CRA Holdings US, Inc. and Subsidiaries v. United States Government

Filing 69

DECISION AND ORDER. The parties are directed to proceed to select a random sample of Plaintiffs' projects in accordance with the Decision and Order. Signed by Hon. Leslie G. Foschio on 8/9/2017. (SDW)

UNITED STATES DISTRICT COURT
WESTERN DISTRICT OF NEW YORK
________________________________________
CRA HOLDINGS US, INC. AND SUBSIDIARIES,

DECISION
and
ORDER

Plaintiffs,
v.

15-CV-239W(F)
UNITED STATES OF AMERICA,
Defendant.
________________________________________
APPEARANCES:

ZERBE FINGERET FRANK & JADAV
Attorneys for Plaintiffs
JOHN H. DIES,
JEFFERSON H. READ,
ROBERT G. WONISH, II, of Counsel
3009 Post Oak Boulevard, Suite 1700
Houston, Texas 77056
DAVID A. HUBBERT
Acting Assistant Attorney General
Tax Division, U.S. Department of Justice
Attorney for Defendant
JAMES M. STRANDJORD, of Counsel
PO Box 55
Ben Franklin Station
Washington, DC 20044

In this R&D tax refund case based on Plaintiffs’ claim that Plaintiffs, a national
environmental remediation firm, are entitled to recover $618,143 in taxes for Plaintiffs’
tax years 2002 and 2003, Plaintiffs’ claim is predicted on various amounts of R&D tax
credits to which Plaintiffs allege Plaintiffs are entitled pursuant to 28 U.S. C. § 41 (“§
41”) arising from Plaintiffs’ Qualified Research Expenses (“QREs”) incurred in
connection with 159 environmental remediation projects undertaken by Plaintiffs during
the tax years at issue. Since the initiation of the case, the parties have disputed the
possible use of a sample of the projects (reduced by Plaintiffs to 159 from 6,100) upon

which Plaintiffs’ amended refund claim is based to facilitate discovery and trial (or
summary judgment) on the merits as a practical means to minimize costs of litigation
and judicial time. The court resolved that issue in its Decision and Order filed January
26, 2017, Dkt. 46 at 14-16, CRA Holdings US, Inc. v. United States of America, 2017
WL 370811 at **5-6 (W.D.N.Y. Jan. 26, 2017), in which the court directed the parties
randomly select 40 projects as a pilot sample to assist the parties and the court in
determining whether the 159 project files were amenable to any sampling method, and
depending on the degree of similarity among the 159 projects, the proper size of such a
sample. The court also directed the parties propose a scheduling order for further
proceedings including potential trial in the case. In accordance with the court’s
direction, Plaintiffs’ proposed a stratified sample based on project size, i.e., contract
costs similar to that developed for the case by an IRS statistical advisor, see Dkt. 33-5,
during the audit process conducted in 2007 using the 40 projects comprising the pilot
sample to increase the likelihood that the resulting sample will be reasonably
representative of all 159 projects. Dkt. 64 at 2-3. Specifically, Plaintiffs suggest
randomly drawing three projects from each of four cost-based groups (strata) for a total
of 12 projects to constitute a random sample of approximately 8% of the 159 projects
which now form the basis of Plaintiffs’ claim. Id. at 3-5. Plaintiffs also submitted a
proposed scheduling order proposing, inter alia, an August 31, 2017 discovery cut-off
and a January 15, 2018 cut-off for dispositive motions. Defendant opposes any
schedule that does not allow for at least six-months of discovery. Dkt. 65 at 6.
In its response, Defendant rejected Plaintiffs’ proposed sampling method
proposing, instead, as a sample for both discovery and determination of the merits of
Plaintiffs’ claim, that the court approve a sample either comprising all 40 projects
2

identified in the pilot sample, or a smaller number of projects if selected by the court
from among all of Plaintiffs’ 159 projects (“the judgment sample”). See Dkt. 65 at 1, 2.
However, because Defendant asserts the amounts of QREs Plaintiffs had previously
submitted to the IRS during the agency’s review of Plaintiffs’ claim has resulted in
significant variations, when compared to data Plaintiffs submitted in this action, Dkt. 65
at 3, and in order to (1) definitively establish the total value of all QREs for Plaintiffs’ 159
projects asserted by Plaintiffs, (2) correctly calculate the amount of refund to which
Plaintiffs are entitled to recover based on the QREs established for the sample projects,
(3) establish Plaintiffs’ methodology for determining Plaintiffs’ QREs, e.g., “estimation,”1
and (4) calculate the “base amounts and fixed-base percentages that plaintiffs claim,”2
Defendant contends discovery directed to these fact questions for all 159 projects is
nevertheless required to assure the sampling process determined by the court does not
“relieve plaintiffs of their burden with regard to their claim for refund.” Dkt. 65 at 2 & n.
1, 5, and 8.
In reply, Plaintiffs state Plaintiffs agree that discovery may be directed to the
qualifications of Plaintiffs’ QREs, quantification of R&D expenses under § 41 for
qualified R&D activities, Plaintiffs’ methodology for determining the quantification of
Plaintiffs’ R&D expenses used to calculate Plaintiffs’ claims § 41 tax credits, and the

1

Plaintiffs state that Plaintiffs “did not calculate the qualified research project [sic] on a project basis,
CRA utilized a hybrid cost center approach as outlined in the original tax study [by Axiom, Plaintiffs’ R&D
refund consultant] and in response to [IRS] information requests during examination.” Dkt. 67 at 5.

2
Under § 41, the R&D tax credit is an amount equal to 20% of any excess of the taxpayer’s QREs for a
taxable, i.e., credit year, over a base amount, 28 U.S.C. § 41(a)(1), which amount is defined as the
product of the fixed-base percentage and the taxpayer’s average gross receipts for the four years
preceding the credit year, § 41(c)(1), but not less than 50% of the QREs for the credit year. § 41(c)(2).
The fixed-base percentage is generally the lesser of 16% of the percentage of the aggregate QREs of the
taxpayer for the base period years, or the aggregate gross receipts of the taxpayer for those years. See
§ 41(c)(3)(A) and (C). Suder v. Comm’r. of Internal Revenue, 2014 WL 4920724, at **12-13 (U.S. Tax Ct.
2014).
3

base amounts and fixed-base percentages Plaintiffs used to calculate Plaintiffs’ claimed
§ 41 credits. See Dkt. 67 at ¶ 1. However, Plaintiffs oppose Defendant’s demand that
such discovery should be directed to all 159 projects as inconsistent with Defendant’s
agreement to utilize a sample and the court’s indicated preference as the basis for
litigation on the merits of Plaintiffs’ claim, and contend such discovery should instead be
limited to those projects comprising the randomized sample as selected by the court,
preferably the stratified random sample drawn from the 40 project pilot sample. Id. ¶¶
7-9. According to Plaintiffs, Defendant’s insistence that some discovery or meritsdetermination be directed against and derived from all 159 projects contradicts the use
and benefits of any sampling technique the court may select to the extent agreed upon
by the parties. Id. ¶¶ 9-10. In particular, Plaintiffs object to any discovery directed to
the 159 projects which now form the basis of Plaintiffs’ tax refund claims for the purpose
of ascertaining the total value of the QREs Plaintiffs claim for all 159 projects as
unnecessary. Dkt. 67 ¶¶ 18-19. As Plaintiffs state, “[d]elineations of the amount of
qualified research expenses on a project-by-project basis across 159 projects is not
necessary to apply the results of a sample as long as the overall/total amount of
qualified research expenses, as claimed by CRA, is discernable.”3 Dkt. 67 ¶ 224
(underlining added). In an R&D tax credit dispute, the amount of QREs for a taxpayer’s
fixed-base period are stated on Plaintiffs’ Form 6765 and the amount of QRE claims for

3

Unless indicated otherwise, underlining is added.

4

The court notes the Amended Complaint, based on the 159 projects at issue, realleges in apparent
error, see Dkt. 51 ¶ 11, Plaintiffs’ allegations as stated in the Complaint that Defendant improperly
disallowed $700,463 in R&D tax credits for Plaintiffs’ 2002 tax years and $795,978 in such credits as the
Complaint alleged, see Dkt. 1 ¶ 11, based on all of the 6,100 projects which are no longer at issue in the
case.
4

Plaintiffs’ credit years as stated in Plaintiffs’ tax returns for the tax years at issue. Thus,
the asserted value of Plaintiffs’ QREs are known quantities.
Defendant’s assertion that discovery (and, presumably, trial) is required to
ascertain the total value of the Plaintiffs’ QRE claim in order to justify applying the
“results of a judgment sample,” Dkt. 65 at 2, is without merit. First, while similar to a
“statistical sample,” the type of sampling contemplated by the court in this case, use of a
“judgment sample,” may facilitate auditing of a taxpayers’ R&D tax credit claims, IRS
Field Directive On The Use of Sampling Methodologies in Research Credit Cases, Dkt.
57-4 at 1, and may provide a valid basis for ascertaining claims of R&D tax credits
generated by a taxpayer’s alleged QRE only if the resulting sample is representative of
all projects upon which Plaintiffs’ claim is predicated.

See Legal Guide to the

Research Credit § 8.8 (Westlaw 2017) (noting that “judgment sampling . . . relies on the
judgment of the person selecting the sample,” and, as such, lacks the “statistical
foundation” and “confidence levels” achievable by random sampling). “If a sample is
selected at random such that every member of the population [Plaintiffs’ 159 projects]5
has a known chance of being selected . . . then based on principles of probability and
statistics, it is possible to estimate characteristics of the population being sampled within
calculable ranges of accuracy.” Id. A representative sample, particularly a random
statistical sample of Plaintiffs’ 159 projects, obviates the risk that all or some portion of
Plaintiffs’ QREs associated with the projects are unsupportable and therefore cannot
satisfy each requirement of § 41. For example, in Suder, 2014 WL 4920724, at *20, the
court found, based on a review of a stipulated sample of 12 of 76 disputed projects, that

5

Unless indicated otherwise, bracketed material is added.
5

11 of the 12 projects constituted valid QREs resulting in a determination that 91.67%
(11/12ths) of the “76 projects constitute qualified research.” Id. Similarly, in the instant
case, requiring a random statistical sample – not a judgment sample – provides greater
assurance that the results of discovery and trial limited to the underlying facts of such
sample will yield results, i.e., compliance with § 41’s requirement and Plaintiffs’
substantiation of Plaintiffs’ expenses related thereto, that may then be applied to the
asserted QRE values of Plaintiffs’ 159 projects upon which Plaintiffs’ amended refund
claims are based. As noted, Plaintiffs do not oppose discovery, based on such a
sample, directed to Plaintiffs’ methods for determining Plaintiffs’ claimed R&D tax
credits including calculation of the required base period and fixed base percentage.
See Dkt. 67 ¶¶ 37, 38. Such sampling has been accepted by courts, see Bayer v.
United States, 2:09-CV-0035 (W.D.Pa. Mar. 23, 2015), Order of Court, Dkt. 210
(authorizing a partial trial to determine whether sampling should be used based on 10
randomly selected plaintiff’s costs centers),6 and the IRS for determination of tax credits
under § 41.
Further and contrary to Defendant’s supposition, Dkt. 65 at 2, use of the
statistical sample in this case excluding discovery to establish the actual amount of
Plaintiffs’ QREs for all 159 projects, does not relieve Plaintiffs of Plaintiffs’ burden to
establish the value of such QREs. Plaintiffs will nevertheless be required to establish
the value of Plaintiffs’ asserted QREs for each of the 12 sampled projects as a required
element of Plaintiffs’ claim and the actual percentage of such asserted values of the
QREs within the sample to the extent the evidence supports and is found by the court.

6

Cited in Legal Guide to the Research Credit § 8.11 n. 15. According to the docket for the Bayer case,
the case was recently, on July 17, 2017, settled without conducting the partial trial.
6

Thus, if the proof (assuming the other QRE requirements are met by Plaintiffs) shows
that the value of Plaintiffs’ QREs within the sample are only substantiated to the extent
of, for example, 50% of the amount of Plaintiffs’ claim, then Plaintiffs’ total QRE claims
for all 159 projects as reflected in Plaintiffs’ tax returns and pleadings will be similarly
reduced by the court. Defendant cites to no caselaw where such a hybrid approach, as
Defendant proposes, to development of the record and trial was required in an R&D tax
credit case nor statistical support from treatises or expert opinion on this issue.
Defendant’s latest objection, filed July 11, 2017, that palpable discrepancies in Plaintiffs’
QRE submissions raise the possibility of abuse by Plaintiffs absent discovery to
Plaintiffs’ asserted QREs for all 159 projects, Dkt. 68 at 2, overlooks that Plaintiffs will
be required to substantiate for the sampled projects every element required for a valid
QRE and that any discrepant data could adversely affect the court’s determination on
the merits. Defendant’s further objection that unless discovery is directed to all 159
projects Plaintiffs may attempt to wrongfully “shift claimed QREs” from projects Plaintiffs
are “unable to support” to others “outside the sample,” id. at 2-3, was addressed by the
court’s May 18, 2017 D&O (Dkt. 62) in which the court found Plaintiffs’ record keeping
system sufficiently to assure the integrity of Plaintiffs’ project records thereby minimizing
such risk to the integrity of the sample as Defendant theorizes. Accordingly,
Defendant’s contention that discovery and trial must be permitted for all 159 projects in
order to require Plaintiffs establish the actual amount of QREs alleged by Plaintiffs and
other issues pertaining to the validity of Plaintiffs’ asserted QREs for all such projects is
inconsistent with the principles underlying statistical sampling, judicial precedent, and
lacks merit. The court therefore turns to the appropriate sampling methodology for
discovery and trial in this case.
7

Plaintiffs request the court require use of 40 projects resulting from the pilot
sample as the basis for a stratified sampling method involving 12 projects. See Dkt. 64
at 3-5; Dkt. 67-1 at 2-4. Defendant opposes and requests limiting any sampling to the
pilot sample and argues that the 40 project sample or a random sample of the 159
projects be used as the basis to determine Plaintiffs’ claims. See Dkt. 65 at 4 n. 2.
Plaintiffs also point out that Defendant has received Plaintiffs’ discovery (interrogatory
responses) directed to the 40 projects comprising the pilot sample. Dkt. 67 ¶ 14.
Additionally, Defendant has not pointed to anything regarding the 40 projects sampled
on a pilot basis that would support a conclusion that for some reason they are not
representative of Plaintiffs’ 159 projects nor referenced any competent authority
advising against the use of the pilot sample to establish a preferred sampling technique.
Indeed, as noted, Defendant proposed as an alternative sample that the projects in the
pilot sample be used as the basis for a determination of whether Plaintiffs’ 159 projects
qualified as QREs. See Dkt. 65 at 1, 4 n. 2. However, the court finds that requiring all
40 projects as the basis for further litigation will significantly increase litigation time and
expense and burden to the court without significantly increasing the likelihood of valid
results. For example, it was necessary for the Tax Court in Suder to conduct a threeweek trial to determine the merits of respondent’s deficiency based on a sample of 12 of
petitioner’s 76 projects at issue. Dkt. 33-9 at 16, Suder, 2014 WL 4920724, at *2.
Guided by such experience, a trial in this case based on 40 projects may require more
than ten weeks. Moreover, a smaller size sample, 12 of 76, or 15.7% was found
acceptable to the court in Suder. See Suder, 2014 WL 4920724, at *6. Here, accepting
Plaintiffs’ proposal that 12 projects be sampled based on size, i.e., QRE labor costs,
from among the 40 projects forming the pilot sample will produce a sample size of 7.5%.
8

Such methodology achieves a reasonable balance between the need to proceed using
a representative sample, one similar to the stratified sample based on project costs
proposed by the IRS statistical consultant to facilitate the audit process in this case,7
and which does not impose unreasonable burdens on both the litigants and the court.
Accordingly, the parties shall proceed to establish a sample of 12 cases in accordance
with Plaintiffs’ proposal as the basis for further proceedings in this case. The court
recommends, that in doing so, in order to avoid further delay, the parties utilize the
same randomization software to which the parties agreed in generating the pilot sample.
A scheduling order governing further proceedings will be entered by the court.

CONCLUSION
The parties are directed to proceed to select a random sample of Plaintiffs’
projects in accordance with the foregoing.
SO ORDERED.
/s/ Leslie G. Foschio
________________________________
LESLIE G. FOSCHIO
UNITED STATES MAGISTRATE JUDGE
Dated: August 9, 2017
Buffalo, New York

7

Such sample was not ultimately used to determine the amount of Plaintiffs’ QREs because the IRS
audit determined Plaintiffs failed to establish accurately Plaintiffs’ base period QRE amounts. See Dkt.
65-5, IRS Form 888A, (concluding that “CRA failed to produce the required factual support for the
assumptions underlying its base period calculations.”), id. at 3.
9