B one would expect to observe, all else equal, a reduction in H&B store revenues following the entry of one of these stores.
By conducting an econometric analysis of the effects of entry, the CC found that entry of both JG and medium and large supermarkets gave rise to negative effects upon H&B store revenues, but that the ”revenue effect” caused by JG was significantly greater than that of supermarkets. Taking the result at face value, this might suggest that JG was a far closer competitor to H&B than supermarkets.
However, the CC also noted that the entry of a supermarket in a given local area is likely to give rise to a ”footfall effect”, whereby more people will visit that area, thus potentially increasing revenues for other stores in the area, including H&B stores. This was again identified by the CC’s econometric analysis, whereby the entry of a supermarket that did not stock any NSF products was found to give rise to statistically significant uplifts in H&B store revenues. From this it can be inferred that the true competition effect of supermarkets exceeds the corresponding effect of new entry by supermarkets on H&B revenues that was identified by the CC’s estimation.
In terms of our analysis in section 3, the CC was able to conclude that although there was evidence to suggest that JG was H&B’s closest competitor, supermarkets were likely to be sufficiently close competitors to provide an effective competitive constraint on the parties post merger.
Overlap analysis is essentially an umbrella term for a range of empirical analyses that seek to identify the extent to which firms constrain each other by comparing firm behaviour in areas where both firms are present (i.e. overlap areas) with behaviour in areas where only one of the firms is present (i.e. non-overlap areas). For example, where firms vary their pricing at a local level one might expect firms to set their prices at lower levels when they face effective competition compared to areas where they face little or no competition. Similarly, where a firm faces a unique competitive constraint from a particularly close competitor, one would expect that firm to set its price at a lower level in local markets where that competitor is present compared to other local markets where that competitor is not present.
Overlap analyses may include comparisons of prices, margins or metrics of quality, range and service in overlap and non-overlap areas, and are typically undertaken using empirical techniques such as regression analysis in order to account for the range of factors (other than competition) that might affect local metrics of competition. Only by taking into account these factors will it be possible reliably to estimate the effect that specific competitors have on the competitive behaviour of a given retailer.24 We describe a number of recent examples where such analysis has been undertaken below.
In StatoilHydro/ConocoPhillips (Jet), an EC merger, an overlap vs. non-overlap type framework was used by the merging parties to gauge whether the presence of Jet petrol stations in local geographic markets was associated with more aggressive Statoil prices. One of the theories of harm to be tested was that despite its low market share, Jet “punched above its weight” being a particularly important constraint on its much larger rival, Statoil.
An interesting feature of the data was that there were periods of price wars (not initiated by Jet but caused by cost cutting exercises by other retailers) and periods of relative price stability. The parties demonstrated that in normal periods of competition (i.e. outside disequilibrium periods characterised by price wars), Statoil was not systematically constrained by Jet. Similarly, the European Commission’s own analysis was not able to detect an economically significant impact on Statoil’s prices from the presence of Jet outside price war periods.
However the Commission considered it appropriate to aggregate together periods of price wars and stable pricing and in so doing came to the view that Jet did pose a material constraint on Statoil. While not explicitly discussed by the Commission, the implied theory of harm would be (the unusual one) that the merger would give rise to a situation where, during price wars, the prices would not be quite as low as they otherwise would have been.
The overlap vs. non-overlap approach can be applied to non-price features as well as price features. Such analyses were undertaken in the merger of two UK book chains, Waterstones plc (Waterstones), owned by HMV group plc, and Ottakars plc (Ottakars). Here the UK Competition Commission considered not only whether the merger would impact on price, but also whether there would be a reduction in the range of books and service quality offered. The CC cleared the transaction unconditionally despite the fact that the parties might appear to be close competitors on the basis of their characteristics (i.e. both specialist high street retailers of books). The CC found that the merged entity would have no incentive to increase prices (which were set nationally) or reduce range or service quality (which were in part set locally) because of the competitive constraints that would continue to be provided by other high street book retailers, supermarkets and online retailers.
Both the parties and the CC were able to gain access to sufficient data to conduct a series of econometric tests designed to examine the extent to which the parties’ stores’ range and service levels varied between overlap and non-overlap areas. The merging parties demonstrated using econometric analysis that there was no evidence of a statistically significant uplift in range when the other party was present.
A similar analysis was conducted with respect to service quality, although this analysis was not straightforward due difficulties in accurately measuring service quality. The CC considered a variety of measures including number of staff, experience of staff, opening hours, refits and refurbishments, and book signings. Broadly speaking, both the CC’s and the parties’ econometric analyses found no evidence that the merging parties increased service in overlap areas with the exception of book signings and refurbishments, where the evidence was mixed. The CC considered the latter features to be of relatively minor importance to overall service quality. Moreover, the parties argued that book signings were not a good indicator of service competition between retailers because (i) they were partly determined by the publisher as opposed to the retailer and (ii) they were not ranked highly by consumers in terms of features that they valued.
Taking all the evidence on range and service quality together, the CC concluded that the Waterstones/Ottakars merger would not give rise to an adverse impact on range and service quality. The interesting aspect of this case was that although the parties appeared similar in terms of their characteristics (specialist brick and mortar booksellers offering a large range of titles), in fact the parties were heavily constrained by competition from different business models – notably supermarket and internet retailers.
Where firms set prices uniformly across their entire estate and/or where range/service quality cannot be easily measured, margins (provided they are measured correctly) may provide an aggregate measure of the amount of competitive ”effort” expended by a particular firm in a given local market.28 A comparison of margins in overlap and non-overlap areas may therefore provide useful insights into the extent to which different firms constrain one another.
A recent prominent example where overlap analysis has been undertaken using margins was by the US Federal Trade Commission (FTC) in the merger between Whole Foods Market, Inc (Whole Foods) and Mild Oats Markets, Inc (Wild Oats), before the US Court of Appeals (the Court). The transaction brought together two of the leading premium, natural and organic supermarkets (PNOS) in the US, which were moreover considered by the FTC to be extremely close competitors. Initially the FTC had its request for a preliminary injunction against the merger denied by the Court, though subsequently this denial was overturned and a preliminary injunction granted by the Court, following a commitment to divest 32 stores and intellectual property associated with the Wild Oats business (including the brand) the FTC approved the transaction.29
The FTC provided evidence that a substantial proportion of consumers in the PNOS sub-market considered that (i) the ability to purchase natural and organic produce (termed ‘high quality perishables’) and (ii) service levels were important, and that such consumers considered Whole Foods and Wild Oats to be superior to other supermarkets in these respects. Moreover, the FTC noted that other grocery retailers did not stock large ranges of high quality perishables.30 Such a finding might therefore suggest that the merging parties were in a position to price discriminate against consumers of high quality perishables, and that although the merging parties would continue to face effective competition from other grocery retailers for standard groceries, they might (i) be particularly close competitors with respect to the sale of high quality perishables and (ii) face extremely limited competitive constraints from other non-PNOs retailers for these products.
In order test this hypothesis the FTC sought to compare the margins earned by Whole Foods across both high quality perishables and other grocery products, and across stores facing differing levels of local competition. The FTC’s analysis identified that, all else equal, the presence of a Wild Oats store in the same local market as a Whole Foods store depressed Whole Foods margins significantly (e.g. due to lower prices and/or improved service) on high quality perishables, but had no effect on the margins earned on those products over which the merging parties also competed with other grocery retailers. These results would therefore appear to support the hypotheses that the parties were able to price discriminate against consumers of high quality perishables, that they were close competitors for the retail of these products, and that they were likely to be relatively unconstrained by competition from other supermarkets with respect to these products. In other words, the parties were found to place a unique competitive constraint upon one another with respect to high quality perishables, such that they would have been likely to be in a position to increase the prices of these products post-merger.
We argued that, in theory, to demonstrate that a merger would be harmful, it is not sufficient to show that firms are “close” (or “closest”) competitors, in the sense that a small price rise by one of the parties leads to a high (or relatively high) diversion of demand to the other (and vice versa). It must also be shown that rivals are currently (and likely to remain in the future) sufficiently “distant” from the merging parties – i.e. that rivals are not currently (and will not soon be) effective at targeting the same customer base as that of the merging parties. This is ultimately an empirical question that can either be answered directly or through an analysis of “practical indicia”, and so we provided examples of cases from Europe and the US where several different empirical approaches to the assessment of closeness of competition have been employed.
1 Our focus is on merger analysis although many of the techniques described in section 4 are relevant for the assessment of market definition and competitive effects in market investigations.
2 In the case of a horizontal merger and the theory of unilateral effects, the issue is often “would the merger lead to higher prices?” In this case, even if pre-existing market power exists, it is not of central importance because the key issue is whether market power is enhanced further. In some special cases, however, pre-existing market power may be of particular importance. For example, where the merger is with a potential entrant, it may prevent pre-existing market power being eroded by the entrant. In this case, it is helpful to identify whether the incumbent firm has market power – if not, the new entrant could not have eroded its market power and so the merger is not problematic. (With non-horizontal mergers, the concern is foreclosure. In these cases pre-existing market power is important to identify – without it, the merger will not be harmful.)
3 For convenience we focus on price rises. However, the concern could be that the merger leads to a reduction of quality, range or service (or even incentives to innovate).
4 See for example the EC Commission Horizontal Merger Guidelines: European Commission, “Guidelines on the assessment of horizontal mergers under the Council Regulation on the control of concentrations between undertakings”, Official Journal of the European Commission, February 5, 2004.
5 This section draws heavily on Majumdar, A. (2009): “Horizontal mergers: Unilateral Effects”, a teaching module written for Kings College London’s Postgraduate Diploma/Masters in Economics for Competition Law.
6 That is to say, this essay does not discuss mergers which may give rise to coordinated effects. In the latter case, the merged firm could profitably sustain higher prices post merger only as part of a coordinated strategy with other rivals, underpinned by credible threats to punish those firms that deviated from the coordinated strategy.
7 There are also “second round” effects which lead to non-merging firms increasing prices. If the merged firm would increase prices due to the internalisation of competition between A and B, then C will gain some extra consumers. This is not a result of coordination – rather, if the prices of A and B go up, some demand is shifted to product C, so its price goes up as well. This is why “unilateral effects” are sometimes referred to as “non-coordinated effects”.
8 A’s share is 20%, so B’s share of the market excluding A equals 20% / (100% - 20%), which equals 25%.
9This begs the question of why such outside goods were not included in the relevant market. This topic merits an essay in its own right but the short answer is as follows. Error could be one (unfortunate) explanation. This highlights “error costs” of defining markets too narrowly at the start of an investigation. That is to say, at the outset, it is better to define a market widely so as to ensure that substitute products are not erroneously omitted. Even if, based on further analysis, products are found to be only weakly substitutable, this fact can be accounted for in the competitive assessment when all the evidence on market power is weighed in the round. Having said this, in some cases substitution could take place to many diverse products such that it would be hard to capture them sensibly in a market definition. As a hypothetical example, suppose that when assessing a newspaper merger it emerges that if the price of a certain newspaper went up, 50% of consumers would stop reading that paper but would switch to reading internet articles, magazines, reading books or watching television. In this case, we should certainly take those constraints into account but it is not clear that to do so would lead us to define a market for (say) “the supply of entertainment in the formats of the internet, newspapers, magazines, books or television”.
10 Demand side substitution assesses how customers respond to small but significant, non-transitory changes in price – it sheds light on how easy it is for customers to switch, to which products they would switch and how quickly they would do so. Supply side substitution assesses how suppliers respond to small but significant, non-transitory changes in price – it sheds light on the ease of introducing new products to the relevant market. Supply side substitution can be thought of the scope for rapid new entry that occurs with relatively little sunk cost, e.g. capacity that can quickly be switched to serve the relevant market.
11 It should also be noted that empirical analyses of switching often impose restrictions on substitution patterns as a way to resolve problems with having insufficient data. The process of market definition helps to provide a reality check as to whether such assumptions are valid.
12 See the discussion of Whole Foods/ Wild Oats in section 5.
13 While the authors acknowledge that it may not always be possible to identify directly the cross-price elasticities between firms, this inability does not render the SSNIP framework to be ineffective, or market definition not useful, as seems to be suggested in the Tribunal’s decision. See our discussion in section 3.2 and see Competition Tribunal, Republic of South Africa, “In the large matter between Massmart Holdings Limited and Moresport Limited”, Case No: 62/LM/Jul05, for the decision.
14 The case examples presented focus on mergers at the retail level, although the techniques applied would equally apply to mergers further up the supply chain. Indeed, the issue of closeness of competition appears central to the South African Competition Commission’s recommendation to prohibit the acquisition of grocery wholesaler Finro by the Massmart group- Competition Tribunal, Republic of South Africa, “In the large matter between Massmart Holdings Limited and Moresport Limited”, Case No: 62/LM/Jul05
15 UK Competition Commission, “GAME Group plc / Game Station Limited: Final Report”, January 16, 2008.
16 It is worth noting that a number of the general criticisms of survey analysis provided above are partly addressed by the facts that: (i) a range of different surveys all presented a consistent picture regarding these results; and (ii) a number of the survey questions related to actual purchasing behaviour (e.g. “where do you purchase preowned games”) rather than hypothetical responses.
17 It can also be used to examine substitutability within a market definition context.
18 Note that this bears a clear ‘real world’ answer to the survey question ““^ ” Indeed, relating this to our discussion of concepts of closeness of competition in section 3, consider the hypothetical example of a local retail market that contains three retailers, A, B and C, each differentiated in terms of their product offerings. Next consider that retailer A closes for several months in order to undertake refurbishment. During this period, consumers who would have originally purchased their goods from retailer A are now faced with the decision (assuming that they wish to continue to buy the relevant product) of purchasing from either retailer B or retailer C. Subsequently, if (say) 70% of A’s customers switch to B while the remaining 30% switch to C, this would appear to suggest that retailer B is a greater constraint than retailer C to retailer A.
19 UK Competition Commission, “Heinz / HP Foods Group: Final Report”, March 24, 2006.
20 The CC used regression analysis to isolate the own effects of promotions on sales volumes by regressing Daddies sales, Heinz sales and own label sales individually on dummy variables for each type of promotion undertaken by Heinz and Daddies. The effect of a competitor promotion on sales was found to be small, unsystematic and likely to vary over a considerable range, which the CC took as evidence that Daddies and Heinz exerted at best only a weak competitive constraint upon one another.
21 Retail level promotions were considered (i.e. consumer reactions to promotions were analysed). The merger related to the manufacturing level, where retailers are the direct customers. The CC thus inferred that consumer behaviour would ultimately influence retailer decisions when dealing with their suppliers.
22 Note that in the Heinz case, consumer switching evidence at the retail level was considered as a way to understand competition at the wholesale level. While such information is often useful, there may often be occasions when evidence on the sensitivity of retail demand is not a good proxy for the sensitivity of wholesale demand.
23 UK Competition Commission, “Holland & Barrett / Julian Graves: Provisional findings report”, July 22, 2009.
24 It should also be noted that there must exist sufficient “variation” in the dataset used for these pieces of analysis for it to be possible accurately capture the effects the presence/absence of different retailers has on the behaviour of firms in a given local market. For example, if retailer B is always present in the same local markets as retailer A, it will not be possible to estimate what effect the presence of retailer B has on retailer A’s prices since there exist no basis to compare the prices of retailer A in areas where retailer B is present against those areas where retailer B is not present. However, if B’s share of local markets differs substantially, it may nonetheless be possible to make meaningful inferences by comparing areas where B has a strong presence against those where B has a weak presence. It is also important to consider be confident that competitive behaviour in a given local market is independent of that in other local markets (it might not be if local markets are defined incorrectly or because potential entry from one local market to a neighbouring market is an important constraint).
25 Commission of the European Communities: “Case No COMP/M.4919 – STATOILHYDRO/ CONOCOPHILLIPS”, October 21, 2008.
26 UK Competition Commission, “HMV Group plc / Waterstone's plc / Ottakar's plc: Final Report”, March 31, 2006
27 United States Court of Appeals, “Federal Trade Commission v. Whole Foods Market, Inc. et.al Appeal from the United States District Court for the District of Columbia (No. 07cv01021)”, July 29, 2008.
28 It is important to note that margins figures that are constructed for accounting purposes may not accurately capture the true economic costs incurred by firms in order to compete with rivals.
29 US Federal Trade Commission, “FTC Consent Order Settles Charges that Whole Foods’ Acquisition of Rival Wild Oats was Anticompetitive”, Press Release, March 6, 2009.
30 Although discussed in the US Court of Appeals’ decision to deny the FTC’s preliminary injunction, little mention of the potential competitive constraints provided by expansions in the range of non PNOS retailers are made in the US Court of Appeals’ reversal of that decision. Although as the Court identifies, its decision is concerned solely with the granting of the preliminary injunction and merely seeks to identify whether there are grounds for more in-depth analysis.
RBB Economics Page