OPINION

SUTTON, Circuit Judge.

"My effort is in the direction of simplicity," once wrote the namesake of the Henry Ford Hospital. Henry Ford, My Life and Work 13 (Garden City Publ'g Co.1922). Mr. Ford apparently had nothing to do with the creation of the Medicare program.

At issue is whether the federal government must reimburse teaching hospitals for the time their residents spent conducting pure research in the 1990s. The answer turns on the meaning of a provision of the Patient Protection and Affordable Care Act of 2010 and the validity of a regulation promulgated under it. In the Act, Congress required the Secretary of Health and Human Services to reimburse teaching hospitals for "all the time spent by an intern or resident ... in non-patient care activities, ... as such time and activities are defined by the Secretary." The Secretary promulgated a regulation excluding from hospitals' Medicare reimbursements the time residents spent conducting pure research. Convinced that the Secretary reasonably exercised the authority delegated to her under the Act, we uphold the regulation.

I.

Under the Medicare program, teaching hospitals receive additional payments, above and beyond the reimbursement rate for treating Medicare patients, to cover the "direct" and "indirect costs of medical education." 42 U.S.C. § 1395ww(d)(5)(B), (h). Direct costs include education-related expenses, such as residents' salaries. St. Mary's Hosp. v. Leavitt, 416 F.3d 906, 909 (8th Cir.2005). Indirect costs, the ones at stake here, include costs incurred by teaching hospitals due to "the general inefficiencies" and "`extra demands placed on other staff'" that result from educating residents. Id. (quoting S.Rep. No. 98-23, at 37 (1983), reprinted in 1983 U.S.C.C.A.N. 143, 192). Unlike direct costs, indirect costs do not lend themselves to simple calculations, as it is not easy to measure "general inefficiencies" and "extra demands." To that end, Congress in 1983 created an elaborate formula for these expenses, which today reads:

42 U.S.C. § 1395ww(d)(5)(B), (d)(5)(B)(ii). The basic idea behind the formula is that the more "full-time equivalent" (FTE) residents a hospital teaches, the larger its Medicare subsidies will be.

This formula, however, does not specify which resident activities count toward a hospital's indirect-cost FTE calculation. Congress left that task to the Secretary, who filled the gap with a regulation providing that the agency would reimburse hospitals for the time residents spend in the "portion of the hospital subject to the prospective payment system or in the outpatient department of the hospital." 42 C.F.R. § 412.105(f)(5), (g)(1)(ii) (1991).

Henry Ford Hospital is a teaching hospital located in Detroit. The hospital applied for Medicare reimbursements for Fiscal Years 1991-96 and 1998-99, but the agency excluded from the hospital's FTE count all of the time residents spent conducting "pure research" — research unrelated to the treatment of a patient. The hospital successfully challenged that determination in federal district court. 680 F.Supp.2d 799 (E.D.Mich.2009).

While the Secretary's appeal of the district court's decision was pending, Congress re-wrote the rules for calculating hospitals' FTE counts in the Patient Protection and Affordable Care Act ("the Act"), Pub.L. No. 111-148, § 5505, 124 Stat. 119, 660-61 (2010). Of import here, it divided residents' activities into "patient care activities" and "non-patient care activities." Patient care activities, it comes as no surprise, "means the care and treatment of particular patients." 42 C.F.R. § 413.75(b) (2010); see id. § 412.105.

The definition of "non-patient care activities" is less straightforward. For the years between 1983 and 2001, the years at issue, the Act says that the Secretary must include in hospitals' indirect FTE counts

PPACA § 5505(b); see id. § 5505(c) (providing that this provision is effective for "cost reporting periods beginning on or after January 1, 1983" through those beginning on September 30, 2001). Exercising her authority to define "such time and activities," the Secretary promulgated a regulation specifying that eligible non-patient care activities do not include the time residents spend conducting pure research. Payments to Hospitals for Graduate Medical Education Costs, 75 Fed.Reg. 71,800, 72,261 (Nov. 24, 2010) (to be codified at 42 C.F.R. § 412.105(f)(1)(iii)(C)).

This new regulation, the Secretary maintains, resolves this dispute for the years in question. The hospital objects, claiming that the Secretary exceeded her authority in promulgating the rule.

II.

A.

The validity of administrative regulations generally raises two questions: (1) has Congress "directly spoken" to the question at hand, and (2) if not, is the agency's answer based on a "permissible construction of the statute"? Chevron, U.S.A., Inc. v. Natural Res. Def. Council, Inc., 467 U.S. 837, 842-43, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). This dispute largely hinges on step one: Does the statute directly answer whether pure research amounts to an eligible "non-patient care activit[y]" and thus must be counted toward a hospital's FTE calculation?

It does not — for several reasons. First, the key words of the provision are not self-defining. Here is the relevant statutory language:

PPACA § 5505(b). In making the FTE calculation, the Secretary thus must consider "all" time spent by residents in "non-patient care activities" that "occur[] in the hospital." The word "all" tells us nothing, until we know what "non-patient care activities" means. The "occurs in the hospital" requirement places a limit on eligibility, to be sure, but it does not tell us which activities in the hospital count toward the FTE count. And the phrase "non-patient care activities" by itself does not tell us which resident activities in the hospitals count. There are many possibilities: activities under the supervision of a doctor? seminars? classes? research? bathroom breaks? meal times? All of these options cast doubt on the idea that the phrase speaks directly to the problem at hand. Looked at in isolation, then, the requirement that "all" of the time residents spend on "non-patient care activities" "in the hospital" does not tell us one way or another whether pure research counts.

Second, context does not remove this ambiguity. The statute offers two examples of eligible "non-patient care activities": "didactic conferences and seminars." The legislative impulse to illustrate what this phrase means confirms that it is not self-defining. Other than proving that these types of activities qualify, the examples say nothing about whether pure research qualifies, and good arguments can be made either way about whether pure research ought to be treated like attendance at seminars and conferences.

Third, the statute expressly delegates to the Secretary the authority to "define[]" eligible "non-patient care activities." PPACA § 5505(b). Had Congress wanted anything that might fall into the category of non-patient care activities to count, no matter how distant from medical care the activity might be, why would it empower the Secretary to define the phrase? The request to define implies a need to define. Otherwise, Congress sent the agency on a fool's errand. An agency is at the apex of its administrative authority when Congress not only gives the agency general authority to implement a statute but also expressly asks the agency to define a specific phrase.

Two D.C. Circuit cases confirm the point. In Wint v. Yeutter, 902 F.2d 76, 77 (D.C.Cir.1990), the court entertained a challenge to an Agriculture Department regulation promulgated under the Immigration Reform and Control Act. The Immigration Act paved a path to lawful permanent residency for undocumented aliens who performed agricultural labor "related to planting, cultural practices, cultivating, growing and harvesting of fruits and vegetables of every kind and other perishable commodities, as defined in regulations by the Secretary of Agriculture." Id. at 78 (quoting 8 U.S.C. § 1160(h)) (emphases added). The agency adopted a "horticultural definition of vegetables" and a "botanical definition of fruits," neither of which covered sugar cane. Id. at 80-81. Sugar cane workers objected because they would not be able to obtain permanent residency. Writing for the majority, then-Judge Ruth Bader Ginsburg reasoned that the "as defined" clause "explicit[ly] delegat[ed] to the USDA ... the authority to define, not merely to apply, the terms in question." Id. at 81. Such language, the court concluded, "signaled judicial deference," leading the court to uphold the regulation. Id.; see also Women Involved in Farm Econ. v. U.S. Dep't of Agric., 876 F.2d 994, 1000 (D.C.Cir.1989) ("Congress explicitly authorized the Secretary to define the term `person,'" which "necessarily suggests that Congress did not intend the word to be applied in its plain meaning sense.").

The hospital reads the statute differently. It construes § 5505(b) to say that "`all... non-patient care activities' must be included in the [indirect cost] FTE count." Henry Ford Hosp. Mar. 24, 2011 28(j) Letter at 1. But this argument fails to account for two key phrases: the examples of covered activities and the delegation of authority to the Secretary to define which other activities count. Why provide examples if the phrase is crystal clear? And why ask the Secretary to define a phrase that needs no definition? We cannot erase the words, least of all Congress's "express delegation of authority to the agency to elucidate a specific provision of the statute by regulation." Chevron, 467 U.S. at 843-44, 104 S.Ct. 2778.

Even if § 5505(b) does not require reimbursement for all non-patient care activities, the hospital adds, it must require reimbursement for pure research in view of an adjacent section, § 5505(a). Section 5505(a) concerns the calculation of a hospital's direct cost reimbursements for the time residents spend in certain "nonprovider setting[s]," such as nursing homes or clinics. PPACA § 5505(a)(1)(B); see id. § 5505(c)(2) (effective date). It says that, for this calculation, the Secretary must count all the time a resident spends in "non-patient care activities, such as didactic conferences and seminars, but not including [pure] research." Id. § 5505(a)(1)(B) (emphasis added). The hospital draws two conclusions from this provision: that Congress understood pure research to be a subset of "non-patient care activities," and that "non-patient care activities" in § 5505(b) must include research because that subsection contains no similar exclusion for pure research.

A legislature's decision to include a phrase in one section but not in a second section, it is true, may imply that the second section should be read not to include the phrase. Russello v. United States, 464 U.S. 16, 23, 104 S.Ct. 296, 78 L.Ed.2d 17 (1983). But this canon creates a potential inference, not a necessary one. "The Russello presumption ... grows weaker with each difference in the formulation of the provisions under inspection." City of Columbus v. Ours Garage & Wrecker Serv., Inc., 536 U.S. 424, 435-36, 122 S.Ct. 2226, 153 L.Ed.2d 430 (2002). That reality by itself defeats the suggestion that this canon precludes the Secretary from defining "non-patient care activities."

No less importantly, it is perfectly sensible for the legislature to specify that pure research is not reimburseable as a category of "direct" costs but to allow the Secretary to make the call for "indirect" costs. Specifying that something should be excluded in one context does not mean it must be included in a different context. See Ours Garage, 536 U.S. at 435-36, 122 S.Ct. 2226. Section 5505(a) covers direct costs incurred outside the hospital, while § 5505(b) covers indirect costs incurred in the hospital. It requires little imagination to appreciate why Congress might want to deny reimbursements for out-of-hospital direct costs yet want to give the Secretary latitude to treat in-hospital indirect costs differently. Perhaps Congress determined that residents learn little when they conduct pure research outside the hospital but trusted the Secretary to ascertain the educational merits of hospital-based research. Perhaps the Medicare budget had no room for the direct costs of research outside the hospital but could absorb the indirect costs of hospital-based research. Either way, § 5505(a) does not speak directly to what Congress meant in § 5505(b).

Because pure research cannot be a patient care activity, the hospital points out, it must be a non-patient care activity. But the two categories do not define the universe of a resident's activities. Just because Medicare does not classify a resident's activity as a patient care activity does not make it a non-patient care activity. Section 5505(a) proves the point. A resident conducting pure research in a nursing home engages in neither a patient care activity nor a non-patient care activity for the purpose of direct cost reimbursements.

University of Chicago Medical Center v. Sebelius, 618 F.3d 739 (7th Cir.2010), we realize, is in some tension with this decision. But we have a benefit that the Seventh Circuit did not — the Secretary's new regulation, which converted a run-of-the-mine statutory interpretation case into a Chevron case. No doubt the Seventh Circuit thought the better reading of § 5505(b) favored the hospital. But the decision does not indicate what the court would have done had the validity of the regulation been presented to it.

B.

That brings us to step two of the Chevron inquiry. When Congress "express[ly] delegat[es] ... authority to the agency to elucidate a specific provision of the statute by regulation," we give the agency's regulations "controlling weight unless they are arbitrary, capricious, or manifestly contrary to the statute." Chevron, 467 U.S. at 843-44, 104 S.Ct. 2778. The regulation satisfies this modest standard.

In issuing the new regulation, the Secretary noted that "didactic conferences and seminars may occur during periods when an intern or resident is otherwise assigned to a rotation primarily requiring the provision of patient care." 75 Fed. Reg. at 72,146. Those conferences and seminars "may involve presentations or discussions related to the treatment of current patients." Id. Didactic conferences and seminars thus differ from pure research rotations in an important way: Conferences and seminars touch on patient care, increasing the benefit to current patients. The Secretary's conclusion that pure research differs from didactic conferences and seminars squares with the statute, and she therefore acted within her authority by excluding pure research from "non-patient care activities, such as didactic conferences and seminars."

How odd, the reader might say: The Secretary refuses to reimburse a non — patient care activity — pure research — because it does not sufficiently affect patient care? Any oddity, however, comes not from the Secretary but from the statute. The idea is to permit compensation for patient care activities and some, but not all, non-patient care activities. The two categories do not cover all resident activities. The Secretary may reasonably believe that Medicare primarily focuses on patient care, not medical research. She thus may be willing to reimburse only those non-patient care activities that seem to benefit current patients, a conclusion that necessarily is not arbitrary given that Congress took the same view in dealing with non-patient care activities outside of a hospital in § 5505(a). Because her exclusion of pure research comports with the relevant statutes, we give the regulation controlling weight.

C.

The hospital complains that the regulation cannot apply retroactively to the fiscal years in question. We disagree.

A rule has retroactive effect if, as here, it "attaches new legal consequences to events completed before its enactment." Landgraf v. USI Film Prods., 511 U.S. 244, 270, 114 S.Ct. 1483, 128 L.Ed.2d 229 (1994). The starting assumption is that Congress intends statutes to operate prospectively, and when Congress delegates rulemaking authority to an agency, we presume that delegation allows the agency to regulate prospectively. Bowen v. Georgetown Univ. Hosp., 488 U.S. 204, 208, 109 S.Ct. 468, 102 L.Ed.2d 493 (1988). Only express congressional authorization for the agency to regulate retroactively will defeat this presumption. Id.

Congress gave just such authorization here, saying that the Secretary may promulgate regulations with retroactive effect. The relevant part of § 5505(b) authorizes the Secretary to "define[]" eligible "non-patient care activities" for 1983 to 2001. Congress confirmed the point in two other places. It instructed the Secretary to "implement the amendments made by this section in a manner so as to apply to cost reporting periods beginning on or after January 1, 1983." PPACA § 5505(c)(1). And it said that the new statute applies only to open cost reports and to those cost reports "as to which there is ... a jurisdictionally proper appeal pending as of the date of the enactment of this Act." Id. § 5506(c). No one doubts that the cost reports at issue fit that definition.

III.

For these reasons, we reverse the judgment of the district court and remand for further proceedings.