The G roceries Groceries Remittal Decision Decision Some concerns 12 November 2010

Size: px
Start display at page:

Download "The G roceries Groceries Remittal Decision Decision Some concerns 12 November 2010"

Transcription

1 The Groceries Remittal Decision Some concerns 12 November 2010

2 Disclaimer My views Not the views of Frontier Or of Tesco

3 My conclusion: the remedy is unsafe Based on a misreading of the analysis in the Groceries Report Doesn t take into account the reality of the market

4 The CC s framework is sensible Size of gap determines scale of benefits Length of delay before replacement determines duration of costs Probability of replacement determines whether there will be benefits CC, Remittal Report, Figure 1, 5.9

5 but overestimates benefits and underestimates t costs Benefits Benefits from econometric analysis in Report Scaled up in Remittal process Now conflict with original i findings and costs Costs depend on whether replacement entry occurs Assumed easy in Remittal But based on different theory of harm to Report

6 and the net benefit is negative after a handful of changes to the CC s assumptions 3,000 2,770m 2,500 t value of reme edy ( m) 2,000 1,500 1,000 Net present m m - 331m -500 Base case Removing the effect of "multiplying the margin concentration results by 4" Using a margin of 15.9% rather than 20% 10% of extensions not replaced Long term delays equal to short term delays -1,000 NPV

7 Benefits Costs Comment

8 Benefits arise when there is replacement by a new fascia Size of gap determines scale of benefits CC, Remittal Report, Figure 1, 5.9

9 based on the CC s margin-concentration analysis Regress store margin on number of rival fascia in area plus controls Large stores: linear specification Store margin falls 3.79% per additional large store fascia (e.g. from 20% to 19.4%) Appendix 4.4 Basis for CC s detriment finding of m excess profits per year across the industry multiply pymargin-concentration relationship by number of concentrated areas (Remittal, paragraph 2.3; Report, 6.54 and footnote 1 p 117) Not for today but I have doubts over validity of CC s analysis. See J Hausman and D Parker, Margin-Concentration Analysis In The UK Groceries Inquiry, Journal of Competition Law and Economics (2010)

10 However, the CC modified the results The CC also investigated the margin concentration relationship for all stores above 280 sq m (with all stores above 280 sq m as competitors) Linear specification: store margin falls 0.96% per additional fascia Quadratic specification: margin falls 4.6% per fascia above 280 sq m but increases by 0.33% per fascia above 280 sq m squared CC argument: quadratic relationship shows higher result for monopoly stores than the average (linear) result, by about 4.5 times; and for duopoly stores, by about 3.8 times; and for triopoly stores by 3.1 times Therefore: Scale up results for large stores in same fashion, reading across from the all stores results

11 which is inappropriate because the two distributions are different Proportion of obs servations 40% 35% 30% 25% 20% 15% 10% The large stores results will be driven by duopoly and triopoly stores while the all stores results are based on a much more even distribution 5% 0% Number of rival fascia Proportion of rival fascia 280 sq m + Proportion of rival fascia 1400 sq m + The CC could have published results from a quadratic specification on large stores, but did not do so

12 leading to perverse results Implie ed percentage decline in marg gin 20% 18% 16% 14% 12% 10% 8% 6% 4% This is the average effect of 3.79% per fascia There are no areas with more than 5 fascia so all the CC s parameters are above the average 2% 0% 0 to 1 1 to 2 2 to 3 3 to 4 Number of rival fascia Modified margin-concentration result

13 and a substantial overestimation of the benefits Annual gross benefits of the competition test m Each year the CC assumes that 4.4% of the problem areas will be solved so after 5 years 22% of the problem is solved, while more than 100% of the detriment is solved 300 Annual detriment the test should rectify ( 105m - 125m) Year

14 Using the original results reduces the NPV of the test by over 90% Net pr resent value of remedy 3, ,500 2,000 1,500 1,000 This change removes 91.4% of the net benefits CC base case Removing the effect of "multiplying the margin concentration results by 4" NPV

15 The CC seems to have used the wrong margin Industry average margins 15% (B48) CC assumes 20% in concentrated areas (B49) Implied margin in unconcentrated areas 13% 27% of stores are in concentrated areas (6.14) 7 percentage point margin differential between concentrated and unconcentrated

16 and the implications are inconsistent with the evidence Implications Marginconcentration analysis 3.79% margin impact per fascia, so monopoly 20% margin duopoly is 19.4% triopoly (i.e. unconcentrated) is 18.8% A 1.2 percentage points difference National prices so no price difference Source of margin differences must be savings in store costs but these must be very large - store payroll costs c. 7%-10% of Source of margin turnover, so a 7 percentage point margin differential is differences equivalent to operating stores with nearly no staff CC found differences were intangible and difficult to measure would have expected such large differences to have been observed at the Report stage

17 Using a margin of 15.9% makes the test NPV-negative Net pre esent value of remedy 3,000 2,500 2,000 1,500 1, A 15.9% margin in concentrated areas implies a 14.7% margin in unconcentrated areas a 1.2 percentage point difference Starting margins matter because the margin-concentration analysis is expressed as a percentage of margin 0 CC base case Removing the effect of "multiplying the margin concentration results by 4" Using a margin of 15.9% rather than 20% -500 NPV

18 Benefits Costs Comment

19 Costs arise when there is delay or no replacement Length of delay before replacement determines duration of costs Probability of replacement determines whether there will be benefits CC, Remittal Report, Figure 1, 5.9

20 Material difference between blocked extensions and blocked new stores New stores Extensions Blocking a new store means a site is potentially available Blocking an extension means Existing site unlikely to be attractive (or available) New site will be difficult to find

21 Key issue is what is known by rival Incumbent knows it is blocked from developing in a particular area, so doesn t search for sites Suppose incumbent would have found a site in a particular area at time t Entrant would have to search in exactly the same place to find the same site Scenario 1: If entrant finds site before e t, no effect of test entrant would develop first Scenario 2: No delay if entrant finds site at exactly the same time t (knife-edge result) Scenario 3: Delay if entrant would oud have found site after t and hence test applies The asymmetry means there must be a delay

22 which is a particular problem for extensions Incumbent knows it is blocked from extending in a particular area, so doesn t search for sites Extension site unlikely to be attractive to new entrant (likely to be small and next door to existing store) Entrant would need to explore area to find completely new site (potentially long delays involved) If anything, the problem gets worse over time you don t know where the incumbent might have tried to extend in the absence of the test

23 Delays from planning can be material and CC s analysis omits search time This is the critical issue especially for extensions Search time to find a suitable site (not measured) Site assembly: average 18 months, 20% of cases up to 4 years (Report, 7.47) Holding time 1 (time between completion of assembly and submission of planning): Varies, no figure given, could be concurrent with outline application Planning permission: months full application, 40 months outline application (Report, ) Call in: adds months (Report, 7.50) Holding time 2 (time between planning and site): not relevant as Competition Test would already have kicked in

24 suggesting that the assumptions on delay and replacement (100%) are optimistic Category of development Short-term delay Long-term delay Transition period New store replaced by another new store Extension replaced by a weak incumbent extension Extension replaced by a new store CC, Remittal Report, Table 2, Annex C, Paragraph 21

25 The CC s evidence on rival entry is very general and ignores problems in finding sites No reason to Grocery retailers suppose site devote considerable assembly and store resource to exploring construction time development would be any different between entrant and incumbent (5.85) opportunities and much of the data is publicly available (5.85) A potential entrant might have an advantage over an incumbent since an incumbent might not want to cannibalise its sales (5.85) Competitors present or interested in entering in 95% of areas where a grocery retailer planned to open a new store (5.87) but this is where retailers would like to enter - no identified sites and no opportunity yet subject to economic evaluation (footnote 65, emphasis added) 85% of planned extensions in areas where there was already a rival fascia note that we can (of which 50% had not always show someone else whether the retailer interested in the area) would be interested in and for the remaining 15%, two-thirds of areas had someone interested (5.95) entering within the isochrone as the retailer had not identified a particular site (footnote 69, emphasis added)

26 If 10% of extensions are not replaced, NPV falls further t value of reme edy ( m) 3,000 2,500 2,000 1,500 1,000 2,770m Both Tesco and Sainsbury put in evidence based on analysis of analogous historical situations which suggested that fewer than 10% of extensions would be replaced this current sensitivity may well be conservative. To go halfway, if 50% of extensions are not replaced, NPV falls to - 693m Net present m - 55m - 182m -500 Base case Removing the effect of "multiplying the margin concentration results by 4" Using a margin of 15.9% rather than 20% 10% of extensions not replaced -1,000 NPV

27 and further still if we assume that delays persist throughout (no transition period ) 3,000 2,770m 2,500 t value of reme edy ( m) 2,000 1,500 1,000 Net present m m - 331m -500 Base case Removing the effect of "multiplying the margin concentration results by 4" Using a margin of 15.9% rather than 20% 10% of extensions not replaced Long term delays equal to short term delays -1,000 NPV

28 The problem is that the CC s theory of harm has changed Report Problem: barriers to entry caused by the planning regime make it difficult to find sites (e.g. sequential test) But there is plenty of demand available (work on population thresholds h suggests UK can sustain double the current number of stores) Theory of harm: a supply problem, not a demand problem Remittal Problem: actual or potential expansion by incumbent soaks up demand and prevents entry But if this is prevented it is easy for a new entrant to come in and find a site Theory of harm: a demand problem, not a supply problem

29 Benefits Costs Comment

30 The remedy affects upstream competition for site development Search for new sites in good areas The CC didn t investigate how this competition works or the effect of the remedy on firms incentives to search for sites Build new store and compete for customers instead focusing on downstream effects for customers once sites have been acquired

31 Model 1: auction approach Landowner decides to sell plot of land Put this up to auction and get all the grocery retailers to bid Award the land to the highest bidder

32 consistent with the remedy assumptions Remedy blocks strong incumbent from bidding If strong incumbent would not have been highest bidder, no effect If strong incumbent would have been highest bidder, second bidder wins Remedy effective y with no delay (new stores only)

33 Model 2: search approach Each firm searches for land that suits its preferences (may include parcels) Use estate agents as intermediaries in local areas Negotiate individual deals with landowners

34 inconsistent with the remedy assumptions Remedy edy blocks strong incumbent from developing land in a particular area, so no search activity carried out by incumbent in that area Rival searching in area may or may not find site (depending on preferences, agents used, luck, timing) Rivals not searching in area Remedy may result in replacement store but with unknown delay No replacement

35 but backed up by the evidence Only 102 successful and unsuccessful bids from 2000 to 2006 matched to the same land sites (50) there were 681 developments over the period (Remittal 5.62) The process of acquiring a site from a developer is far from the simple auction model we might have expected to see (52) Retailers do not often bid together (51) CC Land Holdings working paper July 2007 Instead we see sellers of land sounding out retailers; extensive informal negotiations; non-board approved bids (52) As a result, a straightforward auction of an identifiable site is rarely observed (53) All the largest grocery retailers often assemble their own developments rather than purchase them from developers no bidding necessary (5.85)

36 Frontier Economics Limited in Europe is a member of the Frontier Economics network, which consists of separate companies based in Europe (Brussels, Cologne, London and Madrid) and Australia (Melbourne & Sydney). y) The companies are independently owned, and legal commitments entered into by any one company do not impose any obligations on other companies in the network. All views expressed in this document are the views of Frontier Economics Limited.

37 FRONTIER ECONOMICS EUROPE LTD. BRUSSELS COLOGNE LONDON MADRID Frontier Economics Ltd, 71 High Holborn, London, WC1V 6DA Tel. +44 (0) Fax. +44 (0)