Chivas Products Ltd. v. Owen

864 F.2d 1280, 1988 WL 137801

• Court: Court of Appeals for the Sixth Circuit
• Decided: December 28, 1988
• Docket: No. 87-1784
• Status: Published

Opinions

MERRITT, Circuit Judge.

This is a removal case transferred from the state to the federal court under 28 U.S.C. § 1441.¹ The dispositive question on appeal from the District Court’s dismissal of a Racketeer Influenced and Corrupt Organizations Act (RICO) civil action is one of subject-matter jurisdiction: is civil RICO jurisdiction exclusively federal or concurrent with the states, and if exclusively federal, does the doctrine of “derivative” removal jurisdiction require dismissal in the case? We hold that RICO jurisdiction is exclusively federal and that the state court had no jurisdiction of the RICO claim. Therefore, under the doctrine of “derivative” removal jurisdiction, we have no subject matter jurisdiction and must vacate the judgment below and dismiss the case.

I.

This action originally was brought in the Circuit Court of Michigan in November 1985, seeking damages and equitable relief on a variety of state law theories. On November 18, 1986, plaintiff filed an amended complaint adding a RICO count. Defendants shortly thereafter removed the action to the District Court. The basic allegations were that the combined efforts of defendants, a California automobile components manufacturer and employees of a Michigan distributor, to set up a new entity to distribute the manufacturer’s products were designed to take away the plaintiff distributor’s business and its trade secrets. The plaintiff claims that this conduct constitutes a “pattern of racketeering activity” in violation of the federal mail fraud statute, 18 U.S.C. § 1341, and thus of civil RICO, 18 U.S.C. §§ 1964(c), (d).

In a pretrial conference call with all parties on March 3, 1987, the District Court advised plaintiff’s counsel that “he did not think much of plaintiff’s RICO claim,” and invited defense counsel to file a motion to dismiss or for summary judgment. Plaintiff’s Brief at 6. No motion was forthcoming. According to the District Judge, his law clerk called plaintiff’s counsel subsequently and advised him “of the court’s view that [the RICO count] failed to state a cause of action.” Order of July 15,1987 at 2. At oral argument before this Court, plaintiff’s counsel acknowledged that this conversation occurred, but represented that it consisted basically only of the following:

Law Clerk: “What are you going to do?”

Counsel: “We’re going to engage in discovery.”

None of the Court’s or the clerk’s contacts with counsel were written or otherwise recorded.

On May 27, 1987, the District Court ordered that the RICO count be dismissed because the one discrete course of conduct alleged fails to satisfy the “continuity” aspect of the “pattern” requirement of RICO, 18 U.S.C. §§ 1961(5), 1962(c); see Sedima, S.P.R.L. v. Imrex Co., 473 U.S. 479, 496 n. 14, 105 S.Ct. 3275, 3285 n. 14, 87 L.Ed.2d 346 (1985), relying on the reasoning used in the Eighth Circuit that a RICO “pattern of racketeering” must occur in more than a single criminal episode or scheme. See Superior Oil Co. v. Fulmer, 785 F.2d 252, 257 (8th Cir.1986); see also Northern Trust Bank/O’Hare N.A. v. Inryco, Inc., 615 F.Supp. 828, 831-34 (N.D.Ill.1985). The Supreme Court has granted certiorari in another Eighth Circuit case premised on the Superior Oil reasoning. See H.J. Inc. v. Northwestern Bell Telephone Co., 829 F.2d 648 (8th Cir.1987), cert. granted, — U.S. —, 108 S.Ct. 1219, 99 L.Ed.2d 420 (1988).

In the May 27 order, the District Court also remanded the state claims pursuant to United Mine Workers v. Gibbs, 383 U.S. 715, 726, 86 S.Ct. 1130, 1139, 16 L.Ed.2d 218 (1966). Plaintiff subsequently moved for reconsideration and for leave to file an amended and supplemental complaint that was proposed to allege multiple schemes satisfying the “pattern” standard used by the District Court. The motion was denied, and this appeal followed, seeking review of the procedural propriety of the District Court’s action.

II.

Under our federal system of dual sovereignty, the subject matter jurisdiction of state courts is governed in the first instance by state laws; federal rights are ordinarily enforceable both in federal courts and in state courts. See Gulf Offshore Co. v. Mobil Oil Corp., 453 U.S. 473, 478-79, 101 S.Ct. 2870, 2875, 69 L.Ed.2d 784 (1981). Normally when Congress creates new civil causes of action by statute concurrent jurisdiction in the state courts is presumed, although the first Congress in the Judiciary Act of 1789 committed exclusive jurisdiction of federal criminal cases to the federal courts. Judiciary Act of 1789, ch. 20, § 11, 1 Stat. 79, now codified at 18 U.S.C. § 3231. This presumption in civil cases arises both out of federal recognition of the states’ independent sovereignty and from the states’ duty under the Supremacy Clause to recognize federal law as paramount. Id.; see The Federalist No. 82 (A. Hamilton). For more detailed discussions of the early development of rules of exclusive and concurrent jurisdiction over federal criminal and civil cases in state and federal courts, see H. Friendly, Federal Jurisdiction 8-11 (1973); F. Frankfurter & J. Landis, The Business of the Supreme Court 5-12 (1928); Warren, New Light on the History of the Federal Judiciary Act of 1789, 37 Harv.L.Rev. 49 (1923).

The presumption in civil cases may be rebutted, however. Congress may confine jurisdiction of a federal claim to the federal courts either explicitly or implicitly “[1] by an explicit statutory directive, [2] by unmistakable implication from legislative history, or [3] by a clear incompatibility between state-court jurisdiction and federal interests.” Gulf Offshore, 453 U.S. at 479, 101 S.Ct. at 2875.

District courts and state courts have about equally divided on the issue of RICO concurrent jurisdiction, and the federal courts of appeals have only recently considered the problem. Only the Ninth Circuit, which found jurisdiction concurrent in a case from California, has squarely faced the issue. See Lou v. Belzberg, 834 F.2d 730 (9th Cir.1987), cert. denied, — U.S. -, 108 S.Ct. 1302, 99 L.Ed.2d 512 (1988). The California Supreme Court earlier had held jurisdiction concurrent, Cianci v. The Superior Court of Contra Costa County, 40 Cal.3d 903, 221 Cal.Rptr. 575, 710 P.2d 375 (1985). In dicta, the Seventh Circuit has leaned toward concurrent jurisdiction in a case from Illinois in County of Cook v. MidCon Corp., 773 F.2d 892, 905 n. 4 (7th Cir.1985), although an Illinois intermediate appellate court has held to the contrary, see Thrall Car Mfg. Co. v. Lindquist, 145 Ill.App.3d 712, 718, 99 Ill.Dec. 397, 401, 495 N.E.2d 1132, 1136 (1986).

Two other federal courts of appeals have decided that state courts within their circuits were either likely or unlikely to assume concurrent jurisdiction as a matter of state law. See Dubroff v. Dubroff, 833 F.2d 557, 562 (5th Cir.1987) (Texas likely); Cullen v. Margiotta, 811 F.2d 698, 732 (2d Cir.) (New York courts unlikely, relying on Greenview Trading Co. v. Hershman, 108 A.D.2d 468, 489 N.Y.S.2d 502 (1985), which held civil RICO jurisdiction not concurrent), cert. denied, — U.S. —, 107 S.Ct. 3266, 97 L.Ed.2d 764 (1987). Our Circuit has yet to address this issue in a published case. There are no reported Michigan decisions.

The statutory language of civil RICO does not explicitly commit jurisdiction exclusively to the federal courts, and it is “black letter law” that the “mere grant of jurisdiction to a federal court does not oust the state courts from concurrent jurisdiction.” Gulf Offshore, 453 U.S. at 480, 101 S.Ct. at 2876.

The civil RICO damage action is authorized by 18 U.S.C. § 1964(c), which provides that any person injured “by reason of violation” of criminal RICO § 1962 “may sue therefor in any appropriate United States district court.” It is clear that this language alone does not constitute an “explicit statutory directive” for exclusive jurisdiction of the type described in Gulf Offshore. But we must examine other language in the statute to explore the other two ways that Congress may have designed RICO as an exclusively federal cause of action.

There is no “smoking gun” legislative history in which RICO sponsors indicated an express intention to commit civil RICO to the federal courts. Indeed, Professor G. Robert Blakey, now a Notre Dame law professor who as the chief of the Senate committee staff was the originator of RICO and its principal draftsman, has conceded that “ ‘no one even thought of the issue’ ” but that if members of Congress had, “ ‘they would have made [jurisdiction] exclusive.’ ” Lou v. Belzberg, 834 F.2d at 736 & n. 4 (quoting Prof. Blakey’s remarks reported in Flaherty, Two States Lay Claim to RICO, Nat’l L.J., May 7, 1984, at 3, col. 1-2).

Prof. Blakey’s speculation is a slender reed, cf. Blakey & Cessar, Equitable Relief Under Civil RICO: Reflections on Religious Technology Center v. Wollersheim: Will Civil RICO Be Effective Only Against White-Collar Crime?, 62 Notre Dame L.Rev. 526, 526 n. * (1987), but we believe that there is considerable evidence in the legislative history, though short of an express statement, in support of this view and that the legislative sponsors should be held to have intended an exclusively federal forum for civil RICO.

Although the drafters of RICO consciously avoided placing the provisions of RICO within the Sherman Act, see Sedima, 473 U.S. at 498-99, 105 S.Ct. at 3285-86, they equally consciously borrowed the exact language of the Clayton Act in drafting the jurisdiction and venue sections for civil RICO, see Sedima, 473 U.S. at 486-88, 105 S.Ct. at 3279-80; County of Cook v. Mid-Con Corp., 574 F.Supp. 902, 912 (N.D.Ill.1983), jurisdictional holding questioned in dicta, 773 F.2d 892, 905 n. 4 (7th Cir.1985).

At the time Congress enacted RICO, section 4 of the Clayton Act, 15 U.S.C. § 15 (1964 & Supp. V 1965-69) provided that “[a ]ny person who shall be injured in his business or property by reason of anything forbidden in the antitrust laws may sue therefor in any district court of the United States ... and shall recover threefold the damages by him sustained and the cost of suit, including a reasonable attorney’s fee.”

Civil RICO’s jurisdictional section, 18 U.S.C. § 1964(c), provides that “[a ~\ny person injured in his business or property by reason of a violation of section 1962 of this chapter may sue therefor in any appropriate United States district court and shall recover threefold the damages he sustains and the cost of the suit, including a reasonable attorney’s fee.”

The identical language is emphasized. What little nonidentical language exists consists of necessary substitutions, simplifications, and trivial inversions: deletion of “who shall be;” substitution of “a violation of section 1962 of this chapter” for “anything forbidden in the antitrust laws;” change from “district court of the United States” to “United States district court;” and change from “by him sustained” to “he sustains.”

As noted in Sedima, 473 U.S. at 486-88, 105 S.Ct. at 3279-80, borrowing from the Clayton Act concept was suggested by the American Bar Association and proposed by Rep. Steiger. The principal House sponsor, Rep. Poff, described the treble-damages provision as “another example of the antitrust remedy being adapted for use against organized criminality.” 116 Cong.Rec. 35295 (1970); see also 115 Cong.Rec. 6993 (1969) (statement of Sen. Hruska).

The first noteworthy point here is that the Clayton Act unmistakably provided for exclusively federal jurisdiction. There is explicit legislative history that Congress intended exclusivity and rejected a proposal for concurrent jurisdiction, see 51 Cong.Rec. 9662-64 (1914), and the Supreme Court repeatedly has held against concurrent jurisdiction, see Marrese v. American Academy of Orthopaedic Surgeons, 470 U.S. 373, 380-81, 386-87, 105 S.Ct. 1327, 1331-32, 1335, 84 L.Ed.2d 274 (1985); General Investment Co. v. Lake Shore & M.S. Ry., 260 U.S. 261, 287, 43 S.Ct. 106, 117, 67 L.Ed. 244 (1922); see also Freeman v. Bee Machine Co., 319 U.S. 448, 63 S.Ct. 1146, 87 L.Ed. 1509 (1943); see generally Redish & Muench, Adjudication of Federal Causes of Action in State Court, 75 Mich.L.Rev. 311, 316-18 (1976) (hereinafter Adjudication in State Court).

There is no doubt that the actual intent of Congress in enacting the Clayton Act in 1914 was to follow the precedent of earlier lower federal court rulings that the Sherman Act, enacted in 1890, had called for exclusively federal jurisdiction. See Redish & Muench, Adjudication in State Court, 75 Mich.L.Rev. at 317-18 & n. 34; Note, Exclusive Jurisdiction of the Federal Courts in Private Civil Actions, 70 Harv.L.Rev. 509, 510 n. 13 (1957) (hereinafter Exclusive Jurisdiction). In any event, the Supreme Court later ratified this view of the Sherman Act. See Blumenstock Bros. Advertising Agency v. Curtis Publishing Co., 252 U.S. 436, 440, 40 S.Ct. 385, 386, 64 L.Ed. 649 (1920).

The second point is that there is unmistakably clear evidence that Congress in 1970 closely patterned the civil RICO damages action on § 4 of the Clayton Act, which creates exclusively federal jurisdiction. See Agency Holding Corp. v. Malley-Duff & Associates, 483 U.S. 143, 107 S.Ct. 2759, 2764-65, 97 L.Ed.2d 121 (1987) (noting, for statute of limitations purposes, “similarities in purpose and structure between RICO and the Clayton Act” and “clear legislative intent to pattern RICO’s civil enforcement provision on the Clayton Act”); Sedima, 473 U.S. at 489, 105 S.Ct. at 3281 (“clearest current” in RICO legislative history is “reliance on the Clayton Act model”). Some other courts, noting the absence of evidence that RICO sponsors were actually aware of the exclusive-jurisdiction feature of the Clayton Act, have downplayed this aspect of § 1964(c). We find it significant because § 1964(c), which parallels § 4 of the Clayton Act, is, after all, private civil RICO’s jurisdictional provision. Compare Lou v. Belzberg, 834 F.2d at 736-37; County of Cook v. MidCon Corp., 773 F.2d at 905 n. 4.

The third way in which the Gulf Offshore presumption may be rebutted is by a “clear incompatibility between state-court jurisdiction and federal interests.” RICO is a criminal statute, and, as the Court in Sedima noted, the civil portion of the statute derives its meaning from the first two sections which define criminal liability. 473 U.S. at 488-89, 105 S.Ct. at 3280-81. The federal interests embedded in civil RICO are best assessed not by narrow scrutiny of § 1964(c) in isolation but by examination of the civil RICO damages remedy in the context of the entire RICO statutory scheme.

The legislative history of RICO indicates clearly that the civil damages remedy of § 1964(c) was designed as an integral part of a broad-front attack on organized crime. Like its model, the Clayton Act, civil RICO seeks to enlist “the pressure of ‘private attorneys general’ on a serious national problem for which public prosecutorial resources are deemed inadequate.” Malley-Duff 107 S.Ct. at 2764; see Sedima, 473 U.S. at 487, 105 S.Ct. at 3280 (quoting Rep. Steiger that the civil remedy “would enhance the effectiveness” of the rest of RICO).

The entire scheme has been well described by the commentators: § 1961 defines key concepts, including predicate acts of “racketeering activity,” in the language of criminal liability; § 1962 prohibits different acts; § 1963 establishes criminal penalties; and § 1964 provides civil remedies on top of the criminal liability. Sections 1965-68 govern venue and process, and provide specially for expedition of actions, discovery and closure of proceedings brought by the United States. See generally Blakey & Gettings, Racketeer Influenced and Corrupt Organizations (RICO): Basic Concepts—Criminal and Civil Remedies, 53 Temp.L.Q. 1009 (1980); P. Batista, Civil RICO Practice Manual §§ 2.2-2.6 (1987).

For three principal reasons, we think that features of the RICO structure are incompatible with concurrent state-court jurisdiction. First, virtually all of the predicate offenses defined as “racketeering activity” in § 1961(1) are themselves prosecutable only in the federal courts. The only exceptions are those generic offenses described in § 1961(1)(A) that are “chargeable under State law” and possibly some “fraud in the sale of securities,” § 1961(1)(C). The remaining predicate offenses are enumerated in a long list of federal criminal offenses including a number of highly specialized areas — drug laws, ERISA, labor, bankruptcy, securities, white slavery, mail fraud — whose adjudication is committed to the exclusive jurisdiction of the federal courts by statute. See 18 U.S.C. § 3231 (“offenses against the laws of the United States”); 15 U.S.C. § 78aa (1934 Securities Exchange Act violations). The state courts do not have jurisdiction to interpret and apply these laws, and they are generally unfamiliar with them. The great preponderance of civil RICO cases allege the federal offenses of mail or wire fraud or federal securities fraud. See Report of the Ad Hoc Civil RICO Task Force of the ABA Section of Corporation, Banking and Business Law 57 (1985) (some 80% of civil RICO claims based solely or primarily on mail, wire, or securities fraud). Two courts who have claimed to the contrary that civil RICO usually involves “garden variety state law fraud” have misconstrued these statistics. See Cianci, 710 F.2d at 380 n. 3, and in HMK Corp. v. Walsey, 637 F.Supp. 710, 717 (E.D.Va.1986), aff'd, 828 F.2d 1071 (4th Cir.1987), cert. denied, — U.S. -, 108 S.Ct. 706, 98 L.Ed.2d 657 (1988). Not only are most of the predicate offenses exclusively committed to the federal courts, so is criminal RICO itself, including the often elusive concepts and definitions embedded in §§ 1961 and 1962.

Second, the frequently interstate character of alleged racketeering enterprises means that often predicate acts by the same defendants will occur in several states. See Malley-Duff, 107 S.Ct. at 2766. State-court jurisdiction would multiply the risk of inconsistent decisions in state courts in cases brought by different plaintiffs. The state courts have no multi-district litigation panel and no way to make their decisions on federal law consistent except through review by the Supreme Court. State court jurisdiction over the complex and ambiguous RICO statute would create-the possibility of conflicting interpretations not only in the 12 regional federal courts of appeals but also in 50 states.

Third, the procedural apparatus created by §§ 1965-68 seems inconsistent with concurrent state court adjudication of § 1964(c) claims. Section 1965(a) and (d)’s extended venue and process provisions would be applicable only to civil RICO cases brought in federal court, thus creating either a procedural quagmire of multiple state and federal remedial procedures or a comity-threatening necessity for many state courts to alter radically their normal procedures. In addition, the remedial and procedural sections of civil RICO place great emphasis on the special treatment to be accorded to civil actions brought by the United States. These provisions sometimes limit their application to “any district court of the United States,” see §§ 1965(b)(c); 1966; 1968(g)-(j), but sometimes do not, see §§ 1964(b), 1967. Even if the Attorney General wanted to prosecute a civil RICO case in state court, this procedural scheme, if it does not commit such cases exclusively to the federal courts, certainly creates enormous incentives for the Government to litigate there. We find it difficult to accept that the same statutory scheme which explicitly provides extraordinary procedural vehicles in the federal district courts would silently contemplate concurrent jurisdiction in state courts, many of which either lack those mechanisms or would require considerable revamping to afford them.

In summary, while the statutory language of civil RICO does not say “exclusive jurisdiction,” the legislative history strongly suggests that this was the implicit legislative purpose. Moreover, consideration of several unique features of the RICO definitional and remedial structure demonstrates convincingly that state court jurisdiction would obstruct strong federal interests in uniform substantive and procedural enforcement of this integrated statutory scheme to control organized crime affecting interstate commerce. Thus the state court from which this case was removed lacked subject matter jurisdiction over the RICO claim.

III.

Ordinarily, when a case is removed from a state court which itself lacks subject-matter jurisdiction of the claim that served as the basis for removal, the pre1986 doctrine of “derivative jurisdiction” comes into play. That judicial doctrine holds that a federal district court derives its jurisdiction from the state court from which an action is removed; if the state court had no jurisdiction, the federal court acquires none, even if it would have had jurisdiction had the action been filed in federal court in the first instance. See Franchise Tax Board v. Construction Laborers Vacation Trust, 463 U.S. 1, 24 & nn. 26-27, 103 S.Ct. 2841, 2854 & nn. 26-27, 77 L.Ed.2d 420 (1983); Minnesota v. United States, 305 U.S. 382, 389, 59 S.Ct. 292, 295, 83 L.Ed. 235 (1939); Lambert Run Coal Co. v. Baltimore & O.R.R., 258 U.S. 377, 382, 42 S.Ct. 349, 351, 66 L.Ed. 671 (1922).

The doctrine has been much disapproved, see 14A C. Wright, A. Miller & E. Cooper, Federal Practice & Procedure § 3722 at 284-90, and in 1986 Congress overruled it by statute, see 28 U.S.C. § 1441(e). But the 1986 amendment was applicable only to “civil actions” commenced in state court after June 19, 1986. This action was commenced before the effective date of § 1441(e), although the RICO claim was added by amendment after that date. The operative date under the statute is the date of the commencement of the “civil action,” not the date of the added claim. See 28 U.S.C.A. § 1441(e) note (“The amendment ... shall apply with respect to claims in civil actions commenced in State courts on or after ... June 19, 1986.”)

There is, however, an established exception to this now repealed problem of removal jurisdiction which we must discuss. In Grubbs v. General Electric Credit Corp., 405 U.S. 699, 92 S.Ct. 1344, 31 L.Ed.2d 612 (1972), the Court held that when a case is improperly removed but is tried to judgment on the merits without objection, a party waives his right to later raise the issue of lack of subject-matter jurisdiction at the time of removal. “[T]he issue in subsequent proceedings on appeal is not whether the case was properly removed, but whether the federal district court would have had original jurisdiction of the case had it been filed in that court.” Id. at 702, 92 S.Ct. at 1347.

The Grubbs rule is eminently sensible and conservative of judicial economy. It was the approach adopted long ago in this Circuit, see Handley-Mack Co. v. Godchaux Sugar Co., 2 F.2d 435, 437 (6th Cir.1924); Riggs v. Island Creek Coal Co., 542 F.2d 339, 343 (6th Cir.1976), and has been applied recently in a RICO case to which the doctrine of derivative jurisdiction might otherwise have applied, see Foval v. First National Bank of Commerce, 841 F.2d 126, 129 (5th Cir.1988). See also Sorosky v. Burroughs Corp., 826 F.2d 794, 801 (9th Cir.1987).

Unlike Grubbs and the precedents cited by the Court in Grubbs, however, this action has not been tried “on the merits.” Rather, the District Court dismissed it under F.R.Civ.P. 12(b)(6) at an early stage, when little or no discovery had been done. Courts in the Fourth, Fifth, Ninth and Eleventh Circuits have extended the Grubbs rule to cases on appeal following summary judgment. See, e.g., Able v. Upjohn, Inc., 829 F.2d 1380, 1334 (4th Cir.1987), cert. denied, — U.S. —, 108 S.Ct. 1229, 99 L.Ed.2d 429 (1988); Farina v. Mission Investment Trust, 615 F.2d 1068 (5th Cir.1980); Sorosky v. Burroughs Corp., 826 F.2d 794 (9th Cir.1987); Stone v. Stone, 632 F.2d 740, 742 (9th Cir.1980), cert. denied, 453 U.S. 922, 101 S.Ct. 3158, 69 L.Ed.2d 1004 (1981); Borg-Warner Leasing v. Doyle Electric Co., 733 F.2d 833, 835 n. 2 (11th Cir.1984).

It is not appropriate to extend the Grubbs exception so far as to include a case like this, when dismissal has come so early in the pleadings stage that there has been minimal investment of the parties’ time in discovery or of the court’s time in judicial proceedings or deliberations. The Grubbs “waiver” rule springs from concerns for judicial economy and from a desire to prevent a party that bided its time before raising a defect in removal jurisdiction from getting a second bite at the apple. We think neither concern is significant here, and thus Grubbs is inapplicable. Accord, Dyer v. Greif Brothers, Inc., 766 F.2d 398, 401 (9th Cir.1985) (in case dismissed for failure to prosecute and failure to attend pretrial conference, “concern for judicial economy ... less compelling” when dismissed “at an early stage”).

We express no opinion on the propriety of application of the Grubbs exception to cases resolved on summary judgment, except to note that summary judgment may involve only limited issues with minimal investment of time or may require extensive time and effort exceeding that experienced in many cases “tried on the merits.” Compare, e.g., Foval v. First Nat’l Bank, 841 F.2d 126, 128-29 (5th Cir.1988) (limited proceedings; Grubbs nevertheless held applicable) with Vantine v. Elkhart Brass Mfg. Co., 762 F.2d 511, 518 (7th Cir.1985) (extensive discovery and proceeding; Grubbs held applicable).

IV.

The judgment of the District Court is vacated, and the case remanded to the District Court with instructions to dismiss the case for want of subject-matter jurisdiction.

1 . Section 1441(a) provides:

[A]ny civil action brought in a State court of which the district courts of the United States have original jurisdiction, may be removed by ... the defendants, to the district court of the United States for the district and division embracing the place where such action is pending.