Plaintiffs appeal the order denying attorneys’ fees following lengthy litigation which successfully challenged federal and state regulations that changed the conditions of eligibility for benefits under state Aid to Families with Dependent Children (AFDC) for some three-generation families living together.
The district court granted summary judgment to the plaintiff class.1 We affirmed, adopting the reasoning of the Minnesota district court in Morrison v. Heckler, 602 F.Supp. 1485 (D.Minn.1985), aff'd, 787 F.2d 1285 (8th Cir.1986). Kali v. Bowen, 800 F.2d 971 (9th Cir.1986) (per curiam).
After winning their appeal, the plaintiffs moved the district court for an award of attorneys’ fees and costs against the federal government under the Equal Access to Justice Act (EAJA), 28 U.S.C.A. § 2412(d) (West Supp.1988). The district court denied the motion, observing that only one district court had addressed the question at the time the action was filed and that the Ninth Circuit had never addressed the regulations until this case:
The Court finds that Defendant Bowen's position was reasonable, and one of first impression in this circuit. If the question of law is unresolved and of unclear resolution, then the government's litigation of the issue is reasonable and *331substantially justified. Minor v. United States, 797 F.2d 738, 739 (9th Cir.1986), and Foster v. Tourtellotte, 704 F.2d 1109 (9th Cir.1983).
The same rationale applies with respect to the government’s decision to appeal this case. Because the Ninth Circuit had not yet ruled on this issue, and other circuits had not yet or were in the midst of publishing an opinion on these issues, and because the government’s interpretation of the statute was a reasonable one, attorney’s fees are not justified.
We review the district court’s decision to deny attorneys’ fees under the EAJA for an abuse of discretion. See Pierce v. Underwood, — U.S. -, -, 108 S.Ct. 2541, 2546-49, 101 L.Ed.2d 490 (1988); Minor v. United States, 797 F.2d 738, 739 (9th Cir.1986) (per curiam). The district court abuses its discretion when its “decision is based on an erroneous conclusion of law or when the record contains no evidence on which [it] rationally could have based that decision.” Petition of Hill, 775 F.2d 1037, 1040 (9th Cir.1985). Interpretation of the EAJA presents a question of law reviewable de novo. See id.
The plaintiffs suggest that attorneys’ fees might be available under 28 U.S. C.A. § 2412(b) (West Supp.1988), which authorizes awards of attorneys’ fees against the United States “to the same extent that any other party would be liable under the common law or the terms of any statute which specifically provides for such an award.” Attorneys’ fees for an action brought under 42 U.S.C. § 1983 (1982) are authorized by 42 U.S.C. § 1988 (1982). However, “[fjederal officials who violate federal rights protected by § 1983 generally do not act under ‘color of state law,’ and therefore cannot be held liable for attorneys’ fees under 28 U.S.C. § 2412(b) and 42 U.S.C. § 1988.” Olson v. Norman, 830 F.2d 811, 821 (8th Cir.1987). Because the plaintiffs make no allegation that federal and state officials conspired so that the actions taken by the federal officials could be deemed to have been under “color of state law,” the plaintiffs have not demonstrated a statutory basis for an award of attorneys’ fees under § 2412(b) (West Supp.1988). See id.
The plaintiffs also seek attorneys’ fees under 28 U.S.C.A. § 2412(d)(1)(A) (West Supp.1988). As originally adopted, 28 U.S. C. § 2412(d)(1)(A) (1982) provided that a party prevailing in a suit against the United States or one of its agencies is entitled to attorneys’ fees, costs, and other expenses “unless the court finds that the position of the United States was substantially justified or that special circumstances make an award unjust.” The Ninth Circuit “has long applied a test of reasonableness in determining whether the government’s position was substantially justified.” Oregon Envtl. Council v. Kunzman, 817 F.2d 484, 498 (9th Cir.1987); see, e.g., League of Women Voters of California v. FCC, 798 F.2d 1255, 1257 (9th Cir.1986) (stating that the “substantially justified” requirement of § 2412 requires only a showing that the government’s “ ‘case had a reasonable basis both in law and fact’ ”) (quoting H.R. Rep. No. 1418, 96th Cong., 2d Sess. 10, reprinted in 1980 U.S.Code Cong. & Admin.News 4953, 4984, 4989); Foster v. Tourtellotte, 704 F.2d 1109, 1112 (9th Cir.1983) (per curiam).
The original EAJA contained a “sunset” provision and expired on Sept. 30, 1984. In 1985, Congress reenacted the EAJA without changing the language of 28 U.S.C. § 2412(d)(1)(A). See 28 U.S.C.A. § 2412(d)(1)(A) (West Supp.1988). The House report accompanying the 1985 reenactment suggests that something more than mere reasonableness is required to satisfy the “substantially justified” standard:
Another problem which has developed in the implementation of the Act has been the fact that courts have been divided on the meaning of “substantial justification.” Several courts have held correctly that “substantial justification” means more than merely reasonable. Because in 1980 Congress rejected a standard of “reasonably justified” in favor of “substantially justified,” the test must be more than mere reasonableness.
H.R.Rep. No. 120, 99th Cong., 1st Sess. 9, reprinted in 1985 U.S.Code Cong. & Admin.News 132, 138 (footnote omitted).
*332At least five Ninth Circuit cases have acknowledged that the 1985 legislative history may require a showing of something more than reasonableness to satisfy the “substantially justified” standard, but the cases have found it unnecessary on the facts presented to resolve the issue. See Bresgal v. Brock, 843 F.2d 1163, 1172 (9th Cir.1988); Thomas v. Peterson, 841 F.2d 332, 335 & n. 1 (9th Cir.1988); Andrew v. Bowen, 837 F.2d 875, 878-79 & n. 2 (9th Cir.1988); Edwards v. McMahon, 834 F.2d 796, 802 (9th Cir.1987); Oregon Envtl. Council, 817 F.2d at 498. In Pierce, — U.S. at -, 108 S.Ct. at 2549-51, the Supreme Court rejected the argument that the 1985 legislative history changed the applicable test, finding that the “substantially justified” standard is equivalent to “the ‘reasonable basis both in law and fact’ formulation adopted by the Ninth Circuit and the vast majority of other Courts of Appeals that have addressed the issue.”
The government has the burden of demonstrating that its position was substantially justified. See Hill, 775 F.2d at 1042. The government’s failure to prevail does not raise a presumption that its position was not substantially justified. See id. “In analyzing the reasonableness of the government’s position under the ‘totality of the circumstances’ test, we must look both to the position asserted by the government in the trial court as well as the nature of the underlying government action at issue.” League of Women Voters of California, 798 F.2d at 1258.
The inquiry into the nature of the underlying government action will by definition concern only the merits of that action. The inquiry into the government’s position at trial will encompass the first inquiry to the extent that the government chooses to defend the merits of the challenged action. However, the second inquiry must also focus upon extraneous circumstances bearing upon the reasonableness of the government’s decision to take a case to trial. Perhaps the most important of these extraneous circumstances will be the existence of precedents construing similar statutes or similar facts. See Pierce, — U.S. at -, 108 S.Ct. at 2551-53 (observing that “a string of losses” or “a string of successes” may be “indicative” on the issue of substantial justification).
The inquiry into the existence of substantial justification therefore must focus on two questions: first, whether the government was substantially justified in taking its original action; and, second, whether the government was substantially justified in defending the validity of the action in court. We will address the second question first.
The plaintiffs argue that the Secretary lacked substantial justification because at the time that the government decided to defend the suit the Minnesota district court had already decided Morrison, 602 F.Supp. 1485, in a manner adverse to the government, and the Secretary’s decision to litigate therefore was intended to create an intercircuit conflict on the issue.2 The Supreme Court has recently rejected this argument. “Obviously, the fact that one other court agreed or disagreed with the Government does not establish whether its position was substantially justified.” Pierce, — U.S. at -, 108 S.Ct. at 2552.3
We now turn to the determination whether the government’s decision to promulgate its original regulation was substantially justified. As a threshold matter, we may reject two claims made by the plaintiffs *333that the nature of the district court’s disposition necessarily indicates that the government lacked substantial justification.
First, the plaintiffs argue that the deference given to an agency’s interpretation of its organic statute under 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), demonstrates that a court’s finding that an agency has acted contrary to law implicitly indicates that its action was unreasonable. See Trahan v. Regan, 824 F.2d 96, 101-02 (D.C.Cir.1987) (stating that “only in unusual circumstances can a court find that agency action based on the agency’s interpretation of its own regulations or governing statute is ‘substantially justified’ if the court has previously overruled that interpretation and found the agency action to be contrary to law”), reh’g en banc granted, 832 F.2d 158 (D.C.Cir.1987). We must reject this argument. Morrison expressly stated that little deference was due to the Secretary’s interpretation. See Morrison, 602 F.Supp. at 1491. In any event, this circuit has recognized that “arbitrary and capricious conduct is not per se unreasonable.” Andrew, 837 F.2d at 878.
Second, the plaintiffs claim that the district court’s summary judgment demonstrates that the Secretary’s position was not substantially justified. This argument is meritless, given that under Fed.R.Civ.P. 56 the district court’s finding that summary judgment was appropriate hinged upon the absence of factual disputes, not the clarity of the law. See Pierce, — U.S. at -, 108 S.Ct. at 2552 (observing that where “the dispute centers upon questions of law rather than fact, summary disposition proves only that the district judge was efficient”).
We now reach the primary issue presented upon appeal: Was the government substantially justified in its initial interpretation of the relevant statutes?
The plaintiffs challenged the government regulations on the ground that they exceeded the scope and intent of the enabling statute, 42 U.S.C. § 602(a)(39) (Supp.II 1984).4 This statute, which applies to certain “three-generation” families living together, “requires that in evaluating the need of the grandchild in such families, the state must take into account the income of the grandparent when the grandchild’s parent is “ ‘under the age selected by the State pursuant to section 606(a)(2) of this title.’ ” Kali, 800 F.2d at 972 (quoting 42 U.S.C. § 602(a)(39) (Supp.II 1984)). Section 606(a)(2) gives states the option of defining a “dependent child” as one “who is (A) under the age of eighteen, or (B) ... under the age of nineteen and a full-time student in a secondary school ... if, before he attains age nineteen, he may reasonably be expected to complete the program of such secondary school.” 42 U.S.C. § 606(a)(2) (1982). Hawaii has selected the second of these options. See Hawaii Admin.R. § 17-640-4.
Both the federal government and the state of Hawaii have enacted regulations that construe § 602(a)(39) as applying to parents classified as minors under the regulations implementing § 606(a)(2) “without regard to school attendance.” See Hawaii Admin.R. 17-621-3(e); 45 C.F.R. § 233.20(a)(3)(xviii) (1987). As a result, “the grandparent’s income [was] deemed available to the dependent grandchild in all three-generation families living together in which the intermediate parent is eighteen, rather than only in those families where the eighteen-year-old parent satisfies the school attendance requirement.” Kali, 800 F.2d at 973.
The plaintiffs cite our prior holding that the Secretary’s regulation was against the statutory purpose of the Grandparent Deeming Rule because it ignored the school attendance requirement for children between ages 18 and 19, thereby requiring *334grandparent deeming to a group of 18-year-old non-students who could not under any circumstances receive AFDC funding as children. See Morrison, 602 F.Supp. at 1489. The plaintiffs also observe that every court to have considered this issue has rejected the Secretary’s position.5
However, the government’s failure to prevail does not raise a presumption that its position was not substantially justified. See Oregon Envtl. Council, 817 F.2d at 498. The determination whether the government’s position was substantially justified “requires courts to examine the government’s conduct, not the Secretary’s beliefs with respect to his policies.” Andrew, 837 F.2d at 878-79. We may reverse the district court’s determination that the Secretary’s position was substantially justified only if we find that the court abused its discretion. See Pierce, — U.S. at - -, 108 S.Ct. at 2546-49; Edwards, 834 F.2d at 801.
Applying this highly deferential standard, we uphold the district court’s finding that the Secretary’s interpretation of § 602(a)(39) was substantially justified.
The applicable statutory language could be read to support the Secretary’s regulations. Section 602(a)(39) refers only to “the age selected by the State pursuant to section 606(a)(2),” 42 U.S.C. § 602(a)(39) (Supp. II 1984), while § 606(a)(2) gives states the option of selecting the age of 18 or 19 as the cut-off for dependency status, 42 U.S.C. § 606(a)(2) (1982). This language on its face supports the Secretary’s interpretation even though § 606(a)(2) also includes the school attendance requirement as part of the age 19 cut-off. The Secretary’s interpretation is further bolstered by the fact that § 602(a)(38), which was adopted at the same time, expressly refers to all of the conditions described in § 606(a), thereby incorporating both the age and school attendance requirements of that statute. 42 U.S. C. § 602(a)(38) (Supp.II 1984). Because § 602(a)(39) lacks this general reference to § 606(a), it was not improper to infer that school attendance was not intended to be relevant to the grandparent deeming rule. We therefore cannot say that the Secretary lacked substantial justification in interpreting the statutory language as she did.
Furthermore, we cannot say that the Secretary’s interpretation, although constituting a facially correct reading of the statutes, was so at odds with congressional intent as to render it lacking in substantial justification. Any interpretation denying benefits to the children of 18-year-old students while granting benefits to the children of 18-year-old nonstudents would have given 18-year-old parents a great incentive to drop out of school, thereby contravening the congressional policy of encouraging continued school attendance through the extension of AFDC funding to 18-year-old students. See S.Rep. No. 139, 97th Cong., 1st Sess. 513, reprinted in 1981 U.S.Code Cong. & Admin.News 396, 779-80. It also was reasonable to conclude that Congress did not intend to deny benefits to the children of students while giving benefits to the children of nonstudents; because nonstudents may work full time while students may not, as a general matter nonstudent parents could be expected to earn more and therefore to have less need for AFDC funds than would student parents.6
The district court did not abuse its discretion in finding that the government’s position was substantially justified. The Secretary’s interpretation of the complex statutory scheme was justifiable on the facial language of the relevant statutes and did not clearly contravene congressional intent. “The parties raised difficult issues of statutory interpretation, with substantial sums at stake.” Edwards, 834 F.2d at 803. The Secretary “argued forcefully and well” for *335her position. Trustees for Alaska v. Hodel, 806 F.2d 1378, 1384 (9th Cir.1986). Although the Secretary lost the case on the merits, we cannot say that the district court abused its discretion in finding that the government’s position was substantially justified.
The decision of the district court denying attorneys’ fees is AFFIRMED.
. The plaintiff class consists of 18-year-old parents in Hawaii who either are not full-time students or are not reasonably expected to complete their school program before reaching 19, who live with their parents, and whose AFDC benefits have been or will be reduced, terminated or denied under regulations enacted under the Deficit Reduction Act. Kali v. Bowen, 800 F.2d 971, 973 (9th Cir.1986) (per curiam).
. The plaintiffs also claim that the district court improperly held, in effect, that substantial justification exists per se whenever this circuit has not ruled on an issue. This claim is without merit, given that the court explicitly rested its holding on its finding that the government’s position was reasonable. The court’s observation that the Ninth Circuit had not yet addressed the issue was an appropriate component of the inquiry into substantial justification.
. Keasler v. United States, 766 F.2d 1227, 1232-33 (8th Cir.1985), relied upon by the plaintiffs, did not create a per se rule that the government is never substantially justified in arguing that a court should not follow an adverse precedent issued by a court in another jurisdiction. That opinion held instead that the IRS could not have reasonably believed that the Eighth Circuit would ignore the strong policy of deference to out-of-circuit tax precedents and refuse to follow a long-standing Tenth Circuit case.
. Under 42 U.S.C. § 602(a)(39) (Supp.II 1984): (a) A State plan for aid and services to needy families with children must—
(39) provide that in making the determination ... [of need] with respect to a dependent child whose parent or legal guardian is under the age selected by the State pursuant to section 606(a)(2) of this title, the State agency shall ... include any income of such minor’s own parents or legal guardians who are living in the same home as such minor and dependent child, to the same extent that income of a stepparent is included under paragraph (31).
. These opinions have, for the most part, been unpublished. See Driesens v. Bowen, 670 F.Supp. 1360 (W.D.Mich.1986); Jimenez v. Coken, No. 85-5282 (E.D.Pa. July 17, 1986) [available on WESTLAW, 1986 WL 8141]; Grimesy v. McMahon, No. C 86-0947 SW (N.D.Cal. June 20, 1986); McCarthy v. Heintz, No. H-85-597 (MJB) (D.Conn. Apr. 23, 1986); Topps v. Bowen, No. 85-NC-0187 W (D.Utah Jan. 13, 1986) [available on WESTLAW, 1986 WL 15777]. These cases have split on the determination whether the government's position was substantially justified.
. The Secretary also claims that subsequent legislative history reinforces its position that Congress never intended to make school attendance pertinent to grandparent deeming. The amend*335ment to § 602(a)(39), as originally drafted, adopted the interpretation advocated by the Secretary. See S.Rep. No. 99-313, 99th Cong., 2d Sess. 1075 (1986), 1986 U.S.Code Cong. & Admin.News 4075. The version eventually adopted by Congress eliminated grandparent deeming for all 18-year-old parents, regardless of school attendance. See 42 U.S.C.A. § 602(a)(39) (West Supp.1988). The Secretary argues that the subsequent amendments demonstrate the statute’s ambiguity. We must reject this argument, given that Congress may have deemed the amendments necessary not because of inherent ambiguity in the statute but because of the manner in which the Secretary applied it. Nor can we accept the government’s argument that the initial draft amendment demonstrates the validity of its position, both because Congress declined to accept that draft and because it is unclear whether the draft amendment was intended to clarify existing law or to change it. Although we are unable to find that subsequent legislative history supports the government’s position, neither can we say that such history does anything to demonstrate that the government’s position lacked substantial justification.