With apologies to Twain and Disreali,[1] the title quote may come to mind when, in response to a defense letter outlining an ordinary course of business (OCB) defense comparing consistent and without-undue-collection-effort billing and payment practices in the “preference period”[2] with the period prior thereto (the “historical period”), the plaintiff responds with an OCB position based on a “standard deviation” analysis.
Standard deviation is a concept from statistics and, for most lawyers, best remembered (or forgotten) from that freshman-year statistics class. A standard-deviation analysis quantifies the variation or dispersion of a data set.[3] Picture a “bell curve” showing the grades of that freshman statistics class. It will chart those data points (the grades), most of which will be grouped closer to the mean of the data points, or center of the curve (the Bs and Cs), with “outliers” (As and Fs) on the ends. Deviations from the center, or mean, can be mathematically calculated and grouped as one or more “standard deviations.”
Standard-deviation concepts have been applied in legal contexts, albeit with warnings about their usefulness for the legal issues involved. The Seventh Circuit, in a case involving alleged racial discrimination under the Civil Rights Act, described standard deviation as:
[A] number that quantifies the degree to which disparities spread out above and below the mean of distribution, thus describing the probability that chance is responsible for any difference between an expected outcome and the observed outcome in a sample consisting of two groups (a binomial distribution). The greater the number of standard deviations, the less likely it is that chance is the cause of any difference between the expected and observed results. The Supreme Court noted in Castaneda v. Partida, 430 U.S. 482, 97 S. Ct. 1272, 51 L. Ed. 2d 498 (1977), that ... “[a]s a general rule for ... large samples, if the difference between the expected value and the observed number is greater than two or three standard deviations, then the hypothesis that the [disparity] was random would be suspect to a social scientist.” Id. at 497, n. 17, 97 S. Ct. at 1281, n. 17. The general rule should be applied with caution, however, and its particular applicability may vary from case to case. See D. Baldus & J. Cole, Statistical Proof of Discrimination § 9.03, at 293-97 (1980), 108-12 (Supp. 1984).[4]
In theory, then, preference plaintiffs might be tempted to argue that certain data points (i.e., payments of invoices by a debtor) that fall on the outer edges or further from the center of the curve (i.e., more than one standard deviation away) represent out-of-the ordinary transactions that should not qualify as OCB for preference-defense purposes.
Preference plaintiffs often are chapter 7 trustees, liquidation trusts or plan administrators. Because they might not have ready access to the witnesses (former employees that handled the debtor’s credit and payables issues) necessary to specifically rebut a defendant’s particular OCB defense, they might be inclined to resort to such statistical analyses, especially in the settlement-negotiation stage of a preference litigation, but also at trial.[5] In addition, the cost of the investigation, analysis and presentation to the court, through discovery processes or trial, that a fact-based, as opposed to statistically driven, OCB litigation may require might not make “cost-benefit” sense for an estate or a trust with limited resources (of time and litigation funds). The more mathematical, computer-program-driven statistical approach might be cheaper and easier, but might not completely and accurately depict a debtor’s business relationship with a vendor, and might not be a substitute for the type of OCB analysis that relevant case law suggests must be undertaken.[6]
Case law has dealt with standard deviation-type statistical analyses in myriad ways. An opinion by Bankruptcy Judge Mary Walrath in Delaware fairly recently considered various statistics-based arguments made by the chapter 7 trustee in response to the defendant’s summary-judgment motion based on OCB.[7] The court did not reach a substantive decision on whether the trustee’s analysis was relevant to, or otherwise helpful to the court in determining, OCB because a material issue of fact existed that could not be determined on the summary judgment record before the court with respect to the threshold issue of what period of time was the appropriate one to use for comparison of the historical period to the preference period.[8]
A few years prior to Powerwave Technologies, a decision from the Southern District of New York Bankruptcy Court accepted the chapter 7 trustee’s argument that transfers outside of one standard deviation from the mean were not OCB and thus avoidable, but it did so without substantial discussion about what a standard deviation analysis is, what it shows in the context of an OCB defense, or how it specifically fits into existing case law standards relating to OCB.[9]
Courts also have authored decisions that have simply noted that one of the parties submitted a standard deviation analysis in support of its positions but did not state explicitly it was adopting the analysis as support for its findings of fact or conclusions of law with respect to OCB.[10]
In a case from the Minnesota District Court, the defendants moved to exclude the testimony of a proffered expert witness on the issue of OCB in which, among other things, the witness’ position was that all payments received outside of one standard deviation from the mean of the Historical Period payments were not made within OCB.[11] This court, after noting it was not determining the weight to give the expert’s testimony, nevertheless denied the motion, ruling that arguments warranting exclusion of the testimony on the grounds of reliability had not been presented.[12]
So, what does a standard-deviation analysis really say about OCB, and might it actually present a misleading picture of the entire business relationship between a debtor and a vendor? Assume the following data set: ten invoices each are paid in the historical period in 32, 34, 36, 38 and 40 days, and the exact same data set exists for the preference period. At first, such data logically could prompt the argument and possibly the conclusion that the preference period payments were entirely consistent with the historical period, and thus, all were made in the OCB. However, a standard-deviation approach, mechanically applied to such data sets, counterintuitively could suggest the opposite.
The standard deviation of the above-assumed historical period data set — the data set against which the ordinariness of the preference period data set will be compared — is approximately 2.857.[13] Going back to the bell curve, one standard deviation on either side of the mean of such a curve would produce a standard deviation “range” of from 33.143 to 38.857, or perhaps if rounded to the nearest whole number, 33 to 39. That “range” — one standard deviation on either side of the mean — is the range the Waterford Wedgewood court held was the proper range for its OCB analysis. However, is it really the case that the 20 payments made in 32 and 40 days are not OCB in this scenario simply because they fell slightly outside of one standard deviation from the mean of the historical period data set?
The suggestion here is that a mechanical application of standard-deviation concepts does not address what an OCB litigation is designed to answer: whether an alleged avoidable payment was made by a debtor to a vendor consistently or inconsistently with its pre-preference period conduct? The very nature of what a standard-deviation analysis is suggests it does not provide the answer. At its core, a standard-deviation calculation, and the determination of one standard deviation, simply identifies about 68 percent of the data points in the set but does not say, other than for statistical purposes, how they relate to, or the ordinariness of, the other 32 percent.[14] Why should approximately 32 percent of the data points automatically be excluded from OCB because they fall outside of one standard deviation, especially in the absence of consideration of the specifics of the business relationship between the debtor and its vendor over the historical and preference periods?
Thus, although a standard-deviation analysis provides fodder for an argument that certain portions of a data set might not bear sufficient closeness to the majority of the data points (and thus, that they are not “ordinary”), it does not take any aspects of the debtor/vendor business relationship into account, or otherwise provide a rational basis to conclude that the payments made in 32 and 40 days in the above hypothetical were not made in the ordinary course of business.[15] It certainly does not appear that it satisfies the requirements of the cases interpreting § 547(c)(2)(A) that require an intense factual analysis of a particular business relationship.[16] A standard-deviation analysis seems clearly not to say much if anything at all relevant about whether the 32 percent of payments outside of one standard deviation are or are not OCB, except to the extent that a rote “bell curve” or “unsafe harbor” standard is to be applied, contrary to the established case law.
At bottom, it seems that use of standard deviation analyses in the subjective determination required by § 547(c)(2)(A) is not consistent with the case law that requires an intense fact inquiry into the entire business relationship between the debtor and its vendor.[17] And the standards for a § 547(c)(2)(A) determination consist of intensely factual inquiries, not one of which is determinative, including:
(1) the length of time the parties engaged in the type of dealing at issue;
(2) whether the subject transfers were in an amount more than usually paid;
(3) whether the payments at issue were tendered in a manner different from previous payments;
(4) whether there appears to have been an unusual action by the creditor or debtor to collect on or pay the debt; and
(5) whether the creditor did anything to gain an advantage (such as obtain additional security) in light of the debtor's deteriorating financial condition.[18]
This is not to suggest that a standard-deviation analysis might not have some utility if the defendant asserting OCB is relying on the objective, industry-standard determination under § 547(c)(2)(B) or that expert (or lay opinion) testimony about industry standards is not useful to a court. But most OCB defenses are based on (A) not (B), thus statistical analyses like standard deviation should be of limited utility in most preference litigation in the absence of tying them, if warranted, into the required all-facts-and-circumstances inquiry into the debtor/vendor business relationship.
Accordingly, when defending and trying to settle preference claims in which a subjective OCB defense is applicable, counsel for preference defendants should not be lulled into giving too much credit to a standard deviation (or other statistical) analysis in evaluating a plaintiff’s settlement demand — certainly not to the exclusion of the facts and circumstances of that particular case that control, and that may present a very different picture than the one the statistical expert is trying to portray.
[1] Sources of varying credibility suggest that Mark Twain popularized the phrase, but he gave credit for its origination to Benjamin Disraeli, one-time Prime Minster of Great Britain. But there appears to be no certainty that Disraeli ever uttered or wrote the phrase. According to Wikipedia and other authorities, the phrase “describes the persuasive power of numbers, particularly the use of statistics to bolster weak arguments,” but also may be used as a pithy response when one “doubt(s) statistics used to prove an opponent's point.” See Lies, Damned Lies, and Statistics, Wikipedia, https://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics (last visited Jan. 28, 2019).
[2] Under 11 U.S.C. § 547(b)(4), the 90-day period preceding the bankruptcy filing.
[3] See Standard Deviation, Investopedia, https://www.investopedia.com/terms/s/standarddeviation.asp (last visited Jan. 28, 2019).
[4] Coates v. Johnson & Johnson, 756 F.2d 524 (7th Cir. 1985).
[5] Other statistical analyses that preference plaintiffs often rely on, including “weighted average days late” and the like, are outside the scope of this article but are worthy of scrutiny before being proffered by plaintiffs or given too much weight by defendants.
[6] Some firms that routinely handle preference litigation have developed their own proprietary software that analyzes data in a way to guide both their settlement positions and trial presentations.
[7] Stanziale v. Superior Technical Resources Inc. (In re Powerwave Techs.), No. 13-10134 (MFW), 2017 WL 1373252 (Bankr. D. Del. Apr. 13, 2017) (“Powerwave Technologies”).
[8] The trustee in Powerwave Technologies also posed other statistically related OCB arguments such as “the Batch Method” (using a standard deviation analysis in connection with a transfer that paid several — a “batch” — of invoices) and the “DSO Method” (which considered the “dollar weighted average” of the days outstanding of the dollars paid by the debtor within the historical period and the preference period).
[9] In re Waterford Wedgewood USA Inc., 508 B.R. 821, 836-37 (Bankr. S.D.N.Y. 2014).
[10] See In re AE Liquidation Inc., No. 08-13031 (MFW), 2013 WL 5488476 (Bankr. D. Del. Oct. 2, 2013) (focusing on the ranges of invoicing and payments over the course of the parties’ business relationship rather than on standard deviation analysis); In re Consol. FGH Liquidating Tr., 392 B.R. 648 (Bankr. S.D. Miss. 2008) (focusing more on the actual facts of the parties’ business relationship rather than on the plaintiff trust’s proffered standard deviation analysis); and In re Rhodes Inc., No. 04-78434, 2008 WL 7836409 (Bankr. N.D. Ga. Dec. 1, 2008) (noting, simply and without explanation, that the standard deviation in connection with the parties’ historical period was a certain number but such does not appear to have been a major factor in the court’s decision on OCB, which instead turned on the ranges of the transactions at issue and the parties’ actual business relationship).
[11] Dietz v. Jacobs, No. CIV. 12-1628 JNE/JJG, 2014 WL 1153502 (D. Minn. Mar. 21, 2014).
[12] Id. at 2-8. A challenge to such proffered expert testimony beyond the actual calculation of standard deviations applicable to a particular data set still seems warranted on at least two grounds. Although the concept underlying a standard deviation calculation is not complex, and the calculation itself is fairly straightforward, higher-order math problems typically are not within the skill set of most lawyers. The calculations themselves very well could be the subject of stipulation. After the standard-deviation calculations have been made, however, is the application of such to OCB a proper subject for expert testimony? It seems still squarely for the court to decide, assuming it initially accepts a standard-deviation analysis as probative, whether a transaction inside or outside one (or two or three) standard deviations is or is not OCB, and an “expert” opinion on the application of a standard deviation calculation to the facts and circumstances of a vendor-debtor relationship could be inappropriate, as such would be tantamount to an expert applying the law to a set of facts, which is the job of the court and not a witness, statistics expert or not. Further, the question remains whether expert testimony actually would assist the court in considering OCB issues, in effect: does a standard-deviation analysis say anything much about OCB? As posited hereafter, this article suggests the answer is “not as much as plaintiffs may wish.”
[13] The calculation of standard deviation involves: (1) first determining the mean (the average) of the data set, which in this example would be 36, expressed as: 32 x 10 + 34 x 10 + 36 x 10 + 38 x 10 + 40 x 10 = 1800 ÷ 50 = 36; (2) squaring the difference between each number in the data set and the mean found in part (1), then adding such amounts (42 x 10 + 22 x 10 + 02 x 10 + -22 x 10 + -42 x 10 = 400); (3) then dividing this sum by the amount of the numbers in the data set minus one (400 ÷ (50 - 1) = ); and (4) finally, finding the square root of the number determined in part (3) (square root of 10 is ~ 2.857).
[14] See 68-95-99.7 rule, Wikipedia, https://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule (last visited Jan. 28, 2019); Statistics How To, https://www.statisticshowto.datasciencecentral.com/probability-and-stat… (last visited Jan. 28, 2019).
[15] For example, and certainly without limitation: (1) did the debtor and transferee routinely discuss payment of invoices; (2) did amounts reflecting goods sold or services provided on invoices require any business-level reconciliation or discussion and, thus, take more time for the invoice to be paid; and (3) in the data set assumed above, were the 32- and 40-day payments outside of one standard deviation because, perhaps, the 40-day payment was made over a holiday weekend and took a day or two extra to consummate, or perhaps the 32-day payment was “quick” because the transferor cut a check before the end of a monthly or quarterly accounting or tax period, or perhaps the transferee made a friendly visit to the debtor and was handed a check rather than receiving it in the mail?
[16] See, e.g., Fiber Lite Corp. v. Molded Acoustical Prod. Inc. (In re Molded Acoustical Prod. Inc.), 18 F.3d 217, 219, 220-226 (3d Cir. 1994); In re Tolona Pizza Products Corp., 3 F.3d 1029, 1033 (7th Cir. 1993).
[17] See, e.g., Burtch v. Revchem Composites Inc. (In re Sierra Concrete Design Inc.), 463 B.R. 302, 306 (Bankr. D. Del. 2010). See also Stanziale v. Industrial Specialists Inc. (In re Conex Holdings LLC), 522 B.R. 480, 489-90 (Bankr. D. Del. 2014), in which Delaware Bankruptcy Judge Christopher Sontchi rejected the trustee’s rote statistical approach based on “weighted average days late” in favor of a “holistic” consideration of the parties’ entire relationship.
[18] See, e.g., Burtch v. Detroit Forming Inc. (In re Archway Cookies), 435 B.R. 234, 241-42 (Bankr. D. Del. 2010); Hechinger Liquidation Trust v. James Austin Co. (In re Hechinger Inv. Co. of Del. Inc.), 320 B.R. 541, 548-49 (Bankr. D. Del. 2004).