810 F.2d 1395
UNIVERSITY COMPUTING COMPANY, Plaintiff-Appellant Cross-Appellee,
MANAGEMENT SCIENCE AMERICA, INC., and Larry Smart,
United States Court of Appeals,
March 4, 1987.
Robert F. Henderson, Allen, Knuths & Cassell, Dallas, Tex., for University Computing Co.
Barry F. McNeil, Haynes & Boone, Dallas, Tex., and Robert A. Bartlett, Atlanta, Ga., for Management.
E. Russell Nunnally, Alan B. Rich, Johnson & Swanson, Dallas, Tex., for Larry Smart.
Appeal from the United States District Court for the Northern District of Texas.
Before CLARK, Chief Judge, GOLDBERG and GARWOOD, Circuit Judges.
GOLDBERG, Circuit Judge:
Plaintiff University Computing Company (UCC) appeals an adverse grant of judgment notwithstanding the verdict (j.n.o.v.) in its action alleging breach of contract and tortious interference with contract. In particular, UCC challenges the district court's determination that the proof of lost profits at trial provided insufficient evidence of proximate causation between the breach or interference and the reduced profits. UCC argues that the evidence it adduced was sufficient to support a jury verdict, because exactitude of proof is not required in lost profit determinations.
Although we are very tolerant of the use of statistical proof to show prima facie evidence of causation, there are limits to what constitutes credible statistical evidence. Statistics is a mathematical discipline that requires a definite methodology. In this case, the "statistical" proof relied upon to make out a prima facie case is not capable reasonably of supporting mathematical inferences. Because the direct testimonial evidence of causation also fails to provide the degree of certainty required for recovery to be non-speculative, there was no triable factual issue to be decided by the jury. We thus affirm the decision of the district court.
UCC, through its Applications Software Division (ASD), manufactures software for sale to businesses and institutions. In 1982 and 1983, UCC's ASD "had its problems." Rec. Vol. 6, at 206. The division underwent management changes, and "was actively recruiting to find a person to head up [become a Vice President and General Manager of] the application area." Id. at 207. UCC thus contacted an executive search firm, Resource Dimensions, to conduct the search. One of the persons whom Resource Dimensions located was Larry Smart, who was then working for UCC's competitor, Management Science America, Inc. (MSA). Negotiations between Smart and UCC were conducted in the summer of 1983, and Smart signed an employment contract on Saturday, September 24, 1983.
Three days later, Smart notified UCC that he would not honor the contract, and that he would instead remain with MSA. UCC then hired Don Steele, in December, to head ASD. UCC subsequently sued Smart and MSA for breach of contract and tortious interference with contract, claiming that the failure of Smart to abide by the contract caused UCC to sustain substantial lost profits for the months of October and November. UCC did not claim damages for December, because Don Steele's presence partially served to remedy the problem, and thus UCC could not assess the degree to which the December "lost" profits were due to Smart and MSA.
At trial, UCC attempted to demonstrate a causal relation between Smart's breach and its lost profits, employing a "statistical" damage model developed by an employee of UCC, Paul Newton. Relying upon his familiarity with ASD operations (including his tenure as temporary head of ASD during the search period), and using figures showing the number of contract bookings during the October and November periods for the years 1979-1983, Mr. Newton reasoned: first, that something out of the ordinary must have occurred to cause the magnitude of lost profits in 1983; second, that the only unusual thing that he thought might have caused the lost profits was Smart's failure to honor the contract; and third, that Smart must therefore have caused the extraordinary losses. In particular, Newton posited that Smart's breach created the losses through "the fear and uncertainty and doubt and not being able to say, yes, I am going to be the permanent general manager, or here's where we are going, and so more from sales than anywhere else, but I generally faced those questions about future direction and what does this mean to us." Id. at 209. In short, Smart's failure to appear allegedly caused increased employee uncertainty and dissatisfaction, which in turn caused decreased productivity and increased costs, thus leading to decreased profitability.
After the jury returned a verdict, finding that Smart's breach of contract and MSA's tortious interference proximately caused UCC to lose profits, the district court granted j.n.o.v. to MSA and Smart. In particular, the judge found insufficient evidence under Boeing Co. v. Shipman, 411 F.2d 365 (5th Cir.1969) (en banc), to support the jury's verdict. In the alternative, the trial judge conditionally granted a new trial, were the j.n.o.v. to be overturned on appeal, because the verdict went against the great weight of the evidence. UCC appeals.1
Before assessing the sufficiency of UCC's evidence, we must canvass the applicable substantive law on proof of lost profits. In Southwest Battery Corporation v. Owen, 131 Tex. 423, 115 S.W.2d 1097 (1938), the Supreme Court of Texas established
[t]he rule as announced by the decisions of the courts of this state ... "In order that a recovery may be had on account of loss of profits, the amount of the loss must be shown by competent evidence with reasonable certainty. Where the business is shown to have been already established and making a profit at the time when the contract was breached or the tort committed, such pre-existing profit, together with other facts and circumstances, may indicate with reasonable certainty the amount of profits lost. It is permissible to show the amount of business done by the plaintiff in a corresponding period of time not too remote, and the business during the time for which recovery is sought. Furthermore, in calculating the plaintiff's loss, it is proper to consider the normal increase in business which might have been expected in light of past development and existing conditions."
The rule denying a recovery where the facts show that such profits claimed are too uncertain or speculative, or where the enterprise is new or unestablished, is still enforced, on the ground that the profits which might have been made from such business are not susceptible of being established by proof to that degree of certainty which the law demands.
It is impossible to announce with exact certainty any rule measuring the profits the loss for which recovery may be had. The courts draw a distinction between uncertainty merely as to the amount and uncertainty as to the fact of legal damages. Cases may be cited which hold that uncertainty as to the fact of legal damages is fatal to recovery, but uncertainty as to the amount will not defeat recovery. A party who breaks his contract cannot escape liability because it is impossible to state or prove a perfect measure of damages.
Id. at 1098-99 (citation omitted). The Southwest Battery standard clearly delineates two separate predicates for recovery: that the breach of contract or tortious act cause some damage; and that the amount of damage cannot be speculative, but must be ascertainable by competent evidence permitting reasonably certain assessment. Preexisting profit figures, combined with evidence of relevant "facts and circumstances," are sufficient to establish causation and the reasonably certain amount of damages.
Later Texas lower courts have embellished the reasonable certainty requirement. These cases have held that " 'where there is no basis for determining how much of the damages suffered resulted from the wrongful acts of the defendant and how much resulted from some other causes, a judgment would be based on mere conjecture and could not be upheld.' " Western Minerals, Inc. v. Hill, 441 S.W.2d 677, 679 (Tex.Civ.App.1969, no writ) (quoting Globe Aircraft Corp. v. Thompson, 203 S.W.2d 865 (Tex.Civ.App.1947, no writ)).2
We are required to look to federal law to assess the sufficiency of evidence to prove lost profits (according to state law substantive standards) in diversity cases. Texsun Feed Yards, Inc. v. Ralston Purina Co., 447 F.2d 660, 672 (5th Cir.1971) (citing Boeing, 411 F.2d at 368). The Boeing standard specifies that if "substantial evidence ... of such quality and weight that reasonable and fair-minded men in the exercise of impartial judgment might reach different conclusions, the motions [for directed verdict and j.n.o.v.] must be denied, and the case submitted to the jury." 411 F.2d at 368. Thus, the plaintiff must adduce substantial, competent evidence of a character that can permit reasonable men and women to determine that damage was caused by the breach or interference, and to assess with reasonable certainty the amount of damage and degree of causation of the damage by the breach or interference relative to other factors.
At first blush, UCC would appear to have adduced facts sufficient to present an issue for the jury. UCC provided evidence of recent year "profits" (actually contract bookings) for an ongoing and established business, of lower "profits" for the year 1983, and of a causal mechanism relating the breach and interference to the loss of profits. UCC's "statistical" model further provides evidence of the reasonably expected level of "profit" that might have been obtained, and of an apparent failure of other factors to cause the loss. It would seem that this evidence fulfills the verbal standards established by Southwest Battery for non-speculative recovery.3
Considered under the microscope of judicial scrutiny, however, UCC's proof appears defective and lacking in credibility. Although conceptually we should first separate UCC's prima facie causal showing from a determination of the amount of damages and relative degree of causation, the proof at trial precludes such segregation and collapses the neat, analytic framework established by the courts. UCC's identification of Smart as a cause was itself based upon UCC's ability to preclude other factors as substantial causes. The evidence of such preclusion, however, does not have an adequate statistical foundation to render it sufficiently credible to support the required inferences.
First, we are provided no information about where the "confidence intervals"4 for a sample of four data points5 (the 1979-1982 figures for October and November contract bookings) were obtained. Newton merely derived the mean6 and standard deviation7 for the sample, and listed the confidence intervals for the next point of sample data. We have no information about the relevant population from which the data points were drawn, and suspect that Newton confused populations with samples8 and the sample mean with subsequent data points. Even were Newton to have assumed a "normal" population, his attribution of confidence intervals to four data points likely violates the "Central Limit Theorem" of statistics, which holds, inter alia, that confidence in prediction is low given small samples of data (less than 30 data points). See generally J. Roscoe, Fundamental Research Statistics for the Behavioral Sciences (2d ed. 1975). Thus, as a matter of statistical inference, Newton's conclusion that "something out of the ordinary" had to have happened cannot be considered credible evidence.
Second, Newton's preclusion of other factors cannot be considered credible statistical evidence. Newton did not conduct regressions9 or otherwise provide a statistical test to show that other factors (identified as "considerations" that he rejected) could not have caused the degree of variation of the sample data points. Rather, he assumed that all ordinary factors did not contribute substantially to the resultant decrease in profits (in order to assess whether some extraordinary factor was involved), then mentally checked to see whether any factor had in fact changed sufficiently to cause the variation. To say the least, this does not comport with established norms of statistical methodology; such reasoning violates statistical logic, which first requires a determination as to whether ordinary events could have created the variation before looking to attribute unusual (or any) causes. Small influences can often result in large perturbations when conjoined; it is precisely these influences that adequate statistical tests are designed to assess.
Third, even were some "unusual" event responsible for the variation, Newton offered no direct methodology for testing his causal hypothesis. Although we cannot say that indirect proof (the exclusion of other factors) is illogical or impermissible, such proof requires us to assess the credibility and comprehension of the exclusionary method. Newton based his conclusions on cursory mental consideration of general data regarding market conditions and of operations in the division, and on his observation of the effects on ASD workers. Newton then reasoned that nothing but Smith's breach appeared sufficient to be a cause of UCC's lost profits, and thus that Smith must have been the cause, through the mechanism of employee disaffection. This opinion, although based on data analysis, simply does not provide a credible statistical basis for inferring causation. Thus, Newton and UCC did not provide sufficient prima facie statistical evidence to show a causal link, let alone to attribute the amount or relative degree of damages.
This does not end our analysis, however. We must decide whether Newton's opinion, based on his experience with ASD and his general familiarity with statistical and market data, is sufficient to constitute credible evidence from which reasonable inferences of a causal link, of lack of alternative causation by existing problems, and of lost profits can be drawn. Cf. Fed.R.Evid. 701-05. Common sense notions of causation do not easily permit us to infer a single cause when many factors appear as possible causes. Here, Newton's own testimony established that growth was sporadic, that ASD already was subject to internal staffing and morale problems, and that "profitability" had decreased in 1982. Notwithstanding this difficulty, Newton's testimony again initially appears sufficient to permit a reasonable inference with the requisite degree of certainty by a reasonable juror. The opinion evidence was rendered by one who was well-placed to perform such an analysis, even if the conclusions therefrom were wrong, cursorily made, or have but little credibility for a jury to assess. See Blue Chip Stamps v. Manor Drug Stores, 421 U.S. 723, 743, 95 S.Ct. 1917, 1929, 44 L.Ed.2d 539 (1975) ("We in no way disparage the worth and frequent high value of oral testimony when we say that dangers of its abuse appear to exist in [proof of hazy issues of historical fact] to a peculiarly high degree."), reh'g denied, 423 U.S. 884, 96 S.Ct. 157, 46 L.Ed.2d 114 (1975).
This conclusion is supported by existing case law on lost profit damages, permitting very tenuous causal claims to be sent to the jury. In Texsun Feed Yards, 447 F.2d at 672, we held that a breach of contract, which indirectly may have affected business operations, permitted a lost profit verdict based on scanty direct evidence of actual or relative causation.
Some of Texsun's customers removed their cattle from Texsun's yards because they were dissatisfied with the rate of [weight] gain [of their cattle due to Texsun's use of a new grain provided by Ralston, which breached their contract requiring adequately nutritious feed]. Texsun's customers demanded that Texsun extend cash refunds and account adjustment to compensate them for the additional feeding periods required to bring the cattle up to the desired marketing weight.
Texsun earned a $1,350.72 profit in the fiscal year ending September 30, 1966. The next year it showed a loss of $18,778.97 on lower sales. In fiscal year 1968 it suffered yet another loss. Possibly these losses were occasioned by factors other than the defect in Ralston's ration supplement. But the proved existence of these losses, coupled with the rebates, refunds, and adjustments which Texsun was forced to make to its customers in the spring of 1967, constituted a sufficient predicate for the jury's determination that Texsun suffered damage to the extent of $10,000.00 for lost profits and diminished business reputation. It was error to set aside that finding and enter judgment notwithstanding the verdict as to this issue.
Id. at 663, 672 (emphasis supplied). In Texsun Feed Yards, we did not discuss whether any evidence existed to demonstrate that Ralston's breach (which caused decreased weight gain by cattle, adding time and feed costs to customers, thereby obligating Texsun to provide rebates to customers who demanded refunds) was a cause of Texsun's "lost" business. Nor did we discuss whether the business was "lost" rather than simply "less." Instead, we must have relied upon the inherent nature of the breach to provide the relevant "facts and circumstances" from which a lost profit determination reasonably could be inferred. In short, given the adequate proof of the causal requirement for rebates, the indirect causal mechanism (decreased customer satisfaction and decrease in business) was simply a matter of common experience that could be presumed. The proof relied upon reasonable expectations of people's business behaviors.
Although this method of analysis may conflict with cases discussing proof of damages on markets, see, e.g., Baumstimler v. Rankin, 677 F.2d 1061, 1072-73 (5th Cir.1982); but see Household Goods Carriers Bureau v. Terrell, 417 F.2d 47, 53 (5th Cir.1969) (Terrell I), different results reached on reh'g, 452 F.2d 152 (5th Cir.1971) (en banc), Texsun Feed Yards adds an important dimension when considering sufficiency of causal evidence. It is only thus with regard to hidden assumptions about the inherent plausibility of causal mechanisms that we can assess the nature and amount of evidence required for "reasonable" inference. If a causal mechanism is inherently plausible--if it comports with customary experience--we can make the inference (that the mechanism in fact caused the harm) on less evidence than if the mechanism is less plausible.
Were the breach of contract in this case to support inferences that easily meshed with uniformly agreed-upon and authoritative assumptions about business behavior, we would have no difficulty with Newton's cursory hypothesizing. Common sense notions of causation and human behavior, however, more readily permit the inference that customers who must pay amounts exceeding their expectations are likely to demand rebates and to take their business elsewhere, than the inference that a failure to take charge of an office in disarray will lead to increased levels of employee uncertainty, thereby reducing the employees' abilities to obtain business. Although we might prefer to demonstrate through empirical evidence that our common sense is well grounded in fact, as a practical matter we have no alternative but to use our best current judgment. UCC's proof of the purported causal mechanism, therefore, must be supported by a greater amount of evidence than were that causal mechanism more closely matched with customary expectations and common-sense intuition.
Given these considerations, we hold that Newton's opinion is insufficient as a matter of law to permit a "reasonable" inference of causation. Although judgments by persons well placed to make determinations of causation usually constitute evidence capable of jury assessment, we cannot permit in the instant case a jury verdict of causation based upon such a shaky evidentiary foundation. Newton never directly inquired into the degree (if any) to which Smith's breach actually caused additional uncertainty or disaffection, nor the degree (if any) to which such uncertainty may have caused decreased productivity. Rec.Vol. 6, at 251 ("I didn't go out and ask people did you quit because of Mr. Smart's decision, or did you sit at home rather [than] go out and sell or so forth."); id. at 280 ("I think it would be impossible for me to determine any individual sale, whether it had been delayed, killed or--by that event [Smart's breach]."). The following colloquy shows the tenuous basis for Newton's nonstatistical conclusions:
A. Well, I certainly know people who were very concerned who I had talked with personally and they were considering leaving the company or whether they should stay or, you know, asking a lot of questions during that period about hitting the road.
Q. Do you know people that during that period weren't productive--I guess I'm talking about sales people.
A. Most of the sale force was not productive in terms of actually signing contracts, yes.
Q. But by productive, I mean were they doing their job? Do you know people that just weren't doing their jobs during October and November because they were distracted by Larry Smart?
A. I know people in conversations with me and others, and at least to that extent, they weren't paying full-time attention to their selling job.
Q. Did you fire any of the people that weren't productive?
A. I believe in November we dismissed Linda Moore for lack of production. I don't know of any other people right off.
Q. Was Linda Moore not productive because Larry Smart had not come?
A. I think in Linda Moore's case it had been some period of time that we had been dissatisfied with her performance, and generally, it takes--for us to dismiss someone, it takes a track record of at least several months of unacceptable performance.
Q. Do you know whether Linda Moore knew Larry Smart?
A. I have no idea.
Id. at 270-71.
Although his opinion is certainly credible, Newton's testimony is insufficient as a matter of law to establish causation. It is no more than a well-informed guess, and any inference drawn therefrom is merely speculative. Legal factfinding cannot be performed by hunch.
We hold today that the jury's verdict that Smart and MSA caused UCC to sustain lost damages is not supported by sufficient evidence. It is not surprising that we require in questionable cases a greater evidentiary showing than were the inference to be made obvious and intuitive; in fact, it is only reasonable to do so. Even were Newton's opinion to constitute sufficient prima facie evidence of causation, which it does not, there exist too many inadequately challenged, potential alternative causes to permit attribution, with any degree of certainty, of the amount and degree of lost profits (if any) that Smith and MSA caused. Although Newton's "considered" conclusions regarding causation may be the creative insights of a genius, they equally may be the imaginative speculations of a dolt. UCC's evidence simply does not amount to proof sufficient to permit non-speculative recovery.
Because we are not prepared to radically change longstanding judicial notions of causal proof, we can not impose liability on the basis of evidence as scanty as that provided by Newton, regardless of whether particular individuals (including Newton) consider the inferences drawn therefrom to be reasonable. See Boeing, 411 F.2d at 369-70 ("Federal courts must be able to control the fact-finding process by which the rights of litigants are determined in order to preserve 'the essential character' of the federal judicial system."). Generally, where we have permitted inferences to be made by a jury on evidence not substantially less certain than the evidence provided here, we have done so in the context of relative agreement in the relevant community regarding basic physical and behavioral assumptions. Such commonly-held assumptions are destined to change, as our knowledge of the world and as our behaviors change, but we can do no more than employ the collective wisdom of our time, as gleaned from society and as understood by the members of the judiciary charged with applying the law. That, after all, is the sine qua non of judicial decisionmaking. The judgment of the district court is thus AFFIRMED.
Because we affirm the grant of j.n.o.v., we do not reach UCC's challenge to the district court's conditional decision to grant a new trial
Federal law fully comports with Southwest Battery regarding proof of lost profits. See Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 562-63, 51 S.Ct. 248, 250, 75 L.Ed. 544 (1931). However, federal law conflicts with the later Texas cases requiring proof of relative causation as a prerequisite to recovery. See Terrell v. Household Goods Carriers' Bureau, 494 F.2d 16, 20 (5th Cir.1974) (Terrell II ) (the plaintiff "need not prove that the defendant's wrongful actions were the sole proximate cause of the injuries suffered. He must show only, as a matter of fact and with a fair degree of certainty, that defendant's illegal conduct materially contributed to the injury."), reh'g denied, 496 F.2d 878 (5th Cir.1974), cert. dismissed, 419 U.S. 987, 95 S.Ct. 246, 42 L.Ed.2d 260 (1974). The later Texas cases, however, do not appear to alter the Southwest Battery (or Story Parchment ) requirements for establishing a prima facie case of causation
Newton's conclusion that other factors did not cause the extraordinary magnitude of the losses also apparently fulfills the requirement for proof of relative causation. UCC only claimed damages for losses that, as a matter of "statistical" confidence, exceeded the amount of lost profits that could have resulted from competing causes. If UCC were correct, there could be no competing causes to provide any relative contribution to the damages claimed
See, e.g., D. Baldus & J. Cole, Statistical Proof of Discrimination 310 n. 39 (1980) ("Technically, the confidence interval tells one that if repeated samples were drawn from the universe of interest [population] in the same way as the instant sample was drawn, the means of the samples drawn would fall within the confidence interval a certain percentage, say 95 percent, of the time. On the basis of this information, researchers customarily conclude that the true mean of the universe [population] falls within the confidence limits.") (emphasis added)
A data point is simply a number on some relevant scale that reflects a characteristic of the object of statistical scrutiny. For example, if one inquires as to the height of individuals, the data points 60, 65, and 70 might represent, on a scale of inches, the heights of Curly, Mo, and Larry
A mean (or weighted average) is one method of obtaining an "average" value. Means are obtained by summing the data points on the relevant scale, then dividing that sum by the number of data points. The mean height of the sample in footnote 5 is 65 inches
A standard deviation is one method of determining the "variation" or "distribution" of a group of data points (from the mean). A standard deviation can be obtained by taking the square root of the following quantity: the sum of the squared differences from the mean (the sum of squares), divided by one less than the number of data points. A standard deviation is a dimensionless quantity. The sum of squares for the sample in footnote 5 is 50, one less than the number of data points is two, and the standard deviation is five
Cf. D. Baldus & J. Cole, Statistical Proof of Discrimination 305 (1980) ("Tests of statistical significance and confidence intervals are mathematical procedures which yield probability statements relating the numerical characteristic of a sample population with the characteristics of a larger population or "universe" from which the sample was drawn. Tests of statistical significance and confidence intervals always presuppose a sample and a universe from which the sample is drawn. Consequently, the strict interpretation of results produced by either method requires one to consider the applicability of that presupposition to the case at bar.") (emphasis in original)
A regression is one method of testing the relationship between two characteristics of the objects under statistical consideration. For example, one might regress the intelligence and the height of Curly, Mo, and Larry to see whether a "statistically significant" correlation existed. Only if such a correlation were shown would a statistician attempt to assess the existence of and to identify a causal mechanism between intelligence and height, or vice-versa. For a description of regression techniques, see J. Roscoe, Fundamental Research Statistics for the Behavioral Sciences (2d ed. 1975)