THE COMPETITION TRIBUN,.tf/';~ c'':J'ltC£tf. ~,_, >./·;t( J(
IN THE MATTER OF an Application by the Director of Investigation and Research under sections 92 and 105 of the Competition Act, R.s.c. 1985, c.C-34, as amended:
AND IN THE MATTER OF Imperial Oil Limited
Texaco Canada Inc.
B E T W E E N THE IMPERIAL OIL LIMITED AFFIDAVIT OF DONALD G. McFETRIDGE I, Donald G. McFetridge, of the City of Ottawa, in the Province of Ontario in Canada MAKE OATH AND SAY AS FOLLOWS:
1. I am a Professor of Economics at Carleton University and have been retained by the Director of Investigation and Research, Consumer and Corporate Affairs -Canada, to proviae my opinion on the implications of
CT 89 I 5
Applicant - and -Respondent
- 2 competition from imported motor gasoline from the merger of Imperial Oil Limited and Texaco Canada Inc. Now shown to me and attached as Exhibit "A" to this my affidavit is a copy of my Report.
2. The contents of this Report attached as Exhibit "A" to this my affidavit and the opinions expressed therein are true to the best of my knowledge, information and belief.
3. I make this affidavit pursuant to Rule 42(1) of the Competition Tribunal Rules.
A Commissioner o~~- lrn, and for the Province of Onl-::trio
This is EXHIBIT .A. to the Affidavit of Donald G. McFetridge sworn before me on July 19, 1989
A Commissioner of Oat in an for the Province of Ontario
ASSESSMENT OF THE IMPACT OF POTENTIAL IMPORT COMPETITION ON WHOLESALE GASOLINE PRICING IN ONTARIO AND QUEBEC
D.G. Mcfetridge
SUMMARY The purpose of this report is to assess the extent of potential import competition in the market for motor gasoline in Ontario and Quebec.
I have examined two types of evidence. The first is evidence on the capacity of independent (i.e. non-refiner) marketers to import gasoline. The second is evidence on the effect of import prices on wholesale gasoline prices in Ontario and Quebec.
With respect to the capacity of independents to import, the evidence is that there are large, well-established independents operating in each province with the capacity to import marine cargoes. Ontario independents also have the capacity to import by tanker truck from Buffalo, New York.
Current imports by independents account for about 3.5 per ce~t of Quebec's net supply and 1.5 per cent of Ontario's net supply. Independents have the physical capacity to import about 50 per cent of Quebec's requirements, almost 20 per cent of Ontario's, and about 30 per cent of the combined requirements of the two provinces.
- 2 -They would not require a substantial increase in the wholesale price of gasoline in Quebec or Ontario relative to the off-shore price or United States price to induce them to increase their import volume.
The conclusion which follows from the examination of the import capacity of independents is that at present they have the ability to constrain attempts by domestic refiners to reduce supply and increase wholesale prices.
With respect to the relationship between wholesale gasoline prices in Toronto and Montreal and prices prevailing in the United States, a statistical examination of the evidence reveals that:
(a) The Buffalo wholesale price of gasoline landed in Toronto constitutes an effective ceiling on the Toronto wholesale price. The Toronto price does not adjust instantaneously to changes in the Buffalo price (or the exchange rate) but over a full cycle the average price in Toronto does not differ from the Buffalo price landed in Toronto.
(b) The Montreal wholesale price of gasoline responds to changes in the U.S. Gulf Coast price (landed in Montreal) but the response lag is relatively
- 3 -long. Over a full cycle, however, the average wholesale price in Montreal does not differ statistically from the u.s. Gulf Coast price landed in Montreal.
The price evidence is consistent with a perception by domestic refiners that the elasticity of supply of imports is high and that any sustained excess of domestic over foreign wholesale prices would result in a significant loss of volume. Given the ceiling which the price of imported motor gasoline places on Ontario and Quebec wholesale prices, it is unlikely that the merger could lead to and sustain a substantial increase in wholesale gasoline prices in Ontario and Quebec.
1 I. in---t-r·-o---d---u---c-t·-i-o--n-- The purpose of this memo is Lo assess the extent to which the pricing behaviour of Ontario and Quebec gasoline refiners is constrained by imports or the threat of imports. Specifically, the question is whether an attempt by domestic refiners to restrict their output and thus to raise domestic gasoline prices would be thwarted within a reasonable period of time by an increase in imports. A summary characterization of' import supply conditions 1s the import supply function. TLis ir;dicates the quantity of imports which would be supplied to the domestic market at any given domestic price. The estimation 0f an import supply function for the product, re~ion and time period in question is not possible (motor qasoline imports into 0 n tar i o are av a i 1ab1 c on l y on an an r; u a i bas i s s i n c e 1 9 8 5 ) . Jn the absence of direct staLislicaJ '~stimaLcs, _:mport supply elasticities can he Jn fer red i 1d i rectJ y from two sources. These are: {a) P.stirnates of the capacitv of ir,dependents Lo import gaso}ine and;
2 With respect to the capacity to import, the issues are: (a) Whether there are significant institutional constraints on imports; (b) Whether independent (ie., non refiner) marketers have ability to import enough gasoline from the U.S. and offshore to make it unprofitable for domestic refiners to attempt to restrict. output. and raise price. Issue (b) breaks down further into two basic questjons. These are: (a) What is the pot,ential capacit.y of u1dependenLs to import? (b) How much would the domestic price have to rise before that capacity is utilized?
With respect to in.stituliona-1 constraints, the following questions arise: (a) Are there governmental oarriers Lo importing gasoline?
no quo Las or () t ~l (: j' i} (J ! ' ~asol inP imports. The irnposit.i:n1 of f'PL,\.r-icL:<HlS •1n o f r' P s Lr i c t i on s on i mp o r t s from o f f ~~ h <FT m; 1 :-: \
l' •. i a! I l ' 'lf! T:~ i ':I': i 1. '. \) r· ; d s c f V'
3 regarded as unlikely given the govPrnment's stated commitment to trade liberalization. lt is also considered unlikely that the U.S. would impose any restrictions on gasoline exports to Canada. (b) Are the quality differences between Canadian ~nd imported gasoline such as to constitute an impediment to importing? Gasoline from some lJ.S. sources may have a lower octane rating than Canadian refined gasoline. Blending imported gasoline with domestic gasoline is normally sufficient to remedy this problem. A facility capabJe of raising the octane rating of imported gasoline by one point can also he installed at modest cost.
The current capacity of i nd£~pPndents t.o import ;4nso l i ill! 1s as follows: (a) ~ueJ:?ec Jmports into Quebec com£~ principally via tanker from New York harbour, H.otterdarn and other offshore sourcP.s.
g<·nerally not compct i ti ·,f: >-ti th martn<· c lr~ocs. !!iii !es ~;nTo opPrates a (pr·rni:ial in :-lor;trr:al. The:-;£~ firms
4 market gasoline both through their own networks and through other independents.
CONFIDENTIAL CONFIDENTIEL It is not unreasonable to conclude that the ability of non rPfin1!rs to jmport. is such thn.t the li-indr~d pric<! of import.Pd gasoline constitutes an effpc1,iv1! cPilini:; on thP who1esa1P price of gasoline set by (~ucb(~C' refiners over thf' medium Lo Jon~ term. This should, ;Jf cnursP, hP confirmed with r<!fl•rencf' to thf' price· dat.:c. LhPm'.">(!lvPs. This is donf' 1nSPction Ill.
5 (b) Ontario Imports of gasoline enter Ontario principally via tanker truck from Buffalo and via marine cargo. (i) Imports by Tanker Truck Most of Ontario's independent terminal operators are located within economic truckin~ distance from Buffalo. Gasoline can also be trucked directly from terminals in Buffalo to individual service stations in southern Ontario. The important question is, again, whether given an increase in the wholesale price set by Ontario refiners, independent Ontario marketers could readily increase their gasoline imports from Buffalo thus making the price increase unprofitable. This depends on, among other things: the availability of additional ~asoline in the Buffalo area; - the availability of tanker trucks; - the capacity of border points to handle additional traffic. Gasoline enters the Huffalo area by truck from rt-~f:nPries in Pennsylvania and tl;. pipPl 1nP from ~ew York harbour and Philadelphia.
: CONFIDENTIAL .:_ CONFIDENTJEL
CONFIDENTIAL <C ONFIDENTIEL Current Canadian liftings plus the additional gasoline that could be made available in Buffalo should the demand exist total 20,000 barrels pPr day 11r 7.3 million barrels pPr year. This amount.s to ~3.3 per ce:nt of 1988 Ontario net supply. There doPs not appear to bc~ an:v prob] em obLain ing Lhe Lank e r t r u r k s re q u i red to ha u l th i s add i l i on a ] gas o 1 i rw ! . o Canadian terminals and serv1r1' slat.ions. The number of'
7 additional trucks required is not lar~e (20 trucks making three trips per day could bring in additional 9000 bbl/day). Similarly, the additional number of border crossings involved is small (in percentage terms) and the contribution of additional imports to border congestion would be minimal. Congestion could occur, if anywhere, at U.S. terminals especially if liftings by Canadians were to increase dramatically. This problem could be solved by increasing rack capacity if it appeared that additional Canadian demand would be sustained. ( ii ) Imports of Marine Cargoes There are six major operators of marine accessible terminals in Ontario. These are Universal, Petrocor, McAsphalt, Oleo, Roy-L and Montank. There is presently marine accessible capacity to store 570,000 barrels of motor gasoline. Some 200,000 barrels of addit]onal storage capacity could be obtained at modest cost by converting additional distillate tankage. CONFIDENTIAL CONFIDENTIEL 900,000 barrP]s of mar1ne accp--;sible storage: capacity 1n thP hands of indt:pendPnt marketers ln Ontario. This st.oragf~ t~a~aC'ity will support less in t:-i1! 1..; ay of' annual throughput than it would if it Y.:Prl' ac<·Pssiblf' A reasonable estimatP
8 is that the six major independent terminal operators could handle 7.7 million barrels of marine cargo annually. This amounts to 9.8 per cent of Ontario net supply. Given that the marine terminal capacity exists to expand imports significantly the question remains whether there is sufficient clean cargo tanker capacity suitable for use in the St. Lawrence Seaway. The answer is that, given the current size of the fleet, there may be shortages of vessels at certain times of the year. It js also the case, however, that the number of vessels avajlable for the gasoline trade could readily be expanded over a reasonable period of time. (iii) Marine Cargo plus Trucks Summing the potential capacity of independents to import gasoline via marine cargo and via trucks yields an import capacity of 15 mi] lion barrels of gasoline annually. This amounts to 19 per cent of 1988 Ontario net supply. (iv) Imports via Montreal CONFIDENTIAL ·. CONFJDENTJEL 0 1 co ' s "1 on t r P. ;i 1 t e rm j n a 1 j s c: on n e c t e d w j th Lh e T n1 n :-> -NorthPrn pipeline through which it could ship ga.sol inP to Ott.nwa and as far WP.st into Ontario as Maitland. Thus, in Lhr. event of a supply 'restric:lion in Ontario, addi lional import.s could rPadiJy f]ow into PaslPrn Ontario Yia Montreal. ThP Trans-Northern is most h1·av i 1 ;v used du!' in)!
9 the winter season to move hea Ling o i L from Ontario refineries to Quebec. Capacity would be available for western movement of gasoline during the period of peak demand for that product.
Assessment Independent (ie., non-refiner) terminal operators in Quebec have the capacity to import approximately 51 per cent of that province's motor gasoline requirements. Independent terminal operators, brokers and wholesalers have the capacity to supply almost 20 per cent of Ontario's requirements. At present imports by independents account for roughly 3.5 per cent of Quebec's net supply and roughly 1.5 per cent of Ontario's net supply. lt is clear lhal there is significant potential for expanding the ~asoline imports of the independent sector. J1, js also clear Lhat it would not take a significant increase in the Canadian price relative to foreign prices lo induce this expansion. The reason is that the infrnstrucLure required is in placP..ThPrP could hP somf~ rark cnn,gf~stion in Buffalo and,
''t>risidt'r':ihly \-.iLhon1 af1\H'<•ci;1hl:• imp1iPs Lhat 1,his group is 11• :1 pu:-;iL!t>r; \,tJ frusLr·at(
10 attempts by domf~Rtic refiners to r1~Htrict their output and raise wholesale priceR. 8uppoHe, for example, Ontario refiners attempted to restrict output and raise the Ontario wholesale price by 10 per cent (about 2 cents per litre). According to demand elasticity studies the elasticity of demand for gasoline in the short-run (under two years) is around 0.4. This implies that in order to raise the wholesale price by 10 per cent Ontario refiners would have to reduce their supply of gasoline by about 4.J per cent or abo11t 3. 2 mi 11-i on barrels per year. Ovrr the longer term when vehicle choices and locational patterns can adjust, the required output reduction wou1d he larger. Independents could and would have~ an incentivP tu frustrate any attempt to raise the wholesale price. As thP analysis above has demonstrated, independents have thP capacity to bring much more than :~.2 mi 1 J ion addit.ional barrels of gasoline from the United States or from offshore. Their incentive to do so lies in the higher profit associated with their increased share of the Ontario market. The same reasonin~ applies to quebec ur l,o Ont.ar·io
1 1 fl I. J n_t!:..Q.Quction The examination of the ability of independent (ie., non refiner) gasoline marketers to import product from either the U.S. or offshore has revealed that with existing infrastructure imports could supply as much as 50 per cent of Quebec requirements, 20 per r~ent of Ontario requirements or about 30 per cent of Ontario and Queber· requirements. The threat of imports thus constrains domestic pricing alt.hough it need not eliminate dome:..;tic pricinp; discret.ion entirely. There are a number of po8siliilities here. First, the import supply function could be infinitely elastic. The domestic producers could either be price takers each producing a quantity at which their respective marginal costs were equ<ll t,o t.hP lan1h~d price of imports ur lw fH'lCf' sPt.ters, jointly pricin~ t.o exclude imporh;. 'lhc Jatter is the tradjtionaJ (Eastman-Styko1t) limjt pricing modE~ l . The latter is also consistpnt. with the existence of noni.rans i tory exports wh1 le the:) fornu~r ; s not. In f~ i th er any chan;.r~ in Lhe forei~n nricc
th(· J;1r1dP.d pr1rc of import:.; ;1n1l thr' ;;ur:iv of ir.~rior·t.'...; 1' '>Tl ~; t' c1 t1 () n ( · ( · s an ci (, ~t n n < > 1 !:) r: ~~;, ,, ~ (· ;::. 1 • • . , - < ~ , . , i ~ 1 L: ~
12 factors the cost of import:ing gaHol irw 1s an increasing function of the volume imported. The ela.Hticity of import supply is finite. In this case the 11!..~~gJl}!:l_l landed cost of imports is always equal to the Canadian price but the marginal cost of ~mported gasoline exceeds its average cost. Under these circumstances the Canadian wholesale price could exceed the price in the lJ. S. plus 9-Y_~:r_ag~ transportation costs to Canada on a continuing basis. Canadian producers would have some pr·icing discretion and changes in either the foreign pr·ice or average transportation costs need not be fully reflected jn the domestic price. In sum, evidence that domestic wholesale gasoline prices exceed, on average, the foreign wholesale price plus average transportation costs to Canada and do not n~spond fully lo changes in the foreign price point:; in l.bc direction of a finite supp1y rluslicit.y and at least. some domestjc pricin~ discretion. Evidence that the Canadian price does not exceed, on avera~e, Lhc for·eign who.L\~sale pr1ce plus avera.ge transportation costs and that th<~
d i s c, r· <' L i o '1 •
D-·- ata and Models ~--------------------The product examined here is motor gasoline. Th(~ geographic areas examined are Ontario and Quebec. These geographic areas may or may not he economic markets. This does not matter. The only question here is whether and to what extent sellers in these areas are constrained in their pricing by sellers located outside these areas. 1f they are weakly constrained or unconstrained then the geo~raphin market would coincide with or he smaller than these areas. Tf Lh(~y Hre ti~htly constrai1H~d the geo~raphic market encompasses a wider area. The type of transaction examined 1c the ~holesa]e transaction. The question ·1 s thus thl~ extent to which imports or potential import.s constrain wholesa1c price setting. Posted wholesale prices a1·c called rn.ck pri:~ci:-:. On1y Toronto and Montrpa] rack prict~s art~ c;:;:aminecl. we have assumed that rack prices at, oth(~r de] ivery po·i nts in these two provinces di ff er from e i Lh(•r foronto or Montreal only by transportatjon costs and move in lockstep with
suhm i si:-; ion. I <j p ) ' 'l'llLt. - IOL Toronto raC'k pr·1(·(', t< :ir'('(-j n:otor ':~sr>· in" <lt)~~f'T'V• 1 ti ()f', ~~1 (' f··1~l~i;1\ 1 1J \.;('1 ,,
14 TRlJt = IOL Toronto ruck pric<', unleaded motor gasoline BRLt = Buffalo avera•e (highest and lowest) rack price, leaded motor gasoline pJus transportation observed on Friday of week L fin Canadian dollars/litre] BRUt = Buffalo avera~e rack price, unleaded motor gasoline plus transportation MRLt = IOL Montrea] rack price, leaded motor gaso]:ine MRUt = IOL Montreal rack price, unleaded motor gasoline GCLt = U.S. Gulf Coast leaded gasoline averat;e o-f bid
and asked spot pri~es two weeks prior to we~k t plus transportation GCUt =U.S. Gulf Coa~;t unleaded gasoline pr1cr: plus transportation The second type of data is daily data. Tt :is supplied by TC 1. These data run from January 1, 19HR to April 12, 1989. They include an averagP. !l11ffalo racl< price and sjx Jlptrocan and Turbo).
'.\l~D. = ( ( (HI\ I:~. i ~'.:l '• i \ • . .. I .. • t
wherel BPt = average Buffalo rack in U.8. centf; per U.S. gallon on day t XRt =Canadian dollars/U.S. dollars on day t The basic price model tested is suggested and explained in the IOL submission. To use Toronto as an illustration the hypothesis is that the desired Toronto price is a mark-up on the Buffalo price plus 1.ransportation from Buffalo. Specifically: The actual Toronto price cannot be adjusted to the desired Toronto price instantaneously so that Substituting for TR* we ge~ TH.t =ah+ dbBHt. + (1-b)Tl{t-l When all adjustment Is comp l.ete TH. TR = a + dBH The Toronto price equals the Buffn1o price after fu!1 adjustment if a = 0 and d =
{~onstant. terrn in(:~) dot·~-~ nc: :~1+'ff·1 ·..:.::t.i1:-·ti1·alJ>- ?>~ r·o. ThP hypc\h(~sis that d = 1. Frr~i:..;h1. fr<;m Buffalo 1s assuml•,J to b· r) pr ! i t r p ~ ,\ p 1 (. l\ - 1 I I" I h; 1 !' ~~ (-. : I t f"' i '-0 : l\, [) i d (' d h h I 'fl ~; ~ ~.-; (J I ",I r·a'.h::r· than Tot'on1.o.
1 5 (1) a_?O, d>l ( 2) ( '") t \. .) I and L ( 4)
I ·,I' \ ) I f"r·orn .:·.-. . :->. ('<'nL•; : ) 1 ; I I • : l I • ; i' l ! ; ; n . t \' ~ :..:. l) ~'
16 of thr. cocff i c ients on Blit and TH t- I dof•s not differ statistically from one. Reporting first, the results obtained using the TCI daily data we get TRDt = .0671 + .0346 BRDt + .9632 TlWt-1 (0.96) ( 6. 93) (Jn9.5'1) n = 832 R2 = .99 D.W. = 2.04 h = -0.54 (t-ratios in brackets) The null hypothesis that the constant term is zero cannot be rejected at the usual levels of significance. The estjmates of db and (1-b) sum to .9Y78. The hypothPsis that db + (1-b) = l implying d = J cannot be rejected at
t,h e usua ] s1. gn1. .t' 1. cance ] eve ] s. J 'l'h (~ mo d e,l explains 99 per cent of the var:iatjon in the Toronto rack price and thr>rP is no evidencf~ of a.utocorrcla~ Jon (Durb:n's h 1s not statiRtica]]y significant). The Toroni.o price does not adjust instantaneous 1 y t.n thr: Buffalo price. The mean adjufitment lag is (J-b)/db -963/35 = 27.5 days. This mPans I.hat i~ takes C,7.S days for half cf any di sere pane~-· be"\.;.;('('":: 1 L1· <~1'.'-' red and '.tr'tua I Tc '"OTl t.o arr: t. = - . oo 2 2 I u o :~ 3 = - o . 6 t
17 (a) I,eaded: TRLt = -0.312 + 0.159 BRLt + 0.857 TRl.t-1 (-0.61) (5.26) (28.2]) n = 111 R2 = .94 D.W. = 2.11 h = -0.61 (b) Unleaded: TRUt = -0.214 + 0.173 RRUt + 0.844 TRUt-1 (-0.40) (5.06) (23.81) n = 111 R2 = • 94 D.W. = l • 97 h = 0.20 The results having the following characteristics: (a) the constant terms do nol diffr::r statistjcally from zero (b) the respective sumR of db and (1-b) do not differ statistically from one (c) thPre is no evidf~nce of misspecification (d) the nwan lags are 5. 1 week~. :;.nd 1. :J weeks respectively. Severa] varjations of this modPi i,.;cre run. The first a. l 1 owed for seasonality and fm:nd nonf~. Thf~ second al 1 owed for an asymmetric response to the Buffalo price, to wit,
llBPH = - () oU1Pr'1..'iSf' The basic modi~ l was th('n r·<'--<~s Li :;1:1: 1·d ""i th : h '
18 Tr~L t = -0. 312+0. l 63BRLt-O. 0007 DBPR* URL t +O. 854TRLt-l (0.61) (4.86) (-0.23) (26.01) n = 111 R2 = .94 D.W. = 2.10 h = -0.61 If the asymmetry hypothesis is correct the coefficient of the interaction dummy DBPR*BRLt should be statistically significant and po~itive. It is neither. The asymmetry hypothesis is not supported by this result. The model was also estimated using five different Toronto transaction prices in place of the (posted) Toronto rack price. Only the results for unleaded Masoline are presented here. The transactions prices are defined as: ITTUt = representative IOL Toronto transaction price on Friday of week t for unleaded motor gasoline. TTTUlt = TCI Toronto transaction pricP for unlead1~d motor gasoline - CONFIDENTIAL CONFIDENTIEL TTTU2t = TCl transaction price -TTTU3t = TCI transaction price -The results are as follows: (a) ITTUt = 0.430 + 0.148 I3RUt + 0.869 ITTL~-L (-J.08) (5.48) (:29.:1:)) n = 111 R 2 = .96 D.W. = 2.03 h=-u.Ll mean lap; = 5.9 weeks
(-0.66) ( 2 • 3 3 ) ( z:L 85 l n = l I l mean lag = 9.9 weeks
19 (c) TTTU2t = -0.587 + 0.143 BRUt + 0.889 TTTU2t-l (-1.13) (4.09) (29.15) n = 111 R2 = .95 D.W. = 2.09 h = -0.52 mean lag = 6.2 weeks (d) TTTU3t = 0.023 + .079 BRUt + 0.920 TTTU3t-l (0.03) ( 1. 82) (26.38) n = 111 R2 = .92 D.W. = 1.84 h = 0.88 mean lag = 11.6 weeks These results have the following characteristics: (a) they all have constant terms that are sta~istically zero and the respective sums of the parameters db + (1-bl do not differ from one implying n = 0 and d = l. ln this
they are identical to the results for posted prices. anything, than is the case wit.h posted price1'. Pspecially true of the Texaco transaction prices. are prices at which gasoline is not being Lifted. None of these results imply that transactjon prices respond to Bu ff al o more qui ck l y than post r>ri 'J' hf• mod c 1 was a 1 so ·~ s l i ma t.r·' d (;nt.1,rin)='. s<~paratP1y. Th-P h:vpothcsio~ hPt"(~ 1s th:it th'' f~uf"f"alo rn·1cP serves as a sig-nal Zlf' foc.-1] poi!lt but. d(,r's not imply anything about import comp<:tiLion. in 1.hP (•xchang<' ral<'. Wlli l(' th£' '.;p<'<:i f11'aLion of I.hi~,
1 ' ; f ' Tb is LS These p n c:es. w i th t h,, Bu f' f' a .l 'i p 1 · i c r· I f L hi '.·: l .L.'
20 model leaves something to be desired il does show that the exchange rate matters. 1 Indeed the results show that the short-run response elasticities are virtually identical (at 0.15) and that the long-run response elasticities are both one. The exchange rate matters as much as the Buffalo price in both the short and Long-run. The implication is that it is the threat of imports that matters rather than the Buffalo price as a focal point. Montreal results with the weekly data arc as follows: (a) Leaded MRLt = 0.655 + .044 GCLt + 0.926 MHLt-1 (1.15) ( 2. 04) (28.24) n = 111 R 2 = .91 D.W. - 1.99 h - 0.01 mean lag = 21.0 weeks (b) Unleaded MRUt = 0.102 + 0.033 GCUt + 0.965 Ml?Ut-J ( 0. 23) (l.24) ( 39. 5t:)) n = 111 R2 = .96 D.W. = 1.96 h = O.~J mean lag = 29.2 weeks
contemporaneous <lulf Coast pricf~ and constant terms ar·e not '...; i e;n i I' i can L. + (1-h) = .969. ThP s~.andard rj('\': ;it I ( 11 . 027. ThP nul I hypo t.hPs is I.hat. ·he' :-"1m I. J\t. the time of writ.in11; nnnl i11P;1r· •·:·'.t 1n::1Li1.·ri ha~-· .... y1·t not. yielded ffi{~aningfu1 rTsults.
1.• ; ith,. l•rn~ 1a.c;. The li: tht~ !Padcd mod(:] brl " 1.h 1 '...; :-;um ; ~. , _., orit> (so d :. I )
21 cannot be rejected (t = -1.15). In the unleaded model bd+(l-b) = .998. The standard error of this sum is .022. The null hypothesis that the sum is one cannot be rejected. The Gulf Coast price lagged one week has a statistically stronger effect. The results are: (a) Leaded MRLt = 0.661 + 0.053 GCLt-1 + 0.917 MRLt-1 (0.57) (2.38) (27.75) n = 111 R2 = . 9 2 D.W. = l.97 h = 0.20 mean lag = 18.3 weeks (b) Unleaded MRUt = 0.003 + 0.049 GCUt-1 + 0.956 MRUt-1 (0.01) (1. 86) (39.28) = 111 R 2 n = .96 D.W. = 1.95 h = 0.21 mean lag = 20.5 weeks
In both cases the sume bd+(l-b) does not differ statistically from one. If the Gulf Coast price lagged one period is replaced by the Gulf Coast price lagged two periods the result for leaded ~asoline is virtually identical. The result for unleaded gasol in;~ is man~inai ly weaker statistically but virtually the ~ame in its implications. Turning to MontrPal transaction prirPs for unlPndPd gasoline we have two transact.ion pricP Lime series. These arP: TTMUI =Texaco Canada unlt'ad(~d transact.ion µr·ic;(~ ;.;1th CONFIDENTIAL CONFIDENTIEL
TTMU2 = Texaco Canada unleaded transaction price with
Applying the same model we get: TTMUlt = -0.183 + 0.065CUt + 0.948 TTMUlt-1 (0.42) (2.34) (37.91) n = 111 R2 = .96 D.W. = 1.81 h = 1.03 Mean lag = 14.6 weeks TTMU2t = -0.488 + 0.066 GCUt + 0.960 TTMU2t-l (1.19) (2.34) (4fi.59) n = 111 R2 -- .9 7 D.W. = 2.26 h = -l.40 Mean lag = 14.5 weeks The Montreal transaction prices are somewhat more responsive to the Gulf Coast price than the Montreal rack price. The partial correlation is stronger and the mean lag is almost 15 weeks shorter. Thjs is a contrast to the Toronto results where the transactjon prices and rack price produced very similar results. Insofar as the series provided by Texaco are concerned, the Montreal transaction prices show considerably more variahi lity (more fn~quent chan~es) than the Toronto prices. This may reflect thP frequency with which the Toronto customers lift from Texaco rather than any difference in the market. The weekly model may well hide some of the rf' s pons i v en es s of Mon 1 re a I transact i on r r i r f's t. o for<' i £! n pr i <~e changes. Prices to o f ten t: hang<~ mo r c \. h an CONFIDENTIAL ·~_CONFIDENTIEL
22
once per week. These changes do not show up in the Friday observations which are matched to the Gulf Coast price. Thus Montreal transaction prices are more flexible than is implied by the weekly series used here. The basic result for Montreal is that while Montreal prices respond relatively slowly to Gulf Coast prices they do not differ, on average, from them. models have the counter-intuitive implication that the foreign supply function is less elastic! in the case of Montreal than in the case of Toronto. The results obtained for the weekly data are for the period January 1987-February 1989. the IOL sample period by some six months. lt has been suggested that the additional observations be used to test the "]earning hypothesis", to wit, thai thr' majors thought that deregu]ation would result on a more serious import threat than it actually has. !laving learned t,hat imports do not constitute a serious Lhrc~at., :.he majors begin to raisP their Toronto rack pric<~s ri:lat.ivc: to Buffalo. rPsult. is an upward drift. in eiLh(•r Lhr' :i or· d coeffici(:'1l or ho Lr;. T·~stimat.ion of th(~ daily 1wdcl :·or ;';p pc;·iod .Janiuir·y I, l :rn 8 - ,\ p r i. 1 1 ~ 8 ~ y i e 1 d s t h ' · n · s ul L '',:'J \>, J'',t. - -\'),0.10 + (),(,',' ))~ ( J",'1 \ 'l'1, + (-0.0G) ( 5 . G5 ) r1 = HiH mean lag = 26.] days
23 The Montreal price As such, they extend The 1 l ) , .J , . t . < . ; . .. :.1 ., : 1..; !---l ( IL! . I ;
A comparison of this result with the daily model estimated for the period January l, 1987, reveals that the model estimated for the later period has a smaller constant term but that the two models do not differ statistically. Thus at least as far as Toronto pricing is concerned the evidence does not support the argument that the majors have learned that imports are not a source of concern. If anything it points in the opposite direction. An attempt was made to experiment with alternative Jag distributions. The model repr~sented by equation (3) constrains the weights on lagged values of the independent variable to decline geometrically. That 1 s , th(~ r~ qua t i on TRLt = .01 + .15 13HLt + .85 TRLt-1 can also be written as
+ • 092 BRLt-3 + . 018 BHL1,_4 + .. where the sum of the coefficients on t:hP BRLt-i approaches one as :i becomes very lar·ge. This type of Lag distri">ution a:-;sigri;.; thP ~reaLl'SL va1·iahlP (if'., in rPsponding to the ;luf'f'aio nrjcr: gr Pat. P s t. we i p; h L . l q r 1 r· d (' r· t, ~ ) : 1 I f c) \ ..; 1· r ·, r - \ 1 1 i .~ ..
ntl· ·, \(·n1l ··r: ·. ;•Jay 1,_: ••; ;;ch Ui;1t :1n ·:,tr! 1f·r •.' <..;; ) : :1·
dependent variable. For example: If az > 1 and a3 < 0 this specification implies a lag distrjbution with one hump (ie., an inverted u). Jt proved to be the case, however, that either a 2 < or a3 > 0 or both. As a consequence, this specification also implies geometrically declining weights and does not constitute an improvement on equation (3). A second alternative lag r:! i :-; tr i but ion was emp1 o~d~d w' U1 the daily data. This approach constrains the weights on th e I a g g e d val u es o f th P i IJ ch: pend c n t var i ab] e to Li e a I on g a polynomial of a specified order. Thjs jg called the po] ynomal di std buted ] ag ( PDL) or Almon lag technique. Experiment~ were run wi1h up t.:1 CO la~~ftrd va11Jr!S of' the independent variable~ it!:>elf la~gPd up to t.wo weeks. rrh i_s "extreme" speci fica ti on hypothes i ;,-;es that the Toront.o pr i CP depends on all the values of the Buffalo price betwec-n IS and 75 days earlier. Th(' result of th 1 s ex per] nH:nt 1 s thn l the sum of coefficients of thP (){) lnrrgr~d valuPs nf the ll1;f'falo ~)rJcP 1s 'l'h l' r·r · s i g n i f i <' ;ui t c o n s t a n t. L <~ r m . dist.rihution mah:es 110 s<~n~;( ! t 1 '..:..: ,_t :1 i nvt' r· Lr:d ll ~"1orPnV<' r· : hf' ,Jc; 1 r; i r, i.\a Le.or! ; s lJ, 0); i r,; :; , .·· a serio·.1s mjsspf:cificalion ,,f ::,n;r;(' ~._,_:.r1 < · :i 1v ~ I \1 d ,· · d t h a I. i m p r' o v i n >': < ,; i 1 h <•
26 (equation 3) is goinM to be very difficult if it is possible at all. Some models were also estimated using "monthly" data supplied by TCI. On further inquiry these data proved lo be monthly averages of the daily data described above. The averaging process throws away information leaving only 24 monthly observations to work with. Moreover, the averaging induces autocorrelation which causes further problems. The use of monthly averages became indefensjl1le when the underlying daily and weekly data were made avai]ablr:. While results obtained using the 24 monthly observations are not reported, two experiments performed with the monthly data are worth noting. First, the question arises as to whether the ToronLo rack pric<: is responding 1.o th(' Buffalo rack pri:~(~ <H' to the pr·:icP _,, crude oi I which is vi rtua] l y th<~ same in both Canada and the lJ. S. To test t.he hypothesis that it is the (North American) price of crude oil rather thein the' pricP of L.S. ~asolinc
(~aT-.<icJi ... 1r: t'rlJ(!f\ lri \·r1ic..,;.1~ .. r1 ir1 (':-1n;1,, 1;-1;-' <i,·~\ !;1r·~-:, ·~ 10 r· :i.dded as an Pxpl:1n;~tory vari~;'.i!P :s unPqui\ocal \" \ f t i; () l : \, 1; ~l l ~ h; l;) L: c d .
; 1 ;,t'•. 1 '"·c1s 1r1 cqu.ci.tiun i'.l) ThP 0 I
27 Second, if Ontario refiners are truJy constrained by international factors their respective refining margins should vary with both U.S. gasoline und crude oil prices. Specifically, given the Buffalo rack price, an increase in the Canadian dollar price of crude oil due either to an increase in the U.S. dollar price or a depreciation of the Canadian dollar, should reduce the refJning margin. Given the price of crude oil an increase in its Buffalo rack should increase the refining margin. This hypothesis is tested using , the TCl monthly Buffalo rack and the price of Canadian crude in Chicago. Both the Buffalo rack and the Chicago crude oil prices are highly significant and the correct sign.
CONFIDENTIAL CONFIDENTIEL
The statistical results repor1,ed in the previous section imply that in the case of Toronto that: (a) The Toronto rack and transacUon prices are, on average, just equal to the Buffalo rack price (in Canadian dollars per litre) plus transportation from Buffalo. (b) The Toronto price does not adjust immediately to a change in the Buffalo price. l.t takes a.bout four weeks for hald the required adjustment to occur. /\ typicai adjustment pattern would be: Week l .L 8 2 33 :3 45 4 55 5 G3 6 70 7 75 8 79 9 83 JO 86 ll 88 12 90 13 92 14 93 15 94 16 95 What do t.hPsP results i:np1v :'.b(;:1l I' • J forP i «;11 supply·:' r;' j I'S t , th C l 0 fl .<J: - I' U f; (' <) U <). : l. \ (; f' ·th I 'J'C)r·onto and Buffalo pr-icc~s irripl i1•:-; : !:CJ· ( b ) Th P a c t u a 1 o r p (·' r c e i v f' d (' i;; :-; t 1 .. ! \' : ; t' f ci r c 1 ;.; r i ,;\lf;,il:• 1!-'. !nf'iriilP.
/. !J Second, the lag HI the adjustment of Toronto to Buffa.lo prices implies that the short-run import supply elasticity is finite. That is, the marginal cost of imports is an increasing function of the quantity of imports supplied in the short-run. In the short-run the delivery cost of the marginal litre exceeds the average cost. There are adjustment costs of somP sort and these costs serve to limit the potential irnporL responsf: over the short-term, say, under six weeks. Over lhe longer term, say more than t.we 1 vf: w<~eks the supply of j mports can be regarded as infinitely elastic. The adjustment lag does not imply that the Canadian and UnJted States gasoline markets are not linked or that Canadian producers currently have any r.H1 rket pnwer. J t i rn p l i <~ s on ] y th a t, o v e r a ;· c 1 a t j v c I :- s h o r 1 p P r j o d ( mu c ~1 shorter than the United Stat.es .Justice D<?partment's two year horizon) the supp]} function of i m<ior Ls is upward sloping. On~r the re1evanL JH:riod jt is or is pr~r~eived to
T' \ ;l
issue is not what imports are but what they could be. examination of the capacity of independents to import reveals that as much as 30 per cent of Ontario 'and Quebec requirements could be met from imports by independents. It is highly unlikely that domestic market conditions would be such as to require this level of imports. is there and it is reflected jn the pricing polici~s of domestic refiners. This is most clearly evident jn Ontario. The e v i d P n c e i s v e r y s tr on g are pr j c i ng to meet the landc~d price or i rnports. implication lS that Lhe refiners' perception of the marke~, situation is that higher pric(~S would jnducP additional imports resulting in a loss in volume: and would Tl<)t, as a consequence, be profitable. Thus notwithstanding Lhe rt'=lati\·ely small proportion o:' lhi~ markr~t acc'()\Jn1,cd for' bv imports, thP influence of "imports on dom~:stic madH~t. pr1c1•'.c;; is signif'ican1,, The merger does not change this. ThP situation 1n Quebec differs 1n some respects. Tb£: prov 1 nee has historically lwen open Lo imports of n:d' i ned pr-od111· L. 'T'h erP :1 rr two prov;n1:ia! rF:qu1rem1'nt.s "r()m impc. . ;·t.-.. ( 'o a c; t. p] 1 is 1 r· ans p' l r· t; 1 t. i <: ri ( • (' :., ; '- Cf;ns i di'! rah 1 (' VO l at i i j t .. mo r '' \ \ ;1 i I (} ! I l (· '"' ~· n r·1 Lu pr· 't•'.'-.; ':11 1,.
30 The But the~ threat t ha t 0 n tar i o r«~ f i n c rs The i l, l < ~ ~ l r r~ · 1 t l i l ) ~ 1 () j ~ ; l. ( \ L \ ~ ) r ' f ' ~-) d c) !"1 ( ) t : 1 . ; ~ ! ~ ~-· ; .!. : ' , ~ ) I ; T! r · 1 ( ' f ' t; :-, ; l ' ~ h an l :-~ c; i1 o;.,;:' Ly t he i !'.', '•I. il t :·i :\
31 prices are not as closely cor re I a ted ~,· i th the Gulf Coast as Toronto is with Buffalo. This could be taken to imply lhal Montreal pricing is more insulated from international factors than is Toronto pricin~. IL might also imply that Montreal pricing is simply subject to a more complex set of international determinants. This inh~rpretalion .is more consistent with the import quantity and capaci t;. data and is more persuasive. Taking the two provj ncPs LogPthc'r' \he pot(:ntj al competition from imports is ,_;uch that. dorn1~sL1c whol esal<: prices above the landed price of imports wouJd not be profitable in the short-run, Jet alone over two years. The merger does not change this ci.nd j s, thr'n~fore, unlikely to result in any change in domestic wholr;'.-:alt~ prir.ing hPha...-iour.