0 File No. CT-98/2 0 THE COMPETITION TRIBUNAL IN THE MATTER OF THE COMPETITION ACT, R.S.C. 1985, D c.C-34, as amended, and the Competition Tribunal Rules SOR/94-290, as amended (the "Rules"); u AND IN THE MATTER OF an inquiry pursuant to subsection 10( 1) (b) of the Competition Act relating to the proposed D acquisition of ICG Propane Inc. by Superior Propane Inc. AND IN THE MATTER OF an application by the D Commissioner of Competition for orders pursuant to s. 92 and other provisions of the Competition Act. D BETWEEN: u THE COMMISSIONER OF COMPETITION a Applicant -and ll SUPERIOR PROPANE INC. and ICG PROPANE INC. ~ Respondents D REPLY AFFIDAVIT OF DOUGLAS WEST D TO THE REBUTTAL AFFIDAVIT OF DENNIS W. CARLTON AND GUSTAVO E. BAMBERGER u William J. Miller COMPETmOH TRIBUNAL Jo'Anne Strekaf TRIBUNAL oe LA CONCURRENCE I Steve Robertson Jennifer Quaid ~ o ~ .. r .r 29 t~;.,~ ,A I A ? g 0 e Jamie Bocking 0 AEGISHWl · REGISTRAIRE f Counsel to the Commissioner of Competition c/o Department of Justice Place du Portage, Phase I
50 Victoria Street D Hull, Quebec K1A OC9 Telephone: (819) 997-3325 D Fax: (819) 953-9267 u
0 0 FILE NO. CT-98/2 0 THE COMPETITION TRIBUNAL D IN THE MATTER OF THE COMPETITION ACT, R.S.C. 1985, c.C-34, as amended, and the Competition Tribunal Rules u SOR/94-290, as amended (the "Rules"); AND IN THE MATTER OF an inquiry pursuant to subsection u lO(l)(b) of the Competition Act relating to the proposed acquisition ofICG Propane Inc. by Superior Propane Inc. D AND IN THE MATTER OF an Application by the Commissioner of Competition for an order pursuant to section 92 of the Competition Act D BETWEEN: D THE COMMISSIONER OF COMPETITION 0 Applicant u -and-u SUPERIOR PROPANE INC. and ICG PROPANE INC. Respondents D u REPLY AFFIDAVIT OF DOUGLAS WEST TO THE REBUTTAL AFFIDAVIT OF DENNIS W. CARLT ON AND GUSTAV O E. BAMBURGER I, Douglas West, of the City of Ottawa, in the Province of Ontario, MAKE OATH AND ~ SAY THAT: D 1. I was retained by the Commissioner of Competition to provide expert evidence in this u matter, to review the expert reports and rebuttal reports submitted by the expert witnesses of the Respondents, and to provide my comments with respect to those reports. D
D L:\c3\084\affidavit4.doc
0 0 2. I have reviewed the expert witness rebuttal affidavit of Dennis W. Carlton and Gustavo u E. Bamburger herein. Attached hereto and marked as Exhibit "A" is a true copy of my comments with respect to that Affidavit. The contents of Exhibit "A" and the findings u and opinions expressed therein are true to the best of my knowledge, information and belief. 0 3. Attached hereto and marked as Exhibit "B" is a true copy of my curriculum vitae. D 4. I make this affidavit pursuant to Rule 47(2) of the Competition Tribunal Rules. u SWORN BEFORE ME ) AT CALGARY, ALBERTA ) u TIIIS 20TH DAY OF September 1999 ) u ~~Of[~ -=-~~~~~-:!!:-===--=.:.::=±=-=----- ) . DOUGLAS~-) D ~ 0 u u u u u u 0 L:\c3\084\affidavit4.doc
0 0 D u u u a THIS IS EXHIBIT "A" TO THE AFFIDAVIT OF DOUGLAS u WEST, SWORN BEFORE ME THIS 20TH DAY OF SEPTEMBER, 1999 u A B ~~TER AND SOLlCT TOR AND NOTARY u PUBLIC IN AND FOR THE PROVINCE OF ALBERTA u 0o·frr1n-e &--trfkA-F u u u
D u u u D S:IC3\S02\EXA. WEST.doc
0 I. Introduction u 1. In this Affidavit, I will respond to a criticism of my market share estimates that was made in the Affidavit of Dennis W. Carlton and Gustavo E. Bamberger. Professor Carlton and D Dr. Bamberger state that my market share estimates are less reliable than information u provided to them by Superior (SPI). My share estimates, however, utilize all available market share information supplied by SPI and ICG. In addition, where available, they u use information on actual competitor sales volumes and the sales volumes of SPI and ICG branches. In my view, my market share estimates rely on more comprehensive u information, including objective third party information, than the share estimates made by . SPI, and hence in my opinion are more reliable. SPI's estimates of its own and its u competitors' 1998 market shares appear on templates prepared by branch managers and provided to the Commissioner in response to Undertaking 63. While SPI also reports its 1998 branch sales volume on the template, it does not know with certainty the sales volumes ofICG and its other competitors. Its market share estimates are not based on knowledge of actual sales volumes.
2. The Affidavit of Dennis W. Carlton and Gustavo E. Bamberger also contains a table, Table 3, listing the new entrants into retail propane markets, the year in which they entered, and the 1998 estimated sales volume. This table contains more information than the new entrant table (Table 5) in Professor Carlton's original affidavit. However, it is not clear where the 1998 estimated competitor sales volumes come from. While I have u now received the ANNVOLl and ANNVOL2 data files as part of Professor Carlton's Rebuttal Affidavit, and these files are apparently derived from SPI' s branch templates u (that were provided to the Commissioner in response to Undertaking 63), the competitor sales volumes do not appear to be generated by the branch templates supplied by branch managers. 1 The table is also misleading in that some new entrants are shown entering into multiple markets, but the table does not indicate whether the multi-market entry involves establishing one branch or several branches. The table can give the impression that entry
1 There also appears to be data entry errors in the ANNVOL files that could lead to continuing problems in the regression analysis. For example, in ANNVOL2, row 9, market participants in Guelph are reported as having 22,245,00 litres of sales in 1998, and in columns 10 through 15, their 1998 sales are listed at 22,245 litres. This would appear to be a data entry error.
0 -2-D can occur at small scale. However, the scale of entry is larger when one considers the u total volume being shipped from a given branch. In addition, the table fails to indicate which new entrants are SPI or ICG agents. These new entrants are not fully independent 0 competitors with SPI and ICG. LI 3. Finally, I examine two data files that were supplied by Professor Carlton and Dr. Bamberger in support of their revised regression analysis. They have not clearly u identified the source for the volume sales data that are contained in these two data files, and therefore I cannot test the accuracy of the figures contained therein. u II. Market Share Estimates 0 u 4. Professor Carlton and Dr. Bamberger state, on page 8 of their Affidavit, that they consider my market share estimates to l;>e less reliable than information provided to them u by SPI. Appendix B to their affidavit explains why they believe my market share estimates are unreliable. First, they state that the Superior estimates are based on the u knowledge of an employee of Superior in that area, whereas they allege that I have little or no information on the sales by independents in many of the markets. Second, they u state that in many cases, the SPI/ICG share for a market reported by me appears to be inconsistent with my volume estimates for that market. u 5. With respect to their first point, SPI's share estimates that appear in the branch templates u are based on employee estimates prepared after commencement of this proceeding. I have actual sales volumes for many of the independent propane dealers in Canadian retail u propane markets, making for more accurate market share estimates. Professor Carlton and Dr. Bamberger also rely solely on SPI's market share estimates, ignoring the ICG market u share estimates that are available, whereas I use share estimates provided by SPI and u ICG, in conjunction with sales volume data from SPI, ICG and independents. u
0
D -3-u 6. It is the case that a number of independents chose not to report their sales volumes. However, it is also the case that many of the larger independents, particularly those with u multiple branches, did report their volumes. In addition, for those markets where u sufficient volume information from independents was not available to allow me to calculate market shares using volume sales data, I used SPI and ICG estimates of market u shares in conjunction with volume sales data from SPI and ICG. u 7. With respect to their second point, Professor Carlton and Dr. Bamberger construct an appendix table, Table B-1, to illustrate discrepancies between my reported SPI/ICG u shares and the shares that they calculated using my data on SPI, ICG, and independents' branch sales volumes. The reason they are able to find discrepancies is because they u assume that when, for a given market, I report an independent's branch sales volume, all of that independent's sales volume is in that particular market. This is frequently not the D case. Some independent branches serv~ multiple markets given some are located between SPI and ICG branches. Even in Professor Carlton and Dr. Bamberger's table of new E entrants, they have a number of examples of multi-market entry by an entrant serving u these markets from a single branch. My market share estimates have taken into account, where possible, the markets served by a given branch, and an allocation has been made of u that branch's volume to these markets. In order to see how these allocations were made, I am attaching my worksheets to this affidavit as Appendix A. My worksheets also report u SPI's market share estimates for 1997 and 1998. Comparing figures for 1997 and 1998, one will find some major differences in market share estimates for some SPI branches. u 8. I will now comment on the specific examples of share estimate problems raised by u Professor Carlton and Dr. Bamberger. · u However, for independents I report branch sales volumes, and these volumes do not necessarily correspond to the independents' market sales volumes as D explained above in paragraph 7. u
D
0 -4-u Even SPJ, in its 1998 share estimates, finds that GRG has a 20% share in St. Romuald and a 9% share in u Thetford Mine::~. u 9. u Thetford Gaz does not have an estimated 1998 share in u this market, although it did have an estimated 1997 share .. u u u u J0 . Next. Professor Carlton and Dr. Bamberger focus on Market 21. Here they ~"tate that I estimate an SPI/ICG share of 99% for this market However. because I do not report u sales for King Welding, they define my implied share for SPl/lCG as 100%. They further state that I do not explain why I assume that King Welding has a 1% share of this market. u In the absence of sales information for King Welding, I accepted SPrs 1997 market share estimate for King Welding of 1% . It should be kept in mind that my market share u estimates use 1997 data, and not the 1998 data that Professor Carlton and Dr. Bamberger use in their analysis. u 11. Professor Carlton and Dr. Bamberger also suggest that in some instances I ignore my own u methodology. As an example, they point to Market 33 (LaRonge). They State that I identify six competitive dealers, but that I claim that five of these branches are too far u away from LaRonge to offer any significant competition to SPI/ICG. They further state u u
D
0 -5-D that I do not explain the basis for this claim, even though these branches are captured in u my market polygon for LaRonge. D 12. In Market 33, I identify two possible competing dealers, Federated Coop and Stittco. Federated Coop's branch is located in LaRonge, while the worksheet for Market 33 u indicates that the five Stittco branches are located at distances of217, 321, 331 467, and 771 kms. from La Ronge (see Appendix A). Two of the Stittco branches (at Thompson u and Churchill) distribute propane through a grid system. The long distances between the Stittco branches and LaRonge not in the same market as LaRonge. Stittco does not ship u into the LaRonge area. The Stittco branches are captured by SPI' s LaRonge trade area polygon because there are no SPI locations north of LaRonge to bound the polygon. But u just because the polygon is not bounded in one direction does not imply that it will pay to ship propane hundreds of kilometres to some distant consumers. It should also be noted u that ICG does not identify Stittco as a competitor in LaRonge. u 13. The market share estimates contained in my Affidavit were obtained by using available u information on SPI, ICG, and other propane dealers' sales volumes, in conjunction with market share estimates made by ICG and SPI. For some propane dealers, sales volumes u were available at the branch level, while for others they were not. Branch sales volumes also had to be allocated to different markets if a branch served multiple markets. u 14. As an example, consider Market 40 (Yorkton). u The u reason for this discrepancy is that Federated Coop serves Yorkton, but it did not report a branch sales volume for Yorkton. Yorkton is presumably served by Federated's Regina u branch. Now SPI has a 1997 SPI/ICG estimated share of72% in Regina, Weyburn and Yorkton, while Federated Coop has a 28% share. u A market share u D
io -6-u estimate of 65% for SPI/ICG is obtained for Yorkton. This is the same market share estimate that I report for the Regina market. 0 u 15. As another example, consider Market 27 (Owen Sound). u Once again the discrepancy arises because Table B-1 assumes that branch sales volumes are market sales volumes. This was not my u assumption for independents. In fact, Sparling (Blyth) is found by my market analysis to be in Markets 27, 18 and 23, Hensall Coop is found to be in Markets 27, 18, and 23, UPI u (Barrie) is found to be in Markets 27 and 14, and UPI (Guelph) is found to be in Markets 27, 18, 19, and 23. SPI reports these independents competing in these same markets. u III. New Entry u 16. In this section of my Reply, I will comment on the table (Table 3) that has been u constructed by Professor Carlton and Dr. Bamberger to illustrate the extent of new entry into retail propane markets in Canada since 1990. The table conveys the impression that u some propane dealers have opened multiple branches and are operating them at low sales volumes. In fact, some propane dealers are serving multiple markets from a single branch u location. u 17 . Consider first paragraph 29 of the Carlton/Bamberger Affidavit. There they state that u Canwest has entered at several locations since 1990, and its annual sales appear to be less than two million litres at two of the locations. u
D u 18. In paragraph 29, Lo Cost is reported to have entered three of SPI' s branch areas since 1990, and its sales are reported to be less than two million litres at each location. u 0
io -7-D D 19. Turning now to Table 3 in Professor Carlton's and Dr. Bamberger's Affidavit, it should D first be noted that I only received a fax version of the Affidavit. It was extremely difficult u to read the numbers in Table 3, particularly when I tried to distinguish a 6 from an 8 or a 9. u 20. I might have been able to check the numbers that I could not read by examining the u source document for their numbers. However, the estimates of independents' sales volumes do not appear to come from a source in my possession. In particular, they do not u appear to be generated from SPI' s 1998 branch templates (in response to Undertaking 63). As an example, in the Concord branch template, there is no share of overall propane u business in Concord reported for Cam Carb. But in Table 3, Cam Carb has an estimated sales volume of 360,000. As another example, Praxair in Sudbury has a branch template u market share of2%. However, if one finds the total Sudbury market volume by dividing SPI's market volume by its market share, and then finds Praxair's estimated volume by D multiplying this number by 2%, one gets a volume of 439,677 and not 342,328 as reported in Table 3. D 21. Table 3 lists 103 entries from 1990 to 1998. Eighteen of these entries, however, were by u SPI or ICG agents. Twenty of the entries in the table were not, according to my spatial u market analysis, actually in the markets as indicated in the table. There are, in addition, a number of cases where branch entry can appear overstated because more than one market u is being served by the same branch. a 22. For example, Table 3 has three entries for Lo Cost, but it only has one branch located in Lethbridge. Lo Cost is not in Red Deer, Radium or Regina according to the spatial D market analysis. SPI reports small market shares for Lo Cost in these three markets in 1998 (e.g. 0.5% in Regina). u ,0
-8-23. Table 3 has two entries for Highlands, but it only has one branch located at Fenelon Falls. Two entries are reported for British Empire, but it only has one branch located at Bobcaygeon. Value Propane has two entries in Table 3, but only one branch located at Sunderland. There are three entries for Sparling in Table 3, but these all refer to the same branch. There are two entries for Edpro, but it only has one branch located in Concord. Sunrise has three entries in Table 3, but only one branch located in Toronto. Propane Expert has one branch located at Vars, but it has two entries in the table. Propane du Suroit has two entries in the table, but only one branch located at St. Martine.
24. By reporting a propane dealer's multiple entries at lower volumes, the minimum efficient scale for entry could appear lower than it really is. This would be particularly so if multiple markets are being served from a single branch. The branch might require the combined volumes from serving multiple markets in order to be viable.
25. As another example of how Table 3 contains misleading information, Table 3 shows five new entries into Sudbury (which all took place in 1990) by BOC, Brown, Liquid Air, Praxair, and Weldco. Collectively, these firms are estimated by SPI to have a 13% market share in 1998. These firms were not listed as market competitors in SPl's 1997 branch templates. Instead, SPI reported independents' market shares at 1% in 1997. Interestingly, all of these propane dealers are SPI agents.
IV. Market Share Data Files 26. Professor Carlton and Dr. Bamberger have provided us with a number of data files that were apparently used in the regression analysis. The file called SHARE98.DBF contains SPI and ICG shares and the total share for independents, by branch. The shares appearing in this data file do not correspond to the market shares that were reported in the 1998 branch templates supplied by SPI. For example, for Bancroft the data file reports an SPI share of 44%, an ICG share of 41 %, and a share for independents of 15%. The SPI
-9-template reports an SPI share of 40%, an ICG share of 35%, and a share for independents of 25%. As another example, for Kelowna, the data file reports an SPI share of 28%, an ICG share of 10%, and a share for independents of 62%. The SPI 1998 template for Kelowna reports an SPI share of25%, an ICG share of 33%, and a share for independents of42%.
27. It is also unclear where the figures in the column labelled TVOL98 come from. They are not market volumes that were derived from the templates. For example, on the basis of the template for Bancroft, one would calculate a market volume of 11,787,172 litres. The data file SHARE98.DBF reports a market volume for Bancroft of 10,755,140.
28. In the data file called WESTSHAR.DBF, volumes are reported for SPI, ICG and independents by market. The total volume for a given market will be the same as the total volume that appears in SHARE98._DBF, provided the market definition is the same. Once again, however, it appears that the volume figures did not come from the SPI 1998 templates. I cannot test the accuracy of these numbers without knowing their source.
V. Conclusion 29. In this Reply Affidavit, I have focused on three aspects of the Rebuttal Affidavit of Dennis Carlton and Gustavo Bamberger. First, I have explained why my market share estimates are reliable, based as they are on market share estimates from SPI and ICG, and on volume sales data provided by SPI, ICG and a large number of independents. Second, I have discussed why the new entry table appearing in Professor Carlton's and Dr. Bamberger's affidavit overstates the extent of new entry into retail propane markets and can understate the minimum efficient scale needed to enter and remain viable in a retail propane market. I have also highlighted problems that I fmd with the table, including the unknown source for the independents' sales volumes and the lack of identification of agents. Finally, I also examine two of the data files that were apparently used to support
LI -10-D the regression analysis of Professor Carlton and Dr. Bamberger. Once again, the source D for volume sales data in these data files could not be identified. LI D D D 0 D LI 0 D D 0 D D u u
D