Full opinion text
OPINION EASTERBROOK, Circuit Judge. The Northern Indiana Public Service Company (NIPSCO) generates electric power in coal-fired stations. It serves much of northwest Indiana. Unfortunately, NIP-SCO’s forecasts of its needs for coal have not been accurate — in part because the demand for power in the region has declined rather than increased (about 40% of its sales are related to the steel industry). NIPSCO also did not anticipate the rapid movements in the price of coal, which varies with the price of oil. Coal is valued for its energy content, so as the price of obtaining BTUs from oil declined, the price of coal also declined. The price of high-sulfur Midwest coal, readily available in the spot market, declined most quickly. The transportation cost from mines in Illinois and Indiana also was low. NIPSCO, however, was committed to purchase large quantities of high-priced, low-sulfur western coal. The delivered price of this coal sometimes was more than twice the delivered price of midwestern spot coal. Two contracts with Colorado Westmoreland, Inc. (CWI) signed in 1977 obliged NIPSCO to take a total of 1,050,000 tons of coal per year through 1993. Although CWI and NIPSCO renegotiated these contracts in 1980, reducing NIPSCO’s minimum take (and increasing the price per ton), a large supply of CWI’s coal accumulated in a stockpile at NIPSCO’s Mitchell generating station. The Public Service Commission of Indiana (which I call the PSC, even though its name recently was changed to the Indiana Utility Regulatory Council) thought the accumulation wasteful and excluded $52 million from NIPSCO’s rate base (the “used and useful” investment on which a regulated utility is allowed to earn a regulated rate of return). NIPSCO also had a contract with Carbon County Coal Co. requiring it to take a substantial number of tons per year. Concluding that it did not need so much low-sulfur coal (at least not at Carbon County’s delivered price), NIPSCO broke its contract. The result was a judgment for $181 million in favor of Carbon County Coal. See Northern Indiana Public Service Co. v. Carbon County Coal Co., 799 F.2d 265 (7th Cir.1986). Reeling from the $52 million exclusion— but before the $181 million judgment— NIPSCO decided to renegotiate its contracts with CWI. NIPSCO had been having trouble keeping its generating units at the Mitchell station in compliance with environmental regulations while using CWI’s coal. Seizing on this, NIPSCO told CWI that it considered the contract defunct; at the same time, NIPSCO tested CWI’s coal in Unit 15 at its Schahfer generating station, which was having environmental problems of its own burning coal from Medicine Bow Coal Co. CWI’s coal worked at Unit 15. NIPSCO offered CWI a contract to supply all of the coal Unit 15 needed. Between 1979 (when Unit 15 was placed into service) and 1982 Unit 15 had required about 1 million tons per year. CWI wanted a fixed-take contract; NIPSCO, intent on averting another fiasco caused by excess stockpiles, insisted on a requirements contract. In May 1982 NIPSCO and CWI started negotiating a new contract. They signed the document in April 1983. The important clauses of this new contract (the Contract) provide: SECTION 4 — QUANTITY OF COAL A. Buyer agrees to purchase from Coal Company and Coal Company agrees to sell to Buyer all of the coal required by Buyer during the term of this Agreement to supply Unit No. 15 at Buyer’s R.M. Schahfer Generating Station in Wheatfield, Indiana (Hereinafter called “Schahfer Unit 15”). Buyer represents that its estimated annual requirement of coal for Schahfer Unit 15 is approximately one million tons, as set forth in Exhibit I hereto. B. At least 90 days prior to the beginning of each contract year ... Buyer shall advise Coal Company in writing of the quantity of coal it estimates it will require at Schahfer Unit 15 during that contract year and for each month of that contract year. [If the amount exceeds 1.25 million tons], Coal Company shall have the right [to limit its sales to that amount]. Coal Company shall supply Buyer with its full requirements up to 1,250,000 tons per year and agrees to use its best efforts to supply Buyer with contract year requirements in excess thereof____ E. Coal Company agrees that it will not execute contracts to sell coal from the Mine to others which when added to the amount of tons which Coal Company is at that time obliged to deliver to Buyer, call for a rate or amount of production which exceeds the capacity of the Mine. Coal Company will treat all of its purchasers fairly and will not give preference to other purchasers over Buyer in times of short supply or production delays. F. On or before the 25th of each month, Buyer shall order from Coal Company, in writing, the quantity of coal it will require to be delivered hereunder during the next succeeding month, specifying quantities to be delivered in substantially equal deliveries during such month. The amounts so ordered in each contract year shall be in substantially equal monthly amounts, and shall in the aggregate approximate the tonnage specified pursuant to paragraph B above. G. Nothing contained in this Agreement shall be construed to require the purchase of a quantity of coal greater than is needed for the operation of Buyer’s Schahfer Unit 15; nor shall anything in this Agreement be construed to prevent Buyer from operating any and all of its generating stations, and utilizing other sources of power supply in the most efficient, economical, and prudent manner for the production and supply of electrical energy, nor shall anything in this Agreement be construed to cause Buyer to exceed its reasonable reserve stockpile requirement. Buyer shall treat Coal Company no less fairly than any of its other producers of coal. SECTION 25 — LIMITATION OF WARRANTIES Except as specifically provided above in this Agreement, neither the Coal Company nor Buyer makes any warranty, representation or condition of any kind, express or implied (including no warranty of merchantability or of fitness for a particular purpose). EXHIBIT I SCHAHFER UNIT 15 ESTIMATED ANNUAL REQUIREMENT (TONS) 1982 1,062,000 1983 887,000 1984 1,059,000 1985 981,000 1986 1,029,000 1987 1,042,000 1988 1,102,000 1989 1,179,000 1990 1,052,000 1991 1,117,000 1992 1,000,000 1993 1,000,000 After the parties signed this contract, NIPSCO promptly tendered the annual estimate called for by § 4.B, in an amount similar to the estimate shown in Exhibit I. NIPSCO later reduced its purchases for 1983 because Unit 15 suffered an unexpected outage. All told, NIPSCO took 737,655 tons in 1983. NIPSCO delivered an estimate for 1984 of 1,011,300 tons. It ended up taking only 713,295. It estimated 650.000 tons for 1985 and took 573,305; it estimated 263,600 tons for 1986 and took 314,871. (CWI shut its mine for the last half of 1986 because of a fire, but NIP-SCO’s personnel testified that by then CWI already had delivered all the coal NIPSCO needed for the year, because a steel strike diminished the demand for NIPSCO’s electricity. The parties agree that 1986 was an aberration.) And by mid-1984 NIPSCO’s internal estimates were showing requirements at Schahfer Unit 15 in the vicinity of 300.000 tons per year for 1986-87, though with a recovery to 600,000 tons and up for 1988 and later years. NIPSCO and CWI are at odds about why Unit 15 is burning so much less coal than the contract estimated, about whether this change was predictable in 1983, and about whether NIPSCO is required to take from CWI the tons listed in Exhibit I whether or not Unit 15 burns them. CWI sent NIP-SCO demand letters insisting on compensation for the shortfall. NIPSCO then filed this diversity suit, which the parties agree is governed by Indiana law. NIPSCO sought a declaratory judgment that it has abided by the Contract and that the Contract remains in force through its termination date in 1993. CWI filed a counterclaim for damages for breach, and added claims based on promissory estoppel. A bench trial was held between July 6 and July 20, 1987. This opinion contains the findings of fact and conclusions of law required by Fed.R.Civ.P. 52. The parties have stipulated to a substantial agreed statement of facts; each side also has accepted many propositions in the other’s proposed findings of fact. I adopt all of these agreed-on facts. To the extent the compression necessary to produce a readable opinion yields any inaccuracy, the agreement of the parties controls. But the parties do not agree at all on why Unit 15’s burn has been cut, whether NIPSCO should have foreseen this (or did), and what the negotiators said to each other about the meaning of the terms in the Contract. I concentrate on resolving these contested matters. Part I addresses the nature of (and the reasons for) the change in the operation of Unit 15, Part II discusses the negotiation and meaning of the Contract, and Part III handles some residual matters. As CWI’s post-trial brief accurately puts it, the central question in the case is whether the Contract “permit[s] NIPSCO to take advantage of the lower cost of electric production at its other units to produce more electricity there than at Unit 15 when that cost differential was known at the time NIPSCO represented its coal needs at Unit 15 to be about one million tons annually”. I answer that question “Yes”, and accordingly judgment will be entered for NIPSCO. CWI advances several legal theories. One is essentially fraud: that NIPSCO estimated requirements of one million tons annually while knowing that Unit 15 would not burn this much, and that fraud in the inducement entitles it to repudiate the Contract. The claim of fraud, that NIPSCO was keeping two sets of books, is the subject of Part I, which also covers the related claim that NIPSCO did not determine its requirements in good faith. CWI’s second claim is that even if NIPSCO acted in good faith, its negotiators made representations that are the basis of an estoppel because NIPSCO intended CWI to rely on them, and CWI did. This is the subject of Part II.A. Finally, CWI contends that the Contract itself, independent of collateral promises, requires NIPSCO to take the amounts contained in Exhibit I. I cover this and related claims in Parts II.B and III of this opinion. I A Any electric utility should produce power as cheaply as possible, subject to the constraint that lower costs not unduly reduce the reliability of the system. NIPSCO has consistently attempted to do this, using its own equipment to the maximum possible degree. This has led, ironically, to very high costs. The demand for power on NIPSCO’s system averages about 1,400 MW (megawatts), with the average daily minimum about 1,300 MW, the peak demand in the ordinary week approximately 1,600 MW, and the yearly peak around 2,400 MW. (The demands here do not include power generated and used internally by NIP-SCO’s customers; the steel mills in NIP-SCO’s territory operate more than 300 MW of their own generating equipment.) Modern coal-fired plants are most efficient at about 600-700 MW per unit, with two or more units in a station; 800 MW per station (two 400 MW units) is probably the smallest reasonably efficient size. See Paul L. Joskow & Richard Schmalensee, Markets for Power 48-54 (1983). If NIP-SCO operated only units of this size, however, it would have so few that a combination of ordinary maintenance and unexpected failure might leave it unable to cover the demand for power. The larger the units, the fewer NIPSCO will possess; the fewer units it possesses, the greater the vulnerability of the system to chance events. NIPSCO in fact operates a substantial number of coal fired units, with “net demonstrated capacity” (highest sustainable output) between 60 and 472 MW apiece. It also operates some units fired by oil or gas and two small hydro units; the hydro units are too small to matter, and the gas turbines are used only to cover peak demands. I therefore ignore both kinds of units in this opinion. And NIPSCO has contracted for power from neighboring utilities. Until 1987 NIPSCO took a “firm” 400 MW of power from Indiana & Michigan Power Co. (a subsidiary of American Electric Power Co.). This was a commitment that I & M met from its whole system, so it was exceptionally reliable; it was also exceptionally inexpensive. (For 1987 NIPSCO is taking 200 MW of power from I & M; starting in 1988 it will take none.) NIPSCO operates four “stations” — Michigan City, Bailly, D.H. Mitchell, and R.M. Sehahfer — each with multiple generating units. As of 1983, NIPSCO’s system had these resources, the average cost of which is shown (PX 174): Unit Maximum . Minimum $/MWH “I&M Contract” 400 400 Michigan City 12 469 250 18.98 Bailly 8 320 180 19.72 Bailly 7 160 ? 19.88 Michigan City 2/3 120 30 23.38 Sehahfer 17 344 ? 24.58 Mitchell 4 125 ? 30.38 Mitchell 6 125 ? 30.51 Mitchell 5 125 ? 30.54 Mitchell 11 110 ? 30.97 Sehahfer 15 472 200 34.25 Sehahfer 14 431 220 36.01 In 1986 NIPSCO put Sehahfer 18 into service. This unit is a twin of Sehahfer 17. Two things principally cause the dramatically different costs per megawatt-hour. One is the “heat rate” of the unit — the number of BTUs needed to produce a kilowatt-hour of electricity. The lower the better. Larger and newer units have lower heat rates. The other is the delivered cost of fuel. Some of NIPSCO’s units can use high-sulfur coal and others, because of environmental restrictions, must use the substantially more expensive low-sulfur coal. (It is both more expensive at the mine mouth and more costly to ship, because it must travel longer distances. For extended periods, the rail costs per ton of moving CWI’s coal to Indiana have exceeded CWI’s price per ton.) Even among high-sulfur units burning midwestern coal there is a substantial difference in variable costs of operation. Bailly 8 and Sehahfer 17 burn identical coal. But Sehahfer 17, a brand new unit, has a flue gas desulfurization unit (a “scrubber”), which is costly to operate; Bailly 8 does not. (I refer only to the average variable costs of operation. None of the costs in the table includes the cost of building the station. Throughout this opinion I disregard sunk costs.) The EPA allows Michigan City 12 and the Bailly units to discharge unscrubbed stack gas because they are located in dirtier areas and are older. On how Congress and the EPA have come to induce utilities to burn high-sulfur coal, see Bruce A. Ackerman & William T. Hassler, Clean Coal/Dirty Air (1931). Changes in the price of coal will affect the costs of production. For example, in 1986 Bailly 8 was the cheapest unit, at $17.35 per MWH, because the cost of spot coal had declined, while Michigan City 12, burning coal on a contract, cost $21.19 per MWH. But low-sulfur units have been substantially more expensive throughout than high-sulfur units; in 1986 Sehahfer 15 cost $36.26 per MWH to run. The marginal cost of production may be higher or lower than average variable costs. Each unit has a most efficient output, generally at 80-90% of net demonstrated capacity. Marginal costs of power fall as the unit produces more megawatts and rise after the optimal point. It is therefore advantageous to run any unit — if it is run at all — at its optimally efficient point, even beyond it if the next MW of power from this unit is cheaper than the next MW of power from the next-most-efficient unit. NIPSCO uses principles of “economic dispatch” to decide how much power to obtain from each unit in use (a “spinning” unit). Employees of NIPSCO run a control center, where computers show the demand on the system and the production of each unit. The computers (and the employees) know the marginal costs of each unit at different levels of output. As demand rises during the day, NIPSCO increases the output of the unit then on line that has the lowest marginal costs of producing the needed amount of power. (NIPSCO usually makes changes in 10 MW increments.) If necessary, NIPSCO starts up a new unit to cover demand. Because the load on the system and the amount of power generated must balance, NIPSCO also reduces the output of its units as demand falls. It cuts back on the most expensive power then being produced. All of these decisions are made at the margin, using costs that have been computed for each increment at each unit. NIPSCO used these principles of “economic dispatch” in 1982 and 1983, when NIPSCO and CWI were negotiating; CWI knew that this was how NIPSCO dispatched its spinning units. It is not possible, however, to add and subtract power wholly by reference to marginal cost. Units take time to build up power. Unit 15, for example, takes two hours to go from 200 MW to 472 MW. PX 42, p. 6. An increase in demand therefore may have to be filled from the unit that can get up to power first, rather than from the unit that can produce sustained power at the least cost. Moreover, once a unit is turned off altogether (as opposed to being reduced to its minimum generating level), it may take two days to bring the unit back on line. How long it takes depends on how far the internal temperature of the unit has fallen and the size of the unit. Large units can change temperature only so fast; to turn on the boilers full blast before the turbines have warmed up may cause fractures or worse. It is therefore very difficult to move large units on and off line frequently. NIPSCO, like other utilities, tries to keep on line the units it needs, even if this entails running them at inefficiently low loads. There is also a problem of reliability to consider. A unit on line may fail unexpectedly. A small shortfall may be reflected in a “brownout” (all customers receive power but at lower voltages, a form of rationing). When there is a larger shortfall, the utility must either supply the missing power quickly or “shed load” — that is, terminate service to some customers. Failure to shed load may lead to permanent damage to generation and transmission facilities. The customers whose loads have been “shed” do not appreciate this — especially if they are running blast furnaces or rolling mills that depend on reliable power. Because it is impossible to fire up new plants quickly when a plant on line fails, utilities maintain “spinning reserves”. Spinning reserves are the unused capacity of plants that are on line. These plants may be “spun up” to cover the loss from an unanticipated failure. One standard in the industry, to which NIPSCO tried to conform, is that the utility have enough spinning reserves to cover, within 10-30 minutes, the unexpected failure of the largest plant in the system. See Joskow & Schmalensee, Markets for Power 73. While spinning up its own reserves, the utility will buy “emergency power” from neighboring firms. NIPSCO is a member of the East Central Area Reliability Council (ECAR), a number of connected utilities that agree to supply power to one another in time of need. ECAR’s members collectively can cover a shortfall much faster than any one member can. The ability of NIPSCO to buy power from its neighbors plays another role in the case, to which I return. NIPSCO operated its system to obtain the most efficient output from the units then running, but in 1983 it did not decide which units to run so as to minimize costs. Suppose NIPSCO anticipated a load on a given day of 1,300 MW. The cheapest way to meet that load in 1983 would have been with 400 MW of power from I & M, 420 MW of power from Michigan City 12 (about 90% of its net demonstrated capacity), 288 MW from Bailly 8, 144 MW from Bailly 7, and 108 MW from Michigan City 2 and 3 (twin units with 60 MW maximum apiece). If NIPSCO anticipated a demand of 1,600 MW, it could put Schahfer 17 on line. If demand unexpectedly exceeded estimates, NIPSCO could turn on its gas-fired turbines, which come up to full power quickly; NIPSCO has 241 MW of capacity in these “peaking” units. It would not need the low-sulfur (and high-cost) units at Mitchell and Schahfer Stations unless the demand was anticipated to exceed 1,600 MW or one of the high-sulfur units were out of service. If NIPSCO ran its system this way, however, it would be unprotected against failure. In order to be able to replace the 420 MW of power it would lose if Unit 12 were to fail, NIPSCO needed 420 MW of “spinning reserves”. Only Schahfer Units 14 and 15 were large enough, in combination, to supply this. Schahfer 14, running at its minimum state of 220 MW, has 211 MW of spinning reserves (168 MW if NIPSCO plans to take it only to 90% of net demonstrated capacity); Schahfer 15, running at its minimum state of 200 MW, offers 272 MW of spinning reserves (225 MW if spun up to 90% of net demonstrated capacity). Schahfer 14 and 15 between them cover the requirements for spinning reserves, even when Michigan City 12 is running full blast. So NIPSCO ran them as of course when they were available. In NIPSCO’s parlance, Schahfer 14 and 15 — like Michigan City 12 and Bailly 8 — were “run-when-available” units. Unless they were down for service or repairs, these four units ran every day. Why only Michigan City 12, Bailly 8, and Schahfer 14 and 15? What happened to Bailly 7 and the other inexpensive, high-sulfur units? These units, and the cost savings they offered, were casualties of NIP-SCO’s quest for a reliable system. Once NIPSCO turned on Schahfer 14 and 15, it obtained 420 MW of power even at their minimum state. The generation and load of the system must balance. That meant turning off lower-cost units or reducing the output of (“backing down”) these units. It had 400 MW of firm power from I & M, 420 MW of power from Schahfer 14 and 15 (power that could be sloughed only by turning off the units), 420 MW readily available from Michigan City 12, and 288 from Bailly 8. That is 1,528 MW, with plenty of spare power spinning at Schahfer. Because average loads on the NIPSCO system are less than 1,528 MW, NIPSCO had to back down Michigan City 12 and Bailly 8. In other words, to ensure reliable power for its customers, NIPSCO was running the two most expensive plants in the system full time, and to “make room” for the costly power from these two plants, it had to run its two most efficient plants at less than their optimally efficient levels (and not run other low-cost plants at all). And this was on a system inefficient to start with because of the small average size, and old age, of its plants. By one estimate, NIPSCO’s marginal cost of power is approximately twice that of other utilities in the region. B The steel industry in northern Indiana, hard pressed to meet foreign competition, wants cheap electricity, which is a substantial component of the cost of finished steel. When the PSC granted NIPSCO a 20% increase (on top of its high base) shortly after CWI and NIPSCO signed their contract in the spring of 1983, the steel industry and other large customers sought ways to cut NIPSCO’s prices — or at least moderate future increases. Every quarter the PSC adjusts NIP-SCO’s prices to take account of changes in the cost of fuel. The steel industry opposed NIPSCO’s request for an increase in the fall of 1983, contending that NIPSCO operated its system inefficiently. The industry, joined by the PSC’s own consumer advocacy arm (the Utility Consumers Counselor’s Office), maintained that NIPSCO should not recover the higher costs of fuel, because it was neglecting a cheaper source of electricity — “economy power”. Economy power is electricity available from neighboring utilities on a minute-to-minute basis. All utilities maintain spinning reserves; all face fluctuations in demand. This means that a utility may be able to produce and sell, over the nation’s network of long-distance, high-voltage transmission wires, substantial quantities of electricity. NIPSCO was buying 400 MW of “firm” power from I & M ’round the clock, 365 days a year. I have already mentioned “emergency” power, which is costly but essential in times of need. Still another kind of power is economy power, which one firm will generate using otherwise-idle (but spinning) capacity for the benefit of its neighbor. A firm with a large plant able to make some marginal MW of power for, say, $15/MWH may sell it to a neighbor. If the buyer’s costs of generating the same power are $25/MWH (25 mills/KWH: mills/KWH being the usual pricing unit), both firms gain at any price between the two. Splitting the savings appears to be customary in the industry. For much of NIPSCO’s operating day, economy power is available at about 18-20 mills/KWH. Economy power is not guaranteed. The selling firm may withdraw it on 10 minutes’ notice. The buying firm therefore can purchase economy power only against spinning capacity. If the marginal megawatts of the highest-cost unit on line cost NIPSCO more to produce than economy power costs to purchase, NIPSCO (and its customers) can benefit if NIPSCO “backs down” its expensive plants and replaces the power with economy purchases. If the selling utility withdraws the power, NIP-SCO can spin up its plants to cope. Until early 1984 NIPSCO purchased economy power only to cover peaks. It covered its average load with its own equipment, even when its marginal cost of production was significantly greater than the cost of economy power. The PSC ordered NIPSCO to change its policy. An order entered on December 27, 1983 (the Economy Purchase Order or EPO) directed NIPSCO to buy more power, in order to reduce its costs of fuel. Although the PSC approved NIPSCO’s request for an increase to cover the cost of fuel, it also directed NIPSCO to file a report “detailing its efforts toward developing systematic procedures designed to facilitate to the fullest extent reasonably possible the location, acquisition and utilization of excess electricity from neighboring utilities to be utilized in supplementing NIPSCO coal generation.” NIPSCO sought rehearing of this order, on the ground that it would impair the reliability of its system. As NIPSCO saw things, it could not buy economy power by backing down the output of Michigan City 12 or its other inexpensive high-sulfur coal plants; marginal power was available from these plants at or below the cost of economy power. To save money, NIPSCO had to cut the output of its high-sulfur plants. But as these generally ran at minimum capacity and were used to provide spinning reserves, NIPSCO could not cut them back further without taking them off line— which would impair its ability to meet emergencies. On February 22, 1984, the PSC denied rehearing, making these comments: In our original Order in this cause, this Commission did not envision NIPSCO shutting down a generating facility and replacing its output with purchased power from a neighboring utility. The start up time necessary to bring the shut down facility back “on line” is so great that it would detract from the reliability of NIP-SCO’s service. However, we did envision NIPSCO decreasing its generation output in its more expensive generation facilities, to the degree that with reasonable notice the output from those plants could be raised to the maximum level. Order of Feb. 22 at 3. What the PSC described is physically possible but would not save much money, indeed might cost the consumers some money. The Order of February 22 envisions economy purchases against the output of Schahfer 14 and 15 and the Mitchell units. These were used, at more than minimum levels, only in peak periods. So perhaps the Commission was hinting only that economy power should be purchased in peak periods. But keeping the more expensive units on line just to be able to purchase economy power in peak periods might increase system costs, if it led to more of the expensive generation capacity being on line in slack periods. John Dunn, the vice president of NIP-SCO in charge of generation, decided that the only way to comply with the Economy Purchase Order and ensure a substantial savings was to change the dispatch criteria of the system. Instead of running Schahfer 14 and 15 whenever they were available, Dunn decided to run the entire system according to cost considerations. That meant putting Michigan City 12, Bailly 7 and 8, and Schahfer 17 on line as the principal units, and covering additional needs with the smaller stations at Michigan City before turning on the big units Schahfer 14 and 15. The new dispatch criteria reduced NIPSCO’s spinning reserves, which meant less reliability; it also reduced the percentage of NIPSCO’s power being generated by the low-sulfur units. The combination meant substantial savings — although, as things turned out, NIPSCO purchased only a little more power from outside than it had in 1982 and 1983. (It is not necessary to go into detail, and the figures, PX 231 p. 20, show only net purchases without isolating their nature. NIPSCO purchases non-economy power and also sells power to other utilities, so that net purchase figures may mislead.) Under the new approach, implemented in the spring of 1984, Schahfer 14 and 15 were run principally during high-load periods (weekdays and evenings) and when other units were off line for repairs. NIPSCO purchased economy power against the marginal output of Schahfer 14 and 15, when they were on line, and the Mitchell units when they were on line. NIPSCO declines to take more than 150 MW of economy power at a time (because it believes more than that would be too difficult to cover if withdrawn), and for reasons not fully explored in this case does not buy economy power unless it is available at 6 mills/KWH less than the cost of generation from NIP-SCO’s plants — which essentially means that NIPSCO will not reduce the output of its high-sulfur plants to buy economy power. The PSC reviewed NIPSCO’s performance in subsequent quarterly fuel-cost orders. In a sequence of orders through 1984 and 1985, the PSC praised NIPSCO’s compliance with the Economy Purchase Order. The Utility Consumers Counselor’s Office, which monitored NIPSCO’s compliance (including a review of the hour-by-hour purchase and dispatch decisions), also was satisfied. As in many administrative cases, the “law” is found not so much in the formal orders as in the day-to-day implementation; the regulated firm tries to comply with the views of the regulator’s staff. That process, much in evidence between NIPSCO and the PSC, led to the dispatch criteria NIPSCO has been using since spring 1984. The PSC was entitled to think that the order had done some good, since economy power is cheap relative to power generated by NIPSCO. NIPSCO’s change in dispatch criteria saved consumers even more money, relative to what economy purchases alone would have achieved. NIPSCO uses a computer model, discussed below, to project how its system will operate. Usually NIP-SCO estimates the demand, cost, and operating factors it will encounter and uses the model to predict, for example, how much coal it needs to buy. The model also can perform retrospective “what-if?” calculation. The operators feed the model with the actual data the system encountered (that is, actual costs of fuel, actual consumption of electricity, actual down-time of plants, actual prices of economy power) and change some other things (such as which plants are put on line). The model then estimates what would have changed had the system been operated in the hypothesized way. Frank Venhuizen, the head of computer modeling at NIPSCO, used the model to determine what would have happened after the EPO had NIPSCO started buying economy power (up to 150 MW, whenever it was 6 mills/KWH less than internal cost) and kept Schahfer 14 and 15 on run-when-available status. The model estimated that fuel costs at NIPSCO would have been $3.3 million higher in 1984, $8.4 million higher in 1985, and $19.5 million higher in 1986 than they were. Tr. 2272-73. This is a system-wide increase of 0.35 mills/KWH in 1984, 0.97 mills/KWH in 1985, and 2.37 mills/KWH in 1986. This despite the fact that, with Schahfer 14 and 15 on line more, NIPSCO would have purchased as much as 30% more economy power than it actually did. Tr. 2324. With Units 14 and 15 on run-when-available status, and no economy purchases, NIPSCO’s costs would have been even higher than those in Mr. Venhuizen’s calculations. So the EPO saved NIPSCO (and its consumers) a lot of money; Mr. Dunn’s rearrangement of the dispatch criteria at NIP-SCO in response to the EPO saved them even more; but the consequence was that NIPSCO used more high-sulfur coal and less low-sulfur coal. CWI and Carbon County Coal bore the brunt of this. Carbon County, which had a fixed-quantity contract, has recovered damages. CWI hopes to do as well. C Exhibit I to the contract came from NIP-SCO’s computer model of the long-run operation of its system. Called PROMOD, the model is the work of Energy Management Associates, Inc., of Atlanta, and is used by more than half of the industry for the same purposes for which NIPSCO employed it. The value of a model’s output depends on the data and assumptions fed in. Several witnesses testified, and I find, that NIPSCO used PROMOD in a reasonable way. NIPSCO made reasonable efforts to estimate future costs and demands, the important variables in the model. These estimates turned out to be wrong, but that is the nature of predictions. NIP-SCO consistently overestimated the growth of demand for its electricity; as demand fell in absolute terms, NIPSCO simply reduced its estimate of the rate of growth. But NIPSCO’s estimates were based Pn the best available data and a reasonable methodology. I essentially agree with the testimony of Eugene T. Meehan, an employee of EMA, and Kent P. Anderson, an economist, that NIPSCO made estimates and employed the PROMOD model in a commercially reasonable way. Not perfect, but reasonable and in good faith. See PX 182 —PX 193, comparing NIPSCO’s projections with other forecasters’. There is little reason to spell out the data and the conclusions, given the view I take below of the meaning of the Contract. Mr. Venhuizen produced Exhibit I in July 1982, using the best estimates NIPSCO had about future demand and costs. He assumed that NIPSCO would reduce its purchases from I & M during 1986; this turned out to be wrong (they did not begin to decline until 1987), but the error is unimportant. The modeling also included the assumption that Schahfer 17 would “go commercial” (produce electricity for commercial sale) in mid-1983 and that Schahfer 18 would follow in mid-1985. The latter assumption also proved to be a year off; it offset the mistaken assumption about curtailing the take from I & M. Three assumptions made in mid-1982 account for the large difference between PROMOD’s predictions and what happened. First, Mr. Venhuizen assumed that Schahfer 15 would be run when available, at least at minimum capacity, but that production above minimum would depend on the marginal costs of the units then on line. This assumption, reflected in the PROMOD designation “M” for “must-run”, treated Unit 15 as on line ’round-the-clock for about 325 days per year. (NIPSCO told PROMOD to assume some time for routine maintenance and other planned outages, and PROMOD would make probabilistic computations about the likelihood and duration of unplanned outages.) Unit 15 would burn more than 750,000 tons of coal per year when run round the clock at 200 MW. To the extent it was called on to supply more than 200 MW at peak hours, or the minimum state were more than 200 MW, it would burn more than this amount. (During most, if not all, of the time CWI and NIPSCO were negotiating this contract, NIPSCO treated Schahfer 15 as having a minimum capacity of 280 MW. I find that NIPSCO did not promise CWI that it would retain this minimum. Both minimum and maximum capacities of generating plants change over time, with adjustments, repairs, and the breaking in of parts.) In other words, Yenhuizen told PROMOD to assume that Schahfer would be run in the future just as it had been run in 1980-82, years when it burned about one million tons per year. The output of the PROMOD run that became Exhibit I was almost entirely dependent on this constraint and assumptions about how long the unit would be out of service for maintenance and unplanned repairs each year. The low estimate for 1983, for example, came from the assumption that Unit 15 would be taken off line for an extended period to make alterations to improve its heat rate. If Unit 15 had remained as a run-when-available (PROMOD M or must-run) unit, it would have burned close to the Exhibit I estimates. Two other assumptions were potentially important to the results. One was the rate of growth of demand for NIPSCO’s electricity, the other was the price of coal. Suppose Schahfer 15 were run only when the marginal cost of its electricity was the lowest in NIPSCO’s system. As the table early in this opinion shows, NIPSCO had a lot of capacity with lower variable costs than Schahfer 15, even once the I & M contract ended. Whether Schahfer 15 produced much electricity then would depend on how fast the system grew: if it grew quickly, the output from Unit 15 would be essential; if it grew slowly it would be a decade before Unit 15’s power was needed; and if demand shrunk, Unit 15 could be sold for scrap. Relative cost of coal also was important, for a similar reason: change the relative economics of the units, and you change which units run. If coal at Mitchell (and for Schahfer 14) suddenly became cheap relative to CWI’s coal for Unit 15, then Unit 15 would burn very little coal; reverse the relative prices, and you also reverse the effects. Kent Anderson performed a multiple regression analysis to determine how sensitive the predictions of the coal demands at Schahfer 15 were to assumptions about the must-run condition, the growth of demand, and the relative costs of coal. His conclusion, in outline, is that if the cost of coal and the growth of demand had been what NIPSCO thought in the summer of 1982 they would be, then removing the must-run designation for Schahfer 15 would have made relatively little difference to the estimate of the amount of coal it would burn, but that removing the M designation and reducing the demand forecast and changing the relative delivered prices of coal would make a great deal of difference. Demand on NIPSCO’s system fell substantially during the years in question, principally because the steel industry was closing plants and cutting output at plants that remained open. The decline of steel affected all the ancillary industries in Northern Indiana, as well as the size of the population (and thus residential demand). See PX 231 p. 16; Table at PX 175. I need not trot out the numbers in the regression. See Tr. 1053-86 and PX 237 (revised) — PX 243 (revised). I credit Mr. Anderson’s conclusions. NIPSCO itself generated a raft of PROMOD runs during and after the negotiation of the Contract. Every quarter it prepared a “base case” run, reflecting its best analyses at the time. “Base case” runs from 1982 through the first quarter of 1984 — with a single exception to be discussed later — show essentially the numbers that also appeared in Exhibit I. Mr. Dunn’s decision to revise the order of dispatch in spring 1984 was incorporated into PROMOD, and that, coupled with changes in price and demand assumptions that had been made during the interim, led to a big change in the PROMOD projections. These base case PROMOD projections then formed the basis of the annual estimates NIPSCO furnished to CWI under § 4.B of the Contract. Here are some selected PROMOD runs showing the nature of the change. The name of the run is its date. 832Q means “1983, second quarter” — NIP-SCO’s internal estimate 13 days after signing the Contract. All figures, which are from PX 256, are in thousands of tons. All runs are “base cases” (with the exception of the run that produced Exhibit I), and in each case designations such as BASE or BAS have been omitted. (See also PX 232-34 for graphic presentations.) Year Exh. I 822Q 823QB 1982 1,062 1,062 — 1983 887 887 665 1984 1,059 1,059 769 1985 981 981 1,027 1986 1,029 1,012 1,021 1987 1,042 995 995 1988 1,102 1,102 1,037 1989 1,179 1,180 956 1990 1,052 1,052 1,016 1991 1,117 1,117 1,026 1992 1,000 — 1,039 1993 1,000 Avg. 1,043 1,044 955 832Q 834Q 841QA 843QF2 811 817 1,010 945 824 605 929 1,037 1,000 709 988 877 589 233 985 940 790 252 997 954 1.021 548 994 946 1,005 614 933 950 1,077 738 998 919 1.021 654 1,004 1,028 1,126 662 — 1,024 1,116 690 965 949 957 571 This table shows that NIPSCO was using, for its own purposes, roughly the same numbers it furnished CWI. I conclude that NIPSCO acted in good faith in making available to CWI its best available estimates about Schahfer’s requirements, using the actual operation of Unit 15 at the time and the same assumptions NIPSCO was using for all other corporate purposes. Until PROMOD was revised (in mid-1984) to include both lower growth projections and an alteration of the assumption about how Schahfer 15 would be used, the average estimated burns continued to fluctuate around 1 million tons per year. D There is one curiosity in this table: run 823QB. This was the “budget base case” for fall 1982, the run NIPSCO used as the foundation for its 1983 corporate budget. The figures for 1985 and forward track the figures in Exhibit I, but the figures for 1983 and 1984 do not. How these figures came to be reveals something both about the operation of NIPSCO and the negotiation of the Contract. Thomas Howarth, now retired, was the vice president of NIPSCO for “gas operations and fuel procurement” between 1977 and 1983. He was the principal negotiator of the Contract for NIPSCO and was responsible for other contracts as well. On his head lay the responsibility for the coal pile on the ground at Mitchell Station, the $52 million the PSC had excluded from NIPSCO’s rate base. Mr. Howarth scorned fancy models like PROMOD. He testified at trial that he neither knew nor cared much about PROMOD; all he cared about was the day-to-day burn of coal. This is a strange attitude for someone in charge of negotiating long-term contracts, but Mr. Howarth is not on trial here. Neither Howarth nor any of NIPSCO’s other negotiators had any responsibility for the dispatch (or other actual operation) of NIP-SGO’s units. His responsibility was to have available coal that would burn cleanly in the units, when the operations people decided to run them. Howarth ordered coal, lots of coal, for the Mitchell station. Not all of it burned cleanly. What did not burn piled up, and up, and up. Stung by the exclusion from the rate base, Howarth tried to persuade the operations side of NIPSCO to burn up “his” coal pile. Since the operations people were having trouble with CWI’s coal at Mitchell, they were not cooperating. The operations staff started shipping CWI’s coal from Mitchell to Schahfer, where it worked; the rest of it lay idle at Mitchell. (NIPSCO ultimately moved the coal pile at Mitchell to Schahfer, and the coal pile at Schahfer to Mitchell.) Howarth, who could bear no more, fired off a memo (DX EE) directing the operations staff to burn the coal at Mitchell and cut back the burn at Schahfer — both to accommodate the extra generation at Mitchell and to await the arrival of more CWI coal at Schahfer 15, where Howarth did not think he had an adequate stockpile. Howarth had no authority to direct the operations staff to do anything. Like good bureaucrats, they ignored a memo from outside the chain of command. But a copy of Howarth’s memo reached Venhuizen, who turned it over to his staff. Venhuizen’s staff viewed Howarth’s “directive” as a problem in modeling. How could PROMOD be tweaked to simulate the burn of more coal at Mitchell and less at Schahfer? The staff ultimately reset the designator of Schahfer from M (= must run) to P (= run weekdays and weekday evenings, but not nights or weekends) and changed the assumed costs of coal. Joseph Almasy, one of Venhuizen’s subordinates, asked Howarth if this is what he had in mind; he apparently paid the request (like anything else about PROMOD) little heed and said yes (DX EL) — knowing (but not telling Almasy) that the operations staff had ignored him. Almasy and Venhuizen then ran a budget base case, using the Howarth assumptions because they were (they thought) the best available. So PROMOD, the victim of Garbage In, Garbage Out, dutifully reported a lower burn at Schahfer 15 in 1983 and 1984, but no significant change in later years (when the designation returned to M). Although no one at NIPSCO paid any attention to this, and apparently no one on the negotiation team thought about it, someone at CWI did. The record does not reveal how, but it is clear that CWI’s negotiating team learned of the estimate of 665 tons for 1983 and became concerned. CWI was contemplating a requirements contract and needed to know NIPSCO’s requirements. Moreover, the variance between the 887 tons of Exhibit I and the 665 tons was large enough to suggest a problem in PROMOD, and CWI wanted to know where these numbers were coming from. NIPSCO invited CWI to send people to Hammond, Indiana (NIPSCO’s headquarters) to see what PROMOD was. In January 1983 CWI sent one of the members of its negotiating team (who knew nothing about models and computers) plus Ron Rominiecki, the controller of General Coal Co., who knew something about models and computers but nothing about the negotiations. (General Coal Co., now called Westmoreland Coal Sales Co., is like CWI a subsidiary of Westmoreland Coal Co. During 1982-83 General Coal was the sales agent for CWI’s coal, and the final decisions about the contract were made by General and Westmoreland rather than by CWI.) By the time CWI’s people arrived, the error in the 1982 budget base case had been corrected. Almasy and Venhuizen showed CWI’s employees around and told them about the model. Almasy and Venhuizen, though, apparently did not know (or had forgotten) why 823QB had the glitch in 1983 and 1984; they told Rominiecki that the decline in projected burn was because Unit 15 would be on repair for a longer time than anticipated. And that turned out to be true — though serendipitously. Rominiecki returned to Philadelphia, General Coal’s headquarters, and made both oral and written reports of his visit. The written report (DX GH, PX 157) shows that the NIPSCO staffers were completely open with the CWI team and told CWI everything it wanted to know about the model. Rominiecki saw raw inputs and raw outputs, though he was not allowed to take anything with him. (Inputs and outputs contain commercially sensitive data, such as the prices charged by CWI’s competitors.) The record also contains Rominiecki’s handwritten notes before and during the visit (DX GF-1). I conclude on the basis of these notes and Rominiecki’s report that CWI knew everything about PROMOD that it wanted to find out, and had a solid working understanding of the model. Mr. Rominiecki knew that PROMOD gave only estimates, and he must have known that point estimates (1,022,000 tons is a point estimate) do not reveal an important piece of information: the degree of confidence that the estimate is correct and the probable error if it is not. The “confidence interval” around the estimates is small, so long as Unit 15 is run when available. Relax that constraint and the confidence interval is large indeed. See PX 197 and PX 198, putting standard deviation bands around PROMOD estimates run by NIPSCO itself over the course of a few years. Rominiecki could have requested, but did not, the information essential to allow him to compute the confidence intervals around the point estimates of Exhibit I. The output of BAS823QB (the full name of the fall ’82 run) is a centerpiece of CWI’s case. CWI observes that the numbers in this report diverged substantially from Exhibit I, yet NIPSCO did not furnish BAS823QB to CWI. Howarth never told CWI about these estimates or about his “directive” to change the burn. All of this is irrelevant, however. BAS823QB was modeling something that never happened. Computer models do not govern the burn at any unit. The operations staff pays no attention to the PROMOD outputs; the numbers are meant for budgeters and fuel buyers. PROMOD estimates what will happen when the operations staff dispatches the units in a certain fashion, and the units break down at predictable rates (though at unpredictable times). BAS823QB was a worthless estimate, because it did not reflect accurately how the operations personnel planned to run the system. Howarth, who ignored PROMOD from beginning to end, was not going to make anything of it; Robert Kelly, second in command of the negotiating team, apparently never heard of either Howarth’s directive or this run of computer output. Kelly and Howarth continued to believe throughout the negotiations that Exhibit I was NIPSCO's best corporate estimate. As it turns out, they were right (although the estimate was way off). By the time the contract was signed, BAS823QB had been scrapped, and the best estimate soon to be produced with data current on April 13 was remarkably close to Exhibit I. (The base case in use within NIPSCO on April 13, BASE831Q, is very similar to BASE832Q and therefore was not reproduced in the table.) CWI in fact had NIP-SCO’s best internal estimates. One final point about BAS823QB. CWI implies that NIPSCO clung to that estimate longer than NIPSCO claims. In the spring of 1983 NIPSCO applied for an increase in the rates at which it sold power to a co-op for resale at retail. These rates are regulated by the Federal Energy Regulatory Commission. NIPSCO filed its application, with supporting documentation, in March 1983. The documentation included the budget based on BAS823QB — which was, except for the juggling of the burn between Mitchell and Sehahfer, the best guess going. The FERC’s rules require the utility to give the actual data for a “historical test year” (NIPSCO chose 1981) and then the estimates for a projected future test year. These projections must be filed no later than 15 months after the close of the historical test year, here by the end of March 1983. NIPSCO obviously used the same data that underlay its 1983 budget, which it was submitting to FERC. NIPSCO had been gathering data for the projected test year for more than six months before NIP-SCO revised its budget for 1983 after correcting its PROMOD runs. NIPSCO had two alternatives: to submit the original 1983 budget warts and all (and from the point of view of NIPSCO, and its future rates, the error in BAS823QB was scarcely a pimple) or to recalculate everything, miss the March 1983 deadline, and have to start again with a new test year. The choice was clear: NIPSCO submitted the error and met the deadline. The customer (the Wabash Valley Power Association) then asked NIPSCO for its best available projections. In June 1983, NIPSCO, understanding this as a request for the data it had submitted to FERC, sent Wabash BAS823QB. Wabash could not have cared whether CWI coal was burned at Mitchell and Medicine Bow coal at Sehahfer, or the reverse; there is no indication that it was annoyed by the error. This is the last anyone heard of BAS823QB — until this litigation. I trust that BAS823QB, its place in Federal Supplement secure, can rest in peace. E Shortly after the parties signed the contract, Howarth retired. NIPSCO appointed a new team to procure coal and administer coal contracts. In the fall of 1983 one of the members of the new team told CWI that NIPSCO needed a reduction of $15 per ton in the delivered price of coal (which was then about $65 per ton). Tr. 1321-24, 1820-23, DX HX, DX IS. The reduction would be split evenly between CWI and the railroads hauling the coal. NIPSCO had opened negotiations with its railroads for a contract to carry coal (permitted under the Staggers Rail Act of 1980), and after extended negotiations finally achieved a significant reduction in the price of transportation. CWI takes this demand for a price concession as a sign of NIPSCO’s bad faith in determining its requirements. I do not so understand it. NIPSCO’s negotiators did what one would expect at a time when spot prices were falling and NIPSCO was under increasing pressure from its customers to reduce the price of fuel (fall '83, as I have recounted, marked the first coordinated opposition by the steel industry to a request for a fuel-adjustment increase). CWI had itself raised the possibility of a price reduction for sales over one million tons; NIPSCO could hardly be forbidden to explore the possibility of a reduction for other sales. It is not commercial bad faith to talk price with a supplier. CWI also maintains that even in the fall of 1983 Unit 15 was not being run when available, Tr. 1932, 2072-73, implying that NIPSCO changed the status of Unit 15 before the Economy Purchase Order. This is a misunderstanding. In the fall of 1983 Schahfer 14 was undergoing environmental shake-down tests, which required that it be run near full power. The figures I have already given show that NIPSCO could not run Unit 14 at full power and still have Unit 15 on line during slack periods. Mr. Dunn testified that Unit 15 was taken off the line for something short of 30 days while Unit 14 was running full blast. Accommodations of this sort would not affect the long-run burn of coal at Unit 15 and do not show knavery. A time doubtless would come when Unit 15 would be run full blast and Unit 14 would be turned off during the interim; these things balance out. That one such episode occurred in the fall of 1983 shows nothing. A final point about NIPSCO’s good faith in determining its requirements. Much of the delivered price of CWI’s coal is rail transportation. In April 1983 NIPSCO was paying about $29 per ton under a special tariff negotiated with the Denver & Rio Grande Western Ry., the carrier originating the movement at CWI’s Orchard Valley mine in western Colorado. The D & RGW negotiated with other railroads for their share of the transportation, paying them out of the $29. By 1986 NIPSCO had negotiated separate contracts with each carrier, getting them to bid against one another for larger portions of the movement. (It is possible to interchange the cars at different junctions, so the length of each carrier’s haul is in NIPSCO’s control, within limits.) CWI says that NIPSCO should have negotiated these contracts sooner, reducing the delivered price and increasing the burn of CWI’s coal. NIP-SCO replies that it was hard to negotiate with the D & RGW, which is the only carrier serving the Orchard Valley mine and therefore in a position to extract the consumers’ surplus of the whole movement. The parties and expert witnesses debated for some length whether NIPSCO did the best it could, or even acted prudently; the views are conflicting. Compare Tr. 2206-19 with Tr. 2293-2302. I need not resolve them. NIPSCO acted in good faith toward CWI. The contract with the D & RGW was only the third NIPSCO achieved; even today much of NIPSCO’s coal moves on tariffs rather than contracts. NIPSCO never got a contract to move the Carbon County coal from Wyoming to Indiana. Commercial good faith does not imply optimal conduct. NIPSCO’s ratepayers may regret the time it took NIPSCO to strike a deal with the railroads; alternatively, they may find that by holding out NIPSCO got better terms (as NIPSCO claims). Either way, CWI has no legitimate beef. F I conclude: 1. NIPSCO furnished in good faith, and CWI possessed, NIPSCO’s best estimates about Schahfer 15’s requirements. The estimates in Exhibit I were accurate as of the date the Contract was signed. CWI knew what PROMOD was and the fundamental assumptions that were responsible for the estimates in Exhibit I. 2. The estimates in Exhibit I arose from assumptions about demand for electricity, price of coal, and use of Schahfer 15 that were reasonable at the time. 3. John Dunn, who was and is responsible for dispatching NIPSCO’s system, intended throughout 1982 and 1983 to keep Schahfer 15 on run-when-available status for the foreseeable future. But for the Economy Purchase Order, Mr. Dunn would have done this to maximize the reliability of the supply of energy. 4. Dunn’s alteration of the dispatch criteria was a reasonable response to the EPO. Even though it was not the only possible response, Mr. Dunn’s response reduced the cost of generating power below any alternative that retained Unit 15 on run-when-available status. The PSC was delighted with NIPSCO’s changes and believed that they complied with the EPO. 5. Neither NIPSCO nor CWI foresaw the EPO or the alteration of NIPSCO’s dispatch criteria. Attorneys for both the Utility Consumer Counselors’ Office of the PSC and the industrial intervenors testified that the PSC had never entered any similar order. They, like Dunn, testified that the order surprised them. I conclude that everyone was surprised by the order. To the extent they should not have been surprised — because the high costs of NIP-SCO’s system put handwriting on the wall in light of the pressure from customers to cut costs and the statutory obligation to hold costs down, see Ind. Code § 8-1-2-42(d)(1) — CWI also should not have been surprised. In fact, for reasons covered in Part II, CWI was less surprised than NIP-SCO because Charles Mann, a consultant at Dames & Moore, told CWI about the risk that NIPSCO would find Unit 15 uneconomical to run in a different system of dispatch. No one is required, however, to predict administrative novelties. 6. After changing the dispatch criteria in the spring of 1984, Mr. Dunn operated Unit 15 in the same way as any other. NIPSCO did not manipulate its requirements at Unit 15 to take advantage of the contract with CWI. NIPSCO reduced its burn at Unit 14 by more than it cut the burn at Unit 15 (see PX 231 pp. 18-19) even though it had a take-or-pay contract for coal at Unit 14. (The Carbon County contract, for which NIPSCO has paid dearly.) Had NIPSCO been proceeding in commercial bad faith, it would have done the opposite, running Unit 14 while closing Unit 15. The reduction in the burn at Unit 15 occurred because of a combination of the change in the order of dispatch (a sine qua non), the decision to purchase economy power against the marginal production of the high-cost units, the reduction in the demand for NIPSCO’s power, and a decline in the price of spot, high-sulfur coal compared with CWI’s coal. None of these was anticipated in April 1983, and it was commercially reasonable not to have anticipated them. These findings mean that NIPSCO did not defraud CWI into entering the Contract or setting its price, and that NIPSCO has arrived at its requirements in good faith. See Peoples Trust Bank v. Braun, 443 N.E.2d 875, 877 (Ind.App.1983) (fraud means a representation known to be untrue when made, or an untrue representation recklessly made). II The circumstances that led to the reduction in the use of coal at Schahfer Unit 15 would be irrelevant if NIPSCO had a fixed-take (“take-or-pay”) contract. The Carbon County case establishes that. CWI insists that it has a take-or-pay contract with NIP-SCO, whether because Exhibit I establishes contractual minima (the contract claim) or because NIPSCO’s negotiators promised that the assumptions underlying the Exhibit I numbers would be unchanged and that CWI relied to its detriment on these representations (a promissory estoppel claim). Most of the time at trial was spent on the negotiating history of the Contract. I overruled NIPSCO’s objection, based on the parol evidence rule, to the entire inquiry; I also declined to hold that the limitation on warranties in § 25 of the Contract is such a clear “entire agreement” clause that extrinsic evidence was precluded.