0 0 File No. CT-98/2 io 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 SOR/94-290, as amended (the "Rules"); D AND IN THE MATTER OF an inquiry pursuant to subsection 10( 1) (b) of the Competition Act relating to the proposed 0 acquisition of ICG Propane Inc. by Superior Propane Inc. AND IN THE MATTER OF an Application by the D Commissioner of Competition for an order pursuant to section 92 of the Competition Act C()MPET!TiON TRIBUNAL TRIBUNAL De LA CONCURRENCE D F p I OCT 29 1999 ~· ~ BETWEEN: D & REGISTRAR • P.EGISTRAlRE f 0 THE COMMISSIONER OF COMPEIR+:u:wi11--<.---r-r~-.-1 0 - and -D SUPERIOR PROPANE INC. and ICG PROPANE INC. D Respondents 0 REBUTTAL AFFIDAVIT OF DOUGLAS WEST D I, Douglas West, of the City of Ottawa, in the Province of Ontario, MAKE OATH D AND SAY THAT: 1. I was retained by the Commissioner of Competition to provide expert economic D evidence in this matter, to review the expert reports submitted by the expert a witnesses of the Respondents, and to provide my comments with respect to those reports. D C:\TEMP\0198-affidavit.doc D
0 -.2 -D 2. I have reviewed the expert witness affidavit of Dennis Carlton filed 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 and opinions expressed therein are true to the best of my knowledge, information and belief.
3. Attached hereto and marked as Exhibit "B" is a true copy of my curriculum vitae. 4. I make this affidavit pursuant to Rule 47(2) of the Competition Tribunal Rules. 0 SWORN BEFORE_ ME) ) 0 AT ;/wl/ ( Q'-lr;.hec:.. ) J)'-' THIS f:L DAY OF September 1999 ) c:=>¥ G: 2=£;;-0 Douglas West u -BARR+&fER-AN6-S0l:ICIT0R-· ~~./.-LA.r?-U:d.,, /c'-z.., C a;i,/~ D 0 a u a
D a
D D C:ITEMP\0198-affidavit.doc 0
D 0 u 0 0 D D 0 This is Exhibit "A" to the Affidavit of Douglas West sworn/affirmed D before me at the city of Hull in the Province of Quebec this 14th day of September, 1999.
0 u Commissioner of Oaths for Canada, etc. 0 u a D a
D
0
D I. Introduction 0 1. I have reviewed the evidence of Dennis W. Carlton. I have found that there are significant errors or discrepancies in the preparation of data sets, particularly with respect u to the market share calculations and compilation of branch lists. These errors are significant enough to invalidate the market share comparisons in paragraph 26 of his 0 report, and may affect his regression results. 2. Professor Carlton has also inappropriately characterized the possibility of entry by U.S.-D based propane dealers into a Canadian retail propane market. Such entry has not occurred, and would face the same sort of obstacles as entry and expansion by Canadian D firms. 3. Professor Carlton has constructed a table (Table 5) to help illustrate the extent of new D entry into retail propane markets in Canada in the last three years. This table is reproduced by me, with Superior's (SPI's) estimates of these new entrants'market shares. D This table shows that the majority of the new entrants have only been able to acquire a D one or two percent share of sales in the branch trade area that they enter. Several of the larger market shares estimated for new entrants by SPI are incorrect. 0 4. Professor Carlton may also have compromised his analysis by using 1998 share estimates in his 1997 regressions. SPI has made market share estimates for 1997, and as I show in 0 a table, these estimates differ, sometimes quite substantially, from the 1998 estimates. 5. The next section of my affidavit will critique Professor Carlton's discussion of possible u entry by U.S. retailers. Subsequent sections will focus on problems in Professor Carlton's branch lists and market share data, his description of new entry into Canadian u retail propane markets, and his use of 1998 market share estimates in 1997 regressions. u u
D u 1 D
D D II. U.S. Propane Retailers D 6. Professor Carlton, on page 8 of his Affidavit, has suggested that major U.S. retailers D could expand into Canada if prices in Canada increased relative to U.S. prices. He offers as support for this possibility (1) the availability of propane in Canada at the wholesale D level, (2) the proximity of U.S. retailer locations to the U.S.-Canada border, and (3) the similarity of regulatory requirements in Canada and the U.S.
D 7. Even though the top 10 propane retailers in the U.S. reportedly (by Carlton) have 1500 retail locations in states that border Canada, none of them has a single retail outlet in D Canada. Professor Carlton, in his Table 5, shows that some Canadian retail propane markets have supported new entry in the past three years. If entry by U.S.-based retailers D is easy, why would they not have taken advantage of such entry opportunities? These u large American propane retailers certainly have the expertise to expand into a Canadian market if entry is easy. The fact that th~y have not suggests that there are some entry u barriers that Professor Carlton has not discussed. 8. There are several ways in which an American propane retailer might attempt entry into a u Canadian retail propane market. First, the American retailer could attempt to shi.P. propane into Canada by tanker truck from one of its U.S.-based branches. However, an a American-based retailer would face the same sorts of constraining factors on the ability to serve more distant customers as a Canadian-based retailer. These were listed in my a earlier Affidavit as (i) the distance of the branch to the different points in the candidate market, (ii) the storage capacity of the branch, (iii) the number of customers available to u the branch that are not locked into a supplier through contracts, (iv) the propane dealer's reputation in the area of the candidate market and the costs to overcome any reputational u barriers, (v) the density of customers in different areas of the candidate market and the volumes purchased, (vi) natural barriers to transport such as mountains and rivers, and a (vii) the existing road network. a a 2 D
D 0 9. In addition, an American-based propane retailer would have the additional costs D associated with crossing the U.S.-Canada border. These costs could include ( 1) additional money and time costs because of delays at the border crossing caused by u traffic congestion and the need to clear customs,(2) any costs resulting from the use of different measurement regimes in Canada and the U.S., (3) costs associated with D differences in tax, regulatory, and licensing regimes between Canadian jurisdictions and the U.S., (4) costs associated with dealing with different currencies and fluctuations in D exchange rates, and (5) costs associated with dealing with French language requirements in Quebec. These additional costs could make it unprofitable for a U.S.-based retailer to a serve a Canadian retail propane market even if the market can support profitable Canadian-based retail expansion. a 10. I am aware of only one case of a U.S.-based firm supplying a Canadian retail propane market from a branch located in the U.S. This case involves a company called Lake Gas, 0 which supplies propane to Fort Frances.from a branch in International Falls, Minnesota. a SPI has estimated its 1998 market share in Fort Frances as 2%. 11. The second way that an American retailer could attempt to serve a Canadian retail a propane market is by establishing a branch in a Canadian retail propane market. ~ However, the majority of new entrants in Canadian retail propane markets in the last a three years have only managed to acquire a one or two percent share of the market that they have entered. (See Table 2 in Section IV below.) Furthermore, I am unaware of any D cases where a U.S.-based retailer has opened up branches in Canada. Even in markets where entry has been possible, U.S.-based retailers have not responded to the incentive to D enter. 12. The third way in which an American-based firm could enter a Canadian retail propane D market is by buying an existing firm. Entry or expansion by acquisition has historically e been the quickest and most common means of entry and expansion in the propane industry in the face of entry impediments. E D :o
D a III. Branch Lists and Market Share Data u 13. In paragraph 26 of his report, Professor Carlton carries out some market share comparisons using 1998 data supplied by SPI. According to Appendix B to his report, a these data were obtained from templates that were distributed to branch managers in the spring of 1999. Branch numbers were assigned to the data based on location names in 0 accordance with SPI documents. Some satellites were combined with their respective branches, and Attachment 5 to his report contains the program that carries out this u aggregation. The templates filled out by branch managers that report 1998 sales and market shares have been supplied by SPI to the Commissioner in response to Undertaking u 63. 14. My initial reaction to reading paragraph 26 was to attempt to replicate Professor Carlton's a results. The first step in that replication would be to identify the 64 Superior branches that Professor Carlton refers to in paragraph 26. While Professor Carlton refers to these a 64 branches, he never identifies the brap.ches by name. I first looked at Attachment 5, a which has a list of branches and satellites, some of which are being combined. Attachment 5 contains a list of 63 branch names, but it does not contain the names of the D other 17 SPI branches for which SPI provided branch template information. To ~erify that this was the list that was being used for the comparisons in paragraph 26 as well as a Professor Carlton's regression analysis, I compared the post-aggregation branch list (which should have 50 branches in it) with those appearing in two data files used for the a regression analysis. These two files, regdatab.dbf and regdat7b.dbf, were supplied by Professor Carlton on a disk. a 15. Data file regdat7b.dbf contains the data used in Professor Carlton's 1997 regressions. There are 57 numbered branches in this data file (and not the expected 50). Data file D regdatab.dbf contains the data used in Professor Carlton's 1998 regressions. There are 65 numbered branches in this data file. There are 10 branch numbers listed in the 1998 Q regression data that are not listed in the 1997 data as follows: 15- Smithville D 22- Pembroke I 4 D
0 a 444- Grande Prairie 0 807- Chetwynd 831- Watson Lake u 16- Simcoe 14- Strathroy u 467- Fort McMurray 808- Terrace u 17 - Walkerton. None of the above branches appear to have been new branches in 1998. u 16. In addition, the 1997 regression data file contains two branch numbers, 345- Trenton and 348- Truro, that are not referred to in the 1998 regression data file. These branches were D not on a list of closed branches provided by Superior in response to undertaking 95. a 17. In an effort to resolve these discrepancies in branch lists, I examined the other attachments in Professor Carlton's report, hoping to find the definitive branch list. u Instead, in Attachment 10, which contains a program that sorts data into rural and urban branches, I found branch numbers that were not listed in Attachment 5. These branch 0 numbers are 6, 33, 240, 246, 421, 427, 445, 456, 557, 704, 712. Of these, 33, 24Q, and 456 do not appear in either the 1997 or 1998 regression data files. a 18. In Attachment 5, as well as in the regdatab.dbf file, Professor Carlton assigns some branch numbers to Superior branch names which differ from the branch numbers a assigned by Superior. It is not clear whether Professor Carlton intended to use the data attached to the branch number or to the branch name that has been incorrectly identified a with the branch number. For example, in Attachment 5, branch number 420 is associated with Maidstone when in fact Superior has it assigned to Macklin. In Attachment 5, a branch number 431 is assigned to Fort Nelson when in fact Superior has it assigned to Fort St. John. Further errors of this type are encountered with respect to branches that I Professor Carlton numbers 343, 344, 346, 348 and 350. He calls these branches Dartmouth, Bridgewater, Kentville, Truro, and Sydney, respectively, but Superior assigns D those branch names the numbers 344, 345, 347, 350, and 343, respectively. D 5 0
0 a 19. Because of the discrepancies between branch lists, I am not sure which 64 branches were a selected by Professor Carlton for the market share comparisons that are described in paragraph 26. I am also uncertain as to which branches and satellites have been u aggregated for the regression analysis that Professor Carlton has conducted. This makes it difficult, if not impossible, to replicate Professor Carlton's results by building a a regression data file from the data supplied by SPI. Replication of results by starting with primary data is a useful way of checking to see if data entry errors or data manipulation a errors have been made in building data files. 20. The differing branch lists is not the only problem that is encountered in the process of a trying to replicate Professor Carlton's results. The market share variables that are used by Professor Carlton in his regression analysis are also problematic. These variables are u called SPISHARE, INDSHARE, and ICGSHARE in both regdatab.dbf and regdat7b.dbf. They are used to define when either an ICG share or the independents' share in a a particular market exceeds a certain defined threshold, like 15, 20, 25 or 30 percent. They are also presumably used for the market share comparisons discussed in paragraph 26. E They also enter directly as explanatory variables in some of Professor Carlton's a regressions (see his footnote 38). The problem is that the market share figures in.these variables, which Professor Carlton states come from the branch templates (see page 2 of D Appendix B of his Affidavit), do not correspond to the market shares that SPI has estimated for itself, ICG and its other competitors and reported in its 1998 branch D templates. 21. I have done a branch by branch comparison of Professor Carlton's market share estimates a and those that are reported by SPI in its branch templates. See Table 1. While none of the non-zero numbers match up, some of the share estimates used by Professor Carlton a are not even close to those reported by SPI, and some can lead to unreasonable conclusions. Note that when there are two entries in the table for a given branch number, u the first entry is for the branch name that SPI attaches to the branch number (except for the discrepancies mentioned in paragraph 18 above), while the second entry. is for the I branch that is aggregated with the first entry in Professor Carlton's Attachment 5. 6 6 0
lo D Table 1 0 Comparison of SPl's and Carlton's Market Share Estimates: 1998 SPl's Share Estimates Carlton's Share Estimates D Branch Branch SPI's ICG's Independents' SPI's ICG's Independents' # Name Share Share Share Share Share Share D 1 Concord 35 30 35 32.96 21.53 46.50 2 Stratford 45 12 43 37.79 13.66 48.55 4 Belleville 60 25 15 56.72 21.40 22.88 0 5 Fenelon 25 50 32.04 31.85 36.11 Falls 5 Bancroft 40 35 25 0 6 Barrie 45 30 25 24.66 22.51 52.83 7 Guelph 40 15 45 24.47 22.54 52.99 D 9 Huntsville 42 2 56 62.88 4.89 32.22 9 Bala 70 8 22 14 Strathroy 45 12 43 39.13 15.37 45.51 D 15 Smithville 48 4 48 99.97 0.01 0.02 16 Simcoe 60 5 35 52.67 5.04 42.29 17 Walkerton 60 5 35 45.59 4.90 49.51 u 20 Whitby 30 25 45 32.95 28.42 38.62 22 Pembroke 58 38 4 71.49 27.52 1.00 23 Ottawa 40 31 29 44.07 22.53 33.40 u 30 Porcupine 27 48 25 36.23 41.38 • 22.39 30 Hearst 75 10 15 u 39 Chatham 28 4 68 48.83 3.23 47.94 41 North Bay 74.1 24.6 1.3 67.73 28.92 3.35 41 New 74.2 22.5 a Liskeard 42 Sudbury 45 40 15 39.52 51.09 9.39 91 Thunder Bay 45 55 0 44.69 55.22 0.09 a 91 Dryden 60 40 0 91 Fort Frances 75 23 2 100 Winnipeg 30 37 33 27.45 39.21 u 100 Virden 25 40 35 202 St. Constant 25 30 45 25.53 31.33 43.34 203 Drummond- 16.35 38.02 45.63 21.66 23.13 55.21 D ville ')..,.., -.J.J Gatineau 36 30 34 46.78 30.66 22.56 235 Joliettte 22 16 62 25.29 25.16 49.55 E ..,.., 241 Vimont 15 . -)_ ; .J .J 37.04 _ r. ) .) -_; 37.43 I 7 D
D D SPl's Share Estimates Carlton's Share Estimates D Branch Branch SPI's ICG's Independents' SPI's ICG's Independents' # Name Share Share Share Share Share Share 243 St. Romuald 27 33 40 u 27.66 44.04 28.31 245 Thetford 30 35 35 24.81 32.45 42.74 Mines .., 246 Sept Iles 21 76 ,j 21.22 75.19 3.59 D 248 Les Cedres 40 10 50 94.00 0.84 5.16 250 Riviere du 33 60 7 24.28 49.60 26.12 0 Loup 271 Cap-de-la 30 45 r_ ) 36.17 46.13 17.70 Madeleine u 343 Dartmouth 49 0 51 48.24 0.00 51.76 344 Bridgewater 45 1 54 43.69 0.70 55.61 345 Trenton 65 0 35 0.15 0.00 99.85 D 346 Kentville 46 2 52 38.33 0.34 61.32 348 Truro 38 2 60 0.06 5.75 94.19 350 a Sydney 45 1 54 36.65 0.15 63.20 401 Regina 32 38 30 32.03 ..., .... . . .,.., 34.64 .).) .).) 401 Swift 30 26 44 a Current 406 Saskatoon 30 30 40 36.35 11.69 51.97 407 Prince 20 20 60 27.67 32.02 40.31 a Albert ..,.., ~ 413 Kelowna 25 ,j ,j 42 28.10 9.75 62.15 414 Kamloops 40 32 28 40.75 29.44 29.81 u 420 Maidstone 30 30 40 26.71 34.35 38.95 420 Elk Point 27 43 30 421 Campbell 42 46 12 47.94 40.07 12.00 a River 423 Edson 27 27 46 37.21 34.34 28.45 a 423 Valemont 40 60 0 4"_",j' Whitecourt 20 80 0 427 Bums Lake 48 48 4 54.51 43.56 1.93 a 431 Fort St. John 30 30 40 28.35 48.83 22.82 431 Whitehorse 20 80 0 443 Slave Lake 30 32 38 30.67 33.99 35.35 a D D 8 0
0 D SPl's Share Estimates Branch Branch SPI's ICG's Independents' 0 # Name Share Share Share 444 Grand Prairie 3 9 34 27 D 445 Prince George 3 6 39 25 459 Red Deer 29 27 44 a 467 Fort 49 49 2 McMurray 550 Charlotte- 34 0 66 D Town 557 Moncton 30 5 65 701 Leth bridge 40 38 22 D 704 Calgary 32 30 38 707 Cranbrook 38 38 24 707 Radium na na na 0 709 Edmonton 29 22 49 712 Peace River 46 26 28 801 Coquitlam 34 31 35 D 802 Nanaimo 30 25 45 807 Chetwynd 30 49 21 a 808 Terrace 60 38 2 831 Yukon 20 80 0 841 St. John's 52 0 48 a 843 Bathurst 38 2 60 845 Pasadena 54 0 46 D a a
D D E I 9 0
Carlton's Share Estimates SPI's ICG's Independents' Share Share Share 23.13 32.92 43.94 43.11 36.26 20.63 19.72 14.69 65.58 48.77 51.23 0.00 32.55 0.00 67.45 39.56 2.93 57.51 42.90 41.69 15.42 39.60 34.26 26.15 46.08 38.89 15.04 32.27 28.24 39.49 99.90 0.05 0.05 33.30 30.63 36.06 30.72 28.20 41.08 28.57 48.82 22.61 71.14 25.79 3.07 13.33 86.67 0.00 99.73 0.00 0.27 37.16 1.22 ~ 61.62 38.24 0.00 .61.76
0
D 22. For example, for branch 345, Trenton, Professor Carlton reports an SPI share of0.15%, an independent share of 99.85% and a zero share for ICG. D 0 SPI's 1998 share estimates for Trenton are a 65% share for itself and a 35% share for D Irving. These share estimates, while different from the actual shares based on sales volumes, are much closer to the actual shares than Professor Carlton's numbers. D For a second example, Professor Carlton reports SPI, independent, and ICG shares for branch 348 (Truro) as 0.06, 94.19, and 5.75%, respectively. SPI's 1998 share estimates
D for Truro are a 38% share for itself, a 50% share for Irving and a 2% share for ICG. D ::=ven SPI's share estimates, which give it a smaller share than Irving, u are not consistent with the actual sales volumes. 24. For a third example, for branch 712, Peace River, Professor Carlton reports an SPI share a of 99.90%, an ICG share of .05%, and a .05% share for independents. a a It is also of interest to note that SPI' s 1998 market share estimates for itself and ICG in Peace River are inconsistent with the sales D figures above. SPI estimates that it has a 46% share of Peace River and High Level, while ICG is estimated to have a 26% share. a _r. )_ For a fourth example of where Professor Carlton's market share figures are way out of a line with actual sales volumes, one can look at St. John's (branch 841 ), where Professor Carlton has reported a 99.73% share estimate for SPI and a 0.27% share estimate for D Irving. D D 10 0
D 0 26. The market shares contained in the market share variables would appear to be in error. I do not know where the market share figures contained in regdatab.dbf and regdat7b.dbf D come from. a 27. Besides the fact that Carlton's market share figures are not the same as those reported by SPI on its branch templates, there is also the problem that SPI's market share estimates 0 are not always accurate. 28. There are two types of market share estimate inaccuracies that can be examined. First, D one can look at SPI's estimates of its market share and ICG's market share in specific branch trade areas. It can be difficult to assess the accuracy of these estimates in some a branch trade areas because it is not always clear over what area SPI is estimating a market share, and ICG does not always have a branch in the same city as SPI. One can, however, D look at estimates of SPI and ICG market shares in areas where SPI and ICG have branches close to one another, and where they do not have other branches nearby. B 29. For example, for Thunder Bay, SPI estimates that it had a 45% share in 1998, while ICG had a 55% share. . a u 30. For Saskatoon, SPI has estimated that it and ICG had equal market shares in 199&. a 31. For Swift Current, SPI has estimated that its 1998 share (30%) is larger than ICG's (26%). a 32. For Fort McMurray, SPI has estimated that it and ICG had equal market shares in 1998. D 33. For Kelov,,'Ila, SPI has estimated that SPI had a 25% share in 1998, while ICG had a 33% share.
D 34. In Prince George, SPI has estimated that SPI had a 36% share in 1998, while ICG had a 39% share. D D D 11 ,0
D u 3 5. The second type of market share estimate inaccuracy can arise from inaccurate estimates u of other competitors' market shares. For example, SPI has estimated that Joy's 1998 market share in Fort St. John is 35%. D As another example, SPI estimates its 1998 Moncton share at 30%, ICG's share at 5% and Irving's share at 65%. u Not only is ICG's estimated u share incorrect, but it is inconceivable that Irving would have increased its sales volume from 1997 to 1998 to such an extent as to give it a 65% share of the market. u 36. As a third example, SPI has estimated that Kelly's 1998 market share in Bancroft is 15%. while Algonquin's market share is estimated at 10 percent. 0 u u u IV. New Entry u 37. In paragraph 27, Professor Carlton describes the extent of new entry into Canadian retail propane markets over the past three years. Table 5 of his report lists the names of new u entrants by SPI branch trade area for each year 1996, 1997, and 1998. The purpose of this description is apparently to indicate the ease of entry over the last three years. u However, there is no indication in the table of the extent to which these entrants were successful at capturing large shares of the markets that they entered. For this reason, I u have reproduced Professor Carlton's Table 5 below, but I have also included the 1998 market share for each entrant that has been estimated by SPI. (There are five entrants for E which market share estimates are not available.) ~ ~ 12 D
D u Table 2 D Entrants Making Sales in Competition with Superior and Superior's Estimates of Their 1998 Market Shares: 1996-1998 u Superior Branch D Year Number Superior Location u 1996 1 Concord 5 Fenelon Falls E 9 Huntsville 22 Pembroke 39 Chatham 0 100 Winnipeg 343 Dartmouth D 421 Campbell River ~ 459 Red Deer/Rocky Mountain House 704 Calgary E u 707 Radium ' 1997 1 Concord ~ . E 5 Bancroft 15 Smithville 20 Whitby I ' 22 Pembroke E 23 Ottawa 41 North Bay 202 St. Constant ti . 2.") ".) Gatineau 248 Les Cedres E . 13 0
Market Name of Entrant Share(%) Chemw eld 0.5 CSE 1.0 Sparling Propane 10.0 Leggett 5.0 MacDonald's 1.0 Edp ro 1.0 Aeshu 2.0 BOC Gases <1.0 Praxair <1.0 Columbia Fuels 6.0 Super Save 2.0 Teskey 1.0 Bayne 3.0 Keebee 1.0 ~ Sure Fire 4.0 Lo Cost Edp ro 2.0 SMS Sunrise 5.0 Welder's Source <1.0 Kelly's Fuels 15.0 Edp ro 1.0 Sparling 5.0 Sunrise 5.0 Value 10.0 Budget 3.0 Expert 2.0 Budget <1.0 Propane du Suroit 10.0 Propane Expert 1.0 Propane du Suroit 10.0
0 0 Superior 0 Branch Year Number Superior Location D 250 Riviere du Loup 343 Dartmouth 704 Calgary 0 709 Edmonton/Camrose/ Westlock 0 1998 5 Bancroft 6 Barrie 0 u 23 Ottawa 91 Fort Frances 246 Sept Iles u 346 Kentville 350 Sydney D 401 Regina, Weyburn, Yorkton 406 Saskatoon D 423 Edson 431 F Ort St. John 443 Slave Lake/Red Earth D 444 Grand Prairie 459 Red Deer/Rocky Mountain D House 707 Radium D 802 NanimoNictoria 807 Chetwynd D 841 St. John's n
D D 14 D
Market Name of Entrant Share(%) Milliard 1.0 Action Propane <1.0 Cowboy Propane <0.5 Bo-Li 1.0 Hughes 1.0 Algonquin 10.0 Northmore Budget Propane 3.0 Coop 7.0 Sparling 10.0 Sunrise 2.0 Lake Gas 2.0 Soude R 1.0 Irving 50.0 Thrifty 2.0 Scotia Propane 3.0 Lo Cost 0.5 Canw est 5.0 Neufeld Canw est 2.0 Cal-Gas 1.0 Neufeld 17.0 Canw est 2.0 Cal-Gas 4.0 Keeb ee 1.0 Rex Propane 1.0 Gas Plus Columbia Fuels 25.0 Neufeld Propane 1.0 North Atlantic 8.0
0 u 38. Of the 56 entries for which there are market share data, 32 of the market share estimates u are 2% or less. Twenty of the entries have market shares greater than 2% but less than or equal to 10%. One might ask whether all of these new entrants will be profitable in the 0 long run and whether they will be effective competitors and continue serving the market given the market shares that they have been able to acquire so far. Only four of the entries U· are greater than 10%. Of these, the market share estimate for Kelly's Fuels in Bancroft is an overestimate (see paragraph 36 above). The estimated market share for Neufeld in u Slave Lake (17%) is too high. u Columbia Fuels, which has an estimated NanaimoNictoria market share of 25%, is an ICG bulk dealer. (Indeed, several other firms on the list, including u Chemweld, BOC, Praxair, and Welders Source, are SPI agents.) Irving in Kentville had postive volume in 1997, so it is not a new entrant in 1998. There is also evidence that D Budget, Co-op and Sparling were serving Barrie prior to 1998, that Canwest was serving Saskatoon prior to 1998, that Cal-Gas was serving Red Deer prior to 1998, and that D Neufeld was an established propane dealer prior to 1998. u 39. Also with respect to the new entrant list, it should be noted that on the basis of th~ spatial market analysis in my earlier affidavit, not all of the entrants are in the markets as listed u in the table. For example, there are three entries for Edpro on the list. But Edpro only has one branch located in London, and it is not in two of the markets for which entries appear a in the table. There are three entries for Budget Propane. But it only has one branch in Aurora, and it is not in two of the markets for which entries appear in the table. Lo Cost D has two entries in the table, but its one branch located in Lethbridge is not in the markets as listed in the table. Sunrise has three entries in the table, but only one branch in
·~ Toronto. Sunrise is not in one of the markets as listed in the table. Sparling has three entries in the table, but it is not in one of the markets as listed in the table. D ~ u 15 0
0 0 v. Professor Carlton's Regressions and Market Share Data u 40. Professor Carlton, on p. 14 of his report, states that he uses 1998 market share estimates D for both the 1997 and 1998 regression analysis. He assumes that shares by branch did not change substantially between 1997 and 1998. He presumably assumes this because if D 1997 and 1998 shares are substantially different, using 1998 shares in a 1997 regression could cause an errors in variables problem . The coefficient estimate would then be 0 biased towards zero. The assumption regarding 1997 and 1998 shares could have been a tested given that SPI did provide 1997 market share estimates for a number of its branches. (See the SPI document, dated September 30, 1998, Bates Nos. 186286-186354.) Professor Carlton could also have used the 1997 market share estimates directly 0 in his regressions. 41. Table 3 below provides a list of branches for which SPI has estimated both 1997 and D 1998 market shares for SPI/ICG. The market share estimates are also provided. D
D a
D D a
D D a 16 0
0 D Table 3 SPI /ICG Branch Trade Area Shares: 1997 and 1998 0 SPI/ICG 1997 SPI/ICG 1998 u Branch Branch Area Share Branch Area Share Moncton 58 35 u Riviere du Loup 100 93 Cap-de-la-Madeleine 49 75 Drummondville 75 54.37 0 Joliette 38 38 Les Cedres 48 50 Vimont 48 u 48 St. Romuald 71 60 Sept Iles 82 97 Thetford Mines 46 65 D Barrie 76 75 Pembroke 97 96 Kingston 82 85 D Ottawa 73 71 Strathroy 63 57 u Concord 79 65 Whitby 79 55 Dryden 100 100 D Fort F ranees 100 98 .. Peterborough 79 45 Sudbury 99 85 D North Bay 98 98.7 Porcupine 90 75 Thunder Bay 100 100 D Virden 70 65 Winnipeg 72 67 D Maidstone 73 60 Prince Albert 40 40 Regina/Weyburn/Yo rkton 72 68 D Saskatoon 65 60 Swift Current 59 56 Calgary 61 62 D Edmonton 63 51 Fort McMurray 99 98 Grand Prairie a 88 73 Lethbridge 58 78 Peace River/High Level 86 72 D 17 0 ......
0 D SPI/ICG 1997 SPI/ICG 1998 u Branch Branch Area Share Branch Area Share Slave Lake 69 62 Medicine Hat 0 82 78 Red Deer 61 56 Red Earth Creek 74 62 0 Kami oops 72 72 Burns Lake 98 96 Terrace 87 98 u Campbell River 92 88 Castleg ar 100 70 Chetwynd 81 79 0 Fort St. John 70 60 Coquitlam 72 65 Cranb rook 80 76 0 Kelowna 54 58 N anaimoNictoria 78 55 Prince George 87 75 0 Watson Lake 100 100 Whitehorse 100 100 D
0 D 0 0 0 u
0 a 18 0
0
D 42. Clearly, there are significant differences in SPI's estimated market shares for some of the 0 branches contained in Table 3. For example, the estimated SPI/ICG share in Moncton is 58% in 1997 and 35% in 1998, the estimated SPI/ICG share in \Vhitby is 79% in 1997 u and 55% in 1998, the estimated SPI/ICG share in Peterborough is 79% in 1997 and 45% in 1998, and the estimated SPI/ICG share in Grand Prairie is 88% in 1997 and 73% in 0 1998. Other examples of major differences in market share estimates can also be found in Table 3. As discussed earlier, 1998 estimates should not be used in 1997 regressions. 0 VI. Conclusion 0 43. In this affidavit, I have highlighted some data problems that have appeared in Professor Carlton's analysis, particularly those involving market share estimates and compilations 0 of branch lists. These data problems make it difficult to replicate Professor Carlton's results and to understand how the data sets used in his regression analysis were derived. D In addition, I have found that Professor. Carlton has inappropriately characterized the D possibility of entry by U.S.-based propane dealers into the Canadian retail propane market. Such entry has not occurred and would face the same sort of obstacles as entry u and expansion by Canadian firms. 44. Professor Carlton has constructed a table to illustrate the extent of new entry into retail 0 propane markets in Canada in the last three years. He does not draw any conclusions from the information in the table, and given the absence of volume and share 0 information, it is not clear that any conclusions can be drawn. I have reproduced his new entry table with market shares included. These data reveal that most new entrants have D only acquired relatively small market shares since they entered the market. 45. Finally, Professor Carlton may have compromised his analysis by using 1998 share D estimates in his 1997 regressions. SPI' s market share estimates for 1997 and 1998 are u not the same, and some of the estimates for a given market are substantially different. D D 19 0