OPINION

This case arises out of the 2012 merger of two health care providers in Nampa, Idaho. The Federal Trade Commission (“FTC”) and the State of Idaho sued, alleging that the merger violated § 7 of the Clayton Act, 15 U.S.C. § 18 , and state law; two local hospitals filed a similar complaint. Although the district court believed that the merger was intended to improve patient outcomes and might well do so, the judge nonetheless found that the merger violated § 7 and ordered divestiture.

As the district court recognized, the job before us is not to determine the optimal future shape of the country’s health care system, but instead to determine whether this particular merger violates the Clayton Act. In light of the careful factual findings by the able district judge, we affirm the judgment below.

I. Background

A. The Health Care Market in Nampa, Idaho

Nampa, the second-largest city in Idaho, is some twenty miles west of Boise and has a population of approximately 85,000. Before the merger at issue, St. Luke’s Health Systems, Ltd. (“St. Luke’s”), an Idaho-based, not-for-profit health care system, operated an emergency clinic in the city. Saltzer Medical Group, P.A. (“Saltzer”), the largest independent multi-specialty physician group in Idaho, had thirty-four physicians practicing at its offices in Nam-pa. The only hospital in Nampa was operated by Saint Alphonsus Health System, Inc. (“Saint Alphonsus”), a part of the multistate Trinity Health system. . Saint Alphonsus and Treasure Valley Hospital Limited Partnership (“TVH”) jointly operated an outpatient surgery center.

The largest adult primary care physician (‘TCP”) provider in the Nampa market was Saltzer, which had sixteen PCPs. St. Luke’s had eight PCPs and Saint Alphon-sus nine. Several other PCPs had solo or small practices.

B. The Challenged Acquisition

Saltzer had long had the goal of moving toward integrated patient care and risk-based reimbursement. After unsuccessfully attempting several informal affiliations, including one with St. Luke’s, Saltzer sought a formal partnership with a large health care system.

In 2012, St. Luke’s acquired Saltzer’s assets and entered into a five-year professional service agreement (“PSA”) with the Saltzer physicians (the “merger” or the “acquisition”). Saltzer received a $9 million payment for goodwill. The initial PSA contained hortatory language about the parties’ desire to move away from fee-for-service reimbursement, but included no provisions implementing that goal. An amended PSA, however, contained some quality-based incentives. The merger did not require Saltzer doctors to refer patients to the St. Luke’s Boise hospital, nor did it require that Saltzer physicians use St. Luke’s facilities for ancillary services.

C. Procedural History

In November 2012, the Private Hospitals filed a complaint in the District of Idaho seeking to enjoin the merger under Clayton Act § 7. The complaint alleged anticompetitive effects in the relevant markets for “primary care physician services,” “general acute-care inpatient services,” “general pediatric physician services,” and “outpatient surgery services.” The district court denied a preliminary injunction, noting that: (1) the PSA did not require referrals to St. Luke’s, minimizing any immediate harm to the Private Hospitals; (2) implementation of the PSA was to take place over time; and (3) the PSA provided a process for unwinding the transaction if it were declared illegal.

In March 2013, the FTC and the State of Idaho filed a complaint in the district court seeking to enjoin the merger pursuant to the Federal Trade Commission Act (“FTC Act”), the Clayton Act, and Idaho law. This complaint alleged anticompeti-tive effects only in the adult PCP market. The district court consolidated this case with the one filed by the Private Hospitals, and after a nineteen-day bench trial, found the merger prohibited by the Clayton Act and the Idaho Competition Act because of its anticompetitive effects on the Nampa adult PCP market.

The district court expressly noted the troubled state of the U.S. health care system, found that St. Luke’s and Saltzer genuinely intended to move toward a better health care system, and expressed its belief that the merger would “improve patient outcomes” if left intact. Nonetheless, the court found that the “huge market share” of the post-merger entity “creates a substantial risk of anticompetitive price increases” in the Nampa adult PCP market. Rejecting an argument by St. Luke’s that anticipated post-merger efficiencies excused the potential anticompetitive price effects, the district court ordered divestiture. This appeal followed.

II. Standard of Review

We review the district court’s findings of fact for clear error and its conclusions of law de novo. Husain v. Olympic Airways, 316 F.3d 829, 835 (9th Cir.2002), aff'd, 540 U.S. 644 , 124 S.Ct. 1221 , 157 L.Ed.2d 1146 (2004). The question is whether a finding of fact is “clearly erroneous,” not whether there is a “compelling case” for an alternative finding. California v. Am. Stores Co., 872 F.2d 837, 842 (9th Cir.1989), rev’d on other grounds, 495 U.S. 271 , 110 S.Ct. 1853 , 109 L.Ed.2d 240 (1990).. The district court’s choice of remedy is reviewed for abuse of discretion. Theme Promotions, Inc. v. News Am. Mktg. FSI, 546 F.3d 991, 1000 (9th Cir.2008) (citing United States v. Alisal Water Corp., 431 F.3d 643 , 654 (9th Cir.2005)).

III. The Clayton Act § 7 Analysis

A. Overview of the Clayton Act

The great Yankee catcher Yogi Berra is reputed (likely apocryphally) to have said that it’s “tough to make predictions, especially about the future.” The Perils of Prediction, Economist, June 2, 2007, at 96. Yet that is precisely what this case requires. Because § 7 of the Clayton Act bars mergers whose effect “may be substantially to lessen competition, or to tend to create a monopoly,” 15 U.S.C. § 18 , judicial analysis necessarily focuses on “probabilities, not certainties,” Brown Shoe Co. v. United States, 370 U.S. 294, 323 , 82 S.Ct. 1502 , 8 L.Ed.2d 510 (1962). This “requires not merely an appraisal of the immediate impact of the merger upon competition, but a prediction of its impact upon competitive conditions in the future; this is what is meant when it is said that the amended § 7 was intended to arrest anticompetitive tendencies in their incipiency.” United States v. Phila. Nat’l Bank, 374 U.S. 321, 362 , 83 S.Ct. 1715 , 10 L.Ed.2d 915 (1963) (internal quotation marks omitted).

Section 7 claims are typically assessed under a “burden-shifting framework.” Chi. Bridge & Iron Co. v. FTC, 534 F.3d 410, 423 (5th Cir.2008). The plaintiff must first establish a prima facie case that a merger is anticompetitive. See Olin Corp. v. FTC, 986 F.2d 1295, 1305 (9th Cir.1993) (discussing how plaintiffs establishment of a prima facie case on statistical evidence was the first step in the analysis). The burden then shifts to the defendant to rebut the prima facie case. See id.; Am. Stores, 872 F.2d at 842 (citing United States v. Marine Bancorporation, Inc., 418 U.S. 602, 631 , 94 S.Ct. 2856 , 41 L.Ed.2d 978 (1974)). “[I]f the [defendant] successfully rebuts the prima facie case, the burden of production shifts back to the Government and merges'with the ultimate burden of persuasion, which is incumbent on the Government at all times.” Chi. Bridge & Iron, 534 F.3d at 423 .

B. The Relevant Market

“Determination of the relevant product and geographic markets is a necessary predicate to deciding whether a merger contravenes the Clayton Act.” Marine Bancorporation, 418 U.S. at 618 , 94 S.Ct. 2856 (internal quotation marks omitted). Definition of the relevant market is a factual question “dependent upon the special characteristics of the industry involved and we will not disturb such findings unless clearly erroneous.” Twin City Sportservice, Inc. v. Charles O. Finley & Co., 676 F.2d 1291, 1299 (9th Cir.1982). Although the parties agree that the relevant product market in this case is adult PCPs, St. Luke’s vigorously disputes the district court’s determination that Nampa is the relevant geographic market. We find no clear error in that factual finding.

The relevant geographic market is the “area of effective competition where buyers can turn for alternate sources of supply.” Morgan, Strand, Wheeler & Biggs v. Radiology, Ltd., 924 F.2d 1484, 1490 (9th Cir.1991) (alteration omitted) (quoting Oltz v. St. Peter’s Cmty. Hosp., 861 F.2d 1440, 1446 (9th Cir.1988)) (internal quotation marks omitted). Put differently, “a market is the group of sellers or producers who have the actual or potential ability to deprive each other of significant levels of business.” Rebel Oil Co. v. Atl. Richfield Co., 51 F.3d 1421 , 1434 (9th Cir.1995) (quoting Thurman Indus., Inc. v. Pay ’N Pak Stores, Inc., 875 F.2d 1369, 1374 (9th Cir.1989)) (internal quotation marks omitted). The plaintiff has the burden of establishing the relevant geographic market. See United States v. Conn. Nat’l Bank, 418 U.S. 656, 669-70 , 94 S.Ct. 2788 , 41 L.Ed.2d 1016 (1974).

A common method to determine the relevant geographic market, and the one used by the district court, is to find whether a hypothetical monopolist could impose a “small but significant nontransi-tory increase in price” (“SSNIP”) in the proposed market. See Theme Promotions, 546 F.3d at 1002 ; see also In re Se. Milk Antitrust Litig., 739 F.3d 262, 277-78 (6th Cir.2014) (describing the relevant geographic market as one in which “buyers ... respond to a SSNIP by purchasing regardless of the increase”); U.S. Dep’t of Justice & FTC, Horizontal Merger Guidelines (“Merger Guidelines’”) § 4 (2010). If enough consumers would respond to a SSNIP by purchasing the product from outside the proposed geographic market, making the SSNIP unprofitable, the proposed market definition is too narrow. See Theme Promotions, 546 F.3d at 1002 .

Market definition thus perforce focuses on the anticipated behavior of buyers and sellers. See, e.g., Rebel Oil, 51 F.3d at 1430, 1434-35. In the health care industry, insurance companies effectively act both as buyers and sellers. See FTC v. Freeman Hosp., 69 F.3d 260 , 270 n. 14 (8th Cir.1995); Gregory Vistnes, Hospitals, Mergers, and Two-Stage Competition, 67 Antitrust L.J. 671, 672 (2000). Noting that “the vast majority of health care consumers are not direct purchasers of health care — the consumers purchase health insurance and the insurance companies negotiate directly with the providers,” the district court correctly focused on the “likely response of insurers to a hypothetical demand by all the PCPs in a particular market for a [SSNIP].”

The district court found that a hypothetical Nampa POP monopolist could profitably impose a SSNIP on insurers. Citing testimony that Nampa residents “strongly prefer access to local PCPs,” the court found that “commercial health plans need to include Nampa PCPs in their networks to offer a competitive product.” “Given this dynamic- — -that health plans must offer Nampa Adult PCP services to Nampa residents to effectively compete— Nampa PCPs could band together and successfully demand a [SSNIP] (or reimbursement increase) from health plans.”

St. Luke’s argues that the district court erred by considering only the current behavior of Nampa consumers, not their likely response to a SSNIP. St. Luke’s is of course correct that geographic market definition involves prospective analysis — -it predicts consumer response to a hypothetical price increase. See FTC v. Tenet Health Care Corp., 186 F.3d 1045 , 1053-54 (8th Cir.1999). But that is precisely what the district court did. The court not only examined present Nampa consumer behavior, but also concluded that it would not change in the event of a SSNIP.

This determination was supported by the record. Evidence was presented that insurers generally need local PCPs to market a health care plan, and that this is true in particular in the Nampa market. For example, Blue Cross of Idaho has PCPs in every zip code in which it has customers, and the executive director of the Idaho Physicians Network testified that it could not market a health care network in Nam-pa that did not include Nampa PCPs. Evidence also indicated that consumers would not change their behavior in the event of a SSNIP. Experts testified that because health care consumers only pay a small percentage of health care costs out of pocket, the impact of a SSNIP likely would not register.. Similarly, there was testimony that consumers choose physicians on factors other than price. The court was unconvinced by evidence that insurers could defend against a SSNIP by steering consumers to non-Nampa PCPs.

For similar reasons, there also was no clear error in the district court’s determination that evidence that one-third of Nampa residents travel to Boise for PCPs did not prove that a significant number of other residents would so travel in the event of a SSNIP. Those who traveled generally went to PCPs near their Boise places of employment. Thus, the court reasonably found this statistic not determinative of whether other Nampa residents would be willing to travel.

C. The Plaintiffs’ Case

Once the relevant geographic market is determined, a prima facie case is established if the plaintiff proves that the merger will probably lead to anticompeti-tive effects in that market. See Olin, 986 F.2d at 1305 ; see also Chi. Bridge & Iron, 534 F.3d at 423 . A prima-facie case can be established simply by showing high market share. United States v. Syufy Enters., 903 F.2d 659 , 664 n. 6 (9th Cir.1990); see also FTC v. H.J. Heinz Co., 246 F.3d 708, 716 (D.C.Cir.2001). However, “statistics concerning market share and concentration, while of great significance, [a]re not conclusive indicators of anticompetitive effects .... ” United States v. Gen. Dynamics Corp., 415 U.S. 486, 498 , 94 S.Ct. 1186 , 39 L.Ed.2d 580 (1974); see also FTC v. Warner Commc’ns Inc., 742 F.2d 1156 , 1163 n. 1 (9th Cir.1984). Thus, plaintiffs in § 7 cases generally present other evidence as part of the prima facie case. See Gen. Dynamics, 415 U.S. at 498 , 94 S.Ct. 1186 (“[0]nly a further examination of the particular market — its structure, history and probable future — can provide the appropriate setting for judging the probable anti-competitive effect of the merger.” (quoting Brown Shoe, 370 U.S. at 322 n. 38, 82 S.Ct. 1502 )); see also Chi. Bridge & Iron, 534 F.3d at 431 (noting that market share data was “just one element in the Government’s strong prima facie case”); Carl Shapiro, The 2010 Horizontal Merger Guidelines: From Hedgehog to Fox in Forty Years, 77 Antitrust L.J. 49, 50-60 (2010) (noting the trend in merger enforcement to consider factors in addition to market share).

The district court held that the plaintiffs established a prima facie case because of the post-merger, entity’s: (1) market share; (2) ability to negotiate higher PCP reimbursement rates with insurers; and (3) ability to “charge more [ancillary] services at the higher hospital billing rates.” The court also found that “entry into the market has been very difficult and would not be timely to counteract the anticom-petitive effects of the Acquisition.” St. Luke’s does not challenge the barriers-to-entry finding; we review the others in turn for clear error.

1. Market Share

A commonly used metric for determining market share is the Herfindahl-Hirschman Index (“HHI”). See ProMedica Health Sys., Inc. v. FTC, 749 F.3d 559, 568 (6th Cir.2014); H.J. Heinz, 246 F.3d at 716 . HHI is “calculated by summing the squares of the individual firms’ market shares,” which “gives proportionately greater weight to the larger market shares.” Merger Guidelines § 5.3. The analysis “considers] both the post-merger level of the HHI and the increase in the HHI resulting from the merger.” Id. The Merger Guidelines classify markets as (1) unconcentrated (HHI below 1500); (2) moderately concentrated (HHI between 1500 and 2500); or (3) highly concentrated (HHI above 2500). Id. Mergers that increase the HHI more than 200 points and result in highly concentrated markets are “presumed to be likely to enhance market power.” Id. “Sufficiently large HHI figures establish the FTC’s prima facie case that a merger is anti-competitive.” H.J. Heinz, 246 F.3d at 716 .

The district court calculated the post-merger HHI in the Nampa PCP market as 6,219, and the increase as 1,607. St. Luke’s does not challenge these findings. As the district court correctly noted, these HHI numbers “are well above the thresholds for a presumptively anticompetitive merger (more than double and seven times their respective thresholds, respectively).” See ProMedica, 749 F.3d at 568 (noting that a merger with similar HHI numbers “blew through those barriers in spectacular fashion”).

2. PCP Reimbursements

The district court also found that St. Luke’s would likely use its post-merger power to negotiate higher reimbursement rates from insurers for PCP services. Recognizing that the § 7 inquiry is based on a prediction of future actions, see Phila. Nat’l Bank, 374 U.S. at 362 , 83 S.Ct. 1715 , this finding was not clearly erroneous.

Because St. Luke’s and Saltzer had been each other’s closest substitutes in Nampa, the district court found the acquisition limited the ability of insurers to negotiate with the merged entity. Pre-acquisition internal correspondence indicated that the merged companies would use this increased bargaining power to raise prices. An email between St. Luke’s executives discussed “pressur[ing] payors for new directed agreements,” and an exchange between Saltzer executives stated that “[i]f our negotiations w/ Luke’s go to fruition,” then “the clout of the entire network” could be used to negotiate favorable terms with insurers. The court also examined a previous acquisition by St. Luke’s in Twin Falls, Idaho, and found that St. Luke’s used its leverage in that instance to force insurers to “concede to their pricing proposal.”

3. Ancillary Services

The district court’s finding that St. Luke’s would raise prices in the hospital-based ancillary services market is more problematic. The court found that St. Luke’s would “exercise its enhanced bargaining leverage from the Acquisition to charge more services at the higher hospital-based billing rates.” Because insurers and providers typically negotiate for all services as part of the same contract, the district court found that St. Luke’s increased leverage with respect to PCP services would allow it to demand higher fees for ancillary services.

The problem with this conclusion is that the district court made no findings about St. Luke’s’ market power in the ancillary services market. Absent such a finding, it is difficult to conclude that the merged entity could easily demand anticompetitive prices for such services. Perhaps the court was suggesting that St. Luke’s would engage in tying, “a device used by a seller with market power in one product market to extend its market power to a distinct product market.” Cascade Health Solutions v. PeaceHealth, 515 F.3d 883, 912 (9th Cir.2008). Although various antitrust statutes, including Sherman Act §§ 1 and 2, Clayton Act § 3, and FTC Act § 5, address tying, Clayton Act § 7 does not expressly prohibit the practice. A leading antitrust treatise cautions against condemning a merger for potential tying effects as “superfluous and overdeterrent.” Phillip Areeda & Herbert Hovenkamp, Antitrust Law: An Analysis of Antitrust Principles and Their Application (“Aree-da”) ¶ 1144a (2010).

Wholly aside from these conceptual difficulties, the factual underpinnings of the district court’s conclusion are suspect. The documents cited by the district court merely state that St. Luke’s hopes to increase revenue from ancillary services, not that it plans to charge higher prices. An increase in revenue could occur in a variety of ways not involving increased prices, such as increased Medicare payments or increased volume from Saltzer referrals. The district court did not find that Saltzer physicians would inappropriately label in-house services as hospital-based, or that they would force patients to travel to the St. Luke’s hospital in Boise for services that could be provided in-house in Nampa. And the court did not identify any past actions that would allow it to predict that St. Luke’s would act anti-competitively in the future in the ancillary services market. Indeed, in post-merger negotiations with Blue Shield, St. Luke’s did not do so. We thus find that the ancillary services finding is not supported by the record.

4. The Prima Facie Case

But absent the ancillary services finding, the district court’s conclusion that a prima facie case was established is amply supported by the record. “Section 7 does not require proof that a merger or other acquisition has caused higher prices in the affected market. All that is necessary is that the merger create an appreciable danger of such consequences in the future.” Hosp. Corp. of Am. v. FTC, 807 F.2d 1381, 1389 (7th Cir.1986).

The extremely high HHI on its own establishes the prima facie case. See H.J. Heinz, 246 F.3d at 716 ; United States v. Baker Hughes, Inc., 908 F.2d 981 , 982-83 & n. 3 (D.C.Cir.1990). In addition, the court found that statements and past actions by the merging parties made it likely that St. Luke’s would raise reimbursement rates in a highly concentrated market. See Hosp. Corp., 807 F.2d at 1388-89 (expressing concern that a history of cooperation among hospitals could lead to collusion when a merger caused the market to become' more concentrated). And, the court’s uncontested finding of high entry barriers “eliminates the possibility that the reduced competition caused by the merger'will be ameliorated by new competition from outsiders and further strengthens the FTC’s case.” H.J. Heinz, 246 F.3d at 717 .

The facts found by the district court are similar to those in other cases in which a prima facie violation of § 7 was established. See, e.g., Chi. Bridge & Iron, 534 F.3d at 431-32 (high HHI, limited rivals, high entry barriers, and customer perception); Lucas Auto. Eng’g, Lie. v. Bridge-stone/Firestone, Inc., 140 F.3d 1228, 1236-37 (9th Cir.1998) (reversing summary judgment for defendant because undisputed facts showed high market share and “insurmountable barriers to entry”); FTC v. Univ. Health, Inc., 938 F.2d 1206 , 1219-20 & n. 27 (11th Cir.1991) (high market concentration, high entry barriers, and evidence that defendants intended to eliminate competition with the merger); Am. Stores, 872 F.2d at 841-43 (high market share, and insufficient evidence of low entry barriers to rebut the prima facie case). The district court did not clearly err in its factual findings, which adequately support its ultimate conclusion that the plaintiffs established “a prima facie case that the Acquisition is anti-competitive.”

D. The Rebuttal Case

Because the plaintiffs established a pri-ma facie case, the burden shifted to St. Luke’s to “cast doubt on the accuracy of the Government’s evidence as predictive of future anticompetitive effects.” Chi. Bridge & Iron, 534 F,3d at 423. The rebuttal evidence focused on the alleged procompetitive effects of the merger, particularly the contention that the merger would allow St. Luke’s to move toward integrated care and risk-based reimbursement.

1. The Post-Merger Efficiencies Defense

The Supreme Court has never expressly approved an efficiencies defense to a § 7 claim. See H.J. Heinz, 246 F.3d at 720 . Indeed, Brovm Shoe east doubt on the defense:

370 U.S. at 344 , 82 S.Ct. 1502 . Similarly, in FTC v. Procter & Gamble Co., the Court stated that “[p]ossible economies cannot be used as a defense to illegality. Congress was aware that some mergers which lessen competition may also result in economies but it struck the balance in favor of protecting competition.” 386 U.S. 568, 580 , 87 S.Ct. 1224 , 18 L.Ed.2d 303 (1967).

Notwithstanding the Supreme Court’s statements, four of our sister circuits (the Sixth, D.C., Eighth, and Eleventh) have suggested that proof of post-merger efficiencies could rebut a Clayton Act § 7 prima facie case. See ProMedica, 749 F.3d at 571 ; H.J. Heinz, 246 F.3d at 720-22 ; Tenet, 186 F.3d at 1054-55; Univ. Health, 938 F.2d at 1222-24. The FTC has also cautiously recognized the defense, noting that although competition ordinarily spurs firms to achieve efficiencies internally, “a primary benefit of mergers to the economy is their potential to generate significant efficiencies and thus enhance the merged firm’s ability and incentive to compete, which may result in lower prices, improved quality,, enhanced service, or new products.” Merger Guidelines § 10; see also Oliver E. Williamson, Economies as an Antitrust Defense Revisited, 125 U. Pa. L.Rev. 699, 699 (1977) (“Sometimes ... a merger will ... result in real increases in efficiency that reduce the average cost of production of the combined entity below that of the two merging firms.”). However, none of the reported appellate decisions have actually held that a § 7 defendant has rebutted a prima facie case with an efficiencies defense; thus, even in those circuits that recognize it, the parameters of the defense remain imprecise.

The status of the defense in this circuit remains uncertain. A quarter of a century ago, we rejected an efficiencies defense in RSR Corp. v. FTC, 602 F.2d 1317, 1325 (9th Cir.1979). RSR, however, involved an argument that the merger would allow the defendant to compete more efficiently outside the relevant market. Id. More recent cases focus on whether efficiencies in the relevant market negate the anticompetitive effect of the merger in that market. See Univ. Health, 938 F.2d at 1222. Even after RSR, several district courts in this circuit have suggested that there could be such a defense. See, e.g., United States v. Bazaarvoice, Inc., No. 13-cv-00133-WHO, 2014 WL 203966 , at *64, *72-73 (N.D.Cal. Jan. 8, 2014); United States v. Oracle Corp., 331 F.Supp.2d 1098, 1174-75 (N.D.Cal.2004); but see California v. Am. Stores Co., 697 F.Supp. 1125, 1132-33 (C.D.Cal.1988) (finding that RSR barred an efficiencies defense), rev’d on other grounds, 872 F.2d 837 , rev’d on other grounds, 495 U.S. 271 , 110 S.Ct. 1853 .

We remain skeptical about the efficiencies defense in general and about its scope in particular. It is difficult enough in § 7 cases to predict whether a merger will have future anticompetitive effects without also adding to the judicial balance a prediction of future efficiencies. Indeed, even then-Professor Bork, a sharp critic of Clayton Act enforcement actions, see, e.g., Robert H. Bork and Wade S. Bowman, Jr., The Crisis in Antitrust, 65 Colum. L.Rev. 363, 373 (1965), rejected the efficiencies defense, calling it “spurious” because it “cannot measure the factors relevant to consumer welfare, so that after the economic extravaganza was completed we would know no more than before it began,” Robert H. Bork, The Antitrust Paradox: A Policy at War with Itself 124 (1978). Judge Richard Posner has regularly expressed similar views. See Richard A. Posner, Antitrust Law 133 (2d ed.2001) (“I said back then that there should be no general defense of efficiency. I still think this is right. It is rarely feasible to determine by the methods of litigation the effect of a merger on the costs of the firm created by the merger.”); Richard A. Posner, Antitrust Law: An Economic Perspective 112 (1976) (“I would not allow a generalized defense of efficiency.”); cf. Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L.Rev. 1, 39 (1984) (“[Njeither judges nor juries are particularly good at handling complex economic arguments.... ”).

Nonetheless, we assume, as did the district court, that because § 7 of the Clayton Act only prohibits those mergers whose effect “may be substantially to lessen competition,” 15 U.S.C. § 18 , a defendant can rebut a prima facie case with evidence that the proposed merger will create a more 'efficient combined entity and thus increase competition. For example, if two small firms were unable to match the prices of a larger competitor, but could do so after a merger because of decreased production costs, a court recognizing the efficiencies defense might reasonably conclude that the transaction likely would not lessen, competition. See Merger Guidelines § 10 (“Merger-generated efficiencies may enhance competition by permitting two ineffective competitors to form a more effective competitor, e.g., by combining complementary assets.... [Ijncremental cost reductions may reduce or reverse any increases in the merged firm’s incentive to elevate price.”).

Because we deal with statutory enforcement, the language of the Clayton Act must be the linchpin of any efficiencies defense. The Act focuses on “competition,” so any defense must demonstrate that the prima facie case “portray[s] inaccurately the merger’s probable effects on competition.” Am. Stores, 872 F.2d at 842 . In other words, a successful efficiencies defense requires proof that a merger is not, despite the existence of a prima facie case, anticompetitive.

Courts recognizing the defense have made clear that a Clayton Act defendant must, “clearly demonstrate” that “the proposed merger enhances rather than hinders competition because of the increased efficiencies.” United States v. Long Island Jewish Med. Ctr., 983 F.Supp. 121, 137 (E.D.N.Y.1997). Because § 7' seeks to avert monopolies, proof of “extraordinary efficiencies” is required to offset the anticompetitive concerns in highly concentrated markets. See H.J. Heinz, 246 F.3d at 720-22 ; see also Merger Guidelines § 10 (“Efficiencies almost never justify a merger to monopoly or near-monopoly.”). The defendant must also demonstrate that the claimed efficiencies are “merger-specific,” see United States v. H & R Block, Inc., 833 F.Supp.2d 36, 89-90 (D.D.C.2011), which is to say that the efficiencies cannot readily “be achieved without the concomitant loss of a competitor,” H.J, Heinz, 246 F.3d at 722 ; see also Merger Guidelines § 10 & n. 13. Claimed efficiencies must be verifiable, not merely speculative. See, e.g., FTC v. CCC Holdings Inc., 605 F.Supp.2d 26, 74-75 (D.D.C.2009); Oracle, 331 F.Supp.2d at 1175 ; see also Merger Guidelines § 10.

2. The St. Luke’s Efficiencies Defense

St. Luke’s argues that the merger would benefit patients by creating a team of employed physicians with access to Epic, the electronic -medical records system used by St. Luke’s. The district court found that, even if true, these predicted efficiencies were insufficient to carry St. Luke’s’ burden of rebutting the prima fa-cie case. We agree.

It is not enough to show that the merger would allow St. Luke’s to better serve patients. The Clayton Act focuses on competition, and the claimed efficiencies therefore must show that the prediction of anti-competitive effects from the prima facie case is inaccurate. See Univ. Health, 938 F.2d at 1222 (finding efficiencies relevant to the prediction of “whether the acquisition would substantially lessen competition”). Although the district court believed that the merger would eventually “improve the delivery of health care” in the Nampa market, the judge did not find that the merger would increase competition or decrease prices. Quite to the contrary, the court, even while noting the likely beneficial effect of the merger on patient care, held that reimbursement rates for POP services likely would increase. Nor did the court find that the merger would likely lead .to integrated health care or a new reimbursement system; the judge merely noted the desire of St. Luke’s to move in that direction.

The district court expressly did conclude, however, that the claimed efficiencies were not merger-specific. The court found “no empirical evidence to support the-theory that St. Luke’s needs a core group of employed primary care physicians beyond the number it had before the Acquisition to successfully make the transition to integrated care,” and that “a committed team can be assembled without employing physicians.” The court also found that the shared electronic record was not a merger-specific benefit because data ana-lytics tools are available to independent physicians.

These factual findings were not clearly erroneous. Testimony highlighted examples of independent physicians who had adopted risk-based reimbursement, even though they were not employed by a major health system. The record also revealed that independent physicians had access to a number of analytic tools, including the St. Luke’s Epic system.

But even if we assume that the claimed efficiencies were merger-specific, the de fense would nonetheless fail. At most, the district court concluded that St. Luke’s might provide better service to patients after the merger. That is a laudable goal, but the Clayton Act does not excuse mergers that lessen competition or create monopolies simply because the merged entity can improve its operations. See Procter & Gamble, 386 U.S. at 580 , 87 S.Ct. 1224 . The district court did not clearly err in concluding that whatever else St. Luke’s proved, it did not demonstrate that efficiencies resulting from the merger would have a positive effect on competition.

IV. Remedy

“The key to the whole question of an antitrust remedy is of course the discovery of measures effective to restore competition.” United States v. E.I. du Pont de Nemours & Co., 366 U.S. 316, 326 , 81 S.Ct. 1243 , 6 L.Ed.2d 318 (1961). “[T]he relief must be directed to that which is necessary and appropriate'... to eliminate the effects of the acquisition offensive to the statute ... and assure the public freedom from its continuance.” Ford Motor Co. v. United States, 405 U.S. 562 , 573 n. 8, 92 S.Ct. 1142 , 31 L.Ed.2d 492 (1972) (internal citation and quotation marks omitted). Section 7 remedies should not be punitive, but “courts are authorized, indeed required, to decree relief effective to redress the violations, whatever the adverse effect of such a decree on private interests.” E.I. du Pont, 366 U.S. at 326 , 81 S.Ct. 1243 .

The customary form of relief in § 7 cases is divestiture. See id. at 330, 81 S.Ct. 1243 (noting that most litigated Clayton Act § 7 cases “decreed divestiture as a matter of course”); see also ProMedica, 749 F.3d at 573 ; RSR, 602 F.2d at 1325-26 ; Ash Grove Cement Co. v. FTC, 577 F.2d 1368 , 1379-80 (9th Cir.1978). Divestiture is the “most important of antitrust remedies,” and “should always be in the forefront of a .court’s mind when a violation of § 7 has been found.” E.I. du Pont, 366 U.S. at 330-31 , 81 S.Ct. 1243 ; see also id. at 329, 81 S.Ct. 1243 (“The very words of § 7 suggest that an undoing of the acquisition is a natural remedy.”). .This is especially true when the government is the plaintiff. See Am. Stores, 495 U.S. at 280-81 , 110 S.Ct. 1853 (“[I]n Government actions divestiture is the preferred remedy for an illegal merger or acquisition.”).

St. Luke’s nonetheless argues that the district court erred in ordering divestiture because (1) divestiture will not actually restore competition; (2) divestiture eliminates the transaction’s procompetitive benefits; and (3) a proposed conduct remedy was preferable. We find no abuse of discretion in the district court’s choice of remedy.

Although divestiture may generally be the most straightforward way to restore competition, E.I. du Pont, 366 U.S. at 331 , 81 S.Ct. 1243 , a district court must consider whether it will effectively do so under the facts of each case. “A primary concern is whether the offending line of commerce, if disassociated from the merged entities, can survive as a viable, independent entity.” FTC v. PepsiCo, Inc., 477 F.2d 24 , 29 n. 8 (2d Cir.1973). St. Luke’s argues that Saltzer would no longer be able to compete post-divestiture, and that divestiture therefore would not restore competition in the Nampa PCP market.

The district court had ample basis, however, for rejecting that contention. Indeed, in opposing a preliminary injunction, St. Luke’s assured the court that divestiture was feasible. Moreover, Salt-zer’s employees were assured by management that they would have their jobs no matter the result of the litigation, and a number of them testified that Saltzer would be viable as an independent entity. The district court also noted that “any financial hardship to Saltzer will be mitigated by St. Luke’s payment of $9 million for goodwill and intangibles as part of the Acquisition.... ”

Nor did the district court abuse its discretion in its consideration of the costs and benefits of divestiture. The court expressly determined that divestiture was appropriate because any benefits of the merger were outweighed by the anticompetitive concerns. See Am. Stores, 872 F.2d at 848 . The Supreme Court has specifically stated that “it is well settled that once the Government has successfully borne the considerable burden of establishing a violation of law, all doubts as to the remedy are to be resolved in its favor.” E.I. du Pont, 366 U.S. at 334 , 81 S.Ct. 1243 .

Finally, the district court did not abuse its discretion in choosing divestiture over St. Luke’s’ proposed “conduct remedy”— the establishment of separate bargaining groups to negotiate with insurers. Divestiture is “simple, relatively easy to administer, and sure,” E.I. du Pont, 366 U.S. at 331 , 81 S.Ct. 1243 , while conduct remedies risk excessive government entanglement in the market, see U.S. Dep’t of Justice, Antitrust Division Policy Guide to Merger Remedies § II n. 12 (2011) (noting that conduct remedies need to be “tailored as precisely as possible to the competitive harms associated with the merger to avoid unnecessary entanglements with the competitive process”). The district court, moreover, found persuasive the rejection of a similar proposal in In re ProMedica Health System, Inc., No. 9346, 2012 WL 1155392 , at *48-50 (F.T.C. March 28, 2012), adopted as modified, 2012 WL 2450574 (F.T.C. June 25, 2012). Even assuming that the district court might have been within its discretion in opting for a conduct remedy, we find no abuse of discretion in its declining to do so. See ProMedica, 749 F.3d at 572-73 (holding that the FTC did not abuse its discretion in choosing divestiture over a proposed conduct remedy).

Y. Conclusion

For the reasons stated above, we AFFIRM the judgment of the district court.

. The district court found that "[ajdult PCP services include physician services provided to commercially insured patients aged 18 and over by physicians practicing internal medicine, family practice, and general practice.” . The parties and the district court regarded the PSA as the functional equivalent of an employment agreement, and we assume the same. . The Private Hospitals filed an amended complaint in January 2013. . The Idaho Competition Act is "construed in harmony” with federal antitrust law, Idaho Code §§ 48-102 (3), -106, and the district court held that the antitrust analysis is the same for each. The parties do not contend otherwise. .The court therefore did not address the Private Hospitals’ contentions with respect to the other product markets. . This quotation is not included in the definitive book of Berra quotations, see Yogi Berra, The Yogi Book: “I Really Didn’t Say Everything I Said!” (1998), and its provenance is at best unclear, see, e.g., The Yale Book of Quotations 92 (Fred R. Shapiro ed., 2006) (attributing a variant to Niels Bohr, but noting that the exact authorship is disputed). . The application of this framework in the Ninth Circuit is not rigid. Thus, in determining whether the prima facie case has been rebutted, a district court may consider evidence submitted by the plaintiff in the casein-chief. See Olin, 986 F.2d at 1305 (finding no burden-shifting error because the FTC had determined that the rebuttal evidence was insufficient to overcome the prima facie showing); see also Chi. Bridge & Iron, 534 F.3d at 424-25 (stating that Olin "allows [a court] to preserve the prima facie presumption if the [defendant] ... fails to satisfy the burden of production in light of contrary evidence in the prima facie case”). . Although the Merger Guidelines are "not binding on the courts,” Olin, 986 F.2d at 1300 , they "are often used as persuasive authority,” Chi. Bridge & Iron, 534 F.3d at 431 n. 11. . This "two-stage model” of health care competition is "the accepted model.” John J. Miles, 1 Health Care & Antitrust L. § 1:5 (2014). In the first stage, providers compete for inclusion in insurance plans. See Vistnes, supra, at 674. In the second stage, providers seek to attract patients enrolled in the plans. See id. at 681-82. Because patients are "largely insensitive” to price, the second stage "takes place primarily over non-price dimensions.” Id. at 682. Thus, antitrust analysis focuses on the first stage. Id. at 692. . Extensive evidence was offered about Micron, a Boise employer that created a health care plan including financial incentives for employees to use certain providers; the plan caused a substantial portion of Micron employees residing in Nampa to switch to non-Nampa PCPs. St. Luke’s argues that this evidence proved that Nampa consumers would respond to a SSNIP. But the district court did not clearly err in finding the Micron example unpersuasive. Micron's cost differentials were much higher than a SSNIP, Boise PCPs were close to work for Micron's employees, and it was unclear whether other employers would be willing or able to replicate Micron's program. . Ancillary services, such as x-rays and diagnostic testing, are sometimes performed by doctors in conjunction with PCP examinations. Before the merger, Saltzer provided many ancillary services at its physicians' offices. Insurance companies and Medicare often offer higher reimbursements for ancillary services performed at a hospital-based outpatient facility. . The district court found that a core reason for high health care costs is the prevalent fee-for-service reimbursement model, based on the apparently uncontested opinions of expert witnesses. Experts have recommended moving toward integrated care and risk-based reimbursement. "In an integrated delivery system, [PCPs] and specialty physicians work as a team, with PCPs managing patient care and specialty physicians consulting and providing care as needed.” Risk-based reimbursement (also known as capitation) means that "providers receive reimbursement from insurers in the form of a set amount for each patient rather than a payment for each service rendered. The set amount is based on the average expected health care utilization for the patients given such factors as their age and medical history.” "Capitation motivates providers to consider the costs of treatment as they will share in the savings if they can keep actual costs below the set amount they receive.” . Some courts have attempted to explain why the Supreme Court cases do not recognize an efficiencies defense, see, e.g., H.J. Heinz, 246 F.3d at 720 n. 18 (arguing that the "possible economies” language in Procter & Gamble does not. ban an actual efficiencies defense), but others have simply stated that the defense exists without addressing the language in Brown Shoe and its progeny, see, e.g., ProMedica, 749 F.3d at 571 . . St. Luke’s argues that once a defendant comes forward with proof of efficiencies, the burden shifts to the plaintiff to show that there are ways of achieving those efficiencies without the merger. This tracks the Sherman Act analysis. See, e.g., Bhan v. NME Hosps., Inc., 929 F.2d 1404, 1412-14 (9th Cir.1991). But, in Clayton Act § 7 cases, after a plaintiff has made a prima facie case that a merger is anticompetitive, the burden of showing that the claimed efficiencies cannot be "attained by practical alternatives,” Merger Guidelines § 10 n. 13, is properly part of the defense, see Olin, 986 F.2d at 1305 (explaining that it is the defendant's "burden to rebut a prima facie case of illegality”). That burden, moreover, is not unduly onerous, as the defendant need not disprove alternatives that are "merely theoretical.” Merger Guidelines § 10. . Conduct remedies include "firewall, nondiscrimination, mandatory licensing, transparency, and anti-retaliation provisions, as well as prohibitions on certain contracting practices.” U.S. Dep't of Justice, Antitrust Division Policy Guide to Merger Remedies § II.B (2011); see also Areeda ¶ 990d.