Opinion for the Court filed by Circuit Judge SENTELLE.
SENTELLE, Circuit Judge.
Rosemary Love and nine other female farmers appeal from the denial of their motion for class certification. Appellants claim, on behalf of themselves’ and “not less than 3,000” similarly situated women, that the United States Department of Agriculture (“USDA” or “the Department”) discriminatorily administered its lending programs, and that the Department failed to process and properly inves
tigate women’s discrimination complaints over the last quarter-century. Because we conclude the District Court did not abuse its discretion in denying the Appellants’ motion for class certification and did not err in dismissing the failure-to-investigate claim, we affirm in part. However, because the Appellants’ claim under the Administrative Procedure Act, 5 U.S.C. § 706(2)(A) (“APA”), remains largely un-briefed, we remand in part.
I. INTRODUCTION
The Appellants allege the USDA has engaged in a nationwide “pattern or practice” of discrimination, dating back to 1981.
Specifically, the Appellants’ first claim — hereinafter the “discrimination claim” — is that the Department violated the Equal Credit Opportunity Act, 15 U.S.C. §§ 1691-1691f (“ECOA” or “the Act”) by employing subjective loan-making criteria, which enabled decentralized decision-makers to discriminate amongst loan applicants on the basis of gender. The Appellants’ second claim — hereinafter the “failure-to-investigate claim” — -is that the Department violated both ECOA and the APA by systematically dismantling its complaint-processing systems and failing to investigate discrimination claims filed by women farmers.
A
The USDA administers its farm loan and subsidy programs through the Farm Service Agency (“FSA”).
The FSA makes several different types of loans, in-eluding “farm- ownership” loans, which assist farmers in buying or improving farm property, 7 C.F.R. pt.1943, “operating” loans, which provide credit and management assistance to help farmers run their farms,
id.
pt.1941, and “emergency” loans, which help farmers resume operations after a disaster,
id.
pt.1945. Under the USDA’s regulations, a farmer seeking a farm credit or benefit must first ask the Department for a loan application, which the USDA is required to disburse.
See id.
§ 1910.4(b) (“All persons requesting an application will be provided [one].”).
After receiving an application, a farmer is then required to submit her completed application to a local county committee, the members of which are selected by other farmers from that county.
See Pigford v. Glickman,
206 F.3d 1212, 1214 (D.C.Cir. 2000). The local county committee initially determines whether the applicant is eligible for the program, and USDA staff members ultimately grant or deny the application.
See
7 C.F.R. §§ 1910.5, 1910.4(i). The USDA has promulgated criteria upon which the local committees are to rely in making eligibility decisions, including citizenship, legal capacity, education and farming experience, farm size, inability to obtain sufficient credit elsewhere, and “character” (which emphasizes credit history and reliability).
See id.
§§ 1941.12 (operating loan criteria), 1943.12(a) (ownership loan criteria), 764.4 (emergency loan criteria).
Any farmer who believes the USDA denied her application for a program loan or
benefit on the basis of gender or any other prohibited basis may file a civil rights complaint with the Secretary of the USDA and/or the USDA’s Office of Civil Rights (“OCR”).
See id.
§ 15.6. The Department and/or OCR may then conduct an investigation and institute compliance proceedings, if needed.
See id.
§§ 15.8-.10. To effectuate compliance, the Secretary may refer the matter to the Department of Justice or institute any other applicable proceedings under state or local law.
Id.
§ 15.8(a).
If a farmer is dissatisfied with the USDA’s response to her discrimination complaint, she may sue in federal court under ECOA. The Act makes it “unlawful for any' creditor to discriminate against any applicant with respect to any aspect of a credit transaction ... on the basis of race, color, religion, national origin, sex or marital status, or age.” 15 U.S.C. § 1691(a). ECOA creates a private right of action against creditors, including the United States, who violate its anti-discrimination provisions, and it makes such creditors “liable to the aggrieved applicant for any actual damages sustained by such applicant acting either in an individual capacity or as a member of a class.”
Id.
§ 1691e(a). Exhaustion of administrative remedies is not required before an ECOA suit may be filed, and prevailing Appellants may recover attorney’s fees.
Id.
§ 1691e(d).
B
The Appellants filed a three-count complaint in the United States District Court for the District of Columbia, seeking $3 billion in money damages under ECOA, as well as both compensatory and equitable relief under the APA and the Declaratory Judgment Act, 28 U.S.C. § 2201(a). To buttress their claims, the Appellants filed 1,828 declarations, which purported to show that the Department “allowed, indeed supported, unconscionable disparate impact [through USDA’s lending programs] around the country.” The Appellants’ statistics expert, Patrick M. O’Brien, also filed a report, which purported to provide empirical evidence of “the adverse affects [sic] of a discriminatory system.” The Department moved to dismiss the complaint under Fed. R. Civ. P. 12(b)(6), and the Appellants moved for class certification under Fed. R. Civ. P. 23.
Before ruling on the propriety of class certification, the District Court dismissed the Appellants’ failure-to-investigate claim on three grounds. First, the District Court concluded that the USDA’s failure to investigate the Appellants’ complaints did not constitute a “credit transaction” within the meaning of ECOA, 15 U.S.C. § 1691(a). Second, the court held the APA’s “abuse of discretion” standard does not cover the USDA’s failure to investigate Appellants’ complaints because ECOA provided an alternative “adequate remedy.” Third, the District Court concluded Appellants’ claims for money damages arising from the USDA’s failure to investigate their complaints are barred by the doctrine of sovereign immunity.
In light of the court’s dismissal of the failure-to-investigate claim, the Appellants based their motion for class certification solely upon their discrimination claim. Accordingly, the Appellants filed an amended complaint, along with a renewed motion for certification of two subclasses. “Subclass 1” consists of female farmers who asked for loan application forms but did not receive them. “Subclass 2” consists of female farmers who received, completed, and submitted their loan application forms but did not garner a loan. The amended complaint also dropped the Appellants’ demand for $3 billion in damages and asked
only for “compensatory damages appropriate for proof at trial.”
The District Court denied the Appellants’ renewed motion to certify the class on three grounds. First, the court concluded that the Appellants failed to satisfy Rule 23(a)’s commonality requirement because there had not “been a substantial showing that would permit the inference that members of the class suffered from a
common policy of discrimination
that pervaded all of the challenged decisions.” Second, the court concluded that the class could not be certified under Rule 23(b)(2) because the claims for monetary relief predominated over the claims for equitable relief. Third, the court concluded the class could not be certified under Rule 23(b)(3) for the same reason it did not satisfy Rule 23(a)’s commonality requirement: The Appellants “adduced no evidence ... establishing that it is or ever was actually USDA policy to refuse to give loan application forms to women,” and individual justifications for any loan denials would predominate over questions common to the putative class.
The District Court stayed the proceedings so that the Appellants could seek interlocutory review of the denial of class certification on their discrimination claim, as well as the dismissal of their failure-to-investigate claim. In our discretion, we granted the Appellants’ petitions for interlocutory review of the class certification denial. We have jurisdiction over the District Court’s dismissal of their failure-to-investigate claim under 28 U.S.C. § 1292(b).
See
Fed. R. Crv. P. 23(f);
In re Lorazepam & Clorazepate Antitrust Litig.,
289 F.3d 98, 105-06 (D.C.Cir.2002). The parties (with the help of three
amici curiae)
then fully briefed and argued both issues before a merits panel.
Compare In re Veneman,
309 F.3d 789, 791 (D.C.Cir. 2002) (denying a 23(f) petition “because the critical questions required to resolve it are entirely unbriefed”).
II. DISCRIMINATION CLAIM
Appellants’ first assignment of error is that the District Court erroneously declined to certify their putative class as to the “discrimination claim.” To justify class certification, Appellants must first satisfy four threshold requirements, demonstrating that:
(1) the class is so numerous that joinder of all members is impracticable, (2) there are questions of law or fact common to the class, (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class, and (4) the representative parties will fairly and adequately protect the interests of the class.
Fed. R. Civ. P. 23(a). If all four prerequisites under Rule 23(a) are met, then Appellants must also show that their claims fit within one of the subsections of Rule 23(b).
See Amchem Prods., Inc. v. Windsor,
521 U.S. 591, 613-16, 117 S.Ct. 2231, 138 L.Ed.2d 689 (1997). Rules 23(b)(2) and (b)(3) — the two subsections at issue in this case — have different requirements depending primarily on the nature of the relief sought: Certification under the former is appropriate where a class seeks declaratory or injunctive relief, while certification under the latter is appropriate where,
inter alia,
the “ ‘final relief relates exclusively or predominantly to money damages.’ ”
In re Veneman,
309 F.3d at 792 (quoting Fed. R. Civ. P. 23(b)(2) advisory committee notes).
The District Court declined to certify the class because the Appellants failed to show “commonality” under Rule 23(a)(2), and they failed to meet the requirements of Rule 23(b)(2) or 23(b)(3). Because we conclude the District Court neither erred as a matter of law nor abused its discre
tion in finding the putative class members lack “commonality” under Rule 23(a)(2), we affirm without considering whether certification would also be improper under Rule 23(b).
Appellants argue that the District Court erred by failing to find commonality within Subclass 1 under Rule 23(a)(2). We review the District Court’s decision for abuse of discretion.
See Hartman v. Duffey,
19 F.3d 1459, 1471 (D.C.Cir.1994);
Wagner v. Taylor,
836 F.2d 578, 586 (D.C.Cir.1987).
Each member of Subclass 1 allegedly requested an application for a farm loan and had her request denied. Although the District Court recognized that “[p]laintiffs have assembled an impressive collection of sworn statements asserting that a great many women have been flatly refused ... loan application forms,” it nonetheless declined to certify Subclass 1. The court noted that the Appellants’ declarations “might have been sufficient to support certification of subclasses of women who suffered discrimination at the hands of particular county committees,” but variations between the declarations precluded a finding of “commonality” for purposes of a nationwide class action.
The Appellants concede that the Department offered varying reasons for denying their requests, but Appellants argue these variations are irrelevant “because there is no basis for refusing to distribute loan applications to requestors, pursuant to USDA’s own regulations.”
See
7 C.F.R. § 1910.4(b) (“All persons requesting an application will be provided [one].”). Appellants further concede that they have no statistical proof that women’s requests for loan applications were treated differently from men’s. Nevertheless, Appellants assert that a mountain of anecdotes proves that their claims are “not mere coincidence.” In the absence of any data to the contrary, the Appellants claim, the District Court should have presumed that 622 aggrieved women’s declarations warranted the certification of the class.
We disagree. As we have previously noted:
[Tjhere is more to a showing of commonality than a demonstration that class plaintiffs suffered discrimination on the basis of membership in a particular group.... While in a case alleging intentional discrimination, such as this one, a plaintiff need not isolate the particular practice and prove that such practice
caused
the discrimination, plaintiffs must make a significant showing to permit the court to infer that members of the class suffered from a common policy of discrimination that pervaded all of the employer’s challenged employment decisions.
Hartman,
19 F.3d at 1472 (emphasis in original). Thus, commonality under Rule 23(a) requires a named plaintiff to show (i) discrimination (ii) against a particular group (iii) of which the plaintiff is a member,
plus
(iv) some additional factor that “permit[s] the court to infer that members of the class suffered from a common policy of discrimination.”
Id. Hartman’s
“common policy” requirement is consistent with the Supreme Court’s teaching in
General Telephone Co. of the Southwest v. Falcon,
457 U.S. 147, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982). There the Court noted:
Conceptually, there is a wide gap between (a) an individual’s claim that he has been [discriminated against], and his otherwise unsupported allegation that the company has a policy of discrimination, and (b) the existence of a class of persons who have suffered the same injury as that individual, such that the individual’s claim and the class claims
will share common questions of law or fact .... For respondent to bridge that gap, he must prove much more than the validity of his own claim.
Id.
at 157-58, 102 S.Ct. 2864.
The District Court did not abuse its discretion in concluding that Appellants failed to “bridge th[e] gap” between their individual claims and Hartman’s “common policy” requirement. Some members of Subclass 1 (namely, the 622 declarants) offered anecdotal evidence that they were personally denied loan applications, and some of those same women reported discriminatory treatment by USDA officials. For example, the declarants allege that USDA officials told them “they were too early to apply for a loan, too late to apply for a loan, that they need not bother filling out an application because they were not eligible to receive a loan, or that their husbands should apply.” Appellants’ Br. at 11. Certainly these allegations may give the declarants standing to bring
individual
suits. However, the declarants’ allegations did not require the District Court to infer the existence of a “common policy of discrimination” that affected the
non-declarants,
as well. The bald allegation that the declarants and non-declarants alike are unified by a “common policy” of gender discrimination is insufficient to show the District Court abused its discretion under
Hartman
and
Falcon.
As the Appellants candidly admit, there is no evidence that women were more likely than men to be refused loan applications.
See
Appellants’ Br. at 13. Moreover, the declarants’ anecdotal accounts differ widely, and their complaints of discrimination are interspersed with nondiscriminatory evidence and innocuous explanations.
Compare
Joint Appendix (“J.A.”) 538 (“I was told there were no f[u]nds available.”),
and
J.A. 540 (“I was told that there were no funds or applications available.”),
with
J.A. 536 (A USDA official “told me that farming business was too risky for women.”). Thus, even taking as true all of the declarants’ factual allegations, the District Court was nevertheless left to wonder whether some of the declar-ants (to say nothing of the silent, non-declarants) were victimized by gender discrimination. Because the Appellants failed to “bridge th[e] gap” between their individual and class-wide claims, we cannot fault the District Court for refusing to take a leap of faith.
A recent case from the Sixth Circuit illustrates our point. In
Reeb v. Ohio Dep’t of Rehab. & Corr.,
435 F.3d 639 (6th Cir.2006), four female prison guards filed a putative class action against their employer, alleging it violated Title VII by treating them differently from the male guards. Seeking to represent all past and present female employees at the prison, the named plaintiffs claimed to have been the victim of the defendant’s “general policy” of discrimination on the basis of sex. The district court certified the class and found Rule 23(a)’s commonality requirement met “[b]ecause the issue of whether the Defendant violated Title VII will be common to the class.”
Id.
at 644 (internal quotation marks and citation omitted). The appeals court reversed for an abuse of discretion because “[i]f this were the test, every plaintiff seeking to certify a class in a Title VII action would be entitled to that certification.”
Id.; see also Bacon v. Honda of Am. Mfg., Inc.,
370 F.3d 565, 571 (6th Cir.2004) (“Conclusory allegations and general assertions of discrimination are not sufficient to establish commonality.” (citing
Falcon,
457 U.S. at 157, 102 S.Ct. 2364));
Sprague v. Gen. Motors Corp.,
133 F.3d 388, 397 (6th Cir.1998) (“It is not every common question that will suffice [to show commonality], however; at a sufficiently abstract level of generalization, al
most any set of claims can be said to display commonality: What we are looking for is a common issue the resolution of which will advance the litigation.”)-
After conducting a “rigorous analysis” of the Appellants’ claims under Rule 23(a),
Hartman,
19 F.3d at 1473, the District Court in this case exercised its considered judgment and concluded that class certification would not advance the litigation.
Cf. Pigford v. Glickman,
185 F.R.D. 82 (D.D.C.1999) (chronicling the arduous process of handling a similar case as a class action). Even if Subclass 1 were certified, the District Court concluded, the outcome of this case would nevertheless turn on a series of individualized inquiries into application-distribution practices in more than 2,700 FSA offices across the country over the last quarter-century. Unlike the
Reeb
court, we need not consider whether the contrary conclusion would have constituted an abuse of discretion. For present purposes, it is sufficient to note that while it might not be reversible error to certify a subclass under these or similar circumstances,
see Chiang v. Veneman,
385 F.3d 256, 265-66 (3d Cir.2004) (affirming the certification of a class akin to Subclass 1), it is also not an abuse of discretion for the District Court to refuse to do so,
see Eubanks v. Billington,
110 F.3d 87, 96 (D.C.Cir.1997).
We hasten to emphasize that we do not hold that anecdotal evidence alone is inherently insufficient to justify class certification. In this case, however, the District Court did not abuse its discretion by refusing to certify Subclass 1 for lack of coim monality under Rule 23(a)(2) on the diverse affidavits before it, and the large number of silent, potential class plaintiffs. As a result, we have no occasion to eonsider the District Court’s alternative holdings under Rule 23(b).
See Amchem,
521 U.S. at 613-14, 117 S.Ct. 2231.
Appellants mount a two-prong challenge to the District Court’s analysis of Subclass 2, which comprises women applicants who were allegedly denied farm loans on the basis of their gender. The District Court concluded that the geographic dispersal and decentralized organization of the USDA’s loan offices “cut[] against any inference for class action commonality.” Moreover, the court noted that the Department offered a wide array of objective, highly individualized justifications (such as the “failure to meet collateral requirements, poor credit, [and] insufficient income”) for denying the Appellants’ loan requests. Appellants first claim the District Court erred as a matter of law. Second, Appellants claim the District Court abused its discretion by ignoring the facts. We address (and reject) each claim in turn.
Appellants argue that the District Court erred as a matter of law in its evaluation of commonality within Subclass 2. In Appellants’ view, “[t]he District Court created its own entirely new standard for determining how subjective a system must be in order to demonstrate a common policy of discrimination.” Appellants argue that Subclass 2 is unified by the USDA’s policy of delegating standardless discretion to local loan-making officers, and “[t]his Court” has held that “a common policy [of discrimination] exist[s] when significantly subjective decision-making operates on a national basis with discriminatory results.”
Accordingly, Appellants criticize “the District Court’s substitution” of its own legal standard — which Appellants dub a “subjectivity spectrum” — “in lieu of well-established precedent.” Our review of the question of law is de novo.
See, e.g., Gov’t of Rwanda v. Johnson,
409 F.3d 368, 372 (D.C.Cir.2005).
Appellants are wrong about the law of “[t]his Court.” We have never held that subjective decision-making processes require a district court to find commonality for purposes of class certification.
See, e.g., Hartman,
19 F.3d at 1472 (holding plaintiffs’ challenge to the defendant’s “subjective” decision-making did not warrant a finding of commonality; “While plaintiffs’ statistics may have demonstrated that discrimination against women applicants to the six [] job categories was afoot, nothing in the record so far permits the additional inference that class members suffered a
common
injury.” (emphasis in original));
see also Wagner,
836 F.2d at 594-95. The District Court’s application of
Hartman
was not erroneous as a matter of law.
Appellants next argue that the District Court abused its discretion by failing to properly consider evidence of commonality within Subclass 2. Appellants rely on 859 declarations from disappointed female loan applicants, who “attested that they were denied loans based upon their gender.” However, by the Appellants’ own count, 41% of these declarants have no idea why their loan applications were denied.
See
Appellants’ Br. at 21 (asserting that 27% received “no official written response” to their loan applications, while an additional 14% received “the equivalent of no response”). The District Court was well within the bounds of its discretion to find a lack of commonality where two out of every five of the Appellants’ own declar-ants — to say nothing of the silent non-declarants — would be forced to prove at trial that individual reasons for their loan denials were not pretextual
(e.g.,
one woman had poor credit, another had no collateral, a third was victimized by gender discrimination, etc.). For the same reasons that justified the District Court’s conclusion that Subclass 1 does not exhibit “commonality,” the record in this case does not compel the conclusion that the USDA discriminated against women
across the class,
even if some
individual
declarants in Subclass 2 properly stated a discrimination claim.
See Hartman,
19 F.3d at 1472.
Appellants’ efforts to demonstrate commonality through rudimentary statistics are equally unavailing. Appellants’ statistics expert, Patrick O’Brien, admits that his analysis does not account for the possibility that women receive fewer loans because they apply for fewer loans, and he recognizes that women may receive smaller loans because they operate smaller farms. Moreover, there are countless other, non-discriminatory explanations for any patterns in the USDA’s lending data. For example, women may have relatively less credit than men; women may have relatively less wealth than men; or women may apply for loans at a younger — and thus riskier — age than men. Instead of conducting a relatively simple statistical analysis (such as a multiple regression) to control for any or all of these variables, O’Brien simply reported a series of elementary cross-tabulations, from which it is impossible — as a statistical matter — to draw meaningful conclusions.
See, e.g.,
William H. Greene, Econometric Analysis 100-01, 147-60 (4th ed.2000); Frank T. Denton,
The Significance of Significance: Rhetorical Aspects of Statistical Hypothesis Testing in Economics, in
The Consequences Of Economic Rhetoric 163 (Arjo
Klamer, Donald N. McCloskey & Robert M. Solow eds., 1988). The Appellants’ expert concedes that “[n]o statistical tests for significance were done” on his data.
Compare Bazemore v. Friday,
478 U.S. 385, 399 n. 9, 106 S.Ct. 3000, 92 L.Ed.2d 315 (1986) (holding the Appellants’ empirical evidence of discrimination was admissible to show disparate treatment because it showed “statistically significant” disparities).
Given the numerous shortcomings in the Appellants’ efforts to show commonality within Subclass 2, the District Court did not abuse its discretion in denying the Appellants’ certification motion under Rule 23(a)(2). Accordingly, we have no occasion to consider the District Court’s alternative holdings under Rule 23(b).
See Amchem,
521 U.S. at 613-14, 117 S.Ct. 2231.
III. FailuRE-To-Investigate Claim
Appellants’ final argument is that the District Court erred in dismissing their “failure-to-investigate” claim prior to ruling on their motion for class certification. The District Court concluded the USDA’s alleged failure to investigate Appellants’ complaints is not reviewable under ECOA, which makes it “unlawful for any creditor to discriminate against any applicant with respect to any aspect of a
credit transaction”
on the basis of sex. 15 U.S.C. § 1691(a) (emphasis added). The court concluded, as a matter “of course,” that a failure to investigate does not fit within the plain meaning of a “credit transaction.” The District Court further concluded that Appellants’ failure-to-investigate claim is not reviewable under the APA because ECOA provides an adequate alternative remedy.
See
5 U.S.C. § 704 (limiting the applicability of the APA to a “final agency action for which there is no other adequate remedy in a court”).
Appellants argue the District Court erred under both ECOA and the APA. First, the Appellants urge us to construe liberally the meaning of a “credit transaction” under 15 U.S.C. § 1691(a) and to hold their failure-to-investigate claim is reviewable under ECOA. Alternatively, the Appellants argue their claim is reviewable under the APA because the USDA’s regulations empower aggrieved applicants to file discrimination claims, which the Department is legally obligated to investigate. Our review is de novo.
See, e.g., Johnson,
409 F.3d at 372.
We first reject the Appellants’ argument under ECOA because even a liberal interpretation of “credit transaction” does not encompass “failure to investigate a discrimination complaint.”
Cf. Bissette v. Colonial Mortgage Corp. of D.C.,
477 F.2d 1245, 1247 (D.C.Cir.1973) (noting “ ‘liberal construction’ and interpretation based on legislative purpose can only go so far. Where the meaning of the statute and regulations is clear, a contrary reading would become destruction of the statutory scheme.”). True, the phrase “credit transaction” is not without ambiguity, but it is not a normal understanding of those words in the English language that they would include the failure to investigate complaints of discrimination in an underlying event even if the underlying event falls within the meaning of the phrase “credit transaction.”
Again, there is a certain ambiguity in the term. The Department has gone to the trouble of issuing a defining regulation.
See
12 C.F.R. § 202.2(m). The regulatory scheme of which that is a part may become the stuff of APA litigation upon remand, but so far as an action at law based on this section of the ECOA, that simply is not within the plain meaning of the statute.
This leaves the Appellants’ claim that the USDA’s failures to investigate
discrimination claims are renewable under the APA for another day. We follow this course of nonaction because the record is little developed on that argument, and the parties’ briefs have given short-shrift to the issue before us. As our review of this claim is discretionary in the first place, and as the order certifying the 12(b)(6) ruling for interlocutory review was without prejudice to further decision by the merits panel of this court, we decline to exercise our discretion and remand the APA claim for further proceedings in the District Court.
See In re Veneman,
309 F.3d at 796.
Conclusion
For the reasons stated above, the District Court’s denial of the Appellants’ motion for class certification is affirmed. The District Court’s dismissal of the Appellants’ failure-to-investigate claim under ECOA is affirmed, but its dismissal of the Appellants’ failure-to-investigate claim under the APA is remanded.
So ordered.