Opinion by Judge FERGUSON; Dissent by Judge REINHARDT.

Various municipalities and officials (“Appellants”) filed the instant lawsuit against the Department of Commerce and Secretary of Commerce Donald Evans to compel the Secretary to adopt statistically adjusted population data as the official Census 2000 for redistricting purposes. Secretary Evans decided against its use based on the recommendation of a committee of senior Census Bureau officials and statisticians who, after extensive study, expressed serious concerns regarding possible flaws in the statistical (or sampling) methodology.

The District Court rejected Appellants’ claim that Secretary Evans’ decision not to use the adjusted data (hereinafter “adjustment decision”) violated 13 U.S.C. § 195 , which requires the Secretary of Commerce to use “sampling” in conducting the census if “he considers it feasible.” The District Court found moot Appellants’ claim that Secretary Evans violated the notiee-and-comment requirements of the Administrative Procedure Act (“APA”) in revoking former Secretary of Commerce William Daley’s delegation of the adjustment decision to the Director of the Bureau of Census. We affirm.

I. BackgRound

The Constitution requires Congress to conduct a decennial census “in such Manner as they shall by Law direct” for apportioning congressional representation to the states. U.S..CONST, art. I, § 2, cl. 3; id amend. XIV, § 2. By means of the Census Act, Congress has delegated broad discretion to the Secretary of Commerce to conduct the census “in such form and content as he may determine, including the use of sampling procedures and special surveys.” 13 U.S.C. § 141 (a). The Secretary, in turn, oversees the Bureau of the Census, an agency within the Department of Commerce responsible for conducting the decennial census. 13 U.S.C. §§ 2 , 4 & 21.

Although the Constitution mandates only that the census be taken for reapportionment purposes, the census data is used for myriad other purposes. Wisconsin v. City of N.Y., 517 U.S. 1, 5-6 , 116 S.Ct. 1091 , 134 L.Ed.2d 167 (1996) (describing secondary uses of census data). Appellants challenge the Census 2000 results because of their effect on the allotment of federal and state funds, as well as the redistricting of their respective state legislatures.

A. Sampling and the Census Act

At issue in this case is whether the Secretary was required to statistically adjust the results of Census 2000 pursuant to 13 U.S.C. § 195 , as amended in 1976. Since 1940, the Bureau has employed sampling techniques to gather supplemental information regarding the population. U.S. Bureau of the Census, Dep’t of Commerce, Accuracy and Coverage Evaluation: Statement on the Feasibility of Using Statistical Methods to Improve the Accuracy of Census 2000, 65 Fed.Reg. 38, 374, 38, 382 & n. 21 (June 20, 2000) (“A.C.E. Report”).

In 1957, the Secretary requested that Congress enact a law that would clearly authorize the use of sampling in conducting the census. Utah v. Evans, — U.S. -, -, 122 S.Ct. 2191, 2202 , 153 L.Ed.2d 453 (2002) (citing legislative history); Dep’t of Commerce v. U.S. House of Reps., 525 U.S. 316, 336 , 119 S.Ct. 765 , 142 L.Ed.2d 797 (1999) (amended opinion) (same). In response, Congress amended the Census Act by enacting § 195, which *865 provided that the Secretary “may, where he deems it appropriate, authorize the use of the statistical method known as ‘sampling’ in carrying out the provisions of this title” for purposes other than apportionment. Census Act of 1957, Pub.L. No. 85-207, § 14 , 71 Stat. 481 , 484 (1957) (codified at IS U.S.C. § 195 (1957)); see Deft of Commerce, 525 U.S. at 336 , 119 S.Ct. 765 . In 1976, Congress amended § 195 to provide:

Mid-Decade Census of Population Act of 1976, Pub.L. No. 94-521, § 10 , 90 Stat. 2459 , 2464 (1976) (emphasis added).

Since the enactment and revision of § 195, the Bureau has considered the use of statistical sampling to compensate for the error rate in the decennial census. In particular, it has studied the possibility of implementing statistical adjustments to the census data to compensate for the undercount of the population and the disproportionate undercount (or “differential undercount”) of racial and ethnic minorities, children, and renters. See 65 Fed. Reg. 38 , 376; see also Wisconsin, 517 U.S. at 7-8 , 116 S.Ct. 1091 . However, for both Census 1980 and Census 1990 (and most recently for Census 2000), the Secretary ultimately decided against the use of the statistically adjusted data as the official census and relied instead on the headcount data. Id. at 8 , 116 S.Ct. 1091 .

B. Census 2000

In 1997, the Bureau submitted a plan to Congress for Census 2000. See U.S. BUREAU OF THE CENSUS, DEP’T OF COMMERCE, REPORT TO CONGRESS: THE PLAN FOR CENSUS 2000, 23 (Aug.1997) (“Census 2000 Plan”). The Census 2000 Plan incorporated the findings of the National Academy of Sciences (“NAS”) that sampling could be used as an alternative to traditional means of enumeration. 1 Accordingly, the Census 2000 Plan provided for a “limited use of sampling” in ascertaining the population. Significantly, the Census 2000 Plan would have resulted in one set of census results: the initial headcount data adjusted by statistical sampling techniques. Id. at 52 .

The Census 2000 Plan precipitated legal challenges to the use of statistical sampling in determining the population. In 1999, the Supreme Court held that “the Census Act prohibits the proposed uses of statistical sampling in calculating the population for purposes of apportionment.” Deft of Commerce, 525 U.S. at 343 , 119 S.Ct. 765 . Following Department of Commerce, the Bureau modified its plan to statistically adjust the Census 2000 only for purposes other than apportionment. 65 Fed.Reg. 38,377.

1. The Accuracy and Coverage Evaluation (“A.C.E.”) Report

On June 2000, the Bureau issued the A.C.E. Report, which described the statistical methodology to be used for Census 2000. The Report set forth the Bureau’s preliminary determination that the use of the statistically adjusted data was feasible *866 in terras of both the statutory deadline for releasing redistricting information and the improvement of the census’ accuracy. Id. at 38,374.

The A.C.E. methodology is based on a “dual system estimation,” also referred to as “capture/recapture.” Id. at 38,378. This statistical method compares the results of the initial census with those of post-enumeration surveys from sample blocks of the population to determine the under- or over-count of initial census data according to post-strata characteristics (geography, age, sex, race, and housing status). Id. It then extrapolates the rate of error for each group to the entire population and statistically adjusts the initial data on a nationwide scale. Id.; see generally Wisconsin, 517 U.S. at 8-9 , 116 S.Ct. 1091 (colorfully describing capture/recapture by analogy to counting pumpkins in a large pumpkin patch).

The A.C.E. Report made a preliminary determination that it was feasible to use the A.C.E. methodology to produce statistically adjusted census data by the April 1, 2001, statutory deadline. 2 65 Fed.Reg. 38,374. The Report defined § 195’s statutory term “feasible” as encompassing two components: “operational” and “technical” feasibility. Id. Operational feasibility referred to the logistical possibility of adjustment in light of “other aspects of the census plan, timing, and funding constraints.” Id. at 38,379. In contrast, technical feasibility referred to the determination that “the use of sampling is expected to improve the overall accuracy of census data.” Id.

Although the A.C.E. Report reflected an initial prediction that the A.C.E. would improve the census, the Bureau needed to study the data further to determine whether the A.C.E. had met its expectation of “improv[ing] overall numeric and distributive accuracy and ... reducing] the differential under-count.” Id. at 38,387. The A.C.E. Report recommended that “statistical correction is appropriate absent strong evidence that it will degrade the overall quality of the final census data.” Id. However, it cautioned:

Id. at 38,393.

To ensure that the final decision was insulated from political manipulation, the Bureau proposed that the Executive Steering Committee for A.C.E. Policy (“ESCAP”) — a committee of senior Bureau officials and statisticians — assess the effectiveness of A.C.E. and “make a recommendation to the Census Bureau Director regarding the use of the statistically corrected census data.” Id. at 38,394.

2. Promulgation and Revocation of the “Daley Rule”

On October 6, 2000, former Secretary of Commerce William Daley promulgated a rule in compliance with the notiee-and-comment procedures of the APA, delegating the adjustment decision to the Director of the Bureau. 3 Report of Tabulations of Population to States and Localities Pursuant to 13 U.S.C. [§ ]141(c) and Availability *867 of Other Population Information, 65 Fed. Reg. 59 , 713, 59, 716 (Oct. 6, 2000) (codified at 15 C.F.R. § 101.1 (a) (2000)) (“Daley Rule”). The Daley Rule required that ES-CAP’s recommendation regarding the use of A.C.E. data be published. 15 C.F.R. § 101.1 (b)(1). It also provided that the ESCAP Report would be released to the public and to the Census Director simultaneously. Id. § 101.1(b)(2). In addition, the Daley Rule required that the adjusted data be released to the public if the Census Director decided, against the recommendation of ESCAP, not to use the adjusted data. Id. § 101.2(b).

On February 16, 2001, Secretary Evans revoked the Daley Rule in all respects, except for the release of the ESCAP Report to the public. Report of Tabulations of Population to States and Localities Pursuant to 13 U.S.C. §§ 141 (c) and Availability of Other Population Information; Revocation of Delegation of Authority, 66 Fed. Reg. 11,231 (Feb. 23, 2001). The revocation did not provide for a public comment period, based on Secretary Evans’ finding that it was a “rule of agency organization, procedure and practice.” Id. at 11 ,232 (citing 5 U.S.C. § 553 (b)(3)(A)).

3. The ESCAP Report

In accordance with the A.C.E. Report’s recommendations, ESCAP studied the proposed statistical adjustment to Census 2000. As defined by regulation, ESCAP was chaired by the Bureau’s Associate Director for Decennial Census and included senior statisticians, demographers, and other senior professionals. 15 C.F.R. § 101.1 (b)(3). On March 1, 2001, ESCAP unanimously recommended against the use of the A.C.E. data. See Executive Committee for Accuracy and Coverage Evaluation Policy, Recommendation Concerning the Methodology to be Used in Producing the Tabulations of Population Reported to States and Localities Pursuant to 13 U.S.C. [§ ]141(c) (Mar. 1, 2001) (“ESCAP Report”), Report Notice, 66 Fed.Reg. 14,-004 (Mar. 8, 2001).

ESCAP studied the operation of the A.C.E. according to the criteria set forth in the A.C.E. Report, including: (1) conduct of key operations, (2) consistency with historical measures of undercount, and (3) quality measures. Id. After many weeks of “examining voluminous evidence,” ES-CAP was unable to recommend the use of the A.C.E. data due to unresolved concerns about its reliability. Id. at 14,005. In particular, ESCAP found that the A.C.E. data diverged widely from the demographic analysis’ estimate of the under-count and reflected possible synthetic and balancing errors. 4 66 Fed.Reg. 14,005. Although further research may have proved the A.C.E. data to be more accurate, ESCAP’s “uncertainty due to these concerns is too large at this time to allow for a recommendation to adjust.” Id.

4. Adjustment Decision

Following the release of the ESCAP Report, the Acting Director of the Bureau, William Barron, Jr., adopted ESCAP’s recommendation not to use the A.C.E. data for redistricting purposes. Subsequently, Secretary Evans adopted both the Acting Director’s and ESCAP’s recommendation not to use the A.C.E. data and directed that only the unadjusted data be released for purposes of redistricting.

*868 C. Procedural History

On February 21, 2001, Appellants brought this lawsuit against Secretary Evans, challenging his revocation of the Daley Rule without notice and comment. Following the release of the ESCAP Report and Secretary Evans’ adjustment decision, Appellants moved for a temporary restraining order and injunctive relief on the grounds that Secretary Evans’ adjustment decision deprived them of federal and state funding as well as proper representation in their respective state legislatures. They argued that the adjustment decision violated 13 U.S.C. § 195 because he had decided against the use of sampling even though it was “feasible.”

The District Court denied both preliminary and permanent injunctive relief and dismissed the action. Applying the standard set forth in Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 , 104 S.Ct. 2778 , 81 L.Ed.2d 694 (1984), the District Court found that deference to Secretary Evans’ interpretation of § 195 was appropriate because the statute did not reveal Congress’ intent regarding the Secretary’s obligation to release the adjusted data. In addition, the District Court found Secretary Evans’ interpretation of § 195 to be a permissible construction of the statute. Finally, the District Court concluded that ESCAP’s application of this standard was reasonable in light of the inconsistency of the A.C.E. data with the demographic analysis, as well as concerns with synthetic and balancing errors. Accordingly, the District Court upheld Secretary Evans’ adjustment decision.

II. STANDARD OF REVIEW

We review de novo the District Court’s interpretation of § 195. Silver Sage Partners, Ltd. v. City of Desert Hot Springs, 251 F.3d 814, 819 (9th Cir.2001). We must affirm its “decision to deny in-junctive relief ... unless it incorrectly applied the law, relied on clearly erroneous factual findings, or otherwise abused its discretion.” Contract Servs. Network, Inc. v. Aubry, 62 F.3d 294 , 297 (9th Cir.1995).

III. Discussion

In this case, Appellants seek to enjoin Secretary Evans’ adjustment decision as violative of § 195 because, they contend, the use of the A.C.E. data for redistricting purposes was “feasible.” In addition, Appellants argue that Secretary Evans’ adjustment decision was invalid because the authority to make that decision had been delegated to the Director of the Bureau under the Daley Rule, and his revocation of the rule failed to comply with the APA’s notice-and-comment procedures.

A. Legality of Secretary’s Adjustment Decision under § 195

This is not the first time that the Secretary of Commerce’s adjustment decision has been the subject of a legal challenge. Following Census 1990, the Supreme Court upheld the Secretary’s decision not to statistically adjust the census as “well within the constitutional bounds of discretion over the conduct of the census.” Wisconsin, 517 U.S. at 24 , 116 S.Ct. 1091 . More recently, the Supreme Court held in Department of Commerce that § 195 prohibited the Secretary from statistically adjusting the census for purposes of apportionment. 525 U.S. at 343 , 119 S.Ct. 765 . In this case, we are called upon to answer a question not answered by either Wisconsin or DepaHment of Commerce: whether § 195 imposes an obligation on the Secretary to statistically adjust the census for redistricting purposes. 5

In reviewing Secretary Evans’ interpretation of § 195, we follow the two-part test established in Chevron. First, *869 we must determine whether Congress’ intent is clear as to the precise issue involved. Chevron, 467 U.S. at 842 , 104 S.Ct. 2778 . If we answer this question in the affirmative, our inquiry is at an end. Id. at 843 , 104 S.Ct. 2778 . “However, if the statute is ambiguous as to the question at issue, then we go to the second step and determine the meaning of the statute’s language by giving deference to the governing agency’s interpretation of the statute’s language.” Royal Foods Co. v. RJR Holdings Inc., 252 F.3d 1102, 1106 (9th Cir.2001) (citing Chevron, 467 U.S. at 842-44 , 104 S.Ct. 2778 ).

1. Statutory Analysis of § 195

Our starting point in construing § 195 is the language of the statute itself. Id. (citing Chevron, 467 U.S. at 842-44 , 104 S.Ct. 2778 ). “We look to the text of the statute to ‘determine whether the language at issue has a plain and unambiguous meaning with regard to the particular dispute in the ease.’ ” Id. (quoting Robinson v. Shell Oil Co., 519 U.S. 337, 340 , 117 S.Ct. 843 , 136 L.Ed.2d 808 (1997)).

The relevant provision of § 195 states that “the Secretary shall, if he considers it feasible, authorize the use of the statistical method known as ‘sampling’ in carrying out the provisions of this title.” 13 U.S.C. § 195 . From the statutory language, the parties ascribe very different meanings. Appellants urge us to find that § 195 creates a presumption in favor of the adjusted data and admits of little administrative discretion. Appellees, on the other hand, argue that § 195 confers broad discretion upon the Secretary to determine in what instances and against what standard the use of sampling should be found feasible. 6

a. § 195 Confers Discretion upon the Secretary

First, we disagree with Appellants that § 195 creates a presumption in favor of the adjusted data. Appellants conflate the Secretary’s feasibility determination with what should follow from his finding of feasibility. Thus, Appellants conveniently skip over the critical first step — whether *870 the Secretary considered it feasible to adopt the adjusted data as the official 2000 census for redistricting purposes. Yet, the operation of § 195 hinges entirely on this condition.

Our reading of the plain statutory text is that the combination of terms “shall” and “if he considers it feasible” demonstrates Congress’ intent to strongly encourage the use of sampling while conferring meaningful discretion upon the Secretary to determine whether statistics should be used in any given instance. The language of § 195 is straightforward. It requires the Secretary to incorporate the use of statistical methodologies in the taking of the census if and where he considers it feasible. By its plain terms, the statute directs the Secretary to use sampling, but only “if he considers it feasible.” 13 U.S.C. § 195 . As a corollary, no sampling need be used if the Secretary does not consider it feasible.

Section 195 does not specify under what conditions the Secretary should find the use of sampling feasible. Nowhere does § 195 state, as Appellants would have us read it, that the Secretary should presume that statistical information is more reliable than unadjusted headcount data, or that he should strain to find the use of sampling feasible even though he'is not fully certain about the ramifications of that choice. Instead, the statute vests discretion in the Secretary both to set the standard for feasibility and to decide whether that standard has been met.

Ignoring the clear statutory language “if he considers it feasible,” Appellants argue that § 195 requires the Secretary to statistically adjust the census “if feasible.” 7 However, it is “a cardinal principle of statutory construction that a statute ought, upon the whole, to be so construed that, if it can be prevented, no clause, sentence, or word shall be superfluous, void, or insignificant.” TRW Inc. v. Andrews, 534 U.S. 19 , 122 S.Ct. 441, 449 , 151 L.Ed.2d 339 (2001) (internal quotation marks and citation omitted). Thus, we cannot accept Appellants’ proposed statutory interpretation because it renders the “if he considers it” portion of the phrase meaningless.

In addition, Appellants argue that the statutory language in § 195 creates a presumption in favor of the adjusted data. They cite Board of Pardons v. Allen, 482 U.S. 369, 378 , 107 S.Ct. 2415 , 96 L.Ed.2d 303 (1987), as support for the proposition that the use of the term “shall,” followed by the discretionary language “if he considers it,” creates a “presumption in favor of doing that which is the subject of the statute.” However, Allen itself reveals the error in Appellants’ argument.

In Allen , the Supreme Court held that a Montana statute created a liberty interest in parole when it provided that the parole “ ‘board shall release on parole ... any [prisoner] ... when in its opinion there is reasonable probability that the prisoner can be released without detriment to the prisoner or to the community.’ ” 482 U.S. at 376 , 107 S.Ct. 2415 (first ellipses in original) (quoting MONT. CODE ANN. § 46-23-201 (1985)). In holding that the parole statute created a presumption of *871 release once the predicate conditions had been found, the Supreme Court did not engraft on the statute a presumption that the parole board find the prisoner did not pose a danger to herself or to the community. 8 Rather, the parole board retained “very broad” discretion to determine whether those conditions had been met in the first place. See id. at 381 , 107 S.Ct. 2415 (relying on Greenholtz v. Inmates of Neb. Penal & Corr. Complex, 442 U.S. 1, 13 , 99 S.Ct. 2100 , 60 L.Ed.2d 668 (1979)).

Similarly, § 195 does not stack the deck in favor of a finding of feasibility by creating a presumption that the adjusted data is more accurate than the unadjusted or headcount data. We would be remiss if we were to interpret § 195 in such a way as to require the Secretary to statistically adjust the census even if he and a panel of experts were uncertain as to the reliability of the adjusted data, just as the Montana legislature surely did not intend for the state parole board to release a prisoner whose potential for future detriment could not be ascertained. The better reading of the statute is that Congress intended to strongly encourage the use of sampling; however, it left to the Secretary meaningful discretion whether a sampling methodology should be used in any given instance.

Viewing § 195 in the context of the Census Act as a whole, we are further convinced that Congress has delegated to the Secretary discretion to choose appropriate methodologies and different sampling techniques. Most notably, § 141(a) confers broad discretion to the Secretary to conduct the census “in such form and content as he may determine, including the use of sampling procedures and special surveys.” 13 U.S.C. § 141 (a).

Because Congress conditioned the use of sampling on the Secretary’s consideration of its feasibility, § 195 does not create a presumption in favor of statistical adjustment of the census, nor does it require the Secretary to consider the adjusted data as the default data for Census 2000. Instead, § 195 grants broad discretion upon the Secretary to “consider” as an initial matter what uses of sampling are “feasible.”

b. The Term “Feasible” is Ambiguous

Appellants also argue that § 195 admits of little administrative discretion because the Secretary is empowered to decide only whether a use of sampling is minimally feasible, i.e., “capable of being done, executed, or effected.” However, unlike other cases in which the statutory term “feasible” was limited to physical possibility, this proposed definition would run counter to the statutory objectives of the Census Act. Thus, we find that the term “feasible” as used in § 195 is an ambiguous term left open to the Secretary’s reasonable interpretation.

It is well established that we must consider the context in ascertaining the meaning of the statute. Brower v. Evans, 257 F.3d 1058, 1065 (9th Cir.2001). The context relevant to our analysis “relates to ‘the design of the statute as a whole and to its object and policy.’ ” Id. (quoting Gozlon-Peretz v. United States, 498 U.S. 395, 407 , 111 S.Ct. 840 , 112 L.Ed.2d 919 (1991)). We “ ‘may consider the purpose of the statute in its entirety, and whether the proposed interpretation would frustrate or advance that purpose.’ ” Id. (quoting United States v. Mohrbacher, 182 F.3d 1041, 1049 (9th Cir.1999)).

*872 Appellants do not dispute the fact that the overall goal of the Census Act is accuracy. See Wisconsin, 517 U.S. at 5 , 116 S.Ct. 1091 (so stating). Yet, their proposed definition of “feasible” would run counter to this goal. Their proposed definition would require the Secretary to use sampling so long as it was capable of being done even if the methodology was flawed and produced skewed , results. Thus, “feasible” must mean more than sheer physical or financial possibility. Indeed, the plain meaning of “feasible” incorporates whether a particular action is “capable of .being successfully done or accomplished.” BLACK’S LAW DICTIONARY 739 (4th ed.1957) (emphasis added). Whether a task can be.done “successfully” necessarily derives its meaning from the objective at hand. Because the fundamental objective of the Census Act is accuracy, the use of sampling could not be successful if its methodology or result was inaccurate.

Accordingly, § 195 must be read in light of the overall purpose of the Census Act in order to avoid flawed results. Nevertheless, it is unclear to what extent the Secretary may consider accuracy in his feasibility determination. Must he be 50%, 75%, or 99% certain that the use of sampling would produce accurate results? Section 195 leaves these questions unanswered. Because § 195 does not specify the circumstances under which the Secretary must find sampling feasible, the statutory term remains ambiguous. See Biodiversity Legal Found. v. Babbitt, 146 F.3d 1249 , 1254(10th Cir.1998) (finding that statutory phrase “to the maximum extent practicable” was ambiguous because its ordinary meaning did not “specif[y] the circumstances” under which the agency could forgo its statutory 90-day deadline).

Further, the choice of language “if he considers it” as a pre-condition of “feasible” demonstrates that Congress intended for the Secretary to make such judgment calls. This phrase indicates that Congress did not intend to limit the Secretary’s discretion to a finding of whether a particular use of sampling is capable of being done. Rather, it left the choice to the Secretary as to whether sampling could be used, bringing to bear his expertise on the effectiveness of different statistical methodologies and their compatibility with the other aspects of the census. Thus, unlike other cases in which the agency had “little administrative discretion” in making a feasibility determination, § 195 reflects Congress’ intent for the Secretary to strike a balance as to the feasibility of using sampling in any given instance. See Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 411 , 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971) (holding that discretion was severely constrained when statute forbid construction on “any public parkland unless [ ] there is no feasible and prudent alternative to the use of such land”) (quoting 23 U.S.C. § 138 (1964 ed., Supp. V)); cf. Am. Textile Mfrs. Inst., Inc. v. Donovan, 452 U.S. 490, 508-09 , 101 S.Ct. 2478 , 69 L.Ed.2d 185 (1981) (holding that Congress had struck policy balance in enacting statute that agency “ ‘shall set the standard which most adequately assures, to the extent feasible, ... that no employee will suffer[health detriments]’ ”) (quoting 29 U.S.C. § 655 (b)(5) (1970)).

In sum, by providing an open definition of “feasible,” qualifying it further with the language “if he considers it feasible,” and refraining from the inclusion of any criteria to guide the Secretary’s decision, the statute conferred discretion upon the Secretary to establish his own standard for determining the feasibility of using the adjusted data. 9 13 U.S.C. § 195 (emphasis *873 added). Thus, the Secretary’s interpretation of § 195 need only be “reasonable.” Chevron, 467 U.S. at 843 , 104 S.Ct. 2778 .

2. Secretary’s Interpretation of § 195

Because we have concluded that Congress has not spoken to the precise question at issue, we must defer to the Secretary’s interpretation of § 195. We will uphold his interpretation so long as it is “based on a permissible construction of the statute.” Id. If his “reading fills a gap or defines a term in a reasonable way in light of the Legislature’s design, we give that reading controlling weight, even if it is not the answer ‘[we] would have reached if the question initially had arisen in a judicial proceeding.’ ” Regions Hosp. v. Shalala, 522 U.S. 448, 457 , 118 S.Ct. 909 , 139 L.Ed.2d 895 (1998) (quoting Chevron, 467 U.S. at 843 n. 11, 104 S.Ct. 2778 ).

The Secretary interpreted § 195’s statutory term “feasible” as having two components: (1) operational feasibility (logistical possibility) and (2) technical feasibility (increased accuracy). 65 Fed.Reg. at 38, 379-80. Operational feasibility refers to the “Bureau’s ability to conduct the A.C.E. with available resources and within required deadlines or time frames.” Id. at 38, 380. “Technical feasibility,” on the other hand, refers to whether “the A.C.E. statistical methodology ... will improve the accuracy of the census for non-apportionment uses of the data.” Id.

The A.C.E. Report set forth an objective framework for evaluating the technical feasibility of the A.C.E. data according to: “(1) the conduct of key operations, (2) the consistency of the A.C.E. results with historical measures of undercount, and (3) measures of quality.” 65 Fed.Reg. 38,387. The A.C.E. Report stated that its preliminary analysis had “led it to expect that the A.C.E. will improve overall numeric and distributive accuracy and that it will reduce the differential undercount.” Id. Thus, it tentatively recommended that “statistical correction is appropriate absent strong evidence that it will degrade the overall quality of the final census data.” Id. Nevertheless, it refrained from making a final determination until after ESCAP had studied the data in light of the above criteria. Id. Subsequently, ESCAP adopted the A.C.E. Report’s criteria for evaluating the feasibility of statistical adjustment and studied the effectiveness of the A.C.E. data to see if its expectations for improved accuracy had, in fact, been met. 66 Fed.Reg. 14,004.

We find that this interpretation of § 195 is a permissible reading of the' statute. The Secretary’s definition of “feasible” reflects the plain meaning of the term feasible, i.e., whether the adjustment of Census 2000 was “capable of being successfully done or accomplished.” BLACK’S LAW DICTIONARY 739 (4th ed.1957). The inclusion of accuracy in the Secretary’s interpretation of “feasible” is reasonable in light of Congress’ objective to attaining as accurate a census as possible. See Regions Hosp., 522 U.S. at 457 , 118 S.Ct. 909 . Further, the criteria by which the A.C.E. data were to be evaluated creates an appropriate standard against which the achievement of improved accuracy could be assessed.

*874 Appellants accept the A.C.E. Report as the proper frame-work for guiding the Secretary’s adjustment decision. But, they argue that the A.C.E. Report established a legal standard, which bound the Secretary to order adjustment unless he found “strong evidence” that it would degrade the overall census results.

Appellants’ argument is unavailing. It is clear from the context of the A.C.E. Report that its determinations were tentative in nature. Although the A.C.E. Report suggested that “strong evidence” should be found in order to decide against adjustment, the Report cautioned that the Bureau still needed to “conduct an objective review before making a final determination to release the statistically corrected data.” 65 Fed.Reg. 38,387. It concluded that the Bureau “will not ... release corrected redistricting data until it has brought its technical judgment to bear in assessing the available data to verify that its expectations [of improved accuracy] have been met.” Id. at 38,393.

In addition, the statements made in the A.C.E. Report did not have the force of law because the A.C.E. Report was not the product of a formal adjudication or notice- and-comment rulemaking. See Christensen v. Harris County, 529 U.S. 576, 587 , 120 S.Ct. 1655 , 146 L.Ed.2d 621 (2000) (holding that opinion letters and other interpretations that are not “arrived at after, for example, a formal adjudication or.notice-and-comment rulemaking ... lack the force of law” and “do not warrant Chevron-style deference”). Therefore, we need not defer to its statements implying that statistical adjustment would be presumed unless the Secretary was able to point to strong evidence of unreliability. Id.

Accordingly, the A.C.E. Report did not limit ESCAP’s discretion to a “strong evidence” determination. Instead, the A.C.E. Report set forth an objective framework for guiding ESCAP’s analysis of the operational success of the A.C.E. according to certain specified criteria. Moreover, it made clear that the use of the adjusted data was contingent on ESCAP’s verification of its expectancy that the A.C.E. would improve the census overall.

3. Secretary’s Feasibility Determination

Having found that the Secretary permissibly interpreted § 195, we must next decide whether his adjustment decision was arbitrary or capricious. Under the APA, the Secretary’s adjustment decision must stand unless it was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706 (2)(A).

We have held that a decision is “considered arbitrary and capricious” if the agency relied on irrelevant factors, failed to consider a crucial aspect of the issue before it, offered an explanation that runs counter to the evidence, or “ ‘is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.’ ” Midwater Trawlers Co-op. v. Dep’t of Commerce, 282 F.3d 710 , 720(9th Cir.2002) (quoting Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 , 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983)). Our inquiry must “be searching and careful, but the ultimate standard of review is a narrow one.” Marsh v. Or. Natural Res. Council, 490 U.S. 360, 378 , 109 S.Ct. 1851 , 104 L.Ed.2d 377 (1989) (internal quotation marks and citation omitted). We must not substitute our judgment for that of the Secretary. See id.

In determining that the use of the A.C.E. data was not feasible, i.e., would not improve the accuracy of the census, ESCAP considered the criteria set forth in the A.C.E. Report. The ESCAP Report stated that its study included: “(1) a con *875 sideration of operational data to validate the successful conduct of the A.C.E.; (2) whether the A.C.E. measures of under-count were consistent with historical patterns of undercount and independent demographic analysis benchmarks; and (3) a review of quality measures.” 66 Fed.Reg. 14,004. Following weeks of studying “voluminous evidence,” ESCAP found some disturbing inconsistencies and possible errors in the methodology. Because it was unable to resolve its concerns “in the time available,” it recommended against the release of the adjusted data for redistricting purposes. Id. at 14,005.

In analyzing the A.C.E. data, ESCAP found a “surprising inconsistency between the DA [demographic analysis] and A.C.E. estimates.” Id. at 14,012. Whereas the demographic analysis estimated that the unadjusted census overcounted the population by 1.8 million persons, the A.C.E. indicated that the census had undercount-ed the population by 3.3 million persons. Id. at 14,012-13. Even assuming that undocumented immigration had doubled during the 1990s, the demographic analysis estimated an undercount of only 0.9 million persons. Id. at 14,013. Although the source of the problem could also have been the 1990 census data or the demographic analysis techniques, ESCAP had pinpointed other concerns about corruption of the A.C.E. data and, therefore, concluded that the inconsistency might be attributable to “an as — yet undiscovered problem in the A.C.E. or census methodology.” Id. at 14,005; see also id. at 14,015.

Two factors, in particular, called into question the reliability of the A.C.E. data. First, ESCAP expressed concern that “synthetic error” might have corrupted the data. Id. at 14,015. Synthetic error, in this case, may have resulted from the assumption that persons in a particular group would be undercounted by the initial census consistently nationwide. Id. Based on its studies, ESCAP found that synthetic error might have led to an error in estimating the population of certain groups by as much as 58%. Id. at 14,016.

Second, ESCAP was concerned that “balancing error” had occurred. Id. The A.C.E. was conducted by use of two different surveys: the P-sample (an independent survey used to measure missed persons) and the E-sample (a re-examination of census records used to measure erroneous inclusion of persons). Id. Balancing error may have occurred because the P-samples and the E-samples were handled differently, with more time and effort spent on the P-samples. Id. Accordingly, ESCAP found that the “error could introduce an upward bias” in the A.C.E. data, thereby explaining the greater undereount found by the A.C.E. Id.

Based on these reasons, ESCAP recommended against the use of the A.C.E. data. It explained:

Id. at 14,005.

Appellants argue that the ESCAP Report misapplied the “strong evidence” standard set forth in the A.C.E. Report and reversed the presumption in favor of the adjusted data. Their argument rests upon statements in the ESCAP Report that “the majority of the evi *876 dence” suggested the adjusted data was more accurate. However, Appellants mischaracterize the record. Each of ES-CAP’s statements regarding the improved accuracy of the adjusted data was tempered with ESCAP’s concern regarding possible errors in the A.C.E. data or methodology. 10

Moreover, Appellants unpersua-sively attempt to construct a legal standard out of a scientific inquiry. We agree with the District Court’s reasoning that “[t]he meaning of strong evidence should not be gauged by comparison to legally derived standards of proof, such as ... ‘preponderance of the evidence.’ ” City of Los Angeles v. Evans, No. CV-01-01671GAF, slip op. at 24 (C.D.Cal. Apr. 26, 2001). Rather, considering that ESCAP’s study was a technical assessment of the A.C.E. data, “[s]trong evidence in that context was meant to mean the sort of evidence that would give professionals in the field pause” before recommending the use of the adjusted data. Id.

In the alternative, Appellants challenge the strength of the evidence against the A.C.E. data, asserting that ESCAP was being “overcautious.” They point to one article, which has asserted that ESCAP’s concerns were “weak and derivative in nature.” See (Margo Anderson & Stephen E. Fienberg, Counting and Estimation: Methodology for Improving the Quality of Censuses 14-16 (2001)).

However, because the determination as to whether ESCAP’s concerns were warranted “ ‘requires a high level of technical expertise,’ we must defer to ‘the informed discretion of the responsible federal agene[y].’ ” Marsh, 490 U.S. at 377 , 109 S.Ct. 1851 (quoting Kleppe v. Sierra Club, 427 U.S. 390, 412 , 96 S.Ct. 2718 , 49 L.Ed.2d 576 (1976)). As one district court aptly observed, disputes regarding the Bureau’s use of statistics in taking the census “are best resolved not by the courts but by the Bureau itself, whose experience with prior censuses and expertise in the collection and analysis of statistical information render it especially qualified to make the appropriate decisions.” City of Philadelphia v. Klutznick, 503 F.Supp. 663, 676 (E.D.Pa.1980).

We cannot substitute our judgment for that of the agency charged with administering § 195. Marsh, 490 U.S. at 376 , 109 S.Ct. 1851 . Under our narrow standard of review, it is sufficient that the Bureau’s panel of experts decided, based on their consideration of the relevant factors, that the data carried too high a risk of a fundamental flaw and could not be certain within the time frame allotted that the adjusted data would improve the accuracy of the census. 11 Given the substantial divergence of the A.C.E. data from the demographic analysis, as well as concerns with synthetic and balancing error, we cannot say that the Secretary’s adoption of *877 ESCAP’s recommendation not to use the adjusted data was “arbitrary or capricious.” Therefore, we uphold his adjustment decision.

B. Revocation of “Daley Rule”

Appellants raise an additional challenge to Secretary Evans’ adjustment decision. Appellants argue that he did not have the authority to make the decision because it had been delegated to the Director of the Bureau by the Daley Rule. Further, they assert that Secretary Evans’ attempt to revoke this rule was ineffective because it failed to comply with the APA’s notice-and-comment procedures.

Assuming arguendo that the Daley Rule was substantive and that the revocation was accomplished improperly, the revocation was harmless error because the Director of the Bureau also adopted ES-CAP’s recommendation not to use the statistically adjusted data for purposes of redistricting. See Riverbend Farms, Inc. v. Madigan, 958 F.2d 1479, 1487 (9th Cir.1992) (holding that APA requires us to take “due account” of “harmless error”). In other words, we know that the result would have been exactly the same even if the delegation had remained in place; all decision-makers, including the Director of the Bureau, rejected the use of the adjusted data. Thus, we decline to overturn the Secretary’s adjustment decision on the basis that it was violative of the APA’s no-tiee-and-eomment procedures.

IV. Conclusion

Section 195 confers broad discretion on the Secretary to decide if and when sampling is feasible in the taking of the decennial census. We find that Secretary Evans’ interpretation of the statute, as permitting him to consider accuracy as a component of feasibility, was a permissible construction of the statute. Further, his decision not to statistically adjust Census 2000 was neither arbitrary nor capricious. Finally, no harm flowed from any violation of the APA’s notice-and-comment procedures. Therefore, we uphold Secretary Evans’ adjustment decision. Because we affirm the District Court’s dismissal of the action, we need not address the question whether injunctive relief would have been an appropriate remedy in this case.

AFFIRMED.

. The studies were conducted pursuant to the Decennial Census Improvement Act of 1991, Pub.L. No. 102-135, 105 Stat. 635. The Act required NAS to study the "means by which the Government could achieve the most accurate population count possible," considering, inter alia, "the appropriateness of using sampling methods, in combination with basic data-collection techniques or otherwise, in the acquisition or refinement of population data.” 105 Stat. at 635 § 2(a)(1) (codified at § 141 note (1991)).

. Section 141 requires the Secretary to provide a tabulation of slate populations within one year of the decennial census date. 13 U.S.C. § 141 (c).

. The Daley Rule was originally proposed un *867 der former Secretary Daley in June 2000. However, the final rule was promulgated under his successor, Secretary Norman Mineta. 65 Fed.Reg. 59,716.

. As discussed further below, “synthetic error” occurs when the statistics rely on a faulty assumption that a segment of the population can be extrapolated to the whole of the population. See 66 Fed.Reg. 14,015. "Balancing error” occurs when different surveys are handled differently with disparate time and effort spent in their implementation. Id. at 14,016 .

. In Wisconsin , the Supreme Court explicitly declined to address the question whether *869 § 195 "constrained the Secretaiy’s discretion to statistically adjust the decennial census." 517 U.S. at 20 n. 11, 116 S.Ct. 1091 . In Department of Commerce, the Supreme Court addressed the issue of statistical adjustment only for apportionment purposes. 525 U.S. at 343 , 119 S.Ct. 765 .

. We reject Appellees’ argument that the adjustment decision is unreviewable under the APA because it falls within its exception for actions "committed to agency discretion by law.” 5 U.S.C. § 701 (a)(2). This exception is applicable only "in those rare circumstances where the relevant statute is drawn so that a court would have no meaningful standard against which to judge the agency's exercise of discretion.” Lincoln v. Vigil, 508 U.S. 182, 191 , 113 S.Ct. 2024 , 124 L.Ed.2d 101 (1993) (internal quotation marks and citations omitted). Here, the Secretary's discretion to "consider” whether sampling is feasible does not defy meaningful judicial review. Unlike other cases in which we have found the statutory grant of discretion so broad that no standard of review could be discerned, § 195 contains mandatory language that the Secretary "shall” use sampling "if he considers it feasible.” Cf. Rank v. Nimmo, 677 F.2d 692, 699-700 (9th Cir.1982) (holding that statute, which provided the administrator "may” take action "at [his] option,” could not be reviewed because it vested the "widest discretion possible” in the administrator).

Moreover, although the phrase “if he considers it feasible” confers broad discretion, it has its limits. For example, if the Secretary had considered statistical adjustment but decided against it due to his political disinclina-tions, his decision would violate § 195. See Franklin v. Massachusetts, 505 U.S. 788, 817 , 112 S.Ct. 2767 , 120 L.Ed.2d 636 (1992) (Stevens, J., concurring). Thus, this is not a situation in which there is "no law to apply.” Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 410 , 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971).

. Appellants overly rely on dictum in Department of Commerce in proffering this interpretation. Therein, the Supreme Court stated that the 1976 amendments to the Census Act "changed a provision that permitted the use of sampling ... into one that required that sampling be used for such purposes if ‘feasible.’ ” 525 U.S. at 341 , 119 S.Ct. 765 . However, we cannot seriously weigh this .statement in our statutory analysis because to do so would render the phrase "if he considers it” superfluous. Further, the Supreme Court failed to explain or even acknowledge its disregard for this phrase. Under these circumstances, we doubt the Court would wish us to disregard this language for purposes of statutory interpretation.

. Although we find Allen instructive as to this point, we do not think that Allen or its progeny control. Those cases involved a very different issue — whether a prisoner had a constitutional interest in parole release for purposes of analyzing his due process claim. See Allen, 482 U.S. at 381 , 107 S.Ct. 2415 . Appellants do not argue here that § 195 created a constitutional interest in the release of the adjusted data.

. Because we reach this conclusion from the statutory text alone, we need not consider *873 Appellees' argument that the legislative history of § 195 is silent as to the issue at hand. Nevertheless, we are inclined to agree that § 195 was enacted to authorize the use of sampling for collecting supplemental population data, not the replacement of initial headcount data. See Utah, 122 S.Ct. at 2201-02 ; Dep’t of Commerce, 525 U.S. at 339 , 119 S.Ct. 765 ("As amended, the section now requires the Secretary to use statistical sampling in assembling the myriad demographic data that are collected in connection with the decennial census.”) (emphasis added).

. For example, the ESCAP Report stated that "absent the concerns noted above, adjustment would result in more accurate data at the state, congressional district, and county levels.” 66 Fed.Reg. 14,006 (emphasis added).

. In fact, these concerns were borne out by further research. On October 17, 2001, ES-CAP released a second report recommending against the use of A.C.E. for all other purposes based on its finding that the adjusted data “overstated the net undercount by at least 3 million persons.” See Report of the Executive Steering Committee for Accuracy and Coverage Evaluation Policy on Adjustment for Non Redistricting Uses, 66 Fed.Reg. 56,006 (Nov. 5, 2001). Attributing the problem to balancing error, the Report explained that “the A.C.E. failed to measure a significant number of census erroneous enumerations, many of which were duplicates. This level of error in the A.C.E. measurements of net coverage is such that the A.C.E. results cannot be used in their current form.” Id.