In the

    United States Court of Appeals
                 For the Seventh Circuit
                     ____________________
No. 15-2760
KANSAS CITY SOUTHERN RAILWAY COMPANY, et al.,
                                   Plaintiffs-Appellants,

                                 v.

SNY ISLAND LEVEE DRAINAGE DISTRICT, a political subdivision
of the State of Illinois,
                                       Defendant-Appellee.
                     ____________________

         Appeal from the United States District Court for the
                     Central District of Illinois.
                No. 13-3144 — Richard Mills, Judge.
                     ____________________

      ARGUED APRIL 5, 2016 — DECIDED AUGUST 3, 2016
                 ____________________

    Before WOOD, Chief Judge, and BAUER and WILLIAMS, Cir-
cuit Judges.
    WOOD, Chief Judge. The Sny Island Levee Drainage District
(“the District” or “Sny”) was organized in 1880 in the Circuit
Court of Pike County, Illinois, to protect the District from
flooding and surface water runoff from the Mississippi River.
The Kansas City Southern Railway Company (“KC”) and the
2                                                    No. 15-2760

Norfolk Southern Railway Company (“Norfolk”) both oper-
ate main line railways over the Mississippi River Flood Plain
in the District. The District is permitted under state law to as-
sess properties within its territory in order to maintain the
levees. The Railroads have now sued the District for the sec-
ond time, alleging, as they did in the earlier case, that the Dis-
trict used an assessment calculation formula that discrimi-
nated against them in violation of the Railroad Revitalization
and Regulatory Reform Act (the “4-R Act”), 49 U.S.C. § 11501.
After a 12-day bench trial, the district court found for Sny. Its
finding was supported by the evidence, and so we affirm.
                                I
    The Illinois Drainage Code allows drainage districts to
levy three different types of taxes on entities within the dis-
trict: “original assessments,” “annual maintenance assess-
ments,” and “additional assessments.” 70 ILCS 605/5-1. Orig-
inal assessments are used for construction of new levees. The
District has had in place an annual maintenance assessment,
authorized by the Pike County Circuit Court, since construc-
tion of its original levees; there are approved periodic in-
creases. The annual assessment has been based on a per-acre
formula since its establishment in 1961; it is adjusted for infla-
tion. Additional assessments are permitted for maintenance
projects that exceed the annual maintenance assessment
budget.
                                A
   In 2009, facing a budget shortage, the District adopted a
new methodology—not based on acreage—for calculating the
annual assessment to be levied on interstate properties owned
by railroads, pipelines, and utilities. The new methodology
No. 15-2760                                                     3

was developed by Sny attorney David Human and Klingner
& Associates engineer James Powell. It purported to calculate
assessments based on the benefits the District conferred on
each property. The Railroads sued over the 2009 change, and
the case made its way to this court. We found the annual as-
sessment discriminatory because the District maintained per-
acre taxes for most types of property, reserving the benefit ba-
sis only for the railroads, pipelines, and utilities. Kansas City
S. Ry. v. Koeller, 

(7th Cir. 2011). The latter group
accounted for only 8 out of the 700 properties in the District.
    We found Human’s and Klingner’s numbers, which esti-
mated benefits for the railroads, pipelines, and utilities at $280
per acre, to be unreliable. They assumed property values
without any supporting evidence, implemented an unex-
plained 8% “capitalization rate,” made different assumptions
about flooding when calculating railroad benefits than pipe-
line and utility benefits, and ultimately “refined” the numbers
when they came out too high. 

    As a result, we enjoined the District from collecting the an-
nual assessment, which we found to be a tax within the mean-
ing of subsection (b)(4) of the 4-R Act, pursuant to the new
formula. 49 U.S.C. § 11501(b)(4). The relevant comparison
class, we found, was a “functional” one that included other
commercial and industrial properties but not agricultural or
residential ones. 

. While we acknowl-
edged that reasonable distinctions between different types of
property, such as improved and unimproved land, could sup-
port a taxation rate that reflects the difference, we also cau-
tioned that a discriminatory assessment is one that “imposes
a proportionately heavier tax on railroading than other activ-
4                                                    No. 15-2760

ities.” 

(quoting Burlington N. R. Co. v. City of Su-
perior, Wis., 

, 1187 (7th Cir. 1991)). The District
had not taxed other commercial and industrial properties pro-
portionally to the Railroads. To the contrary, it was charging
eight of the other entities in the comparison class nothing and
treating six as though they were agricultural properties. We
stated, “[i]f … the Drainage Code requires [the commission-
ers] to assess all property on a ‘benefit basis’ then [the] entire
scheme should reflect that.” 

    Since our July 2011 decision, the District has chosen not to
collect annual assessments from the Railroads at all, rather
than revert back to collection under the per-acre formula. It
has continued to collect its annual assessment from non-rail-
road and non-interstate properties in the District on a per-acre
basis.
                                B
    In 2011, with the 2009 case pending, the District began the
process for a one-time additional assessment. The Pike
County Circuit Court has the authority to approve, and his-
torically has approved, occasional “additional assessments”
in accordance with the Drainage Code, for things such as ex-
tra repair work, construction, enlargement or repair of pump-
ing plants, and the payment of legal obligations incurred by
the District. 70 ILCS 605/5-1. The circuit court approved an
additional assessment of $5,853,162 in December 2011, and
asked Sny to file an assessment roll showing how it would be
distributed. Sny again relied on Klingner to develop the as-
sessment roll. But this time, the engineers spent a full year cal-
culating the benefit amount for every property in the District.
Sny filed an assessment roll based on the new benefit calcula-
tions that distributed the assessment according to the benefit
No. 15-2760                                                       5

that each property obtained from the existence of the levee
and drainage works. After the Railroads and the Illinois Rural
Electric Cooperative filed objections, the circuit court denied
the Cooperative’s objection and approved the assessment roll
in July 2013 as to all properties except the Railroads, reserving
the question of the Railroads for the federal district court to
determine. The roll identified the tax on KC as $91,084.59 if
paid in one installment or $103,612.52 if paid in five annual
installments, and on Norfolk as $102,976.18 in one installment
or $117,139.71 in five.
   The federal district court took 12 days to evaluate the com-
peting expert evidence from the Railroads and the District, ul-
timately finding that the Railroads had not shown that the
District’s methodology was discriminatory. Kansas City S. Ry.
Co. v. Sny Island Levee Drainage Dist., 

(C.D. Ill. 2015).
    The Klingner engineers used five categories of property
for their benefit calculations: agricultural, residential, wet-
lands/recreational, not-for-profit, and commercial and indus-
trial. The latter included electric utilities, pipelines, railroads,
and other commercial structures. They adopted a software
program used by the Army Corps of Engineers called the Hy-
drologic Engineering Center Flood Damage Reduction Anal-
ysis (HEC-FDA), which predicts expected annual flood dam-
age to different types of properties. The program uses the
“Monte Carlo” simulation, which analyzes water flow
throughout a 50-year simulation period, based on historical
information, to determine the likelihood that different in-
stances and degrees of flooding will occur. Both the Army
Corps and the Federal Emergency Management Agency
(FEMA) use the Monte Carlo simulation for estimating flood
6                                                 No. 15-2760

damage. Klingner used replacement cost estimates that took
into account depreciation for structures that would be dam-
aged in flooding. They used water-surface profiles from a
2004 Army Corps study to estimate flood levels.
    Klingner engineers used depth-damage curves, which
chart damage in relation to water depth, to predict property
damage. For some categories of property, Klingner was able
to use published depth-damage curves, including for residen-
tial and some types of commercial properties. Its engineers
developed their own depth-damage curves for railroads,
pipelines, and utilities, even though there are published
curves for pipelines, and those curves indicate that floods do
not cause pipeline damage. The Klingner engineers knew
from personal experience during the 1993 levee breach that
flooding does cause pipeline damage. They did not calculate
any damage to local and municipal pipelines, however, find-
ing that those pipelines were far enough away from the river
that no damage would occur in a breach scenario. Their cal-
culations for agricultural properties focused on crops and did
not include structures, based on an assumption (backed by
testimony) that structures could be moved before a flood.
These three data points were put into the HEC-FDA system
to come up with benefit numbers.
    Klingner then reduced its original estimates of damage to
the Railroads by 75%, explaining that it was being “conserva-
tive” with the estimate because of the inherent uncertainties
of flood prediction. It maintains that 25% to 30% is the “mini-
mum” amount of damage that would occur to the Railroads
from flooding. Sny then assessed each property at 25.3% of its
theoretical benefit level by category.
No. 15-2760                                                  7

    The Railroads’ experts chose a different methodology for
assessing benefits. They used FEMA’s loss estimation soft-
ware, known as “Hazus-MH,” to estimate flood damage for
all types of property except the railroads. Hazus-MH esti-
mates loss based on flood depth, property location, property
value, property type, and accepted depth-damage curves. The
Railroads’ experts used county assessor fair-market values for
residential and commercial properties, and Hazus-MH de-
faults for pipelines and utilities. For flood depth, the Rail-
roads’ experts employed hydraulic modeling software from
the Netherlands Delft Institute of Hydraulics. Because Hazus-
MH has no established depth-damage curves for railroads,
the Railroads’ expert David Brookings walked the length of
the railroad and estimated damages using the hydraulic-
model estimated flood depths, based on his own experience
with water damage to railroad lines.
    The Railroads’ experts calculated the estimated hypothet-
ical flood damage to District properties from flooding if there
were no levees, minus the loss to the properties from flooding
with the levees present in the event of levee breach, times the
probability of such a breach. (One might say P (NL-WL) = B,
where P is the probability of a breach, NL is damage with no
levee, WL is damage with a levee, and B is the benefit of hav-
ing a levee.) This gave them what they considered a net ben-
efit. Ultimately, they came up with negative numbers for the
Railroads’ properties, suggesting that the presence of the lev-
ees would actually lead to more damage to the Railroads dur-
ing a flood than they would experience if there were no lev-
ees. This is because, in Brookings’s experience, floodwater
that rises at the same level on both sides of an embankment
(as he said it would from a flood absent any levees) causes
8                                                   No. 15-2760

less damage than floodwater rising on one side of an embank-
ment (as it would if there were a levee breach). The Railroads’
experts’ calculation of hypothetical benefits to agricultural
and residential properties contained an assumption that
farmers would continue to plant crops each year, even if
flooding occurred every year, and that residential property
owners would also continue to remain at their homes and re-
build every year.
   The district court found problems with the methodology
used by the Railroads’ experts. Given those problems, and its
positive findings about the credibility and reliability of the
District’s expert witnesses, it ruled that the additional assess-
ment was not discriminatory.
                               II
    The Railroads raise four arguments on appeal: (1) that the
district court did not properly apply the burden-shifting
framework from section 11501(b)(4); (2) that the court erred in
finding the assessment not discriminatory; (3) that the court
erred by admitting testimony from Sny’s engineers; and (4)
that the court erred by limiting the comparison class to other
commercial and industrial properties.
                               A
    We review de novo whether the district court erred in its
allocation of the burdens of production and persuasion. Ka-
wasaki Kisen Kaisha, Ltd. v. Plano Molding Co., 

, 361
(7th Cir. 2015), cert. denied, 

(2015). Subsection
(b)(4) of the 4-R Act, the catch-all provision, prohibits a state
from “[i]mpos[ing] another tax that discriminates against a
rail carrier providing transportation subject to the jurisdiction
No. 15-2760                                                    9

of the Board under this part.” 49 U.S.C. § 11501(b)(4). The Su-
preme Court has identified a two-step burden-shifting in-
quiry for determining whether conduct is discriminatory un-
der subsection (b)(4). See BNSF Ry. Co. v. Tennessee Dep’t of
Revenue, 

, 271 (6th Cir. 2015) (citing CSX Transp.,
Inc. v. Alabama Dep’t of Revenue, 

, 288 n.8 (2011))
(“CSX I”). The first step involves the plaintiff’s “establishing
a prima facie case of discriminatory tax treatment.” 

den then shifts to the defendant to offer a sufficient justifica-
tion. 

contend that the district court failed to fol-
low the two-step analysis. They are wrong. Because the dis-
trict court found that the Railroads did not demonstrate a
prima facie case of discriminatory tax treatment, there was no
need to move to step two. Even if the Railroads presented
enough evidence to suggest that discrimination was plausi-
ble, the district court was entitled to weigh the evidence how-
ever it deemed appropriate. This was a suit for an injunction,
not a matter that was eligible for a jury trial. Evidence tending
to show that discrimination is plausible does not automati-
cally constitute a prima facie case of discrimination.
    Even if the district court had explicitly considered the ev-
idence of discrimination first, as the Railroads think it should
have, and then considered the evidence from the District, the
outcome would have been the same. The court found that the
assessments were not discriminatory, and it would have
found so at the second step of the two-step process even if the
Railroads are correct that it failed separately to analyze the
first step. The court’s findings indicate that had it considered
the methodology discriminatory for any reason, it would
have found any such discrimination justified by the testimony
10                                                     No. 15-2760

and evidence presented that described the Railroads’ benefits.
We find no error of law.
                                 B
    The Railroads also attack the court’s conclusion that the
assessment was not discriminatory. We review the interpre-
tation of the law de novo, but the question whether discrimi-
nation exists is a factual one on which our review is for clear
error. See, e.g., Breneisen v. Motorola, Inc., 

, 704 (7th
Cir. 2011) (noting that “[a] district court’s interpretation of a
federal statute such as the FMLA is a question of law which
we review de novo”); United States v. P.H. Glatfelter Co., 

, 676 (7th Cir. 2014) (factual findings reviewed for
clear error). See generally Pullman-Standard v. Swint, 

, 287–88 (1982) (question whether discrimination existed is
a pure question of fact subject to clear-error review).
    The Railroads would like us to characterize the central
question in the case—whether the District’s methodology for
calculating the benefit to the Railroads and other properties
was valid—as a legal one, but it is not. We outlined the rele-
vant legal principles in Koeller, 

, where we held
that a per-benefit assessment system is acceptable under the
4-R Act as long as it is proportional and the system for calcu-
lating benefits is grounded in evidence. The only debate in the
present case is how that rule applies to the new facts.
     Where a lower court makes “detailed findings” based on
a choice between the testimony of conflicting experts, “[i]f
they find support in the evidence we are bound thereby” even
if the record would support contrary findings. Wahl v. Carrier
Mfg. Co., 

, 3 (7th Cir. 1966). That rule applies here.
The district court had to choose between Sny’s experts and the
No. 15-2760                                                   11

Railroads’, and it found Sny’s experts to be more credible and
accurate. There is ample support in the record for this finding.
    The Railroads would like us to find discrimination inher-
ent in the change from the per-acre to the benefit method and
the fact that the Railroads make up 0.3 percent of the land in
the district and paid 6 percent of the benefit assessment. At
oral argument, counsel for the Railroads stated, “our claim is
that we were discriminated against on a per-acre basis.” But
there is no reason to think that benefits bear a direct relation
to acreage, regardless of the use to which the land is put. And
we already have held that Sny is not obligated to conduct a
per-acre assessment. If the Railroads’ only claim of discrimi-
nation is that the District assessed properties within its pur-
view on a basis other than acreage, their claim fails. A benefits
assessment is discriminatory only if it does not assess all enti-
ties in the class in a proportionate manner according to benefit
obtained. The Railroads claim that Sny used the same meth-
odology as the one we enjoined from 2009. But Sny presented
evidence to the contrary, which the district court credited.
    The Railroads next say that the judge erred in accepting
Sny’s experts, because of five asserted flaws: (1) that the
Klingner engineers did not include damages to agricultural
structures; (2) that the engineers used original cost of con-
struction as the basis of calculating the value of such struc-
tures; (3) that the engineers used recent construction costs in-
stead of original construction costs for the Railroads; (4) that
the engineers made “assumptions of catastrophic damages
that were unsupported by hydraulic modeling, their personal
experience, or information or documentation from the 1993
flood”; and (5) that the engineers calculated projected flood
damage to the railroad right-of-way “as a percentage of the
12                                                    No. 15-2760

entire length of each rail line in the district instead of specify-
ing locations where damage might be expected.” Addition-
ally, they presented evidence that they believe indicated dis-
criminatory intent, including the similarity between the num-
bers arrived at in this assessment and the 2009 one. These sim-
ilarities, they say, suggest that the numbers were manipulated
to reach a “pre-conceived result.” Finally, they complain that
Sny treated local and municipal water and sewer facilities dif-
ferently from interstate pipelines.
    The Railroads charge that Klingner improperly alternated
between using original cost-to-construct values and adding
30 percent to those values as of the date of a remodel where
applicable, and that it used “different years of cost-to-con-
struct data for different categories of property.” They com-
plain that the engineers did not sufficiently show how they
arrived at their railroad damage estimates, and did not break
down unit costs by number of laborers, hours of labor, type
of equipment, or milepost locations where work would be
performed. And they contend that the 75 percent “discount”
and Sny’s acknowledgement about the uncertainties inherent
in predicting hypothetical flood damage reveal that Sny was
working backwards to desired numbers rather than accu-
rately calculating benefits.
    Sny responds that (1) Klingner did not include value of
agricultural structures because those structures would be re-
moved prior to a flood; (2) it used the original cost of construc-
tion because the structures were mostly old and in disrepair,
and this number was a reasonable estimate of replacement
cost; (3) it used more recent numbers for the Railroads be-
cause they undergo constant improvements pursuant to strict
No. 15-2760                                                   13

maintenance requirements and federal regulation; (4) the cal-
culated damage was based on reliable software that predicts
water events, combined with analysis from the 1993 flood (alt-
hough it appears that Sny included the estimates from the
Norfolk damages but did not weight the fact that the KC line
suffered no damages in 1993); and (5) their depth-damage
curve for estimating Railroad damages was reasonable.
   Sny defends the 30 percent for remodels as necessary to
get an accurate estimate of the value of updated structures.
Sny notes that it checked the replacement costs against the
Railroads’ own cost estimates. And it claims that the 75 per-
cent reduction was used to arrive at a “minimum” number.
Sny also points out that the district court had many valid rea-
sons for discounting the Railroads’ experts, who maintained
that the Railroads incurred a negative benefit from the levees.
    The district court heard and evaluated all of these points.
It concluded that although “there might be other ways of as-
sessing properties that are as good or better,” the approach of
the District’s engineers was “reasonable” and they were “gen-
erally credible in discussing [the] process.” Kansas City S. Ry.

. The court was “unable to con-
clude there was any discriminatory impact upon the Rail-
roads ... [t]he District used an appropriate method to deter-
mine that the Railroads’ assessments reflected a proportion-
ate and just share of the District’s levee protection costs.” 

no clear error in the district court’s assessment of this
battle of the experts. Its accompanying conclusion that the
Railroads had not demonstrated discrimination was therefore
also not error.
14                                                   No. 15-2760

                                C
    Another arrow in the Railroads’ quiver is an attack on the
court’s decision to admit the testimony of the District’s ex-
perts in the first place. They contend that it failed to use the
correct legal standard in doing so, and thus that the whole
case must be re-done. The Railroads claim legal error in the
district court’s failure to make an explicit statement on each
Daubert criterion (although more properly they should have
paid more attention to Federal Rule of Evidence 702, which
superseded Daubert many years ago): that the proffered ex-
pert testimony (1) be relevant and reliable, (2) be offered by a
qualified witness, (3) be based in a scientifically reliable meth-
odology, and (4) assist the trier of fact in understanding the
facts in issue. Daubert v. Merrell Dow Pharm., Inc., 

,
589 (1993).
    But the district court was not required to make these ex-
plicit findings before allowing the testimony, and so his fail-
ure to do so does not require us to review de novo his decision
on admissibility. Where a trial judge conducts a bench trial,
the judge need not conduct a Daubert (or Rule 702) analysis
before presentation of the evidence, even though he must de-
termine admissibility at some point. Metavante Corp. v. Emi-
grant Sav. Bank, 

, 760 (7th Cir. 2010). The Railroads
moved to exclude the District’s expert testimony, and the
court entered an order stating that it would defer considera-
tion of admissibility. In the end, it never issued a separate rul-
ing on the motion, but its opinion makes clear that it consid-
ered the admissibility of the expert testimony and ruled that
it could come in. The district court described all of the wit-
nesses’ qualifications in detail, and explained that their meth-
odology was consistent with FEMA and the Army Corps of
No. 15-2760                                                      15

Engineers. The district court did not commit legal error, nor
did it abuse its discretion, in admitting the testimony of Sny’s
experts.
                                D
    Last, the Railroads contend that the comparison class
against which their assessment should be measured is all
other properties in the District, instead of the narrower class
of commercial and industrial properties used by the district
court. They argue that Alabama Dep't of Revenue v. CSX
Transp., Inc., 

(2015) (“CSX II”), is inconsistent
with our 2012 decision in Koeller, which held that the appro-
priate comparison class for the Railroads was other RPU and
commercial and industrial properties. In CSX II, the Court
found that competitors with the plaintiff railroad were the ap-
propriate comparison class for a (b)(4) claim because the com-
parison class should be “based on the theory of discrimination
alleged in the claim.” 

It also stated that “all general
and commercial taxpayers is an appropriate comparison
class,” but “not the only one,” and that “all the world, or at
least all the world within the taxing jurisdiction, is [a rail-
road’s] comparison-class oyster” under (b)(4). 

The
Court clarified that the “similarly situated” requirement is not
as narrow in the 4-R Act as for tax claims under the Equal
Protection Clause. 

And the Court noted that rail-
roads cannot easily “hand-pick” their comparison class based
on CSX II, because “it is not easy to establish that the selected
class is ‘similarly situated’ for purposes of discrimination in
taxation.” 

    We read CSX II as emphasizing the importance of the
“similarly situated” requirement for subsection (b)(4) claims;
so understood, it is consistent with our “functional approach”
16                                                    No. 15-2760

in Koeller. Although Koeller’s holding is no longer sound to the
extent that it rejects all universal classes, the appropriate class
for the Railroads’ claim here is one that includes only other
“similarly situated” entities: the railroads, pipeline, and utili-
ties, plus commercial and industrial properties, excluding ag-
ricultural properties. CSX II does not allow the Railroads to
define their class as “all taxpayers in the district” simply by
alleging that they were discriminated against as compared to
all taxpayers in the district.
    The Railroads argue that because Sny takes the position
that it has assessed all types of property on the same basis, all
types of property are therefore similarly situated. But that
renders the “similarly situated” language meaningless: by
that logic, the class would always be all taxpayers, not simi-
larly situated taxpayers.
    The Railroads conceded at oral argument that they cannot
prove their discrimination claim other than by referring to a
universal class. Even that would not help them on this record.
The district court found, as a matter of fact, that the calcula-
tion of benefits to all types of properties was appropriate and
acceptable. We find no clear error in that assessment, and
thus no violation of the 4-R Act.
                                III
   Because the Railroads have not identified any legal error,
abuse of discretion, or clearly erroneous factual finding on the
part of the district court, that court’s judgment is AFFIRMED.