In the

       United States Court of Appeals
                  For the Seventh Circuit
                      ____________________
No. 15-1827
MIDWEST FENCE CORPORATION,
                                                   Plaintiff-Appellant,

                                  v.

UNITED STATES DEPARTMENT OF TRANSPORTATION, et al.,
                                   Defendants-Appellees.
                      ____________________

          Appeal from the United States District Court for the
            Northern District of Illinois, Eastern Division.
           No. 10 C 5627 — Harry D. Leinenweber, Judge.
                      ____________________

   ARGUED JANUARY 12, 2016 — DECIDED NOVEMBER 4, 2016
                      ____________________

   Before BAUER and HAMILTON, Circuit Judges, and
PETERSON, District Judge. *
    HAMILTON, Circuit Judge. Plaintiff Midwest Fence Corpo-
ration challenges federal and state programs that offer ad-
vantages in highway construction contracting to disadvan-
taged business enterprises, known as DBEs. For purposes of

   * The Honorable James D. Peterson, United States District Judge for
the Western District of Wisconsin, sitting by designation.
2                                                   No. 15-1827

federally funded highway construction, DBEs are small busi-
nesses that are owned and managed by “individuals who are
both socially and economically disadvantaged,” 49 C.F.R.
§ 26.5, primarily racial minorities and women, who have his-
torically faced significant obstacles in the construction indus-
try due to discrimination, § 26.67(a). Pursuant to the federal
DBE program, states that accept federal highway funding
must establish DBE participation goals for federally funded
highway projects and must attempt to reach those goals
through processes tailored to actual market conditions.
    Plaintiff Midwest Fence is a specialty contractor that fo-
cuses its business on guardrails and fencing. Because of its
size and specialization, it usually bids on projects as a subcon-
tractor. Midwest Fence is not a DBE. It alleges that the defend-
ants’ DBE programs violate its Fourteenth Amendment right
to equal protection under the law. Midwest Fence named as
defendants the United States Department of Transportation
(USDOT), the Illinois Department of Transportation (IDOT),
and the Illinois State Toll Highway Authority (the Tollway).
    Under the defendants’ DBE programs, government con-
tracting decisions may be made with reference to racial clas-
sifications, so these programs are subject to strict scrutiny.
They can survive an equal protection challenge only if the de-
fendants show that their programs serve a compelling gov-
ernment interest and are narrowly tailored to further that in-
terest. Adarand Constructors, Inc. v. Pena, 

, 227
(1995). Remedying the effects of past or present discrimina-
tion can be a compelling governmental interest. Shaw v. Hunt,

, 909 (1996), citing City of Richmond v. J.A. Croson
Co., 

, 498–506 (1989).
No. 15-1827                                                       3

    The district court granted the defendants’ motions for
summary judgment. Midwest Fence Corp. v. U.S. Dep’t of Trans-
portation, 

(N.D. Ill. 2015). We affirm. We join
other circuits in holding that the federal DBE program is fa-
cially constitutional. The program serves a compelling gov-
ernment interest in remedying a history of discrimination in
highway construction contracting. The program provides
states with ample discretion to tailor their DBE programs to
the realities of their own markets and requires the use of race-
and gender-neutral measures before turning to race- and gen-
der-conscious ones. The IDOT and Tollway programs also
survive strict scrutiny. These state defendants have estab-
lished a substantial basis in evidence to support the need to
remedy the effects of past discrimination in their markets, and
the programs are narrowly tailored to serve that remedial
purpose.
I. Legal and Factual Background
       The Federal DBE Program
    Because we review a grant of summary judgment, we base
our decision on facts that are either undisputed or reflect dis-
puted evidence in the light reasonably most favorable to the
non-moving party, Midwest Fence. Stevens v. Interactive Finan-
cial Advisors, Inc., 

, 739 (7th Cir. 2016); see also An-
derson v. Liberty Lobby, Inc., 

, 255 (1986). We lay out
in broad strokes the federal DBE program, which dates back
to 1983 and has been reauthorized as recently as 2015 in the
Fixing America’s Surface Transportation Act, Pub. L. No. 114-
94, § 1101(b), 129 Stat. 1312, 1323–25 (2015); see also Moving
Ahead for Progress in the 21st Century Act, Pub. L. No. 112-
141, § 1101(b), 126 Stat. 405, 414–16 (2012). After reviewing
substantial data, testimony, and studies regarding highway
4                                                  No. 15-1827

construction markets across the United States, Congress de-
termined that discrimination continued to “pose significant
obstacles for minority- and women-owned businesses seek-
ing to do business” in those markets. § 1101(b)(1)(A), 126 Stat.
at 415. Congress found there was a strong basis to continue
the DBE program to try to remedy the ongoing effects of dis-
crimination. § 1101(b)(1)(E), 126 Stat. at 415.
    The DBE program establishes a national goal of spending
at least 10% of federal highway funds in contracting with dis-
advantaged businesses. 49 C.F.R. § 26.41. DBEs are small busi-
nesses owned and controlled by socially and economically
disadvantaged individuals. See § 26.5. Women and racial and
ethnic minorities are presumed to be socially and economi-
cally disadvantaged, but they still must certify their disadvan-
taged status and provide economic evidence. § 26.67(a). The
presumption can be rebutted. § 26.67(b). Presumption or not,
no business can qualify as a DBE if the controlling owner’s net
worth exceeds $1.32 million or if the firm’s gross receipts for
the previous three fiscal years average more than $23.98 mil-
lion per year. §§ 26.65(b), 26.67(a)(2)(i).
    The federal program provides a framework for states to
implement their own programs. States establish their own
goals for DBE participation in federally funded transportation
projects by (1) determining the relative availability of DBEs
“ready, willing and able” to participate in those projects; and
(2) examining local conditions to adjust the base figure if nec-
essary. See § 26.45.
   The regulations require states to use race- and gender-
neutral means to the maximum extent possible to meet their
goals, providing a non-exhaustive list of techniques for pro-
No. 15-1827                                                    5

moting DBE participation. See § 26.51; see also § 26.5 (defin-
ing “race-neutral” to include gender-neutral). If a state cannot
meet its goal through neutral means, it must set contract-spe-
cific DBE subcontracting goals on projects with subcontract-
ing possibilities. § 26.51(d)–(e). The federal program expects
states to monitor their DBE participation continuously. If a
state is on track to exceed its DBE goal, it must reduce or elim-
inate contract goals as necessary, § 26.51(f)(2), and states may
adjust their overall goals at any time to reflect changed cir-
cumstances, § 26.45(f)(1)(ii).
    The federal program allows states to seek waivers of goal-
setting provisions, § 26.15, and permits states themselves to
decide on an individual basis whether bidders have made
good faith efforts to satisfy a specific contract goal when they
fall short, § 26.53(a)(2). Guidance as to what constitutes good
faith efforts is found in 49 C.F.R. pt. 26 app. A. The federal
regulations also require states to address overconcentration,
meaning that states must ensure that the use of DBEs in a par-
ticular sector does not unduly burden non-DBEs in that sec-
tor. 49 C.F.R. § 26.33. The federal program also requires peri-
odic reauthorization by Congress.
       IDOT’s Implementation of the DBE Program
   In exchange for federal funding, IDOT has implemented a
DBE program. State law requires Illinois to follow the federal
program for solely state-funded projects and federally funded
projects alike. 30 Ill. Comp. Stat. 575/6(d).
    IDOT uses a number of race- and gender-neutral initia-
tives to facilitate DBE participation in its contracts, including
a DBE mentoring program, a highway construction training
program, a Small Business Initiative to encourage smaller
6                                                   No. 15-1827

firms to participate in competitive bidding on prime con-
tracts, a Small Business Advisory Committee to provide IDOT
with input on small business issues, and an “unbundling” ef-
fort to reduce contract size so that a greater range of busi-
nesses can bid. IDOT also operates DBE resource centers,
works to eliminate barriers in the process required to qualify
firms to bid on prime contracts, hosts networking and indus-
try forums, and engages in numerous other efforts to draw
DBEs into the market.
    These race- and gender-neutral initiatives have never ena-
bled IDOT to reach its participation goal (which is 22.77%), so
it also sets individual contract goals on many contracts. In
compliance with the federal regulations, IDOT sets those
goals on contracts that have subcontracting possibilities. An
IDOT engineer identifies the types of work included in the
prime contract and estimates the value of each line item. An
IDOT compliance officer then determines whether at least
two DBEs are available to perform the work for each line item.
The compliance officer then adds up the estimated cost of the
line items and divides it into the total contract cost to set a
DBE participation goal for the contract. For example, if
$100,000 of line items on a $1 million contract could be per-
formed by DBEs, the contract goal would be 10%. IDOT’s Of-
fice of Business and Workforce Diversity examines the goal to
determine whether it is realistic.
    These individual contract goals are important but not
rigid. Prime contractors can still win contracts if they are un-
able to meet DBE participation goals. The bidding process re-
quires each prime contractor to submit a sealed bid accompa-
nied by a list of the DBEs the contractor will use to achieve the
DBE goal. The lowest bidder is awarded the contract if it
No. 15-1827                                                   7

meets the DBE goal or if IDOT determines that the bidder has
made good faith efforts to do so. In the latter instance, IDOT
grants a “front-end waiver” to the prime contractor. Whether
a front-end waiver is appropriate depends on a variety of fac-
tors, including the contractor’s method of soliciting DBEs, the
follow-up and assistance the contractor offered DBEs, work
with local officials, and so on.
    Price also matters. According to IDOT, contractors need
not use a more expensive DBE subcontractor if the price dif-
ferential is too great. The unwritten “rule of thumb,” accord-
ing to IDOT’s former deputy director of business and work-
place diversity is 5%.
   If a front-end waiver is denied, the prime contractor can
appeal to IDOT’s reconsideration officer. If reconsideration is
denied, the contract goes to the next lowest bidder or is rebid.
The practical availability of these front-end waivers is a focus
of Midwest Fence’s challenges to the Illinois program.
       The Tollway Program
    The Tollway adopted its own DBE program in 2005. Un-
like IDOT, the Tollway received no federal funding during the
relevant time, so it was not subject to the federal regulations.
For the most part, though, its program mirrors IDOT’s pro-
gram. The Tollway treats a business as a DBE if it is certified
under the federal regulations or if the City of Chicago or Cook
County deems it a “Minority or Women-Owned Business.”
   Like IDOT, the Tollway uses race- and gender-neutral
measures such as unbundling contracts, implementing a
Small Business Initiative, partnering with other agencies to
provide supportive services, conducting seminars on doing
business with the Tollway, and making information available
8                                                   No. 15-1827

on its website. Those measures have not produced substantial
DBE participation, however, so the Tollway also sets DBE par-
ticipation goals. It does so one contract at a time. Most of its
goals are achieved through subcontract dollars. It sets those
goals by comparing line items in its contracts to an availability
table in a 2006 National Economic Research Associates study,
which identifies DBE availability by industry code and serves
as evidence supporting the Tollway’s DBE program.
    In theory, at least, the Tollway grants good faith efforts
waivers based on criteria like those in the federal regulations.
These criteria include whether the contractor solicited DBEs
through all reasonable and available means, unbundled con-
tract work items into economically feasible units, provided
DBEs with sufficient information, engaged in good faith ne-
gotiations with DBEs, and so on. Cf. 49 C.F.R. pt. 26 app. A
(federal guidance concerning good faith efforts). Midwest
Fence again disputes the availability of those waivers in prac-
tice, calling the Tollway’s contract goals de facto quotas.
       Procedural History
   Midwest Fence filed this lawsuit in 2010 alleging that the
federal DBE program, IDOT’s implementation of it, and the
Tollway’s own program violate its Fourteenth Amendment
right to equal protection. Against the federal program, Mid-
west Fence asserts the following primary theories:
       •   The federal regulations prescribe a method
           for setting individual contract goals that
           places an undue burden on non-DBE sub-
           contractors, especially certain kinds of sub-
           contractors, including guardrail and fencing
           contractors like Midwest Fence.
No. 15-1827                                                    9

       •   The presumption of social and economic dis-
           advantage is not tailored adequately to re-
           flect differences in the circumstances actu-
           ally faced by women and the various racial
           and ethnic groups who receive that pre-
           sumption.
       •   The federal regulations are unconstitution-
           ally vague, particularly with respect to
           “good faith efforts” to justify a front-end
           waiver.
    Midwest Fence also asserts that IDOT’s implementation of
the federal program is unconstitutional for essentially the
same reasons. Finally, Midwest Fence challenges the Tollway’s
program on its face and as applied. Midwest Fence sought de-
claratory and injunctive relief from all defendants, and mon-
etary relief from IDOT and the Tollway.
   Both sides moved for summary judgment, and the district
judge granted the defendants’ motions. Midwest Fence 

. The judge found that Midwest Fence
had standing to bring most of its claims. 

On the
merits, the district court upheld the facial constitutionality of
the federal DBE program. 

    Turning to IDOT, the court concluded that Midwest Fence
did not rebut the evidence of discrimination that IDOT of-
fered to justify its program, 

and that Midwest Fence
had presented no “affirmative evidence” that IDOT’s imple-
mentation unduly burdened non-DBEs, failed to make use of
race-neutral alternatives, or lacked flexibility, 

  The court noted that Midwest Fence’s challenge to the Toll-
way’s program paralleled the challenge to IDOT’s program.
10                                                    No. 15-1827

that the Tollway, like IDOT, had established a
“strong basis in evidence” for its program. 

Finally,
the district court concluded that, like IDOT’s program, the
Tollway’s program imposed a minimal burden on non-DBEs,
employed a number of race-neutral measures, and offered
substantial flexibility. 

Midwest Fence has appealed
the final judgment. We have jurisdiction under 28 U.S.C.
§ 1291.
II. Standing
       Standing to Challenge the DBE Programs Generally
    Before addressing the merits of Midwest Fence’s equal
protection claims, we must address the defendants’ argument
that Midwest Fence lacks standing under Article III of the
Constitution to bring them at all. See Northeastern Florida
Chapter of the Associated General Contractors v. City of Jackson-
ville, 

, 663 (1993) (standing is “an essential and un-
changing part of the case-or-controversy requirement of Arti-
cle III”), quoting Lujan v. Defenders of Wildlife, 

, 560
(1992). The district court found that Midwest Fence has stand-
ing, with one narrow exception.
    Our standard of review depends on the procedural pos-
ture of the case, whether we are dealing with bare allegations
in the pleadings, a paper record of evidence on summary
judgment, or evidence at trial. The party invoking federal ju-
risdiction bears the burden of establishing standing, and the
elements of standing must be supported with the quantum of
evidence required at each successive stage of litigation. “At
the summary-judgment stage, ‘the plaintiff can no longer rest
on … mere allegations, but must set forth by affidavit or other
evidence specific facts.’” Edgewood Manor Apartment Homes,
No. 15-1827                                                    11

LLC v. RSUI Indemnity Co., 

, 771 (7th Cir. 2013)
(internal quotation marks omitted), quoting Defenders of Wild-

; accord, Six Star Holdings, LLC v. City of
Milwaukee, 

, 801–02 (7th Cir. 2016); Hummel v. St.
Joseph County Board of Comm’rs, 

, 1016 (7th Cir.
2016). Here, defendants moved for summary judgment in part
on the basis that Midwest Fence lacks standing, so the legal
issue is whether Midwest Fence offered evidence that could
support a finding of standing.
    To invoke a federal court’s jurisdiction, a plaintiff must al-
lege and then show (1) injury in fact, (2) a causal relationship
between the injury and the challenged conduct, and (3) a like-
lihood that the injury will be redressed by a favorable deci-
sion. Northeastern 

–64. Northeastern Flor-
ida involved a race-conscious ordinance enacted to increase
participation in public contracting. The issue was whether
standing requires a plaintiff challenging such a law to show
that it would have been awarded the contract but for the ex-
istence of the race-conscious law. The Court said no:
       When the government erects a barrier that
       makes it more difficult for members of one
       group to obtain a benefit than it is for members
       of another group, a member of the former group
       seeking to challenge the barrier need not allege
       that he would have obtained the benefit but for the
       barrier in order to establish standing. The “in-
       jury in fact” in an equal protection case of this
       variety is the denial of equal treatment resulting
       from the imposition of the barrier, not the ulti-
       mate inability to obtain the benefit.

(emphasis added).
12                                                  No. 15-1827

    The plaintiff “need only demonstrate that it is able and
ready to bid on contracts and that a discriminatory policy pre-
vents it from doing so on an equal basis.” 

redressability follow from this definition of injury: causation,
because the theory is that the policy prevents equal competi-
tion; redressability, because invalidating the policy will again
place the plaintiff on equal footing for competitive purposes.
See 

n.5.
    The district court correctly found that Midwest Fence has
standing under the Northeastern Florida approach. By alleging
and then offering evidence of lost bids, decreased revenue,
and difficulties keeping its business afloat as a result of the
DBE program and its inability to compete for contracts on an
equal footing with DBEs, Midwest Fence showed both causa-
tion and redressability. Midwest Fence Corp. v. U.S. Dep’t of
Transportation, No. 10 C 5627, 

, at *5–6, *9
(N.D. Ill. June 27, 2011); Midwest Fence 

.
    Defendants contend that our decision in Dunnet Bay Con-
struction Co. v. Borggren, 

(7th Cir. 2015), defeats
Midwest Fence’s standing here. While some language in Dun-
net Bay might be read that way, the standing holding in that
decision turned on an unusual and complex set of facts under
which it would have been impossible for the plaintiff to have
won the contract it sought and for which it sought damages.
Dunnet Bay’s low bid on an IDOT highway resurfacing con-
tract had been rejected for failing to meet the contract’s DBE
goal, but its bid was also 16% above the program estimate.
Moreover, Dunnet Bay had been inadvertently omitted from
IDOT’s authorized bidder list, an omission that may have cost
it some DBE subcontractor quotes. 

692. Under
No. 15-1827                                                  13

those circumstances, IDOT chose to rebid the contract. 

Since IDOT “did not award the contract to anyone under
the first letting and re-let the contract,” Dunnet Bay “suffered
no injury because of the DBE program in the first letting.” 

    This case is readily distinguishable from Dunnet Bay. Here,
Midwest Fence seeks prospective relief that would enable it
to compete with DBEs on an equal basis more generally. That
is sufficient under Northeastern Florida. Any suggestion in
Dunnet Bay to the contrary would be inconsistent with North-
eastern Florida and must be read as dicta. We need not consider
here the sorts of details of particular bidding procedures for
particular contracts as in Dunnet Bay.
       Standing to Challenge the Target Market Program
    Although the district court determined that Midwest
Fence has standing generally, the court carved out one narrow
exception, finding that Midwest Fence lacks standing to chal-
lenge the IDOT “target market program.” That program was
set up “to remedy particular incidents and patterns of egre-
gious race or gender discrimination.” 20 Ill. Comp. Stat.
2705/2705-600. Under the program, certain contracts may be
designated “target market contracts” to remedy that discrim-
ination. Target market remedial measures can include reserv-
ing certain work for performance by minority- or female-
owned businesses; implementing formation and bidding pro-
cedures that encourage bidding by DBEs; setting separate
participation goals for a particular contract or contracts; es-
tablishing incentives for achieving those goals; and setting
aside particular contracts exclusively for DBEs. 44 Ill. Admin.
Code § 6.830(a).
14                                                   No. 15-1827

    No evidence in the record shows that Midwest Fence bid
on or lost any contracts subject to the target market program.
IDOT has not yet set aside any guardrail and fencing contracts
under the target market program. Midwest Fence has there-
fore not shown that it has suffered from an “inability to com-
pete on an equal footing in the bidding process” with respect
to contracts within the target market program. Northeastern

. The possibility that IDOT might some-
day include a guardrail and fencing contract in the target mar-
ket program does not give Midwest Fence standing now. The
possible injury Midwest Fence fears is not one of “sufficient
immediacy and ripeness to warrant judicial intervention.” 

quoting Warth v. Seldin, 

, 516 (1975).
III. Claims Against the United States Department of Transportation
       Facial Versus As-Applied Challenge
    The federal DBE program authorizes and to some extent
requires state governments to rely on racial classifications in
awarding government contracts. Accordingly, the equal pro-
tection challenge requires the government to show that the
program can survive strict scrutiny, meaning that the pro-
gram serves a compelling government interest and is nar-
rowly tailored to advance that interest. Adarand 

; Northern Contracting, Inc. v. Illinois, 

, 720 (7th Cir. 2007). Remedying the effects of past or pre-
sent discrimination can be a compelling governmental inter-
est. 

. In this appeal, Midwest Fence does
not challenge the national compelling interest in remedying
past discrimination. We therefore focus on whether the fed-
eral program is narrowly tailored.
No. 15-1827                                                     15

     Before digging into the merits of Midwest Fence’s chal-
lenge to the federal program, we must address another pre-
liminary issue: whether Midwest Fence can maintain an as-
applied challenge against USDOT and the federal program or
whether, as the district court held, the claim against USDOT
is limited to a facial challenge. Midwest Fence sought a decla-
ration that the federal regulations are unconstitutional as ap-
plied in Illinois. The district court rejected the attempt to bring
that claim against USDOT, treating it as applying only to
IDOT. Midwest 

, citing Sherbrooke
Turf, Inc. v. Minnesota Dep’t of Transportation, 

, 973
(8th Cir. 2003). We agree.
    A facial challenge ordinarily requires a plaintiff to show
there is no set of circumstances under which the challenged
statutes or regulations can operate constitutionally. See United
States v. Salerno, 

, 745 (1987). Midwest Fence’s ar-
gument seeks to turn this standard on its head, for it amounts
to an argument that because (Midwest Fence contends) the
federal regulations are applied unconstitutionally in Illinois,
the regulations themselves are unconstitutional.
    A principal feature of the federal regulations is their flexi-
bility and adaptability to local conditions. See, e.g., 49 C.F.R.
§ 26.45 (states to set DBE participation goals); § 26.33 (states
to devise “appropriate measures” to address overconcentra-
tion). That flexibility is important to the constitutionality of
the program. As noted, such a race- and gender-conscious
program must be narrowly tailored to serve the compelling
governmental interest. See United States v. Paradise, 

, 171 (1987) (plurality opinion) (in determining whether
race-conscious remedies are appropriate, courts consider,
16                                                    No. 15-1827

among other factors, the “flexibility and duration of the re-
lief”); Western States Paving Co. v. Washington State Dep’t of
Transportation, 

, 993 (9th Cir. 2005) (citing Paradise
factors, including flexibility, in analyzing narrow tailoring of
federal DBE program); Sherbrooke 

(same).
The flexibility in the regulations makes the states, not USDOT,
primarily responsible for implementing their own programs
in ways that comply with the Equal Protection Clause. It
makes sense, then, that a state, not USDOT, is the correct party
to defend a challenge to its implementation of its program.
The district court did not err by treating the claims against
USDOT as only a facial challenge to the federal regulations.
       Narrow Tailoring
    The Eighth, Ninth, and Tenth Circuits have all found the
federal DBE program constitutional on its face. See Western
States 

; Sherbrooke 

–
68; Adarand Constructors, Inc. v. Slater, 

, 1155 (10th
Cir. 2000). We agree with these other circuits.
    The disputed issue here is whether the program is nar-
rowly tailored to further a compelling governmental interest.
Adarand 

. Narrow tailoring re-
quires “a close match between the evil against which the rem-
edy is directed and the terms of the remedy.” Builders Ass’n of
Greater Chicago v. County of Cook, 

, 646 (7th Cir.
2001). Like the district court, we look to the Paradise factors:
(a) “the necessity for the relief and the efficacy of alternative
[race-neutral] remedies,” (b) “the flexibility and duration of
the relief, including the availability of waiver provisions,” (c)
“the relationship of the numerical goals to the relevant labor
[or here, contracting] market,” and (d) “the impact of the relief
on the rights of third parties.” 

; see also
No. 15-1827                                                    17

Western States 

(citing Paradise factors in
analyzing narrow tailoring of federal DBE program); Sher-
brooke 

(same); Adarand 

(same). The Tenth Circuit in Adarand added to the
mix the question of over- or 

, citing 

.
    First, the federal DBE program requires states to meet as
much as possible of their overall DBE participation goals
through race- and gender-neutral means. 49 C.F.R. § 26.51(a).
If a state projects it can meet its entire DBE participation goal
for the year through neutral means, it “must implement [its]
program without setting contract goals during that year,” un-
less such goals later become necessary. § 26.51(f)(1) (emphasis
added). A state that projects it is likely to exceed its overall
DBE participation goal “must reduce or eliminate the use of
contract goals” as necessary to avoid exceeding its overall
goal. § 26.51(f)(2) (emphasis added).
    Next, on its face, the federal program is both flexible and
limited in duration. Quotas, one hallmark of an inflexible sys-
tem, are flatly prohibited, see § 26.43, and states may apply
for waivers, including waivers of “any provisions regarding
administrative requirements, overall goals, contract goals or
good faith efforts,” § 26.15(b). The regulations also require
states to remain flexible as they administer the program over
the course of the year. For instance, they must continually re-
assess their DBE participation goals and whether contract
goals are necessary to meet those overall goals. § 26.51(f). If a
state falls behind its DBE participation goal, it may increase
its use of contract goals—but if it is on track to exceed its par-
ticipation goal, it “must reduce or eliminate” its reliance on
those measures. § 26.51(f)(2).
18                                                   No. 15-1827

    States need not set a contract goal on every USDOT-
assisted contract, nor must they set those goals at the same
percentage as the overall participation goal. § 26.51(e)(2).
“The goal for a specific contract may be higher or lower than
that percentage level of the overall goal, depending on such
factors as the type of work involved, the location of the work,
and the availability of DBEs” for that particular contract. 

broader scale, states may also adjust their overall partic-
ipation goals if circumstances change. § 26.45(f)(1)(ii).
    Together, all of these provisions allow for significant and
ongoing flexibility. States are not locked into their initial DBE
participation goals. Their use of contract goals is meant to re-
main fluid, reflecting a state’s progress toward its overall DBE
goal. As for duration, Congress has repeatedly reauthorized
the program after taking new looks at the need for it. See, e.g.,
Fixing America’s Surface Transportation Act, Pub. L. No. 114-
94, § 1101(b), 129 Stat. 1312, 1323–25 (2015). As noted above,
states must monitor progress toward meeting DBE goals on a
regular basis and alter the goals if necessary. They must stop
using race- and gender-conscious measures if those measures
are no longer needed.
    The numerical goals are also tied to the relevant markets.
The federal program features a national “aspirational goal” of
10 percent of funds going to DBEs, but that goal “does not
authorize or require recipients to set overall or contract goals
at the 10 percent level, or any other particular level, or to take
any special administrative steps if their goals are above or be-
low 10 percent.” 49 C.F.R. § 26.41(c); see also § 26.45(b). Ra-
ther, the regulations prescribe a process for setting a DBE par-
ticipation goal that focuses on information about the specific
No. 15-1827                                                    19

market and that is intended to reflect “the level of DBE partic-
ipation you would expect absent the effects of discrimina-
tion.” § 26.45(b). The regulations thus instruct states to set
their DBE participation goals to reflect actual DBE availability
in their jurisdictions, as modified by other relevant factors like
DBE capacity.
    Midwest Fence focuses on the burden on third parties and
argues the program is over-inclusive. But the regulations in-
clude mechanisms to minimize the burdens the program
places on non-DBE third parties. A primary example is sup-
plied in § 26.33(a), which requires states to take steps to ad-
dress overconcentration of DBEs in certain types of work if
the overconcentration unduly burdens non-DBEs to the point
that they can no longer participate in the market. Standards
can be relaxed if uncompromising enforcement would yield
negative consequences. For instance, states can obtain waivers
if special circumstances make the state’s compliance with part
of the federal program “impractical,” § 26.15(a), and contrac-
tors who fail to meet a DBE contract goal can still be awarded
the contract if they have documented “good faith efforts to
meet the goal,” § 26.53(a)(2).
    Against this backdrop, Midwest Fence argues that a “mis-
match” in the way contract goals are calculated results in a
burden that falls disproportionately on specialty subcontrac-
tors. Under the federal regulations, states’ overall goals are set
as a percentage of all their USDOT-assisted contracts. See
§ 26.45. However, states may set contract goals “only on those
[USDOT]-assisted contracts that have subcontracting possi-
bilities.” § 26.51(e)(1) (emphasis added). This makes sense: if
a contract has no subcontracting possibilities, then either it
will be performed by a DBE or it will not; there is nothing in
20                                                 No. 15-1827

between. Midwest Fence argues that because DBEs must be
small, see § 26.65, they are generally unable to compete for
prime contracts. This, Midwest Fence argues, is the “mis-
match.” Where contract goals are necessary to meet an overall
DBE participation goal, those contract goals are met almost
entirely with subcontractor dollars, which places a heavy bur-
den on non-DBE subcontractors while leaving non-DBE
prime contractors in the clear.
    To understand how, imagine a state’s DBE participation
goal is 10%, so that DBEs should receive 10% of “all Federal-
aid highway funds [the state] will expend in FHWA-assisted
contracts in the forthcoming three fiscal years.” § 26.45(e)(1).
Let’s say the state will spend $100 million. To meet its overall
goal, $10 million should go to DBEs. Let’s also assume the
state cannot achieve that goal with race-neutral measures and
so must turn to contract goals. Then suppose that half the con-
tracts have no subcontracting possibilities, rendering them in-
eligible for contract goals. Now, instead of drawing the $10
million in DBE funds from the total $100 million, the state
looks only to the $50 million in contracts with subcontracting
possibilities to meet its DBE contract goals. Furthermore, if of
that $50 million, only $20 million of work can be performed
by subcontractors (and thus by DBEs), the $10 million will be
drawn from that pool and that pool alone. Thus, in this exam-
ple, although the participation goal calls for DBEs to receive
10% of total funds, in practice it requires the state to award
DBEs fully half of the available subcontractor funds while tak-
ing no business away from prime contractors. This, according
to Midwest Fence, is the “mismatch” that imposes a dispro-
portionate burden on specialty subcontractors like itself.
No. 15-1827                                                    21

    This prospect is troubling. The DBE program can impose
a disproportionate burden on small, specialized non-DBE
subcontractors, especially when compared to larger prime
contractors with whom DBEs would compete less frequently.
This potential for a disproportionate burden, however, does
not render the program facially unconstitutional. The consti-
tutionality of the program depends on how it is implemented.
    Moreover, some of the suggested race- and gender-neutral
means that states can use under the federal program are de-
signed to increase DBE participation in prime contracting and
other fields where DBE participation has historically been
low. See § 26.51(b)(1), (7). For instance, the regulations specif-
ically encourage states to make contracts “more accessible to
small businesses.” § 26.51(b)(1). Suggested means include es-
tablishing “a race-neutral small business set-aside for prime
contracts under a stated amount,” § 26.39(b)(1); facilitating
the “ability of consortia or joint ventures consisting of small
businesses, including DBEs, to compete for and perform
prime contracts,” § 26.39(b)(4); and “ensuring that a reasona-
ble number of prime contracts are of a size that small busi-
nesses, including DBEs, can reasonably perform,”
§ 26.39(b)(5). As USDOT argues, the federal program “explic-
itly contemplates [DBEs’] ability to compete equally by re-
quiring States to report DBE participation as prime contrac-
tors and makes efforts to develop that potential.”
    These measures to increase the participation of DBEs in
prime contracting will not always be sufficient. They may in
fact rarely be sufficient. States will continue to resort to con-
tract goals that open the door to the type of mismatch that
Midwest Fence describes. But the program on its face does not
22                                                  No. 15-1827

compel an unfair distribution of burdens. Small specialty con-
tractors may have to bear at least some of the burdens created
by remedying past discrimination under the federal DBE pro-
gram, but the Supreme Court has indicated that innocent
third parties may constitutionally be required to bear at least
some of the burden of the remedy. Wygant v. Jackson Board of
Education, 

, 281 (1986) (plurality opinion).
    Midwest Fence argues that the federal program is over-in-
clusive because it grants preferences to groups without ana-
lyzing the extent to which each group is actually disadvan-
taged. This is essentially an attack on a report by Dr. Jon Wain-
wright, which reviewed 95 disparity and availability studies
across 32 states between the years 2000 and 2012. Many of
Midwest Fence’s criticisms are best analyzed as part of its as-
applied challenge against the state defendants, but we men-
tion two federal-specific arguments briefly.
    First, Midwest Fence contends that nothing proves that the
disparities Dr. Wainwright relied upon were caused by dis-
crimination. But to justify its program, USDOT does not need
definitive proof of discrimination. Instead it must have a
“strong basis in evidence” that remedial action is necessary to
remedy past discrimination. 

; see also
H.B. Rowe Co. v. Tippett, 

, 241 (4th Cir. 2010) (“A
state need not conclusively prove the existence of past or pre-
sent racial discrimination to establish a strong basis in evi-
dence for concluding that remedial action is necessary.”).
   Second, Midwest Fence attacks what it perceives as the
one-size-fits-all nature of the program, suggesting that the
regulations ought to provide different remedies for different
groups because those groups are disadvantaged to different
degrees. The table in Dr. Wainwright’s report on which the
No. 15-1827                                                   23

district court relied to find widespread disparities shows var-
ying levels of disparity for different groups, Midwest Fence
argues, so the federal program should offer different solu-
tions to each group. Instead, it offers a single approach to all
the disadvantaged groups, regardless of degree.
    Midwest relies on Builders Ass’n of Greater 

, in which we held unconstitutional a set-aside ordi-
nance that used a presumption of discrimination “merely on
the basis of having an ancestor who had been born in the Ibe-
rian peninsula,” where there was simply no evidence in the
record of past discrimination against that group. This case is
different. Though the table in question reflects some variation
in the extent of disadvantage, Midwest has not argued that
any of the groups in the table were not in fact disadvantaged
at all. Though the minority groups in question are given pre-
sumptive DBE status across the board, the federal regulations
ultimately require individualized determinations. Each pre-
sumptively disadvantaged firm owner must certify that he or
she “is, in fact, socially and economically disadvantaged,” 49
C.F.R. § 26.67(a)(1), and that presumption can be rebutted,
§ 26.67(b). In this way, the federal program requires states to
extend benefits only to those who are actually disadvantaged.
   We agree with the district court and with the Eighth,
Ninth, and Tenth Circuits that the federal DBE program is
narrowly tailored on its face, so it survives strict scrutiny.
IV. Claims Against IDOT and the Tollway
       Motion to Exclude Supplemental Record on Appeal
   Before we evaluate the merits of Midwest Fence’s equal
protection claims against IDOT and the Tollway, we must re-
24                                                No. 15-1827

solve a dispute about the evidence properly before us on ap-
peal. Before oral argument, Midwest Fence took the unusual
step of supplementing the record with evidence that was not
before the district court when it made its decision. The state
defendants have moved to exclude that evidence from the ap-
pellate record. We decided to take up that motion with the
merits, and we now grant the motion to exclude.
    The evidence in question is a disparity study commis-
sioned by the Tollway in 2013 and performed by Colette
Holt & Associates. The Holt study was underway during the
district court proceedings. Parts of the study had been com-
pleted, and in the district court the Tollway relied on some of
that completed analysis to support its claim that women and
minorities continued to face barriers in the Illinois construc-
tion market, including wage differentials and decreased like-
lihood of forming their own construction businesses.
    The district court decision of March 24, 2015 referred to
portions of the Holt study in the record. See Midwest 

–38. On November 13, 2015, after the de-
fendants had filed their briefs on appeal but before Midwest
Fence filed its reply brief, the Holt study was completed. Mid-
west Fence filed a motion in the district court to supplement
the record with the complete Holt study, and on December 10,
2015, the court granted the motion to that effect. Five days
later, the Illinois defendants moved in this court to exclude
the supplemental evidence.
    Midwest Fence explained at oral argument that it wants
one specific piece of information in the record: a chart that
lays out disparity ratios in the construction industry for each
of several demographic groups. A disparity ratio, also called
No. 15-1827                                                     25

a disparity index, reflects a group’s participation in contract-
ing opportunities.
    Federal Rule of Appellate Procedure 10(e) governs correc-
tion or modification of the record on appeal. It provides no
legal basis for this appellate court to consider the complete
Holt study. Borden, Inc. v. Federal Trade Comm’n, 

,
788 (7th Cir. 1974) (“Rule 10(e) does not give this court author-
ity to admit on appeal any document which was not made a
part of the record in the district court.”). The record without
the complete Holt study “truly discloses what occurred in the
district court,” and nothing was “omitted from or misstated
in the record by error or accident,” Fed. R. App. P. 10(e).
    As a general rule, we will not consider evidence on appeal
that was not before the district court when it rendered its de-
cision. Adding new evidence would essentially convert an ap-
peal into a collateral attack on the district court’s decision.
United States v. Banks, 

, 567 (7th Cir. 2005); see also
Shasteen v. Saver, 

, 934 n.2 (7th Cir. 2001) (“[T]he
purpose of Rule 10(e) is to ensure that the court on appeal has
a complete record of the proceedings leading to the ruling ap-
pealed from, not to facilitate collateral attacks on the ver-
dict.”), quoting United States v. Hillsberg, 

, 336 (7th
Cir. 1987); EEOC v. Sears, Roebuck & Co., 

, 337 (7th
Cir. 1988) (“Even the results of simple math are inappropriate
for this court to consider on appeal, however, if the results
were not initially presented to the district court.”).
    Midwest Fence urges us to make an exception for two rea-
sons: (1) the information was within not its control but the
Tollway’s; and (2) the Tollway should not be permitted to
cherry-pick portions of the Holt study that support its posi-
tion when other portions cut against its use of race-conscious
26                                                  No. 15-1827

measures. Midwest Fence’s frustration is understandable. But
this is no reason to permit a collateral attack on the district
court’s judgment based on data that the district court never
considered and that were not even available to the Tollway
when it began to defend this lawsuit and its DBE program.
The Illinois defendants deserve an opportunity to scrutinize
their program in light of the Holt study, which itself should
be considered in the context of all the evidence they have
about the market in Illinois, and to take action—or not—ac-
cordingly. The entire Holt study may be useful to one side or
another in a future challenge to the IDOT and Tollway pro-
grams, but it is not available here. We grant the motion to ex-
clude the Holt study and do not consider it in this appeal.
       Void for Vagueness?
    Midwest Fence argues that the federal regulations are un-
constitutionally vague as applied by IDOT. The regulations
provide that if a state establishes a DBE contract goal, it must
award the contract to a bidder who makes “good faith efforts”
to meet that goal, meaning either actually obtaining sufficient
DBE participation or documenting adequate efforts to do so.
49 C.F.R. § 26.53(a). Midwest Fence contends that the regula-
tions fail to specify what good faith efforts a contractor must
make to qualify for a waiver.
    Appendix A of 49 C.F.R., Part 26, provides direction. It
says that a state’s “determination concerning the sufficiency
of the firm’s good faith efforts is a judgment call.” It provides
a list of actions a state should consider as part of a bidder’s
good faith efforts, emphasizing that it is “not intended to be a
mandatory checklist, nor is it intended to be exclusive or ex-
haustive.” 

Midwest Fence specifically attacks
¶ IV(D)(2), which addresses possible cost differentials in the
No. 15-1827                                                     27

use of DBEs: “the fact that there may be some additional costs
involved in finding and using DBEs is not in itself sufficient
reason for a bidder’s failure to meet the contract DBE goal, as
long as such costs are reasonable.” At the same time, the reg-
ulation assures prime contractors that they need not accept
higher price quotes from DBEs “if the price difference is ex-
cessive or unreasonable.” 

on when a differ-
ence in price is significant enough to justify falling short of the
DBE contract goal is too vague, according to Midwest Fence.
    Vagueness doctrine grew out of the due process clauses of
the Fifth and Fourteenth Amendments. It seeks to enforce the
principle that “laws which regulate persons or entities must
give fair notice of conduct that is forbidden or required.” FCC
v. Fox Television Stations, Inc., 567 U.S. —, —, 

,
2317 (2012). This doctrine is a poor fit for Midwest Fence’s
complaint. The regulations on good faith efforts do not
threaten non-DBE subcontractors or prime contractors with
punishment. At worst, a prime contractor who misjudges the
extent of its good faith efforts will lose out on a contract, but
even then only after an opportunity for administrative recon-
sideration, 49 C.F.R. § 26.53(d).
   Midwest Fence’s approach is problematic for another rea-
son. It argues that the lack of more specific guidance decreases
the program’s flexibility, but we do not see how this is so. Ap-
pendix A allows a bidder to use “good business judgment” to
decide whether it should select a DBE for subcontracting; it
cannot rely exclusively on the existence of “some additional
costs” to reject the DBE, but it can reject an excessively high
quote from a DBE. 49 C.F.R. pt. 26 app. A, ¶ IV(D)(2). If the
standard seems vague, that is likely because it was meant to
be flexible. What price differential is “unreasonable” may
28                                                   No. 15-1827

well vary from project to project—even between subcontrac-
tors on the same contract—based on capability, capacity, or
other factors. A more rigid standard could easily be too arbi-
trary and hinder prime contractors’ ability to adjust their ap-
proaches to the circumstances of particular projects.
    Midwest Fence’s real argument seems to be that in prac-
tice, prime contractors err too far on the side of caution, grant-
ing significant price preferences to DBEs instead of taking the
risk of losing a contract for failure to meet the DBE goal. Mid-
west Fence contends this creates a de facto system of quotas
because contractors believe they must meet the DBE goal in
their bids or lose the contract. But Appendix A to the regula-
tions cautions against this very approach. See 

(“The
Department also strongly cautions you against requiring that
a bidder meet a contract goal (i.e., obtain a specified amount
of DBE participation) in order to be awarded a contract, even
though the bidder makes an adequate good faith efforts
showing. This rule specifically prohibits you from ignoring
bona fide good faith efforts.”).
    Flexibility and the availability of waivers affect whether a
program is narrowly tailored. The regulations caution against
quotas; provide examples of good faith efforts prime contrac-
tors can make and states can consider; and instruct a bidder to
use “good business judgment” to decide whether a price dif-
ference between a DBE and a non-DBE subcontractor is reason-
able or excessive in a given case. For purposes of contract
awards, this is enough to “give fair notice of conduct that is
forbidden or required,” Fox Television Stations, 567 U.S. at 

.
No. 15-1827                                                   29

       Equal Protection Challenge
       1. Compelling Interest with Strong Basis in Evidence
    Turning to the merits of Midwest Fence’s equal protection
claims based on the actions of IDOT and the Tollway, the first
issue is whether the state defendants had a compelling inter-
est in enacting their programs, one with a “strong basis in ev-
idence.” See 

, quoting 

. We and other circuits have held that a state agency is
entitled to rely on the federal government’s compelling inter-
est in remedying the effects of past discrimination to justify
its own DBE plan for highway construction contracting.
Northern 

–21; see also Western
States 

; Sherbrooke 

.
However, not all of IDOT’s contracts are federally funded, and
the Tollway did not receive federal funding at all during the
period covered by the complaint. With respect to those con-
tracts, at least, we must consider whether IDOT and the Toll-
way have established a strong basis in evidence to support
their programs. We start with IDOT’s evidence and then turn
to the Tollway’s evidence. We then address Midwest’s criti-
cisms of the defendants’ evidence.
          a. IDOT Program
   IDOT relies first on a DBE availability study performed in
2004 by National Economic Research Associates (NERA). The
study, led by Dr. Wainwright, defined IDOT’s construction
market by identifying all highway and road contracts be-
tween 1996 and 2000 and identifying the item-level cost esti-
mates for each contract; categorizing line items by Standard
Industrial Classification codes and calculating a dollar-based
30                                                            No. 15-1827

weight to apply to those categories; eliminating Standard In-
dustrial Classification codes that offered negligible opportu-
nities; and calculating geographic weights in Illinois based on
contractor and subcontractor ZIP codes. It then estimated
businesses in the market using Dun & Bradstreet’s Market-
Place, a database of U.S. businesses; collected DBE lists from
IDOT and other entities to find any DBEs that MarketPlace
might have missed; and adjusted the figures to account for
possibly erroneous DBE classifications. Applying the Stand-
ard Industrial Classification and geographic weights to the
number of DBEs yielded a DBE availability of 22.77% across
all projects. The 2004 NERA study would have adjusted the
availability figure upward slightly to account for artificially
low totals caused by discrimination that keeps minorities and
women out of the market or makes their businesses more
likely to fail, but IDOT ultimately adopted the 22.77% figure. 1
   Next, IDOT commissioned Mason Tillman Associates to
perform a full disparity study, as distinct from an availability


     1IDOT later updated its availability data to reflect a 2013 availability
study by Mason Tillman Associates. The study reviewed contracts
awarded between July 1, 2010 and June 30, 2011. Using the same methods
it employed in its disparity study, discussed later in this section, Mason
Tillman calculated that 33.01% of willing and able construction subcon-
tractors were DBEs. The updated availability study also included addi-
tional anecdotal evidence of race and gender barriers derived from inter-
views with business owners in Illinois and a regression analysis of other
non-highway construction industries that found “statistically significant
disparities in the likelihood of owning a business … for minorities and
females,” “statistically significant disparities in business earnings for mi-
norities, females, and persons with disabilities,” and statistically signifi-
cant disparities in the likelihood of being denied a business loan for Afri-
can Americans and Hispanics.
No. 15-1827                                                    31

study. The study was published in 2011. Mason Tillman col-
lected information about prime contractors and subcontrac-
tors involved with construction, architecture, and engineering
contracts awarded between January 1, 2006 and December 31,
2008. It began by (1) collecting utilization records to deter-
mine the extent to which IDOT used DBEs and non-DBEs and
(2) defining IDOT’s market area. It then identified businesses
that were willing and able to provide needed services. Firm
availability was weighted to reflect IDOT’s contracting pat-
tern, with weights assigned to different areas based on the
percentage of dollars expended in those areas. The study then
determined whether there was a statistically significant un-
der-utilization of DBEs by calculating the dollars each group
would be expected to receive based on availability; calculat-
ing the difference between the expected and actual amount of
contract dollars received; and ensuring that results were not
attributable to chance.
    The study found that DBEs were significantly under-uti-
lized as prime contractors. They made up 25.55% of prime
contractors in the construction field. They received only 9.13%
of prime contracts valued below $500,000 and only 8.25% of
the available prime contract dollars in that range. With respect
to even smaller prime contracts—those valued at $25,000 or
less—DBEs received just 8.96% of contracts, amounting to
10.38% of available dollars. For prime contracts under
$500,000, the disparity ratio was 0.32; for prime contracts for
$25,000 or less, the disparity ratio was 0.41. Both ratios were
statistically significant. A figure below 0.80 is generally con-
sidered solid evidence of systematic under-utilization calling
for affirmative action to correct it. In the realm of construction
subcontracting, the study showed that DBEs made up 29.24%
of available subcontractors. In the construction industry, they
32                                                  No. 15-1827

received 44.62% of available subcontracts, but those subcon-
tracts amounted to only 10.65% of available subcontracting
dollars. This yielded a statistically significant disparity ratio
of 0.36, again low enough to signal systemic under-utilization.
    IDOT has relied on additional data to justify its program.
For example, IDOT conducted a “zero-goal” experiment in
2002 and 2003, when it did not apply DBE goals to contracts
that would otherwise have featured them. Without contract
goals, the share of the contracts’ value that DBEs received
dropped dramatically, to just 1.5% of the total value of the
contracts. See Northern 

. Another
informal analysis in IDOT’s 2012–15 goal-setting report
showed that of 4,129 construction, architecture, and engineer-
ing contracts IDOT awarded from 2006 through 2008, more
than half were advertised without a DBE goal. In those con-
tracts advertised without a DBE goal, the DBE subcontractor
participation rate was just 0.84%. The 2004 NERA availability
study did not itself analyze market disparities, but it did re-
view a number of past Illinois-area disparity studies commis-
sioned in the wake of Croson. All found “pervasive and sys-
temic discrimination against minorities and women.” While
less rigorous, the data from these studies help show that
IDOT had a strong basis in evidence to adopt its program.
          b. Tollway Program
    To show a basis for its program, the Tollway relies primar-
ily on a 2006 NERA study that was limited to the Tollway’s
contracting market area. NERA first determined DBE availa-
bility in the Tollway’s marketplace via what it calls a “custom
census” process, creating a database of representative pro-
jects, identifying geographic and product markets, counting
businesses in those markets, identifying and verifying which
No. 15-1827                                                    33

businesses are minority- and woman-owned, and verifying
the ownership status of all other firms. NERA then examined
the Tollway’s historical contract data and reported its DBE uti-
lization between 2000 and 2005 as a percentage of contract
dollars. Finally, it compared DBE utilization and DBE availa-
bility, coming up with disparity indices divided by race and
sex, as well as by industry group.
    In the construction industry, the 2006 NERA study found
that out of 115 disparity indices, 80 showed statistically sig-
nificant under-utilization of DBEs. In fact, in 61 cases, the dis-
parity index was both zero and statistically significant, mean-
ing the Tollway did not use any DBEs at all in that category.
    The study then discussed statistical disparities in earnings
and the formation of businesses by minorities and women,
concluding: “In almost every instance examined, a statisti-
cally significant adverse impact on earnings is observed in
both the economy at large and in the construction and con-
struction-related professional services sector.” The study also
found that women and minorities are not as likely to start
their own businesses “and that minority business formation
rates would likely be substantially and significantly higher if
markets operated in a race- and sex-neutral manner.”
    The 2006 NERA study also used regression analysis to as-
sess differences in wages, business-owner earnings, and busi-
ness-formation rates between white men and minorities and
women in the wider construction economy. The study found
statistically significant disparities remained between white
men and other groups, even controlling for various independ-
ent variables such as age (as a proxy for experience), educa-
tion, location, industry affiliation, and time. The disparities,
34                                                   No. 15-1827

according to NERA, were consistent with a market affected
by discrimination.
   Like IDOT, the Tollway has also presented additional evi-
dence to justify its DBE program. Before adopting the pro-
gram, the Tollway set aspirational participation goals on a
small number of contracts. Those attempts failed. In 2004, the
Tollway did not award a single prime contract or subcontract
to a DBE, and the DBE participation rate in 2005 was 0.1%
across all construction contracts. Finally, in adopting its pro-
gram, the Tollway also considered anecdotal evidence that
was presented in Northern Contracting—namely, the disturb-
ing testimony of several DBE owners regarding barriers that
they themselves faced. See Northern Contracting, Inc. v. Illinois,
No. 00 C 4515, 

, at *13–14 (N.D. Ill. Sept. 8,
2005), aff’d, 

(7th Cir. 2007).
          c. Midwest Fence’s Criticisms
    Most of Midwest Fence’s criticisms of the defendants’ evi-
dence stem from a report prepared by its expert, labor econo-
mist Dr. Jonathan Guryan. Midwest Fence argues that Dr.
Guryan’s attacks on the various reports in the record create a
triable issue of fact as to whether IDOT and the Tollway had
the requisite “strong basis in evidence” to conclude that race-
and gender-conscious remedial action was necessary.
    Two sections of Dr. Guryan’s report provide the most sub-
stantial bases for criticism of the defendants’ evidence. In Sec-
tion 6.1, Dr. Guryan argues that NERA has failed to account
for DBEs’ readiness, willingness, and ability to do business
with IDOT and the Tollway, and that Mason Tillman’s method
of assessing readiness and willingness is flawed. In Section
6.2, Dr. Guryan argues that the Mason Tillman study failed to
No. 15-1827                                                                35

account for DBEs’ relative capacity, meaning a firm’s ability to
take on more than one contract at a time. 2 See Rothe Develop-
ment Corp. v. Dep’t of Defense, 

, 1044 (Fed. Cir.
2008) (disparity studies took account of firm size but failed to
account for relative capacities of firms to bid for more than
one contract at a time). Other sections of the report present
less substantial critiques: (1) the existence of the DBE program
may bias availability upward because it creates an incentive
for DBEs to do business with the public sector; (2) the exist-
ence of the DBE program may cause participants to alter their
behavior, such that “[a]nything observed about the public sec-
tor may be affected by the DBE program”; and (3) the data
may be stale.
    One major problem with Dr. Guryan’s report is that he did
not perform any substantive analysis of his own. As the dis-
trict court noted, the evidence he offered is “speculative at
best.” Midwest 

. For example, he
noted only that it was “unclear” whether IDOT’s assumption
of willingness is appropriate and/or “how this assumption af-
fects the measure of availability of DBE firms.” His relative
capacity analysis was similarly speculative, arguing that the


    2  The NERA studies do not account for firm capacity. According to
Dr. Wainwright, that is because excluding firms based on capacity would
preclude the government from instituting a remedy, since discrimination
itself has prevented minority- and women-owned firms from achieving
the capacity required to compete for contracts. See Rothe Development Corp.
v. Dep’t of Defense, 

, 1045 (Fed. Cir. 2008) (recognizing that “a
minority-owned firm’s capacity and qualifications may themselves be af-
fected by discrimination”). The Mason Tillman study found no effective
way to account for capacity, but it performed a regression analysis to
measure relative capacity and limited its disparity analysis to contracts
under $500,000 to take capacity into account to the extent possible.
36                                                No. 15-1827

assumption that firms have the same ability to provide ser-
vices up to $500,000 “may not be true in practice,” and that if
the estimates of capacity are too low, “the resulting disparity
index overstates the degree of disparity that exists.” And so
on: he argued that the existence of the DBE program “may”
cause an upward bias in availability, that any observations of
the public sector in general “may” be affected by the DBE pro-
gram’s existence, and that data become less relevant as time
passes. Given the substantial utilization disparities shown in
the defendants’ reports, these speculative critiques do not
raise a genuine issue of fact as to whether the defendants had
a substantial basis in evidence to believe that action was
needed to remedy discrimination.
    Midwest Fence argues that requiring it to provide an in-
dependent statistical analysis places an “impossible burden”
on it due to the time and expense that would be required. Re-
call, though, that the burden is initially on the government to
justify its programs. Since the state defendants offered evi-
dence to do so, the burden then shifted to Midwest Fence to
show a genuine issue of material fact as to whether the state
defendants had a substantial basis in evidence for adopting
their DBE programs. Speculative criticism about potential
problems will not carry that burden.
    Even the capacity question, which is Dr. Guryan’s strong-
est criticism (and which Rothe recognized as a serious prob-
lem in analyzing the Department of Defense’s nationwide
contract evaluation preferences, 

–45), does
not undermine the substantial basis in evidence here. For one
thing, Dr. Guryan and Midwest Fence have not explained
how to account for relative capacity. In addition, even the
No. 15-1827                                                  37

court in Rothe recognized that defects in capacity analyses are
not fatal in and of themselves:
       To be clear, we do not hold that the defects in the
       availability and capacity analyses in these six
       disparity studies render the studies wholly un-
       reliable for any purpose. Where the calculated
       disparity ratios are low enough, we do not fore-
       close the possibility that an inference of discrim-
       ination might still be permissible for some of the
       minority groups in some of the studied indus-
       tries in some of the jurisdictions. And we recog-
       nize that a minority-owned firm’s capacity and
       qualifications may themselves be affected by
       discrimination. But we hold that the defects we
       have noted detract dramatically from the proba-
       tive value of these six studies, and, in conjunc-
       tion with their limited geographic coverage,
       render the studies insufficient to form the statis-
       tical core of the “strong basis in evidence” re-
       quired to uphold the 

(emphases in original) (footnote omitted).
Here, the defendants’ studies do not suffer from the geo-
graphic problems in Rothe. They show striking utilization dis-
parities in specific industries in the relevant areas. And they
are consistent with the anecdotal and less formal evidence de-
fendants have offered. Dr. Guryan’s speculation that failure to
account for relative capacity might have biased DBE availa-
bility upward does not undermine the “statistical core of the
‘strong basis in evidence’ required.” 

15-1827

    In addition to the Guryan report, Midwest Fence repeats
its argument that the disparity studies do not prove discrimi-
nation but are at best merely consistent with it. As noted
above, a “state need not conclusively prove the existence of
past or present racial discrimination to establish a strong basis
in evidence for concluding that remedial action is necessary.”
H.B. Rowe 

; see also 

(citing a Title VII case, Hazelwood School District v. United
States, 

, 307–08 (1977), for proposition that where
“gross statistical disparities can be shown, they alone in a
proper case may constitute prima facie proof of a pattern or
practice of discrimination”). Midwest Fence also attacks the
anecdotal evidence defendants have offered. For the sake of
argument, even if we assume that such anecdotal complaints
by themselves could not supply the strong basis in evidence
needed to institute a race- or gender-conscious remedy, those
complaints here just bolster defendants’ statistical analyses.
    In an argument limited to the Tollway, Midwest Fence also
attacks the Tollway’s supporting data from before it instituted
its DBE program, suggesting that those data are unreliable be-
cause: (1) trial testimony in Northern Contracting, 

, at *10, revealed that 1.5% to 1.7% of Tollway contract
dollars went to DBEs in 2002 and 2003; and (2) the disparity
study the Tollway eventually relied upon in enacting its own
program found that overall, between 2000 and 2005, DBEs re-
ceived 11.4% of contract dollars in construction. Alternatively,
Midwest suggests, the Tollway must have had excellent DBE
participation in 2000 and 2001 to come up with an 11.4% av-
erage in light of its dismal numbers from 2002 through 2005.
(Recall that the Tollway awarded no contracts to DBEs in 2004
and only 0.1% of contracts to DBEs in 2005.) The Tollway re-
sponds that during Northern Contracting, it simply used the
No. 15-1827                                                          39

best data available and that in any event both data sets show
disparities, albeit to varying degrees.
    The latter point is persuasive. We assume some of the Toll-
way’s data are not exact. This is likely since the 2004–05 “ex-
periment” was not formal, nor was it meant to be statistically
rigorous. But the NERA numbers, which are likely the most
reliable given the method, show significant disparities in the
construction industry. The numbers from other sources are
even more dismal. While every single number in the Tollway’s
arsenal of evidence may not be exact, the overall picture still
shows beyond reasonable dispute a marketplace with sys-
temic under-utilization of DBEs far below the “disparity in-
dex lower than 80 as an indication of discrimination,” H.B.
Rowe 

. Dr. Guryan’s abstract criticisms do
not undermine that core of evidence.
        2. Narrow Tailoring
   Finally, we come to the question of narrow tailoring:
whether IDOT’s and the Tollway’s implementation of their
DBE programs yielded “a close match between the evil
against which the remedy is directed and the terms of the
remedy,” Builders Ass’n of Greater 

. We
return to the so-called Paradise factors: (a) “the necessity for
the relief and the efficacy of alternative [race-neutral] reme-
dies,” (b) “the flexibility and duration of the relief, including
the availability of waiver provisions,” (c) “the relationship of
the numerical goals to the relevant labor [here, construction]
market,” and (d) “the impact of the relief on the rights of third
parties.” 

. 3


    3 With regard to the issue of narrow tailoring, Midwest Fence argues
that the district court improperly read our cases—particularly Milwaukee
40                                                            No. 15-1827

    The necessity of relief overlaps our analysis of IDOT’s and
the Tollway’s strong basis in evidence for believing their pro-
grams were needed to remedy lingering effects of discrimina-
tion. As discussed above, Midwest Fence has not undermined
the defendants’ strong combination of statistical and other ev-
idence to show that their programs are needed to remedy dis-
crimination. Both IDOT and the Tollway use race- and gen-
der-neutral alternatives. The undisputed facts show that
those alternatives have not been sufficient to remedy discrim-
ination. For IDOT, even the combination of race-neutral and
race-conscious measures has not been enough to meet its
overall goal. See also Northern 

(“[T]he record makes clear that IDOT uses nearly all of the
methods described in [49 C.F.R.] § 26.51(b) to maximize the
portion of the goal that will be achieved through race-neutral
means.”).
   As for flexibility, both IDOT and the Tollway make front-
end waivers available when a contractor has made good faith

County Pavers Ass’n v. Fiedler, 

(7th Cir. 1991), and Northern
Contracting, 

—as allowing the federal regulations to
“preempt” the Constitution. The problem for plaintiffs challenging this
sort of race-conscious program run by state or local governments subject
to federal regulations is to identify the responsible level of government.
As we explained above at pages 16–18, the federal regulations necessarily
give states considerable flexibility so that their programs can be tailored
narrowly to local history and markets. Milwaukee County Pavers and North-
ern Contracting should not be read to foreclose claims that a state or local
government violates the Equal Protection Clause in the way that it imple-
ments those flexible federal regulations. In terms of those opinions, a state
or local government would “exceed[] its federal authority,” Northern Con-

, citing Milwaukee County 

–
25, if it implemented federal law in a manner that violates the Equal Pro-
tection Clause.
No. 15-1827                                                   41

efforts to comply with a DBE goal. Midwest argues that the
low number of waivers granted, coupled with contractors’
fears of having a waiver denied and thereby losing a contract,
shows the system is a de facto quota system. We do not agree.
IDOT and the Tollway have not granted large numbers of
waivers, but there is also no evidence that they have denied
large numbers of waivers. The evidence here does not show
that defendants are responsible for failing to grant front-end
waivers that contractors do not request.
     In the absence of evidence that defendants fail to adhere
to the general good faith effort guidelines and arbitrarily deny
or discourage front-end waiver requests, Midwest Fence’s
contention that contractors fear losing contracts if they ask for
a waiver does not make the system a quota system—even in
the face of Illinois’s supposedly waiver-averse political cli-
mate. See Dunnet 

–99 (rejecting argument
that IDOT had implemented a no-waiver policy based on po-
litical pressures; evidence showed “unbroken record of grant-
ing waivers”). Midwest Fence’s own evidence shows that in
2007, IDOT granted 57 of 63 front-end waiver requests, and in
2010, it granted 21 of 35 front-end waiver requests. This is not
evidence of a quota system.
    We know, as well, that the Tollway granted at least some
front-end waivers between January 2006 and July 2012, in-
volving 1.02% of contract dollars. Without evidence that far
more waivers were requested, even this low total does not
raise a genuine dispute of fact.
   Midwest Fence also argues that we ought to look at the
dollar value of waivers granted rather than the raw number
of waivers granted. This argument is underdeveloped and in
42                                                No. 15-1827

any event does not support a different outcome here. The de-
fendants grant more front-end waiver requests than they
deny, regardless of the dollar amounts those requests encom-
pass. If Midwest Fence intends to suggest that IDOT and the
Tollway have an unwritten policy of granting only low-value
waivers, it has presented no evidence to that effect.
    Midwest Fence’s final and best argument against narrow
tailoring is its “mismatch” argument, discussed briefly above.
Its broad condemnation of the IDOT and Tollway programs
as failing to create a “light” and “diffuse” burden for third
parties is not persuasive. In Wygant, the case from which it
draws this standard, the Court’s plurality wrote that “hiring
goals impose a diffuse burden, often foreclosing only one of
several opportunities,” while “layoffs impose the entire bur-
den of achieving racial equality on particular individuals,”
making that burden “too 

. The DBE
programs, which set DBE goals on only some contracts and
allow those goals to be waived if necessary, may end up fore-
closing “one of several opportunities” for a non-DBE specialty
subcontractor like Midwest Fence. But there is no evidence
that they “impose the entire burden” on that subcontractor by
shutting it out of the market entirely, as a layoff of an em-
ployee would do. See 

Fence’s point that subcontractors
appear to bear a disproportionate share of the burden as com-
pared to prime contractors is troubling. The evidence shows
disparities in both the prime contracting and subcontracting
markets. But under the federal regulations, individual con-
tract goals are set only for contracts that have subcontracting
possibilities. (This makes some sense. A “contract goal” on a
contract without subcontracting possibilities would be the
No. 15-1827                                                 43

equivalent of a set-aside with respect to that contract, which
opens up a new set of problems.) Although some DBEs are
able to bid on prime contracts, the necessarily small size of
DBEs makes that difficult in most cases.
    In the end, however, the record shows that the problem
Midwest Fence raises is largely theoretical. Not all contracts
have DBE goals, so subcontractors are on an even footing for
those contracts without such goals. IDOT and the Tollway
both use neutral measures, including some designed to make
prime contracts more accessible to DBEs. There seems to be
no dispute that DBE trucking and material suppliers count to-
ward fulfillment of a contract’s DBE goal—even though they
are not used as line items in calculating the contract goal in
the first place—which opens up even contracts with DBE
goals to non-DBE subcontractors. Nor do we understand
Midwest Fence’s argument with respect to the 50% of the con-
tract that Midwest Fence says must be performed by the prime
contractor, per IDOT regulations. Illinois sets its contract
goals with reference to subcontracting line items that can be
performed by DBEs. That the goals are expressed as a percent-
age of the total contract does not seem to affect that fact.
    If Midwest Fence had presented evidence rather than the-
ory on this point, the result might be different. Evidence that
subcontractors were being frozen out of the market or bearing
the entire burden of the DBE program would likely require a
trial to determine at a minimum whether IDOT and the Toll-
way were adhering to their responsibility to avoid overcon-
centration in subcontracting. See 49 C.F.R. § 26.33. Midwest
Fence has shown how the Illinois program could yield that re-
sult but not that it actually does so.
44                                                No. 15-1827

    In light of the IDOT and Tollway programs’ mechanisms
to prevent subcontractors from having to bear the entire bur-
den of the DBE programs—the use of DBE materials and
trucking suppliers in satisfying goals, efforts to draw DBEs
into prime contracting, and so on—Midwest Fence has not
shown a genuine dispute of fact on this point. The theoretical
possibility of a “mismatch” could be a problem, but we have
no evidence that it actually is. See Dorsey v. Morgan Stanley,

, 627 (7th Cir. 2007) (inferences supported only by
speculation or conjecture will not defeat a summary judgment
motion).
    We conclude that the IDOT and Tollway DBE programs
are narrowly tailored to serve the compelling state interest in
remedying discrimination in public contracting. They include
race- and gender-neutral alternatives, set goals with reference
to actual market conditions, and allow for front-end waivers.
So far as the record before us shows, they do not unduly bur-
den third parties in service of remedying discrimination. Mid-
west Fence has failed to present a genuine dispute of fact on
this point.
     The judgment of the district court is AFFIRMED.